Professional Documents
Culture Documents
Main Index
Corporate Europe, Middle East, Africa
MSC Software Corporation MSC Software GmbH
4675 MacArthur Court, Suite 900 Am Moosfeld 13
Newport Beach, CA 92660 81829 Munich, Germany
Telephone: (714) 540-8900 Telephone: (49) 89 431 98 70
Toll Free Number: 1 855 672 7638 Email: europe@mscsoftware.com
Email: americas.contact@mscsoftware.com
Japan Asia-Pacific
MSC Software Japan Ltd. MSC Software (S) Pte. Ltd.
Shinjuku First West 8F 100 Beach Road
23-7 Nishi Shinjuku #16-05 Shaw Tower
1-Chome, Shinjuku-Ku Singapore 189702
Tokyo 160-0023, JAPAN Telephone: 65-6272-0082
Telephone: (81) (3)-6911-1200 Email: APAC.Contact@mscsoftware.com
Email: MSCJ.Market@mscsoftware.com
Worldwide Web
www.mscsoftware.com
Disclaimer
MSC Software Corporation reserves the right to make changes in specifications and other information contained
in this document without prior notice.
The concepts, methods, and examples presented in this text are for illustrative and educational purposes only,
and are not intended to be exhaustive or to apply to any particular engineering problem or design. MSC Software
Corporation assumes no liability or responsibility to any person or company for direct or indirect damages resulting
from the use of any information contained herein.
User Documentation: Copyright 2016 MSC Software Corporation. . All Rights Reserved.
This notice shall be marked on any reproduction of this documentation, in whole or in part. Any reproduction or
distribution of this document, in whole or in part, without the prior written consent of MSC Software Corporation is
prohibited.
This software may contain certain third-party software that is protected by copyright and licensed from MSC
Software suppliers. Additional terms and conditions and/or notices may apply for certain third party software. Such
additional third party software terms and conditions and/or notices may be set forth in documentation and/or at
http://www.mscsoftware.com/thirdpartysoftware (or successor website designated by MSC from time to time).
PCGLSS 8.0, Copyright 1992-2016, Computational Applications and System Integration Inc. All rights reserved.
PCGLSS 8.0 is licensed from Computational Applications and System Integration Inc.
MSC, Dytran, Marc, MSC Nastran, Patran, the MSC Software corporate logo, e-Xstream, Digimat, and Simulating
Reality are trademarks or registered trademarks of the MSC Software Corporation and/or its subsidiaries in the
United States and/or other countries.
NASTRAN is a registered trademark of NASA. LS-DYNA is a trademark or registered trademark of Livermore
Software Technology Corporation. FLEXlm and FlexNet Publisher are trademarks or registered trademarks of
Flexera Software. All other trademarks are the property of their respective owners.
Use, duplicate, or disclosure by the U.S. Government is subjected to restrictions as set forth in FAR 12.212
(Commercial Computer Software) and DFARS 227.7202 (Commercial Computer Software and Commercial
Computer Software Documentation), as applicable.
U.S. Patent 9,361,413
Revision 0. November 25, 2016
NA:V2017:Z:Z:DC-USR-PDF
Main Index
Contents
MSC Nastran Nonlinear Users Guide
MSC Nastran Nonlinear
Users Guide
Contents
1 Introduction
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Capabilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Truss Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Main Index
4 MSC Nastran Nonlinear Users Guide
Parameter Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Linear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Nonlinear Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Main Index
Contents 5
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Main Index
6 MSC Nastran Nonlinear Users Guide
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5 Computational Methods
Nonlinear Computational Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Material Nonlinearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Large Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Contact. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Stiffness Matrix Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Element Assembly Parallelization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Element Memory Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Stiffness Matrix Memory Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Main Index
Contents 7
Main Index
8 MSC Nastran Nonlinear Users Guide
Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Determining if Your Simulation has failed to Converge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Determining if Your Simulation is Converging Slowly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Rigid Body Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Load Controlled Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Contact Convergence Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Follower Force Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Material Nonlinearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Mesh Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Instability/Buckling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Main Index
Contents 9
Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Single-Point Constraints (SPC, SPC1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Main Index
10 MSC Nastran Nonlinear Users Guide
9 Contact
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Main Index
Contents 11
Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
Tolerance on Nodal Reaction Force or Nodal Stress Before Separation Occurs. . . . . . . . . . . . . . . . 388
Release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Defining Separation in Patran. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
Bilinear Friction Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Friction User Subroutine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
Friction Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading. . . . . . . . . . 407
Problem Statement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
The Structure and Its Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Model Description. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
Solving the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
Highlighted Sections of the MSC Nastran Bulk Data File. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
Inspecting the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
Main Index
12 MSC Nastran Nonlinear Users Guide
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
10 Materials
Material Model Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
Main Index
Contents 13
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements . . 557
Interlaminar Stresses for Continuum Composite Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559
Main Index
14 MSC Nastran Nonlinear Users Guide
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614
11 Element Library
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616
Main Index
Contents 15
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
Main Index
16 MSC Nastran Nonlinear Users Guide
13 Advanced Topics
Superelements and Modal Neutral Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724
BSQUEAL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 740
RESTART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
File Management Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
Case Control Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
Bulk Data Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757
Main Index
Contents 17
Main Index
18 MSC Nastran Nonlinear Users Guide
Main Index
Chapter 1: Introduction
Nonlinear Users Guide
1 Introduction
Introduction 20
MSC Software Products 22
MSC Nastran Implicit Nonlinear (SOL 400) 25
Capabilities 27
Using the Finite Element Method (FEM) 29
Clevis Pin Lug Design 30
General Analysis Tips 37
Frequently Asked Questions 38
Main Index
20 Nonlinear Users Guide
Introduction
Introduction
This book is one in a series of users guides describing the use of MSC Nastran. The first book in the series, Getting
Started with MSC Nastran Users Guide, is intended for those readers who are unfamiliar with the finite element
method (FEM), those who have never run MSC Nastran, or those who are interested in a general overview of MSC
Nastrans capabilities. It is assumed that you are familiar with the material in Getting Started with MSC Nastran Users
Guide and have MSC Nastran available.
This manual provides a complete background to MSC Nastran Implicit Nonlinear (SOL 400) and describes using SOL
400 within the MSC Nastran environment. The theoretical aspects of nonlinear analysis methods, types, and techniques
are included as well as thorough descriptions for nonlinear material models, properties, and loads and constraints. The
goal is to provide material relevant to this subject in such a manner that this book can be used both as a learning tool
and as a reference text.
Where appropriate, Patran forms and menus are shown so you can easily use SOL 400 from the Patran pre- and
postprocessing software environment.
The contents of this manual as follows:
Introduction (Ch. 1) discusses the FEM and shows a sample linear and nonlinear analyses. The following seven
chapters will take you through choices and formats needed for building and loading a model.
MSC Nastran Files (Ch. 2) discusses the input and output files and file management, executive, case control, bulk data
and parameter sections of an MSC Nastran analysis input file through the use of small sample problems.
Nonlinearity and Analysis Types (Ch. 3) discusses the main purposes of the Solution Methods and Strategies including
a brief overview of the finite element displacement method for solving linear and nonlinear structural problems in
statics, a description of the theory, techniques and algorithms specifically used in SOL 400, and outlines some
guidelines for selecting appropriate analysis methods.
Solution Strategies for Nonlinear Analysis (Ch. 4) covers the three sources of nonlinearity: material (results from the
nonlinear relationship between stresses and strains), geometric (large deformations change the load path and thus the
way the structure carries the load), and nonlinear boundary conditions (such as contact) and describes the analysis
procedures that have been implemented in SOL 400 to accurately simulate those nonlinear behaviors. The analysis
types refer to different categories of use cases that might be modeled such as static (no changes with time), transient
(loading or material response changes with time), buckling (loading causes a loss of stability), and modal (uses the
natural frequencies of the system which may vary upon changes in loading or material response) simulations.
Computational Methods (Ch. 5) includes numerical methods and running large models.
Setting Up, Monitoring, and Debugging the Analysis (Ch. 6) discusses the steps required to prepare the input file for the
purpose of performing a linear or nonlinear analysis, how to submit the job, how to monitor the progress and
completion of the analysis, as well as what steps to take if you do not get the desired results solution.
Results Output from the Analysis (Ch. 7) provides and overview of the available MSC Nastran output file formats and
shows how to specifying the appropriate output requests in the input file to obtain the desired output.
Load and Constraint Conditions (Ch. 8) focuses on how to specify the appropriate boundary conditions for your model
to properly constrain, as well as prevent any rigid body motion, of your structure. This chapter also discusses how to
load the structure using various types of load and boundary conditions such as accelerations, pressures, forces and
Main Index
Chapter 1: Introduction 21
Introduction
thermal loads. This chapter also covers symmetry and has an expanded description on how to exploit symmetry in a
structure to reduce problem size.
Contact (Ch. 9) covers the use of 3-D contact to simulate physical reality between model components that come into
contact. It includes standard touching contact, glued contact (relative motion is prevented), friction, and relative
sliding. The objective of contact modeling is to detect the relative motion of the bodies, apply a constraint to avoid
penetration, and apply appropriate boundary conditions to simulate the contact behavior, which may include frictional
behavior and/or heat transfer. The analysis of contact behavior is complex because of the requirement to accurately
track the motion of multiple geometric bodies, and the motion due to the interaction of these bodies after contact
occurs. This includes analysis of interference fits, rubber seals, tires, crash, and manufacturing processes among others.
Materials (Ch. 10) begins by discussing SOL 400 support for isotropic, orthotropic and anisotropic linear and
composite materials, and then goes through nonlinear material modeling including nonlinear elastic, elasto-plastic,
hypo, hyper and visco-elastic including rate and temperature dependencies.
Element Library (Ch. 11) discusses the MSC Nastran element library which consists of numerous elements that can be
categorized as scalar, one dimensional (1-D), two-dimensional (2-D), or three- dimensional (3-D). This includes
deformable elements as well as R-Type Constraint Elements, which are elements that impose fixed constraints between
components of motion at the grid points or scalar points to which they are connected. The last part of the chapter covers
modeling guidelines and techniques such as element choice, density and transition.
Rigid Elements, Connectors, and Multi-Point Constraints (Ch. 12) discusses Rigid Elements, Connectors and Multi-
Point Constraints. These techniques are used to make special connections between parts of the model that model
specific physical connections such as welds, hinges, or moment-carrying joints.
Advanced Topics (Ch. 13) discusses restarts, brake squeal analysis, and user defined services and subroutines
Main Index
22 Nonlinear Users Guide
MSC Software Products
Main Index
Chapter 1: Introduction 23
MSC Software Products
The MSC Nastran Demonstrations Problems Manual provides example problem and includes description of the input,
procedures, and results information that relates to the practical use of the MSC Nastran inputs, element libraries, and
loads and boundary conditions.
Below is a complete list of MSC Nastran documents. You may find any of these documents from MSC Software at
http://simcompanion.mscsoftware.com/infocenter/index?page=home.
Users Guides
Getting Started
Linear Static Analysis
Dynamic Analysis
MSC Nastran Demonstration Problems
MSC Nastran Embedded Fatigue Users Guide
Superelements
Design Sensitivity and Optimization
Nonlinear User's Guide (SOL 400)
Implicit Nonlinear (SOL 600)
Explicit Nonlinear (SOL 700)
Utilities Guide
User Defined Services
Thermal Analysis
Aeroelastic Analysis
Main Index
24 Nonlinear Users Guide
MSC Software Products
Patran Documentation
Three key books from the Patran library may be of assistance in running MSC Nastran Nonlinear:
Patran Users Guide this introductory guide gives you the essential information you need to immediately
begin using Patran for MSC Nastran Nonlinear projects. Understanding and using the information in this
guide requires no prior experience with CAE or finite element analysis.
Patran Reference Manual a counterpart to the MSC Nastran Reference Manual, this manual provides
complete descriptions of basic functions in Patran, geometry modeling, finite element modeling, material
models, element properties, loads and boundary conditions, analysis, and results.
MSC Nastran Preference Guide gives specific information that relates to using Patran with MSC Nastran as
the intended analysis code. All application forms and required input are tailored to MSC Nastran.
Main Index
Chapter 1: Introduction 25
MSC Nastran Implicit Nonlinear (SOL 400)
Nonlinear Analysis
Linear analysis assumes a linear relationship between the load applied to a structure and the response of the structure.
The stiffness of a structure in a linear analysis depends on its initially undeformed state. Linear static problems are
solved in one step, by a single decomposition of the stiffness matrix. A number of important assumptions and
limitations are inherent in linear static analysis. Materials behavior is such that the stress is directly proportional to
strain (linear) and to loads that do not take the material beyond its permanent yield point (the material remains elastic).
Linear analysis is restricted to small displacements; otherwise, the stiffness of the structures changes and must be
accounted for by regenerating the stiffness matrix. Lastly, loads are assumed to be applied slowly as to keep the
structure in equilibrium.
It becomes necessary to consider nonlinear effects in structures when modeling materials with nonlinear behavior and
where large deformations (rotations and/or strains) occur. In addition, contact problems exhibit nonlinear effects due
to changes in boundary conditions.
In a nonlinear problem, the stiffness of the structure depends on the displacement, and the response is no longer a linear
function of the load applied. As the structure displaces due to loading, the stiffness changes, and as the stiffness
changes, the structures response changes. As a result, nonlinear problems require incremental solution schemes that
divide the problem up into steps calculating the displacement, then updating the stiffness. Each step uses the results
from the previous step as a starting point. As a result, the stiffness matrix must be generated and decomposed many
times during the analysis adding time and costs to the analysis.
Nonlinear problems present many challenges. A nonlinear problem does not always have a unique solution.
Sometimes, a nonlinear problem does not have any solution, although the problem can seem to be defined correctly.
Nonlinear analysis requires choosing a solution strategy which includes dividing the loading into logical steps,
controlling the numerical processing, and planning for the possibility of changing the solution strategy during the
analysis using restarts. Which solution method to use depends on the structure itself, the nature of the loading, and the
anticipated nonlinear behavior. In some cases, one method can be advantageous over another; in other cases, the
converse might be true.
If a solution is obtainable, there is also the issue of efficiency. Each solution procedure has pros and cons in terms of
matrix operations and storage requirements. In addition, a very important variable regarding overall efficiency is the
Main Index
26 Nonlinear Users Guide
MSC Nastran Implicit Nonlinear (SOL 400)
size of the problem. The time required to assemble a stiffness matrix, as well as the time required to recover stresses
after a solution, vary roughly linearly with the number of degrees of freedom of the problem. On the other hand, when
using a direct solver the time required to go through the solver varies roughly quadratically with the bandwidth, as well
as linearly with the number of degrees of freedom.
Main Index
Chapter 1: Introduction 27
Capabilities
Capabilities
The most important capabilities of SOL 400 are presented in the following list.
1. SOL 400 solves linear and nonlinear (material, contact and/or geometric) static, heat transfer, modal
(vibration), buckling, and transient dynamic structural finite element problems.
2. Eigenvalue solutions are available in SOL 400 for solving linear or nonlinear modal analyses and linear
buckling analysis using either Lanczos or Inverse Power Sweep methods of iteration. Through the use of
parameters you can control the convergence of the eigenvalues, and the modes to retain.
3. MSC Nastran has advanced numerical techniques including parallelization to support very large models.
4. SOL 400 supports the following elements types:
Scalar Elements
Beams
Shells
2-D Plane Strain
2-D Plane Stress
Axisymmetric
3-D Solids
Lower-order Elements
Higher-order Elements
5. RBE elements and multi-point constraint equations are supported in SOL 400 to tie specific nodes or degrees
of freedom to each other. Special MPC entities are supported, (e.g. rigid links) which can be used to tie two
nodes together or equate the motion of two degrees of freedom. Both small and large rotations are supported.
6. SOL 400 supports the following loads and boundary conditions:
Constrained nodal displacements (zero displacements at specified degrees of freedom). Enforced nodal
displacements (nonzero displacements at specified degrees of freedom in the nodal coordinate system).
Both constrained and enforced displacements can be specified as relative or absolute.
Forces applied to nodes in any coordinate system.
Follower force distributed loads.
Temperature applied to nodes. Temperature can be applied as a load in a structural analysis. The reference
temperature is user definable.
Inertial body forces, acceleration and velocity can be applied in the global coordinate system.
Contact between two bodies can be defined by selecting the contacting bodies and defining the contact
interaction properties. Gluing and ungluing are provided. Enforced motion or velocity of rigid contacts
surfaces is available.
Main Index
28 Nonlinear Users Guide
Capabilities
7. SOL 400 supports both temperature independent and dependent, isotropic, orthotropic, and anisotropic material
properties. They can be defined for elastic, elastic-plastic, hyper-elastic, hypo-elastic, visco-elastic, and creep
constitutive models. In addition, SOL 400 also supports cohesive material, gasket material, and thermo-
mechanical shape memory material. Nonlinear elastic-plastic materials can be defined by specifying piecewise
linear stress-strain curves, which may be temperature and / or rate dependent.
8. Physical properties can be associated with SOL 400 elements such as the cross-sectional properties of the beam
element, the area of the beam and rod elements, the thickness of shell, plane stress, plane strain, and membrane
elements, spring parameters, and point masses among others.
9. Fracture Mechanics capabilities include VCCT crack propagation and cohesive zone interface and closure
analysis, and a large number of failure index criterion for analyzing delamination of composite elements.
10. Laminated composite solid and shell elements are supported in SOL 400 through the PCOMP, PCOMPG, and
PCOMPLS entries of the materials capability. Each layer has its own material, thickness, and orientation and
may represent linear or nonlinear material behavior. Failure index calculations are also supported. Fast
integration techniques are available with the PCOMPF entry. Equivalent material models may be incorporated
using PSHELL.
11. Analysis jobs may consist of (possibly) complicated loading histories (such as would occur in a multi-step
manufacturing process). A single SOL 400 analysis (subcase) may consist of multiple steps that specify the
loading sequence.
Main Index
Chapter 1: Introduction 29
Using the Finite Element Method (FEM)
Main Index
30 Nonlinear Users Guide
Clevis Pin Lug Design
Figure 1-1 Clevis Pin Lug Assembly with a Lug Base Load
Main Index
Chapter 1: Introduction 31
Clevis Pin Lug Design
Required Solution
Returning to the clevis-pin-lug structure, assume that the primary design criteria are to have an adequate margin- of-
safety on yield, or possibly estimate the stresses for the purpose of fatigue life calculation. The load for this structure
consists of a force applied to the lug at the center.
A good first step in any analysis is to identify the output quantities that you are trying to determine. In the example,
the output quantities are the stresses and interaction stresses or forces on each of the components. Using a summation
of forces, it might be possible to estimate whether the normal forces and stresses at the points of contact seem
reasonable. Once you have an estimate, you can identify modeling errors quickly. For many engineering problems, you
cannot look in a handbook for a formula. However, you can always make some simplifying assumptions and obtain a
crude estimate. Having a good estimate, simplifies the model checkout.
This application example evaluates the performance of contact algorithms at curved boundaries between deformable
bodies. A cylindrical pin is located in the cylindrical hole of a lug-clevis assembly. The diameters of the hole and the
pin are slightly different. Point forces are applied to the center of the top surfaces of the lug and the base of the clevis
is assumed to be clamped down. It is assumed that the tangential (frictional) contact forces are negligible; e.g., can be
ignored for this purpose. Due to the symmetry condition, a half model of the assembly would be sufficient for the finite
element analysis if we assumed that the loading was symmetrical, however we will not make that assumption for this
model.
The displacement components and contact normal and tangential forces are of interest. One analysis is conducted with
SOL 400 with standard CTETRA elements. In the current version of SOL 400, the advanced large strain elements are
defined by a PSOLID entry pointing to an auxiliary PSLDN1 entry.
Now you are ready to generate the finite element model. At this point, you need to decide what kind of elements should
be used and how much detail is needed.
To model the structure with CTETRA elements, mesh each of the components using Patran or other meshing program.
Each of the 3 components is represented as a mesh of CTETRA elements. The CTETRA elements are connected to their
neighbor elements at points known as grid points. The physical locations of the grid points determine the quality of
each of the CTETRA elements. The quality of the mesh will be described in a later chapter. Associated with each set
of CTETRA elements is a property entry with the material (MAT1) and element formulation characteristics (PSOLID +
PSLDN1). The properties represent the physical properties of the elements and must be entered for each set of elements
with distinct characteristics. The material properties, consisting of the modulus of elasticity and Poisson's ratio, are
entered on a MAT1 material entry.
This concept of representing the real world structure with finite elements is generally referred to as meshing. Hence,
you can say that the pin - lug - clevis structure is meshed with CTETRA elements. For most models, meshing is
performed with the use of a preprocessor such as Patran. The relative number of elements used to represent the
structure or part of a structure is commonly referred to as mesh density. As the mesh density is increased, more
elements are needed and the size of the elements is decreased. The process of increasing the mesh density to improve
the accuracy of the results is also referred to as refining the mesh.
Main Index
32 Nonlinear Users Guide
Clevis Pin Lug Design
FEM Solution
Numerical solutions have been obtained with MSC Nastran solution sequence 400 for the 3-D case. First, the advanced
3-D elements are used to conduct the analysis with contact and friction. In comparison, the same analysis is also
conducted with the standard 3-D solid elements.
The contact, material/geometry, solution/convergence schemes and other parameters are explained below.
Contact Parameters
The element mesh using the 3-D solid element is shown in Figure 1-1. The contact body named as Pin represents the
pin. The contact bodies named as Clevis and Lug define those components. A point load (red arrow) is applied at the
center point of top end of the pin. It should be noted that the symmetry has been taken into consideration.
In the input data file, the contact bodies are defined as below:
$ Deform Body Contact LBC set: clevis
BCBODY 1 3D DEFORM 5 2 -1
ADVANCE 1061
BSURF 5 100001 100002 100003 100004 100005 100006 100007
$ Deform Body Contact LBC set: lug
BCBODY 2 3D DEFORM 6 2 -1
ADVANCE 1061
BSURF 6 200001 200002 200003 200004 200005 200006 200007
$ Deform Body Contact LBC set: pin
BCBODY 3 3D DEFORM 7 2 -1
ADVANCE 1061
BSURF 7 300001 300002 300003 300004 300005 300006 300007
The BCBODY with ID 1 defines the clevis as a three-dimensional deformable body. The BCBODY with ID 2 defines
the lug, and ID 3 the pin, also as a three-dimensional deformable bodies.
The BCTABLE bulk data entries shown below define the touch conditions between the bodies:
BCTABLE 1 3
SLAVE 1 0. 0. 0. 0. 0
1 3 0
FBSH 1.+20 .99 0.
MASTERS 2
SLAVE 1 0. 0. 0. 0. 0
1 3 0
FBSH 1.+20 .95 0.
MASTERS 3
SLAVE 2 0. 0. 0. 0. 0
1 3 0
FBSH 1.+20 .95 0.
MASTERS 3
As shown above, BCTABLE with ID 1 is used to define the contact touching conditions for the first step of the analysis.
A similar BCTABLE with ID 0 is written out by Patran to define the contact touching conditions at the start of the
analysis. Zero (0) identifies the case number. The BCTABLE entry is mandatory for the contact analysis with SOL 400.
Also, the options (BCONTACT with ID 0 and BCPARA with ID 0) are all applied at the start of the analysis that are
required in the case of initial stress free contact, if there is rigid body positioning required, or in the case where there
Main Index
Chapter 1: Introduction 33
Clevis Pin Lug Design
is permanent glued contact specified. For the loading analysis defined as load case 1 under the case control section, the
contact touching conditions are redefined by options of BCTABLE, BCPARA, and BCONTACT with ID 1. In this
example, the BCPARA is only defined once because the parameters specified herein are applied through the analysis
from the beginning unless some of these parameters are redefined by BCTABLE entry with ID 1. It should be mentioned
is that the BIAS parameter is defined as 0.99 (the default value is 0.9).
BCPARA 0 NLGLUE 0
BIAS .99
Material/Geometry Parameters
The both bodies in this analysis are defined as isotropic elastic materials (steel). The Young's modulus and Poisson
ratio are defined as:
MAT1 1 3.+7 .3 .1
As shown above, the material ID is 1.
Main Index
34 Nonlinear Users Guide
Clevis Pin Lug Design
BCPARA 0 NLGLUE 0
PARAM LGDISP 1
NLSTEP 1 1.
GENERAL 10 1 10
ADAPT .01 1.-5 .5 4 1.2 0
6 2.-4
MECH PV PFNT
.2
$ Direct Text Input for Bulk Data
$ Elements and Element Properties for region : solids
PSOLID 1 1 0
PSLDN1 1 1
C10 SOLID Q
The element type is defined by the PSOLID and PSLDN1 bulk data options as shown above where (C10 SOLID Q)
defines the 3-D tet10 continuum solid element with quadratic integration scheme.
Results
Numerical solutions have been done with MSC Nastran SOL 400. As seen in Figure 1-2, a relatively coarse mesh is
used for the lug and clevis and a finer mesh is used for the pin. Since curvature based meshing is used by default in
Patran the mesh will be finer near the curved parts of the components.
The overall solution results shown in the above figure represent the von mises stress, which is an appropriate stress
measure (measures how close the material is to yield) for crystaline isotropic materials like metals.
For the comparison, similar plots of the von mises stresses in each component are shown in the following plot. Again,
the element type is defined by PSOLID including PSLDN1 option which defines large displacement, large strain
elements.
Main Index
Chapter 1: Introduction 35
Clevis Pin Lug Design
One of the primary items of interest in a contact analysis is the contact stresses between the components. No friction
was used is this simulation since very little relative motion is expected and the normal forces would be much larger
than any frictional component of stress. Plots of the contact normal stresses (both contact normal stress and contact
normal forces are available) are shown in the following plots. Mesh density is very important to getting a smooth
accurate stress contour plot. For this problem, the load balancing requirement imposed by the equilibrium requirement
requires that the overall stresses be reasonably accurate, but if the goal of this simulation is to get an accurate stress
plot of the contact stresses, a much finer mesh is recommended.
Main Index
36 Nonlinear Users Guide
Clevis Pin Lug Design
The resulting contact normal nodal stress plots are shown in Figure 1-4. The peak value in the contact normal stress is
found to be in the neighborhood of 30 KSI which is in the range where local yielding would be of concern for many
metals.
Main Index
Chapter 1: Introduction 37
General Analysis Tips
It is your responsibility to choose the appropriate element types and number of elements to obtain the best results. If
you use a one-dimensional element to model a two-dimensional plate structure, you can expect poor results. If, on the
other hand, you choose a two-dimensional element type for your plate structure, the approximation will be
considerably better. This choice requires engineering judgment. It is easy to use a graphics preprocessor to generate a
good-looking model of your structure that contains many elements, but the model can be a poor model for the analysis.
This problem is further compounded by the fact that today's postprocessors can take the most meaningless results and
present them in a very convincing manner. To gauge the fidelity of the stress distribution, it is often useful to plot the
gauss point stresses extrapolated to the nodes (for the purpose of the fringe plot), but unaveraged so the variations of
the stress contributions from each of the elements connected to the node can be assessed. The topic of fringe plot value
averaging and extrapolation will be discussed further in Chapter 7: Results Output from the Analysis.
Main Index
38 Nonlinear Users Guide
Frequently Asked Questions
Main Index
Chapter 1: Introduction 39
Frequently Asked Questions
and let the graphical pre- postprocessor write out the correct input. See Chapter 6: Setting Up, Monitoring, and
Debugging the Analysis for an example of the process required to convert a SOL 101 model to a SOL 400 model.
Q7: Is it possible to take an existing Abaqus model and convert it into a SOL 400 model? If so what is the best
way to do this? Will I get the same answer?
Beginning with the 2014 release, there is a translator that can be used to convert an Abaqus input file into an MSC
Nastran input file. Prior to that, the best strategy is to read the file into Patran, change the preference to MSC Nastran,
and add the additional data.
Getting the same answer depends on a lot of factors, including the compatibility of the element types selected, the
analysis capabilities and algorithms (such as contact algorithms), and the analysis procedures and controls selected. In
particular, characterization of contact interactions may affect the results.
Q8: Does Patran support SOL 400?
While Patran does not support 100% of SOL 400 capabilities, it provides fairly complete support of the most
commonly used features of SOL 400 including nonlinear materials, contact, composites, and multi
stepping/perturbation analysis. MSC is continuing to develop Patran to support new SOL 400 capabilities.
Q9: Does SimXpert support SOL 400?
The answer to the previous question also applies to SimXpert.
Q10: If I am new to SOL 400 what's the best way to learn it?
The MSC documentation system has a complete set of solved SOL 400 problems in a document called the MSC
Nastran Demonstration Problems. The input files from these solved problems are included in the documentation
system. Also, Patran has a set of demo problems that can be found under the Analysis menu, under the "Run a Demo"
tab. Once these demos are run, Patran can be used to interrogate the model and investigate how the models are set up.
Finally, the SimCompanion website has a knowledge base with answers to commonly asked questions, as well as links
to the MSC training courses on SOL 400.
Q11: Where can I find example input files for SOL 400 demo problems?
Example problems can be found in the MSC Nastran Demonstration Problems Manual. These example problems also
include input files. There is also a whole test suite of example input files in the MSC Nastran test problem library. Also,
the "Run a Demo" problems in Patran will leave the input files that were used to run the example.
Q12: What kind of results are supported in the op2, xdb, HDF5, and/or dball/master?
While you can request SOL 400 output in any valid MSC Nastran output format, only the MASTER/DBALL and new
OP2 (PARAM,POST,1) and HDF5 formats will contain all of the results data blocks of SOL 400. Specifically, the
nonlinear stress/strain and contact results data blocks will be missing if any other format is requested. Thus, MSC
strongly recommends that you use MASTER/DBALL, new OP2,or HDF5 output formats for SOL 400 solutions.
Q13: What do I do if I run the analysis and do not get a solution? What if it does not converge?
The first thing to do is to check the solution files (*.f06, *.log, *.f04) to look for error messages. Typically there
will be a message telling why the solution did not converge.
Main Index
40 Nonlinear Users Guide
Frequently Asked Questions
There are a couple of answers to this question depending on why the analysis did not get a solution. Reasons why an
analysis does not result in a solution can be grouped into two categories. The first group relates to having a correctly
formatted input file. These tend to be the most common, and also the easiest to fix. The second group relates to having
what might be called a well-posed problem. This means that our problem does not violate any of the laws of physics,
and that there actually is a valid numerical solution to the problem we are trying to solve. A typical example of this
type of error would be trying to come up with a static solution to a problem that is not statically determinate. There
are methods used to determine if a model is properly constrained, such as running a modal analysis to look for
unconstrained rigid body modes.
Q14: How do I know if I NEED to do a nonlinear analysis?
While this is not an exact answer as loading conditions will affect this, in general, if your strains are approaching 5%,
or the deflection of any node in your model approaches 5% of the smallest dimension, you should do a nonlinear
analysis.
Q15: How do I know if linear contact will give me the right answer, or if I need to do a nonlinear analysis?
The terms linear and contact are almost an oxymoron. The very nature of contact problems are that finite deformation
of the nodes must be tracked to determine if the notes are in contact or not in contact. This deformation is almost always
large enough to invalidate linear contact solutions. The best use for linear contact is to use it to connect dissimilar
meshes or full assembly modeling where no separation occurs.
Q16: What is the difference between linear buckling analysis and nonlinear buckling analysis?
A linear buckling analysis is based on the un-deformed configuration of the structure. A nonlinear buckling analysis
is based on a deformed shape of the structure. The deformation and stresses make the structure stiffer (if in tension)
and cause the natural frequencies to increase. This is why it is often necessary to do a nonlinear buckling analysis to
get accurate mode shapes and frequencies.
Q17: What is a "stress stiffened model " analysis?
Before extracting the eigen-values from the structural system, the model may be pre-loaded prior to the modal
extraction. Initial stress effects are then included in the stiffness which tend to raise the eigenvalues. This effect tends
to affect the eigenvalues more than the eigenvectors (mode shapes).
Q18: Will RBE's rotate with the rest of the model?
If a large displacement, nonlinear analysis is performed, then yes, the RBEs rotates with the model. In a linear analysis,
the displacement and rotation of the nodes is assumed to be infinitesimally small.
Q19: What are Marc elements; also called Advanced nonlinear elements or large strain elements? When would
I use them? How do I activate them? Is there any penalty for always using Marc elements instead of MSC
Nastran elements?
Most MSC Nastran elements were originally formulated for linear analysis. At some point, some of those original
elements were modified for use in nonlinear analysis (SOLs 106 and 129), but certainly not all of them. Rather than
develop new elements and capabilities from scratch for SOL 400 MSC Nastran decided to copy the well-proven,
robust, mature large deformation/large strain analysis procedures and element formulations from Marc. Rather than
develop new nonlinear element designations, which would've made it a lot of work to convert a linear SOL 101 model
to a SOL 400 nonlinear model, MSC Nastran decided to use the same element designations and simply allow you to
Main Index
Chapter 1: Introduction 41
Frequently Asked Questions
control the formulation used through secondary property entries such as PBEMN1, PSHLN1, PSHNL2, and PSLDN1.
While this method gives you direct control over the element formulation, it should be noted that SOL 400 automatically
selects the proper formulation for the problem. This means that if the problem is a large deformation, large strain
problem, SOL 400 automatically uses the large strain element formulation. Thus, unless you want to directly control
the formulation (such as use reduced integration or some other special formulation), you are best off allowing MSC
Nastran to select the formulation used, which also means no additional input is required.
Q20: Do shell and beam offsets rotate with the model?
Yes, if MDLPRM,OFFDEF,LROFF is set in the deck when parameter LGDISP>=0.
Q21: Will CBAR elements rotate with the model?
Yes, if CBAR is converted to a CBEAM.
Q22: Will CGAP elements rotate with the model?
No. Contact bodies should always be used for contact involving large deformations.
Q23: Will CBEAM pin flags rotate with the model?
No.
Q24: What is the difference between the Stress Tensor and the Nonlinear Stress Tensor quantities I see listed
in my Patran results?
There is no difference. The label "stress tensor" was used for linear analysis, and the label "nonlinear stress tensor" for
nonlinear analysis. But in fact, they should be same
Q25: Which stress quantity does MSC Nastran use when it looks up the stress/strain curve I provided?
For small strain elements (traditional MSC Nastran elements), engineering stress/strain is used, but for large strain
nonlinear element (new Marc elements), Cauchy stress is used.
Q26: Will the force I apply rotate with the model?
These are referred to as "follower forces" and they will be included if: 1) a nonlinear analysis procedure is used; 2) a
FORCE1 or FORCE2 entry is used for loading; and 3) and appropriate PARAM,LGDISP value is used. Only Follower
Forces applied by FORCE1, FORCE2, MOMENT1, MOMENT2 will rotate with model. Forces applied by RFORCE
will rotate according to the specified angular velocity/acceleration rules.
Q27: What if I don't want the force to rotate with the model?
Short answer: use the static force FORCE bulk data entry.
Q28: Will the pressure I apply to a surface rotate with the model?
Only Follower Pressure applied by PLOAD, PLOAD2, and PLOAD4 will rotate with model. See follower force
notes under the PLOAD and PLOAD4 entries in the QRG for more details.
Q29: Is it possible to create the input such that the pressure I apply to a surface will not rotate with the model?
Yes. With PLOAD4, if the load direction is given, the pressure applied will be fixed in the given direction and not rotate
with model. See remark 2 of PLOAD4 in the QRG:
Main Index
42 Nonlinear Users Guide
Frequently Asked Questions
Remark 2. The continuation entry is optional. If fields 2, 3, 4, and 5 of the continuation entry are blank, the load is
assumed to be a pressure acting normal to the face. If these fields are not blank, the load acts in the direction defined
in these fields. Note that if CID is a curvilinear coordinate system, the direction of loading may vary over the surface
of the element. The load intensity is the load per unit of surface area, not the load per unit of area normal to the direction
of loading.
Q30: If I am familiar with running linear MSC Nastran analyses but have never run a nonlinear SOL 400
analysis, how can I learn what I need to do differently?
Many important features and concepts in this book are illustrated with examples. As you read this guide, it is highly
recommended that you run the example problems and make variations of these problems. When learning a new MSC
Nastran feature, engineers all too often generate a large problem using several hundred or several thousand degrees of
freedom as a test case. This practice has become the norm in recent years with the advent of graphics preprocessors
and automatic meshing. Rarely is such a large model necessary to learn a new feature; in most cases, it just adds
unnecessary complexity. For this reason most of the examples in this book are small generally less than 100 degrees
of freedom. To facilitate the use of these example problems, example problems referenced in this guide are delivered
with MSC Nastran system under the MSC_DOC_DIR/doc/linstat with the extensions of .dat. It is recommended you
copy any example problem to your local directory, so you can see the files created, and you don't inadvertently create
files in the delivery directory.
Q31: How can I see a plot of load vs deflection for the rigid body that loaded my model?
Only available for load controlled rigid bodies, you can plot the results of the control grid of the rigid body.
Q32: How can I monitor the nonlinear solution to see if it is progressing toward convergence?
See Chapter 6: Setting Up, Monitoring, and Debugging the Analysis for more details, but the short answer is: monitor
the jobname.sts file with a text editor that automatically updates. As each load increment converges a new line
with relevant information will appear in the .sts file.
Q33: Is it possible to see the results of any of the intermediate loading steps before the analysis is complete?
Yes. It is possible to have MSC Nastran write out the results of converged increments to intermediate OP2 files. See
the INTERM option of the OUTCTRL parameter under the NLOPRM entry in the QRG.
Q34: How can I find out what the normal and frictional forces or stresses are between the contact bodies?
Use the BOUTPUT case control output request and these values will be in the MASTER/DBALL or new OP2 files along
with the other output requested.
Q35: Is an RBE spider the best way to fix a surface of my structure?
It depends on the constraints on the surface. If all the grids on the surface have the same behavior in some direction,
an RBE is easy to use. It is also easy to "glue" the area to a rigid surface and use the rigid surface controls. You may
also use SPC/SPC1.
Q36: What are alternatives to an RBE spider for displacement control of part of my structure?
SPC/SPC1/SPCD/SPCR may be combined together for any complicated displacement control. As mentioned in the
previous question, glueing the area to a rigid surface and using the rigid surface controls is also easy.
Main Index
Chapter 1: Introduction 43
Frequently Asked Questions
Q37: Will I get the same answer if I apply my loads and constraints using a glued-on rigid surface as I would
using an RBE spider or displacement constraint?
Yes, if both loading methods are done correctly you will get the same results.
Q38: How do a simulate a complex sequence of loading, such as the process of an engine head going through
the sequence of bolt tightening, then heating up?
Break the loading up into discrete loadcases. For more information on this process, see Chapter 3: Nonlinearity and
Analysis Types in this manual for more information on this topic.
Q39: How do I know if Patran supports a certain type of modeling process or feature?
The MSC Nastran Preference Guide for Patran describes all of the entities (including elements, materials, properties,
loads and constraints) that Patran support for use in MSC Nastran SOL 400.
Q40: What is the simplest way to include contact in my model?
If you simply create contact bodies (Patran has a tool under Tools - Modeling that automatically creates contact bodies
based on groups, materials, properties, etc.) and runs a SOL 400 nonlinear analysis, the contact interaction is included
by default, but does not include friction between the components. To include friction, you need to specify the friction
model and friction coefficients (you can specify a global model value, a value for each contact body, or for each contact
pair the most localized value, in the most general to most specific order just provided, will be used)
Q41: What Thermal Solutions are supported in SOL 400?
SOL 400 supports steady state and transient thermal analysis procedures (ANALYSIS = HSTAT and HTRAN),
updated to include thermal contact and coupled thermal-structural interaction, plus the Sinda RC network approach.
Chapter 3: Nonlinearity and Analysis Types has a description of the SOL 400 thermal analysis capabilities, plus a
description of the coupled multi-physics capabilities that are supported.
Q42: How do the RCNS & RCNT RC Network analysis options in SOL 400 work?
These RC analysis options run a subset of MSC's Sinda Solver. Standalone Sinda is a modern equivalent of
traditional Resistor-Capacitor type solver. Also called Finite Difference Lumped Parameter Network Solver.
Simulation can include various aspects of conduction, convection, radiation and with optional Fortran access
predefined or custom loading function can be specified. Using Patran with Sinda preference, or in this case from MSC
Nastran, an RC network is generated from traditional shaped elements. MSC Nastran implementation does not include
the Fortran access nor standard/custom loading function. Loading must be manually defined but the same effect can
be achieved. Additionally, MSC Nastran implementation does include automated radiation refinement techniques
available in Sinda.
Main Index
44 Nonlinear Users Guide
Frequently Asked Questions
Main Index
Chapter 2: MSC Nastran Files
Nonlinear Users Guide
Main Index
46 Nonlinear Users Guide
Introduction to Organization of MSC Nastran Files
Main Index
Chapter 2: MSC Nastran Files 47
Introduction to Organization of MSC Nastran Files
Element and
Material
Bulk Data Properties,
Fixed Displ,
Etc.
MSC Nastran Implicit Nonlinear
Load Incrementation,
Complete Input Deck
Applied Loads,
Case Control
Control Information
Applied Displacements
Etc.
Input Conventions
SOL 400 performs all data conversion internally so that the system does not abort because of data errors made by you.
The program reads all input data options alphanumerically and converts them to integer, floating point, or keywords,
as necessary. SOL 400 issues error messages and displays the incorrect option image if it cannot interpret the option
data field according to the specifications given in the manual. When such errors occur, the program attempts to scan
the remainder of the data file and ends the run with a FATAL ERROR message.
Two input format conventions can be used: fixed and free format. You can mix fixed and free format options within a
file.
The syntax rules for fixed fields are as follows:
Give floating point numbers with or without an exponent. If you give an exponent, it must be preceded by the
character E or D and must be right-justified. If data is double precision, a D must be used.
The syntax rules for free fields are as follows. See the Format of Bulk Data Entries in the QRG for more details.
Main Index
48 Nonlinear Users Guide
Introduction to Organization of MSC Nastran Files
Check that each option contains the same number of data items that it would contain under standard fixed
format control. This syntax rule allows you to mix fixed field and free field options in the data file because the
number of options you need to input any data list are the same in both cases.
Separate data items on a option with a comma. The comma can be surrounded by any number of blanks.
Within the data item itself, no embedded blanks can appear.
Give floating point numbers with or without an exponent. If you use an exponent, it must be preceded by the
character E or D and must immediately follow the mantissa (no embedded blanks).
Give keywords exactly as they are written in the manual.
All data can be entered as uppercase or lowercase text.
Small field format is limited to 8 columns per field. Large field is 16 columns.
For most bulk data entries, MSC Nastran does not make the distinction between zero and blank. Thus, if a zero is
entered and the default is some other value, the default is normally used. If you wish to use zero, enter a small number
such as 1.0E-12 instead.
Section Descriptions
The NASTRAN Statement
The NASTRAN statement is an optional statement used to specify system parameters that control internal solution
processing or provide specific diagnostics. The NASTRAN statement, if used, must be placed before the Executive
Control Statement. The NASTRAN statement is useful when you need to specify system-related parameters, such as
setting the BUFFSIZE for the I/O. For a description of the NASTRAN statement options, refer to the nastran Command
and NASTRAN Statement in the MSC Nastran Quick Reference Guide.
Executive Control
This group of entries provides overall job control for the problem and sets up initial switches to control the flow of the
program through the desired analysis.
The Executive Control Section contains statements that
Select the solution sequence to be run (e.g., statics, normal modes).
Request various diagnostic outputs.
Contain user-written DMAP. DMAP is an advanced topic and beyond the scope of this users guide.
Main Index
Chapter 2: MSC Nastran Files 49
Introduction to Organization of MSC Nastran Files
This set of input must be terminated with an CEND parameter. See The Executive Control Section Overview in this
chapter for more details. See Chapter 3: Executive Control Statements, Executive Control Section in the QRG for a
complete list and required syntax.
Case Control
This group of options provides the loads and constraints and load incrementation method and controls the program
after the initial elastic analysis. Case Control options also include blocks which allow changes in the initial model
specifications. Case Control options can also specify print-out and postprocessing options. In linear solution
sequences, including linear analyses in SOL 400, each set of load sets must be begin with a SUBCASE command and
be terminated by another SUBCASE or a BEGIN BULK command. If there is only one load case, the SUBCASE entry
is not required. In SOL 400, the STEP case control commands are required for the nonlinear and perturbation analyses
under each SUBCASE. In addition, the SUBSTEP case control command is used in SOL 400 for the coupling analyses
under each STEP. See The Case Control Section in this manual for more discussion. See Chapter 4: Case Control
Commands, The Case Control Section in the QRG for a complete list of case control options and syntax.
Main Index
50 Nonlinear Users Guide
Introduction to Organization of MSC Nastran Files
Main Index
Chapter 2: MSC Nastran Files 51
Introduction to Organization of MSC Nastran Files
Using Patran, you can run a MSC Nastran analysis or you may generate the MSC Nastran Input File to run externally.
For information on generating the MSC Nastran Input file from within Patran, see Chapter 3: Running an Analysis,
Analysis Form in the Patran Interface to MSC Nastran Preference Guide.
Patran finds the bulk data file with the current job name and displays the file for editing in a text editing window.
Main Index
52 Nonlinear Users Guide
Truss Example
Truss Example
The purpose of this section is provide an overview of the input; it is not meant to describe all of the options available
for static analysis. The next section provides an overview of the output files. The goal of these two sections is to show
how a typical MSC Nastran input file is used to represent a typical model.
To illustrate the form of input and output files, you can use the simple SOL 101 truss structure shown in Figure 2-1.
2 Y
A = 4.0 in
6 2
E = 30.0 10 lb in
1 3
4
J = 1.27 in 2
3
= 0.3 2
5 10 ft
Elements
1 X
Grid points
4 4
F =1000 lb
Pin joints
50 ft
The truss structure consists of five members, each with a cross-sectional area (A) of 4 in2 and a polar moment of inertia
(J) of 1.27 in4. The material is steel with a Youngs modulus E of 30 x 106 psi and a Poissons ratio of 0.3. All the
connections between the members are pin joints (i.e., they can transmit forces but not moments). The goal of the
analysis is to determine the displacement of the grid points and the stresses and forces within the members.
The MSC Nastran input file listing for the truss structure is given in Listing 2-2. (See
MSC_DOC_DIR/doc/linstat/truss1.dat.) Although the input file shown represents a fairly simple structure, it is typical
of all MSC Nastran input files. The purpose of the example is not to show how to perform an analysis or discuss the
specifics of the model, such as the grid points or elements. Rather, the purpose is to present the form of the input file
itself. Discussion of the actual commands and entries in the input file begins in Chapter 3: Grid Points, Scalar Points,
and Coordinate Systems in the MSC Nastran Linear Statics Analysis Users Guide.
Main Index
Chapter 2: MSC Nastran Files 53
Truss Example
$ FILENAME - TRUSS1.DAT
$
ID LINEAR,TRUSS1 Executive
SOL 101 Control
TIME 2 Section
CEND
TITLE = LINEAR STATICS USERS SAMPLE INPUT FILE
SUBTITLE = TRUSS STRUCTURE
LABEL = POINT LOAD AT GRID POINT 4 Case
LOAD = 10 Control
SPC = 11 Section
DISPLACEMENT = ALL
ELFORCE = ALL
ELSTRESS = ALL
BEGIN BULK
$
$ THE GRID POINTS LOCATIONS
$ DESCRIBE THE GEOMETRY
$
GRID 1 0. 0. 0. 3456
GRID 2 0. 120. 0. 3456
GRID 3 600. 120. 0. 3456
GRID 4 600. 0. 0. 3456
$
$ MEMBERS ARE MODELED USING
$ ROD ELEMENTS
$
CROD 1 21 2 3
CROD 2 21 2 4
CROD 3 21 1 3
CROD 4 21 1 4
CROD 5 21 3 4 Bulk Data
$ Section
$ PROPERTIES OF ROD ELEMENTS
$
PROD 21 22 4. 1.27
$
$ MATERIAL PROPERTIES
$
MAT1 22 30.E6 .3
$
$ POINT LOAD
$
FORCE 10 4 1000. 0. -1. 0.
$
SPC1 11 123456 1 2
$
ENDDATA
As mentioned in the previous sections, this input file consists of five distinct sections: the NASTRAN statement, the
File Management Section (FMS), the Executive Control Section, the Case Control Section, and the Bulk Data Section.
Input in the File Management Section and the Executive Control Section is referred to as a statement; input in the Case
Control Section is referred to as a command; and input in the Bulk Data Section is referred to as an entry.
Main Index
54 Nonlinear Users Guide
Truss Example
NASTRAN Statement
The truss example is so simple that a Nastran statement is not required.
Main Index
Chapter 2: MSC Nastran Files 55
Truss Example
Grid Entries
Elements use the grid
points for their
geometry.
Elements C
Type Entries
Element entries
specify property
entries.
Properties P
Type Entries
Property entries
specify material
entries.
Material M
Type Entries
Case Control Section
SPC
SPC Entries
Command
LOAD
LOAD
Entries
Command
Returning to our truss example, the members of the truss are modeled using rod elements. The MSC Nastran notation
for a rod is the CROD bulk data entry. The C preceding ROD denotes that the entry is for an element connection;
i.e., it connects grid points. This pattern is true for all elements, for example, the bulk data entry notation for a element
connection of a QUAD4 plate element is CQUAD4. This notation also makes it convenient when trying to find a
description of a particular element in the MSC Nastran Quick Reference Guide since they are in alphabetical order
under the Cs.
Main Index
56 Nonlinear Users Guide
Truss Example
All elements in the input file must have a unique ID with respect to all the other elements. The IDs of the CROD
elements in the truss example are 1, 2, 3, 4, and 5.
Each of the CROD entries refers to the PROD entry 21 (Field 3), which is the property definition of the CROD elements.
The PROD entry specifies a material ID of 22 (field 3), the cross-sectional area of 4.0 in2, and the torsional stiffness
coefficient of 1.27 in4. All property entries begin with P.
The material ID of 22 refers to a MAT1 entry. The MAT1 entry defines an isotropic material. The MAT1 entry is defined
formally in Material Properties and Composites (Ch. 5) in the MSC Nastran Linear Statics Analysis Users Guide; but
for now, it is sufficient to say that an isotropic material requires only two constants to fully describe the material. In
this case the material property is described as having a Youngs modulus of 30 x 106 psi and a Poissons ratio of 0.3.
All material entries start with M.
The FORCE and SPC entries, as mentioned previously, are included in the analysis because they were called out in the
Case Control Section by their ID number. The FORCE entry in this example specifies a point load of 1000 pounds
acting at grid point 4 in the -Y direction. The SPC1 entry specifies that all six degrees of freedom are constrained at
grid points 1 and 2.
If you have not performed an MSC Nastran run before, you are encouraged to copy the truss model truss1.dat into
your working directory and submit it as an MSC Nastran job. See MSC_DOC_DIR/doc/linstat/truss1.dat.
Main Index
Chapter 2: MSC Nastran Files 57
Overview of the MSC Nastran Output Files
Table 2-1 Files Resulting from Submitting the Job Named run1.dat
run1.dat The input file describing the model, the type of solution, the output requests, etc.
Generated with a text editor or preprocessor.
run1.f06 The main output file containing the printed output such as displacements, element
forces, stresses, etc.
run1.f04 A history of the assigned files, disk space usage, and modules used during the analysis.
Useful for debugging.
run1.log A summary of the command line options used and the execution links.
run1.DBALL A database containing the input file, assembled matrices, and solutions. Used also for
restarting the run for additional analysis or output.
run1.MASTER The file containing the master directory of the files used by the run and the physical
location of the files on the system. This file is also needed for a restart run.
run1.IFPDAT The file containing the bulk data entries for IFP restart processing
run1.plt Contains the plot information requested by the NASTRAN plotter commands specified
in input file.
run1.pch Contains the punch output as requested in the input file.
run1.op2 Output database used for postprocessing of the results.
run1.h5 Results database in HDF5 format.
run1.sts Contains nonlinear statistics, if linear contact is utilized.
miscellaneous scratch files Several scratch files are generated during the analysis which MSC Nastran
automatically deletes upon completion of the run.
Print Files
The print file jobname.f06 file contains a complete text output of solution information, including an input
summary, solution diagnostics and results if requested.
Because of the potential size of the print file, certain information is optional. Instead of printing out a complete echo
of the input deck, a summary can be printed. You can select which stresses and strains are included in the output file.
Main Index
58 Nonlinear Users Guide
Overview of the MSC Nastran Output Files
A complete listing of the truss1.f06 output file (minus the page numbers and date) is shown in Figure 2-3. This is
one of the few occasions that an output file is shown in its entirety; only selected portions of the output will be shown
in the remaining examples in this users guide. Historically, the .f06 suffix is used because the program is writing
this file to FORTRAN unit 6.
Many of the .f06 output formats are shown in the Ap. A: Sample Output Listings in the MSC Nastran Linear Statics
Analysis Users Guide with notes, so it is a good referenced for understanding your output requests.
Status File
SOL 400 produces a status file jobname.sts that is very useful for assessing if the analysis has completed
successfully, and, if not, how much of it was completed before the analysis was terminated. This is extremely useful
because the most common problem with obtaining nonlinear solutions is avoiding, detecting, and managing
convergence problems.
Message Files
The message files jobname.msg (if it is run from Patran) contain diagnostic error and warning messages output by SOL
400. The message file is the best way to immediately check an analysis for successful execution if the job is run from
Patran. Otherwise, check JID.f04 and JID.f06.
SOL 400 has two levels of messages:
1. Exit number messages;
2. User Fatal Error Message (UFM)
Main Index
Chapter 2: MSC Nastran Files 59
Overview of the MSC Nastran Output Files
Output2 Format
The Patran's MSC Nastran interface supports several different OUTPUT2 file formats. The interface, running on any
platform can read a binary format OUTPUT2 file produced by MSC Nastran running on any of these same platforms.
Patran may be able to read binary format OUTPUT2 files from other platforms if they contain 32 bit, IEEE format
entities (either Big or Little Indian).
For platforms that do not produce OUTPUT2 files in these formats, Patran's MSC Nastran preference can read
OUTPUT2 files created with the FORM=FORMATTED option in MSC Nastran. This option can be selected from the
Analysis/Translation Parameters form in Patran and directs MSC Nastran to produce an ASCII format OUTPUT2 file
that can be moved between any platforms. The Patran MSC Nastran interface detects this format when the OUTPUT2
file is opened, automatically converts it to the binary format, and then reads the model and/or results into the Patran
database.
Main Index
60 Nonlinear Users Guide
Overview of the MSC Nastran Output Files
N A S T R A N E X E C U T I V E C O N T R O L E C H O
ID LINEAR,TRUSS1
SOL 101
TIME 2
CEND
C A S E C O N T R O L E C H O
COMMAND
COUNT
1 TITLE = LINEAR STATICS USER'S GUIDE
2 SUBTITLE = TRUSS STRUCTURE
3 LABEL = POINT LOAD AT GRID POINT 4
4 LOAD = 10
5 SPC = 11
6 DISPLACEMENT = ALL
7 ELFORCE = ALL
8 ELSTRESS = ALL
9 BEGIN BULK
INPUT BULK DATA ENTRY COUNT = 33
S O R T E D B U L K D A T A E C H O
ENTRY
COUNT . 1 .. 2 .. 3 .. 4 .. 5 .. 6 .. 7 .. 8 .. 9 .. 10 .
1- CROD 1 21 2 3
2- CROD 2 21 2 4
3- CROD 3 21 1 3
4- CROD 4 21 1 4
5- CROD 5 21 3 4
6- FORCE 10 4 1000. 0. -1. 0.
7- GRID 1 0. 0. 0. 3456
8- GRID 2 0. 120. 0. 3456
9- GRID 3 600. 120. 0. 3456
10- GRID 4 600. 0. 0. 3456
11- MAT1 22 30.E6 .3
12- PROD 21 22 4. 1.27
13- SPC1 11 123456 1 2
ENDDATA
TOTAL COUNT= 14
M O D E L S U M M A R Y
NUMBER OF GRID POINTS = 4
NUMBER OF CROD ELEMENTS = 5
*** USER INFORMATION MESSAGE 7310 (VECPRN)
ORIGIN OF SUPERELEMENT BASIC COORDINATE SYSTEM WILL BE USED AS REFERENCE LOCATION.
RESULTANTS ABOUT ORIGIN OF SUPERELEMENT BASIC COORDINATE SYSTEM IN SUPERELEMENT BASIC SYSTEM COORDINATES.
OLOAD RESULTANT
SUBCASE/ LOAD
DAREA ID TYPE T1 T2 T3 R1 R2 R3
1 FX 0.000000E+00 ---- ---- ---- 0.000000E+00 0.000000E+00
FY ---- -1.000000E+03 ---- 0.000000E+00 ---- -6.000000E+05
FZ ---- ---- 0.000000E+00 0.000000E+00 0.000000E+00 ----
MX ---- ---- ---- 0.000000E+00 ---- ----
MY ---- ---- ---- ---- 0.000000E+00 ----
MZ ---- ---- ---- ---- ---- 0.000000E+00
TOTALS 0.000000E+00 -1.000000E+03 0.000000E+00 0.000000E+00 0.000000E+00 -6.000000E+05
F O R C E S I N R O D E L E M E N T S ( C R O D )
ELEMENT AXIAL ELEMENT AXIAL
ID. FORCE TORQUE ID. FORCE TORQUE
1 2.495156E+03 0.0 2 2.554449E+03 0.0
3 -2.544570E+03 0.0 4 -2.504844E+03 0.0
5 4.990313E+02 0.0
S T R E S S E S I N R O D E L E M E N T S ( C R O D )
ELEMENT AXIAL SAFETY TORSIONAL SAFETY ELEMENT AXIAL SAFETY TORSIONAL SAFETY
ID. STRESS MARGIN STRESS MARGIN ID. STRESS MARGIN STRESS MARGIN
1 6.237891E+02 0.0 2 6.386123E+02 0.0
3 -6.361426E+02 0.0 4 -6.262109E+02 0.0
5 1.247578E+02 0.0
Main Index
Chapter 2: MSC Nastran Files 61
Overview of the MSC Nastran Output Files
After the input file echo, the OLOAD RESULTANT is printed which shows the resultant of the loads about the origin
of the basic coordinate system in this case. Next is the USER INFORMATION MESSAGE (UIM) 5293 which
shows the epsilon and external work for the run. The epsilon is a measure of the numeric conditioning of the model,
while the external work is the work due to the applied loads. Both of these values are useful for model checkout and
are discussed in more detail in Model Verification (Ch. 10) in the MSC Nastran Linear Statics Analysis Users Guide.
In the truss example, output is requested for displacements, element forces, and element stresses. The displacement
request is the only grid point output. Element forces and stresses are element output. The last summary is the maximum
of the applied loads. After the summaries, the displacement for all of the grid points, the element forces, and the
element stresses is printed as requested in the Case Control Section.
For static analysis of the truss model, just one applied load is used. You can, however, specify multiple loading
conditions to be analyzed in a single input file if desired. If you do specify multiple loading conditions, the OLOAD
RESULTANT and EXTERNAL WORK are shown for each selected loading conditions. The external work is calculated
as . In the case of enforced displacement, the equivalent load is calculated as the constraint force required to impose
the enforced displacement. The external work print-out in this case is meaningless.
The information in the output file is extremely important for model checkout and debugging. Make a habit of reviewing
these diagnostics for all of your runs. Experiment with the truss1.dat file; run the input file and view truss1.f06
in your text editor.
After this overview of the MSC Nastran input and output files, the next three sections contain the Executive Control
Section, Case Control Section, and Bulk Data Section. The File Management Section is discussed in Database
Management (Ch. 15) in the MSC Nastran Linear Statics Analysis Users Guide.
Main Index
62 Nonlinear Users Guide
SOL 400 Output
Main Index
Chapter 2: MSC Nastran Files 63
SOL 400 Output
Results
Like the enormous amount of data needed to define the simulation model to an analysis code, there is a large volume
of data returned from the simulation analysis. And just as it is virtually impossible to construct a model with a text
editor alone, it is equally as difficult to read and interpret the results by hand. Using a postprocessor with a graphical
user interface such as Patran or SimXpert is highly recommended.
The Patran Results application gives you control of powerful graphical capabilities to display results quantities in a
variety of ways:
Deformed structural plots
Color banded fringe plots
Marker plots (vectors, tensors)
Freebody diagrams
Graph (XY) plots
Animations of most of these plot types.
The Results application treats all results quantities in a very flexible and general manner. In addition, for maximum
flexibility results can be:
Sorted
Reported
Filtered
Derived
Deleted
All of these features help give meaningful insight into results interpretation of engineering problems that would
otherwise be difficult at best.
The Results application is object oriented, providing postprocessing plots which are created, displayed, and
manipulated to obtain rapid insight into the nature of results data. The imaging is intended to provide graphics
performance sufficient for real time manipulation. Performance will vary depending on hardware, but consistency of
functionality is maintained as much as possible across all supported display devices.
Capabilities for interactive results postprocessing also exist. Advanced visualization capabilities allow creation of
many plot types which can be saved, simultaneously plotted, and interactively manipulated with results quantities
reported at the click of the mouse button to better understand mechanical behavior. Once defined, the visualization
plots remain in the database for immediate access and provide the means for results manipulation and review in a
consistent and easy to use manner.
For more information, see Chapter 1: Introduction to Results Postprocessing in the Patran Reference Manual.
Main Index
64 Nonlinear Users Guide
Parameter Discussion
Parameter Discussion
Parameters are used throughout MSC Nastran to provide you with controls over many aspects of the solution. They
are powerful tools for setting corporate, section, or user defaults. Each solution sequence is designed with a specific
set of default values and solution paths. In the majority of the cases, the default values have been found well suited for
most models. There are occasions, however, when you may want to deviate from the default values.
There are multiple ways of setting parameters for MSC Nastran. For the NASTRAN Statement they are set by the rc
file and the execution statement. (See Configuring the Runtime Environment (Ch. A) in the MSC Nastran Installation
and Operations Guide and Chapter 1: nastran Command and NASTRAN Statement, The NASTRAN Statement in the
QRG). For the Case Control and Bulk Data Section the selection is extensive and controls for a parameter can come
from each section. (See Chapter 5: Parameters (Ch. 5) in the QRG). For the advanced users, there is also the DMAP
(Direct Matrix Abstraction Program) module language. (See the MSC Nastran DMAP Programmers Guide)
The majority of the parameters referenced in this guide are case control or bulk data parameters. These will be the ones
controlling default tolerances or controls for the type and amount of output. Depending on the parameter, it can be
specified either as a PARAM bulk data entry and/or a PARAM case control command. (See Chapter 5:Parameters,
Parameter Applicability Tables in the QRG) The format of each is as follows: PARAM,xname,xvalue where xname and
xvalue are the name and value of the PARAMeter, respectively, such as PARAM,WTMASS,0.00259, for the terms of the
structural mass matrix to be multiplied by the value of this parameter when they are generated.
Main Index
Chapter 2: MSC Nastran Files 65
The Executive Control Section Overview
The format of the executive control statements is free-field format between columns 1-72. The statements may begin
in any column (except for comments). The operands that is, the input data following the name are separated from
the name using a comma and/or one or more blanks. If more than one operand is needed, they are also separated by a
comma and/or one or more blanks.
Comments may be placed anywhere in the input file and are denoted by placing a $ at the start of the comment. If an
input line is to be a comment only, it must start in column 1. Comments are ignored by MSC Nastran and have no affect
on the results. You should use comments generously; they greatly improve the readability of the input file.
The ID statement can be used to help identify a job; this statement is optional.
The SOL statement specifies the desired solution sequence to be used. To invoke a particular analysis type, the SOL
name is followed by the ID number or name of the solution sequence. Linear static analysis is the Solution Sequence,
SOL 101 or SOL SESTATIC. The Solution Sequences are numbered in the range of 101 through 700. Linear static
analysis can also be selected with the SOL 400 statement and the addition of the Case Control command
ANALYSIS=STATICS.
The ECHOON and ECHOOFF statements are used to control the printing of the Executive Control Section. Since the
Executive Control Section is generally quite small, it is recommended that it be echoed (the default choice). If the
Executive Control is not printed, readability is hindered, especially if you have to come back to the output file several
months after the original run was made. Since ECHOON is the default, this statement is not widely used. However, one
useful application for the ECHOOFF statement is security. By using ECHOOFF, portions of the Executive Control
Section that you may consider proprietary (such as user-written DMAP) may be eliminated from the output file.
Multiple ECHOON and ECHOOFF statements may be used.
The DIAG statement in the QRG is used to request additional diagnostics or output and/or to modify the solution.
For more details, see Chapter 3: Executive Control Statements, Executive Control Statement Descriptions in the QRG.
Main Index
66 Nonlinear Users Guide
The Case Control Section
2 Y
A = 4.0 in
6 2
E = 30.0 10 lb in
1 3
4 Subcase 2
J = 1.27 in 2 F =2000 lb
3
= 0.3 2
5 10 ft
Elements
1 X
Grid points
4 4
As an example, a second load case is added to the previous truss structure as shown in Figure 2-4. The two separate
analyses are performed during a single run. The first subcase consists of the 1000-pound force acting at grid point 4 in
the negative Y-direction. The second subcase consists of the 2000-pound force acting at grid point 3 in the X-direction.
The displacements are requested for all the grid points for both load cases, but only the element forces are desired for
the second load case.
The Case Control Section up to the Bulk Data Section, meeting these requirements, is shown in Listing 2-3. (See
MSC_DOC_DIR/doc/linstat/truss2.dat.)
Main Index
Chapter 2: MSC Nastran Files 67
The Case Control Section
$
SUBCASE 1
LABEL = POINT LOAD AT GRID POINT 4
LOAD = 10
$
SUBCASE 2
LABEL = POINT LOAD AT GRID POINT 3
LOAD = 11
ELFORCE = ALL
$
BEGIN BULK
$ Same as truss1.dat shown earlier
$
The Case Control Section consists of two subcases identified with ID numbers 1 and 2. The choice of the ID numbers
is arbitrary, but they must be strictly in ascending order. For example, if the first subcase is 10, the ID number of the
second subcase can be any integer number greater than 10.
Any title information, output requests, or bulk data selections made prior to the first subcase are the defaults for all the
subcases unless they are overridden by a request within a specific subcase. For the Case Control Section shown in
Listing 2-3, a displacement request is placed above the first subcase but not within either of the two subcases.
Therefore, the displacements are output for all of the grid points for both subcases. Since an element force request
appears only in Subcase 2, no element force output is generated for Subcase 1. The element forces are output for all
elements for the loading condition defined in Subcase 2.
It is often desirable to limit the output to a select number of grid points or elements; this is accomplished with the use
of the SET command. The SET case command identifies a group of grid points and/or elements as a specific set ID.
For example, suppose you want to output all of the grid point displacements for Subcase 1 and the grid point
displacements for grid points 3 and 4 for Subcase 2. In addition, you want the element force output for element 3 for
Subcase 1 and elements 3 and 4 for Subcase 2. A Case Control Section that meets these requirements is shown in
Listing 2-4. (See MSC_DOC_DIR/doc/linstat/truss3.dat.)
Main Index
68 Nonlinear Users Guide
The Case Control Section
DISPLACEMENT =1
ELFORCE = 1
BEGIN BULK
The ECHO case command in the Case Control Section is used to control the printing of the Bulk Data Section. There
are five options for this command:
SORT Prints the Bulk Data Section in sorted form; free field format is converted to small field format
and is sorted alphabetically. This is the default.
UNSORT Prints the Bulk Data Section exactly as it is input.
BOTH Prints the Bulk Data Section in the sorted and unsorted format.
PUNCH Punches the bulk data to an ASCII file.
NONE Suppresses printing of the bulk data.
For example, to echo the bulk data in the unsorted form, you use the following ECHO command:
ECHO = UNSORT
The output requests, such as DISPLACEMENT, FORCE, STRESS, etc., are needed whenever data recovery quantities
are to be computed, even if they are not printed. Such is the case when you use a postprocessor to view the results.
Typical output requests are as follows:
SET 1 = 5, 6, 7
SET 3 = 1, 5, 9
STRESS = ALL
DISP(PLOT) = 1 $ WHERE 1 IS THE ID OF A SET OF GRID POINTS
ELFORCE(PUNCH) = 3 $ WHERE 3 IS THE ID OF SET OF ELEMENTS
The PLOT option suppresses the printed output, but the data recovery is still performed. This option is useful for large
models where the quantity of printed output would be excessive, but the data recovery is still required for
postprocessing. The PUNCH option writes the results to an ASCII file instead of the .f06 file. This option makes it
convenient for exporting the results to other programs.
Main Index
Chapter 2: MSC Nastran Files 69
The Case Control Section
The results output from large simulations can quickly consume enormous amounts of computer disk storage space. Not
only does MSC Nastran store the output for use by the postprocessing program, the postprocessor itself can use large
quantities of disk space when it creates different views of the data for use in a graphical display environment. If the
postprocessor program does not require results from one (or more) subcase(s), both computer time and disk space are
wasted reading and storing the unnecessary data. Use of the POST command can be used to eliminate the unwanted
results data from the OUTPUT2 file. This can substantially reduce the file processing time and disk space used by the
postprocessor.
Another benefit of the POST command is that it allows MSC Nastran to write data into output files by specifying their
file names instead of Fortran unit reference numbers. This capability removes the limitation imposed by the MSC
Nastran file management system on the number of ASSIGN statements.you are now able to write solution data into as
many output files, as desired.
The amount of data recovered for each subcase is dependent upon the case control commands. If a request is placed
above all subcases, that request applies to all of the subcases. Data are generated and produced one time for all subcases
by examining the case control requests in each subcase. Each separate type of response is stored in its own data block.
For example, displacement results are computed and stored separately from element stress results. These data blocks
include results for any subcase that contains a data recovery request with a print, plot, or punch destination.
The POST command provides control over the contents of the output file produced by the MSC Nastran program when
a PARAM,POST bulk data entry is present in the input. It does not affect the contents of the results data blocks, only
the amount of data that is transferred from those data blocks to the output file. The POST command is examined for
the output requested in each subcase. Only that output requested by the POST command is placed on the output file
for the subcase. Note that the data must be made available by the presence of a data recovery command in the subcase.
For example, the POST command cannot output stress data if no stress data have been requested.
The general format of the POST command is:
POST TOFILE furn
ppname oplist
TOCASE filename
Example
This sample shows the usage of the POST command to eliminate unwanted output from being placed on the
postprocessor data file generated by MSC Nastran. The model data is not important for this example. For example,
consider the case of a static simulation performed on a model requiring three load cases. The required output is
displacements at three grid points in the model. Furthermore, the output for the second load case must also include
punched displacement data for all grid points. Displacement results for subcases 100 and 300 are to be placed on the
default OUTPUT2 file for use by the Patran postprocessor. No output for subcase 200 is required for postprocessing by
Patran. The subcase structure might look something like:
SET 1000 = 10, 11, 12
DISP = 1000
$
POST TOFILE 12 $ use defaults: Patran
$
SUBCASE 100
LOAD = 100
$
Main Index
70 Nonlinear Users Guide
The Case Control Section
SUBCASE 200
LOAD = 200
DISP(PUNCH) = ALL
POST NODISP $ stop any displacement output from going to POST file
$
SUBCASE 300
LOAD = 300
In this example, the presence of the POST command above all of the subcases indicates that all output requested (DISP
for SET 1000 is the only output requested and generated) is to be stored on the default OUTPUT2 FORTRAN unit for
use by Patran. The POST command in SUBCASE 200 with the NODISP option prevents the large output produced by
the DISP(PUNCH)=ALL request in SUBCASE 200 from being placed on the OUTPUT2 file.
A complete description of all of the case control commands is available in the QRG.
Main Index
Chapter 2: MSC Nastran Files 71
The Bulk Data Section
1 2 3 4 5 6 7 8 9 10
8 8 8 8 8 8 8 8 8 8
The name of the bulk data entry is entered in field 1 beginning in column 1. Fields 2 through 9 are used for data input.
The only limitation is that the data within these fields must lie completely within the field, have no embedded blanks,
and must be of the proper type. There are three distinct types of input used for bulk data entries: real, integer, and
character; in general, they are not interchangeable. If a particular field requires an integer, such as an ID number on the
GRID entry, entering a real number causes a fatal error.
Field 10 of the bulk data entry is used for two purposes. If the bulk data entry does not have a continuation line, Field
10 may be used as an optional comment field. If the bulk data entry has a continuation line, field 10 is used for the
continuation identifier. The continuation identifier must be unique with respect to all the other identifiers in your Bulk
Data Section.
Real numbers may be entered in various ways; however, the number must contain a decimal point. For example, the
real number 7.0 may be entered as 7., 7.0, 7.E0, .7E+1, 700.E-2, etc. Integer numbers do not have decimal points.
Character input consists of one to eight alphanumeric characters, the first of which must be alphabetic.
A blank may be used in many of the data fields, in which case the appropriate default is used. The defaults for data
fields, if they are applicable to a particular field, are described in the QRG. There are certain situations that allow either
an integer or a real number in a data field. In these situations, the choice of an integer or a real number in these data
fields affects the problem solution. Fortunately, there are only a few such situations, and they are discussed in the
appropriate sections of this users guide.
Consider the definition for grid point 2 of the truss model shown in Figure 2-6. The name of the entry is GRID, which
begins in column 1 of field 1. The grid point ID (2 in this case) must be an integer (no decimal point) greater than 0.
Fields 3 and 7 represent coordinate system IDs and also must be integers. Since these fields are blank, the default of 0
is used. Fields 4, 5, and 6 represent the physical location of the grid point, and they must be entered as real numbers.
Main Index
72 Nonlinear Users Guide
The Bulk Data Section
Optional fields are field 8, which is used to define permanently constrained degrees of freedom, and field 9, the
superelement ID field. If they are used, only integers are acceptable. Since the GRID entry does not have a continuation
line, field 10 may be used as a comment, if desired.
1 2 3 4 5 6 7 8 9 10
GRID ID CP X1 X2 X3 CD PS SEID
GRID 2 0.0 120. 0. 2345
Figure 2-6 A Bulk Data Example: the GRID Entry
If a bulk data entry requires more than eight data fields, continuation lines must be used. When a continuation line is
used, field 10 contains the continuation identifier that must be unique with respect to all other continuation identifiers.
Column 1 of field 10 is always ignored and is not considered part of the identifier. The small field continuation line is
denoted by a + symbol in column 1 of field 1. The continuation line is matched to its parent entry by the identifier in
columns 2 through 8 of Field 1. An example is shown in Figure 2-7.
1 2 3 4 5 6 7 8 9 10
TYPE +ID1
+ID1 +ID2
+ID2
Figure 2-7 Small-Field Entry with Continuation Lines
The identifiers used in Figure 2-7 are arbitrary. The only restriction is that the alphanumeric character strings used as
the continuation identifiers do not contain the symbols *, =, or $. The + symbol used in column 1 of field 10 is for
clarity.
It is often more convenient to let MSC Nastran generate the continuation identifiers automatically. To do this, leave
both field 10 of the parent entry and field 1 of the continuation line blank, and place the continuation line directly
following the parent entry. The blank field indicates that it is a small field continuation of the previous line. The
advantage of using continuation identifiers is that the input file may be unsorted, that is, the continuation lines do not
have to follow the parent line. However, this benefit is generally not worth the inconvenience of using the continuation
identifiers.
The most common form of input when generating the bulk data entries manually is the free field format. In the free
field format, commas are used to separate the fields (blank spaces are also acceptable, but not recommended). An entry
using the free-field format is identified by a comma or an equal sign in any of the first eight columns of the entry.
Internally, the free field format is converted to small field format, so the same rules apply to free field format entries
as for the small field format entries. Integers and character entries must be eight characters or less, and real numbers
longer than eight characters are rounded to eight characters. Free field format entries must start in column 1. If
automatic continuation is to be used, the continuation line starts with a comma in field 1, indicating that the first field
is blank. As long as less than 80 columns are used, the free field format can have more than 8 data fields and
connections will be automatic.
You may find it necessary to input integers larger than eight characters or real numbers with more significant digits
than can be placed in an eight-character field. In such cases, a large field format option is available. With large field
Main Index
Chapter 2: MSC Nastran Files 73
The Bulk Data Section
format, as shown in Figure 2-8, Fields 2 through 9 are 16 characters in length and generally require two lines instead
of one.
1a 2 3 4 5 10a
TYPE* *ID1
b 6 7 8 9 10a
*ID1
The large field format entry is identified by the * symbol after the entry name. The large field format continuation is
denoted by the * symbol in column 1.
It is possible to mix small field and large field format continuations; however, it is generally not recommended because
it becomes difficult to discern the locations of the fields. There is no automatic continuation feature for the large field
format. If the large field format is used, the continuation identifiers must be used.
Figure 2-9 shows a CORD2R entry using the three different formats. Knowledge of the CORD2R entry is not necessary
for this example except to note that this entry requires a continuation line. Coordinate systems are discussed in Grid
Points, Scalar Points, and Coordinate Systems, 61.
1 2 3 4 5 6 7 8 9 10
CORD2R 1 0 0.0 0.0 0.0 1.0 0.0 0.0 +COR1
+COR1 1.0 1.0 0.0
Small-Field Format with Continuation Identifier
Large-Field Format
Figure 2-9 Input Format Example
Main Index
74 Nonlinear Users Guide
The Bulk Data Section
In the examples shown in Figure 2-9, column 1 of field 10 contains a + symbol for the small-field format and a *
symbol for the large-field format. Since this column is ignored, the choice of the + and the * is arbitrary; however, these
choices do improve readability because they serve as a reminder for the format of the continuation line. This
convention is commonly used in the MSC Nastran user community.
The majority of bulk data entries are shown in small-field or large-field format throughout this users guide whenever
possible. This does not mean that free-field format should not be used; it is just less readable for this guide.
Main Index
Chapter 3: Nonlinearity and Analysis Types
Nonlinear Users Guide
Main Index
76 Nonlinear Users Guide
Introduction
Introduction
The FEM uses different numerical techniques to solve the equations depending on whether the system of equations
describes a linear or nonlinear system. In this chapter, we look at the different types of analysis problems that may be
solved, categorized according to the numerical procedure used to solve the problem. From an engineering point of
view, the numerical method used determines the limits on the behavior of the physical problem under consideration.
For example, a linear static analysis is based on the assumption of a linear force-deformation and stress-strain response
of the structure and material. This restriction is imposed by the solution method chosen. If we find that any of these
assumptions are no longer valid once the load exceeds a certain level, a linear static solution is no longer valid and we
must consider nonlinear numerical methods to obtain a solution that incorporates all of the physics of the problem.
While SOL 400 is valid for use on linear problems, it has been designed to solve nonlinear problems that include large
deformation, material, and contact/boundary nonlinearities. There are many cases where the solution types may be
mixed. For example, to get an accurate representation of mode shapes and natural frequencies of a structure under load,
you must first do a nonlinear static analysis to get the correct stiffness matrix to do the eigenvalue extraction on to
provide the natural frequencies and mode shapes. A similar situation occurs with buckling where the nonlinear effects
reduce the actual critical buckling load. These types of solutions are called perturbation analyses and are covered in
the sections titled Linear Perturbation Analysis and Multi Stepping in SOL 400. Another important use of multi-
step analyses is to control load history. Loads and constraints are grouped into load cases that are then assigned in
sequence to SOL 400 nonlinear steps and subcases. In this way, the user has control over the loading sequence applied
in the analysis. Complex loading histories that simulate processes such as extrusions and other manufacturing
processes can then be simulated.
The goal of this chapter is to inform the user of the various analysis types available using MSC Nastran SOL 400 and
provide the information they would need to formulate the solution to the engineering structural or thermal problems
they need to solve, including the types of analysis controls that might be required. The analysis overview of those
procedures is covered in this chapter and the Patran user interface that would be the recommended way to create the
model and set up the solution procedure is covered in Chapter 6: Setting Up, Monitoring, and Debugging the Analysis.
Main Index
Chapter 3: Nonlinearity and Analysis Types 77
Linear Analysis
Linear Analysis
In a linear analysis, we implicitly assume that the deflections and strains are very small and the stresses are smaller
than the material yield stresses. Consequently, there is assumed to be a linear relationship between the applied loads
and the response of the structure. The stiffness can be considered to remain constant (i.e., independent of the
displacements and forces) and the finite element equilibrium equations
P = Ku
are linear where the stiffness matrix K is independent of both u , the generalized displacement vector, and P , the
generalized force vector. This linearity implies that any increase or decrease in the load will produce proportional
increase or decrease in displacements, strains, and stresses. Because of the linear relationship, you only need to
calculate the stiffness of the structure once. From this stiffness representation, you can find the structures response to
other applied loads by multiplying the load vectors by the decomposed stiffness matrix. Linear static problems are
solved in a single step. In addition, solutions can be combined using the principle of superposition.
A linear analysis is the simplest and most cost effective type of analysis to perform. Because linear analysis is simple
and inexpensive to perform and often gives satisfactory results, it is the most commonly used structural analysis.
Nonlinearities due to material, geometry, or boundary conditions are not included in this type of analysis. The behavior
of an isotropic, linear, elastic material can be defined by two material constants: Youngs modulus, and Poissons ratio.
In actuality, linear analysis is merely an approximation to the true behavior of a structure. In some cases the
approximation is very close to the true behavior, in other cases linear analysis may provide highly inaccurate results.
The following is a summary of the main steps in a linear static analysis:
1. Input: The problem geometry (nodes and elements), physical and material properties, and loads and boundary
conditions are taken from the SOL 400 input file and put into the MSC Nastran database.
2. Element stiffness matrix and force vector calculation: The element stiffness matrices and equivalent nodal
forces for distributed forces are computed. The detailed descriptions in the Chapter 11: Element Library provide
the kinds of forces that each element can support.
3. Global stiffness matrix and load vector assembly: The global stiffness matrix and the combined nodal force
vectors are assembled. Boundary and constraint conditions are incorporated by modifying the element stiffness
matrices and force vectors.
4. Solution of equations: The nodal displacement vector is computed by solving the system of simultaneous.
5. Strain energy and reaction force calculation: The strain energy and reaction forces (unbalanced grid point
forces) are computed using the displacement vector, the element stiffness matrices and the force vectors.
6. Stresses and strains calculation: The strains and stresses are computed at selected points for each element. See
the Chapter 11: Element Library for a detailed descriptions of the stress recovery points for the SOL 400
elements.
SOL 400 allows you to perform linear elastic analysis using any element type in the program. Various kinematic
constraints and loadings can be prescribed to the structure being analyzed; the problem can include both isotropic,
orthotropic, and anisotropic elastic materials.
The principle of superposition holds under conditions of linearity.
Linear analysis does not require storing as many quantities as does nonlinear analysis; therefore, it uses the memory
more sparingly.
Main Index
78 Nonlinear Users Guide
Nonlinear Analysis
Nonlinear Analysis
However, we know that in many structures the deflections and the stresses do not change proportionately with the
loads. In these problems the structures response depends upon its current state and the equilibrium equations reflect
the fact that the stiffness of the structure is dependent on both u and P .
P = K P u u
As the structure displaces due to loading, the stiffness changes, and as the stiffness changes the structures response
changes. As a result, nonlinear problems require incremental solution schemes that divide the problem up into steps
calculating the displacement, then updating the stiffness. Each step uses the results from the previous step as a starting
point. As a result the stiffness matrix must be generated and inverted many times during the analysis adding time and
costs to the analysis.
In addition, because the response is not proportional to the loads, each load case must be solved separately and the
principle of superposition is not applicable.
Main Index
Chapter 3: Nonlinearity and Analysis Types 79
Nonlinear Effects and Formulations
Material Nonlinearity
Recall that linear analysis assumes a linear relationship between stress and strain. Material nonlinearity results from
the nonlinear relationship between stresses and strains. In addition, large strain can influence the material behavior.
Considerable progress has been made in attempts to derive the continuum or macroscopic behavior of materials from
microscopic backgrounds, but, up to now, commonly accepted constitutive laws are phenomenological. Difficulty in
obtaining experimental data is usually a stumbling block in accurately simulating material behavior. A plethora of
models exist for more commonly available materials like elastomers and metals. Material models of considerable
practical importance are: composites, viscoplastics, creep, soils, concrete, powder, and foams. Figure 3-1 shows
representation of the elastoplastic, elasto-viscoplasticity, and creep.
Examples of material nonlinearities include metal plasticity, materials such as soils and concrete, or rubbery materials
(where the stress-strain relationship is nonlinear elastic). Various plasticity theories such as von Mises or Tresca (for
metals), and Mohr-Coulomb or Drucker-Prager (for frictional materials such as soils or concrete) can be selected by
you. Three choices for the definition of subsequent yield surfaces are available in SOL 400. They are isotropic
hardening, kinematic hardening, or combined isotropic and kinematic hardening. With such generality, most plastic
material behavior, with or without the Bauschinger effect, can be modeled.
Elasto-Plastic Behavior Elasto-Viscoplastic Behavior
Creep Behavior t
Main Index
80 Nonlinear Users Guide
Nonlinear Effects and Formulations
Geometric Nonlinearity
Geometrically nonlinear problems involve large displacements; large means that the displacements invalidate the
small displacement assumptions inherent in the equations of linear analysis. For example, consider a classical thin plate
subject to a lateral load; if the deflection of the plates midplane is anything close to the thickness of the plate, then the
displacement is considered large and a linear analysis is not applicable.
Geometric nonlinearity results from the nonlinear relationship between strains and displacements on the one hand and
the nonlinear relation between stresses and forces on the other hand. If the stress measure is conjugate to the strain
measure, both sources of nonlinearity have the same form. This type of nonlinearity is mathematically well defined,
but often difficult to treat numerically. Three important types of geometric nonlinearity occur:
1. Problems where large rotation occur.
2. The analysis of buckling and snap-through problems (see Figure 3-2 and Figure 3-3).
3. Large strain problems such as manufacturing, crash, and impact problems. In such problems, due to large strain
kinematics, the mathematical separation into geometric and material nonlinearity is not unique.
P Linear
P Stable
u Pc Neutral
Unstable
u
Figure 3-2 Buckling
Main Index
Chapter 3: Nonlinearity and Analysis Types 81
Nonlinear Effects and Formulations
conservative loading system when integrated over the structure. A pressurized cylinder (see Figure 3-6) is an example
of this.
Main Index
82 Nonlinear Users Guide
Nonlinear Effects and Formulations
Contact and friction problems lead to nonlinear boundary conditions. This type of nonlinearity manifests itself
including assembly modeling, metal forming, gears, interference of mechanical components, pneumatic tire contact,
and crash (see Figure 3-7). Loads on a structure cause nonlinearity if they vary with the displacements of the structure.
If there is a change in constraints due to contact during loading, the problem may be classified as a boundary nonlinear
problem and would require the use of BCTABLE/ BCTABLE1, BCONECT, BCONPRG, BCONPRP, BCBODY, or BSURF
bulk data entry options. CGAP elements would have been used in the traditional nonlinear sequences of SOL 106 or
129; however, the use of GAP elements is strongly discouraged in SOL 400.
Geometric Nonlinearities
Geometric nonlinearity leads to two types of phenomena: change in structural behavior and loss of structural stability.
There are two natural classes of large deformation problems: the large displacement, small strain problem and the large
displacement, large strain problem. For the large displacement, small strain problem, changes in the stress-strain law
can be neglected, but the contributions from the nonlinear terms in the strain displacement relations cannot be
neglected. For the large displacement, large strain problem, the constitutive relation must be defined in the correct
frame of reference and is transformed from this frame of reference to the one in which the equilibrium equations are
written.
The kinematics of deformation can be described by the following approaches:
Lagrangian Formulation
In the Lagrangian method, the finite element mesh is attached to the material and moves through space along with the
material. In this case, there is no difficulty in establishing stress or strain histories at a particular material point and the
treatment of free surfaces is natural and straightforward.
The Lagrangian approach also naturally describes the deformation of structural elements; that is, shells and beams.
Shortcomings of the Lagrangian method are that flow problems are difficult to model and that the mesh distortion is
as severe as the deformation of the object.
The Lagrangian approach can be classified in two categories: the total Lagrangian method and the updated Lagrangian
method. In the total Lagrangian approach, the equilibrium is expressed with the original undeformed state as the
Main Index
Chapter 3: Nonlinearity and Analysis Types 83
Nonlinear Effects and Formulations
reference; in the updated Lagrangian approach, the current configuration acts as the reference state. The kinematics of
deformation and the description of motion is given in Table 3-1 and Figure 3-8.
Previous f
t=n u
Current
t=n+1
Fn
un + 1
un F
Reference
t=0
Fn+1 = Fn
0 0
S i j E ij dV = b i i dV + t i i dA (3-1)
V V A
0 0 0
Main Index
84 Nonlinear Users Guide
Nonlinear Effects and Formulations
0
Here S ij is the symmetric second Piola-Kirchhoff stress tensor, E i j , is the Green-Lagrange strain, b i is the body force
0
in the reference configuration, t i is the traction vector in the reference configuration, and i is the virtual
displacements. Integrations are carried out in the original configuration at t = 0 . The strains are decomposed in total
strains for equilibrated configurations and the incremental strains between t = n and t = n + 1 as:
n+1 n
Eij = E i j + E i j (3-2)
n
while the incremental strains are further decomposed into linear, E i j and nonlinear, E i j parts as:
n
E i j = Ei j + E i j
1 u u 1 u n u
E = --- -----------i + -----------j + --- --------k ------------k (3-3)
2 X j X i 2 X i X j
The second term in the bracket in equation (3-3) is the initial displacement effect. E n is the nonlinear part of the
incremental strain expressed as:
1 u u u n u
= --- ------------k ------------k + --------k ------------k
n
E (3-4)
2 X i X j X j X i
K 0 + K 1 + K 2 u = F R (3-5)
0 0
K 0 ij = imn D mnpq pqj dV
V
0
0 u
in the above equations, imn and imn are the constant and displacement dependent symmetric shape function gradient
matrices, respectively, and D mnpq is the material tangent,
K 2 ij = N i k N j l S kl dV
V
0
Main Index
Chapter 3: Nonlinearity and Analysis Types 85
Nonlinear Effects and Formulations
in which S kl is the second Piola-Kirchhoff stresses and N i k is the shape function gradient matrix.
Also, u is the correction displacement vector. F and R are the external and internal forces, respectively.
This Lagrangian formulation can be applied to problems if the undeformed configuration is known so that integrals
can be evaluated, and if the second Piola-Kirchhoff stress is a known function of the strain. The first condition is not
usually met for fluids, because the deformation history is usually unknown. For solids, however, each analysis usually
starts in the stress-free undeformed state, and the integrations can be carried out without any difficulty.
For elastic-plastic and viscoplastic materials, the constitutive equations usually supply an expression for the rate of
stress in terms of deformation rate, stress, deformation, and sometimes other (internal) material parameters. The
relevant quantity for the constitutive equations is the rate of stress at a given material point.
It, therefore, seems most obvious to differentiate the Lagrangian virtual work equation with respect to time. The rate
of virtual work is readily found as
v
S ij E i j + S i j --------k ------------k dV =
X i X j b i i dV + t i i dA (3-6)
V V A
0 0 0
This formulation is adequate for most materials, because the rate of the second Piola-Kirchhoff stress can be written as
S i j = S i j E kl S mn E pq (3-7)
For many materials, the stress rate is even a linear function of the strain rate
S i j = D i j kl S mn E pq E kl (3-8)
Equation (3-6) supplies a set of linear relations in terms of the velocity field. The velocity field can be solved
noniteratively and the displacement can be obtained by time integration of the velocities.
The second Piola-Kirchhoff stress for elastic and hyperelastic materials is a function of the Green-Lagrange strain
defined below:
S i j = S i j E kl (3-9)
S i j = D i j kl E kl (3-10)
the resulting set of equations is still nonlinear because the strain is a nonlinear function of displacement.
Main Index
86 Nonlinear Users Guide
Nonlinear Effects and Formulations
analysis of shell and beam structures in which rotations are large so that the nonlinear terms in the curvature
expressions may no longer be neglected, and
large strain plasticity analysis, for calculations which the plastic deformations cannot be assumed to be
infinitesimal.
contact problems where it is easier to express the constraints with respect to the current state.
In general, this approach can be used to analyze structures where inelastic behavior (for example, plasticity,
viscoplasticity, or creep) causes the large deformations. The (initial) Lagrangian coordinate frame has little physical
significance in these analyses since the inelastic deformations are, by definition, permanent. For these analyses, the
Lagrangian frame of reference is continuously being redefined.
It is instructive to derive the stiffness matrices for the updated Lagrangian formulation starting from the virtual work
principle in equation (3-9).
Direct linearization of the left-hand side of equation (3-9) yields:
S i j d E ij dV = i k kj u i j dv (3-11)
V V
0 n+1
where u and are actual incremental and virtual displacements respectively, and kj is Cauchy stress tensor.
dS i j E i j dV
s s
= ij L i jk l u kl dv (3-12)
V V
0 n+1
denotes the symmetric part of , which represents the gradient operator in the current configuration. Also, in
s
1
i j = --- F i m S mn F jn
J
s
E i j = F mi mn F nj
(3-13)
and
1
L i jk l = --- F i m F j n F kp F l q D mnpq
J
in which D mnpq represents the material moduli tensor in the reference configuration which is convected to the current
configuration, L i j k . This yields:
K 1 + K 2 u = F R (3-14)
K 1 ij = im n L m n pq pqj
V
n+1
Main Index
Chapter 3: Nonlinearity and Analysis Types 87
Nonlinear Effects and Formulations
in which im n is the symmetric gradient operator-evaluated in the current configuration and kl is the Cauchy stresses
and K 2 is the geometric stiffness matrix written as
K 2 ij = kl N i k N j l dv
V
n+1
Keeping in view that the reference state is the current state, a rate formulation analogous to equation (3-6) can be
obtained by setting:
F i j = i j E i j = d i j -------- = ------- Sij = ij (3-15)
X i x i
v k k
i j d i j + i j ------- ------------ dv = b i i dv + ti i da (3-16)
x i x j
V V A
n+1 n+1 n+1
in which b i and t i is the body force and surface traction, respectively, in the current configuration
In this equation, i j is the Truesdell rate of Cauchy stress which is essentially a Lie derivative of Cauchy stress
obtained as:
1 1
i j = F i n JF nk k l F ml F mj (3-17)
The Truesdell rate of Cauchy stress is materially objective implying that if a rigid rotation is imposed on the material,
the Truesdell rate vanishes, whereas the usual material rate does not vanish. This fact has important consequences in
the large deformation problems where large rotations are involved. The constitutive equations can be formulated in
terms of the Truesdell rate of Cauchy stress as:
i j = L i jk d k
Main Index
88 Nonlinear Users Guide
Nonlinear Effects and Formulations
M = -1: turns off large displacement effects for advanced nonlinear elements (supercedes PARAM,LGDISP
setting).
M = 0: defaults to setting PARAM,LGDISP,N.
M = 1: generally sets flags for Updated Lagrange, Additive. Internally switches to Total Lagrange or Updated
Lagrange Multiplicative for specific material/element types (shown in the following table).
M = 2: generally sets flags for Updated Lagrange, Multiplicative. Internally switches to Updated Lagrange,
Additive for specific material/element types (shown in the following table).
Main Index
Chapter 3: Nonlinearity and Analysis Types 89
Nonlinear Effects and Formulations
Material Nonlinearities
In a large strain analysis, it is usually difficult to separate the kinematics from the material description. The following
table lists the characteristics of some common materials.
Main Index
90 Nonlinear Users Guide
Nonlinear Effects and Formulations
A complete description of the material types mentioned in the table is given in Chapter 10: Materials. However, some
notable characteristics and procedural considerations of some commonly encountered materials behavior are listed
next.
Inaccuracies in experimental data, misinterpretation of material model parameters and errors in user-defined material
law are some common sources of error in the analysis from the materials viewpoint. It is useful to check the material
behavior by running a small model with prescribed displacement and load boundary conditions in uniaxial tension and
shear (single element tests are not recommended).
Elastomers
Structures composed of elastomers, such as tires and bushings, are typically subjected to large deformation and large
strain. An elastomer is a polymer, such as rubber, which shows a nonlinear elastic stress-strain behavior. These
materials are characterized by the form of their elastic strain energy function.
For the finite element analysis of elastomers, there are some special considerations that do not apply for linear elastic
analysis. These considerations include:
Mesh Distortion
Incompressible Behavior
Instabilities
Existence of Multiple Solutions
Mesh Distortions
When extremely large deformations occur, the element mesh should be designed so that it can follow these
deformations without complete degeneration of elements. This problem is more prevalent when the updated Lagrange
procedure is used. For problems involving extreme distortions, the Marc global adaptive remeshing capability should
be used.
Main Index
Chapter 3: Nonlinearity and Analysis Types 91
Nonlinear Effects and Formulations
Incompressible Behavior
One of the most frequent causes of problems analyzing elastomers is the incompressible material behavior. Lagrangian
multipliers (pressure variables) are used to apply the incompressibility constraint. The result is that the volume is kept
constant in a generalized sense, over an element.
The Lagrange formulation is implemented with appropriate constraint ratios for lower- and higher-order elements in
2-D and 3-D.
The large strain elasticity formulation may also be used with conventional plane stress, membrane, and shell elements.
Because of the plane stress conditions, the incompressibility constraint are automatically satisfied.
Instabilities
Under some circumstances, materials can become unstable. This instability can be real or can be due to the
mathematical formulation used in the calculations.
Instability can also result from the approximate satisfaction of incompressibility constraints. If the number of
constraints is insufficient, local volume changes can occur. Under some circumstances, these volume changes can be
associated with a decrease in total energy. This type of instability usually occurs only if there is a large tensile
hydrostatic stress. Similarly, over constraints give rise to mesh locking and inordinate increase in total energy under
large compressive stresses.
Main Index
92 Nonlinear Users Guide
Nonlinear Effects and Formulations
When incompressible material is being modeled, the basic linearized incremental procedure is used in conjunction with
mixed variational principles similar in form to the Herrmann incompressible elastic formulation. These formulations
are incorporated in plane strain, axisymmetric, and 3-D elements. These mixed elements may be used in combination
with other elements in the library (suitable constraint equations may be necessary) and with each other. Where different
materials are joined, the pressure variable at the corner nodes must be uncoupled to allow for mean pressure
discontinuity. MPCs must be used to couple the displacements only.
Plasticity
In recent years there has been a tremendous growth in the analysis of metal forming problems by the FEM. Although
an Eularian flow-type approach has been used for steady-state and transient problems, the updated Lagrangian
procedure, pioneered by McMeeking and Rice, is most suitable for analysis of large strain plasticity problems. The
main reasons for this are: (a) its ability to trace free boundaries, and (b) the flexibility of taking elasticity and history
effects into account. Also, residual stresses can be accurately calculated.
The large strain plasticity capability in SOL 400 allows you to analyze problems of large-strain, elastic-plastic material
behavior. These problems can include manufacturing processes such as forging, upsetting, extension or deep drawing,
and/or large deformation of structures that occur during plastic collapse. The analysis involves both material,
geometric and boundary nonlinearities.
In performing finite deformation elastic-plastic analysis, there are some special considerations which do not apply for
linear elastic analysis. These considerations include:
Choice of Finite Element Types
Nearly Incompressible Behavior
Main Index
Chapter 3: Nonlinearity and Analysis Types 93
Nonlinear Effects and Formulations
Instabilities
Elastic-plastic structures are sometimes unstable due to necking phenomena. Consider a rod of a rigid-plastic
incompressible workhardening material. With the current true uniaxial strain rate and H the current workhardening,
the rate of true uniaxial stress is equal to
= H (3-18)
The applied force is equal to F = A , where A is the current area of the rod. The rate of the force is therefore equal to
F = A + A (3-19)
On the other hand, conservation of volume requires that
A + A = 0 (3-20)
Hence, the force rate can be calculated as
F = H A (3-21)
Main Index
94 Nonlinear Users Guide
Nonlinear Effects and Formulations
Instability clearly occurs if H . For applied loads (as opposed to applied boundary conditions), the stiffness matrix
becomes singular (nonpositive definite).
For the large strain plasticity option, the workhardening slope for plasticity is the rate of true (Cauchy) stress versus
the true (logarithmic) plastic strain rate. The workhardening curve must, therefore, be entered as the true stress versus
the logarithmic plastic strain in a uniaxial tension test.
In SOL 400, two formats are provided for data entry:
a. true stress vs. logarithmic plastic strain
b. true stress vs. logarithmic total strain.
Data provided in format b is internally converted to format a by the program.
Creep
Creep is a time-dependent inelastic behavior that can occur at any stress level, either below or above the yield stress
of a material. Creep is an important factor at elevated temperatures. In many cases, creep is also accompanied by
plasticity, which occurs above the yield stress of the material.
Conventional creep behavior is based on a von Mises creep potential with isotropic behavior described by the
equivalent creep law:
cr cr
= f , ,T ,t
The material behavior is therefore described by:
cr c r
= -------- t
cr
where -------- is the outward normal to the current von Mises stress surface and is the equivalent creep strain rate.
There are two numerical procedures used in implementing creep behavior. The default is an explicit procedure in which
the above relationship is implemented in the program by an initial strain technique. In other words, a pseudo-load
vector due to the creep strain increment is added to the right-hand side of the stiffness equation.
Ku = P + D dv
T cr
where K is the stiffness matrix, and u and P are incremental displacement and incremental nodal force vectors,
respectively. The integral:
T cr
D dv
V
Main Index
Chapter 3: Nonlinearity and Analysis Types 95
Nonlinear Effects and Formulations
is the pseudo-load vector due to the creep strain increment in which is the strain displacement relation and D is the
stress-strain relation. When plasticity is also specified through a suitably defined yield criterion and yield stress in
MSC Nastran, the plasticity is treated implicitly while the creep is treated explicitly.
As an alternative, an implicit creep procedure can be requested. In this case, the inelastic strain rate has an influence
on the stiffness matrix. Using this technique, significantly larger steps in strain space can be used. This option is only
to be used for isotropic materials with the creep strain rate defined by a creep constant.
Viscoelasticity
In certain problems, structural materials exhibit viscoelastic behavior. Two examples of these problems are quenching
of glass and time-dependent deformation of polymeric materials. The viscoelastic material retains linearity between
load and deformation; however, this linear relationship depends on time. Consequently, the current state of deformation
must be determined from the entire history of loading. Different models consisting of elastic elements (spring) and
viscous elements (dashpot) can be used to simulate the viscoelastic material behavior described in Chapter 10:
Materials. Both the equation of state and the hereditary integral approaches can be used for viscoelastic analysis.
A special class of temperature dependence known as the Thermo-Rheologically Simple behavior (TRS) is also
applicable to a variety of thermal viscoelastic problems. To model the thermo-rheologically simple material behavior,
MATTVE can be used to choose the Williams-Landel-Ferry equation or the power series expression or Narayanaswamy
model.
In SOL 400, two options are available for small strain viscoelastic analysis. The first option uses the equation of state
approach and represents a Kelvin model. The second option is based on the hereditary integral approach and allows
the selection of a generalized Maxwell model. The thermo-rheologically simple behavior is also available in the second
option for thermal viscoelastic analysis. For additional details, see Viscoelastic in Chapter 10: Materials.
The Simo model for large strain viscoelasticity can be used in conjunction with the damage and hyperelastic Mooney,
Ogden, Gent, or Arruda Boyce material model. The large strain viscoelastic material behavior can be simulated by
incorporating MATVE.
Viscoplasticity Procedures
There are two procedures in SOL 400 for viscoplastic analysis: explicit and implicit. A brief description of each
procedure follows:
Explicit Method
The elasto-viscoplasticity model in SOL 400 is a modified creep model to which a plastic element is added. The plastic
element is inactive when the stress is less than the yield stress of the material. You can use the elasto-viscoplasticity
model to solve time-dependent plasticity and creep as well as plasticity problems with a nonassociated flow law.
The NLMOPTS,CREEP entry in SOL 400 has been modified to enable solving problems with viscoplasticity. The
method is modified to allow solving elastic-plastic problems with nonassociated flow rules which result in
nonsymmetric stress-strain relations if the tangent modulus method is used.
NLMOPTS,CREEP controls allow you to select the procedure used to perform the time integration. If the explicit
method is used, then the time step must be small, but if the material is elastic and small deformation, no reassembly of
Main Index
96 Nonlinear Users Guide
Nonlinear Effects and Formulations
the stiffness matrix is required. If the implicit method is chosen, then larger time steps may be used, but reassembly
occurs at every increment.
In thermal creep simulations, it is necessary that the time step be chosen to satisfy accuracy of both the rate independent
thermal stress problem and the rate dependent creep problem. To insure that this occurs, the NLSTEP, GENERAL,
CREEP bulk data option has been introduced.
The viscoplastic approach converts an iterative elastic-plastic method to one where a fraction of the initial force vector
is applied at each increment with the time step controls. The success of the method depends on the proper use of the
automatic creep time step controls. This means that it is necessary to select an initial time step that will satisfy the
tolerances placed on the allowable stress change.
The allowable stress change is specified in the creep controls. The most highly stressed element usually yields the
maximum strain rate. It is also important to select a total time that gives sufficient number of increments to work off
the effects of the initial force vector. A total time of 30 times the estimated t is usually sufficient.
SOL 400 does not distinguish between viscoplastic and creep strains.
Since the viscoplasticity model in SOL 400 is a modified creep model, you can learn more about the details of
viscoplastic analysis control by reading the discussion later in this chapter, and elsewhere in this manual, on the creep
analysis procedure.
Nonlinear Loading
When the structure is deformed, the directions and the areas of the surface loads are changed. For most deformed
structures, such changes are so small that the effect on the equilibrium equation can be ignored. For some structures,
such as flexible shell structure with large pressure loads, the effects on the results can be quite significant so that the
surface load effects have to be included in the finite element equations.
SOL 400 forms both pressure stiffness and pressure terms (follower force) based on current deformed configuration
with MSC Nastrans PARAM,LGDISP1 option. Point forces may also be updated with deformation.
Main Index
Chapter 3: Nonlinearity and Analysis Types 97
Overview of Analysis Types
Main Index
98 Nonlinear Users Guide
Overview of Analysis Types
Main Index
Chapter 3: Nonlinearity and Analysis Types 99
Overview of Analysis Types
Linear Analysis
Linear analysis is obtained by considering the response in the step as the linear perturbation response about the base
state. The base state is the current state of the model at the end of the last general, nonlinear analysis step prior to the
linear perturbation step. Thus, the concept of linear analysis in MSC Nastran SOL 400 is rather general. A simple
example of the value of this generalization is the natural frequencies of a violin string under increasing tension. In this
case, geometrically nonlinear analysis of the string can be done in several steps, in each of which the tension is
increased. At the end of each such step, the frequencies can be extracted in a linear perturbation analysis step.
Load magnitudes (including the magnitudes of prescribed boundary conditions), during a linear perturbation analysis
step, are defined as the magnitudes of the load perturbations only.
Likewise, the value of any solution variable is output as the perturbation value only the value of the variable in the
base state is not included.
During a linear perturbation analysis step, the models response is defined by its linear elastic stiffness at the base state.
Plasticity and other inelastic effects are ignored. For hyperelastic materials, the tangent elastic moduli in the base state
are used. Contact conditions cannot change during a perturbation analysis they remain as they are defined in the base
state. Frictional slipping is not allowed during perturbation analyses all points in contact are assumed to be sticking
if friction is present. If geometric nonlinearity is included in the general, nonlinear analysis upon which the linear
perturbation study is being based, stress stiffening or softening effects and (pressure and other follower force) load
stiffness effects are included in the linear perturbation analysis. In this case, perturbation stresses and strains are
defined relative to the base state configuration. The effects of temperature and field variable perturbations are ignored
for materials that are dependent on temperature and field variables. However, temperature perturbations will produce
perturbations of thermal strain. Some procedures are purely linear perturbation procedures. These are:
Linear Statics ANALYSIS = STATICS
Bifurcation Buckling ANALYSIS = BUCK
Natural Frequency ANALYSIS = MODES
Modal Linear Transient ANALYSIS = MTRAN
Modal Complex Eigenvalue ANALYSIS = MCEIG
Linear perturbation analysis may be performed from time to time during a fully nonlinear analysis. This is done by
continuing the nonlinear response steps between the linear perturbation steps. The linear perturbation response has no
effect as the nonlinear analysis is continued. Generally, dynamic analyses may not be interrupted to perform
perturbation analyses: before performing the perturbation analysis, MSC Nastran SOL 400 requires that the structure
be brought into static equilibrium.
Modal linear transient analysis and linear static analysis are done in the time domain. The step time of linear
perturbations is never accumulated into the total time. For linear static perturbations the step time always begins at zero
for each new step.
Main Index
100 Nonlinear Users Guide
Overview of Analysis Types
Pertubation analysis implies a linearized solution about a nonlinear, preloaded state. The preloaded state may be either
a nonlinear static, nonlinear transient analysis, or a thermo-mechanically coupled analysis. When performing a
perturbation analysis in SOL 400 with large displacement activated:
The preload will be performed including large displacements and other material nonlinearities.
The stress obtained will be included in the initial stress stiffness of the perturbation step.
Main Index
Chapter 3: Nonlinearity and Analysis Types 101
Overview of Analysis Types
Direct and Modal Linear Transient Dynamic ANALYSIS = DTRAN and MTRAN
Depending upon the structure and the nature of the loading, two different numerical methods can be used for a transient
response analysis: direct and modal. The direct method (DTRAN) performs a numerical integration on the complete
coupled equations of motion. The modal method utilizes the mode shapes of the structure to reduce and uncouple the
equations of motion (when modal or no damping is used); the solution is then obtained through the summation of the
individual modal responses. ANALYSIS = MTRAN (modal transient) is used when the transient dynamic response of a
linear system, which includes inertial effects, is being studied. Since the use of modal transient analysis is covered
completely in the MSC Nastran Dynamic Analysis Users Guide, it will not be covered in this manual.
Main Index
102 Nonlinear Users Guide
Overview of Analysis Types
Nonlinear Single Physics: Nonlinear structures or nonlinear heat: For nonlinear structures, a ANALYSIS = NLSTAT
must come before a ANALYSIS = NLTRAN. Only Statics to Transient is allowed for structures. For heat transfer, a
nonlinear steady state heat transfer to a nonlinear transient heat transfer is not allowed.
A structural example:
SUBCASE 1
STEP 10
ANALYSIS=NLSTAT
STEP 20
ANALYSIS=NLSTAT
STEP 30
ANALYSIS=NLTRAN
Nonlinear Chained Physics: A nonlinear steady state heat with results used for a nonlinear static structural analysis.
Only a steady state heat to a structural nonlinear statics is allowed.
SUBCASE 1
STEP 10
ANALYSIS=HSTAT
STEP 20
ANALYSIS=NLSTAT
Nonlinear Coupled Physics: This allows for four combinations of STEP/SUBSTEP within the SUBCASE.
ANALYSIS = HSTAT for the first SUBSTEP and ANALYSIS = NLSTAT for the second SUBSTEP
ANALYSIS = HTRAN for the first SUBSTEP and ANALYSIS = NLTRAN for the second SUBSTEP
ANALYSIS = HTRAN for the first SUBSTEP and ANALYSIS = NLSTAT for the second SUBSTEP
ANALYSIS = HSTAT for the first SUBSTEP and ANALYSIS = NLTRAN for the second SUBSTEP
Main Index
Chapter 3: Nonlinearity and Analysis Types 103
Overview of Analysis Types
An example is:
SOL 400
CEND
TITLE=MSC Nastran SOL 400, Linear Perturbation Analysis
SUBTI=3D General Contact with Large Displacement Turned on
$
SUBCASE 1
STEP 1
LABEL=Nonlinear Static Analysis with Contact
ANALYSIS = NLSTATIC
NLPARM = 1
BCONTACT = 1
BOUTPUT=ALL
SPC = 2
LOAD = 3
DISPLACEMENT(SORT1,REAL)=ALL
STEP 2
LABEL=Linear Perturbation, DFREQ
ANALYSIS = DFREQ
DLOAD=200
FREQ =10
AUTOSPC=YES
SPC = 2
DISPLACEMENT = ALL
STEP 3
LABEL=Linear Perturbation, MFREQ
ANALYSIS = MFREQ
NLIC STEP 1 LOADFAC 1.0
METHOD = 30
DLOAD=200
FREQ =10
AUTOSPC=YES
RESVEC =NO
SPC = 2
DISPLACEMENT = ALL
Nonlinear Chained Analysis - with mesh/time change physics: A standard single physics nonlinear steady state
ANALYSIS = HSTAT or transient heat transfer ANALYSIS = HTRAN with either SCRATCH=NO or SCRATCH=MINI on
the job submittal. This is followed by a mechanical job submittal with an ASSIGN
hrun='name_of_heat_run.MASTER' and DBLOC DATABLK=(HEATDB) LOGI=hrun in the File Management
Section of the MSC Nastran executive and an ANALYSIS = NLSTAT or ANALYSIS = NLTRAN with a
TEMP(LOAD,HSUB,HSTEP,HTIME) in the subsequent mechanical job STEP. The subsequent mechanical job can have
both a different mesh than the heat job and different time steps.
An example:
SOL 400 (submitted with SCRATCH=MINI from a bulk file named Course_Mesh_Heat.dat)
CEND
SPC = 1
IC = 10
THERMAL=ALL
SUBCASE 3
STEP 4
ANALYSIS=HSTAT
NLPARM = 1
LOAD = 202
SUBCASE 10 (Note if STEP not provided STEP=1 defaulted)
Main Index
104 Nonlinear Users Guide
Overview of Analysis Types
ANALYSIS=HTRAN
TSTEPNL = 2
DLOAD = 404
BEGIN BULK
SOL 400 (Mechanical run using solution from
Course_Mesh_Heat.dat)
ASSIGN hrun= 'Course_Mesh_Heat.MASTER' (the ticks (') are
required)
DBLOC DATABLK=(HEATDB), LOGI=hrun
CEND
TEMPERATURE(INITIAL) = 1
SUBCASE 1
ANALYSIS=NLTRAN
STEP 1
TSTEPNL= 1
SPC = 2
TEMP (LOAD, HSUBC=3) = 3
DISPLACEMENT(SORT1,REAL)=ALL
NLSTRESS = ALL
STRESS = ALL
STEP 2
TSTEPNL= 2
SPC = 2
TEMP (LOAD, HSUBC=10, HTIME=0.80) = 4
DISPLACEMENT(SORT1,REAL)=ALL
NLSTRESS = ALL
STRESS = ALL
SUBCASE 2
ANALYSIS=NLTRAN
STEP 3
TSTEPNL= 3
SPC = 2
TEMP (LOAD, HSUBC=10, HTIME=ALL) = 5
DISPLACEMENT(SORT1,REAL)=ALL
NLSTRESS = ALL
stress = all
BEGIN BULK
Main Index
Chapter 3: Nonlinearity and Analysis Types 105
Static Analysis
Static Analysis
Static stress analysis is used when inertia effects can be neglected. The problem may still have a real time scale, for
example when the material has a viscoplastic response, such as rate dependent yield. The analysis may be linear or
nonlinear. Nonlinearity may arise from large displacement effects, material nonlinearity and boundary nonlinearity
(such as contact and friction).
Linear static analysis involves the specification of load cases and appropriate boundary conditions. Solutions may be
combined in a postprocessing mode.
Nonlinear static analysis requires the solution of nonlinear equilibrium equations, for which the program uses Full
Newton-Raphson or Modified Newton-Raphson. Many problems involve history dependent response, so that the
solution is usually obtained as a series of increments, with iteration within each increment to obtain equilibrium.
Increments must sometimes be kept small (in the sense that rotation and strain increments must be small) to assure
correct modeling of history dependent effects, but most commonly the choice of increment size is a matter of
computational efficiency if the increments are too large, more iteration will be required. Each solution method has a
finite radius of convergence, which means that too large an increment can prevent any solution from being obtained
because the initial state is too far away from the equilibrium state that is being sought it is outside the radius of
convergence. Thus, there is an algorithmic restriction on the increment size. For most cases, the automatic
incrementation scheme is preferred, because it will select increment sizes based on these considerations. Direct user
control of increment size is also provided because there are cases when you have considerable experience with his
particular problem and can therefore select a more economic approach. A complete discussion of the numerical
methods used to solver nonlinear static problems is included in this chapter; to get a converged solution, see Chapter 6:
Setting Up, Monitoring, and Debugging the Analysis.
Geometrically nonlinear static problems frequently involve buckling or collapse behavior, where the load-
displacement response shows a negative stiffness, and the structure must release strain energy to remain in equilibrium.
Several approaches are possible in such cases. One is to treat the buckling response dynamically, thus actually
modeling the kinetic response with inertia effects included as the structure snaps. This is easily accomplished by using
the restart option to terminate the static procedure and switch to a dynamic procedure when the static solution goes
unstable. In some simple cases, displacement control can provide a solution, even when the conjugate load (the
reaction force) is decreasing as the displacement increases. More generally, static equilibrium states during the unstable
phase of the response can be found by using the modified Riks method. This method is for cases where the loading
is proportional that is, where the load magnitudes are governed by a single scalar parameter. The method obtains
equilibrium solutions by controlling the path length along the load-displacement curve within each increment (rather
than controlling the load or displacement increment), so that the load magnitude becomes an unknown of the system.
The method can provide solutions even in cases of complex, unstable response. The Riks method cannot be used in
contact, heat transfer, coupled, or enforced motion.
Post-buckling
Geometrically nonlinear static problems frequently involve buckling or collapse behavior, where the load displacement
response shows a negative stiffness, and the structure must release strain energy to remain in equilibrium. Several
approaches are possible in such cases. One is to treat the buckling response dynamically; thus, actually modeling the
kinetic response with inertia effects included as the structure snaps. This is easily accomplished by using a transient
dynamic procedure to include inertial effects when the solution goes unstable. In some simple cases, displacement
Main Index
106 Nonlinear Users Guide
Static Analysis
control can provide a solution, even when the conjugate load (the reaction force) is decreasing as the displacement
increases. More generally, static equilibrium states during the unstable phase of the response can be found by using an
arc length method. This method is for cases where the loading is proportional that is, where the load magnitudes are
governed by a single scalar parameter. The method obtains equilibrium solutions by controlling the path length along
the load-displacement curve within each increment (rather than controlling the load or displacement increment), so that
the load magnitude becomes an unknown of the system.
The method can provide solutions even in cases of complex or unstable response.
Main Index
Chapter 3: Nonlinearity and Analysis Types 107
Normal Modes ANALYSIS = MODES
Main Index
108 Nonlinear Users Guide
Normal Modes ANALYSIS = MODES
conditions change, then the natural frequencies and mode shapes both change. For example, if the cantilever beam is
changed so that it is pinned at both ends, the natural frequencies and mode shapes change.
z x
1
z x
2
z x
3
z x
4
Figure 3-11 The First Four Mode Shapes of a Cantilever Beam
Modal quantities can be used to identify problem areas by indicating the more highly stressed elements. Elements that
are consistently highly stressed across many or all modes will probably be highly stressed when dynamic loads are
applied.
Modal strain energy is a useful quantity in identifying candidate elements for design changes to eliminate problem
frequencies. Elements with large values of strain energy in a mode indicate the location of large elastic deformation
(energy). These elements are those which most directly affect the deformation in a mode. Therefore, changing the
properties of these elements with large strain energy should have more effect on the natural frequencies and mode
shapes than if elements with low strain energy were changed.
SOL 400 contains two methods for eigenvalue extraction and three time integration operators. Nonlinear effects,
including material nonlinearity, geometric nonlinearity, and boundary nonlinearity, can be incorporated.
In addition to distributed mass, you can also attach concentrated masses associated with each degree of freedom of the
system. You can include damping in either the modal superposition or the direct integration methods. You can also
Main Index
Chapter 3: Nonlinearity and Analysis Types 109
Normal Modes ANALYSIS = MODES
include (nonuniform) displacement and/or velocity as an initial condition, and apply time-dependent forces and/or
displacements as boundary conditions.
Eigenvalue Analysis
SOL 400 uses either the inverse power sweep method or the Lanczos method to extract eigenvalues and eigenvectors.
The inverse power sweep method is typically used for extracting a few modes while the Lanczos method is optimal
for a few or many modes.
In dynamic eigenvalue analysis, we find the solution to an undamped linear dynamics problem:
K 2 M = 0
where K is the stiffness matrix, M is the mass matrix, are the eigenvalues (frequencies) and are the eigenvectors.
In SOL 400, K is the tangent stiffness matrix, which can include material and geometrically nonlinear contributions.
The mass matrix is formed from both distributed mass and point masses.
Main Index
110 Nonlinear Users Guide
Eigenvalue Extraction Method
Lanczos Method
The Lanczos method overcomes the limitations and combines the best features of the other methods. It is efficient, and
if an eigenvalue cannot be extracted within the range that you specify, a diagnostic message is issued. This method
computes accurate eigenvalues and eigenvectors. Furthermore, it prints meaningful user diagnostics and supports
parallel processing computers.
Comparison of Methods
The best method for a particular model depends on four factors:
1. The size of the model (the total number of degrees of freedom).
2. The number of eigenvalues desired.
3. The available real memory on your computer.
4. How well you can estimate the range of your eigenvalues.
Main Index
Chapter 3: Nonlinearity and Analysis Types 111
Eigenvalue Extraction Method
The enhanced inverse power method (SINV) can be a good choice if only a few modes are needed, and you have a
reasonable idea of your eigenvalue range of interest. It is useful for models in which only the lowest few modes are
desired.
For medium to large models, the Lanczos method is the recommended method.
User Interface
A METHOD command is required in a subcase of the case control section to select the appropriate eigenvalue
extraction method in the bulk data section. The bulk data entry is different depending on whether you are using the
inverse power (INV), enhanced inverse power (SINV), or Lanczos method. The EIGRL entry is used for the Lanczos
method, and the EIGB entry is used for the INV and SINV methods.
Main Index
112 Nonlinear Users Guide
Panel Buckling
Panel Buckling
This problem illustrates the nonlinear post-buckling analysis of a curved panel. This simulation highlights several SOL
400 capabilities including large displacement, material plasticity, contact between multiple deformable bodies and is
included in this chapter as an example of a perturbation buckling analysis.
Model Description
The model consists of a stiffened panel approximately 14 x 22 inches. The skin is constructed from aluminum and has
a slight curvature. A longitudinal Z stiffener is attached to the skin by a single row of rivets. The stiffener has a joggle
in it, and a doubler is used to fill the gap between the stiffener and the skin. Two additional longitudinal doublers are
attached to the outside edges of the skin.
This model was originally constructed in metric units and was converted to English units. As a result of the conversion,
some of the dimensions are rounded off.
Main Index
Chapter 3: Nonlinearity and Analysis Types 113
Panel Buckling
The rivets are modeled using CBAR elements. The element property is defined using the PBAR entry.
The three doublers are modeled as three deformable contact bodies. The stiffener lower flange is defined as a
deformable contact body. The skin is in contact with several components. Each contact area within the skin is defined
as a deformable contact body.
An RBE2 rigid body element is used to connect all the nodes on one end of the panel to a single node. This node is
used to apply the panel compressive loading.
Material
The skin and doublers are made of aluminum alloy 2024 which is modeled as an elastic-plastic material using MAT1
and MATEP entries. The stiffener is made of aluminum alloy 7349 and is also modeled as an elastic-plastic material
using MAT1 and MATEP entries. The plastic stress-strain relationships are defined by pairs of stress-strain values using
TABLES1 entries.
The NLPARM entry is used to specify the nonlinear analysis iteration strategy. Forty fixed time increments are specified
for this problem. Output requests are made for stress and strain using the STRESS and NLSTRESS entries. These
results are included in the MASTER/DBALL output file.
Main Index
114 Nonlinear Users Guide
Panel Buckling
Main Index
Chapter 3: Nonlinearity and Analysis Types 115
Panel Buckling
Main Index
116 Nonlinear Users Guide
Panel Buckling
Results Plots
The following plot shows the final deformation of the panel.
The following plot shows a fringe plot of the von Mises stress at full load.
Main Index
Chapter 3: Nonlinearity and Analysis Types 117
Panel Buckling
The following plot shows a graph of the total longitudinal reaction force vs. time increments.
Main Index
118 Nonlinear Users Guide
Dynamic Analysis of Linear Systems By Modal Methods
Main Index
Chapter 3: Nonlinearity and Analysis Types 119
Dynamic Analysis of Linear Systems By Modal Methods
where p = + i
This equation is similar to that for normal modes analysis (see Real Eigenvalue Analysis (Ch. 3) in the MSC Nastran
Dynamic Analysis Users Guide) except that damping is added and the eigenvalue is now complex. In addition, the
mass, damping, and stiffness matrices may be unsymmetric, and they may contain complex coefficients.
Complex eigenvalue analysis is controlled with the EIGC bulk data entry (similar to the EIGRL or EIGR bulk data
entries for normal modes analysis). There are three methods of solution: upper Hessenberg, complex Lanczos, and
inverse power. Complex eigenvalue analysis is available as a direct method (SOL 107), in which the equations are of
the same size as the number of physical variables. Complex eigenvalue analysis is also available as a modal method
(SOL 110), in which undamped modes are first computed and then are used to transform the matrices from physical to
modal variables.
There are three methods of solution: upper Hessenberg, complex Lanczos, and inverse power. Complex eigenvalue
analysis is available as a direct method (SOL 107, SOL 400 ANALYSIS = DCEIG), in which the equations are of the
same size as the number of physical variables. Complex eigenvalue analysis is also available as a modal method (SOL
110, SOL 400 ANALYSIS = MCEIG), in which undamped modes are first computed and then are used to transform
the matrices from physical to modal variables.
Main Index
120 Nonlinear Users Guide
Transient Dynamic Analysis
Main Index
Chapter 3: Nonlinearity and Analysis Types 121
Transient Dynamic Analysis
damping the dominant frequencies of the response may be far removed from those of the excitation because of
resonances at higher frequencies. Thus, for instance, assume that the excitation in a damped problem has three
dominant frequencies, one at 500 Hz, another at 600 Hz and yet another at 800 Hz. The highest dominant frequency is
800 Hz. This corresponds to a period of (1/800) seconds or 0.00125 seconds. So, a time step of (1/10)th of this period,
which is 0.000125 seconds, would be a good initial value.
Dynamic Integration
Direct integration is a numerical method for solving the equations of motion of a dynamic system. It is used for both
linear and nonlinear problems. In nonlinear problems, the nonlinear effects can include geometric, material, and
boundary nonlinearities. For transient analysis, MSC Nastran offers a family of direct integration operators listed
below.
Newmark-beta Operator
Generalized-alpha Operator
Direct integration techniques are imprecise; this is true regardless of which technique you use. Each technique exhibits
at least one of the following problems: conditional stability, artificial damping, and phase errors.
Newmark-beta Operator
This operator is probably the most popular direct integration method used in finite element analysis. For linear
problems, it is unconditionally stable and exhibits no numerical damping. The Newmark-beta operator can effectively
obtain solutions for linear and nonlinear problems for a wide range of loadings. The procedure allows for change of
time step, so it can be used in problems where sudden impact makes a reduction of time step desirable. This operator
can be used with adaptive time step control. Although this method is stable for linear problems, instability can develop
if nonlinearities occur. By reducing the time step and/or adding (stiffness) damping, you can overcome these problems.
Generalized-alpha Operator
One of the drawbacks of the existing implicit operators is the inability to easily control the numerical dissipation. While
the Newmark-beta method has no dissipation and works well for regular vibration problems, the Single-Step Houbolt
method has numerical dissipation and works well for implicit dynamic contact problems. A single scheme that easily
allows zero/small dissipation for regular structural dynamic problems and high-frequency numerical dissipation for
dynamic contact problems is desirable. In [Ref. 1], a Generalized-alpha method has been presented as an
unconditionally stable, second-order algorithm that allows user-controllable numerical dissipation. The dissipation is
controlled by choosing either the spectral radius S of the operator or alternatively, two parameters f and m . The
choice of the parameters provides a family of time integration algorithms that encompasses the Newmark-Beta,
Single-Step Houbolt, Hilber-Hughes-Taylor (HHT) [Ref. 2], and Wood-Bossak-Zienkiewicz (WBZ) [Ref. 3] time
integration methods as special cases.
Technical Background
Consider the equations of motion of a structural system:
Ma + Cv + Ku = F (3-22)
Main Index
122 Nonlinear Users Guide
Transient Dynamic Analysis
where M , C , and K are mass, damping, and stiffness matrices, respectively, and a , v , u , and F are acceleration,
velocity, displacement, and force vectors. Various direct integration operators can be used to integrate the equations of
motion to obtain the dynamic response of the structural system. The technical background of the direct integration
operators available in MSC Nastran is described below.
Newmark-beta Operator
The generalized form of the Newmark-beta operator is
u n + 1 = u n + tv n + 1 2 t 2 a n + t 2 a n + 1 (3-23)
v n + 1 = v n + 1 ta n + ta n + 1 (3-24)
n
where superscript denotes a value at the nth time step and u , v , and a take on their usual meanings.
1
---------- 1
- M + --------- C + K u = F n + 1 R n + M a n + --------- v n + Cv n (3-25)
t 2 t t
The particular form of the dynamic equations corresponding to the trapezoidal rule
= 12, = 14
results in
4
------- 2 4
M + ----- C + K u = F n + 1 R + M a n + ----- v n + Cv n
n
(3-26)
t 2 t t
R = T dv (3-27)
V
u n + 1 = u n + u (3-28)
Notice that the operator matrix includes K , the tangent stiffness matrix. Hence, any nonlinearity results in a
reformulation of the operator matrix. Additionally, if the time step changes, this matrix must be recalculated because
the operator matrix also depends on the time step.
Main Index
Chapter 3: Nonlinearity and Analysis Types 123
Transient Dynamic Analysis
where
n+1+ n+1
u f = 1 + f u f un (3-30)
n+1+
f n+1 n
v = 1 + f v f v (3-31)
n+1+
m n+1 n
a = 1 + m a m a (3-32)
The displacement and velocity updates are identical to those of the Newmark algorithm
u n + 1 = u n + tv n + 1 2 t 2 a n + t 2 a n + 1 (3-33)
v n + 1 = v n + 1 ta n + ta n + 1
where, as shown in [Ref. 1], optimal values of the parameters and are related to f and m by
1 2
= --- 1 + m f (3-34)
4
1
= --- + m f (3-35)
2
It is seen that the f and m parameters can be used to control the numerical dissipation of the operator. A simpler
measure is the spectral radius S. This is also a measure of the numerical dissipation; a smaller spectral radius value
corresponds to greater numerical dissipation. The spectral radius of the generalized alpha operator can be related to the
f and m parameters as follows
S
f = ------------- (3-36)
1+S
1 2S
m = ---------------- (3-37)
1+S
S varies between 0 and 1. Accordingly, the ranges for the f and m parameters are given by 0.5 f 0.0 and
0.5 m 1 . f = -0.5, m = -0.5 corresponds to a spectral radius of 1.0 and f = 0, m = 1 corresponds to a
spectral radius of 0.0. It can also be noted that the case of S = 1 has no dissipation and corresponds to a mid-increment
Newmark- beta operator.
Main Index
124 Nonlinear Users Guide
Transient Dynamic Analysis
accelerations. With even larger time steps, the velocities start oscillating. With still larger steps, the displacement
eventually oscillates. In nonlinear problems, instability usually follows oscillation. When using adaptive dynamics,
you should prescribe a maximum time step.
As in the Newmark-beta operator, the time step in Houbolt integration should not exceed 10 percent of the period of
the highest frequency of interest. However, the Houbolt method not only causes phase errors, it also causes strong
artificial damping. Therefore, high frequencies are damped out quickly and no obvious oscillations occur. It is,
therefore, completely up to the engineer to determine whether the time step was adequate.
For the Generalized-alpha operator, depending on the chosen parameters, the integration scheme can vary between the
Newmark-beta operator and the Single-step houbolt operator. For spectral radii < 1, there is artificial damping in the
system. Depending on the type of problem, the Generalized-alpha parameters and the associated time step should be
carefully chosen to reduce phase errors and effects of artificial damping.
In nonlinear problems, the mode shapes and frequencies are strong functions of time because of plasticity and large
displacement effects, so that the above guidelines can be only a coarse approximation. To obtain a more accurate
estimate, repeat the analysis with a significantly different time step (1/5 to 1/10 of the original) and compare responses.
Defaults
If no contact is present and no nonlinearities and no damping, by default the HHT procedure is used with f = 0.05
and m = 0.0 .
If contact is present or nonlinearities or damping, by default the WBZ procedure is used with f = 0.0 and
m = 1.0 .
Newmark-Beta
The Newmark-beta procedure is default when using the TSTEPNL bulk data entry. When using the NLSTEP bulk data
entry, the Generalized-apha method is the default. The exact Newmark-Beta scheme is not derivable from the
Generalized-alpha Method, but a very close operator can be obtain by using f = 0 and m = 0 which results in
1 1
the classic unconditionally stable (for linear systems) of = --- and = --- , and equalibrium is satisfied at the end
4 2
of the increment. For the mid increment Newmark-beta operator, one defines f = 0.5 and m = 0.5 ; in which
case, the equalibrium is satisified at the mid increment.
Single-Step Houbolt
3 1
The single-step Houbolt operator may be obtained by choosing f = 0 and m = 1 with 1 = --- and = --- .
2 2
Main Index
Chapter 3: Nonlinearity and Analysis Types 125
Transient Dynamic Analysis
HHT
The HHT scheme is obtained by using 0.33 f 0.0 and m = 0 .
WBZ
The WBA scheme is obtained by using f = 0 and 0 m 1 .
Generalized Alpha
The Generalized Alpha scheme is obtained by using 0.5 f 0.0 and 0.5 m 1.0 .
Heat Transfer
Numerical damping may be specified through the value of NDAMP only. NDAMPM is not used. NDAMP can be varied
in the range of (-2.414,0.414). At these outer limits, the transient scheme reduces to the Backward-Euler method. Any
value that is outside this range is automatically reset to the closest outer limit. For NDAMP = 0.0, the transient scheme
reduces to the Crank-Nicholson scheme. The default value of NDAMP is -0.05.
G 1
B = B + B + ------- K + ------- G E K E
1 2
W3 W4
1
B damping elements (CVISC, CDAMPi) + B2GG
2
B B2PP direct input matrix + transfer functions
Main Index
126 Nonlinear Users Guide
Transient Dynamic Analysis
W4 frequency of interest in radians per unit time (PARAM,W4) for conversion of element structural damping into
equivalent viscous damping
KE element stiffness matrix
Transient response analysis does not permit the use of complex coefficients. Therefore, structural damping is included
by means of equivalent viscous damping. To appreciate the impact of this on the solution, a relation between structural
damping and equivalent viscous damping must be defined.
The viscous damping force is a damping force that is a function of a damping coefficient b and the velocity. It is an
induced force that is represented in the equation of motion using the B matrix and velocity vector.
M u t + B u t + K u t = P t (3-38)
The structural damping force is a displacement-dependent damping. The structural damping force is a function of a
damping coefficient G and a complex component of the structural stiffness matrix.
M u t + 1 + iG K u t = P t (3-39)
Assuming constant amplitude oscillatory response for an SDOF system, the two damping forces are identical if
Gk = b (3-40)
or
Gk
b = ------- (3-41)
Therefore, if structural damping G is to be modeled using equivalent viscous damping b , then the equality equation
(3-41) holds at only one frequency (see Figure 3-12).
Main Index
Chapter 3: Nonlinearity and Analysis Types 127
Transient Dynamic Analysis
Equivalent
Viscous
b = Gk 3 (or 4 )
Damping
3 (or 4 )
f v = bu = ibu
Figure 3-12 Structural Damping Versus Viscous Damping (Constant Oscillatory Displacement)
Two parameters are used to convert structural damping to equivalent viscous damping. An overall structural damping
coefficient can be applied to the entire system stiffness matrix using PARAM,W3,r where r is the circular frequency at
which damping is to be made equivalent. This parameter is used in conjunction with PARAM,G, GFL. The default value
for W3 is 0.0, which causes the damping related to this source to be ignored in transient analysis.
PARAM,W4,r is an alternate parameter used to convert element structural damping to equivalent viscous damping.
PARAM,W4,r is used where r is the circular frequency at which damping is to be made equivalent. PARAM,W4 is used
in conjunction with the GE field on the MATi entry. The default value for W4 is 0.0 which causes the related damping
terms to be ignored in transient analysis.
Units for PARAM,W3 and PARAM,W4 are radians per unit time. The choice of W3 or W4 is typically the dominant
frequency at which the damping is active. Often, the first natural frequency is chosen, but isolated individual element
damping can occur at different frequencies and can be handled by the appropriate data entries.
Initial conditions u 0 and u 0 are used to determine the values of u 1 , P 0 , and P 1 used in equation (3-42)
to calculate u 1 .
u 1 = u 0 u 0 t (3-42)
P 1 = K u 1 + B u 0 (3-43)
Main Index
128 Nonlinear Users Guide
Transient Dynamic Analysis
P 0 = K u 0 + B u 0 (3-44)
Regardless of the initial conditions specified, the initial acceleration for all points in the structure is assumed to be zero
(constant initial velocity).
Initial conditions should only be specified in the a-set.
Main Index
Chapter 3: Nonlinearity and Analysis Types 129
Nonlinear Transient Response Analysis
Case Control
Each subcase and step defines a time interval starting from the last time step of the previous subcase or subcase,
subdivided into smaller time steps using the NLSTEP entry. The output time is labeled by the cumulative time,
including all previous subcases. The data blocks containing solutions are generated at the end of each subcase for
storage in the database for output processing and restarts. As such, converged solutions are apt to be saved at many
intermediate steps in case of divergence and more flexible control becomes possible with multiple subcases. Results
from converged increments can be output to *.op2 files using the NLOPRM case control entry.
The input loading functions may be changed for each subcase or continued by repeating the same DLOAD request.
However, it is recommended that one use the same TLOAD bulk data for all the subcases in order to maintain the
continuity between subcases, because TLOADi data defines the loading history as a function of cumulative time. Static
loads (PLOADi, FORCEi, MOMENTi) may be associated with time-dependent functions by matching the EXCITEID
on the TLOADi entries. Nonlinear forces as functions of displacements or velocities (NOLINi) may be selected and
printed by the case control commands NONLINEAR and NLLOAD, respectively. Each subcase may have a different
Main Index
130 Nonlinear Users Guide
Nonlinear Transient Response Analysis
time step size, time interval, and iteration control selected by the NLSTEP request. Case control requests that may not
be changed after the first subcase are SPC, MPC, DMIG, and TF.
Output requests for each subcase are processed independently. Requested output quantities for all the subcases are
appended after the computational process for actual output operation. See Chapter 8 for a discussion on output requests
and see the QRG for a complete list of output requests.
Initial conditions (displacement or velocity) can be specified by the bulk data input, TIC, selectable by the case control
command IC. If initial conditions are given, all of the nonlinear element forces and stresses must be computed to satisfy
equilibrium with the prescribed initial displacements. On the other hand, initial conditions can be generated by
applying static analysis for the preload using PARAM,TSTATIC in the first subcase. Then the transient analysis can be
performed in the ensuing subcases. Associated with the adaptive time stepping method, the PARAM,NDAMP is used to
control the stability in the ADAPT method. The NDAMP parameter represents the numerical damping (a recommended
value for usual cases is 0.01), which is often required to improve the stability and convergence in contact problems.
Main Index
Chapter 3: Nonlinearity and Analysis Types 131
Creep in SOL 400
Parameters A , R , and K are specified in the following form, as recommended by Oak Ridge National
Laboratory:
R ce
d j=1 c
d j=2
K e sinh f
g k=1 ee
f k=2
Type ijk where i , j , and k are digit equal to 1 or 2, according to the desired function in the table above. For example,
b d f
TYPE=122 defines A = a , R = c , and K = ee
Main Index
132 Nonlinear Users Guide
Creep in SOL 400
More details are provided in the MATVP description of the QRG. The options offered by MSC Nastran for modeling
creep are as follows:
Creep data can be entered directly through the MATVP bulk data entry. The form of the creep is designated
with either POWER for empirical creep law or TABLE for a tabular input of creep model parameters.
An automatic time stepping scheme can be used via NLSTEP to maximize the time step size in the analysis.
Main Index
Chapter 3: Nonlinearity and Analysis Types 133
Heat Transfer
Heat Transfer
Heat transfer problems including conduction, forced convection, and boundary radiation and convection can be solved
with MSC Nastran SOL 400. The problems can be transient or steady-state, linear or nonlinear. The heat transfer
element library includes elements that allow for heat storage (specific heat) and heat conduction, and elements that also
allow for forced convection caused by fluid flowing through the mesh. Heat interface elements are also provided, to
model the heat transfer across the boundary layer between a solid and a fluid, or between two closely adjacent solids.
Shell-type heat transfer elements are included, since so many structures are of this type. The second-order elements
usually give more accurate results for the same number of nodes in the mesh.
This section discusses pure heat transfer problems, where the temperature field can be found without knowledge of the
stress and deformation of the bodies being studied.
Time integration in transient problems is done with the backward Euler method (the modified Crank-Nicholson
operator) in the pure conduction elements. This method is unconditionally stable for linear problems. The
convection/diffusion elements use the trapezoidal rule for time integration, and include numerical diffusion control
(the upwinding Petrov-Galerkin method) and, optionally, numerical dispersion control. The elements with
dispersion control offer improved solution accuracy in cases where the transient response of the fluid is important but,
because artificial dispersion control introduces a stability limit on the size of the time increment, it may be more
economical to use the elements without this feature in transient cases where transient effects in the fluid itself are not
a critical part of the solution.
For all transient heat transfer cases, the time increments may be specified by you or selected automatically based on a
user prescribed maximum nodal temperature change in a step. Automatic time incrementation is generally preferred.
The system of equations for the convective/diffusive elements is not symmetric, and so the nonsymmetric matrix
storage and solution scheme is automatically invoked when these elements are used. Heat transfer problems are very
often solved to provide temperature fields for subsequent thermal stress analyses. MSC Nastran SOL 400 provides a
very simple interface which uses the temperatures at the nodes from the heat transfer analysis to define the temperature
field at different times in the stress analysis. The interface assumes that corresponding nodes in the stress analysis mesh
have the same numbers as they have in the heat transfer analysis mesh. Nodes may be removed for the stress problem;
for example, elements that represent non-structural parts of the heat transfer mesh (such as insulation or cooling fluid)
may be omitted in the stress analysis. When these temperatures are read into the stress analysis problem, temperatures
Main Index
134 Nonlinear Users Guide
Heat Transfer
at nodes that are not in the mesh are ignored. Heat transfer problems may be nonlinear because the material properties
are temperature dependent or because the boundary conditions are nonlinear. Usually the nonlinearity associated with
temperature dependent material properties is mild, because the properties do not change rapidly with temperature
except when latent heat effects are included, in which case the analysis may be severely nonlinear. Boundary
conditions are very often nonlinear: film coefficients may be functions of surface temperature, or radiation may be
occurring. Again, the nonlinearities are often mild and cause little difficulty. An exception is the boiling film
condition, in which the film coefficient can change very rapidly because the fluid adjacent to the surface boils.
Generally, such problems are handled effectively by MSC Nastran SOL 400 when they are modeled as transient
problems. For this reason, steady-state cases involving severe nonlinearities are sometimes more effectively solved as
transient cases, obtaining the required steady-state solution as the very long time response and using the transient
simply to stabilize the solution for that long time response. MSC Nastran SOL 400 uses an iterative scheme to solve
nonlinear heat transfer problems. The scheme is basically a Newton method, with some modification to ensure stability
of the iteration process. The iteration is governed by a user set parameter giving the maximum allowable change in the
temperature field from iteration to iteration, hence controlling the error in property or boundary condition evaluation.
Boundary conditions provided for heat transfer analysis include prescribed flux (distributed or concentrated),
prescribed temperature, film, and radiation conditions. Any of these can be specified as a function of time.
Main Index
Chapter 3: Nonlinearity and Analysis Types 135
References
References
1. Chung, J. and Hulbert, G.M., A time integration algorithm for structural dynamics with improved numerical
dissipation: The Generalized-Method, Journal of Applied Mechanics, Vol. 60, pp. 371 - 375, June 1993.
2. Hilber, H.M., Hughes, T.J.R., and Taylor, R.L., Improved Numerical Dissipation for Time Integration
Algorithms in Structural Dynamics, Earthquake Engineering and Structural Dynamics, Vol. 5, pp. 283-292,
1977
3. Wood, W.L., Bossak, M., and Zienkiewicz, O.C., An Alpha Modification of Newmarks Method,
International Journal for Numerical Methods in Engineering, Vol., 15, pp. 1562-1566, 1981.
Main Index
136 Nonlinear Users Guide
References
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis
Nonlinear Users Guide
Main Index
138 Nonlinear Users Guide
Introduction
Introduction
MSC Nastran SOL 400 provides a comprehensive FE solution for multi-physics problems such as structure analysis,
thermal analysis, as well coupled analysis. Even though it is originally targeted to deal with various nonlinearities as
geometry nonlinearity, material nonlinearity, as well boundary nonlinearity (contact), it is also embedded with the
powerful capabilities of MSC Nastran on linear analyses.
The FEM is a powerful tool for analyzing complex problems in structural and continuum mechanics. The analysis of
a structure using the FEM has four basic steps:
1. Modeling, in which the structure is subdivided into an assemblage of discrete volumes called finite elements,
and properties are assigned to each element.
2. Evaluation of element characteristics, such as stiffness and mass matrices, followed by assembling the element
characteristic matrices to obtain the assembled or global matrices characteristic of the entire structure. A
similar process is followed to obtain the total loads, in vector form, applied to the structure.
3. Solution of the system equations for displacements, natural frequencies and mode shapes, or buckling load
factors.
4. Calculating other quantities of interest, such as strains, stresses and strain energy.
SOL 400 uses the finite element displacement method, in which a system of equations is solved to obtain the
displacements at all node points of the structure. Comprehensive presentations of the FEM together with numerous
applications are available in textbooks and the research literature.
Solving the equations of finite element analysis involves the manipulation of large matrices of numbers, which is best
done using computers. While modern computers are extremely fast and have vast amounts of memory, the bottom
line in finite element analysis is that it is very easy to discretize a structure to the point that it is considered to be a
large problem. Typically, the geometry of the problem will dictate how fine of a mesh is required to get an acceptable
solution. While modern graphical user interfaces (GUI's) such as Patran can provide some guidance and tools for
evaluating the mesh density, in the end it is up to the user to ensure that the problem is adequately discretized. The size
of FEA problems is typically measured by the number of degrees of freedom (or degrees of freedom) in the finite
element mesh. For nonlinear problems, the size is even more important because the methods used to solve these
nonlinear equations are usually iterative in nature, meaning that the system of equations must be solved many, many
times to follow the behavior of the structure as it changes. The changes characterized may be the shape (called large
deformation), or the status of the material (metal materials typically yield when a certain stress level is exceeded), or
loading changes (the load orientation may follow the deformation of the structure, or different parts of the structure
may come into contact changing the load path). The purpose of this chapter is to describe the numerical methods and
procedures required to solve the linear and nonlinear equations used to perform finite element simulations.
We begin this chapter with some general observations and recommendations on how to approach the solution to
nonlinear FEA problems, and then move through the details of how to get an accurate, efficient solution using the
algorithms and methods available in MSC Nastran SOL 400. Along with the efficiency, we discuss the concept of
robustness, which means choosing alogorithms that will tolerate changes in the problem (sometimes abrupt, such as
contact characterizing impact) with out causing the solution algorithms to fail. A robust numerical method automates
changes in the solution parameters, such as load increment size, as required by the problem yet still provides a
reasonably accurate and efficient solution.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 139
Introduction
The characteristics of the solution methods and strategies in this chapter include: solution iteration methods, transient
analysis effects such as time step selection, and time integration method, arc-length methods for post-buckling, and
convergence measurements and controls. This chapter may be read as a tutorial, but is probably best used as a reference
when considering specific problems encountered while trying to solve a nonlinear finite element problem.
Main Index
140 Nonlinear Users Guide
Nonlinear Characteristics and General Recommendations
Nonlinear analysis requires deeper insight into structural behavior. First of all, the type of nonlinearities involved must
be determined. If there is a change in constraints due to contact during loading, the problem may be classified as a
boundary nonlinear problem and would require contact body creation and contact modeling. The material nonlinearity
is characterized by material properties. However, the material nonlinear effects may or may not be significant
depending on the magnitude and duration of the loading, and occasionally on environmental conditions. The
anticipated stress level would be a key to this issue. The geometric nonlinearity is characterized by large rotations
which usually cause large displacements. Intuitively, geometric nonlinear effects should be significant if the deformed
shape of the structure appears distinctive from the original geometry without amplifying the displacements. There is
no distinct limit for large displacements because geometric nonlinear effects are related to the dimensions of the
structure and the boundary conditions. Strain is nondimensional, and a strain greater than 2% indicates a geometrically
nonlinear problem.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 141
Differences Between Linear and Nonlinear Analysis
Main Index
142 Nonlinear Users Guide
Differences Between Linear and Nonlinear Analysis
You are guaranteed a solution if the boundary conditions and material properties are set up correctly. The stiffness
matrix is assembled and solved only once in the entire analysis.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 143
Differences Between Linear and Nonlinear Analysis
Newton-Raphson Iteration scheme begins: Apply a portion of the total load to start: (1% in this case):
Assembly of stiffness Matrix.
Solution of the Stiffness Matrix.
Check for convergence (this is an important step which is only seen in nonlinear analysis).
If converged, the solution/structure is in equilibrium. Go to step 3.
If not converged, update information and reassemble, resolve stiffness matrix.
Keep iterating till convergence is achieved.
3. (After convergence) Get displacements, strains, stresses.
4. Apply the next increment of load and go to Step 2. Keep doing this until all the load is applied.
5. View Results.
Iterations
In the incremental solution process, the unbalanced forces that occur during a load increment are reintroduced
internally into the solution until the solution has converged. The process of redistributing the unbalanced force within
a load increment is known as an iteration. The iteration is the lowest level of the solution process. Iterations continue
within a load increment until the solution converges or any of the specified convergence parameters are exceeded. A
complete description of the numerical procedure used to solve the nonlinear problem is given in this chapter.
Main Index
144 Nonlinear Users Guide
Differences Between Linear and Nonlinear Analysis
Input Phase:
Read Input Data
Space Allocation
Data Check
Incremental
Loads
Equivalent Nodal
Load Vector
Matrix Assembly
Iteration Loop
Matrix Solution
Time Step Loop
Stress Recovery
No
Convergence
Yes
Output Phase
Yes Next
Increment
No
Stop
The important point to note is that the total load is applied gradually in steps (or increments) and for each load step,
the solution is arrived at after one or more iterations. If the behavior of the model is generally linear, few iterations are
required to solve that load step. If the model behavior is complex/nonlinear, many iterations might be required. Each
iteration involves an assembly and solution of the stiffness matrix. Hence, nonlinear problems inherently take longer
than linear models (of the same size) to solve. At the end of each iteration, a check is made to see if the solution has
converged. If the convergence check fails, the iteration is repeated with the new information. This process repeats until
convergence is achieved. Following that, the next increment of load is applied. The load increments are applied until
the full load of the model is solved.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 145
Differences Between Linear and Nonlinear Analysis
To achieve accurate results, three key points have to be paid attention to:
Iteration method;
Load increment control; and
Convergence criterion.
The model may contain superelements, but only the residual structure (superelement 0) may consist of nonlinear
elements, mixed with any type of linear elements. As aforementioned, other potentially nonlinear elements in the
residual structure become actively nonlinear only if a LGDISP or LRGSTRN is used and/or if they use the nonlinear
material data specified on any one of the entries: MATEP, MATVE, MATVP, MATF, MATS1, MATS3, MATS8,
MATSORT, MATG, MATHE, MATHP, MATSMA, and MCOHE, and/or perturbation analyses.
Main Index
146 Nonlinear Users Guide
Differences Between Linear and Nonlinear Analysis
Normal rotation for the "drilling degree of freedom" of shell elements restrained by the default value of 100 on
the K6ROT parameter when the geometric nonlinearity is involved. In rare cases it may be necessary to adjust
this value. This can help with convergence, but may also affect the results.
Understand the basic theory of plasticity, creep, or rubber elasticity before using these capabilities.
The time step size for a transient response analysis should be carefully considered based on the highest natural
frequency of interest because it has significant effects on the efficiency as well as the accuracy of the solution.
The automated procedures used by NLSTEP is adequate for this purpose.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 147
Equations Governing Nonlinear Analysis
The constitutive relations represented by stress-strain relations, e.g., for linear elasticity,
ij = D ij kl kl
1 u i u j
ij = --- -------- + --------
2 x j x i
where u i are displacements.
These systems of governing differential equations must be satisfied for every infinitesimal element throughout the
domain of the continuum. The complete set of state variables, namely displacements, may be determined by solving
these systems of equations supplemented by boundary conditions, and in dynamic situations by initial conditions as
well. For the nonlinear problems, the governing equations should be satisfied throughout the history of load
application. The material nonlinearity is manifested in the constitutive relations. The geometric nonlinearity is
pronounced in the strain-displacement relations, but it also affects the equilibrium equation by changing applied loads.
Changes in constraints affect the boundary conditions, which constitute contact problems.
Most of the known analytical solutions for the solid mechanics problems are based on ideal geometry and linear
approximations. However, the real nature is more complicated and inherently nonlinear. The linear system is a very
particular case of a general problem. Even the nonlinear solutions that we seek deal with only a small subset of special
cases in a general category of nonlinear problems. When the nonlinear system is confronted, no general mathematical
solutions exist and superposition no longer applies. The system may even be non-conservative.
The first phase of the structural analysis is the idealization of a physical system into a simpler and more manageable
engineering problem. The idealization process involves simplifications of the geometry, boundary and joint conditions,
and loading conditions, etc. using engineering intuitions, experimental data, empirical observations, and classical
solutions. If the idealized structural system renders a problem that cannot be resorted to a classical method of analysis,
further idealization is required, namely discretization, for numerical analysis.
For the discrete system, governing differential equations are converted to algebraic equations. The finite element
model represents a structure by an assemblage of finite elements interconnected at nodal points. State variables are the
displacements (displacement method or stiffness approach) of the nodal points which carry fictitious forces
representing distributed stresses actually acting on the element boundaries. The equilibrium requirements are satisfied
at nodal points by the nodal force balance. The material constitutive laws are satisfied at the integration points of the
element. The compatibility is ensured by the displacement continuity between elements.
Main Index
148 Nonlinear Users Guide
Equations Governing Nonlinear Analysis
= U+W (4-1)
where U is the strain energy of the system and W is the potential energy of the external loads. The equilibrium
equations can be obtained by invoking the principle of virtual work or the Ritz method, i.e.
= 0 or -------------- = 0 (4-2)
u
which implies that the total potential of the system must be stationary with respect to the state variables (displacement)
for equilibrium to be ensured. The functional II is so called because it involves the integral of implicit functions of the
state variables, u .
Considering a three-dimensional continuum for a nonlinear problem, the stationary condition results in
i j ij dV = b i u i dV + t i u i dS + p i u i (4-3)
V V S i
where the dots and denote infinitesimal increments and arbitrary variations, respectively. The left-hand side
represents variations in the strain energy increment and the right-hand side represents variations in the external work
which consists of body forces b i (such as a gravity load), traction forces t i at the boundary surface (such as pressure
loads), and concentrated forces p i . Now it remains to determine admissible functions expressing the arguments of the
functional II in terms of state variables u , which are valid throughout the whole region and satisfy the boundary
conditions.
The FEM can be characterized by the following features distinguished from the conventional Ritz methods or the
matrix method for frame structures:
The whole region of the system is divided into numerous sub-domains, called finite elements, which have
simple geometrical shapes.
The variational process is limited to each finite element, which aggregates into a whole region when
assembled.
The admissible displacement field within each element, u , can be expressed in terms of nodal
displacements using interpolation functions known as shape functions, N, i.e.,
u = N u (4-4)
The strain displacement relations for the element can then be established in terms of nodal displacements using
the shape functions in equation (4-4), i.e.,
= B u (4-5)
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 149
Equations Governing Nonlinear Analysis
where
T
= x y z x y y z zx (4-6)
and the element matrix B consists of derivatives of the shape functions, evaluated at the current deformed
geometry. Notice that the geometric linear problem requires that the element matrix be evaluated only at the
initial geometry. In SOL 400, when LGDISP is used in conjunction with the advanced elements, a consistent
updated Lagrange technique is used. MSC Nastran SOL 106 and SOL 129 employ an approximate updated
Lagrangian approach for geometric nonlinear problems, by which linear strains are computed in the updated
element coordinate system in order to eliminate the effects of the rigid body rotation but the equilibrium is
established at the final position in the stationary coordinate system. This method does not require
reevaluation of the element matrix [B] (constant in the absence of large strains) while the element coordinates
are reevaluated continuously.
Equilibrium equations for an element may be obtained by reducing equation (4-3) after the substitution of
equations (4-4) and (4-5), based on the small deformation theory. Then the element boundary stresses are
statically equivalent to the nodal forces which balance the applied external loads, i.e.,
e e
F = P (4-7)
with
e T
F = B dV (4-8)
V
and
e T T
P = N b dV + N s t dS + p (4-9)
V S
where N s is an appropriate interpolation function for the traction force. equilibrium equation for an
incremental load may be expressed as
F = dV = P
T
B (4-10)
V
where should be components of co-rotational stress which is independent rotation.
The element stiffness matrix can be obtained by substituting the constitutive relations into equation (4-10); i.e.,
= D (4-11)
where
T
= x y z x y yz z x
Main Index
150 Nonlinear Users Guide
Equations Governing Nonlinear Analysis
and D is a material tangent matrix. The nodal forces of an element can then be expressed as
e
F
T e
= B dV = K u (4-12)
V
e T
K = B D B dV (4-13)
V
Notice that this expression represents an element stiffness due to the material stiffness without geometric
nonlinear effects. As will be shown later, an additional stiffness K d due to initial stresses should be included
for an incremental process because the initial stresses exist from the second increment.
The equilibrium must be satisfied in the whole region throughout the complete history of load application. Equilibrium
equations for the global discrete system are obtained when all the elements are assembled, i.e.,
T e
B dV = P (4-14)
m V m
where over m denotes a summation over all elements. For the incremental process, the equilibrium equation may
be rewritten as
T 0
B dV = P (4-15)
m V
with
e T 0
P = P B dV (4-16)
m m V
0
where represents an initial stress or the stress state at the preceding load step.
Because of the approximations involved in the interpolation functions, the finite element model provides an
approximate solution even if the equilibrium (equation (4-14)) is satisfied exactly. Consequently, the differential
equations of equilibrium are not satisfied exactly even for linear problems, but the error decreases as the finite element
mesh is refined. This convergence condition is ensured by using conforming elements that satisfy the patch test. With
a displacement approach, the finite element model is generally known to produce a stiffer structure than in reality.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 151
Load Incrementation and Iteration
Main Index
152 Nonlinear Users Guide
Load Incrementation and Iteration
operation and element force calculation. From the efficiency point of view, the number of stiffness matrix updates and
the number of iterations should be minimized, which may be conflicting requirements.
Adaptive features implemented in MSC Nastran nonlinear capabilities are:
Newton's iteration for static and implicit dynamic analysis
Adaptive stiffness matrix update strategies
Selective BFGS updates
Selective line search processes
Adaptive bisection and recovery of load increment
Arc-length methods for static post-buckling and snap-through problems
Crisfield's arc-length method
Riks and modified Riks methods
Adaptive arc-length adjustment
Selective BFGS Updates
Adaptive switching algorithm for limiting cases
Adaptive correction in case of path reversal
Direct time integration for transient response analysis
Quadratically accurate dynamic operators
Automatic time step adjustment
Pg + Qg Fg = 0 (4-17)
where P g , Q g , and F g represent vectors of applied loads, constraint forces, and element nodal forces,
respectively. Element nodal forces are nonlinear functions of displacements for nonlinear elements. Since the
equilibrium condition is not immediately attained in the presence of nonlinear elements, an iterative scheme such as
the Newton-Raphson method is required. Since the error vanishes at constrained points and the constraint forces vanish
at free points, the unbalanced forces acting at nodal points at any iteration step are conveniently defined as an error
vector by
Ra = Pa Fa (4-18)
Notice that the a-set is equivalent to the 1-set in the nonlinear analysis because the r-set does not exist. The subscript a
will be dropped for simplicity in the following discussion.
Based on Newton's method [equations (4-17) and (4-18)], a linearized system of equations is solved for incremental
displacements by Gaussian elimination in succession. The Jacobian of the error vector emerges as the tangential
stiffness matrix. The equation to solve at the i-th iteration is
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 153
Load Incrementation and Iteration
i i1
K T u = R (4-19)
where
R F
K T = ------- = ------- (4-20)
u u = u' u u = u'
i i i1
u = u u (4-21)
and
i i
R = P Fu (4-22)
The iteration continues until the residual error R and the incremental displacements u become negligible, which
is signified by the convergence criteria.
The tangential stiffness consists of the geometric stiffness in addition to the material stiffness, i.e., without regard to
the coordinate transformation,
T m d
K T = ------ B dV = K + K (4-23)
u
V
m d
where K and K refer to the material and the differential stiffnesses, respectively. The material stiffness is given in
equation (4-13) with a material tangential matrix for D . The differential stiffness, which is caused by the initial
stress, is defined as follows:
d B T T
K = ----------N- dV = G M G dV (4-24)
u
V V
where B N represents the second order effects in the strain-displacement relations, G consists of derivatives of shape
functions and M is a function of stresses. Notice that the initial displacement stiffness is not included in K T because
its effects are already eliminated in the element formulation.
Then the residual load error is automatically carried over to the next incremental process. The merit of the Newton-
Raphson method is the quadratic rate of convergence, i.e.,
2
u* ui + 1 q u* ui (4-25)
where u* is a true value of {u}, q is a constant, and represents a vector norm. From a practical standpoint,
however, determination of the tangential stiffness and its inverse at each iteration entails a considerable amount of
computation. As Figure 4-3 suggests, one may resort to the modified Newton's method which requires the tangential
stiffness to be evaluated just once at the initial position, u 0 , and used thereafter to solve for u i . However, more
iterations are required for a given accuracy by the modified Newton's method.
Main Index
154 Nonlinear Users Guide
Load Incrementation and Iteration
Load
3
2
Iteration Count
1
Deflection
(a) Newton-Raphson Method
Load
9
5 678
34
Iteration Count
2
1
Deflection
(b) Modified Newton Method
The Gaussian elimination method is better suited for this approach than the iterative solver method because the
decomposition is performed only once.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 155
Load Incrementation and Iteration
Initial Tangent
Starting
Point
u
0
u u * u2 u1
Figure 4-4 Hybrid Newton Method based on Stiffness Matrix Update Strategy
Update Principles
Newton's method could be trapped in an infinite loop, oscillating about the local maximum as illustrated in
d
Figure 4-5(a). This difficulty is overcome during the Newton's iteration by discarding the differential stiffness, K ,
when the tangential stiffness is not positive definite as shown in Figure 4-5(b). However, the non-positive definite
stiffness matrix will be retained and used for iteration if the arc-length method is used for the static solution.
From the users point of view, the stiffness update process is primarily controlled by the input data in KMETHOD
(PFNT, FNT, SEMI, ITER, or AUTO) and KSTEP (integer) fields of the NLSTEP bulk data entry.
Main Index
156 Nonlinear Users Guide
Load Incrementation and Iteration
Load
P
2 4
5 Iteration
sequence
1
Deflection
(a) Iteration stops at 3 with PARAM,TESTNEG = -1 if the stiffness becomes negative.
Iteration will be trapped if continued with TESTNEG = +1 or 0.
Load
P
Km
1
3
Kd omitted for these
wo iterations 5
4
Deflection
(b) Iteration sequence drops the differential stiffness with PARAM,TESTNEG = -2
when the negative diagonal terms appear in the stiffness matrix.
Figure 4-5 Iteration Trap Condition
Quasi-Newton Method
The quasi-Newton update method is employed as an option to accelerate the convergence and improve the overall
effectiveness of the modified Newton's iteration. Experience shows that the BFGS method has a significant impact on
the program's performance.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 157
Load Incrementation and Iteration
Note: The BFGS method cannot be used in a simulation that includes contact.
Quasi-Newton update methods are considered the most sophisticated methods of the Newton's iterative solution
schema and represent the culmination of extensive algorithm development for Newton's iterative process. They have
been developed for efficient nonlinear iterations by approximating the inverse Hessian matrix in place of the true
inverse that is required in Newton's method. Quasi-Newton methods have been widely and successfully used in
nonlinear optimization applications. One of the quasi-Newton techniques, known as BFGS (Broyden-Fletcher-
Goldfarb-Shanno) update, was introduced into the FEM by Matthies and Strang.
By the BFGS update, the information acquired during the iteration is used to modify the inverse stiffness matrix. This
approximate update to the inverse stiffness matrix results in a secant modulus in the search direction. As these updates
accumulate, the BFGS method renders a stiffness matrix resembling tangential stiffness in the limit. When combined
with the line search, the performance of the BFGS update with respect to effectiveness and efficiency depends largely
on the implementation. It is extremely difficult to tune the tolerances and parameters to the optimal condition, which
may be problem dependent, for the general class of problems.
where R represents an error vector to be minimized and K 0 is a Hessian matrix. In the absence of line searches or
quasi-Newton updates, the second iteration would lead to the next feasible direction:
2 1 1
d = K0 R u (4-27)
with
1 0 1
u = u +d (4-28)
i
Consider a Taylor series expansion of the load error R about u , expressed as
i i
R = Ru Ku u + Oh (4-29)
where
Ru = P Fu (4-30)
with F u being the aggregate vector of element nodal forces. Assuming that the load stiffness due to follower forces
is negligible, the stiffness matrix is formed by
Main Index
158 Nonlinear Users Guide
Load Incrementation and Iteration
R F
K = ------- = ------- (4-31)
u u
i1 i
In view of equation (4-29), the data from two points, u and u , should provide some information about [IC],
because they should satisfy:
1
= K or = K (4-32)
with
i i1
= u u (4-33)
and
i1 i
= R R (4-34)
1
It is natural to attempt constructing successive approximations to K based on the data obtained during the iterative
procedure.
The earliest quasi-Newton scheme was proposed by Davidon and later elaborated by Fletcher and Powell. By this
scheme which is referred to as the DFP method, the inverse Hessian is updated by adding two symmetric rank one
matrices at each iteration. Therefore, the scheme is a rank two correction procedure, i.e.,
1 T 1
1 1
T K i K i
Ki + 1 = Ki + --------- --------------------------- (4-35)
T T 1
Ki
1
Notice that the formula simply satisfies equation (4-32) while preserving positive definiteness and symmetry of K .
It is also possible to update approximations to the stiffness matrix itself, rather than its inverse. Recalling the
1
complementary roles of K and K with respect to the quasi-Newton vectors in equation (4-32), the formula for
[K] is found by interchanging 7 and b, i.e.,
1 T 1
T K i K i
K i + 1 = K i + -------- ---------------------------- (4-36)
T T
Ki
Another way of finding the formula for the inverse stiffness matrix is to invert equation (4-36). This can be done by
applying the general inversion identity known as the Sherman-Morrison formula. The resulting formula is referred to
as the Broyden-Fletcher-Goldfarb-Shanno (BFGS) update for the inverse Hessian, i.e.,
T 1 T 1 1 T
1 1 K i T K i + K i
Ki + 1 = Ki + 1 + ------------------ --------- -------------------------------------------- (4-37)
T T T
Numerical experiments have indicated that the performance of the BFGS method is superior to that of the DFP method.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 159
Load Incrementation and Iteration
An equivalent form of equation (4-37) can be expressed for the j-th BFGS update as:
1 T 1 T
Kj = Cj Kj 1 Cj + zj j j (4-38)
where[
T
Cj = I zj j j (4-39)
and
1
z j = ------------------------- (4-40)
T
j j
This is a recurrence formula employed in MSC Nastran, which is applicable to every pair of quasi-Newton (QN)
vectors. Notice that the index j for the BFGS update may be different from the iteration index i. The stiffness matrix
(therefore, its inverse) is assumed to be symmetric and positive definite throughout the derivation.
Case 4. When two successive data points involved in QN vector generation are too closely spaced.
1
Case 5. When the change in K due to BFGS update is negligible.
Main Index
160 Nonlinear Users Guide
Load Increment Size
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 161
Load Increment Size
for coupled analysis, the NLSTEP entry must be above the first SUBSTEP command. A single NLSTEP entry is used
for all SUBSTEPs of the STEP.
NLSTEP
Describes the control parameters for mechanical, thermal, and coupled analysis in SOL 400 and for linear contact
analysis in SOL 101. Specifies the convergence criteria, step size control and numerical procedure for time/load
stepping in SOL 400. For multi-physics, it controls both structural and thermal analysis. Defines analysis preference
and control parameters for linear contact analysis in SOL 101. There are three groups of data that can be entered
through this option:
1. General data which defined parameters that may be used for a variety of simulations. This data is provided by
the GENERAL keyword.
2. Selecting the type of procedure used to control the time/load stepping procedure. These procedures are
activated by the keywords: LCNT, FIXED, ADAPT, or ARCLN (arc length or continuation method). Only one
of the keywords may be chosen in a loadcase.
3. Data associated with the physics type that are activated by the keywords: MECH, HEAT, COUP, and RCHEAT.
One can enter as many as necessary.
The NLSTEP is selected by the NLSTEP=ID case control command.
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
GENERAL MAXITER MINITER MAXBIS CREEP
FIXED NINC NO
ADAPT DTINITF DTMINF DTMAXF NDESIR SFACT INTOUT NSMAX
IDAMP DAMP CRITTID IPHYS LIMTAR RSMALL RBIG
ADJUST MSTEP RB UTOL
ARCLN TYPE DTINITFA MINALR MAXALR SCALEA NDESIRA NSMAXA
HEAT CONVH EPSUH EPSPH EPSWH KMETHODH KSTEPH
MAXQNH MAXLSH LSTOLH
MECH CONV EPSU EPSP EPSW KMETHOD KSTEP MRCONV
MAXQN MAXLS LSTOL FSTRESS
COUP HGENPLAS HGENFRIC
Main Index
162 Nonlinear Users Guide
Load Increment Size
NLSTEP - GENERAL
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
GENERAL MAXITER MINITER MAXBIS CREEP
ID: invoked by NLSTEP case control command
TOTTIME: total step time to be reached for this step (Default=1.0)
CTRLDEF: the values QLINEAR, MILDLY, and SEVERELY are available along with the option of leaving
blank. These values are used to set the convergence defaults.
MAXITER, MINITER: maximum and minimum number of iterations in one increment (Defaults = 10, 1)
MAXBIS: maximum number of bisections (Default=10). When negative and reached without convergence the
step continues with the next increment (be careful)
CREEP: for creep step (1= creep, 0= no creep)
NLSTEP - Fixed
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
FIXED NINC
NINC: number of increments for fixed time stepping (default = 50). The time step will equal
TOTTIME/NINC.
NLSTEP - MECH
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
MECH CONV EPSU EPSP EPSW KMETHOD KSTEP MRCONV
MAXQN MAXLS LSTOL FSTRESS
CONV: Flags to select convergence criteria. U, P, W,V,N,A and combinations; Default = PV).
U: for displacement error control
P: for residual (load) error control
W: for energy (work) error control
V: for vector component norm used
N: for length norm used
A: for automatic switching of error measure
EPSU, EPSP, EPSW: Error tolerances for displacement (U), load (P) and work (W) criteria. (Defaults = 0.1, 0.1,
0.1). The defaults are adjusted based upon the value of CTRLDEF.
KMETHOD: Method for controlling stiffness updates.
PFNT is pure full Newton-Raphson iteration method which is the default and strongly recommended.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 163
Load Increment Size
ITER is a quasi-Newton method where the stiffness matrix is updated every KSTEP iteration. It should not
be used if contact or severe nonlinearities are included in the model.
KSTEP: Number of iterations before the stiffness update; only used for the ITER method. (Default = 10)
MRCONV: Flag to specify if rotations and moments should be included in the convergence testing when CONV
is set to UPV or UPN: (Default = 3)
0: check on forces, moments, displacements and rotations
1: check on forces, moments and displacements
2: check on forces, displacements and rotations
3: check on forces and displacements
MAXQN: Maximum number of quasi-Newton vectors. (Default = MAXITER). Not used for PFNT.
MAXLS: Maximum number of line searches allowed for each iteration. (Default = 4). Not used for PFNT.
LSTOL: Line Search tolerance. (0.01 < Real < 0.9; Default = 0.5). Not used for PFNT.
FSTRESS: Fraction of effective stress used to limit the sub increment size in material routines (0.0 < Real <
1.0; Default = 0.2) For non material enhanced elements only.
NLSTEP - ARCLN
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
GENERAL MAXITER MINITER MAXBIS CREEP
ARCLN TYPE DTINITFA MINALR MAXALR SCALEA NDESIRA NSMAXA
TYPE: Constraint type for Arc Length Method. (Character: CRIS, RIKS, or MRIKS; Default =CRIS)
CIRS is the Chrisfield method
RIKS is the Riks method
MRIKS is the modified Riks method (recommended)
DTINITFA: Initial time step defined as a fraction of the load step time (TOTTIME). (Default = .01)
MINALR, MAXALR: Minimum, maximum allowable arc-length adjustment ratio between increments. (Default
= 0.25, 4.0)
SCALEA: Scale factor (w) for controlling the loading contribution in the arc-length constraint. (Default = 0.0)
NDESIRA: Desired number of iterations for convergence to be used for the adaptive arc-length adjustment.
(Default = 4)
NSMAXA: Maximum number of increments. (Default = 1000). The job will stop if this limit is reached.
NLSTEP - ADAPT
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
ADAPT DTINITF DTMINF DTMAXF NDESIR SFACT INTOUT NSMAX
Main Index
164 Nonlinear Users Guide
Load Increment Size
DTINITF: Initial time step defined as fraction of total load step time (TOTTIM). (Real; Default 0.01). If
DTINITF>=DTMAXF then, DTINITF is reset to DTMAXF. If CTRLDEF is set to QLNEAR, the user should
set DTINITF equal to TOTTIM.
DTMINF: Minimum time step defined as fraction of total load step time (TOTTIM). (Real; Default 1e-5).
DTMAXF: Maximum time step defined as fraction of total load step time (TOTTIM). (Real; Default 0.5).
NDESIR: Desired number of iterations per increment. See Remark 2. (Integer; Default = 4)
SFACT: Factor for increasing time steps due to number of iterations. See Remark 4. (Real; Default = 1.2).
INTOUT: Output flag. Integer > -1. (Default = 0)
-1 - Only the last increment of the step will be output.
0 - Every computed load increment will be output.
> 0 - The output will be obtained at INTOUT equally spaced intervals. The time step will be temporarily
adjusted if necessary in order to reach these points in time.
NSMAX: Maximum number of increments in the current load case. (Integer; Default = 99999). The job will
stop if this limit is reached.
IDAMP: Flag for activating artificial damping for static analysis.
Enter 4 for time step control using damping.
Enter 5 for damping based time step control but no damping added.
Enter 6 for adding damping when the minimum time step has been reached. Recommended for robust and
stable analysis in a nonlinear analysis with advanced element.
Enter 0 for not adding damping. (Integer, Default = 0).
DAMP: Damping ratio. (Real; Default = 2.e-4).
CRITTID: ID of TABSCTL Bulk Data entry which defines the user criteria to use. See Remark 5. (Integer;
Default 0)
IPHYS: Flag to determine if automatic physical criteria should be added and how analysis should proceed if a
user criterion is not satisfied. (Integer; Default = 2)
2 Do not add automatic physical criteria; stop when any user criterion is not satisfied
-2 Do not add automatic physical criteria; continue when user criteria are not satisfied
1 Add automatic physical criteria; stop when any user criterion is not satisfied
-1 Add automatic physical criteria; continue when any user criterion is not satisfied
LIMTAR: Enter 0 to treat user criteria as limits, 1 to treat user criteria as targets. (Integer; Default = 0). Only
used if a user criterion is given through CRITTID. See Remark 5.
RSMALL: Smallest ratio between time step changes due to user criteria. (Real; Default = 0.1)
RBIG: Largest ratio between time step changes due to user criteria. (Real; Default = 10.0)
ADJUST: Time step skip factor for automatic time step adjustment. Only for dynamics. (Integer; Default = 0).
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 165
Load Increment Size
MSTEP: Number of steps to obtain the dominant period response. (10 < Integer < 200 or = -1; Default = 10).
RB: Define bounds for maintaining the same time step for the stepping function during the adaptive process.
(0.1 < Real <1.0; Default = 0.6).
UTOL: Defines tolerance on displacement. (.0001 < Real < 1.0; Default = 1.0)
NLSTEP - HEAT
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
HEAT CONVH EPSUH EPSPH EPSWH KMETHODH KSTEPH
MAXQNH MAXLSH LSTOLH
CONVH: Flags to select convergence criteria. (Character = U, P, W, V, N, A or any combination.
(Default UPW)
U: for displacement error control
P: for residual (load) error control
W: for energy (work) error control
V: for vector component norm used
N: for length norm used
A: for automatic switching of error measure
EPSUH: Error tolerance for temperature (U) criterion. (Real; Default =1.0E-2)
EPSPH: Error tolerance for heat flux (P) criterion. (Real; Default =1.0E-2)
EPSWH: Error tolerance for work (W) criterion. (Real; Default =1.0E-2)
KMETHODH: Method for controlling stiffness updates. (Character = PFNT, AUTO or ITER).
Default = AUTO.
KSTEPH: Number of iterations before the stiffness update for the ITER method. (Integer; Default = 1).
MAXQNH: Maximum number of quasi-Newton correction vectors to be saved on database. (Integer;
Default = MAXITER). Not used for PFNT.
MAXLSH: Maximum number of line searches allowed for each iteration. (Integer; Default = 4). Not used for
PFNT.
LSTOLH: Line Search tolerance. (0.01 < Real < 0.9; Default = 0.5)
NLSTEP - COUP
1 2 3 4 5 6 7 8 9 10
NLSTEP ID TOTTIME CTRLDEF
COUP HGENPLAS HGENFRIC
HGENPLAS: Conversion factor for heat generated due to plasticity. (Real; Default = 0.0)
HGENFRIC: Conversion factor for heat generated due to friction. (Real; Default = 0.0)
Main Index
166 Nonlinear Users Guide
Convergence Controls
Convergence Controls
Three methods are available for determining if convergence is obtained on any given iteration: residual force,
displacement, and strain energy. You can select one of these three criteria for convergence or you may specify a
combination of residual and displacement. The AND combination signals that both residual and displacement must be
met, while the OR combination specifies that either one can satisfy convergence criteria. If you are using residual there
may be cases in which the force residuals are null in which case is it necessary to switch over to displacement. An
Autoswitching option (on by default) allows for this switching.
The default measure for convergence in SOL 400 is residual which is based on the magnitude of the maximum residual
load compared to the maximum reaction force. This method is appropriate since the residuals measure the out-of-
equilibrium force, which should be minimized. This technique is also appropriate for Newton methods, where zero-
load iterations reduce the residual load. The method has the additional benefit that convergence can be satisfied without
iteration. You have complete control over how convergence is defined through the Iterations Parameters form in Patran
or through the options on the NLSTEP entry.
The basic procedures are outlined below.
1. RESIDUAL CHECKING
F r e s id u a l
-------------------------------- TOL 1 (4-41)
Fr e a c t i o n
F r e s id u a l Mr e s i d u a l
-------------------------------- TOL 1 and ---------------------------------- TOL 2 (4-42)
Fr e a c t i o n Mr e a c t i o n
where F is the force vector, and M is the moment vector. T OL 1 and T OL 2 are control tolerances. F
indicates the component of F with the highest absolute value. Residual checking has one drawback. In some
special problems, such as free thermal expansion, there are no reaction forces. If the value of the residuals and
reactions is less than 1.e-6, this test is ignored. If the AUTOSW flag on the NLSTRAT entry is ON, the program
automatically uses displacement checking in this cases.
2. DISPLACEMENT CHECKING
u
- TOL 1
--------------- (4-43)
u
u
- TOL 1 and ---------------
--------------- - TOL 2 (4-44)
u
where u is the displacement increment vector, u is the displacement iteration vector, is the incremental
rotation vector, and is the rotation iteration vector. With this method, convergence is satisfied if the
maximum displacement of the last iteration is small compared to the actual displacement change of the
increment. If the value of the incremental and iterative displacement is less than 1.e-8, this test is ignored. A
disadvantage of this approach is that it results in at least one iteration, regardless of the accuracy of the solution.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 167
Convergence Controls
k
j
j = 0
0 k+1
un + 1 un + 1
u
Figure 4-6 Displacement Control
Auto-Switch
In several types of analyses, maximum reactions or displacements are extremely small (even close to the round-off
errors of computers). In such circumstances, not all types of relative convergence criteria may work properly. For
example, in a problem with stress-free motion, the convergence check based on relative displacement increments
works correctly but not the convergence check based on relative residual or strain energy. In this situation, it is
necessary to check the convergence with absolute values of reactions or strain energy; otherwise, the analysis may
terminate prematurely. Similarly, this kind of situation may happen for problems with spring back and free thermal
expansion or constraint thermal expansion. The details for the cases where convergence checking with relative values
Main Index
168 Nonlinear Users Guide
Convergence Controls
may encounter difficulties are listed in the table below. The AUTO SWITCH option is designed to switch to the proper
convergence check scheme automatically if any of the situations mentioned above occur during the analysis. This
optional convergence check is activated by adding character A into the CONV field of the NLSTEP entry. This AUTO
SWITCH option allows automatic switching of the convergence check scheme to check as required on either residuals
or displacements if small reactions or displacements are detected, or to use the absolute strain energy checking if
necessary. If AUTO SWITCH is turned on, it:
1. Switches on the relative residual checking if the relative displacement criterion is used (which fails when the
maximum incremental displacement becomes very small Max._incremental_displacement/Smallest_element
size < 1.0e-8)
2. Switches on the relative displacement checking if the relative residual force (moment) criterion is used (which
fails when the maximum reaction force becomes very small < 1.0e-6)
3. To switch on the absolute energy checking if the structure is free of stress and deformation (strain energy
density < 1.0e-15).
Note that if both residual and incremental displacement criteria are already chosen (like UP), the AUTO SWITCH
feature will not be activated even if character A is specified. In this case, SOL 400 will ignore it.
.
Convergence Variable
Displacement/ Residual Strain Energy
Analysis Type Rotation Force/Torque
Stress-free motion Yes No No
Springback No Yes No
Free Thermal Expansion Yes No No
Constraint Thermal Expansion No Yes Yes
Yes relative tolerance testing works.
No relative tolerance testing does not work.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 169
Convergence Controls
2. The default tolerance for thermal analysis is 1% relative testing on displacements, residuals and energy (CONV
= UPW). The default iteration scheme followed is Modified Newton-Raphson with Automatic Stiffness Update
(KMETHOD = AUTO).
NLSTEP has the following limitations:
1. NLSTEP is not supported for the CREEP material option. It is only supported for elements using MATVP.
CREEP should still be specified through NLPARM with DT > 0.0.
2. NLSTEP with ARCLN does not support contact. This is also an existing limitation of the arc-length scheme
specified through NLPCI.
3. Quasi-static damping (IDAMP) specified on the ADAPT keyword of NLSTEP is only available for enhanced
elements. Automated element defaults automatically map all possible elements to the enhanced ones in
conjunction with IDAMP.
Recycling Criterion
The basic algorithm works with a user input desired number of recycles. The time step is increased for the upcoming
increment if the current increment uses fewer recycles than desired. If the number of recycles during an increment
exceeds the desired number, a time step cutback (bisection) is performed and the time step is reduced. If the increment
is converging the time step decrease is postponed until the next increment. Special care is taken for contact when
changes in contact occurs, like new contact, sliding or separation. A distinction is made between Newton-Raphson
iterations and Contact-Induced iterations only the former is used for controlling the time step changes. Without this,
the time step would often be reduced excessively.
In addition to this algorithm, there is also a scheme based upon artificial damping in a static analysis. Estimations of
strain energy changes are used for applying artificial damping for unstable situations (with sudden reductions of strain
energy). These estimates are also used for modifying the time step.
For adaptive stepping, user-defined criteria for controlling the time step are also available. The user-criteria ID is
specified through the CTITTID field on the ADAPT keyword of the NLSTEP entry and this in turn refers to a TABSCTL
bulk data entry where the actual criteria as specified. You can set a limit on the incremental displacements, rotations,
stresses, temperature etc. If a criterion would be violated a bisection and time step reduction is done. The user-criteria
thus work as limits. There is also an option to treat the criteria as targets, in which case the time step for the next
increment will be increased in order to reach the specific displacement increment for instance. The time step is never
increased during an increment.
The default recycle based criterion works as follows: You specify a desired number of recycles, NDESIR, default=4.
For problems with severe nonlinearities or for problems with very small convergence tolerances, it may be necessary
to increase this number. This number is used as a target value for the load stepping scheme. If the number of recycles
required in the current increment is less than the desired number, the load step for the next increment is increased. The
time step increase is based on a factor, S u , that you can also specify. Typical values for S u are in the range of 1.2 to
1.5. While the time step increase is obviously more aggressive with larger scale factors, it should be noted that there
may be excessive recycling and cutbacks if sudden nonlinearities are encountered. In order to avoid this, the following
logic is used for higher scale factors : If the actual number of recycles in an increment is greater than 60% of the desired
number of recycles (i.e., the current increment did not converge easily), the increased scale factor for the next
Main Index
170 Nonlinear Users Guide
Convergence Controls
increment is limited to 1.25 for scale factor values between 1.25 and 1.5625, and to 80% of the value for scale factors
above 1.5625.
T 2 N r m N rm + 1
s = ------s- (4-46)
Tm
where N rm is the maximum number of recycling related cutbacks for the increment and is calculated from
5
10 T s 10 5 T s
N rm = log 10 ------------------- N rm = log 10 ------------------- (4-47)
Tm Tm
T s is the time increment before any recycling related cutbacks occur for the increment and T m is the minimum
possible time step for the increment. T m is equal to the value set by you ( 10 5 by default) if there is no quasi-static
inertial damping and is equal to 10 3 times the value set by you ( 10 8 by default) if there is quasi-static inertial
damping. The scale-back factor for any cutback is the smaller of S N r ,1 S u . This scheme guarantees that no matter
what the starting time step for an increment, the minimum time step is reached in a reasonable number of cutbacks if
the increment consistently fails to converge.
With this feature a damping factor, F d , is introduced, which at the start of the loadcase, is set to 0. The time step for
the first increment is set equal to the user defined initial time step. During the assembly of the stiffness matrix K and
the right-hand side vector F , the contributions from damping are added to both sides of the equation system as K d a m p
and F d a m p , respectively.
With artificial damping option, the adaptive time stepping scheme is still used to control the time step size, however,
the adjustment will be made based on the damping energy of the system. For the first increment of the loadcase, the
calculation of F d and predicted energy is based on the estimated strain energy and damping energy for the loadcase.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 171
Convergence Controls
For the subsequent increments of the loadcase, F d and the time step size are modified according to the total strain
energy and estimated strain energy.
For a given step t , a solution is obtained and SOL 400 finds the largest values of stress change per stress, and
cr el
creep strain change per elastic strain, . It compares these values to the tolerance values, T s (stress change
tolerance) and T e (strain change tolerance), for this period. The value p is calculated as the larger of T and
cr el
T . If p 1 , the solution is continued. Upon the completion of the existing time stepping, the time
stepping will chosen for the next step as t new = t old , where is a factor calculated according to the criteria for
the creep analysis. The criteria are the tolerances you entered through the optional word CREEP of NLADAPT entry.
When you enter the tolerances and controls, the following conventions apply:
All stress and strain measures in tolerance checks are second invariants of the deviatoric state (that is,
equivalent von Mises uniaxial values).
You can reset all the tolerances and control upon the completion of one load step sequence.
Since the time increment is adjusted to satisfy the tolerances, it is impossible to predetermine the total number of time
increments for a given total creep time.
Exceptions
There are some exceptions to the basic scheme outlined above. If an increment is consistently converging with the
current load step and the number of recycles exceeds the desired number, the number of recycles is allowed to go
beyond the desired number until convergence is achieved or up to the user specified maximum number. The time step
is then decreased for the next increment by 1 S u . An increment is determined to be converging if the convergence
ratio was decreasing in three previous recycles.
Special rules also apply in a contact analysis. During the recycles, the contact status can keep changing (new nodes
come in contact, nodes slide to new segments, separate etc.). Whenever the contact status changes during an increment,
a new set of contact constraints are incorporated into the equilibrium equations and more recycles are necessary in
order to find equilibrium. These extra recycles, due to contact changes, are not counted when the recycle number is
checked against the desired number for determining if the load step needs to be decreased within the increment. Thus,
only true Newton-Raphson iterations are taken into account. For the load step of the next increment, the accumulated
Main Index
172 Nonlinear Users Guide
Convergence Controls
number of recycles during the previous increment is used. This ensures that the time step is not increased when there
are many changes in contact during the previous increment.
Defaults
The defaults of the NLSTEP option are carefully chosen to be adequate in a wide variety of applications. There are
cases, however, when the settings may need to be modified. Assume that the default settings are used, which means
that the recycle based control is active with an initial load of one per cent of the total. If the structure is weakly
nonlinear, convergence is obtained in just a few recycles and the time steps for successive increments get progressively
larger. This can lead to problems if the initially weakly nonlinear structure suddenly exhibits stronger nonlinearities;
for instance, occurrence of plasticity or parts coming into contact. Possible remedies to this problem include:
1. Decrease the time step scale factor to a smaller number so the step size does not grow so rapidly.
2. Use the maximum time step to limit large steps.
3. Decrease the desired and maximum number of recycles to decrease the load step if more recycles are needed.
Another situation is if the structure is highly nonlinear and convergence is slow. In this case, it may be necessary to
increase the desired number and maximum number of recycles. In general, there is a close connection between the
convergence tolerances used and the desired number and maximum number of recycles.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 173
Arc-Length Methods for Post-buckling Analysis
The ARCLN entry is applicable to MECH only and is ignored for creep analysis. The available constraint types are as
follows.
TYPE = CRIS:
T
i O i O 2 i O 2
U n U n U n U n + w 2 = l n
TYPE = RIKS:
T
i i 1 i O 2 i
U n U n U n U n + w = 0
TYPE = MRIKS:
T
i i 1 i 1 O 2 i i1 O
U n U n U n U n + w = 0
where:
The constraint equation has a disparity in the dimension by mixing the displacements with the load factor. The scaling
factor w is introduced as user input so that the user can make constraint equation unit-dependent by a proper scaling
of the load factor (). As the value of is increased, the constraint equation is gradually dominated by the load term. In
the limiting case of infinite, the arc-length method is degenerated to the conventional Newtons method
The MINALR and MAXALR fields are used to limit the adjustment of the arc-length from one increment to the next by:
Main Index
174 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
The arc length adjustment is based on the convergence rate (i.e., number of iterations required for convergence) and
the change in stiffness. For constant arc length during analysis, use:
MINALR = MAXALR = 1
The arc-length l for the variable arc-length strategy is adjusted based on the number of iterations that were required for
convergence in the previous increment I m ax and the number of iterations desired for convergence in the current
increment (NDESIRA) as follows:
R u = P u F u (4-48)
where F represents the internal forces, and the total external load P is expressed as
P u = P 0 + P (4-49)
where P 0 denotes the applied load at the end of the preceding subcase, P represents the load increment in the current
subcase, and is the load factor varying from 0 to 1 (but not to be limited to this range) within the subcase. Linearizing
R about u , equation (4-48) can be expressed as
R R
R u + u + R u + ------- u + ------- (4-50)
u
Based on equations (4-48) and (4-50), the equilibrium condition at u + u + dictates that
P F P
R u + ------- u ------- u + ------- (4-51)
u u
P F P
where ------- is the follower matrix, ------- is the stiffness matrix K , and ------- = P .
u u
P
K ------- u = R u + P (4-52)
u
where the follower matrix may be omitted. The iterative process can be established by decomposing equation (4-52)
into two parts:
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 175
Arc-Length Methods for Post-buckling Analysis
Ku R = R u and Ku P = P (4-53)
u = u R + uP (4-55)
1 T 1 1 2 2
u u + w = l (4-56)
T
2 1 2 1 1 2
l = u P u P + w (4-57)
if the initial value of is given. The constraint of equation (4-56) has a disparity in the dimension by mixing the
displacements with the load factor. For this reason, the scaling factor w is introduced so that you can scale to the
appropriate dimension or delete the term. The default value of w is zero. The iteration follows the path on the
plane normal to the initial tangent as shown in figure. Therefore, the subsequent iterations i 1 must satisfy
1 T i 2 1 i
u u + w * = 0 (4-58)
Recalling that the first iteration should result in
1 1 1
u = u P (4-59)
from which the load factors for the subsequent iterations are determined by
Main Index
176 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
T
1 i
u P R
i
= ---------------------------------------------------- (4-61)
T
2 1 i
w + u P u P
and
i i1 i
= + (4-62)
Notice that the normal plane does not change during the iteration by Riks' method. In addition, u p remains
constant if the iteration process is the modified Newton's method.
Alternatively, the normal plane may be updated at every iteration. If the normal plane is to be normal to the cumulative
incremental displacements for the preceding iterations as shown in figure, the orthogonality condition in equation
(4-58) should be modified to:
i1 0 T i 2 i1 0 i
u u u + w (4-63)
The increment in the load factor for i 1 is obtained by solving equation (4-63), i.e.,
i1 0 T i
i u u u R
= -------------------------------------------------------------------------------------------------- (4-64)
2 i1 0 i1 0 T i
w + u u u P
This variation of Riks' method has an advantage over Crisfield's method as it avoids the solution of a quadratic
equation.
Crisfield's Method
Instead of iterating on the normal plane, the solution is sought on the surface defined by equation (4-56) with an arc-
length of l as depicted, i.e.,
i 0 T i 0 2 i 0 2 2
u u u u + w = l (4-65)
This constraint can be interpreted as keeping the incremental displacement constant, if w = 0 . Substituting equation
(4-55) into the preceding equation, we obtain a quadratic equation in terms of :
i 2 i
a + b + c = 0 (4-66)
where
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 177
Arc-Length Methods for Post-buckling Analysis
T
i i 2
a = u P u P + w (4-67)
T
i i1 0 i 2 i1 0
b = 2 u P u u + u R + 2w (4-68)
T
i1 0 i i1 0 i 2 i1 0 2 2
c = u u + u R u u + u R + w l (4-69)
Since Crisfield's method leads to a quadratic equation, the selection of the proper root of this equation becomes the
most critical process for the success of this method. There are two roots to equation (4-66),
2
i b b 4ac
= ----------------------------------------- (4-70)
2a
i1 0 i 0
The root is chosen so that the angle between two vectors u u and u u is less than 90; i.e.,
i 0 T i1 0
u u u u
cos = ----------------------------------------------------------- (4-71)
2
l
1 i1 0 T i
= 1 + -------- u u u 0 (4-72)
2
l
There are cases where no roots can be found. Such is the case when the trial solution is far from the true
solution and stays outside the region covered by the arc length. In this case, the trial solution vector is scaled
so that the direction vector intersects with the surface defined by equation (4-65).
The wrong choice of the root could cause an unintentional loading path reversal, by which the solution returns
to the previous state. Such cases can be detected by checking the orthogonality of the incremental
displacements of the two successive solutions. If this case is detected, the root is chosen so that the angle
1 0 i 0
between u u and u u is an acute angle.
with
Main Index
178 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
1 1 1
= = --------------- (4-74)
NINC
Notice that NINC is a user-specified value for the number of increments in the subcase.
The arc length should be continuously updated at every increment using the information gathered during the preceding
increment. One method is to reduce the arc length if it requires an excessive number of iterations to attain a converged
solution, i.e.,
Id
l new = - l
---------- (4-75)
I m a x old
where I d is the desired number of iterations for convergence and I m a x is the number of iterations required for
convergence at the preceding step. It is also desirable to decrease the arc length if the stiffness ratio increases; i.e.,
1
l new = --------------- l old (4-76)
K r a t io
with
T
K new uj 1 uj 2 Fj Fj 1
K rati o = ABS ------------- = ABS -------------------------------------------------------------------------- (4-77)
K old T
uj uj 1 Fj 1 Fj 2
where the subscripts j , j 1 , j 2 denote the converged solution steps. In terms of the arc length, the computation
is simplified as
n l n 1
K rati o = ABS ----------------------------------------- (4-78)
n 1 l n 1
The adaptive process should be based on the arc length ratio, i.e.,
l n e w
ALRATIO = -------------- (4-79)
l ol d
Combining two criteria, the new arc length ratio may be adapted to the nonlinearity by
Id 1 Id 1
- and --------------- GT 1 then ALRATIO = MIN ----------- ---------------
If ---------- (4-80)
Im a x K r a t io I m a x K rati o
Id 1 Id 1
- and --------------- LT 1 then ALRATIO = MAX ----------- ---------------
If ---------- (4-81)
Im a x K r a t io I
m a x rati oK
else
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 179
Arc-Length Methods for Post-buckling Analysis
ALRATIO = 1 (4-82)
In order to maintain the stability for the adaptive process, ALRATIO should also be bounded, i.e.,
i
If the iteration produces i which exceeds the value of for output points, should be adjusted to the specified
value of for the output; i.e.,
* * * i1
= MIN (4-85)
and
* i * i
u = u R + u P (4-86)
*
where is the specified value for output (end of subcase or intermediate points). In the stiffening case this
limit may be exceeded at the first increment of the subcase. This limiting process may be implemented as
follows:
Process equation (4-85) at every iteration with
* 1
= --------------
NINC
- (4-87)
*
if INTOUT=ALL, otherwise = 1 . Then if
i * 0
u + u u l (4-88)
i * i *
= and u = u ; (4-89)
* * i i
otherwise, ignore and u and proceed to update u and .
The arc length should be updated upon convergence whenever this limiting process takes place. Notice that this
limiting process in effect reduces the arc length method to the conventional Newton's iteration.
In the unstable regime where the stiffness is negative, the load factor decreases with a forward step. When this
1
happens, the sign of should be reversed. This possibility should be examined at the beginning of each
increment. The sign can be determined by the sign of a dot product; i.e.,
Main Index
180 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
1 T 1
j = j j 1 if u j 1 u j 2 u P 0 (4-90)
An adaptive bisection algorithm is also incorporated to cope with divergent cases. If the iterative process using the arc
length method tends to diverge, the arc length is bisected. The bisection is combined in concert with the stiffness matrix
update strategy. The bisection procedure will continue until the iterative process is stabilized and a converged solution
is found. However, the number of contiguous bisections is limited by a user-specified parameter, MAXBIS in the
NLPARM bulk data entry. The variable arc length at every increment invokes the recovery from the bisection process
once the difficulties in convergence are overcome.
1
2
3
u
Figure 4-7 Riks Method
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 181
Arc-Length Methods for Post-buckling Analysis
u
(a) Arc Length in Terms of Combined Variables
u
(b) Arc Length in Terms of Displacements
The solution methods described above involve an iterative process to achieve equilibrium for a fixed increment of load.
None of them have the ability to deal with problems involving snap-through and snap-back behavior except the
NLSTEP (Case) method with artificial damping. An equilibrium path as shown in Figure 4-10 displays the features
possibly involved.
Main Index
182 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
2
6
F
3
Force
5
u
Displacements
Figure 4-10 Snap-through Behavior
The issue at hand is the existence of multiple displacement vectors, u , for a given applied force vector, F . This method
provides the means to ensure that the correct displacement vector is found. If you have a load controlled problem, the
solution tends to jump from point 2 to 6 whenever the load increment after 2 is applied. If you have a displacement
controlled problem, the solution tends to jump from 3 to 5 whenever the displacement increment after 3 is applied.
Note that these problems appear essentially in quasi-static analyses. In dynamic analyses, the inertia forces help
determine equilibrium in a snap-through problem.
Thus, in a quasi-static analysis sometimes it is impossible to find a converged solution for a particular load (or
displacement increment):
n + 1 F n F = F (4-91)
This is illustrated in Figure 4-10 where both the phenomenon of snap-through (going from point 2 to 3) and snap-back
(going from point 3 to 4) require a solution procedure which can handle these problems without going back along the
same equilibrium curve.
As shown in Figure 4-11, assume that the solution is known at point A for load level n F . For arriving at point B on
the equilibrium curve, you either reduce the step size or adapt the load level in the iteration process. To achieve this
end, the equilibrium equations are augmented with a constraint equation expressed typically as the norm of incremental
displacements. Hence, this allows the load level to change from iteration to iteration until equilibrium is found.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 183
Arc-Length Methods for Post-buckling Analysis
Path
B
Load (F)
th
rc leng
l: a
n F
A
Displacement (u)
The augmented equation, c u , describes the intersection of the equilibrium curve with an auxiliary surface g for
a particular size of the path parameter :
r u = F R u = 0
(4-92)
c u = g u = 0
Variations of the parameter moves the surface whose intersection with the equilibrium curve r generates a sequence
of points along the curve. The distance between two intersection points, denoted with 0 and , denoted by l is the
so-called arc length.
Linearization of equation (4-92) around point A in Figure 4-11 yields:
K P u r
T = r (4-93)
n n 0 0
where:
r r
K = ------ : P = ------ (4-94)
u
T c c
n = ------ : n 0 = ------ (4-95)
u
r = F R (4-96)
r 0 = g u (4-97)
It can be noted that a standard Newton-Raphson solution procedure is obtained if the constraint condition is not
imposed. The use of the constraint equation causes a loss of the banded system of equations which would have been
Main Index
184 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
obtained if only the K matrix was used. Instead of solving the N + 1 set of equations iteratively, the block elimination
process is applied.
i1
Consider the residual at iteration i to which the fraction of load level corresponds
i i1 i1 i i1
r = F R u (4-98)
i
The residual for some variation of load level, , becomes
i i1 i i i i1
r + = F + r (4-99)
which can be written as:
i i1 i i i1 i i
u + = u + u * (4-100)
i i1 i 1
where u = K r (4-101)
i i 1
and u * = K F (4-102)
i
Notice that u * does not depend on the load level. The equation above essentially establishes the influence of a change
i
in the load level during one iteration on the change in displacement increment for that iteration. After one iteration
is solved, this equation is used to determine the change in the load level such that the constraint is followed. There are
several arc length methods corresponding to different constraints.
Among them, the most well-known arc length method is one proposed by Crisfield, in which the iterative solution in
displacement space follows a spherical path centered around the beginning of the increment. This requirement is
translated in the formula:
2 i i
c = l = u u (4-103)
where l is the arc length. The above equation with the help of equations (4-102) and (4-103) is applied as:
i T i i 2 i1 i i1 T i i
u * u * + 2 u + u u * +
(4-104)
i1 i i1 T i1 i i1 2
u + u u + u l = 0
1
The equation above is interpreted with i = 1 and u = 0 in the prediction phase while retaining the full form of
equation (4-104) in the correction phase. Two solutions for are available. We choose the one that maintains a
positive angle of the displacement increment from one iteration to the next.
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 185
Arc-Length Methods for Post-buckling Analysis
i i
The two roots of this scalar equation are 1 and 2 . To avoid going back on the original load-deflection curve,
i1 i
the angle between the incremental displacement vectors, u and u (before and after the current iteration,
i i i i
respectively) should be positive. Two alternative values of u (namely, u 1 and u 2 corresponding to 1
i
and 2 are obtained and the cosine of two corresponding angles ( 1 and 2 ) are given by
i T i1
u n + 1 1 u n + 1
cos 1 = ----------------------------------------------------
- (4-105)
l
i T i1
u n + 1 2 u n + 1
and cos 2 = ----------------------------------------------------
- (4-106)
l
0
Once again, the prediction phase is interpreted with i = 1 and u n + 1 = u n , while equations (4-105) and (4-106)
retain their full form in the correction phase.
i i
As mentioned earlier, the appropriate root, 1 or 2 is that which gives a positive cos . In case both the
angles are positive, the appropriate root is the one closest to the linear solution given as:
i1 i i1 i 2
i u + u u + u l
= -------------------------------------------------------------------------------------- (4-107)
i1 i i
2 u + u u *
Crisfields solution procedure, generalized to an automatic load incrementation process, has been implemented in MSC
Nastran Implicit Nonlinear as one of the options using (NLPCI and NLSTRAT). Various components of this process are
shown in Figure 4-12.
1 F
Load (F)
F2 = 2 F
r1
B
Displacement (u)
Main Index
186 Nonlinear Users Guide
Arc-Length Methods for Post-buckling Analysis
The constraints in equations (4-103) and (4-104) are imposed at every iteration. Disadvantage of the quadratic equation
suggested by Crisfield is the introduction of an equation with two roots and thus the need for an extra equation to solve
1 1
the system for the calculated roots if two real roots exists. This situation arises when the contribution u (or u ) is
very large in comparison to the arc-length. This can be avoided in most cases by setting sufficiently small values of the
error tolerance on the residual force. In case the above situation still persists despite the reduction of error tolerance,
SOL 400 has two options to proceed:
1. To attempt to continue the analysis with the load increment used in the initial step of auto increment process.
2. Use the increment resulting from the linear constraint for the load.
P
K ------- u = R u + P (4-108)
u
where the follower matrix may be omitted. The iterative process can be established by decomposing equation (4-108)
into two parts:
Ku R = R u (4-109)
and
Ku P = P (4-110)
u = u R + u P (4-112)
The residual error, R in equation (4-109), is evaluated at every iteration and equation (4-110) is solved again by
Forward-Backward Substitution (FBS). However, equation (4-110) needs to be solved only once upon stiffness matrix
update in the absence of the BFGS update. Two different approaches for adopting BFGS updates can be contemplated.
The BFGS update may be applied to equations (4-109) and (4-110). Then, FBS has to be processed on both equations
at every iteration. It is also possible to adopt the BFGS update only on equation (4-110), i.e.,
Main Index
Chapter 4: Solution Strategies for Nonlinear Analysis 187
Arc-Length Methods for Post-buckling Analysis
K u R = R u (4-113)
where K implies that the BFGS update is incorporated into K . The BFGS does not have to be implemented in equation
(4-110), or u P does not have to be reevaluated unless a new stiffness matrix is formulated. The solution should not be
altered by introducing the BFGS, but the convergence will be expedited. This is because the effect of u P diminishes
as the solution converges ( u approaches zero).
It is necessary to modify the quasi-Newton vector y for the arc-length method. For a constant external applied load,
the quasi-Newton vector is defined. The applied load varies as a function of the load factor (p) at every iteration in the
arc length methods. The changes in p during the iteration can be accommodated in the quasi-Newton vector by
i1 i i
= R 1 NINC u R (4-114)
where NINC is the number of increments for the subcase.
Various strategies of the BFGS updates for the arc-length methods are implemented with a control by SYSTEM(145).
The MSC Nastran definition at the top of the input data may include
NastranSystem 145 = n$
Main Index
188 Nonlinear Users Guide
References
References
1. Zienkiewicz, O. C. and R. L. Taylor. The Finite Element Method (4th ed.) Vol. 1. Basic Formulation and Linear
Problems (1989),) Vol. 2. Solid and Fluid Mechanics, Dynamics, and Nonlinearity (1991) McGraw-Hill Book
Co., London, U. K.
2. Bathe, K. J. Finite Element Procedures, Prentice-Hall, Englewood Cliffs, NJ, 1995.
3. Hughes, T. J. R. The Finite Element MethodLinear Static and Dynamic Finite Element Analysis, Prentice-
Hall, Englewood Cliffs, NJ. 1987.
4. Ogden, R. W. Large Deformation Isotropic Elasticity: On The Correlation of Theory and Experiment for
Incompressible Rubberlike Solids, Proceedings of the Royal Society, Vol. A (326), pp. 565-584, 1972.
5. Cook, R. D., D. S. Malkus, and M. E. Plesha, Concepts and Applications of Finite Element Analysis (3rd ed.),
John Wiley & Sons, New York, NY, 1989.
6. Bathe, K. J. Finite Element Procedures, Prentice Hall, Englewood Cliffs, NJ, 1996.
7. Riks, E. An incremental approach to the solution of solution and buckling problems, Int. J. of Solids and
Structures, V. 15, 1979.
8. Riks, E. Some Computational Aspects of the Stability Analysis of Nonlinear Structures, Comp. Methods in
Appl. Mech. and Eng., 47, 1984.
9. Crisfield, M. A. A fast incremental iterative procedure that handles snapthrough, Comput. & Structures, V.
13, 1981.
10. Ramm, E. Strategies for tracing the nonlinear response near limit points, in K. J. Bathe et al (eds), Europe-
US Workshop on Nonlinear Finite Element Analysis in Structural Mechanics, Ruhr University Bochum,
Germany, Springer-Verlag, Berlin, pp/ 63-89. Berlin, 1985.
Main Index
Chapter 5: Computational Methods
Nonlinear Users Guide
5 Computational Methods
Nonlinear Computational Costs 190
Solution of Matrix Equations 194
Computer Hardware 206
Main Index
190 Nonlinear Users Guide
Nonlinear Computational Costs
Material Nonlinearity
Elastic-plastic make sure that the material has sufficient strain hardening to support the load. Note strain softening
materials often require additional iterations. Materials in the elastic plastic range require additional iterations when
temperature increases, and the yield stress is temperature dependent and decreases.
Hyperelastic material Often the strain/load levels reached in a simulation are greater than those used to evaluate
the material constants. This can result in an improper representation of the material which results in a large number of
iterations. Make sure the strains calculated are within the range of the experimental test data used for the curve fitting.
Large Deformation
Compressive loads may result in buckling which results in the need for additional increments and iterations. The
numerical problem may occur even before the structure buckles. The reason is that the initial stress is dominated by
negative values that reduces the numerical conditioning of the stiffness matrix. The best method to reduce the number
of iterations is to use NLSTEP and include damping in the system.
In models involving compression of rubber seals, while classical buckling does not occur, numerical problems may
still arise which results in the need for additional iterations. Because rubber and rubber like materials have small
resistance in shear and are nearly incompressible again the initial stress may dominate (in an unfavorable manner) the
conventional structural stiffness due to the material. The result is an instability; that often leads to penetration and/or
large number of iterations. The initial stress is automatically on when one includes Large Displacements, but one can
control the contribution using the MDLPRM,NLDIFF parameter.
In such models, it is recommended to use either:
Main Index
Chapter 5: Computational Methods 191
Nonlinear Computational Costs
Contact
Contact results in a large number of iterations either because MSC Nastran is not able to satisfy the geometric aspect
of contact given the increment of the applied load resulting in penetration, or because of separation of bodies which
may lead to both a geometric problem, and because of the need to redistribute the load over different paths.
The geometric problem may be both due to large step sizes and due to a contact tolerance distance that is inappropriate
for the particular model. The first problem can be addressed by taking either smaller load steps and/or using the
NLSTEP entry to control the step size.
By default, the user does not need to enter the contact tolerance distance, and the program calculates one based upon
the user geometry. In models that combine shells, beams, and the program, tolerance may be too small. You should
explicitly define ERROR on the BCONPRG option.
Main Index
192 Nonlinear Users Guide
Nonlinear Computational Costs
There are two modes that may be invoked which controls the implementation and the coordination between the
different groups of elements activated with the RUNOPT keyword, MULTIMST and MSTSLV. Indicating multi-master
and master-slave, respectively.
MSTSLV Specifies "master-slave" Parallel execution mode. Only the Master DMP process executes the full
MSC Nastran solution. This is optimal for single-host DMP execution.
MULTIMST Specifies "multiple-master" Parallel execution mode. All DMP processes execute the full MSC
Nastran solution. This is optimal for multiple-host DMP execution.
MSC Nastran detects single- and multiple-host DMP execution and sets the default run option accordingly.
Note that the VCCT bulk data entry for fracture mechanics is not available with parallel element assembly. Also, note
that DMP and SMP cannot be combined, and if a user selects both DMP and SMP, then the DMP parallel assembly
will be used.
Note that the recommended approach as of the MSC Nastran 2017.0 release is to use SMP as it is simpler and leads to
better performance.
Main Index
Chapter 5: Computational Methods 193
Nonlinear Computational Costs
Main Index
194 Nonlinear Users Guide
Solution of Matrix Equations
Numerically, we never invert the stiffness matrix, but either use techniques that fall under the classes of direct solution
using a variation of Gaussian elimination or iterative schemes.
The stiffness (or tangent stiffness) matrix may be either symmetric or nonsymmetric, and MSC Nastran has the
following procedures that may be used.
Non-symmetric,
Symmetric Symmetric Non- Non-symmetric Non-positive
Positive Definite Positive Definite Positive Definite Definite
Real Complex Real Complex Real Complex Real Complex
Direct Serial Yes Yes Yes Yes Yes Yes Yes Yes
Direct SMP Yes Yes Yes Yes Yes Yes Yes Yes
Direct DMP Yes Yes Yes Yes Yes Yes Yes Yes
Direct DMP / Yes Yes Yes Yes Yes Yes Yes Yes
SMP
Element Iterative Yes No No No No No No No
Serial
Matrix Serial Yes Yes No No No No No No
Iterative
Matrix DMP Yes Yes No No No No No No
Iterative
Singularity Ratio
The singularity ratio, R , is a measure of the conditioning of the system of linear equations. R is related to the
conditioning number, C , which is defined as the ratio between the highest and lowest eigenvalues in the system. The
singularity ratio is an upper bound for the inverse of the matrix conditioning number.
1RC (5-1)
C and R establish the growth of errors in the solution process. If the errors on the right-hand side of the equation are
less than E prior to the solution, the errors in the solution will be less than , with
Main Index
Chapter 5: Computational Methods 195
Solution of Matrix Equations
CE (5-2)
The singularity ratio is a measure that is computed during the Crout elimination process of MSC Nastran Implicit
Nonlinear using the direct solver. In this process, a recursive algorithm redefines the diagonal terms
k1
k k 1
K kk = K kk K mk K m k 1 ik1 (5-3)
m = i
where i is a function of the matrix profile. K kk is a diagonal of the kth degree of freedom. The singularity ratio is
defined as
k k 1
R = min K kk K kk (5-4)
k k 1
If all K kk and K kk are positive, the singularity ratio indicates loss of accuracy during the Crout elimination
process. This loss of accuracy occurs for all positive definite matrices. The number of digits lost during the elimination
process is approximately equal to
The singularity ratio also indicates the presence of rigid body modes in the structure. In that case, the elimination
k
process produces zeros on the diagonal K kk 0 . Exact zeros never appear because of numerical error; therefore, the
singularity ratio is of the order
n
R = O 10
d i g i t
(5-6)
where n digit is the accuracy of floating-point numbers used in the calculation. For most versions of SOL 400,
k
n digit 12 . If rigid body modes are present, K kk is very small or negative. If either a zero or a negative diagonal is
encountered, execution of SOL 400 is terminated because the matrix is diagnosed as being singular.
You can force the solution of a nonpositive definite or singular matrix. In this case, SOL 400 does not stop when it
k
encounters a negative or small term K kk on the diagonal. If you use Lagrangian multiplier elements, the matrix
k
becomes non-positive definite and MSC Nastran Implicit Nonlinear automatically disables the test on the sign of K kk .
However, it still tests for singular behavior.
MSC Nastran SOL 400 also supports the PARAM,AUTOSPC, in which case the rigid body mechanism is suppressed
by putting a large number on the diagonal. The value of EPS on the AUTOSPC case control is set to 1.E-8.
Note: The correctness of a solution obtained for a linearized set of equations in a nonpositive definite system is
not guaranteed.
Main Index
196 Nonlinear Users Guide
Solution of Matrix Equations
Matrix Conditioning
Matrix conditioning for linear systems such as those encountered in SOL 101 is governed by a few phenomenom which
can be controlled by the user.
1. Rigid body modes the system is not fully constrained. In the case of a structure such as a building, this is a
user error, and one should use SPC to insure that there are no rigid body models. In the case of a plane or rocket,
this is by design, so the inertia relief (INREL parameter) should be used. If you are not sure if one has
constrained the rigid body modes, one can use the AUTOSPC case control option.
2. Poorly conditioned matrices because continuum elements are used with very high aspect ratios. Typically
aspect ratio > 100 cause problems in structural analysis. This is less of a problem in heat transfer. In such case,
either improve the mesh or switch to beam or shell elements.
3. Shell or beam element that are very thin. The problem is due to the mismatch of the (high) membrane and (low)
bending stiffness. This is more likely to occur for shell models, in which case it may be beneficial to switch to
membrane elements.
4. Flat shells, because of the formulation in the classic CQUAD4 and CTRIA3 shell elements if they are flat a
singularity occurs associated with the normal to the surface. The K6ROT parameter may be used to eliminate
this problem.
5. Disparity in material moduli, for example if the Youngs modulus of one material is 1.e7 and another material
is 1.e-1. Be sure one does not have data errors in model.
6. Poissons ratio =0.5 for conventional solid elements, either set Poissons ratio to 0.49 and/or use advanced
elements.
7. Orthotropic materials where material data is close or beyond the stability limits.
8. Conflicting constraints, implying that degrees of freedom are treated as retained after previously defining them
as constrained can lead to poor conditioning. Occasionally this can be manually corrected, but the
recommended procedure is to use the AUTOMSET,YES parameter. It should be noted that this process has a cost
associated with it, but it is strongly recommended especially when using contact using the Node-to-Segment
perspective and the LMT2MPC=1 on the MDLPRM bulk data entry.
For Nonlinear analysis, additional sources of poor matrix conditioning may include:
9. Use of rigid elements (RBAR, RBAR1, RJOINT, RROD, RTRPLT, RTRPLT1, RBE1, RBE2, and RBE3) with
Lagrange multipliers may lead to poor numerical conditioning.
The RIGID control case command selects the type of rigid element. It has the following format:
LINEAR
RIGID = LAGRAN
LGELIM
There are several strategies:
LINEAR will select the linear rigid elements, LAGRAN will select the Lagrange rigid element with the Lagrange
multiplier method, and LGELIM will select the Lagrange element with the Lagrange elimination method.
Main Index
Chapter 5: Computational Methods 197
Solution of Matrix Equations
If RIGID control case command does not exist in the user case control file for SOL 400, the selection
RIGID=LAGRAN is assumed.
If only small rotations occur, and all the nonlinearity is due to material behavior then RIGID=LINEAR may
be used. Else, one can use RIGID=LGELIM.
10. When materials exist in the model that exhibit strain softening, the conditioning number will be adversely
effected. This may occur due to plasticity or rate effects.
11. When structural buckling occurs. This may be global buckling such as in classical Euler beam or column
buckling or the wrinkling of aerospace panels. The user can use NLSTEP to obtain a solution. The alternative
is to incorporate dynamics into the model.
12. Contact, especially when a body separates and hence now has multiple rigid body modes.
13. Contact, when using Node-to-Segment the default implementation is to use Lagrange multipliers. Lagrange
multipliers result in ill conditioned systems because theoretically they result in an exact zero on the diagonal.
This issue can be resolved by using the option LMT2MPC=2 on the MDLPRM bulk date entry. This is not
necessary when activating the Pardiso solver through the SPARSESOLVER executive control statement. This
technique cannot be used in a heat transfer or a coupled thermal-mechanical analysis.
14. The numerical issues associated with Lagrange multipliers can be mitigated using the PARAM, LMFACT. The
default value of LMFACT is given in the output. If the conditioning number is poor, it is suggested that the user
enter the printed value divided by 100.
15. Similar to LMFACT, one may want to reduce the value of the penalty factor by using the PARAM,PENFN.
Hence, for nonlinear models, it is very useful to examine the conditioning number during the incremental process.
The SPARSESOLVER executive control statement may be used to control the amount of information regarding the
numerical pivoting.
Numerical Stabilization
MSC Nastran provides linear springs for the modeling of support conditions. These springs are activated via the
following two parameters and can help stabilize the model and allow user's to get solutions in cases that would
otherwise not converge. These parameters should be used with care as they may add non-existent energy into the model
that may significantly affect the solution.
AUTOSPC causes an SPC to be applied automatically to components of motion with zero or very small stiffness.
KDIAG
Default = -1.0 (SOL 400 with non-contact analysis), or 0.0 (SOL 400 with contact analysis)
In SOL 400 (nonlinear static and transient analysis), KDIAG may be used to eliminate spurious mechanisms and
singularities in the nonlinear stiffness matrix. The absolute value of KDIAG will be added to some or all of the diagonal
terms in the nonlinear stiffness matrix as follows:
< 0.0 For SOL 400, the absolute value of KDIAG is added to the diagonal term of null columns only.
Main Index
198 Nonlinear Users Guide
Solution of Matrix Equations
Direct Solver
The direct solvers are robust for all solutions, though they may need more resources for large solid models. If poor
numerical conditioning is present (due to very thin shells, mixture of materials or buckling), then there is slight but
negligible increase in computational costs due to the need to perform more pivoting.
The MSC Nastran direct solvers (MSCLDL and MSCLU) can also benefit from parallelization and GPGPU
technology. The MSC solvers provides solutions for Distributed Memory, Shared Memory, and combinations. Because
of the continuous evolution of hardware and variable model size, it is difficult to predict the improvement in the
performance.
The Intel MKL Pardiso solver was introduced in the MSC Nastran 2014.0 release. It shows greatly improved
scalability with SMP if there is enough memory available. The solver is activated using FACTMETH=PRDLDL on
the SPARSESOLVER executive control statement.
The SPARSESOLVER executive control statement can also be used to select other solver methods as well as other
options. However, it is recommended for direct solvers to use one of the MSC Nastran direct solvers or Pardiso.
Main Index
Chapter 5: Computational Methods 199
Solution of Matrix Equations
The DMP implementation is done using Message Passing Interface (MPI), and while the default MPI is acceptable,
you do have the freedom to choose a particular MPI.
The DMP direct solver only scales to DMP=2 and typically at DMP=4 and beyond the performance suffers.
Today the recommendation is to use SMP only, and to use Pardiso if there is enough memory available to have Pardiso
run in-core. The DMP implementation will continue to be supported and is the best available option when using SOL
400 with the MATDIGI bulk data entry.
Memory Usage
The user can specify the amount of memory used for the job, by specifying the memory on the command line. The
default is MEM=MAX, which indicates that the memory is equal to the maximum amount of memory allowed. This
amount of memory set to by memorymaximum keyword. The default value of memorymaximum is 50% of the
physical memory on the machine.
When DMP is used for parallel simulation, then the amount of memory used for each process is set to memorymaxium
divided by the number of processes desired.
For example, for DMP=2, the user is given one quarter of the available memory for each DMP process. The amounts
given to User Open Core, Executive System, Master, Scratch Memory (smem), and buffer Pool (bpool) are problem
dependent.
For SOL 400 using MSCLDL, the standard division of memory to the various systems is designed to provide as much
of memory to bpool to reduce the I/O costs of element matrix computations. For SOL 400 using Pardiso, the division
of memory is designed to ensure that Pardiso is running in-core. While Pardiso solver can run out-of-core, it is
recommended to use MSCLDL if insufficient memory is available to use Pardiso in-core.
For this reason, it is often necessary to use more than half of the memory system when using Pardiso solver. This can
be achieved by setting memorymaximum=100gb for a machine with 128 Gb of memory. In the following, we
provide rough details for a 4M degrees of freedom automobile body and the memory set by mem=max with Pardiso
followed by the amount to keep the job completely in-core.
mem=max, smp=2,mode=i8
The amount required by Pardiso to be running in-core was 12.45 gb implying User OpenCore had enough memory for
in-core performance.
In some instances, it may be necessary to use more than half of the memory of the system to keep Pardiso in-core, and
in this case, one may use mem=max along with memorymaximum=80gb for the example described above.
Main Index
200 Nonlinear Users Guide
Solution of Matrix Equations
Iterative Methods
Iterative solvers are a viable alternative for the solution of large systems. There are two types of iterative solver,
element based and matrix based. The SMETHOD case control command specifies which type, if any, is to be used and
further control is specified using the ITER bulk data entry. These iterative methods are based on preconditioned
conjugate gradient methods. The single biggest advantage of these iterative methods is that they allow the solution of
very large systems at a reduced computational cost. This is true regardless of the hardware configuration. The
disadvantage of these methods is that the solution time is dependent not only upon the size of the problem, but also the
numerical conditioning of the system. A poorly conditioned system leads to slow convergence hence, increased
computation costs.
When discussing iterative solvers, two related concepts are introduced: fractal dimension and conditioning number.
Both are mathematical concepts, although the fractal dimension is a simpler physical concept. Fractal dimension, the
range of which is between 1 and 3, is a measure of the chunkiness of the system. For instance, a beam has a fractal
dimension of 1, while a cube has a fractal dimension of 3. In general, iterative methods are better when the fractal
dimension of the problem is > 2.5; i.e., dominated mostly by 3-D elements.
The conditioning number is related to the ratio of the lowest to the highest eigenvalues of the system. This number is
also related to the singularity ratio, which is reported in SOL 400 output when using a direct solution procedure. In
problems involving beams or shells, the conditioning number is typically small, because of large differences between
the membrane and bending stiffnesses.
There exist certain types of analyses for which the iterative solver is not appropriate. These types include:
Elastic analysis
Explicit creep analysis
Eigenvalue analysis
Elastic or explicit creep analysis involves repeated solutions using different load vectors. When a direct solver is used,
this is performed very efficiently using back substitution. However, when an iterative solver is used, the stiffness
matrix is never decomposed, and the solution associated with a new load vector requires a complete re-solution.
The iterative solvers are activated using the SMETHOD case control command. One can activate the element based
(CASI) by using the SMETHOD=ELEMENT or the matrix based iterative solver, using SMETHOD=MATRIX. The
element based solver is recommended. Both iterative solvers result in a significant decrease in the memory
requirements, disk I/O and scratch disk space.
An iterative solver has additional parameters that govern the accuracy, and hence, convergence of the linear solution
phase. Recall, that the Newton-Rhapson controls still govern the accuracy of the total structural calculations.
Furthermore, the iterative solvers implemented are not designed to solve poorly conditioned problems or non-positive
definite models. For further details, see the section on Matrix Conditioning.
Iterative solvers use a preconditioner to improve the conditioning which reduces the number of iterations. This is set
on the ITER bulk data option.
For the matrix-based iterative solver, there are many preconditions that one can use where the default is the Block
Incomplete Cholesky (BIC). For well-conditioned systems, the Jacobi method may be faster.
Main Index
Chapter 5: Computational Methods 201
Solution of Matrix Equations
For the Element based iterative solver there is only one choice:
ITER, PRECOND,CASI
The element based iterative solver has the following restrictions.
No GENEL elements
DMIG matrices must be small
Reduction to superelements is not possible.
No eigenvalue analysis
No Inertia Relief
No Scalar points, use conventional grid points
No heat transfer or thermal-mechanically coupled analysis
AUTOSPC is not supported.
For a positive-definite system, the maximum number of iterations is equal to the number of active degrees of freedom.
For a well-conditioned system, the number of iterations is less than the square root of the number of of equations.
If the system is non-positive definite, a fatal error will occur in the NLSOLV module. This can occur for multiple
reasons. While switching to the direct solver will obtain a numerical solution, extra care is required that a meaningful
engineering solution is obtained.
Note that neither iterative solver can utilize GPGPU hardware accelerators.
The element based iterative solver automatically takes advantage of SMP if it is specified on the command line.
The matrix based iterative solver can be used in conjunction with DMP; see MDLPRM, DMPITER for more details.
Main Index
202 Nonlinear Users Guide
Solution of Matrix Equations
Ku = F (5-7)
And a nonlinear system is expressed as:
T
K u = F R = r (5-8)
where K is the elastic stiffness matrix, K T is the tangent stiffness matrix in a nonlinear system, u is the displacement
vector, F is the applied load vector, and r is the residual.
The function decreases most rapidly in the direction of the negative gradient.
u = F Ku = r (5-11)
One method to solve both linear and nonlinear problems is to use iterations. The objective of the iterative techniques
is to minimize the function, , without decomposing the stiffness matrix. In the simplest methods,
uk + 1 = uk + k rk (5-12)
where
T T
k = r k r k r k Kr k (5-13)
The problem is that the gradient directions are too close, which results in poor convergence.
An improved method led to the conjugate gradient method, in which
uk + 1 = uk + k Pk (5-14)
T T
k = P k r k 1 P k KP k (5-15)
The trick is to choose P k to be K conjugate to P 1 P 2 P k 1 . Hence, the name conjugate gradient methods. Note
the elegance of these methods is that the solution may be obtained through a series of matrix multiplications and the
stiffness matrix never needs to be decomposed.
Main Index
Chapter 5: Computational Methods 203
Solution of Matrix Equations
Certain problems which are ill-conditioned can lead to poor convergence. The introduction of a preconditioner has
been shown to improve convergence. The next key step is to choose an appropriate preconditioner which is both
effective as well as computationally efficient. The easiest is to use the diagonal of the stiffness matrix. The incomplete
Cholesky method has been shown to be very effective in reducing the number of required iterations.
Perturbation
The perturbation methods that require eigenvalue analysis cannot be performed when using the matrix based iterative
solver. However, the element based iterative solver can be used with any of the ANALYSIS case control commands.
MSC Nastran has multiple methods to perform the eigenvalue analysis required for either modal dynamics or buckling.
The METHOD case control is used to indicate that an eigenvalue analysis will be performed.
where
where
Main Index
204 Nonlinear Users Guide
Solution of Matrix Equations
Buckling
The linear buckling simulation is defined as obtaining the eigenvalues to the problem
K i *K d i = 0
cr
pi = i *Pa
where
Pa is the prescribed load to obtain the stress used in the differential stiffness matrix
cr
pi is the critical load
K T i *K i d i = 0
cr
pi = Pn + i * Pn + 1 Pn
where
K id is the incremental differential stiffness matrix due to a change in stress due to the change in Load
from P n to P n + 1
Main Index
Chapter 5: Computational Methods 205
Solution of Matrix Equations
Note if the K or K T matrix is non symmetric, then complex eigenvalues will be obtained.
Only elements which have a differential stiffness matrix will contribute to the buckling simulation.
This includes CONROD, CROD, CTUBE, CBAR, CBEAM, CBEAM3, CBEND, CQUAD4, CQUADR, CQUAD8,
CTRIA3, CTRIAR, CTRIA6, CHEXA, CPENTA, and CTETRA.
One should note that the interface elements used for cohesive analysis do not have a differential stiffness matrix but
may be included in a buckling simulation.
Main Index
206 Nonlinear Users Guide
Computer Hardware
Computer Hardware
Memory Systems
MSC Nastran is available on a variety of hardware and operating systems, including Windows 7, Windows 10, and
various Linux flavors. MSC Nastran is only available for what has historically been referred to as i8 mode. The MSC
Nastran 2017 release is the first release where this is the case. The i8 mode effecively gives the user access to all of he
hardware memory on the system.
Note that allocating all of the hardware memory to your job is highly discouraged. This will result in memory conflict
with the operating system and would adversely affect other users jobs. If you are running on a standalone machine,
than memorymaxium can be set to 80-90% of the real memory on the computer.
Buffer Pool
When MSC Nastran performs I/O, it buffers the process through a region of memory. This is called the buffer pool or
BPOOL. The purpose of this is to reduce the amount of physical disk I/O during the calculations. In todays computing
environment where CPU speeds have improved at a greater rate than disk speeds and memory costs have been reduced,
this is highly beneficial.
For most solution sequence, the buffer pool is 25% of the memory. When SOL 101 or SOL 400 is used, the estimate
program is used to determine BPOOL size. The remaining memory would be used for storing the engineering data, the
stiffness matrix, and scratch space for the decomposition.
In general, the default values chosen by MSC are adequate, but as an alternative, the user can set size of the buffer pool
using the BPOOL command on the command line of the BUFFPOOL system cell.
Specify BPOOL in one of three ways:
a. bpool=N specifies number of GINO blocks for buffer pooling
b. bpool=size specifies a memory size for buffer pooling
c. bpool=Nx specifies a percentage of MEM to be used for buffer pooling
After a similar job has been run, the Run Statistics in the .f04 file can serve as a guideline for setting the BPOOL in
subsequent runs.
Main Index
Chapter 5: Computational Methods 207
Computer Hardware
Resource Recommendations
It is difficult to know exactly the amount of resources required for a simulation, the requirements are based upon the
size of the model (number or elements and grids), the element type, the connection of the stiffness matrix, the number
of modes, etc. Neglecting the number of modes and focusing on the size of the model, the recommended memory is:
Up to 1 million degrees of freedom 2 GB
Up to 5 million degrees of freedom 8 GB
Up to 10 million degrees of freedom 16 GB
Up to 20 million degrees of freedom 64 GB
BUFFSIZE
BUFFSIZE is the length of an MSC Nastran buffer in terms of 64-bit words. It controls the physical record size for data
storage/transfer that is contained in many MSC Nastran logical units. The default buffer size is 8193. The
recommended buffer size for large models is 65537 which is the largest allowed.
MPI Selection
The types of MPI available are:
On platforms which support more than one MPI implementation, a supported MPI implementation may be selected
using the mpiimplementation keyword. This keyword may be abbreviated as mpiimp.
For example, to select hpmpi on linux64,
mpiimp=hpmpi
should be set on the command line or in the rc file.
Main Index
208 Nonlinear Users Guide
Computer Hardware
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis
Nonlinear Users Guide
Main Index
210 Nonlinear Users Guide
Introduction
Introduction
This chapter describes the practical steps involved in setting up and running MSC Nastran SOL 400 jobs, including
the use of Patran. The Patran UI is set up to guide you through the process of setting up the SOL 400 analysis, including
the job information for the executive section (select the solution type and solution parameters), the subcases and steps
(create subcases and steps and use subcase parameters to specify the required step/subcase control information), the
job submission (use Analyse - Entire Model - Full Run), monitoring the job while it is running and once it is done (user
Monitor - Job - View sts to monitor and Monitor - Job - View f06 to debug). It covers the Patran user interface (UI) and
its capabilities for setting up, submitting, and monitoring the job. It shows the user how to specify the analysis type,
but does not go into any detail about the analysis types of the individual steps being set up. See Nonlinearity and
Analysis Types for the details of analysis types.
In addition to describing the job set-up and execution, methods for monitoring and de-bugging the job are described.
The middle of the chapter describes the methods available for monitoring the job while it is running. Also, the Patran
capabilities for monitoring and viewing the various MSC Nastran analysis files such as .sts, .f06, etc. are covered.
The last part of the chapter describes the recommended approach for debugging a job that has failed. This includes
guidance for interpreting the error messages that are produced from simple formatting errors as well as resolving
convergence problems that may occur if the analysis ever gets to a point where it cannot arrive at an equilibrium
condition and terminates.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 211
Overview of Running a SOL 400 Job
MSC Nastran is executed with a command called nastran. (Your system manager may assign a different name to the
command.) The nastran command permits the specification of keywords used to request options affecting MSC
Nastran job execution. The format of the nastran command is:
nastran input_data_file [keyword1 = value1 keyword2 = value2 ...]
where input_data_file is the name of the file containing the input data and keywordi=valuei is one or more
optional keyword assignment arguments. For example, to run an a job using the data file example1.dat, enter the
following command:
nastran example1
See The nastran Command in the QRG .
The details of submitting a MSC Nastran job are specific to your computer system contact your computer system
personnel or the MSC Nastran Installation Guide for further information.
Main Index
212 Nonlinear Users Guide
Overview of Running a SOL 400 Job
When the Action is set to Analyze, the Method is set to Full Run or Analysis Deck, and the Apply button is selected from
the Analysis form, a jobname.bdf file is created and a MSC Nastran job is launched.
When the analysis is successfully completed, one or more output file is produced. These output files can be directly
imported or attached into the Patran database for postprocessing by setting the Action menu to Access Results.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 213
Overview of Running a SOL 400 Job
Main Index
214 Nonlinear Users Guide
Setting Up a SOL 400 Job
Solution Type
MSC Nastran SOL 400 can simulate many different types of structural and thermal responses. In general, a structural
analysis can be either static or dynamic. Both static and dynamic analysis may simulate linear response or nonlinear
response. SOL 400 incorporates the formulations and functionality to simulate nonlinear static and dynamic structural
responses.
Entry Description
ANALYSIS Specifies the analysis procedure to be used for the step or subcase being set up.
Specifying Subcases
Each subcase is designated with the following case control command.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 215
Setting Up a SOL 400 Job
Entry Description
SUBCASE Delimits and identifies a subcase.
STEP Delimits and identifies a section of a subcase, typically delineating one load step, or
perturbation step in the analysis (see Chapter 4: Solution Strategies for Nonlinear Analysis,
on multi-stepping for a more detailed explanation of how STEPs and SUBCASEs work
together).
Main Index
216 Nonlinear Users Guide
Analysis Procedures
Analysis Procedures
The STEP is the SOL 400 mechanism for associating loads and boundary conditions, output requests, and various other
parameters to be used during part of a complete run. Each step can be designated with one of the analysis types listed
below. For each Analysis Type, you will define the Solution Parameters and Output Requests; these collectively
constitute the Analysis Procedures.
In MSC Nastran, case control commands provide the loads and constraints, and load incrementation method, and
controls the program after the initial elastic analysis. Case control commands also include blocks which allow changes
in the initial model specifications. Case control commands can also specify print-out and postprocessing options. Each
set of load sets must begin with a SUBCASE/STEP command and be terminated by another SUBCASE/STEP or a
BEGIN BULK command. If there is only one load case, the SUBCASE/STEP entry is not required. The SUBCASE
option requests that the program perform another increment or series of increments. The input format for these options
is described in QRG.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 217
Analysis Types
Analysis Types
Analysis Types for steps/subcases in SOL 400 include the following:
Type Description
Linear Static Static stress analysis is used when inertia effects can be neglected. During a linear static step,
the models response is defined by the linear elastic stiffness at the base state, the state of
deformation and stress at the beginning of the step.
Nonlinear Static Nonlinear static analysis requires the solution of nonlinear equilibrium equations, for which
SOL 400 uses Newtons method. Many problems involve history dependent response, so that
the solution is usually obtained as a series of increments, with iteration within each increment
to obtain equilibrium. For most cases, the automatic incrementation provided by SOL 400 is
preferred, although direct user control is also provided for those cases where you have
experience with a particular problem. This includes creep, viscoelastic and body approach
Normal Modes This solution type uses eigenvalue techniques to extract the frequencies of the current system,
Modal Transient The Duhamel Integral method integrates all of the equations of motion through time. The
Dynamic accuracy is based upon the number of modes extracted.
Nonlinear This solution type is used when nonlinear dynamic response is being studied. For most cases,
Transient Dynamic the automatic incrementation provided is preferred, although direct user control is also
provided for those cases where you have experience with a particular problem. The
Generalized-alpha method has been presented as an unconditionally stable (for linear systems),
second-order algorithm that allows user-controllable numerical dissipation.
Direct Frequency Frequency response analysis is a method used to compute structural response to steady-state
Response oscillatory (such as rotating machinery) excitation. In frequency response analysis the
excitation is explicitly defined in the frequency domain. The direct method solves the coupled
equations of motion in terms of forcing frequency
Modal Frequency The modal method utilizes the mode shapes of the structure to reduce and uncouple the
Response equations of motion. The solution for a particular forcing frequency is obtained through the
summation of the individual modal responses.
Direct Complex Complex eigenvalue analysis is used to compute the damped modes of structures and assess the
Eigenvalue stability of systems modeled with transfer functions. Direct Complex eigenvalue analysis
solves the coupled equations of motion.
Modal Complex The modal method utilizes the mode shapes of the structure to reduce and uncouple the
Eigenvalue equations of motion.
Main Index
218 Nonlinear Users Guide
Analysis Types
Steady State Heat This solution type is for heat transfer problems where the temperature field can be found for the
Transfer current contact and deformation states of the bodies being studied. (e.g., it is assumed that there
are no changes in the structure during the heat transfer analysis). For cases where there are
mechanical changes in the structure a coupled thermal-structural solution must be performed.
Transient Heat This solution type is for transient heat transfer problems where the temperature field can be
Transfer found based on the current state of stress and deformation of the bodies being studied (e.g., it
is assumed that there are no changes in the structure during the heat transfer analysis). For cases
where there are mechanical changes in the structure a coupled thermal-structural solution must
be performed. For all transient heat transfer cases, the time increments may be specified
directly, or will be selected automatically based on a user prescribed maximum nodal
temperature change in a step. Automatic time incrementation is generally recommended.
A complete description of the available analysis types is provided in Chapter 3: Nonlinearity and Analysis Types.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 219
Setting Up the Analysis in Patran
Main Index
220 Nonlinear Users Guide
Setting Up the Analysis in Patran
6. Under the Subcase Select... menu, select the steps you have set up in the correct order and deselect the Default
step/subcase unless you used it.
Note: There are two ways to set up SOL 400 analysis jobs in Patran. The first is to select Implicit Nonlinear
as the Solution Type. This option gives you access to most of the various analysis capabilities and
numerical controls available through SOL 400. The second option uses the same menus as if you
were setting up a non-SOL 400 analysis, such as a SOL 106 or 129, but simply changes the executive
command line. To use this second option, simply set up your analysis the same way you would a
non-SOL 400 job and click on the SOL 400 Run toggle on the Solution Parameters form. The need
to use the non-SOL 400 menus is is particularly true for thermal and frequency response analysis
where the ONLY way to set up the step is to use, respectively, the thermal preference and the
Frequency Response solution type menus to set up the analysis.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 221
Translation Parameters
Translation Parameters
Translation parameters define file formats, numerical tolerances, processing options, numbering offsets, and include
files used by Patran.
Main Index
222 Nonlinear Users Guide
Translation Parameters
Number of Tasks Represents the number of processors to be used to run an analysis. It is assumed
that the environment is configured for distributed parallel processing.
Numbering Options... Subform used to indicate offsets for all IDS to be automatically assigned during
translation.
Bulk Data Include File... Prompts you for the filename of the include file.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 223
Solution Parameters
Solution Parameters
Solution parameters control a range of functions in the SOL 400 analysis. Functions such as selecting the solver type,
establishing a restart, specifying domain decomposition are all part of the solution parameters.
Solver Options Specifies the solver to be used in numerically inverting the system of linear
equilibrium equations.
Contact Parameters Defines options for detecting and handling contact.
Direct Text Input This subform is used to directly enter entries in the File Management, Executive
Control, Case Control, and bulk data sections of the MSC Nastran input file.
Restart Parameters Includes a Restart option in the MSC Nastran input file.
Advanced Job Control Sets alternate versions of the solver and alternate formats for the results file.
Domain Decomposition Designates that domain decomposition be done manually, semi-automatically, or
automatically.
Main Index
224 Nonlinear Users Guide
Solution Parameters
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 225
Defining Subcases in Patran
Main Index
226 Nonlinear Users Guide
Defining Subcases in Patran
Subcase Parameters
The subcase parameters represent the settings in MSC Nastran Case Control and Bulk Data Section that take effect
within a subcase and do not affect the analysis in other subcases. Subcase parameters are dependent on the type of
analysis being performed. The set of subcase parameters applicable for each analysis type are described in the
following sections. For more information, see Chapter 3: Solution Methods and Strategies in Nonlinear Analysis in the
MSC Nastran Implicit Nonlinear Users Guide.
Entry Description
NLSTEP Defines parameters for automatic load/time stepping used in SOL 400.
NLPARM Nonlinear Static Analysis Parameter Selection.
NLPCI Defines a set of parameters for the arc-length incremental solution strategies in nonlinear static
analysis.
Note: The arc length method cannot be used with contact.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 227
Defining Subcases in Patran
Entry Description
NLSTEP Nonlinear Static Analysis Parameter Selection for doing creep analysis.
MATVPMA Defines creep characteristics based on experimental data or known empirical creep law. This
material definition should be used with advanced elements.
CREEP Defines creep characteristics based on experimental data or known empirical creep law. This
material definition should be used with conventional elements.
Main Index
228 Nonlinear Users Guide
Defining Subcases in Patran
Entry Description
EIGRL Defines data needed to perform real eigenvalue (vibration or buckling) analysis with the
Lanczos method (recommended).
EIGR Defines data needed to perform real eigenvalue analysis.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 229
Defining Subcases in Patran
Extraction Method Defines the method to use to extract the real eigenvalues.
Lancozs Parameters
Number of Modes Indicates an estimate of the number of eigenvalues to be located.
Lowest/Highest Frequency Defines the lower and upper limits to the range of frequencies to be
examined.
Sequence Checking Requests that Sturm sequence checking be performed on the extracted
eigenvalues.
Entry Description
NLSTEP Parameters for automatic load/time stepping for both static and transient nonlinear analysis
(recommended).
TSTEPNL Traditional way to specify Nonlinear Dynamic Analysis Parameter Selection. Has been replaced by
NLSTEP
Main Index
230 Nonlinear Users Guide
Defining Subcases in Patran
Entry Description
BCMOVE Specifies movement of rigid surfaces.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 231
Defining Subcases in Patran
Main Index
232 Nonlinear Users Guide
Defining Subcases in Patran
Initial Condition Load Factor... Multiplier applied to loads and constraints used as initial conditions of
complex eigenvalue analysis.
Enable Rotor Dynamics... Toggle this ON to enable Specify Spinning Properties form which
allows user to input rotordynamics properties.
Contact Table... Toggle Use Contact Table ON to use of contact table rather than
contact pairs in the analysis. Once toggled ON, the Contact Table
button brings up contact table input form.
Break Squeal... Allows user to input parameters used for break squeal analysis.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 233
Defining Subcases in Patran
Main Index
234 Nonlinear Users Guide
Defining Subcases in Patran
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 235
Defining Subcases in Patran
Main Index
236 Nonlinear Users Guide
Defining Subcases in Patran
In linear analysis, subcases refer to a separate analysis case constituted by substituting a different load vector into the
equation, making each subcase the equivalent of a separate analysis. SOL 400 has kept this terminology and allows
you to run multiple separate analyses in the same job by treating each subcase as a separate analysis. This treats the
structure at the beginning of the first step of the subcase as an un-deformed, unloaded structure with no residual stresses
or strains carried over from the previous subcase. This is in opposition to a step, where the loads and conditions in place
at the end of the previous step constitute the starting point of the next step. Thus, a series of steps in one subcase
constitute a set of loading steps in a separate analysis.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 237
Editing a MSC Nastran Input File
Main Index
238 Nonlinear Users Guide
Monitoring the Analysis
The first column shows the procedural step, while the second column shows the increment within the step. Note that
not every increment size is equal, as can be seen in the TIME STEP OF INC column. The third column (cycl #)
indicates the number of attempts made during this increment.
Patran provides a convenient way to access the status file using the Monitor option of the Analysis form In addition to
opening jobname.sts (Patran knows the working directory and jobname already), you can also view the other MSC
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 239
Monitoring the Analysis
Nastran job files including the .f06, .f04, and .log. The keyword search option makes it easy to find errors in the
SOL 400 run by searching the .f06 file (search for fatal to immediately find the fatal error that terminated the job).
Main Index
240 Nonlinear Users Guide
Monitoring the Analysis
The Analysis Manager provides interfaces within Patran to submit, monitor and manage analysis jobs on local and
remote networked systems. It can also operate in a stand-alone mode directly with MSC Nastran and Marc.
At many sites, engineers have several computing options. Users can choose from multiple platforms or various queues
when jobs are submitted. In reality, the resources available to them are not equal. They differ based on the amount of
disk space and memory available, system speed, cost of computing resources, and number of users. In networked
environments, users frequently do their modeling on local workstations with the actual analysis performed on compute
servers or other licensed workstations.
The Analysis Manager automates the process of running analysis software even on remote and dissimilar platforms.
Files are automatically copied to where they are needed; the analysis is performed; pertinent information is relayed
back to you; files are returned or deleted when the analysis is complete even in heterogeneous computing
environments. Time consuming system housekeeping tasks are reduced so that more time is available for productive
engineering.
The Analysis Manager replaces text-oriented submission scripts with a Motif-based menu-driven interface (or
windows native interface on Windows platforms), allowing you to submit and control his job with point and click ease.
No programming is required. Most users are able to productively use it after a short demonstration.
The Analysis Manager can also be used to monitor job progress (even non-jobs)
Shows all job and host information
Abort function can be used to stop running jobs
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 241
SOL 400 Analysis Messages
MSC Nastran provides many user fatal error messages (UFM). A normal/successful run exit message is as follows:
Note that for a job that terminates abnormally you are referred to the Fatal Error Message in the jobname.f06 file.
UFMs will be covered in a later discussion in this chapter.
One way to debug convergence issues is to use the NLOPRM bulk data entry, such as NLDBG = ADVDBG, DBGPOST
options and so on, to load increment diagnostics in the .f06 file, which may help debug the model. Each iteration of
each load increment generates a report in the .f06 file giving the convergence and stiffness update information.
Main Index
242 Nonlinear Users Guide
SOL 400 Analysis Messages
If the job terminates before 100% of the loading is applied this convergence information can be used to identify
possible reasons. The default numerical method used by SOL 400 is the full Newton method with the load increment
size adjusted according to how many iterations are required to achieve convergence. The incremental load size will be
adjusted up if convergence is achieved easily, and will be reduced if more than the target number of iterations is
required to achieve convergence. Convergence is achieved once the error factors on the selected criteria are below the
required values (this is called the convergence criteria). In this convergence and stiffness update information, the
following values are given:
LOAD STEP percentage of total load reached at this incremental load value
NO. INC increment number of this increment relative to all increments in this step
MAT DIV number of iterations that diverged due to material issues in this increment
TOTL WORK integration of the forces over the displacements (e.g. total work) over the model
AT GRID tells the grid/node number at which the maximum displacement occurred
C tells the degree of freedom (component) of grid/node at which the maximum displacement occurred
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 243
SOL 400 Analysis Messages
NO number of line searches (does not apply with full Newton method)
Main Index
244 Nonlinear Users Guide
Convergence
Convergence
From a users perspective, when performing a nonlinear simulation, the most difficult thing is to resolve convergence
problems encountered in the analysis.
Convergence in Optimization
Convergence in Equilibrium.
In this section, we will only be focusing on convergence equilibrium of the structural system. Figure 6-1 gives a quick
summpar of steps you should undertake.
For structural analysis, a lack of convergence implies that the numerical solution has not reached equilibrium to the
desired level of accuracy. Even if convergence is achieved, it is strongly recommended to determine how many
iterations were required and to consider if this was a reasonable number or an excessive number.
Recall that the number of iterations has a significant influence on the computational costs, so not only is there the
requirement that the solution converges, but that it also converges efficiently.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 245
Convergence
Main Index
246 Nonlinear Users Guide
Convergence
A value of 1 iteration per increment may be deceiving. It means that the testing is based upon relative residual
force/moment, and there could be an exceptionally high value of the
reaction force.
A value of 2 iterations per increment is very good.
A value between 3-6 iterations is good; perhaps a smaller time step should be used, but it could also
indicate that the nonlinearity is changing location/character in the
simulation.
A value of 7-8 indicates that caution should be taken. The causes are discussed below.
A value greater than 8 indicates problems with the simulation. You should take steps to modify
your model.
There are multiple reasons for why a job does not converge and often multiple phenomena may be present in the same
model. The following section provides an overview of the possible steps to remove the problem.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 247
Convergence
One may use the MAXRATIO parameter to set the value which is considered to be an excessive pivot. The default is
1.e7 and this value can be increased to 1.e10 if necessary.
AUTOSPC
The AUTOSPC case control command may be used to automatically apply SPC to the system to remove the
singularities. This is recommended when there are contact bodies that are not in contact at the beginning of the analysis.
In a SOL 400, the recommended approach is to use the AUTOSPC in each step where it is appropriate, such as:
SUBCASE 1
STEP 10
AUOSPC (RESIDUAL,SPC)=YES
LOAD = 10
STEP 20
AUOSPC (RESIDUAL,SPC)=YES
LOAD = 20
STEP 30
LOAD = 30
Main Index
248 Nonlinear Users Guide
Convergence
Contact Friction
Friction is a highly nonlinear process and almost always induces the need to take more iterations in the simulation.
Many problems are friction dominated (such as insertion of clips or brake squeal), and additional iterations are required
for these simulations. If it is not a friction dominated problem, it is often beneficial to first suppress friction using glued
contact to achieve convergence, and then add it in a subsequent simulation. Or one can set the friction coefficient to
zero to neglect the friction.
Material Nonlinearity
Material behavior may cause convergence problems if the material no longer has the ability to carry the load or the
stress. There are multiple physical and numerical reasons for this.
The temperature increases and the material physically softens. This is true for virtually all materials, but for some, the
reduction in load carrying capacity occurs over a small temperature range. Hence, one may need to reduce the time
step.
Elastic-Plastic
While textbooks often discuss elastic pefectly plastic material, and this model only requires one additional material
constant, from a stability and convergence perspective a work-hardening material is preferred. The simplest example
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 249
Convergence
of this is in frame structures a plastic hinge may form if the material is perfectly plastic. To improve convergence, if
no other data is available, one can set the work hardening to 1% of the Youngs modulus.
Composites
Material damage occurs in the composite material which reduces the load carrying capability. This potentially occurs
for MATF or MATDIGI materials. If sufficient material damage as occurred such that there are no longer load paths, then
failure to converge indicates that one has reached the maximum load.
Fracture
The cohesive material used in conjunction with Cohesive Zone Model fails leading to delamination and the need to
redistribute the load. This is a highly nonlinear process.
User Subroutines
User subroutines are very powerful tool to customize the material behavior, but caution is required when using them.
It is very easy to construct a user subroutine that does not represent the physical problem and hence, leads to an unstable
solution. It is important to test material user subroutines to insure stability and proper convergence. When using
MATUSR and user subroutines UMAT or HYPELA, it is important that the tangent stress-strain law is consistent with
the stress to insure quadratic convergence of the Newton-Raphson procedure. User subroutines should be tested with
small load controlled problems to verify their accuracy and robustness before being applied to engineering problems.
Incompressibility
Nonlinear structural analysis presents many scenarios that result in incompressible or nearly incompressible behavior,
including small strain elasticity (with Poissons ratio close to 0.5), rubber elasticity, plasticity, and creep. When using
SOL 400 for these materials, it is strongly recommended to use the Advanced elements that are activated with the
additional Property options.
Nearly incompressible elements store significant strain energy when under compression. If the boundary conditions
are changed, it is likely that there will be a dramatic motion of the rubber material leading to convergence issues.
Mesh Distortion
For large deformation problems, especially for those undergoing large shear strains, the element may become inside
out which is discussed below. But, even before this point, the shape functions will be less effective to represent the
deformation which may lead to poorer convergence. This puts additional constraints on the mesh generation process.
Not only must the mesh to be good for the initial loading, but also good for the complete duration of the event. Lower
order elements are preferred when large deformations occurs.
Instability/Buckling
Many structures are loaded such that an instability may occur. Numerically, this may appear as either a message
(Excessive Pivot Ratio) or an increase in the number of iterations. In general, these are very difficult models
which require multiple approaches to solve. The fundamental problem is that the compressive load is too large and this
results in a non-positive definite system or an excessive pivot.
Main Index
250 Nonlinear Users Guide
Convergence
The first thing to check at this point is: Has the solution gone as far as it can?; i.e., maybe you already have the
complete solution. Sometimes, nonlinear loading causes structures to buckle which may take your analysis into the
post-buckled region. Depending on the type of buckling, you may be simply trying to drive your analysis farther into
the post-buckling range which may not give you the information you are after. For example, if you load a frame
structure with a load that exceeds the critical buckling load, the analysis is simply working to drive the structure further
into a plastic hinge.
There are several approaches to resolve buckling and instability issue:
The traditional way of solving buckling simulations is to use arc-length methods as defined in Arc-Length
Methods for Post-buckling Analysis in Chapter 4: Solution Strategies for Nonlinear Analysis. Unfortunately, this
method is not effective when the buckling modes are in close proximity. This problem occurs in typical
aerospace panels with multiple spars and struts. The arc-length methods also cannot be used with contact.
Use the NLSTEP method and activate the artificial damping by setting IDAMP=4.
Add transient dynamics and use a large time period. The inclusion of dynamics improves the numerical
stability by insuring that the system is always positive definite.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 251
General Hints for Starting out with Nonlinear Analysis
Output Messages
Additional Output
When convergence is a problem it is necessary to obtain as much information as possible on the numerical process.
The NLOPRM case control command is used to obtain additional result information.
In particular:
NLOPRM OUTCTRL=STD,INTERIM, NLDBG=NLBASIC,NRDBG, DBGPOST=LTIME
will provide the additional information to evaluate the performance. If contact bodies are present, it may be preferable
to use:
NLOPRM OUTCTRL=STD,INTERIM, NLDBG=NLBASIC,NRDBG, DVDBG,N3DMED DBGPOST=LTIME
Exit 0 success. The job ran to completion and did everything you asked.
Main Index
252 Nonlinear Users Guide
General Hints for Starting out with Nonlinear Analysis
Exit 1 syntax error in the input file. You should check the input syntax of the line the
error message points to, but it is likely that the actual error was in the input
block prior to where the message points.
Exit 1 with user fatal typically means no convergence due to rigid body motions or a numerically ill-
EXCESSIVE PIVOT RATIO conditioned system. See recommendations for equilibrium below.
in .f06
Exit 1 with other user fatal means the analysis ran into convergence problems part way through and did not
complete. Interpret the error message, look for model problems
For SOL 400 User Fatal http://simcompanion.mscsoftware.com/infocenter/
Messages (UFMs) and their
Interpretation see:
If you receive a Nastran input from another source, it is strongly recommended that you read it into either Patran or
SimXpert and display the model to check for completeness.
Any .sts file with nonzero total time lines means there are converged increments. Display the converged increments
to see what is going on.
Main Index
Chapter 6: Setting Up, Monitoring, and Debugging the Analysis 253
General Hints for Starting out with Nonlinear Analysis
User-specified Upper Limit on Number of Iterations: The solution may fail and give this message if SOL 400
reaches the user-specified upper limit on the number of (Newton-Raphson) iterations within a load increment. The
default is set to 10. One of the first things to try, if you think the problem is otherwise well posed (e.g., no other
problems are known) would be to increase the value (specified on NLSTEP) to a value of 20... What is happening is
that SOL400 keeps iterating and tries to converge to a solution for that increment. If that does not happen, SOL 400
will cut back the load (by half) and re-solve that increment. Sometimes this is not enough to get convergence, and it
will exit with a UFM.
Check the residual values in the ADVDBG section of the .f06 file to confirm this.
Minimum Time Step Size Exceeded/Time-Step Size Too Small: This exit message indicates that SOL 400 cuts back
to a time-step size too small for the analysis to continue. The load stepping algorithm has a cut-back feature where the
load step is automatically reduced (when the time step is decreased, the factor is calculated internally based upon the
minimum time step). When an increment runs into these problems, it automatically cuts the load-step size and resolve
that increment. If the problem persists, it cut-back the load-step again. This happens until the limit of the number of
cut-backs is reached. This can result in a very small time step. In such a case, SOL 400 stops the analysis with a UFM.
Allowable Number of Bisections Exceeded: SOL 400's automatic load stepping scheme is set up such that the applied
load in an increment scales up (or down) depending on how easy (or difficult) the solution was in the previous
increment. The degree of difficulty is determined based on the NDESIR parameter of NLSTEP: 'desired number of
iterations per increment' (default = 4). SOL 400 will scale down the step size until it reaches a lower limit on the step
size (default = 0.001% of total time step) and then exit with a UFM worded similar to the above. This is an indication
to you that the analysis encountered some difficulty at that stage.
Main Index
254 Nonlinear Users Guide
General Hints for Starting out with Nonlinear Analysis
Main Index
Chapter 7: Results Output from the Analysis
Nonlinear Users Guide
Main Index
256 Nonlinear Users Guide
Introduction
Introduction
This chapter describes the process of requesting results from a SOL 400 analysis. It includes the topics of: case control
input which specifies the specific results quantities included in the output; how to make these requests in the SOL 400
input file as well as when setting up the job in Patran; which results quantities (such as contact output) are included in
the requested output data blocks; what output format or file type is recommended for the various solution sequences;
as well as the available and recommended.
The interpretation of the results requested is beyond the scope of this chapter, but is in Chapter A: Interpreting the
Results. In that chapter, a complete description of the results, including the mathematical descriptions of the various
types of results, quantities such as Cauchy stress and engineering strain definitions, are provided. Appendix A also
includes details on interpreting the fringe plots made by the graphical user interfaces like Patran, and the numerical
calculations they use to arrive at the exact value that shows up in the fringe plot. Understanding these calculations is
essential when interpreting the results of a fringe plot, particularly in areas where a course mesh means that the values
are extrapolated a significant distance between the element integration point, where the stress is calculated, and the
node point where the fringe value is represented.
Main Index
Chapter 7: Results Output from the Analysis 257
Output Requests in Case Control
Main Index
258 Nonlinear Users Guide
Output Requests in Case Control
Main Index
Chapter 7: Results Output from the Analysis 259
Output Requests in Case Control
For a complete list of the output requests available via the case control commands, see Case Control Command
Summary in the QRG.
NLOPRM
Case control command, NLOPRM, can be used to better control the nonlinear solution output during solution process,
to provide MSC Nastran users a direct access to nonlinear solutions even the job is still running, to give you some tools
to monitor and debug the nonlinear solution process and gain some insight of nonlinear solution procedure, and to
allow users to print out MPC and MPCY equations from contact constraints before and during a contact process.
Main Index
260 Nonlinear Users Guide
Output Requests in Case Control
Input
Examples
NLOPRM OUTCTRL=STD,SOLUTION DBGPOST=LTIME
NLOPRM OUTCTRL=(SOLUTION,INTERM), MPCPCH=(OTIME,STEP)
Output
The output from NLOPRM case control command is basically controlled by four keywords of OUTCTRL, NLDBG,
DBGPOST, and MPCPCH. Each keyword has a group of describers that are assigned to either by one at a time or more
in juxtaposition. The output is destined to almost all MSC Nastran output media, such as .f06, .pch, .op2, and
DBALL, depending on which keyword is used.
Guidelines
1. NLOPRM case control command may only appear above all SUBCASE, STEP, and SUBSTEP delimiters.
2. For OUTCTRL=SOLUTION, only nonlinear solutions, such as nonlinear stresses, strains, contact status and so
on, are output at the user-specified output intervals. Any solution results in superelements are not computed and
recovered in what we called Phase 3 data recovery. The job is terminated as soon as all nonlinear iterations are
completed.
The nonlinear solution results are also saved in DBALL when a job is launched with scratch=post. They
are ready for postprocessing when the job is completed.
3. When OUTCTRL=INTERM is specified, the nonlinear solutions, such as stresses, strains, contact status and so
on, are output into individual .op2 files for postprocessing, at the user-specified output intervals. You are able
to access these files while the job is still running. The name of a typical .op2 file is the job name followed by
a suffix name of eight-digit number, for instance, my_job.00000008. In .f06 file, the relevant information
of an .op2 file is indicated as corresponding to the load or time increment, STEP, SUBSTEP, and SUBCASE.
4. NLDBG is advanced contact output request for SOL400 users. Users can look into details of nonlinear iteration
for more contact details.
Main Index
Chapter 7: Results Output from the Analysis 261
Output Requests in Case Control
Among the debugging options, N3DBASE can be used to print out some basic contact information in the .f06
file. It includes the contact condition of touching node on the touched patch and separation contact forces.
Standard contact status output can be requested by the BOUTPUT case control command.
5. DBGPOST is used to select the output of nonlinear iterations for debugging purpose. When DBGPOST is
activated, a MSC Nastran data block, OFDBGDT, is created to store both residual and displacement vectors at
user-specified iterations pertaining to load or time increments, STEP and SUBCASE.
6. MPCPCH allows you to punch out multipoint constraint equations from a contact process in the format of either
MPC or MPCY bulk data entries. This is probably the most useful tool for you to gain some insight of how a
MSC Nastran contact job proceeds.
Main Index
262 Nonlinear Users Guide
MSC Nastran Results Quantities
This allows the definition of what data is desired from the analysis code in the form of results. For most solution
sequences, the form consists of two formats: Basic and Advanced. The Basic form retains the simplicity of being able
to specify the output requests over the entire model and uses the default settings of MSC Nastran case control
commands. There is a special set defined in Patran called ALL FEM which represents all nodes and elements associated
with Object defined on the Analysis Form in the Patran Interface to MSC Nastran Preference Guide. This default set
is used for all output requests in the Basic Output Requests form in the Patran Interface to MSC Nastran Preference
Guide.
The Advanced version of this form allows the user to vary these default options. Since output requests have to be
appropriate to the type of analysis, the form changes depending on the solution sequence. The Advanced Output
Requests in the Patran Interface to MSC Nastran Preference Guide also adds the capability of being able to associate
a given output request to a subset of the model using Patran groups. This capability can be used effectively in
significantly reducing the results that are created for a model, optimizing the sizes and translation times of output files.
The creation of Patran groups are documented in Group>Create in the Patran Reference Manual.
The results types that will be brought into Patran due to any of these requests, are documented in Supported OUTPUT2
Result and Model Quantities in the Patran Interface to MSC Nastran Preference Guide. In that chapter, tables are
presented that correlate the MSC Nastran results block, and the Patran primary and secondary results labels with the
various output requests
SOL 400 produces stress and strain results that differ from those results available with other solution sequences.
Main Index
Chapter 7: Results Output from the Analysis 263
Specifying Results Output Requests in Patran
This listbox displays the selected Output Requests for the subcase
shown at the top of the form.
The Delete button deletes the output request highlighted in the Output
Requests listbox.
The TITLE, SUBTITLE, and LABEL are written to the MSC Nastran
output file.
Figure 7-1 Output Requests are Made From the Patran Subcase - Output Request Form
Main Index
264 Nonlinear Users Guide
Specifying Results Output Requests in Patran
Selected Output Results show up in The selected Output Requests will be written to the associated subcase
Output Requests Data Box when the input file is written.
Delete Selecting the Output Request from the data box and hitting Delete
removes that output request.
Main Index
Chapter 7: Results Output from the Analysis 265
Specifying Results Output Requests in Patran
Use this listbox to select the This listbox is used to select the group to
result type to be created. which the output requests relate.
Main Index
266 Nonlinear Users Guide
Result Files
Result Files
When a SOL 400 analysis has been successfully completed, a status file (jobname.sts), a print file
(jobname.f06) and the requested results files are created and saved.
Main Index
Chapter 7: Results Output from the Analysis 267
Result Files
job_id.MASTER Part of the recommended indexing modules are tied to a system cell. That is, an MSC
Nastran database is indexed and saved by MSC Nastran by setting system cell 316 to a
value 7. This system cell tells MSC Nastran executable to create index files for IFP and
OFP datablocks and move the indexed datablocks to the MASTER file. This means that
one can even delete the DBALL file after the MSC Nastran run completes. For example,
if you would like to get an indexed MASTER data file for the job, some_job.bdf, the
following must be executed:
miscellaneous scratch Several scratch files are generated during the analysis which MSC Nastran will
files automatically delete upon completion of the run.
Main Index
268 Nonlinear Users Guide
Specifying Output File Format
MSC Nastran
The following table indicates the options required to select the type of postprocessing file.
Patran
The Results Output Format form is used to request results from the MSC Nastran analysis for use in postprocessing
and verification (output file).
1. Click on the Analysis Application button to bring up the Analysis Application form.
2. On the Analysis Application form, select Solution Type button to bring up the Solution Parameters form.
3. On the Solution Parameters form, select the Results Output button to bring up the Results Output Format form.
4. From this form you can select the output file format(s) you wish to use. The next sections of this chapter will
provide more information about each format to help you select the best format for your purpose.
Figure 7-2 Output Requests are Made From the Patran Subcase - Output Request Form
After the input file echo, the OLOAD RESULTANT is printed which shows the resultant of the loads about the origin
of the basic coordinate system in this case. Next is the USER INFORMATION MESSAGE (UIM) 5293 which
shows the epsilon and external work for the run. The epsilon is a measure of the numeric conditioning of the model,
while the external work is the work due to the applied loads. Both of these values are useful for model checkout and
are discussed in more detail in Model Verification (Ch. 10) in the MSC Nastran Linear Static Analysis Users Guide.
Main Index
Chapter 7: Results Output from the Analysis 269
Postprocessing SOL 400 with Patran
3. After selecting a MASTER or .op2 file, you may want to specify translation parameters.
Tolerances
Division Prevent division by zero errors.
Numerical Compares real values for equality.
Additional Results to be Imported
Rotational Nodal Results Indicates whether Rotational Nodal Results are skipped or included in
translation.
Stress/Strain Invariants Indicates whether Stress/Strain Invariants are skipped or included in
translation.
Main Index
270 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
In addition to these standard results quantities, several Global Variable results can be created. Global Variables are
results quantities where one value is representative of the entire model at a particular load increment.
Results
Like the enormous amount of data needed to define the simulation model to an analysis code, there is a large volume
of data returned from the simulation analysis. And just as it is virtually impossible to construct a model with a text
editor alone, it is equally as difficult to read and interpret the results by hand. Using a postprocessor with a graphical
user interface such as Patran or SimXpert is highly recommended.
Main Index
Chapter 7: Results Output from the Analysis 271
Postprocessing SOL 400 with Patran
The Patran Results application gives you control of powerful graphical capabilities to display results quantities in a
variety of ways:
Deformed structural plots
Color banded fringe plots
Marker plots (vectors, tensors)
Freebody diagrams
Graph (XY) plots
Animations of most of these plot types.
The Results application treats all results quantities in a very flexible and general manner. In addition, for maximum
flexibility results can be:
Sorted
Reported
Filtered
Derived
Deleted
All of these features help give meaningful insight into results interpretation of engineering problems that would
otherwise be difficult at best.
The Results application is object oriented, providing postprocessing plots which are created, displayed, and
manipulated to obtain rapid insight into the nature of results data. The imaging is intended to provide graphics
performance sufficient for real time manipulation. Performance will vary depending on hardware, but consistency of
functionality is maintained as much as possible across all supported display devices.
Capabilities for interactive results postprocessing also exist. Advanced visualization capabilities allow creation of
many plot types which can be saved, simultaneously plotted, and interactively manipulated with results quantities
reported at the click of the mouse button to better understand mechanical behavior. Once defined, the visualization
plots remain in the database for immediate access and provide the means for results manipulation and review in a
consistent and easy to use manner.
Postprocessing Contact
Rigid geometry results that exist in the MSC Nastran results file contain translation and rotation information per
increment. The rigid body NURB data (rigid geometry) can be imported into an empty database, but any translation or
Main Index
272 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
rotation of that rigid geometry is only visible, viewable, or able to animate within Patran under the following
conditions:
The standard output includes, at the end of each increment, a summary of information regarding each body. This
information reports the increment's rigid body velocity, the position of the center of rotation, and the total loads on the
body. These last values are obtained by adding the contact forces of all nodes in contact with the rigid body.
Deformable bodies being in equilibrium have no load reported.
Additional information can be obtained by means of the NLPOPRM case control command, NLDBG keyword. In such
cases, all the contact activity is reported. Namely, every time a new node touches a surface or separates from a surface,
a corresponding message is issued.
The MSC Nastran results file contains the results for both the deformable bodies and the rigid bodies. In performing a
contact analysis, you can obtain three types of results. The first is the conventional results from the deformable body.
This includes the deformation, strains, stresses, and measures of inelastic behavior such as plastic and creep strains. In
addition to reaction forces at conventional boundary conditions, you can obtain the contact forces and friction forces
imparted on the body by rigid or other deformable bodies. By examining the location of these forces, you can observe
where contact has occurred, but MSC Nastran also allows you to select the contact status as a results file variable:
A value of 0 means that a node is not in contact.
A value of 0.5 means the node is in near thermal contact.
A value of 1 means that a node is in contact.
A value of 2 means the node is on a cyclic symmetry boundary.
It is also possible to obtain the resultant force following from contact on the deformable bodies and the resultant force
and moment on the rigid bodies. The moment is taken about the user-defined centroid of the rigid body. The time
history of these resultant forces are of significant issues in many engineering analysis. Of course, if there is no resultant
force on a rigid body, it implies that body is not in contact with any deformable body.
Finally, if the additional print is requested via the NLPOPRM case control, NLDBG keyword, the output file reflects
information on when a node comes into contact, what rigid body/segment is contacted, when separation occurs, when
a node contacts a sharp corner, the displacement in the local coordinate system, and the contact force in the local
coordinate system. For large problems, this can result in a significant amount of output.
The motion of the rigid bodies can be displayed as well as the deformable bodies. Rigid bodies which are modeled
using the piecewise linear approach are displayed as line segments for flat patches.
Main Index
Chapter 7: Results Output from the Analysis 273
Postprocessing SOL 400 with Patran
Main Index
274 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
Main Index
Chapter 7: Results Output from the Analysis 275
Postprocessing SOL 400 with Patran
Main Index
276 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
Main Index
Chapter 7: Results Output from the Analysis 277
Postprocessing SOL 400 with Patran
Main Index
278 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
Main Index
Chapter 7: Results Output from the Analysis 279
Postprocessing SOL 400 with Patran
Main Index
280 Nonlinear Users Guide
Postprocessing SOL 400 with Patran
Main Index
Chapter 8: Load and Constraint Conditions
Nonlinear Users Guide
Main Index
282 Nonlinear Users Guide
Introduction
Introduction
The purpose of this chapter is to describe the procedures for applying loads and constraints to SOL 400 models. Each
type of load or constraint is described, including how it is used in finite element modeling and in the MSC Nastran
input. Further information on how the various types of loads and constraints are used can be found in standard finite
element text books, and further information on the MSC Nastran input format can be found in the QRG.
Main Index
Chapter 8: Load and Constraint Conditions 283
Constraints
Constraints
In static analysis, the rigid body modes must be restrained in order to remove the singularity of the stiffness matrix.
The required constraints may be supplied with single point constraints, multipoint constraints, or free body supports.
If free body supports are used, the rigid body characteristics will be calculated and a check will be made on the
sufficiency of the supports.
This process involves specifying the appropriate boundary conditions for your model. Boundary conditions are
imposed in the form of constraints on selected degrees of freedom on the model. Typically, several degrees of freedom
(at least six) are constrained to ground using either SPC bulk data entries or the PS field of the GRID entry.
Besides single-point constraints, MSC Nastran provides a method of creating linear constraint relationships between
several degrees of freedom known as multiple-point constraints or MPCs. This chapter discusses both the single-point
and multiple-point constraints.
Main Index
284 Nonlinear Users Guide
Constraints
y
n
CROD 1
x
G
CROD 2
Main Index
Chapter 8: Load and Constraint Conditions 285
Enforced Motion Constraints (SPCD and SPCR)
Main Index
286 Nonlinear Users Guide
Applying Constraints
Applying Constraints
Once you have constructed a model of your structure, constraints are added that force selected portions of your model
to remain fixed or to move by a specified amount. These constraints can be either:
Single Point Constraints
Multipoint Constraints
i = u (8-1)
where u is the value of the prescribed displacement on the degree of freedom i . The case of u = 0 is the most
common case, and is often used as a boundary condition, to fix or ground the movement of a point in a certain
direction.
Since the value of i is known, one could, in principle, eliminate the specified degree of freedom from the other
degrees of freedom to be solved for as unknowns. This would reduce the size of the system of equations to be solved,
but on the other hand it would take time to perform the elimination, and this approach adds complexity to the code.
SOL 400 uses a different technique. A number which is large compared to the stiffness coefficients (say, for discussion,
10 20 ) is added to the diagonal term K 11 of the equation for the degree of freedom to be constrained. Also, if the degree
20
of freedom is to be constrained to a nonzero value u, then u 10 is added to the right hand side of the modified
equation. This modified equation is now:
K i1 1 + + K ii 10 20 i + + K in n = F i + u 10 20 (8-2)
Assuming all K ij to be small with respect to 10 20 , the solution of the system of equations is obtained with negligible
error.
The modified system of equations remain well conditioned. The value used by SOL 400 for the large number is 10 10
times the largest stiffness coefficient found on the diagonal of the stiffness matrix.
Main Index
Chapter 8: Load and Constraint Conditions 287
Spatial, Time, Temperature, and Other Dependencies
Main Index
288 Nonlinear Users Guide
Static Loads
Static Loads
A load is applied to the model only if it is specifically called out in the case control section. For reasons discussed
earlier, this provides you with greater versatility. If, however, you forget to request any load in the case control section,
the problem will be solved with zero loads applied. There are no error or warning messages indicating that there is no
load being applied. Forgetting to specify a load request in the case control section can be a common mistake many new
users make. An indication of this problem is when all of the displacements and stresses come out to be zero.
Load Sets
Table 8-1 Bulk Data Entries for Loads
Loads
FORCEi Defines concentrated load at grid point.
MOMENTi Defines moment at a grid point.
NOLIN1i Defines nonlinear transient load.
PLOAD Defines pressure loads on CQUAD4, CTRIA3, CHEXA, CPENTA, and CTETRA. Should not be used
for hyperelastic plane elements CQUAD4, CQUAD8, CQUAD, CTRIA3, and CTRIA6 or for
hyperelastic CHEXA, CPENTA, CTETRA with midside nodes.
PLOAD2 Defines pressure loads on shell elements, CQUAD4 and CTRIA3. Not available for hyperelastic
elements.
PLOAD4 Defines pressure loads on surfaces of CHEXA, CPENTA, CTETRA, CTRIA3, and CQUAD4 elements.
Not available for hyperelastic plane elements CQUAD4, CQUAD8, CQUAD, CTRIA3, and CTRIA6.
PLOADX1 Defines pressure loads on axisymmetric elements CQUADX and CTRIAX.
RFORCE Defines load due to centrifugal force field.
Subcases
In linear analysis, a subcase contains all the loads and boundary conditions used within a single linear analysis case.
For computational efficiency, it is possible to run multiple subcases (using the same inverted stiffness matrix) in the
same analysis run and simply resolve using the different loads in the multiple subcases.
For example, one load case may represent the loads and boundary conditions for each time point in a time-dependent
nonlinear analysis. Multiple load cases can be applied to the same model as subcases for linear analysis to examine
how the model reacts to different loading conditions. Load cases are central to the ability to perform complex analyses
on an individual model. For nonlinear analysis, multiple loadcase runs are used to define the load history on the model.
The ending point of the last step is the starting point of the next step.
For SOL 400 runs, the Subcase Select form looks the same, except for the Select Steps for New Subcase button is
not grey (it can be picked).
For SOL 400, you can select subcases just as in any other solution case, except for one difference. SOL 400
differentiates between load steps and subcases. A number of load steps make up a subcase. The list of load steps in the
Selected Load Steps list box shows you which Patran subcases are designated as load steps for the analysis. A
Main Index
Chapter 8: Load and Constraint Conditions 289
Static Loads
subordinate form shows you which of those selected load steps are designated as the first load steps in the subcases to
be run in the analysis. This is somewhat confusing terminology as it conflicts with Patran's terminology of a subcase.
To help you visualize the setup of subcases versus load steps, a tree widget is used to show the subcases and which
load steps are selected for each subcase. You can interact with either the tree widget or the list boxes. To add and delete
subcases, load steps, or to otherwise manipulate them on the tree widget, use the right mouse button after selecting a
tree item.
Load Steps
A load step (or analysis step) is defined by associating a load case, an analysis procedure, output requests, and any
associated parameters that guide the solution path for the chosen analysis procedure. Whereas a load case is a collection
of loads and boundary conditions for a particular load step, a load step is a collection of relevant analysis parameters
including the associated load case.
The load for a subcase is often subdivided into the number of increments specified for the subcase. The solution
strategy in nonlinear analysis is to apply the loads in an incremental fashion until the desired load level is reached.
Load Types
The static loads in nonlinear analysis consist of concentrated loads, distributed loads, and thermal loads as well as
applied displacements. Most of the relevant loads data applicable to the linear static analysis are also applicable to
nonlinear static analysis. Transient loads define the loadings as functions of time and the location. A load can be
applied at a particular degree of freedom, pressure over the surface area, or the body force simulating an acceleration.
MSC Nastran is based on the stiffness method and deals primarily with force-displacement relations. In a linear elastic
system, force and displacement are related through the constant stiffness of the system; the governing equation of such
a system can be expressed as
Ku = F (8-3)
where K is the stiffness matrix and u and F are nodal displacement and nodal force vectors, respectively. Equation
(8-3) can be solved either for unknown displacements subjected to prescribed forces or for unknown forces (reactions)
subjected to prescribed displacements. At any nodal point, the nodal forces and nodal displacements cannot be
simultaneously prescribed as boundary conditions for the same degree of freedom.
Note: You must prescribe at least a minimum number of boundary conditions to insure that rigid body motion does
not occur.
The prescribed force boundary conditions are often referred to as loads and the prescribed displacement boundary
conditions as boundary conditions.
Note: Boundary conditions can be prescribed in either the global or a local coordinate system. A nodal
transformation between the global and the local coordinate systems must be carried out if the boundary
condition is prescribed in a local system.
Main Index
290 Nonlinear Users Guide
Static Loads
In a nonlinear stress analysis problem, MSC Nastran carries out the analysis incrementally and expresses the governing
equation in terms of the incremental displacement vector u and the incremental force vector F .
Ku = f (8-4)
Consequently, you must also define both the loads and the prescribed nodal displacements incrementally.
In addition to the prescribed displacement boundary conditions, constraint relations can exist among the nodal
displacements. For example, the first degree of freedom of node i is equal to that of node j at all times. The expression
of this constraint relation is
ui = uj (8-5)
Generally, a homogenous linear constraint equation can be expressed as
ut = a1 u1 + a2 u2 + + an un (8-6)
where u represents the degrees of freedom to be constrained, u 1 u n are other retained degrees of freedom in the
structure, and a 1 a n are constants.
You can enter constraints through either the MPC, RBE2, or RBE3 bulk data entries.
Loading Types
You can categorize a particular type of load as either a point (concentrated) load or surface/volumetric (distributed)
load, depending on application conditions. The spatial distribution of the load can be uniform or nonuniform. Special
loading types also exist in various analyses. For example, centrifugal loading exists in stress analysis, and convection
and radiation exist in heat transfer analysis. You can add point loads directly to the nodal force vector, but equivalent
nodal forces first must be calculated by MSC Nastran from distributed loads and then added to the nodal force vector.
These distinguishing features are described below.
A point (or nodal) load of either a vector (force, moment) or a scalar (heat flux) quantity is a concentrated load that is
applied directly to a nodal point (see Figure 8-2). Mechanical point loads can be defined as fixed direction forces or as
follower forces. In a global or local coordinate system, a force vector must be defined in terms of its vector components
(see Figure 8-3). If the force vector is defined in a local coordinate system, then a global-to-local coordinate
transformation matrix must be defined for the nodal point (see Figure 8-4 and Figure 8-5). For axisymmetric elements,
the magnitude of the point load must correspond to the ring load integrated around the circumferences.
Q F
y y
x x
Point (a) Point (b)
Heat Flux Q (Scalar) Force F (Vector)
Figure 8-2 Schematic of a Point Load
Main Index
Chapter 8: Load and Constraint Conditions 291
Static Loads
Fy F F
Fy
Fx Fx
y y x
x x
Figure 8-3 Force Components
y x
Fy = 0
y F = Fx
x
Figure 8-4 Special Selection of Local (x, y) Coordinate System Force Components: F y'' = 0
SLOAD
The SLOAD entry is used to apply loads to scalar points only it cannot be used with grid points.
Main Index
292 Nonlinear Users Guide
Distributed Loads
Distributed Loads
PLOADB3
The PLOAD3 bulk data entry in the QRG defines a distributed load over the entire length of a CBEAM3 element axis.
The load is applied along the line of the shear center.
PLOAD2
The format for the PLOAD2 bulk data entry is similar to the PLOAD entry except that the PLOAD2 references the
element ID instead of the grid points where the element is attached.
The PLOAD2 is usually preferred over the PLOAD entry because it is easier to use. The PLOAD2 entry, similar to the
PLOAD entry, is limited to pressure acting normal to the element surface. The direction of the pressure load is
determined by the numbering sequence of the grid points on the connectivity entry (CQUAD4, CTRIA3, etc.).
PLOAD4
The PLOAD4 entry possesses the most general pressure definition. It can be used to apply pressures and/or tractions to
any of the two-dimensional elements and the surfaces of the 3-D elements. The PLOAD4 entry can be used for
everything that a PLOAD2 entry is used for. The direction of the pressure can be set by one of two methods. By default,
the direction of the pressure is determined by the element normal. For 2-D elements, the direction of positive pressure
is in the direction of the outward normal as is determined by applying the right-hand rule to the ordering sequence of
Main Index
Chapter 8: Load and Constraint Conditions 293
Distributed Loads
the grid points on the connectivity entry. For surfaces of solid elements, the direction of positive pressure is inward
toward the center of the element. The face of the solid to which you are applying the pressure is determined by
specifying the appropriate corner grid points.
Surface/volumetric loads, such as pressure, distributed heat flux, and body force, are distributed loads that are applied
to the surface (volume) of various elements (see Figure 8-5). A surface/volumetric load is characterized by the
distribution (uniform/nonuniform) and the magnitude of the load, as well as the surface to which the load is applied
(surface/volume identification). The total load applied to the surface (volume) is, therefore, dependent on the area
(interior) of the surface (volume).
P1 P2 P1 P2
4 3 4 3
Px
Q
Py
1 2 1 2
y y
P q
x x
(a) Distributed Mechanical Load (b) Distributed Heat Flux
Surface 2-3:
Uniform Normal Pressure p Uniform Heat flux q
Surface 3-4:
Nonuniform Normal Pressure p1 - p2 Nonuniform Heat flux q1 = q2
Whole Volume:
Volumetric Loads PxPy Volumetric Heat Flux Q
Equivalent nodal forces first must be calculated from surface/volumetric loads and then added to the nodal force vector.
MSC Nastran carries out this computation through numerical integration. The calculated equivalent nodal forces for
lower-order elements are the same as those obtained by equally lumping the uniformly distributed loads onto the nodes.
However, for high-order elements, the lumping is no longer simple. As a result, the surface/volumetric loads should
not be lumped arbitrarily.
Main Index
294 Nonlinear Users Guide
Distributed Loads
1/4
This consistent load is generated automatically inside MSC Nastran by using the PLOAD4 entry.
Main Index
Chapter 8: Load and Constraint Conditions 295
Distributed Loads
P/4 P/4
CPENTA
P/4 (6 Nodes) P/4
P/4 P/4
CHEXA
(8 Nodes)
CPENTA
P/4 (6 Nodes)
P/4
(a) (b)
.08333P .08333P
.3333P .3333P
.3333P .3333P
CPENTA .08333P
.08333P
(15 Nodes)
.08333P .08333P
CHEXA .3333P .3333P
(c) (d)
Figure 8-7 Consistent Loads Due to Uniform Loads on a Solid Element Face
Main Index
296 Nonlinear Users Guide
Inertial and Dynamic Loads
Centrifugal (RFORCE)
The RFORCE entry is used when you need to apply a force to your structure due to rotational velocity and/or
acceleration. On the RFORCE entry, you input the components of a spin vector that are used internally to compute
centrifugal forces. Each component of the spin vector is multiplied by the same scale factor.
You must select one of two methods for the internal calculation of the loading vector.
Method=1 yields correct results only when there is no coupling in the mass matrix. This occurs when the lumped mass
option is used with or without the ZOFFS option (see the CQUAD4 entry for a description of ZOFFS). Method=2 yields
correct results for lumped or consistent mass matrix only if the ZOFFS option is not used. The acceleration terms due
to the mass offset (X1, X2, X3) on the CONM2 entry are not computed with method=2. All the possible combinations
of mass matrices and offset and the correct method to be used are shown in Table 8-2.
No Offset Offset
Lumped Method=1 or Method=1
Method=2
Coupled Method=2 Neither
In addition, for problems with elements that have edge grid points (CQUAD8, CTRIA6, CTRlAX6, CHEXA, CPENTA,
and CTETRA), correct centrifugal loads are produced only if the parameter PARAM,COUPMASS,x (where x is greater
than 1), is included in the input file and Method 2 is used.
Note for PARAM,COUPMASS=-1 (the default) the generation of lumped mass matrices that contains only translational
components for the elements listed above. Notable exceptions to this are the CBAR and CBEAM elements, both of
Main Index
Chapter 8: Load and Constraint Conditions 297
Inertial and Dynamic Loads
which will yield rotational and coupling terms in order to preserve the mass center when element offsets are defined.
This offset mass is lumped in the sense that it has low matrix rank, and is coupled in the sense that there are nonzero
off diagonal terms in the mass matrix. The CBEAM element will also yield a mass moment of inertia about the local X
axis of the element, and if NASTRAN BARMASS > 0, then this is also true of the CBAR element.
In order to yield a lumped mass matrix containing translational components only for the CBAR and CBEAM elements,
set SYSTEM(414) = 1, along with the default value for PARAM,COUPMASS (-1). The default value (0) for
SYSTEM(414) produces the coupled mass matrices for CBAR and CBEAM.
Initial Stress and Initial Plastic Strain Mapping from Previous Results
MSC Nastran allows you to enter a set of initial stresses through the ISTRESS bulk data entry that simulates the stress
state in the structure at the beginning of an analysis. A typical example is prestress in a tensioned fabric roof. The set
of initial stresses must be self-equilibrating and should not exceed the yield stress of the material.
MSC Nastran allows you to define the equivalent plastic strain using the IPSTRAIN bulk data entry throughout the
model. This is useful in metal forming analysis in which the previous amount of equivalent plastic strain is often
required. This history dependent variable represents the amount of plastic deformation that the model was subjected
to, and is used in the work (strain) hardening model. This is only used to determine the value of the strain hardening
once plasticity occurs.
Preprocessors such as Patran and SimXpert are very useful in mapping stress states between analyses.
Main Index
298 Nonlinear Users Guide
Thermal Loads (TEMP and TEMPD)
Alternatively, temperature data can be specified on an element-by-element basis as shown in Table 8-3.
Table 8-3 Bulk Data Entries Used for Temperature Definition on Elements
Elements Temperature Data Bulk Data Entry
CROD, CONROD, CTUBE Average temperature at ends A and B. TEMPRB
CBAR, CBEAM, CBEND Average temperature and cross-sectional temperature TEMPRB
gradients at ends A and B.
CBEAM3 Temperature field and gradients along the beam TEMPB3
CQUAD4, CTRIA3, Average temperature and gradient in the thickness direction. TEMPP1
CQUAD8, CTRIA6
Average temperatures specified directly for an element take precedence over the temperatures interpolated from the
elements connected grid points. Solid elements obtain their temperatures only by interpolation from connected grid
points. Note that interpolated grid point temperatures provide temperature gradients over the neutral surface of shell
elements, whereas the TEMPPi entries do not.
The temperature data and the thermal expansion coefficients are used internally to calculate equivalent forces and
moments acting at the grid points.
The TEMPERATURE (Case) (INIT) and TEMPERATURE (Case) (LOAD) case control commands specify the initial
temperature and applied temperature, respectively. The TEMP(INIT) command must appear either above the first
subcase or inside the first subcase.
Main Index
Chapter 8: Load and Constraint Conditions 299
Export of Static Loads
Main Index
300 Nonlinear Users Guide
Export of Static Loads
Object Type
Displacement/Velocty/Acceleration
Nodal
Element Uniform
Element Variable
Force Nodal
Pressure Element Uniform
Element Variable
Main Index
Chapter 8: Load and Constraint Conditions 301
Export of Static Loads
Temperature Nodal
Element Uniform
Element Variable
Inertial Load Element Uniform
Initial Displacement Nodal
Initial Velocity Nodal
Distributed Load Element Uniform
Element Variable
CID Distributed Load Element Uniform
Element Variable
Total Load Element Uniform
Contact Element Uniform
Crack (VCCT) Nodal
Initial Plastic Strain Element Uniform
Initial Stress Element Uniform
Initial Temperature Nodal
The Loads and Boundary Conditions application controls which loads and boundaries and contact information will be
created in the MSC Nastran input file.
Main Index
302 Nonlinear Users Guide
Export of Static Loads
The Loads and Boundary Conditions application in Patran provides the ability to apply a variety of static and dynamic
loads and boundary conditions including contact surfaces to finite element models. Loads/BCs may be associated with
geometric entities as well as FEM entities. When associated with geometric entities, they can be transferred to finite
elements created on the geometry. Loads and boundary conditions are intended to be created in multiple single purpose
groups referred to as load sets. These sets are grouped into load cases in the Load Cases application.
One of the most elegant features in Patran is its ability to create fields that describes the variation of loads and boundary
conditions. The way in which Loads and BCs vary may be defined spatially, by previous analysis results, based on
time, or associated with material properties.
Sets can be visually displayed on the screen by markers which show the location, type, magnitude, and direction of the
applied loads or boundary condition. Only the static portion of a dynamic Loads/BCs set is reflected in the marker
display. Sets can also be displayed as tables.
A powerful capability is the display of any set scalar data directly on the model as a fringe plot. For display purposes,
data are treated as results, with full user control over the spectrum, method, shading, etc. Data display is scalar, but
the data can be pressures, vector component magnitudes, and vector resultant magnitudes. Fringe plots can only be
displayed on finite elements. Fringes of a dynamic Loads/BCs set may be displayed at user-specified times.
Main Index
Chapter 8: Load and Constraint Conditions 303
Export of Static Loads
Even if you do not create any load cases, your load and boundary conditions will still be placed into a default current
load case, named default. If you create a special load case and make it the current load case, then all subsequent LBCs
will be placed in that load case as long as it is current.
Main Index
304 Nonlinear Users Guide
Export of Static Loads
Displacement LBCs
Boundary conditions can be used to specify the value of the displacements at nodes. To create a boundary condition
for displacement, you need to specify the node number, the degree of freedom(s), and the magnitude of the
displacement.
Displacements can be imposed directly on nodes using SPC1 and SPCD bulk data entries. All non-blank entries will
cause an SPC1 entry to be created. If the specified value is not 0.0, an SCPD entry will also be created to define the
non zero enforced displacement or rotation.
Zero or nonzero displacements can also be applied across elements in a uniform or variable fashion. The primary use
of this boundary condition is to apply constraints to solid elements.
Main Index
Chapter 8: Load and Constraint Conditions 305
Export of Static Loads
Entry Description
Translations Defines the enforced translational displacement values. These are in model length units.
(T1,T2,T3)
Rotations (R1,R2,R3) Defines the enforced rotational displacement values. These are in radians.
Force LBCs
Concentrated forces and moments can be applied directly to nodes with the ability to define the direction as well as the
magnitude.
Forces and moments are specified with FORCEi and MOMENTi bulk data entries, where:
Entry Description
FORCE Defines a static concentrated force/moment at a grid point by specifying the magnitude and direction.
MOMENT
FORCE1 Defines a static concentrated force/moment at a grid point where the direction of the force/moment
MOMENT1 is defined to be parallel to a vector between two defined grid points.
FORCE2 Defines a static concentrated force/moment at a grid point where the direction of the force/moment
MOMENT2 is parallel to the cross product of vectors from G1 to G2 and G3 to G4.
Entry Description
Force (F1,F2,F3) Defines the applied forces in the translation degrees of freedom. This defines the N vector
and the F magnitude on the FORCE entry.
Moment (M1,M2,M3) Defines the applied moments in the rotational degrees of freedom. This defines the N
vector and the M magnitude on the MOMENT entry.
Pressure LBCs
Pressure loads can be applied to edges or surfaces of 2-D and 3-D elements. Several bulk data entries are used to apply
pressure loading depending on the element topology.
Main Index
306 Nonlinear Users Guide
Export of Static Loads
Entry Description
PLOAD Defines pressure loads on CQUAD4, CTRIA3, CHEXA, CPENTA, and CTETRA. Should not be used for
hyperelastic plane elements CQUAD4, CQUAD8, CQUAD, CTRIA3, and CTRIA6 or for hyperelastic
CHEXA, CPENTA, and CTETRA with midside nodes.
PLOAD2 Defines pressure loads on shell elements, CQUAD4 and CTRIA3.
PLOAD4 Defines pressure loads on surfaces of CHEXA, CPENTA, CTETRA, CTRIA3, and CQUAD4 elements.
PLOADX1 Defines pressure loads on axisymmetric elements CQUADX and CTRIAX.
These pressures are applied to 2-D and 3-D elements only. Pressures for 1-D elements are applied using the Total
Load LBCs object.
Entry Description
Top Surf Pressure Defines the top surface pressure load on shell elements using a PLOAD4 entry. The negative
of this value defines the P1 through P4 values. These values are all equal for a given element,
producing a uniform pressure field across that face.
Bot Surf Pressure Defines the bottom surface pressure load on shell elements using a PLOAD4 entry. This value
defines the P1 through P4 values.These values are all equal for a given element, producing a
uniform pressure field across that face.
Edge Pressure For Axisymmetric Solid elements (CTRIAX6), defines the P1 through P3 values on the
PLOADX1 entry where THETA on that entry is defined as zero. For other 2-D elements, this
will be interpreted as a load per unit length (i.e., independent of thickness) and converted into
equivalent nodal loads (FORCE entries). If a scalar field is referenced, it will be evaluated at
the middle of the application region.
Main Index
Chapter 8: Load and Constraint Conditions 307
Export of Static Loads
Entry Description
Pressure Defines the face pressure value on solid elements using a PLOAD4 entry. This defines the P1 through P4
values. If a scalar field is referenced, it will be evaluated once at the center of the applied region.
Entry Description
Top Surf Pressure Defines the top surface pressure load on shell elements using a PLOAD4 entry. The negative of
this value defines the P1 through P4 values. If a scalar field is referenced, it will be evaluated
separately for the P1 through P4 values.
Bot Surf Pressure Defines the bottom surface pressure load on shell elements using a PLOAD4 entry. This value
defines the P1 through P4 values. If a scalar field is referenced, it will be evaluated separately
for the P1 through P4 values.
Edge Pressure For axisymmetric solid elements (CTRIAX6), defines the P1 through P3 values on the
PLOADX1 entry where THETA on that entry is defined as zero. For other 2-D elements, this will
be interpreted as a load per unit length (e.g., independent of thickness) and converted into
equivalent nodal loads (FORCE entries). If a scalar field is referenced, it will be evaluated
independently at each node.
Entry Description
Pressure Defines the face pressure value on solid elements using a PLOAD4 entry. This defines the P1 through P4
values. If a scalar field is referenced, it will be evaluated separately for each of the P1 through P4 values.
Temperature LBCs
Temperatures can be defined directly at nodes or temperature fields can be defined across element surfaces.
Main Index
308 Nonlinear Users Guide
Export of Static Loads
Temperatures are specified with TEMP, TEMPPi, or TEMPRB bulk data entries, where:
Entry Description
TEMP Defines temperature at grid points.
TEMPP1i Defines temperature field for surface elements.
TEMPRB Defines temperature field for line elements.
Entry Description
Temperature Defines the T fields on the TEMP entry.
Entry Description
Temperature Defines a uniform temperature field using a TEMPRB entry. The temperature value is used for both
the TA and TB fields. The T1a, T1b, T2a, and T2b fields are all defined as 0.0.
Entry Description
Temperature Defines a uniform temperature field using a TEMPP1 entry. The temperature value is used for the T
field. The gradient through the thickness is defined to be 0.0.
Main Index
Chapter 8: Load and Constraint Conditions 309
Export of Static Loads
Entry Description
Centroid Temp Defines a variable temperature file using a TEMPRB entry. A field reference will be evaluated at
either end of the element to define the TA and TB fields.
Axis-1 Gradient Defines the temperature gradient in the 1 direction. A field reference will be evaluated at either
end of the element to define the T1a and T1b fields.
Axis-2 Gradient Defines the temperature gradient in the 2 direction. A field reference will be evaluated at either
end of the element to define the T2a and T2b fields.
Entry Description
Top Surf Temp Defines the temperature on the top surface of a shell element. The top and bottom values are used
to compute the average and gradient values on the TEMPP1 entry.
Bot Surf Temp Defines the temperature on the bottom surface of a shell element. The top and bottom values are
used to compute the average and gradient values on the TEMPP1 entry.
Entry Description
Temperature Defines the temperature or temperature distribution in the element.
Main Index
310 Nonlinear Users Guide
Export of Static Loads
Entry Description
GRAV Defines acceleration vectors for gravity or other acceleration loading.
RFORCE Defines load due to centrifugal force field.
The IDRF field on the RFORCE entry can be used to allow different portions of the structure to rotate with different
angular velocities or in different directions.
Entry Description
Trans Accel (A1,A2,A3) Defines the N vector and the G magnitude value on the GRAV entry.
Rot Velocity (w1,w2,w3) Defines the R vector and the A magnitude value on the RFORCE entry.
The acceleration and velocity vectors are defined with respect to the input analysis coordinate frame. The origin of the
rotational vectors is the origin of the analysis coordinate frame. In generating the GRAV and RFORCE entries, the
interface produces one GRAV and/or RFORCE entry image for each Patran load set.
Velocity LBCs
Velocities can be defined for transient analysis using the TLOAD entry.
:
Entry Description
TLOAD1 Defines a time-dependent dynamic load or enforced motion
Entry Description
Trans Veloc (v1,v2,v3) Defines the velocity values for the translational degrees of freedom.
Rot Veloc (w1,w2,w3) Defines the velocity values for the rotational degrees of freedom.
Main Index
Chapter 8: Load and Constraint Conditions 311
Export of Static Loads
Acceleration LBCs
Accelerations can be defined for transient response analysis using the TLOAD entry.
:
Entry Description
TLOAD1 Defines a time-dependent dynamic load or enforced motion
Entry Description
Trans Accel (A1,A2,A3) Defines the acceleration values for the translational degrees of freedom.
Rot Accel (a1,a2,a3) Defines the acceleration values for the rotational degrees of freedom.
Entry Description
PLOAD Defines a uniform static pressure load on a triangular or quadrilateral surface comprised of surface
elements and/or the faces of solid elements.
PLOAD1 Defines concentrated, uniformly distributed, or linearly distributed applied loads to the CBAR or
CBEAM elements at user-chosen points along the axis. For the CBEND element, only distributed loads
over an entire length may be defined
PLOAD2 Defines a uniform static pressure load applied to CQUAD4, CSHEAR, or CTRIA3 2-D elements.
PLOAD4 Defines a pressure load on a face of a CHEXA, CPENTA, CTETRA, CTRIA3, CTRIA6, CTRIAR,
CQUAD4, CQUAD8, or CQUADR element.
PLOADX1 Defines surface traction to be used with the CQUADX, CTRIAX, and CTRIAX6 axisymmetric element.
Main Index
312 Nonlinear Users Guide
Export of Static Loads
Defines distributed force or moment loading along beam elements using MSC Nastran PLOAD1 entries. The
coordinate system in which the load is applied is defined by the beam axis and the Bar Orientation element
property. The Bar Orientation must be defined before this Distributed Load can be created. If the Bar
Orientation is subsequently changed, the Distributed Load must be updated manually if necessary.
Entry Description
Distributed Load (f1,f2,f3) Defines the FXE, FYE, and FZE fields on three PLOAD1 entries.
Distributed Moment (m1,m2,m3) Defines the MXE, MYE, and MZE fields on three PLOAD1 entries.
For the element variable type, a field reference is evaluated at each end of the beam to define a linear load
variation.
2. Uniform and Variable Loads on 2-D Elements
Defines a distributed force or moment load along the edges of 2-D elements. The coordinate system for the load
is defined by the surface or element edge and normal. The x direction is along the edge. Positive x is determined
by the element corner node connectivity. See Patran Element Library (Ch. 15) in the Reference Manual - Part
III. For example, if the element is a CQUAD4, with node connectivity of 1, 2, 3, 4. The positive x directions for
each edge would be from nodes 1 to 2, 2 to 3, 3 to 4, and 4 to 1. The z direction is normal to the surface or
element. Positive z is in the direction of the element normal. The y direction is normal to x and z. Positive y is
determined by the cross product of the z and x axes and always points into the element. The MSC Nastran
entries generated, depend on the element type.
Entry Description
Edge Distributed Load For axisymmetric solid elements (CTRIAX6), the PA, PB, and THETA fields on
(f1,f2,f3) the PLOADX1 entry are defined. For other 2-D elements, the input vector is
interpreted as load per unit length and converted into equivalent nodal loads
(FORCE entries).
Edge Distributed Moment For 2-D shell elements, the input vector is interpreted as moment per unit length
(m1,m2,m3) and converted into equivalent nodal moments (MOMENT entries).
For the element variable type, a field reference is evaluated at each end of the beam to define a linear load
variation.
Main Index
Chapter 8: Load and Constraint Conditions 313
Export of Static Loads
Entry Description
PLOAD4 Defines a pressure load on a face of a CHEXA, CPENTA, CTETRA, CTRIA3, CTRIA6, CTRIAR,
CQUAD4, CQUAD8, or CQUADR element.
PLOADX1 Defines surface traction to be used with the CQUADX, CTRIAX, and CTRIAX6 axisymmetric element.
Note: Currently only 1-D element types are supported with this Object even though the form allows for other
types.
Main Index
314 Nonlinear Users Guide
Export of Static Loads
Initial Conditions
Initial conditions provides various ways of initializing the state variables throughout the model.
Entry Description
TIC Defines values for the initial conditions of variables used in structural transient analysis. Both displacement
and velocity values may be specified at independent degrees of freedom.
Entry Description
TIC Defines values for the initial conditions of variables used in structural transient analysis. Both displacement
and velocity values may be specified at independent degrees of freedom.
Main Index
Chapter 8: Load and Constraint Conditions 315
Export of Static Loads
Entry Description
Trans Veloc (v1,v2,v3) Defines the V0 fields for translational degrees of freedom on the TIC entry. A unique TIC
entry will be created for each non-blank entry.
Rot Veloc (w1,w2,w3) Defines the V0 fields for rotational degrees of freedom on the TIC entry. A unique TIC
entry will be created for each non-blank entry.
Main Index
316 Nonlinear Users Guide
Export of Static Loads
Main Index
Chapter 9: Contact
Nonlinear Users Guide
9 Contact
Introduction 319
Contact Methodology 321
Defining Deformable Contact Bodies 323
Rigid Contact Bodies 334
Patran Graphical User Interface for Contact 340
Movement of Contact Bodies 343
Contact Body Interaction 346
Controlling Contact Interactions 353
Time Step Control 367
Linear Contact 368
Contact Detection 374
Separation 388
Delayed Slide Off 391
Friction 392
Dynamic Impact 398
Accuracy 400
Thermal Contact 402
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid
Body Loading 407
Main Index
318 Nonlinear Users Guide
Numerical Procedures 419
Implementation of Constraints in Node-to-Segment 421
Segment-to-Segment Contact 425
Segment-to-Segment Contact Theory 428
References 443
Main Index
Chapter 9: Contact 319
Introduction
Introduction
The simulation of many physical problems requires the ability to model the contact phenomena. This includes analysis
of assembly modeling, interference fits, rubber seals, tires, crash, and manufacturing processes among others. The
analysis of contact behavior is complex because of the requirement to accurately track the motion of multiple
geometric bodies, and the motion due to the interaction of these bodies after contact occurs. This includes representing
the friction between surfaces and heat transfer between the bodies if required. The numerical objective is to detect the
motion of the bodies, apply a constraint to avoid penetration, and apply appropriate boundary conditions to simulate
the frictional behavior and heat transfer. Several procedures have been developed to treat these problems including the
use of Perturbed or Augmented Lagrangian methods, penalty methods, and direct constraints. Furthermore, contact
simulation has often required the use of special contact or gap elements. SOL 400 allows contact analysis to be
performed automatically without the use of special contact elements.
In structural problems, the contact often occurs between deformable bodies, which is called deformable-deformable
contact, and sometimes one of the bodies is significantly stiffer than the deformable bodies of interest and can be
modeled as rigid, which is called rigid-deformable contact. In the case of structural problems the contact characterizes
the stresses and forces transmitted between the contact bodies, and in the case of thermal analysis the contact
algorithms characterize the amount of heat that is transferred between the contact bodies. There is also a class of
problems called thermal-structural problems where both force and heat is transferred between the contact bodies.
Contact problems involve a variety of different geometric and kinematic situations. Some contact problems involve
small relative sliding between the contacting surfaces, while others involve large sliding. Some contact problems
involve contact over large areas, while others involve contact between discrete points. The general contact body
approach adopted by SOL 400 to model contact can be used to handle most contact problem definitions.
The contact body approach provides two formulations for modeling the interaction between surfaces of structures. One
formulation is a small-sliding formulation, in which the surfaces can only undergo small sliding relative to each other,
but may undergo arbitrary rotation. An example of this type of application is the classical Hertz contact problem. The
second formulation is a large-sliding formulation, where separation and sliding of large amplitude, and arbitrary
rotation of the surfaces, may arise. An example is the modeling of a rubber tire rolling on the ground.
This chapter describes the contact capabilities in MSC Nastran SOL 400 that may be used in solving nonlinear
structural and thermal analysis problems. First, by describing how we create the contact bodies in a finite element
model; then how we characterize the interaction between the bodies using either a contact table or the contact pair
capabilities. Setting up the details of contact problems involves a significant amount of information and detail, making
the use of a graphical user interface like Patran almost a requirement. After describing the basics required to set up a
contact problem, we will look at how Patran may be used to help set up the analysis model, including how to define
the contact bodies as well as how to input the data needed to characterize the contact interaction capabilities
(interference fits, friction, etc.) between the contact bodies. The latter part of this chapter is used to go through detailed
descriptions of the various contact capabilities the may be included in the simulation in order to provide the most
accurate characterization possible. These capabilities include: touching or glued contact, contact algorithms such as
node-to-segment or segment-to-segment, friction, small vs. finite sliding, over-closures and interference fits, delayed
slide-off, topology (such as including beam radius or shell thickness), near contact heat transfer coefficient, and
providing the results of the contact analysis including contact status, shear/normal forces and stresses. Often it is
advantageous to use the contact algorithms to glue components together and not let them separate. One of the most
useful capabilities of glued contact is to connect different parts of the structure that are not allowed to separate, but
have non-congruent meshes. This is sometimes referred to as assembly modeling. Immediately following the Patran
Main Index
320 Nonlinear Users Guide
Introduction
contact GUI section in this chapter is a brief demonstration problem illustrating how rigid-deformable contact can be
used to simulate the crushing of a circular pipe section. This problem provides insight as to the types of problems that
can be solved using the MSC Nastran SOL 400 contact capabilities.
Main Index
Chapter 9: Contact 321
Contact Methodology
Contact Methodology
In MSC Nastran, there are two methods used to implement contact constraints. Historically, the first method is known
as node-to-segment. In which case, a node makes contact with a rigid surface or an edge or face of a deformable body.
The second method is known as segment-to-segment contact. In which case, a fraction of an element edge or face may
contact either a rigid-to-surface or a fraction of an edge or face.
Main Index
322 Nonlinear Users Guide
Contact Methodology
of contact are the same when using these different methods, such as contact body definition. While other concepts,
such as separation, are treated differently.
You can select between node-to-segment and segment-to-segment using the METHOD keyword on the BCPARA bulk
data entry.
Main Index
Chapter 9: Contact 323
Defining Deformable Contact Bodies
Main Index
324 Nonlinear Users Guide
Defining Deformable Contact Bodies
In such a simple model, there are multiple ways to define the contact bodies, including:
Putting elements into groups or sets and associating these elements with a particular contact body.
Its unique property IDs are associated with each plate; then associates each property ID with a contact body.
If unique material IDs are associated with each plate, then associate each material ID with a contact body.
Let the GUI scan the complete model and associate each connected region with a contact body.
The last approach is very powerful when obtaining assembly models directly from CAD systems, where hundreds if
not thousands of parts (bodies) are present. See Automatic Contact Body Pair Creation User Interface.
Each node and element should be in, at most, one body. It is not necessary to identify the nodes on the exterior surfaces
as this is done automatically. The algorithm used is based on the fact that nodes on the boundary are on element edges
or faces that belong to only one element. Each node on the exterior surface is treated as a potential contact node. In
many problems, it is known that certain nodes never come into contact; in such cases, the BCHANGE bulk data entry
can be used to identify the relevant nodes. As all nodes on free surfaces are considered contact nodes, if there is an
error in the mesh generation such that internal holes or slits exist, undesirable results can occur.
Using this approach and applying a distributed load on the plates results in the following information.
Main Index
Chapter 9: Contact 325
Defining Deformable Contact Bodies
Contact Stresses
Deformations (Scaled)
You would define the contact interaction using BCTABLE which would define the search order, physical properties,
and numerical behavior such as the contact tolerance and penalty factors.
Main Index
326 Nonlinear Users Guide
Defining Deformable Contact Bodies
You would define the contact interaction using the BCTABL1 bulk data entry. This would give the contact table
identifier referenced by the BCONTACT case control command. Furthermore, it would point to one or more BCONECT
bulk data entries that defines pairs of bodies that come into contact and points to other options as follows:
BCONPRG defines the behavior of the contact interaction, such as the glue conditions
BCONPRP defines the physical properties associated with the contact interaction, such as the friction
coefficient and the thermal coefficients
Element Considerations
MSC Nastran allows contact with almost all of the available elements, but the use of certain elements has a
consequence on the analysis procedure. Contact analysis can be performed with all of the structural continuum
elements, either lower- or higher-order. Friction modeling is available in all of these elements. Higher-order
isoparametric elements use shape functions which, when the elements are loaded by a (for example) uniform pressure,
lead to equivalent nodal loads that oscillate between the corner and midside nodes. This has a detrimental effect on
determining contact separation.
Since the above mentioned oscillating nodal loads cannot be used for separation, the decision whether or not a node
should separate is based on the contact normal stress rather than the contact normal force.
In many manufacturing and rubber analyses, the lower-order elements behave better than the higher-order elements
because of their ability to represent the large distortion; hence, these lower-order elements are recommended.
The constraints imposed on the nodal degrees of freedom are dependent upon the type of element.
1. When a node of a continuum element comes into contact, the translational degrees of freedom are constrained.
Main Index
Chapter 9: Contact 327
Defining Deformable Contact Bodies
2. When a node of a shell element comes into contact, the translational degrees of freedom are constrained and no
constraint is places on the rotational degrees of freedom. The exception to this is when a shell contacts a
symmetry surface. In this case, the rotation about the element edge is also constrained.
Element Dimensionality
In MSC Nastran, a 2-D contact analysis means that plane strain, plane stress or axisymmetric elements are used and
rigid bodies are lines or curves only. Thus the problem itself is truly a 2-D problem in that everything is planar in the
XY plane. A 3-D contact problem can use solids, shells, and even beams. All of these elements are truly 3-D in that
the displacement results can be in any coordinate direction. That is what DIM means in the QRG description of
BCBODY1. All MSC Nastran cares about is whether the analysis is a 2-D analysis or a 3-D analysis. The analysis
DIMensionality needs to be defined correctly. You cannot mix 2-D and 3-D in the same analysis.
When defining the bodies in Patran, 2-D and 3-D have different meanings. 3-D means a 3-D element (solid). 2-D
means a 2-D element (shell or planar type element). 1-D means a 1-D element (beam or bar). So 1-D/2-D/3-D in Patran
when creating contact bodies is just a way of filtering the type of elements to be selectable when defining the contact
bodies. So if your contact body has solids and shells, you define the solids with the 3-D dimensionality filter, and if
they are shells, you define it with the 2-D dimensionality filter. But the analysis itself is still a 3-D analysis. A 2-D
analysis would not have both solids and shells. A 2-D analysis would only have plane strain, plane stress, or
axisymmetric elements.
The same is also true for rigid bodies. For defining them in Patran, you pick the dimensionality 1-D or 2-D depending
on whether it is a line/curve or surface. But the dimensionality of the analysis that is written for the MSC Nastran deck
is either 2-D or 3-D. For 2-D, only rigid lines/curves are allowed with plane strain/stress or axisymmetric elements.
For 3-D, either rigid lines/curve and/or surfaces are allowed.
Beam Elements
The node-to-segment procedure has limited capabilities to simulate beam contact. Contact of beams with non-beams
is established at the beam nodes without any allowance made for the beam dimensions. For simulating beam-beam
contact, an equivalent beam radius has to be defined by you, and multi-point constraints across the beam segments are
established based on a proximity check that includes the beam radii.
2-D Beams
All nodes on beams are potential contact nodes. Beam elements can be used in contact in two modes.
1. The 2-D beams can come into contact with rigid bodies composed of curves in the same x-y plane. The normal
is based upon the normal of the rigid surface.
2. The 2-D beams can come into contact with deformable bodies either of continuum elements or other beam
elements. As the beams are in two dimensions, they do not intersect one another.
3-D Beams
Three-dimensional beam elements can be used in contact in three methods.
1. The nodes of the beams can come into contact with rigid bodies composed of surfaces. The normal is based
upon the normal of the rigid surface.
Main Index
328 Nonlinear Users Guide
Defining Deformable Contact Bodies
2. The nodes of the beam elements can also come into contact with the faces of 3-D continuum elements or shell
elements. The normal is based upon the normal of the element face.
3. The 3-D beam elements can also come into contact with other beams (beam-to-beam contact).
The first two methods are activated by default if contact bodies consisting of beam elements are defined using the
BCBODY bulk data entry. The third method must be activated explicitly by additionally switching on the beam-to-
beam contact flag on the BCPARA bulk data entry.
In the beam-to-beam contact model, a beam element is viewed as a conical surface with a circular cross section. The
radius of the cross-section can vary linearly between the start and end node of a beam element. For each beam element
of a contact body, a contact radius must be entered via the BCBMRAD bulk data entry. The contact radius at a node
follows from the average contact radius of the elements sharing that node. Hence, the start and end node of an element
may have different contact radii.
Contact is detected between two beam elements if the associated conical surfaces touch each other; that is, if the
distance d between the closest points on the conical surfaces is smaller than the distance below which bodies are
considered touching each other. This is outlined in Figure 9-2, where beam elements and their contact body
representation are given. It should be emphasized that the contacting points are points on the conical surfaces and not
nodes of the finite element model.
X Z
Figure 9-2 Beam-to-Beam Contact: Finite Element Model (top) and Contact Body Representation
(bottom)
If two beam elements are in contact, a multipoint constraint equation (tying) is automatically imposed to ensure that
the conical surfaces will not penetrate. This constraint equation involves the displacements of the begin and end nodes
of both elements. The tied node in that equation is automatically selected, taking into account the location of the
contacting points with respect to the elements, any boundary conditions applied to the nodes and any contact between
the nodes and rigid surfaces or faces of continuum or shell elements.
During the iteration procedure, the contacting points of two beam elements can change if the elements slide with
respect to each other. In addition, the points in contact can move from one element to another. In that case, the nodes
involved in the multipoint constraint equations are automatically updated. During sliding, friction may be taken into
Main Index
Chapter 9: Contact 329
Defining Deformable Contact Bodies
account. Since for beam elements the normal stress in the contact points is not available, only Coulomb friction based
on nodal forces (either the arctangent or the bilinear model) is supported for beam-to-beam contact. The glue model,
which imposes that there is no relative tangential motion, is also available.
A limitation of the beam-to-beam contact model is that a contact body cannot contain branches; that is, every element
in the contact body must have a unique successor and predecessor.
By default, the top and bottom face of a shell element (which follow from the mid face by using the shell thickness and
the shell offset) are separately taken into account. This implies that a node may be found to be in contact based on the
top or the bottom face of an element. However, due to the nature of the shell formulation, it is not possible to
simultaneously apply contact conditions using the node-to-segment approach. Touching a shell element may also occur
at its top or bottom face using the segment-to-segment approach.
You have the option to set per pair of contact bodies on the BCTABLE/BCTABL1 bulk data entry as to how the geometry
of the shell elements should be handled. This means both the contacting and the contacted body can be defined and
that the contact description will be based on one of the following options:
both top and bottom (default),
top only (with or without the thickness offset) or
bottom only (with or without the thickness offset).
For glued contact, it is possible to use the top and bottom face, but without the thickness offset.
The thickness offset is one-half the shell thickness plus the user-defined shell offset.
The beam-to-beam contact model discussed above can also be used to model shell edge-to-edge contact. This requires
switching on the beam-to-beam contact flag on the BCPARA bulk data entry and, for a shell contact body combination,
the edge-to-edge contact flag on the BCTABLE/BCTABL1 bulk data entry. For edge-to-edge contact, half of the
thickness of the shell is used to set the contact radius.
Main Index
330 Nonlinear Users Guide
Defining Deformable Contact Bodies
Main Index
Chapter 9: Contact 331
Defining Deformable Contact Bodies
: nodal points
: auxiliary points
Figure 9-3 2-D Contact Bodies: Continuum Elements (Upper); Shell/beam Elements (Lower)
The SPLINE bulk data entry, as discussed before for node-to-segment contact, is available for segment-to-segment
contact. When this option is selected for a shell element or a 2-D beam element, then the transition from the top and
bottom contact segments to the segment at the free ends is automatically marked to have a normal vector discontinuity.
Main Index
332 Nonlinear Users Guide
Defining Deformable Contact Bodies
polyline/polygon points are monitored and once a threshold value d max is exceeded, new points will be created. The
default threshold value equals five times the contact tolerance , but it can also be defined via the BCPARA bulk data
entry.
For a 2-D configuration, the update process for large relative displacements is illustrated in Figure 9-4. The initial
position of the polyline points is shown on the left-hand side. The blue and green areas represent the common areas
between the upper and the left-hand and right-hand lower element. A possible relative displacement d is given in the
middle of Figure 9-4. As long as d d max , the contact simulation is based on the initial position of the polyline points.
If d d max , new polyline points are created, resulting in a situation as shown on the right-hand side. Redefining
polyline/polygon points will be done during an analysis as often as needed. After having created new polyline/polygon
points, important contact data (like the contact stresses) is mapped from the old on the new polyline/polygon points
and used as a starting point to continue the analysis.
Figure 9-4 Large Relative Displacements: Initial Position (left), Sliding (middle) and New Position
(Right)
The METHOD contact parameter on the BCPARA bulk data entry allows you to switch on the flag for small relative
displacements; in which case, there is no update of the position of the polyline/polygon points.
Main Index
Chapter 9: Contact 333
Defining Deformable Contact Bodies
This smoothed representation is used to calculate the contact normal and the distance between surfaces. An accurate
surface is especially important when calculating friction. It should be noted that activating this option has a minor
influence in the computational cost because of improved convergence. If higher-order elements are used, it is
recommended that mid node projection be activated (MIDNOD).
As an example in the interference problem show below, the two minor rings use a coarser mesh face than the outer
ring. While the segment-to-segment procedure would be more accurate, the figure on the left shows the equivalent
stress using the PWL representation; while the figure on the right, uses the analytical approach. Because a cubic spline
is used, the circular contact surface is represented exactly.
Main Index
334 Nonlinear Users Guide
Rigid Contact Bodies
Main Index
Chapter 9: Contact 335
Rigid Contact Bodies
2
1
3 2
3
1
Interior
Side
4
Interior Side
Figure 9-5 Orientation of Rigid Body Segments
For three-dimensional analysis, the interior side is formed by the right-hand rule along a patch. The interior side is
visualized in Patran as the side with markers; the exterior side is visualized in Patran as the side without markers.
It is not necessary for rigid bodies to define the complete body. Only the bounding surface needs to be specified. You
should take care, however, that the deforming body cannot slide out of the boundary curve in 2-D (Figure 9-6). This
means that it must always be possible to decompose the displacement increment into a component normal and a
component tangential to the rigid surface.
Incorrect Correct
Figure 9-6 Deformable Surface Sliding Out of Rigid Surface
Main Index
336 Nonlinear Users Guide
Rigid Contact Bodies
The variable ITYPE defines the type of surface entity to be used for a rigid surface. Since most of the three-
dimensional surfaces can be easily and adequately represented by a finite element mesh of 4-node plate (patch)
elements, the ITYPE = 7 option is a very convenient way of representing three-dimensional rigid surfaces. Both the
connectivities and the coordinates of the 4-node patches can be generated using Patran.
For the (PWL) approach, note that all geometrical data in 3-D space is reduced to 4-node patches. The four nodes will
probably not be on the same plane. The error in the approximation is determined by the number of subdivisions of the
defined surfaces. Note that the normal to a patch is defined by the right-hand rule, based on the sequence in which the
four points are entered.
Note: Patran produces a NURBS description for all 3-D rigid surfaces, even when patches or other geometrical
shapes are specified. If rigid bodies made of patches are desired, then the geometry should be meshed
and the elements specified as the application region.
Bezier Surface
When the Bezier Surface option is chosen, a Bezier surface is defined by the coordinates (x, y, z) of NPOINT1 x
NPOINT2 control points. NPOINT1 points are entered along the first direction and then repeated NPOINT2 times to
fill through the second direction of the surface. NPOINT1 and NPOINT2 have to be at least equal to 4. Number of
subdivisions (NDIV1, NDIV2) entered has to be equal or greater than NPOINT1 and NPOINT2 for Bezier surface.
(NPOINT1-1) x (NPOINT2-1) 4-node patches are created by SOL 400 for the definition of a Bezier surface. Figure 9-7
shows a typical Bezier surface. If it can be treated as an analytical surface, an exact conversion to NURBS is
performed.
Main Index
Chapter 9: Contact 337
Rigid Contact Bodies
r13
r23
r22 r33
2
r20
r01 r10
r30
z
1
y r00 NPOINT1 = 4
NDIV1 = 4
x 1: First direction
2: Second direction
Four-node Patch
When the Four-Node Patches option is chosen, you enter directly all the 4-node patches that comprise this surface.
They are entered following the same format SOL 400 would use to specify connectivities and coordinates of a mesh
of CQUAD4 elements. In this way, a finite element preprocessor can be used to create surfaces. Figure 9-8 shows a
typical 4-node patch surface. It cannot be used as an analytical surface.
Main Index
338 Nonlinear Users Guide
Rigid Contact Bodies
z
Number of patches = 12
12 Number of nodes = 20
y
Nodal coordinates can be entered
x using user subroutine DIGEOM
13
7
12 13
2 Rigid body 7 2
8
12
13
1 Rigid body
7 8
1: First direction
2: Second direction
Main Index
Chapter 9: Contact 339
Rigid Contact Bodies
+10
+8 +6
+9
+7 +5
+4
+2
+3
Z
Y X
+1
Main Index
340 Nonlinear Users Guide
Patran Graphical User Interface for Contact
The Application Region form for contact is used to select the contact bodies whether they be deformable or rigid.
Deformable contact bodies are always defined as a list of elements, the boundary of which defines the contact surface.
Rigid bodies are translated as ruled surfaces (2-D) or straight line segments (1-D) if a mesh or geometry with an
associated mesh is selected. Otherwise, if no mesh is associated with the selected geometry, the contact definition will
be written as geometric NURB surfaces during translation.
Deformable Body
Defining a deformable contact body requires the following data via the Input Properties subform on the
Loads/Boundary Conditions Application form.
Main Index
Chapter 9: Contact 341
Patran Graphical User Interface for Contact
Description
Friction Coefficient (MU) Coefficient of static friction for this contact body. For contact between two bodies with
different friction coefficients, the average value is used.
Heat Transfer Coefficient to Heat transfer coefficient (film) to environment. This is only necessary for coupled
Environment analysis.
Environment Sink Environment sink temperature. This is only necessary for coupled analysis.
Temperature
Contact Heat Transfer Contact heat transfer coefficient (film). This is only necessary for coupled analysis.
Coefficient
Boundary By default a deformable contact body boundary is defined by its elements (Discrete).
However, you can use an Analytic surface to represent the deformable body. This
improves the accuracy for deformable-deformable contact analysis by describing the
outer surface of a contact body by a spline (2-D) or Coons surface (3-D) description.
Exclusion Region This is an optional input. The Analytic surface of a deformable body can be described
by a spline (2-D) or Coons surface (3-D) and by default the entire outer surface will be
included unless an Exclusion Region is selected. For instance, you may not want to
represent locations of a body that never come in contact with the SPLINE option. Select
either Geometry entities of the contact body that have element associated to them, or
select individual FEM nodes along the outer surface. Care should be take when
selecting Exclusion Regions that actual outer surface or edge geometry is selected. If
nodes are being selected that describe a 3-D edge of a solid, the nodes must be in order
(it is safer to select a geometric entity in this case as the nodes could get
reordered incorrectly).
Rigid Body
Defining a rigid contact body requires the following data via the Input Properties subform on the Loads/Boundary
Conditions Application form. The input data form differs for 1-D and 2-D rigid bodies. One dimensional rigid surfaces
are defined as beam elements, or as curves (which may be meshed with beam elements prior to translation) and used
in 2-D problems. The lines or beams must be in the global X-Y plane. Two dimensional rigid surfaces must be defined
as Quad/4 or Tri/3 elements, or as surfaces (which may be meshed with Quad/4 or Tri/3 elements prior to translation)
and are used in 3-D problems. The elements will be translated as 4-node patches if meshed or as NURB surfaces if not
meshed.
Input Description
Flip Contact Side Upon defining each rigid body, Patran displays normal vectors or tic marks. These
should point inward to the rigid body. In other words, the side opposite the side with the
vectors is the side of contact. Generally, the vector points away from the body in which
it wants to contact. If it does not point inward, then use the modify option to turn this
toggle ON. The direction of the inward normal will be reversed.
Symmetry Plane This specifies that the surface or body is a symmetry plane. It is OFF by default.
Main Index
342 Nonlinear Users Guide
Patran Graphical User Interface for Contact
Input Description
Null Initial Motion This toggle is enabled only for Velocity and Position type of Motion Control. If it is ON,
the initial velocity, position, and angular velocity/rotation are set to zero in the
CONTACT option regardless of their settings here (for increment zero).
Motion Control Motion of rigid bodies can be controlled in a number of different ways: velocity,
position (displacement), or forces/moments.
Velocity (vector) For velocity controlled rigid bodies, define the X and Y velocity components for 2-D
problems or X, Y, and Z for 3-D problems.
Angular Velocity (rad/time) For velocity controlled rigid bodies, if the rigid body rotates, give its angular velocity
in radians per time (seconds usually) about the center of rotation (global Z axis for 2-D
problems) or axis of rotation (for 3-D problems).
Velocity vs Time Field If a rigid body velocity changes with time, its time definition may be defined through a
non-spatial field, which can then be selected via this widget. It will be scaled by the
vector definition of the velocity as defined in the Velocity widget. The Angular Velocity
will also be scaled by this time field.
Friction Coefficient (MU) Coefficient of static friction for this contact body. For contact between two bodies with
different friction coefficients, the average value is used.
Rotation Reference Point This is a point or node that defines the center of rotation of the rigid body. If left blank,
the rotation reference point will default to the origin.
Axis of Rotation For 2-D rigid surfaces in a 3-D problem, aside from the rotation reference point, if you
wish to define rotation, you must also specify the axis in the form of a vector.
First Control Node This is for Force or SPCD controlled rigid motion. It is the node to which the force or
SPCD is applied. A separate LBC must be defined for the force, but the application
node must also be specified here. If both force and moment are specified, they must use
different control nodes even if they are coincident. If only 1 control node is specified,
the rigid body will not be allowed to rotate.
Second Control Node This is for Moment controlled rigid motion. It is the node to which the moment is
applied. A separate LBC must be defined for the moment, but the application node must
also be specified here. It also acts as the rotation reference point. If both force and
moment are specified, they must use different control nodes even if they are coincident.
Note: After defining rigid bodies in your model, you can preview the rigid body motion by selecting Preview
Rigid Body Motion...
Main Index
Chapter 9: Contact 343
Movement of Contact Bodies
Associated with the rigid body is a point labeled the center of rotation. When the first two methods are chosen, you
define the translational motion of this point, and the angular motion about an axis through this point. The direction of
the axis can be defined for three-dimensional problems. For two-dimensional problems, it is a line normal to the plane.
For complex time-dependent behavior, the MOTION user subroutine can be used to prescribe the motion as an
alternative to the input. The motion during a time increment is considered to be linear. The position is determined by
an explicit, forward integration of the velocities based upon the current time step. A time increment must always be
defined even if a static, rate-independent analysis is performed.
When load controlled (a more accurate name would be control node rigid bodies) rigid bodies are used, two
additional nodes, called the control nodes, are associated with each rigid body. In 2-D problems, the first node has two
translational degrees of freedom (corresponding to the global x- and y-direction) and the second node has one
rotational degree of freedom (corresponding to the global z-direction). In 3-D problems, the first node has three
translational degrees of freedom (corresponding to global x-, y-, and z-direction) and the second node has three
rotational degrees of freedom (corresponding to the global x-, y-, and z-direction). In this way, both forces and
moments can be applied to a body for the control nodes. Alternatively, one may prescribe one or more degrees of
freedom of the control nodes by using the SPCD bulk data entries. Generally speaking, load-controlled bodies can be
considered as rigid bodies with three (in 2-D) or six (in 3-D) degrees of freedom. The prescribed position and
prescribed velocity methods (see Figure 9-10) have less computational costs than the prescribed load method (see
Figure 9-11), however it is possible to change the loads and constraints on the control node from one subcase to the
next to prescribe more complex motion of the rigid body.
2
Centroid
1 3
V
2
1
Figure 9-10 Velocity Controlled Rigid Surface
Main Index
344 Nonlinear Users Guide
Movement of Contact Bodies
Fy
Mz
Fx
Extra Node
If the second control node is not specified, the rotation of the body is prescribed to be zero.
Velocity Definition
For velocity control, on BCBODY or BCRIGID bulk data entry define:
CONTROL=0, velocity controlled rigid surface
ANGVEL, angular velocity w about the local axis R through the center of rotation CGID, in radians/time
DCOSi, direction cosines of local axis R
VELRBi, translational velocity V C GI D of the center of rotation CGID in direction i
Main Index
Chapter 9: Contact 345
Movement of Contact Bodies
As a demonstration of the movement of a position controlled rigid surface, see Closure of a Rubber Seal using
Segment-to-Segment Contact with Friction (Ch. 78) in the MSC Nastran Demonstration Problem Manual.
Initial Conditions
At the beginning of the analysis, bodies should either be separated from one another or in contact. Bodies should not
penetrate one another at the start of the analysis unless the objective is to perform an interference fit calculation. Rigid
body profiles are often complex, making it difficult for you to determine exactly where the first contact is located.
Unlike other MSC Nastran solution sequences, some calculations take place before a SOL 400 analysis begins. This
is defined as increment zero. During increment zero, if a rigid body has nonzero motion, the initialization procedure
brings it into first contact with a deformable body. No motion or distortion occurs in the deformable bodies during this
process. In a coupled thermal mechanical analysis, no heat transfer occurs during this process. If more than one rigid
body exists in the analysis, each one with a nonzero initial velocity is moved until it comes into contact. Because
increment zero is used to bring the rigid bodies into contact only, you should not prescribe any loads (distributed or
point) or prescribed displacements initially. For multistage contact analysis (often needed to simulate manufacturing
processes), the BCMOVE bulk data entry in conjunction with the BCTABLE bulk data entry allows you to model contact
bodies so that they just come into contact with the workpiece. This procedure is called a Body Approach subcase in
Patran.
Cautions
In static analysis, it is necessary to artificially connect (for instance, by very low stiffness springs) deformable bodies
that during an analysis might be completely separated from other deformable bodies and have no kinematic boundary
conditions applied.
Main Index
346 Nonlinear Users Guide
Contact Body Interaction
Note: The search order is NOT relevant when using the segment-to-segment contact approach.
The concept of MASTER and SLAVE bodies or nodes is also NOT relevant.
MSC Nastran has two input mechanisms for defining the contact interaction:
Main Index
Chapter 9: Contact 347
Contact Body Interaction
Whichever mechanism is used, the variable ISEARCH is used to control the search procedure
ISEARCH=0 (default) Double-sided search
First, the lower ID body is checked against the higher ID body for contact. If contact is found, contact
constraints are created.
Next the higher ID body is checked against the lower ID body and additional contact constraints are created
without conflicting with the existing constraints.
ISEARCH=1 Single-sided search
Search order is from slave to master
Slave and master are defined in the contact table
ISEARCH=2 Automatic
Search order is from the body with smaller average element edge size to body with larger average element
edge size. The search is single sided.
Automatic is ISEARCH=2
Single-sided is ISEARC=1
Double-sided is ISEARCH=0
Main Index
348 Nonlinear Users Guide
Contact Body Interaction
Deformable-to-Deformable Contact
Contact Constraint
In the case of node-to-segment contact, when a node is detected to be in contact with a segment a MPC is created. Note
that this is similar to an RBE3 constraint in the sense that the nodes on the segment are not constrained to move like a
rigid body. Furthermore, when used in SOL 400 in a large displacement simulation the large rotations are included
Main Index
Chapter 9: Contact 349
Contact Body Interaction
As one can observe, penetration may occur when different mesh density are used. For many problems, one simply does
not know which body has a finer mesh, in which case ISEARCH=2 should be used; in which case, for this example,
it will chose the finer mesh as the contacting body.
For some models, such as the closing of the seal shown below, because of large sliding, the contact may transition over
many mesh dimensions, an additional choice called Optimized contact is possible by setting ISEARCH=0 and
ISTYP=2. SOL 400 optimizes the constraints based upon stiffness and mesh densities. This is the recommended
method when using node-to-segment contact.
Main Index
350 Nonlinear Users Guide
Contact Body Interaction
0.0975
Using ISEARCH =0, it will search the top surface vs the bottom surface, and then search the bottom search vs the top
search based upon the body numbering scheme. It will not use any knowledge of the mesh density in the evaluation.
In this particular case, it does a poor job. What happens is shown below, the first pass is
Main Index
Chapter 9: Contact 351
Contact Body Interaction
In the first pass, the coarse lower body contacts the fine upper body and makes constraint equations. In the second pass,
the fine body attempts to make constraints with the coarse body; but, in fact, it cannot make a constraint with an already
tied node. So it results in a poor contact constraint.
In this case using ISEARCH=1 and changing the order of the contact bodies so the fine mesh is the 1st body results in
a better system.
0 is a retained node
1 is a tied node
This model illustrates an interesting behavior of the Automatic search detection procedure as well. The search order is
from the body with the smaller contact tolerance to the body with the larger contact tolerance. This contact tolerance
has two geometric components for shell elements, namely:
Main Index
352 Nonlinear Users Guide
Contact Body Interaction
Contact tolerance based element edge length Contact tolerance based upon property (thickness)
To overcome this problem, one can indicate that the thickness aspect should be neglected for ISEARCH=2.
For node-to-segment contact, using BCPARA,0,THKOFF,1 is recommended.
Note: When using Segment-to-Segment contact, the value of ISEARCH and the ordering of contact bodies is
immaterial.
Beginning with the 2013 release, an alternative method was introduced for defining contact interaction that utilizes the
BCTABL1, BCONECT, and BCONPRG bulk data entries. Utilizing this input style one would have used:
As a demonstration of the use of BCTABLE to define search order, see Double-sided Contact (Ch. 17) and Bolted Plates
(Ch. 23) in the MSC Nastran Demonstration Problem Manual.
Main Index
Chapter 9: Contact 353
Controlling Contact Interactions
Main Index
354 Nonlinear Users Guide
Controlling Contact Interactions
Cohesive Contact -- This is a special case of glued contact. In the modern industries, the product structure like
airplane, automobile, and so on, is becoming more and more complicated with many parts. The assembly process may
be done through a variety of processes such as rivets, bolts, spot weld, seam welds, or adhesives. In the numerical
simulations of the assemblies of many parts into a single structure, it is often too costly to model each one of these
discrete connectors, and the glued contact capability provides an effective and efficient way to simplify and reduce the
computation costs. While this method is easy to use, however, it often results in too stiff of a structure because
effectively the connection is rigid. To alleviate this, a flexible glued contact capability is available. With cohesive glued
contact, it is necessary to model the detailed connectors but provides the stiffness of the connectors in the glued contact.
Moment Carrying Glue by default, the contact constraint for glued contact with shells includes only translational
degrees of freedom. In other words, the moments were not carried across the contact interface. When a node of a shell
or beam element is glued to a load controlled rigid body or to the face of a shell or solid element, the rotation of the
shell or beam can be suppressed. This allows a true moment carrying glued connection. For the case that a connection
is made to the face of a solid element, the rotations of the touching node are connected to the translations of the nodes
of the contacted patch by a constraint relation based upon the large rotation RBE3 theory. Full moment carrying glue
is also supported for the following contact types: Shell-to-Shell, Shell-to-Solid, Beam-to-Shell, and Beam-to-Solid.
The moment carrying feature is optional and is activated through the BCTABLE/BCTABL1 bulk data entry.
Moment carrying glue is not supported for the following types of contact:
Beam-to-Beam
Shell Edge-to-Shell Edge (with BEAMB=1 on BCPARA)
Symmetry Contact deformable contact with a rigid Symmetry surface. In this case, no friction is allowed.
Furthermore, no separation is allowed, and finally, if the rigid surface is contacted by beams or shells, the rotations are
automatically constrained to satisfy the symmetry constraints. You do not need to specify any additional input, other
than specifying that the rigid surface is a SYMM body.
In SOL 400, one can have a multiple contact interaction types in the model meaning general contact, glued contact,
step contact on a contact pair basis. These can change from step to step by activating a new BCTABLE or BCTABL1.
In summary, the glue condition between bodies can be defined via BCTABLE or BCONPRG through the IGLUE
keyword as follows:
IGLUE keyword
0 no gluing
1 Activates the glue option. In the glue option, all degrees-of- freedom of the contact nodes are tied in case
of deformable-deformable contact once the node comes in contact. The relative tangential motion of a contact
node is zero in case of deformable-rigid contact. The node will be projected onto the contact body.
2 Activates a special glue option to insure that there is no relative tangential and normal displacement when
a node comes into contact. An existing initial gap or overlap between the node and the contacted body will not
be removed, as the node will not be projected onto the contacted body. To maintain an initial gap, ERROR
should be set to a value slightly larger than the physical gap.
3 Insures full moment carrying glue when shells contact. The node will be projected onto the contacted
body.
Main Index
Chapter 9: Contact 355
Controlling Contact Interactions
4 Insures full moment carrying glue when shells contact. The node will not be projected onto the contact
body and an existing initial gap or overlap between the node and the contacted body will not be removed, as
the node will not be projected onto the contacted body.
In SOLs 101 and 400, if contact is initially not true set NLGLUE on BCPARA to 1.
For SOL 400 with a mixture of glued and non-glued bodies, BCPARA,0,NLGLUE,1 must also be used
Note: The use of IGLUE=1 or IGLUE=3 can have negative repercussions because of the projecting of nodes
onto the surface. This may result in the loss of the rigid body modes, which effectively grounds the
structure. For large assembly models, it is recommended that either IGLUE=2 or 4 is used, or that
IGLUE=1 or 3 be used in conjunction with Stress Free Initial contact, ICOORD=1.
IGLUE=1 IGLUE=2
IGLUE=3 IGLUE=2
Main Index
356 Nonlinear Users Guide
Controlling Contact Interactions
If one considers an eigenvalue simulation of the model with PARAM, LGDISP activated so the calculation captures the
initial stress effects, using this technique will result in the six rigid body modes that one expects. If this is not done, the
imperfection in the mesh would induce a stress, which would influence the rigid body modes.
For Glued contact, the program forms a constraint between the slave node (A) and the master nodes B,C,D, and E based
upon the isoparametric projection of the node A on the nodes B,C,D,E face.
Main Index
Chapter 9: Contact 357
Controlling Contact Interactions
N ode
ve
E Sla
E(-1,1) D(1,1)
A
d
D
B
B(-1,-1) C(1,-1) Master
Surface
C
Dw A = 1 4 1 x A 1 h A Dw B + 1 4 1 + x A 1 h A Dw C + 1 4 1 + x A 1 + h A Dw D
+ 1 4 1 x A 1 + h A Dw E d
UNGLUE
With UNGLUE, you can select some nodes of the contact body for regular contact instead of glue contact even if the
contact table (BCTABLE) says that they should be glued. Those selected nodes will ignore any glue condition and
follow regular contact rules instead (normal constraint, no tangential constraint).
Breaking Glue
In engineering problems that involve delamination, it is often useful to indicate that two surfaces are glued together,
but may separate if a certain stress level is reached. The simplest is based upon a normal stress (preferred) or a force
condition. For problems like tape peeling, it is useful to include both the normal and shear stress condition. This can
be invoked by JGLUE on BCTABLE or BCONPRG:
0 glued contact nodes will stay in contact. Default.
1 to invoke the standard separation behavior
2 breaking glued with a breaking criterion
Main Index
358 Nonlinear Users Guide
Controlling Contact Interactions
t BG M n BGN
----------------
- + ------------------ 1.0
BGST BGSN
Main Index
Chapter 9: Contact 359
Controlling Contact Interactions
Deformable-Deformable Method In Double-Sided method, for each contact body pair, nodes of both bodies
will be checked for contact. In Single-Sided method, for each contact body
pair, only nodes of the lower-numbered body will be checked for contact.
Results are dependent upon the order in which contact bodies are defined.
Penetration Check This controls contact penetration checking. sometimes referred to as the
increment splitting option. Available options are:
Per Increment, Per means penetration is checked at the end of a load
Iteration (default), increment.
Suppressed (Fixed or means that penetration is checked at the end of
Adaptive) every iteration within an increment. If penetration is
detected, increments are split.
Per Increment suppress this feature for Fixed and Adaptive load
stepping types.
Reduce Printout of Surface Definition This controls reduction of printout of surface definition.
Note: Patran will only write out the contact table if something on the contact table form is changed. The
default is to write out BCONTACT=ALLBODY which does not require that the BCTABLE entry be
written out
Main Index
360 Nonlinear Users Guide
Controlling Contact Interactions
Input Description
Contact Detection Default (by body #) This is the default where contact is checked in the order the
bodies are written to the input deck. In this scenario, the
most finely meshed bodies should be listed first. There will
be contact checks first for nodes of the first body with
respect to the second body and then for nodes of the second
body with respect to the first body. If Single Sided contact
is activated on the Contact Parameters subform, then only
the first check is done
Automatic Unlike the default, the contact detection is automatically
determined and is not dependent on the order they are listed
but determined by ordering the bodies starting with those
having the smallest edge length. Then there will be only a
check on contact for nodes of the first body with respect to
the second body and not the other way around.
First ->Second Blanks the lower triangular section of the table matrix such
that no input can be accepted. Only the contact bodies from
the upper portion are written, which forces the contact
check of the first body with respect to the second body.
Second-> First Blanks the upper triangular section of the table matrix such
that no input can be accepted. Only the contact bodies from
the lower portion are written. Contact detection is done
opposite of First->Second.
Double-Sided Writes both upper and lower portions of the table matrix.
This overrules the Single Sided contact parameter set on the
Contact Parameters subform.
Touch All Places a T to indicate touching status for all deformable-deformable or rigid-
deformable bodies.
Glue All Places a G to indicate glued status for all deformable-deformable or rigid-
deformable bodies.
Deactivate All Blanks the spreadsheet cells.
Body Type Lists the body type for each body; either deformable or rigid.
Release This cell can be toggled for each body to Y or N (yes or no). If Y, this indicates
that the particular contact body is to be removed from this subcase. The forces
associated with this body can be removed immediately in the first increment or
gradually over the entire load step with the Force Removal switch described
below.
Main Index
Chapter 9: Contact 361
Controlling Contact Interactions
Input Description
Touching Body These are informational or convenience list boxes to allow you to see
Touched Body which bodies an active cell references and to see what settings are active for
Distance Tolerance and other related parameters below. You must click on the
touched/touching bodies to see what values, if any, have been set for the
pair combination.
Distance Tolerance Set the Distance Tolerance for this pair of contact bodies. You must press the
Enter or Return key to accept the data in this data box. A nonspatial field can be
referenced that will write this data in table format, if this parameter varies with
time, temperature, or some other independent variable. This overrides any other
settings for Distance Tolerance.
Separation Force Set the Separation Force for this pair of contact bodies. You must press the Enter
or Return key to accept the data in this data box. A nonspatial field can be
referenced that will write this data in table format, if this parameter varies with
time, temperature, or some other independent variable. This overrides any other
settings for Separation Force.
Friction Coefficient Set the Friction Coefficient for this pair of contact bodies. You must press the
Enter or Return key to accept the data in this data box. A nonspatial field can be
referenced that will write this data in table format, if this parameter varies with
time, temperature, or some other independent variable. This overrides any other
settings for Friction Coefficient.
Interference Closure Set the Interference Closure for this pair of contact bodies. You must press the
Enter or Return key to accept the data in this data box. A nonspatial field can be
referenced that will write this data in table format, if this parameter varies with
time, temperature, or some other independent variable. This overrides any other
settings for Interference Closure.
Heat Transfer Coefficient Set the Heat Transfer Coefficient for this pair of contact bodies. You must press
the Enter or Return key to accept the data in this data box. A nonspatial field can
be referenced that will write this data in table format, if this parameter varies
with time, temperature, or some other independent variable. This overrides any
other settings for Heat Transfer Coefficient. This is only used in coupled analysis.
Retain Gaps/Overlaps This is only applicable for the Glued option. Any initial gap or overlap between
the node and the contacted body will not be removed (otherwise, the node is
projected onto the body which is the default). For deformable-deformable
contact only.
Stress-free Initial Contact This is only applicable for initial contact in increment zero, where coordinates
of nodes in contact can be adapted such that they cause stress-free initial contact.
This is important if, due to inaccuracies during mesh generation, there is a small
gap/overlap between a node and the contacted element edge/face. For
deformable-deformable contact only.
Main Index
362 Nonlinear Users Guide
Controlling Contact Interactions
Input Description
Delayed Slide Off By default, at sharp corners, a node will slide off a contacted segment as soon as
it passes the corner by a distance greater than the contact error tolerance. This
extends this tangential tolerance. For deformable-deformable contact only.
Option Description
Method: Method to automatically create bodies:
Automatic Contact Body This method creates bodies based on continuous element connectivity of like elements.
Creation Based On: "Like elements" means all solid 3d elements (hex/tet), shell or 2d element (quad/tri), and
bar/beam (1-D) elements.
Property Sets
Group
Membership
Element
Topology
Based on
Connectivity
Main Index
Chapter 9: Contact 363
Controlling Contact Interactions
This is a simple utility to automatically create Contact Bodies and Contact Body Pairs based on some user criterion.
The utility is found under the Tools pull down menu and is a single utility form with both this automatic Contact Body
Pair creation and automatic Contact Body creation.
Main Index
364 Nonlinear Users Guide
Controlling Contact Interactions
Option Description
Create Options are Body Pair or Deformable Bodies. The Deformable Bodies selection is
described in the next section.
Distance Tolerance: This is the distance between body boundaries, below which two bodies are considered
within distance to create a Body Pair.
All Bodies If this switch is set to All Bodies, then all bodies, both deformable-rigid and deformable-
deformable body pairs will be created.
Deformable Only If this switch is set to Deformable Only, then only deformable-deformable body pairs will
be created.
Create From: The options are to consider bodies in the Entire Model, only those in the Current Viewport,
or by selecting existing bodies manually. If the Selected Bodies option is selected, the List
Box icon appears and the user can select this to get a list of bodies. Only those bodies
based on the All Bodies or Deformable Only switch setting will be listed in the List Box. A
refresh function is included so the list box contents can be dynamically updated if the user
creates bodies while this form is opened. The list box allows for filtering.
Apply Creates the body pairs and issues the PCL command to do so.
Defaults Resets the form to the defaults.
Cancel Closes the form without creating the body pairs.
Main Index
Chapter 9: Contact 365
Controlling Contact Interactions
When the form is originally opened or if being opened and not specifically associated with an existing subcase then
Patran will detect the presence of contact body pair LBCs. If they are found, then the toggle is set to OFF and the Use
Contact Table button is disabled. If no body pair LBCs are present, then the toggle is ON and the button enabled.
Note that this is only valid if the MSC Nastran version is 2013 or higher. If less than 2013, then the toggle to Use
Contact Table will not be visible and the Contact Table must be used. If this is the case, any body pairs in the database
are ignored.
Main Index
366 Nonlinear Users Guide
Controlling Contact Interactions
When the job is saved, a database parameter Use Contact Table=YES/NO is saved.
When initial contact is ON and the user is using body pair LBCs, BCTABL1 with ID 0 is written.
Note that this scenario is only valid if the MSC Nastran version is 2013 or higher. If less than 2013, then the Use
Contact Table toggle is not visible and the Contact Table must be used. In this case, body pairs in the database are
ignored.
For an example of using BCTABLE to define contact interaction, see Tube Flaring (Ch. 15) in the MSC Nastran
Demonstration Problem Manual.
For an example of using BCTABL1 to define contact interaction, see Ball Joint Rubber Boot (Ch. 13) in the MSC
Nastran Demonstration Problem Manual.
Main Index
Chapter 9: Contact 367
Time Step Control
Main Index
368 Nonlinear Users Guide
Linear Contact
Linear Contact
SOL 101 may also be used to preform linear analysis. You can reduce the computational time by using LCPERF or
LCACCU keyword on the NLSTEP bulk data entry. This entry also allows you to customize values of specific control
parameters through the new LCNT keyword.
Example (tpl/3dcnt101/rg_lcdf.dat)
This model consists of a pin as the contact body ID 1 and a clevis as the contact body ID 2. The pin is the master body
and has the SPLINE option on (BCBODY IDSPL). The contact table was set up to use single sided contact
(ISEARCH=1). Stress free initial contact is enabled.
To specify the Performance preference, define NLSTEP case control command and NLSTEP bulk data entry as follows.
NASTRAN SYSTEM(316)=19
$
SOL 101
CEND
$
TITLE = THIS IS A DEFAULT SUBCASE.
ECHO = NONE
BCONTACT = 0
SUBCASE 1
TITLE=This is a default subcase.
Main Index
Chapter 9: Contact 369
Linear Contact
SUBTITLE=Linear_contact
NLSTEP = 1
BCONTACT = 1
SPC = 2
LOAD = 2
DISPLACEMENT(PLOT,SORT1,REAL)=ALL
SPCFORCES(PLOT,SORT1,REAL)=ALL
STRESS(PLOT,SORT1,REAL,VONMISES,BILIN)=ALL
BOUTPUT(PLOT,SORT1,REAL)=ALL
$
BEGIN BULK
$
PARAM PRTMAXIM YES
NLSTEP,1,,LCPERF
:
Similarly, the Accuracy preference can be specified by replacing the following NLSTEP bulk data entry
NLSTEP,1,,LCPERF
with
NLSTEP,1,,LCACCU
The von Mises stress results are shown as follows.
Running on a Linux LX8664 machine, the Performance preference shows 7.3 times improvement in CPU time and 8.2
times improvement in convergence comparing with the baseline model that uses original default values of control
parameters. The Accuracy preference shows 5.5 times improvement in CPU time and 5.5 times improvement in
convergence. The following table lists detail comparisons of these three models.
Main Index
370 Nonlinear Users Guide
Linear Contact
By selecting LCPERF or LCACCU, one effectively modifies the convergence control variables as shown in the
following table.
Main Index
Chapter 9: Contact 371
Linear Contact
Main Index
372 Nonlinear Users Guide
Linear Contact
Main Index
Chapter 9: Contact 373
Linear Contact
This leads to the conclusion that neglecting the effects of nonlinearity can lead to significant errors.
Main Index
374 Nonlinear Users Guide
Contact Detection
Contact Detection
During the incremental procedure, each potential contact node is first checked to see whether it is near a contact
segment. The contact segments are either edges of other 2-D deformable bodies, faces of 3-D deformable bodies, or
segments from rigid bodies. By default, each node could contact any other segment including segments on the body
that it belongs to. This allows a body to contact itself. To simplify the computation, it is possible to use the BCTABLE
bulk data entry to indicate that a particular body will or will not contact another body. This is often used to indicate
that a body will not contact itself. During the iteration process, the motion of the node is checked to see whether it has
penetrated a surface by determining whether it has crossed a segment.
Because there can be a large number of nodes and segments, efficient algorithms have been developed to expedite this
process. A bounding box algorithm is used so that it is quickly determined whether a node is near a segment. If the
node falls within the bounding box, more sophisticated techniques are used to determine the exact status of the node.
During the contact process, it is unlikely that a node exactly contacts the surface. For this reason, a contact tolerance
is associated with each surface.
The contact distance is used to determine the how geometrically close a node needs to be to a segment. This is called
the contact tolerance or the ERROR distance.
ct
nta
Co
ct
Slave nta
Body Co
Master
Body
By default, a node slides off the boundary of a deformable body at a sharp corner if the distance is greater than the
contact tolerance, the contact between the node and the contacted body is lost. An example is two cubes which are
slightly misaligned, where 99+% of the surface is in contact, but the node is not in contact. By invoking the delayed
sliding off option, the tangential contact tolerance is increased by a factor of 10. Note this dilemma does not occur with
segment-to-segment contact.
Main Index
Chapter 9: Contact 375
Contact Detection
Patran Interface
The contact tolerance can be specified either on a global basis or on a contact pair basis. The GUI interface is shown
as follows:
While the default values may be used successfully for most problems, if one has inside knowledge than it is much better
to utilize this on a contact pair basis. A good compromise is use the defaults based upon each contact pair. It will then
use the smaller value associated with each pair, by activating BCPARA ERRBAS=1. This is the recommended
procedure.
It is very useful to know what the program is using for contact tolerances. This can be obtained by using the case
control command NLOPRM NLDBG=N3DBAS.
Main Index
376 Nonlinear Users Guide
Contact Detection
When using Contact Table or Contact Pair, one can obtain additional information regarding the contact tolerance using
the case control command NLOPRM NLDBG=N3DMED.
rance
2 x Tole
If a node is within the contact tolerance, it is considered to be in contact with the segment. The contact tolerance is
calculated by the program as the smaller of 5% of the smallest element side or 25% of the smallest (beam or shell)
element thickness. It is also possible for you to define the contact tolerance through the input.
t trial trial
During an increment, if node A moves from A to A t + t , where A t + t is beyond the contact
tolerance, the node is considered to have penetrated. In such a case, either the increment is divided into subincrements
or the increment is reduced in size.
Main Index
Chapter 9: Contact 377
Contact Detection
A(t)
Atrial (t + Dt)
Figure 9-17 Trial Displacement with Penetration
The size of the contact tolerance has a significant impact on the computational costs and the accuracy of the solution.
If the contact tolerance is too small, detection of contact is difficult, leading to higher costs. Also many nodes are more
likely to be considered penetrating leading to reduction of the time step; therefore, increasing the computational costs.
If the contact tolerance is too large, nodes are considered in contact prematurely, resulting in a loss of accuracy.
Furthermore, nodes might penetrate the surface by a large amount.
An effective compromise is to bias the tolerance area so that a smaller distance is on the outside surface than on the
inside surface. This is done by entering a bias factor. The bias factor should have a value between 0.0 and 1.0. The
default in SOL 400 is 0.9. This results in good accuracy and reasonable computational costs. In analyses involving
frictional contact, a bias (recommended value: 0.95 - 0.99) to the contact core is also found beneficial to
facilitate convergence. When using glued contact, the default value is 0.0.
In some instances, you might wish to influence the decision regarding the deformable segment a node contacts (or does
not contact). This can be done using the EXCLUDE variable on the BCBODY bulk data entry.
(1 - Bias)* tolerance
(1 + Bias)* tolerance
Shell Contact
A node on a shell makes contact when the position of the node plus or minus half the thickness projected with the
normal comes into contact with another segment. In 2-D, this can be shown as:
x1 = A + n t 2
x2 = A n t 2
Main Index
378 Nonlinear Users Guide
Contact Detection
Shell
Midsurface
2
x A t
rance
2x tole
x
Figure 9-19 Default Shell Contact
If point x or y falls within the contact tolerance distance of segment S, node A is considered in contact with the segment
S. Here x 1 and x 2 are the position vectors of a point on the surfaces 1 and 2 on the shell, A is the position vector of a
point (node in a discretized model) on the midsurface of the shell, n is the normal to the midsurface, and t is the shell
thickness.
As the shell has finite thickness, the node (depending on the direction of motion) can physically contact either the top
surface, bottom surface, or mathematically contact can be based upon the midsurface. You can control whether
detection occurs with either both surfaces, the top surface, the bottom surface, or the middle surface. In such cases,
either two or one segment will be created at the appropriate physical location. Note that these segments will be
dependent, not only on the motion of the shell, but also the current shell thickness.
S1
n n S1
2
S2
2
1
1
Include Both Segments Top Segment Only
2
n
S1
S2 2
1
1
Bottom Segments Only Ignore Shell Thickness
Figure 9-20 Selective Shell Contact
Main Index
Chapter 9: Contact 379
Contact Detection
Interference Fit
This feature can be used to simulate situations where bodies with overlaps or gaps in the geometry are required to fit
together. Some examples of interference fit usage include spur gears that are mounted tightly on a shaft, couplings,
collar rings, tire mounted on wheel rim, bearing bushings in hubs, valve seats, rubber seals, bushings made of synthetic
resin pressed material, etc.
The interference fit capability is treated in MSC Nastran using the contact algorithm. The penetration associated with
interference fit can be much larger compared to any contact analysis. Ideally, interference fit computations are
performed in a separate load case, which might take more than a one increment. The user can control the number of
increments or time required to accomplish the fit. The user is equipped with many methods which can be adopted
depending on the geometry.
There are four methods available in MSC Nastran for the interference fit option. These are applicable to both Node-
to-Segment Contact (N2S) and Segment-to-Segment Contact (S2S). These methods are as follows:
1. Contact normal
2. Translation
3. Scaling
4. Automatic
1. Contact Normal
Main Index
380 Nonlinear Users Guide
Contact Detection
This method is generally recommended for situations with small values of interference to be resolved along the
normal direction to the touched interface. The other methods are more suitable when the interference is large.
In this case, the nodes (N2S) or auxiliary points (S2S)of the touching body are projected in a direction normal
to the segments of the touched body. The individual touching entities are positioned at a certain distance (over-
closure or gap) from the touched surface along the normal of the touched surface. The user needs to provide
this interference distance which signifies the maximum overlap or gap between the bodies. This provided value
is specified as a negative value for an overlap and as a positive value for a gap. The detection algorithm works
as follows:
(Overlap/Gap + Interference) at a node (N2S) or auxiliary point (S2S) < D (1+B) on the inside or < D (1-B) on
the outside
Where D = regular distance tolerance and B = bias factor.
To control the number of increments or time during which the interference fit is achieved, the user can
optionally specify a table in which the magnitude of interference distance is ramped down from the possible
maximum value of interference to zero. Additionally, the user can specify the project stress-free option to
correct the existing geometry to make the interference uniform across all the touching nodes. Using this option,
interference can be gradually resolved as shown in Figure 9-21.
Main Index
Chapter 9: Contact 381
Contact Detection
Methods 2 to 4 are more general and allow larger amounts of interference (overlap or gap) between the touching and
touched bodies. The general scheme for these methods is to internally compute an initial shift vector between the nodes
of the touching body and the segments of the touched body. This initial vector is also referred to as a pseudo
displacement vector. The following points should be noted about the pseudo displacement vector:
The magnitude of the pseudo movement is varied based on a table provided by the user, which is normally a
ramp down table from unity to zero over the loadcase. Note that the projection of the touching entity to a
touched segment is done for each iteration but the position of the touching entity is always based on a scaling of
the pseudo displacement vector computed at the start of the loadcase. Therefore, large rotations of the
interfering bodies should not be permitted during the interference loadcase.
The treatment of the interference across two successive loadcases should be noted. Note that the interference
can be specified for any loadcase, though normally, it is specified at the start of the analysis.
Each interference method provides a different scheme for the computation of this pseudo vector.
2. Translation
In this method, for each interfering body pair in the contact table, the user specifies the following:
Magnitude of the vector
Direction cosines of the vector
The coordinate system in which the direction cosines are specified (default is the global Cartesian system)
The body to which to apply the interference vector and
A table specifying the variation of the interference vector during the loadcase
The algorithm performs an initial pseudo movement of the interfering body along the direction specified by the
user such that the overclosure is removed. This is followed by a removal of the pseudo movement such that the
interference fit is achieved. This method is especially useful for large initial over-closure between contacting
bodies in a specific direction.
3. Scaling
In this method, for each interfering body pair in the contact table, the user specifies the following:
Position of the Centroid about which the scaling is to occur
X, Y and Z Scaling Factors
The coordinate system in which the scaling should occur (default is the global Cartesian system)
The body to which to apply the interference vector and
A table specifying the variation of the interference resolution during the loadcase
The pseudo adjustment is done by scaling one of the bodies such that the overlap is removed. The rest of the
algorithm is similar to that of the Translation option.
4. Automatic
In this method, for each interfering body pair in the contact table, the user specifies the following:
A penetration tolerance
The body to which to apply the interference vector and
A table specifying the variation of the interference resolution during the loadcase
Main Index
382 Nonlinear Users Guide
Contact Detection
The overlap distance vectors are found between the nodes of the overlapping body (specified by the user) into
the other body. The penetration tolerance should be a little more than the highest penetration between contact
body pair. Nodes within this tolerance are considered for the interference calculation. The pseudo displacement
vector at each node is established by taking the perpendicular distance between the node and the closest touched
segment. The rest of the algorithm is similar to that of the Translation option. The following additional points
should be noted for the AUTOMATIC method:
In single sided contact, touching body has to be selected as the interfering body.
In general, use of analytical boundary description for the touched body is recommended while using this
method.
For shell contact bodies, user needs to specify the top or bottom surface for detecting contact.
Limitations of Interference Fit
Cannot be used for self contact.
Cannot be used with beam-to-beam contact.
In Methods 2 to 4, large rotations should not be present during the interference loadcase.
Interference is not supported for non-mechanical analyses.
Clearance
Note that if a clearance or overlap needs to be maintained for the entire job, then the interference value needs to be
maintained as a constant value in all the loadcases of the job. This method may not perform adequately when the
interference distance is large in this case, the touching entities may find multiple touched segments within the given
closures distance.
Neighbor Relations
When a node is in contact with a rigid surface, it tends to slide from one segment to another. In 2-D, the segments are
always continuous and so are the segment numbers. Hence, a node in contact with segment n slides to segment n 1
or to segment n + 1 . This simplifies the implementation of contact.
n-1
n
n+1
In 3-D, the segments are often discontinuous. This can be due to the subdivision of matching surfaces or, more likely,
the CAD definition of the under lying surface geometry.
Main Index
Chapter 9: Contact 383
Contact Detection
Nonmatching
Segments
Continuous surface geometry is highly advantageous as a node can slide from one segment to the next with no
interference (assuming the corner conditions are satisfied). Discontinuous surface geometry results in additional
operations when a node slides off a patch and cannot find an adjacent segment. Hence, it is advantageous to use
geometry clean-up tools to eliminate small sliver surfaces and make the surfaces both physically continuous and
topologically contiguous.
Shell Edge-to-Face
As an example, two shell meshes are connected edge-to-edge. The contact bodies are located in the same plane with
the same thickness. The large plate (yellow), coarse mesh is body 1, the small plate(green), fine mesh is body 2.
Main Index
384 Nonlinear Users Guide
Contact Detection
Notice the checking order is flipped so fine mesh will contact coarse mesh.
Main Index
Chapter 9: Contact 385
Contact Detection
Turn on moment carrying capability to enable the glued joint to transfer moments
The resultant MSC Nastran file looks like:
Main Index
386 Nonlinear Users Guide
Contact Detection
When using this configuration, the contacting body (slave) should be the one containing the shell edge.
To model these problems, turn on moment carrying glue. When using the node-to-segment approach, turn on Ignore
Thickness. When using segment-to-segment, leave on the thickness.
For the case of shell node at t/2 away, no special modeling technique is required. Contact is found between the nodes
on the edge and the fact that it is t/2 off the master surface. Recall that shell offsets are also taken into consideration.
For the case where the mesh is such that one edge is in geometric contact with the midplane of the other face, then one
must ignore the shell thickness. Note, which this is common engineering practice, it is not actually the best practice.
The challenge with edge to face contact is in the case where the slave shell is thick and the elements on the elements
on the master surface are smaller than the shell thickness. In this case, the moment carrying connection is not correct
using the node-to-segment technique. The Segment-to-Segment technique must be used.
As an example of this, consider the thick beam (shown as follows) where the left is modeled with tetrahedral elements
and the right with shell elements, where the thickness is the same as the left. Using the node-to segment approach, the
constraints would be between the nodes on the edge of the shell and only the nodes on the faces that it contacts. While
with segment-to-segment contact, the constraint would be between all nodes on the tetrahedral surface and the shell
nodes, and the moment transfer would be correct as illustrated in the following figure.
Main Index
Chapter 9: Contact 387
Contact Detection
Because of the restrictions on multipoint constraints and the fact that there is only one node through the thickness using
node-to-segment, contact is not possible with this self-contacting spring shown below. This is achievable using
segment-to-segment contact.
Main Index
388 Nonlinear Users Guide
Separation
Separation
After a node comes into contact with a surface, it is possible for it to separate in a subsequent iteration or increment.
Mathematically, a node should separate when the reaction force between the node and surface becomes tensile or
positive. Physically, you could consider that a node should separate when the tensile force or normal stress exceeds the
surface tension. Rather than use an exact mathematical definition, you can enter the force or stress required to cause
separation.
Main Index
Chapter 9: Contact 389
Separation
4 separation if the contact pulling stress exceeds FNTOL (default=0.1) times the maximum contact stress in
the model.
For higher-order elements given a stress distribution on a face, the effective force is shown below. This leads to
problems when using a force-based separation criteria. For these elements, a stress-based criteria must be used.
MSC Nastran calculates this value as the maximum residual force in the structure. Note that the maximum residual
force can be specified in the BCPARA bulk data entry. Default for this value, a 10 percent of the maximum reaction
force is used. Consequently, if locally high reaction forces at a particular point are present, the separation force is large
as well. In most cases, however, the default value is a good measure.
If you indicate that separation is to be based upon stresses, a value of the separation stress should be entered. The
default value is the maximum residual force at a node divided by the contact area of node n.
For NTS contact, a zero value of the absolute separation stress can increase the computational effort considerably, for
STS contact this is not the case.
Equilibrium When contact occurs, a reaction force associated with the node in contact balances the internal stress
of the elements adjacent to this node. When separation occurs, this reaction force behaves as a residual force (as the
force on a free node should be zero). This requires that the internal stresses in the deformable body be redistributed.
Depending on the magnitude of the force, this might require several iterations.
You should note that in static analysis, if a deformable body is constrained only by other bodies (no explicit boundary
conditions) and the body subsequently separates from all other bodies, it would then have rigid body motion. For static
analysis, this would result in a singular or nonpositive definite system. This problem can be avoided by appropriate
boundary conditions.
Release
A special case of separation is the intentional release of all nodes from a rigid body. This is often used in manufacturing
analysis to simulate the removal of the workpiece from the tools. After the release occurs in such an analysis, there
might be a large redistribution of the loads. It is possible to gradually reduce the residual force to zero, which improves
the stability, and reduces the number of iterations required. The BCMOVE bulk data entry allows the release
Main Index
390 Nonlinear Users Guide
Separation
(separation) of all the nodes in contact with a particular surface at the beginning of the increment. The rigid body
should be moved away using the BCMOVE bulk data entry or deactivated using the BCTABLE entry to ensure that the
nodes do not inadvertently recontact the surface they were released from.
Maximum Separations Maximum number of separations allowed in each increment. Default is 9999.
Retain Value on NCYCLE Turn ON this button if you do not want to reset NCYCLE to zero when separation
occurs. This speeds up the solution, but might result in instabilities. You can not set
this and Suppress Bounding Box simultaneously.
Increment Specifies whether chattering is allowed or not.
Chattering Specifies the separation criterion (forces or stresses) and the critical value at which the
separation will take place.
Separation Criterion Specifies in which increment (current or next) the separation is allowed to occur.
Force Value Specifies the force or separation for which separation occurs.
Stress Value
As a demonstration of separation based upon forces, see Cup Forming Simulation (Ch. 16) in the MSC Nastran
Demonstration Problem Manual.
As a demonstration of separation based upon stresses, see Friction Between Belt and Pulley (Ch. 24) in the MSC
Nastran Demonstration Problem Manual.
Main Index
Chapter 9: Contact 391
Delayed Slide Off
When using node-to-segment, a node separates if the node goes around a sharp corner. Sliding off a sharp corner may
lead to convergence problems or undesired solutions. Delayed slide off option increases the size of the contact patches
at corners/edges. As shown in the following figure, the node separates when it reaches a valued that is equal to the
error tolerance. Using delayed slide off, one defines the value e t , and the node will not separate until this value is
reached. This useful to ensure initial contact as typically found in assembly modeling problems.
Main Index
392 Nonlinear Users Guide
Friction
Friction
Friction is a complex physical phenomena that involves the characteristics of the surface such as surface roughness,
temperature, normal stress, and relative velocity. The actual physics of friction continues to be a topic of research.
Hence, the numerical modeling of the friction has been simplified to the following idealistic models.
Friction Type
(FTYPE) Description
0 No friction
6 Bilinear Coulomb
7 Bilinear Shear
The most popular friction model is the Adhesive Friction or Coulomb Friction model. This model is used for most
applications with the exception of bulk forming such as forging. The Coulomb model is:
fr n t
where
vr
t = --------
vr
f t f n t
where
Main Index
Chapter 9: Contact 393
Friction
Quite often in contact problems neutral lines develop. This means that along a contact surface, the material flows in
one direction in part of the surface and in the opposite direction in another part of the surface. Such neutral lines are,
in general, not known a priori.
For a given normal stress, the friction stress has a step function behavior based upon the value of v r or u .
ft or fr
Stick
vr
Slip
Figure 9-24 Coulomb Friction Model
= f t f n
where f t is the tangential force and f n is the normal force. is the friction coefficient (See BCTABLE or BCBDPRP
bulk data entry field FRIC). Stick is given by 0 .
The rate of relative tangential displacement is split into elastic and plastic contributions according to:
e p
ut = ut + ut
and the rate of change of friction force vector is related to the elastic tangential displacement by:
e
f t = Du t
Main Index
394 Nonlinear Users Guide
Friction
f n
-------- 0
D =
f
0 -------n-
where is the slip threshold below which sticking is simulated. The algorithm determines as 0.0025 times the
average edge length of the elements defining the deformable contact bodies.
The physically impossible case where 0 , implies that the plastic or slip contribution must be determined.
p
ft = D ut ut
The bilinear Coulomb model then assumes a slip flow potential given by = f t and further assumes that the
direction of the slip displacement rate is given by the normal to the slip flow potential. Note that this is similar to a
nonassociative rule, as and are different functions.
p
u t = --------
f t
where is the slip displacement rate. Also, since the friction force must always lie on the slip surface:
T
= ------ ft = 0
f t
Combining the preceding two equations the slip displacement rate is determined as:
T
------ Du t
f t
= ----------------------------
T
----- - D -------
f t f t
Main Index
Chapter 9: Contact 395
Friction
T
D -------- ------ D
f t f t
f t = D ---------------------------------- u t = D D u t
- T
-----
D --------
f t f t
The bilinear Coulomb model also uses the check on convergence to determine if convergence is achieved when
previous
Ft Ft
----------------------------------------------- e
Ft
previous
where F t is the current total friction force vector and F t is the total friction force vector from the previous
iteration. e is the friction force tolerance, which has a value of 0.05.
The bilinear Coulomb model may not correlate well with experiment if the frictional tractions exceed the flow stress
or the stress of the material. If this occurs, you can modify the frictional stress limit (see BCTABLE entry field FRLIM).
limit
In the case of the frictional stress limit t if the surface traction reaches the limit value, the applied friction force
limit
is reduced, so that the maximum shear traction is given by min n t . You may also switch to the bilinear shear
friction model.
The bilinear shear friction model states that the frictional stress is a fraction of the equivalent stress in the material:
t ------- for stick and t = ------- t for slip
3 3
where t is the surface tangent vector. is the friction coefficient (See BCTABLE entry field FRIC). t is computed by:
t = min n -------
3
While numerically there are differences in the implementation of N2S and S2S for modeling friction, the user input
associated with N2S is also used for S2S. The friction types for both methods are:
The coefficient of friction (FRIC on the BCTABLE or BCBDPRP bulk data entry) has the same meaning.
The distance below which sticking occurs (RVCNST entered on the BCPARA bulk data) is ignored.
The STS procedure allows a non-symmetric formulation to be used, which is activated by setting SEGSYM to 1 on the
BCPARA bulk data . While this may reduce the number of iterations, it may increase the amount of memory to store
the global stiffness matrix and for factorization.
Main Index
396 Nonlinear Users Guide
Friction
= x f n T v r y
or
m = m x f n T v r y
vr relative sliding velocity between point at which friction is being calculated and surface
Friction Parameters
On the Contact Control Parameters subform, select Friction Parameters....
Type Available options for friction Type are: None (default), Shear (for metal forming),
Coulomb (for normal contact default), Shear for Rolling, Coulomb for Rolling,
Stick-Slip.
Note: Unless this pulldown is changed from None, no friction will be active.
Main Index
Chapter 9: Contact 397
Friction
Method For Coulomb type of friction models, there are two methods for computing
friction: Nodal Stress (by default), Nodal Forces.
Relative Sliding Velocity Critical value for sliding velocity below which surfaces will be simulated as
sticking.
Transition Region Slip-to-Stick transition region.
Multiplier to Friction Coefficient Friction coefficient multiplier.
Friction Force Tolerance Friction Force Tolerance.
For a demonstration of the use of friction, see 3-D Loaded Pin with Friction (Ch. 4), Bilinear Friction Model:
Sliding Wedge (Ch. 5) and Friction Between Belt and Pulley (Ch. 24) in the MSC Nastran Demonstration Problem
Manual.
Main Index
398 Nonlinear Users Guide
Dynamic Impact
Dynamic Impact
The automatic contact procedure can also be used in dynamic analyses to model impact problems. High frequency
vibration or impact where wave propagation is important should use SOL 700.
The Newmark-beta/General-alpha procedure combined with NLSTEP have the capability to allow variable time steps
and, when using the user-defined fixed time step procedure, the time step is split by the algorithm to satisfy the contact
conditions.
When a node contacts a rigid surface, it is given the velocity and acceleration of the rigid surface in the normal
direction. The rigid surfaces are treated as if they have infinite mass, hence, infinite momentum.
A dynamic penetration cutback scheme has been implemented. The default iterative penetration scheme that is used
for statics does not work well for dynamics since high-frequency oscillations are excited by this process. Instead, a
time step cutback is triggered when dynamic penetration is detected. The increment is repeated with a smaller time
step. This time step t c is defined by the penetration algorithm as a factor of the original time step t o . The scheme
is depicted in Figure 9-25:
tc
to
Multiple penetrations are possible in a single increment. After the penetration cutbacks, time step for the subsequent
increment is restored to the pre-penetration time step. The penetration cutback is indicated in the .f06 file by
*** USER INFORMATION MESSAGE 4550 (nl3con)
*** TIME-STEP REDUCTION IS ACTIVATED BY DYNAMIC PENETRATION.
The limitations of the dynamic contact enhancements are as follows:
a. The Generalized-alpha scheme with zero spectral radius is a damped operator. The accuracy of the operator
is a function of the time steps used. Large time steps can cause frequency ranges of interest also to be
damped out. A general recommendation would be to use time steps about 2 to 5% of the dominant period
of the system.
Main Index
Chapter 9: Contact 399
Dynamic Impact
b. There is no special code to deal with momentum/energy conservation for impact problems. While the
elimination of high-frequency content through the mechanisms described previously and satisfaction of the
dynamic equilibrium equations given in (1) generally suffices for most contact/impact problems, it may not
suffice for systems where large amounts of energy conversion (kinetic energy to strain energy and vice
versa) occur during the contact process.
As a demonstration of dynamic impact, see Dynamic Impact of a Rigid Sphere on a Woven Fabric (Ch. 47) in the MSC
Nastran Demonstration Problem Manual.
Main Index
400 Nonlinear Users Guide
Accuracy
Accuracy
While both the node-to-segment and segment-to-segment approach can give an accurate solution, the node to segment
technique relies on a discrete numerical implementation while the segment-to-segment approach uses a continuous
approach. In the first example of bolt tightening, the control zone is distributed over a larger and smoother area. As a
force is applied to the bolt, the consequences will be that the contact stresses are lower.
Bolted connection
Node-to-Segment SegmenttoSegment
Displacements agree, note discontinuous mesh glued connection
Main Index
Chapter 9: Contact 401
Accuracy
The third problem shows an interference fit between two concentric disks with different mesh discretization. As seen,
the segment-to-segment approach provides a smoother stress distribution.
Main Index
402 Nonlinear Users Guide
Thermal Contact
Thermal Contact
In thermal contact, a distinction is made between hard contact glued, hard contact convective and near contact. Hard
contact glued means that two bodies contact each other mechanically and the temperature fields on both sides of the
interface must be made identical. Hard contact convective means that two bodies contact each other mechanically, but
the heat flow between the two sides of the interface is through some hypothetical medium, so the temperatures on both
sides do not have to be identical. The heat flow through this hypothetical medium is controlled by a one-dimensional
linear heat flow law. Near contact means that two bodies are not in contact mechanically, but are close enough to
exchange heat. The heat flow between the two interfaces is controlled by a linear or nonlinear one-dimensional heat
flow law, so it is assumed that the bodies are already relatively close to each other as compared to their sizes.
Two types of bodies are considered in a thermal contact analysis, meshed bodies, and rigid bodies. Meshed bodies are
defined by finite elements and can conduct heat, so they can have a nonuniform temperature distribution. Meshed
bodies can also exchange heat with an environment in regions that are not contacting other bodies. Various convective
heat transfer laws are possible. Rigid bodies are defined by geometrical entities and have a constant temperature
everywhere. Rigid bodies can have a heat capacity, in which case they can exchange heat with meshed bodies or they
have a prescribed temperature, so they act as a heat source when contacted by meshed bodies.
When a meshed body contacts another meshed body, the grids of the slave (or touching) body touch entities of a master
(or touched) body. These entities are element edges in two-dimensional analysis, like plane strain or axisymmetric
analysis or they are element faces in three-dimensional analysis. In either case each constraint relation controlling the
heat transfer is between one slave grid and a set of master grids of the touched entity. When a meshed body contacts a
rigid body the grids of the slave body touch geometric entities of the master body. These geometric entities all have
the same temperature and each constraint relation is between one slave grid and the rigid body having the constant
temperature.
Main Index
Chapter 9: Contact 403
Thermal Contact
Mesh-to-Mesh Contact
Mesh-to-Mesh Hard Contact Glued
In this case, a grid is in contact with a face of an element. If the contacting grid temperature is T A and the face grid
temperatures are T B , the condition is a simple linear MPC:
i
N
T A = T B = i T B
i
(9-1)
i = 1
T A is the dependent grid, T B are the independent grids. The weight factors i depend on the location where the T A
i
grid touches the face and are simply the shape function values to interpolate the temperatures inside the face. This may
be written as
TA
TA 1 0 0 TB
= 1 = ST (9-2)
T B 0 1 N .
TB
N
TA
TB T
1 1 N 1 = HM P C T = 0 (9-3)
.
TB
N
Q A = q = h T A T B (9-4)
where is the area contribution for grid A , q the heat flow per unit area and h is the heat transfer coefficient, which
may be composed of several contributions. The heat entering body B is Q B = Q A . For each grid that is in contact
with a master face, a conductive heat transfer element matrix gets created of the following form:
Main Index
404 Nonlinear Users Guide
Thermal Contact
Kc o n t a c t = h S
T 1 1 S , (9-5)
1 1
TA
TB
where the element degrees of freedom of vector is: 1
.
TB
N
This can be interpreted as a one dimensional heat link with one grid being grid A and its other grid B being on the
touching element face, so its temperature is computed from the face grid temperatures by the linear MPC equation.
The heat transfer coefficient h can have several contributions, coming from different heat flow types:
q 1 = h 1 T A T B (9-6)
EXPF
q 2 = h 2 T A T B T A T B (9-7)
2 2
q 3 = T A + T B T A + T B T A T B (9-8)
EX P F E X PF
TA T B
q 4 = h 4 ---------------------------------------------------------------- T A T B (9-9)
T A T B
d d
q 5 = h 5 A 1 ------------- + h 5 B ------------- T A T B (9-10)
d n e a r dn e a r
Equation (9-6) is just a linear heat flow law with a constant heat transfer coefficient. Equations (9-7), (9-8), and (9-9)
represent nonlinear heat flow laws, where equation (9-8) is a radiation law. Equation (9-10) allows to have a distance
dependent heat flow between the two bodies in case of near contact, where d n e a r is the near contact distance tolerance.
When bodies are separated by a distance larger than d n e a r , they are not in contact.
The total flux per unit area is the sum of these contributions:
5
q = qi (9-11)
i = 1
For hard convective contact, only the first flux type is used. For near contact, it may be a combination of all five. All
relations are of the form:
q = h T A T B T A T B (9-12)
Main Index
Chapter 9: Contact 405
Thermal Contact
where the heat transfer coefficient is some nonlinear function of the body temperatures in the contact point. Note that
for the radiative condition q 3 the temperatures should always be absolute temperatures.
Mesh-to-Geometry Contact
Mesh-to-Geometry Hard Contact Glued
A grid of the meshed body is contacting a rigid surface. In this case, the grid temperature is set equal to the rigid
geometry temperature. This can be done in two ways:
as a SPCD condition: T A = T B OD Y
as a MPC condition: T A T B OD Y = 0
In the latter case, an extra scalar point for the rigid contact body is created, which may obtain either a prescribed
temperature or it may be loaded with an external flux. A simple MPC between this scalar point and each grid that
contacts the rigid body is created. With this extra scalar point, it is possible to associate a total heat capacity C MB (S.I.
units: J/K) of the rigid body, which simply represents a lumped heat capacity contribution to the scalar point.
Q A = h T A T B OD Y (9-13)
Again two ways are distinguished as discussed in Mesh-to-Geometry Hard Contact Glued:
1. The body temperature is applied directly. In this case a concentrated conductivity matrix term:
K c o n t a c t = h T A and a load term Q c o n t a c t = h T B O DY for grid A are obtained.
2. An extra scalar point is created. In this case a simple conductivity matrix arises:
This latter relation is similar to the one in Mesh-to-Mesh Convective Heat Transfer (Hard or Near Contact), but now
their is only one master degree of freedom, which is the temperature of the scalar point associated with the rigid body.
No Contact
A face of a meshed body not being contacted or only partially being contacted may exchange heat with its environment
according to similar laws as the first four in Mesh-to-Mesh Convective Heat Transfer (Hard or Near Contact), but the
Main Index
406 Nonlinear Users Guide
Thermal Contact
temperature T B is replaced by an ambient temperature T A M B , which is the same for all faces of a contact body and
T A now represents the surface temperature. The flux per unit are is:
q = h TA TA M B (9-15)
where is an area contribution factor for the face 0 1 . The total contribution is determined by the
integration over the element face (this is similar to an elastic foundation in mechanical terms):
T
Q fi lm = h T A T A M B dA = H A T A Q A M B (9-16)
A
face
where is the vector of shape functions that interpolate the temperatures inside the element face. From the above
expression, a conductivity matrix contribution and a load vector contribution arise.
The conductivity matrix is:
T
HA = h dA (9-17)
A
fac e
QA M B = hT A M B dA (9-18)
A
fac e
Mechanical Coupling
Frictional Heat
The heat generated by sliding friction in the contact interfaces gets evenly distributed over the two bodies that contact
each other and can be expressed as:
1
Q A = Q B = --- F w rel , (9-19)
2
where F w is the magnitude of the friction force at grid A and r el is the magnitude of the relative velocity between
the two bodies at grid A. Q A can be applied directly as a concentrated heat flux to the contacting grid. Q B is distributed
over the grids of the master face being contacted: Q B = i Q B The scale factor may be used to
i
accommodate different unit systems in the heat transfer and mechanical analysis or to accommodate a loss to the
environment, in which case not all heat generated is available as a loading heat source.
Main Index
Chapter 9: Contact 407
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Problem Statement
A model of a cylindrical pipe is subjected to crushing as rigid bodies above and below the pipe move inward toward
each other. The model is created using 2-D shell elements to model the pipe, and rigid surfaces above and below the
pipe. This exercise illustrates several SOL 400 capabilities including large displacement analysis, contact analysis
between rigid and deformable bodies, and plasticity modeled with an elastic perfectly plastic material model.
Main Index
408 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
under external pressure (if submerged) and internal pressure (fluid flow). However, in this case we examine failure due
to external loading from adjacent structures.
In this problem the pipe section is subjected to offset lateral loading from external structures on the top and bottom.
The problem attempts to quantify whether the movement of these external structures cause the plastic collapse of the
pipe. Initial contact with the external structures is expected to cause elastic deformation of the steel pipe. Additional
incremental movement potentially subjects the structure to stresses beyond the proportional limit of the material. The
yield stress defines the onset of plastic strains that may initiate the collapse of the structure walls. The resulting
deformation data can establish the conditions for catastrophic collapse of the pipe section and define allowable limits
to contact of external structures.
Model Description
The model includes a 3-D representation of the deformable pipe structure and two semi-circular sections of rigid pipes
sections.
Main Index
Chapter 9: Contact 409
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Material
The entire pipe section is made of steel and modeled as an elastic-plastic material using the MAT1 and MATEP entries
for large displacement analysis. The linear elastic behavior model is effective up to a yield stress of 36000psi. In this
range imposed stresses induce linear proportional strains. The stress-strain relationship is defined using an elastic
modulus of 3.0E+6 and a Poisons ratio of 0.3. Beyond the yield point perfect plasticity causes complete yielding of the
material upon any incremental stress.
Main Index
410 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Step-by-step instructions for creating and running the analysis on the pipe crush model can be accessed by linking to
the pipe_crush instructions file below. Once you are done creating the model and running the SOL 400 analysis,
proceed by accessing the results file to generate visual displays of the simulated deformations and stresses.
Output Requests
This section of the MSC Nastran .bdf specifies the type of output requested.
$ NASTRAN input file created by the Patran 2013 64-Bit input file
$ translator on March 14, 2014 at 08:15:56.
$ Direct Text Input for Nastran System Cell Section
NASTRAN SYSTEM(316)=19
$ Replace the 19 in the above line with 7
$ to get a restart DBALL instead of what is currently written
$ Direct Text Input for File Management Section
Main Index
Chapter 9: Contact 411
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Main Index
412 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Material Entries
Contact Bodies
Material Entries
This section of the MSC Nastran .bdf describes the material behavior for the pipe. The MAT1 entry defines
the elastic part of the material, and the MATEP entry describes the perfectly plastic behavior.
Contact Bodies
The BCBODY entry describes the rigid surface as defined by NURBS geometry surfaces. The value of analytical
form is 1 (used with all NURBS rigid geometry). Note the -2 describing the rigid body motion.
Main Index
Chapter 9: Contact 413
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
subcase inc cycl sepa cut cycl split separ cut rmesh time step
/step # # # # # # # # # # of
1 |--of the inc--|-----------of the analysis-----------| the inc
1 0 0 0 0 0 0 0 0 0 0.0000E+00
1 1 5 1 0 5 0 1 0 0 1.0000E-02
1 2 1 0 0 6 0 1 0 0 1.0000E-02
1 3 2 0 0 8 0 1 0 0 1.2000E-02
1 4 1 0 0 9 0 1 0 0 1.4400E-02
1 5 1 0 0 10 0 1 0 0 1.7280E-02
1 6 1 0 0 11 0 1 0 0 2.0736E-02
1 7 2 0 0 13 0 1 0 0 2.4883E-02
1 8 1 0 0 14 0 1 0 0 2.9860E-02
1 9 16 3 0 30 0 4 0 0 3.5832E-02
1 10 3 1 0 33 0 5 0 0 2.9860E-02
1 11 2 1 0 35 0 6 0 0 3.5832E-02
1 12 2 1 0 37 0 7 0 0 4.2998E-02
1 13 3 1 0 40 0 8 0 0 5.1598E-02
1 14 3 0 0 43 0 8 0 0 6.1917E-02
1 15 8 2 0 51 0 10 0 0 7.4301E-02
1 16 4 1 0 55 0 11 0 0 8.9161E-02
1 17 3 1 0 58 0 12 0 0 1.0699E-01
1 18 5 0 0 63 0 12 0 0 1.2839E-01
1 19 5 0 0 68 0 12 0 0 1.2839E-01
1 20 2 0 0 70 0 12 0 0 7.5565E-02
Job ends with exit number : 1
total wall time: 15.00
total cpu time: 15.15
Main Index
414 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Results Plots
Figure 9-28 shows the final deformed shape of the pipe section after 1 second of loading. Maximum total strains of
6.0E-2 inches develop in rigid body contact areas. Slightly higher total strains are evident on the lower pipe section in
this region due to the difference in radius of the external rigid body structures.
Plasticity has occurred over most of the model as shown in Figure 9-29. Maximum stresses reaching 4.25E+4 psi occur
as expected in the contact region. Also evident from this figure are high stresses equivalent to those in the contact
region occurring at the constrained boundary condition. This is a consequence of the boundary condition. In a full
length pipe section, we would not expect to see a stress concentrated in this area.
Main Index
Chapter 9: Contact 415
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Examples
Below is an example for glued contact and linear contact. Other examples of contact can be found in the MSC Users
Guide.
Main Index
416 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Main Index
Chapter 9: Contact 417
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
Main Index
418 Nonlinear Users Guide
Elastic-Plastic Collapse of a Cylindrical Pipe under External Rigid Body Loading
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 -3.343832E-05 5.782588E-03 9.203274E-04 1.000000E+00 -3.343832E-05
2 2 -1.705077E-05 4.129258E-03 6.571918E-04 1.000000E+00 -1.705077E-05
3 3 -1.452048E-06 1.205010E-03 1.917832E-04 1.000000E+00 -1.452048E-06
4 4 2.856250E-06 1.690044E-03 2.689789E-04 1.000000E+00 2.856250E-06
5 5 1.440756E-05 3.795729E-03 6.041091E-04 1.000000E+00 1.440756E-05
6 6 2.289412E-05 4.784780E-03 7.615214E-04 1.000000E+00 2.289412E-05
7 7 6.371218E+07 7.981991E+03 1.270373E+03 1.000000E+00 6.371218E+07
8 8 4.260703E+08 2.064147E+04 3.285192E+03 1.000000E+00 4.260703E+08
9 9 1.158678E+09 3.403937E+04 5.417533E+03 1.000000E+00 1.158678E+09
10 10 1.949990E+09 4.415869E+04 7.028074E+03 1.000000E+00 1.949990E+09
11 11 4.246696E+09 6.516668E+04 1.037160E+04 1.000000E+00 4.246696E+09
12 12 4.305021E+09 6.561266E+04 1.044258E+04 1.000000E+00 4.305021E+09
13 13 4.692808E+09 6.850407E+04 1.090276E+04 1.000000E+00 4.692808E+09
14 14 6.424801E+09 8.015486E+04 1.275704E+04 1.000000E+00 6.424801E+09
15 15 7.120649E+09 8.438394E+04 1.343012E+04 1.000000E+00 7.120649E+09
16 16 9.036405E+09 9.506001E+04 1.512927E+04 1.000000E+00 9.036405E+09
17 17 9.057384E+09 9.517029E+04 1.514682E+04 1.000000E+00 9.057384E+09
18 18 9.837994E+09 9.918666E+04 1.578605E+04 1.000000E+00 9.837994E+09
19 19 1.080527E+10 1.039484E+05 1.654390E+04 1.000000E+00 1.080527E+10
20 20 1.156105E+10 1.075223E+05 1.711271E+04 1.000000E+00 1.156105E+10
1 * * * END OF JOB * * *
Main Index
Chapter 9: Contact 419
Numerical Procedures
Numerical Procedures
Lagrange Multipliers
In performing contact analyses, you are solving a constrained minimization problem where the constraint is the no
penetration constraint. The Lagrange multiplier technique is the most elegant procedure to apply mathematical
constraints to a system. Using this procedure, if the constraints are properly written, overclosure or penetration does
not occur. Unfortunately, Lagrange multipliers lead to numerical difficulties with the computational procedure as their
inclusion results in a nonpositive definite mathematical system. This requires additional operations to insure an
accurate, stable solution which leads to high computational costs. Another problem with this method is that there is no
mass associated with the Lagrange multiplier degree of freedom. This results in a global mass matrix which cannot be
decomposed. This precludes the used of Lagrange multiplier techniques in explicit dynamic calculations which are
often used in crash simulations. The Lagrange multiplier technique has often been implemented in contact procedures
using special interface elements such as the SOL 400 gap element. This facilitates the correct numerical procedure, but
puts a restriction on the amount of relative motion that can occur between bodies. The use of interface elements
requires an apriori knowledge of where contact occurs. This is unachievable in many physical problems such as crash
analysis or manufacturing simulation.
Penalty Methods
The penalty method or its extension, the Augmented Lagrangian method, is an alternative procedure to numerically
implement the contact constraints. Effectively, the penalty procedure constrains the motion by applying a penalty to
the amount of penetration that occurs. The penalty approach can be considered as analogous to a nonlinear spring
between the two bodies. Using the penalty approach, some penetration occurs with the amount being determined by
the penalty constant or function. The choice of the penalty value can also have a detrimental effect on the numerical
stability of the global solution procedure. The penalty method is relatively easy to implement and has been extensively
used in explicit dynamic analysis although it can result in an overly stiff system for deformable-to-deformable contact
since the contact pressure is assumed to be proportional to the pointwise penetration. The pressure distribution is
generally oscillatory.
Direct Constraints
Another method for the solution of contact problems is the direct constraint method. In this procedure, the motion of
the bodies is tracked, and when contact occurs, direct constraints are placed on the motion using boundary conditions
both kinematic constraints (MPC and SPC) on transformed degrees of freedom and nodal forces. This procedure can
be very accurate if the program can predict when contact occurs. This is the procedure that is implemented in SOL 400
Main Index
420 Nonlinear Users Guide
Numerical Procedures
when using the node-to-segment method. No special interference elements are required in this procedure and complex
changing contact conditions can be simulated since no apriori knowledge of where contact occurs is necessary.
Notes: The automatic penetration checking procedure is always used with the NLSTEP bulk data entry.
The arc length methods are not available for contact simulation.
Main Index
Chapter 9: Contact 421
Implementation of Constraints in Node-to-Segment
K a a K a b u a f a
=
K b a K b b u b fb
where a represents the nodes in contact which have a local transformation, and b represents the nodes not in contact
and, hence, not transformed. Of the nodes transformed, the displacement in the normal direction is then constrained
such that u a n is equal to the incremental normal displacement of the rigid body at the contact point.
t
P
n
Figure 9-31 Transformed System (2-D)
As a rigid body can be represented as either a piecewise linear or as an analytical (NURB) surface, two procedures are
used. For piecewise linear representations, the normal is constant until node P comes to the corner of two segments as
shown in Figure 9-32. During the iteration process, one of three circumstances occur. If the angle is small
smooth smooth , the node P slides to the next segment. In such a case, the normal is updated based upon the
new segment. If the angle is large ( smooth or smooth ) the node separates from the surface if it is a
convex corner, or sticks if it is a concave corner. The value of smooth is important in controlling the computational
costs. A larger value of smooth reduces the computational costs, but might lead to inaccuracies. The default values
are 8.625 for 2-D and 20 for 3-D. These can be reset using the ANG2D or ANG3D fields on the NLSTRAT entry for
each subcase.
Main Index
422 Nonlinear Users Guide
Implementation of Constraints in Node-to-Segment
In 3-D, these corner conditions are more complex. A node (P) on patch A slides freely until it reaches the intersection
between the segments. If it is concave, the node first tries to slide along the line of intersection before moving to
segment B. This is the natural (lower energy state) of motion.
These corner conditions also exist for deformable-to-deformable contact analysis. Because the bodies are continuously
changing in shape, the corner conditions (sharp convex, smooth or sharp concave) are continuously being re-evaluated.
When a rigid body is represented as an analytical surface, the normal is recalculated at each iteration based upon the
current position. This leads to a more accurate solution, but can be more costly because of the NURB evaluation.
When a node of a deformable body contacts a deformable body, a multipoint constraint (MPC) is automatically
imposed. Recalling that the exterior edges (2-D) or faces (3-D) of the other deformable bodies are known, a constraint
expression is formed. For 2-D analysis, the number of retained nodes is three two from the edge and the contacting
node itself. For 3-D analysis, the number of retained nodes is five four from the patch and the contacting node itself.
The constraint equation is such that the contacting node should be able to slide on the contacted segment, subject to
the current friction conditions. This leads to a nonhomogeneous, nonlinear constraint equation. In this way, a
contacting node is forced to be on the contacted segment. This might introduce undesired stress changes, since a small
gap or overlap between the node and the contacted segment will be closed. During initial detection of contact
(increment 0), the stress-free projection option avoids those stress changes for deformable contact by adapting the
coordinates of the contacting nodes such that they are positioned on the contacted segment. This stress-free projection
can be activated using the BCTABLE entry. A similar option exists for glued contact; however, in this case, overlap will
not be removed.
During the iteration procedure, a node can slide from one segment to another, changing the retained nodes associated
with the constraint. A recalculation of the bandwidth is automatically made. Because the bandwidth can radically
change, the bandwidth optimization is also automatically performed.
A node is considered sliding off a contacted segment if is passes the end of the segment over a distance more than the
contact tolerance. As mentioned earlier, the node separates from the contacted body if this happens at a convex corner.
For deformable contact, this tangential tolerance at convex corners can be enlarged by using the delayed sliding off
option activated via the BCTABLE bulk data entry.
Main Index
Chapter 9: Contact 423
Implementation of Constraints in Node-to-Segment
TA
TB T
gT = 1 1 N .
1 = HM P C T = 0 (9-20)
..
.
TB
N
For each MPC constraint, an additional Lagrange multiplier degrees of freedom is created. The constraint is
furthermore augmented with a penalty term. The system matrix contribution for this constraint becomes:
T T
pH M P C H M P C H M P C
KM P C = , (9-21)
T
HM P C 0
where p is the penalty factor used in the constraint. The associated incremental degrees of freedom are:
dT ,
d
where is the Lagrange multiplier degree of freedom. The residual vector contribution for this constraint becomes:
+ pg H M P C
RM P C = (9-22)
g
When a grid touches a rigid body that has no scalar point associated with it, the SPCD-like constraint is simulated by
a simple MPC equation with a non-homogeneous term C n o n . In this MPC, there are no independent grids.
g = T A T B O DY = 0 (9-23)
dT A = g = T A T B O DY = C n o n (9-24)
and H M P C is the trivial unit vector with one component, so computing K M P C and R M P C becomes almost trivial.
Main Index
424 Nonlinear Users Guide
Implementation of Constraints in Node-to-Segment
Main Index
Chapter 9: Contact 425
Segment-to-Segment Contact
Segment-to-Segment Contact
Instead of using automatically generated multi-point constraint equations to enforce the contact constraints in a
point-wise manner, the segment-to-segment contact algorithm tries to enforce the contact constraints on the common
area of contacting bodies. The methodology used is that of augmented Lagrangians, which can be seen as a
combination of a penalty method and a Lagrange multiplier method. In contrast to a Lagrange multiplier method, there
are no extra degrees of freedom in the system of equations to be solved, but there may be extra iterations needed to
improve the fulfillment of the contact constraints. An important parameter in the algorithm is the augmented Lagrange
penalty factor. Typically, a too high value of this factor can cause an ill-conditioned stiffness matrix, where a too small
value can cause a relatively large penetration after the first iteration in the Newton-Raphson process and many
iterations to remove this penetration. The default value of the penalty factor is determined by the program, but the
BCPARA and BCTABLE bulk data entries allow you to overrule the default value. The default value of the penalty factor
E n depends on the body stiffness of the two contact bodies involved and a characteristic length (note that the
dimension of the penalty factor is force per cubic length). The body stiffness is either defined by the average trace of
the initial stress-strain law of the elements defining the two contact bodies or by the average bulk modulus for (nearly)
incompressible rubber materials, whichever of the two is the largest. For continuum elements, the characteristic length
is given by one half of the average length of all the edges being part of the contact boundary. For shell elements, the
characteristic length is given by half of the average thickness of all the shell elements being part of a contact body.
When there is contact between a solid and a shell element, then the characteristic length is defined by the shell element.
If there is contact between the deformable bodies k and l , with body stiffnesses S k and S l , and the characteristic
0.5 S k + S l
E n = ------------------------------
L
In case of contact with a rigid body, the default value of E n is related to the deformable body k only and is given by:
1000S k
E n = -----------------
L
Since the contact constraints are generally fulfilled in an iterative fashion, a threshold value of the penetration distance
is needed to decide whether or not the augmentation procedure will be invoked. Like the penalty factor discussed
above, this penetration distance beyond which an augmentation is applied has a default value determined by the
program. but can also be user-defined via the BCPARA and BCBODY bulk data entries. The default value is based on
the characteristic length L according to:
a u g = 0.001L
Main Index
426 Nonlinear Users Guide
Segment-to-Segment Contact
Augmentation based on a constant penetration field. This is the recommended augmentation procedure for
linear finite elements.
Augmentation based on a (bi-) linear penetration field. This augmentation procedure should be used only for
quadratic finite elements.
It is also possible to let the program decide on the use of a constant or (bi-) linear penetration field, based on the element
types involved (linear or quadratic).
Contact Tolerance Distance The contact tolerance difference (ERROR entered on the BCPARA or BCTABLE)
behaves in the same manner for both procedures, but in the case of STS it has two additional consequences:
1. If the Sliding Limit (SLDLMT) is not entered on the BCPARA option, then SLDLMT will be 5* ERROR and,
for small values of ERROR, this may unnecessarily result in the message Recalculating Segments, thus
increasing the computational costs.
2. If the penetration distance for which Augmentation is triggered is not entered, then this distance is the value of
ERROR/1000, which will be used for deformable-to-deformable contact (only if the augmentation method is
requested).
Searching Order When using the STS contact algorithm, the search order is virtually immaterial, and the value of
ISEARCH on the BCTABLE bulk data entry and ISTYP on the BCBODY bulk data entry is ignored.
Because the searching order is not used with STS, the value of the Hard-Soft ratio (HARDS on the BCTABLE bulk data
entry) is also ignored.
Large Deformation/Sliding Large sliding behavior is treated differently using NTS and STS. The NTS algorithm
always assumes finite sliding.
Numerical Procedures The S2S method has a few additional parameters that govern the numerical implementation.
Penalty Factors The constraint in the normal direction and the tangential direction (if friction is present) are imposed
using a penalty method or an Augmented Lagrange method. In both cases, penalty factors are required for the normal
and the shear behavior. These values are entered as PENALT and TPENALT on the BCPARA or the BCTABLE bulk data
entry. It is not necessary to enter these values, though it may be appropriate if the default values lead to unsatisfactory
behavior. In general, a too small value of the normal penalty value will lead to penetration, while a too large value will
lead to an increased number of iterations and possibly a non-converging solution. The default value of the normal
penalty is based upon the material behavior and a characteristic length. For shell contact, the default penalty maybe
too large, which results in an overly stiff response or in poor convergence behavior; in such cases, it is recommended
to reduce the magnitude. Note that the penalty factors have units of FL-3.
Augmentation To minimize the amount of penetration associated with too low Penalty Factors the Augmentation
procedure may be used. You can control this for deformable-to-deformable contact using the AUGMENT parameter on
the BCPARA bulk data entry. Note that for deformable-to- rigid contact augmentation is always included. The values
of AUGMENT that can be used are
Main Index
Chapter 9: Contact 427
Segment-to-Segment Contact
The augmentation is not activated unless the amount of the penetration is greater than AUGDIST (entered on the
BCPARA or BCTABLE bulk data entry). The default is the value of the contact tolerance distance divided by 1000. In
general, this is a very small value.
When friction is included in the STS contact algorithm, the Sticking Stiffness is given by TPENALT on the BCPARA
bulk data entry. If TPENALT is not given, PENALT/1000 is used. If PENALT is also not given, then it will use the
default value of PENALT (see above) divided by 1000 will be used.
Alternate Defaults For some recent problems it has been determined that better default values of the penalty should
be used. There may be activated by using System Cell 664 = 1.
This has the effect of forming a penalty factor based upon contact pairs as opposed to global behavior
When deformable-to-deformable contact occurs use the smallest penalty factor as opposed to the average
penalty.
For rubber materials do not incorporate the bulk modulus, if the bulk modulus is very large.
Contact/Separation Output The contact behavior for the STS procedure is based upon a detection and separation
of a continuous set of polygons on the surface as compared to the NTS procedure which detects nodes to segments.
The result is that the contact status generally appears to be both larger and smoother using the STS method. When
considering separation, iterations may occur when any pair of polygons separate and this separation may cause the
convergence criteria to be not fulfilled. The fact that a polygon point has a pulling force which leads to separation
causes a change in the residual load vector and hence may cause the convergence criteria not to be fulfilled.
Since separation of polygon points will not necessarily cause extra iterations, the number of separations using the STS
procedure is not output to the status file (jid.sts).
Compatibility The contact capability in MSC Nastran is undergoing active enhancements with each release.
Extensive testing is being performed to ensure both an accurate and computationally efficient solution. The
consequences are that unlike linear elastic analysis, the numerical results may change from release-to-release. System
cells may be provided to obtain backward compatibility, but that is not always achievable.
Both for the node-to-segment as well as the segment-to-segment contact algorithm, an optimization technique has been
introduced, which adjusts the element thickness and offset (if defined) at hard edges in order to get a correct description
of the top and bottom surfaces.
Main Index
428 Nonlinear Users Guide
Segment-to-Segment Contact Theory
where u is the displacement field and u the kinematically admissible variation of this field. Contact conditions can
be incorporated by adjusting this virtual work equation by two surface integrals:
T
G u u + n g n d + t g t d = 0 (9-26)
Here is that part of boundary of the bodies being in contact and the subscript n is used to indicate the normal
direction to the contact boundary, where the subscript t indicates the tangential direction. The function g n is called a
gap function and expresses the distance between a point and its closest point projection on the contact boundary. If
g n 0 , a point is outside the contact boundary, if g n = 0 , a point is on the contact boundary and if g n 0 , a point
would be beyond the contact boundary, which is physically inadmissible. The Lagrange multiplier n represents the
contact normal stress. Similarly, g t is the tangential gap vector and t is the tangential stress vector.
The tangential or frictional behavior is assumed to be governed by Coulombs friction law, which, as already stated
before for node-to-segment contact, can be expressed as:
= t n 0 (9-27)
Since the gap functions g n and g t depend on the geometry of the bodies (defining the normal and tangential vectors)
and the displacement field u , the contact part of equation (9-26) can be expressed as:
g n T g t
G c u u = n -------
u
- u d + t ------- u d
u
(9-28)
Since equation (9-26) has to be valid for all kinematically admissible variations u , the contact contribution to the
equilibrium equation is given by:
T g n T g t
Fc = n -------
u
- d + t ------- d
u
(9-29)
The overall solution procedure will be based on an augmented Lagrangian approach. It is assumed that the solution at
step n 1 is known and that the solution at step n has to be determined in an iterative fashion, using a Newton-
Raphson process. Based on the iterative displacement solution obtained at iteration i , the following trial solutions for
the Lagrange multipliers are introduced (note that the superscript n to indicate the step number is omitted to simplify
the notation):
Main Index
Chapter 9: Contact 429
Segment-to-Segment Contact Theory
tr ial i1 i t ria l i1 i
n = pn + En gn ; t = tt + Et gt (9-30)
in which p n and t t are fixed estimates of n and t , E n and E t are penalty factors for the contact behavior in the
normal and tangential directions (actually E t is a diagonal matrix with the penalty factor E t on the diagonal) and g n
and g t follow from the global iterative displacement solution. The default values of the penalty factors are derived
from the contact body properties (this will be discussed later on), but can also be user defined.
Looking first at contact in the normal direction, it can easily be seen that if a contact condition is exactly fulfilled, then
i tr ial i1
the gap function g n = 0 and n = p n . The idea of the augmentation procedure is to better approximate the
contact conditions by adjusting the contact normal stress. The evolution of the contact normal stress during the
0
Newton-Raphson process is illustrated in Figure 9-34. The first displacement solution is obtained using p n = p n
(which initially can be zero, but in an arbitrary increment is given by the value at the end of the previous increment).
1
Given the displacement solution, the gap function g n is calculated and the contact normal stress update is given by
1 0 1
p n = p n + E n g n . This new contact normal stress is used in a subsequent iteration. If the augmentation is not
performed, then the method reduces to a penalty method with a constant penalty factor E n . As for the contact
conditions in the tangential direction, the augmentation procedure can be applied also if there is sticking contact. Then
the sticking conditions (no relative tangential displacements) will be better approximated by adjusting the tangential
stresses (see Figure 9-35).
pn
exact value of p n
2
pn
1
pn
atan E n
0
pn
2 1 gn
gn gn
Figure 9-34 Updating Contact Normal Stress according to the Augmented Lagrange Procedure
Main Index
430 Nonlinear Users Guide
Segment-to-Segment Contact Theory
tt
n
1
tt
atan E t
0
tt
2 1
gt gt gt
Figure 9-35 Updating Contact Friction Stress according to the Augmented Lagrange Procedure (2-D)
Within the Newton-Raphson solution procedure, not only the contact contributions to the global force vector (as given
by equation (9-29)), but also the contributions to the global stiffness matrix are needed:
K c = ------ F c (9-31)
u
Further evaluation of equations (9-29) and (9-31) depends on whether there is sticking or slipping. At the very
beginning of an analysis, an assumption has to be made (typically sticking), but after having determined the trial
solutions, it is possible to check if there will be sticking or slipping:
t ria l trial t ria l t ria l t ria l
= t n ; 0 : sticking; 0 : slipping (9-32)
and previously made assumptions on the contact state can be adjusted accordingly. Now the states of sticking and
slipping will be discussed separately.
In case of sticking, the predicted solution for the Lagrange multipliers is:
n tri a l n t ria l
n = n ; t = t (9-33)
For the contact contributions to the global stiffness matrix, the second order derivatives of the gap functions with
respect to the displacements are neglected, so that the following terms are needed:
n g g
-------- = E n -------n- ; -------t = E t -------t (9-34)
u u u u
In case of slipping, the tangential stresses need to be adjusted in order to agree with Coulombs friction law:
t ria l
n tri a l n t tri a l
n = n ; t = ------------------- n (9-35)
t ria l
t
Main Index
Chapter 9: Contact 431
Segment-to-Segment Contact Theory
and the contact contribution to the global stiffness matrix now involves also changes in contact in the normal direction:
tri a l tri a l tr ial
n g t n tri a l t
-------- = E n -------n- ; -------t = ------------------
- ---------------
- + n ------ ------------------
- =
u u u t
tri a l u u t tr ial
2
tri a l
tri a l tr ial
1 ---------------------
t1
- t1 t2
------------------------------
-
t ria l tr ial t ria l 2 t ria l 2
t g n n t t g t
= ------------------- E n -------- + ------------------- E t ------- (9-36)
t
t ria l u t
tri a l
2 u
tt1ria l trial t2 t 2
t ria l
------------------------------
- 1 --------------------- -
tri a l 2 tri a l 2
t t
The first term in the expression for the frictional behavior in equation (9-36) results in a non-symmetric stiffness
matrix. If the assumption is made that the contact normal stress is constant during a Newton-Raphson iteration, then
this term cancels and a symmetric matrix is retained. Using a non-symmetric matrix will significantly increase the
memory requirements. It should be noted that a non-symmetric matrix may only have impact on the convergence
behavior; if convergence will be obtained, the final solution wont be affected.
In Force Vector and Stiffness Matrix Contribution below, it is shown how the equation (9-29) is numerically integrated
in a 2-D or 3-D contact analysis.
Contact Detection
As also shown in Figure 9-3, each contact segment will get a number of auxiliary points. These auxiliary points are
only used during the contact detection phase of the analysis and they are defined at fixed positions on the contact
segments. Since the auxiliary points are defined on a per segment basis, there are multiple points with possibly different
normal vectors at adjacent contact segments. Especially at sharp corners, this allows for multi-body contact in such a
way that each (part of a) segment can touch only one other segment.
The actual contact detection phase consists of two passes. During the first pass, there is a check whether an auxiliary
point is sufficiently close to a contact segment and if the corresponding normal vectors are almost opposite. Hence,
there is a distance check followed by a direction check (see Figure 9-36). During the distance check, the distance
between an auxiliary point and its closest point projection on a contact segment is calculated and compared with a
reference value. This reference value is determined by the contact tolerance and the bias factor B , where 0 B 1 .
The default contact tolerance is calculated by the program as the smaller of 5% of the smallest element side or 25% of
the smallest (beam or shell) element thickness, based on all the elements defined in any contact body. Alternatively,
you can define the contact tolerance. The default bias factor is zero. A nonzero bias factor, entered by you, implies that
the outside zone becomes smaller and the inside zone larger and usually avoids unnecessary iterations following from
physically unrealistic contact detection. If the position of an auxiliary point is x a and that of its closest point projection
on a segment s of contact body is x s , then the distance check is passed if x a x s 1 B when the auxiliary
point lies outside the contact body, and if x a x s 1 + B , when the auxiliary point lies inside the contact body.
If the distance check is passed, then the direction check is performed. This direction check is passed if the angle
Main Index
432 Nonlinear Users Guide
Segment-to-Segment Contact Theory
between the normal vector at the auxiliary point and the normal vector at the potentially contacted segment is larger
than a threshold value , so n a n s . The default value of is 120. The angle between the normal vectors
is defined according to Figure 9-37. Once both the distance and the direction check are passed, the corresponding
segments are marked as a pair of potentially contacting segments and for each of these pairs, the projection of one
segment on the other is determined. In 2-D, this projection defines a polyline; in 3-D, it defines a polygon. In
Figure 9-38, a 2-D and a 3-D example is given. Each polyline/polygon has a number of polygon/polygon points, which
will serve as the basis for the numerical integration of equation (9-29). It should be noted that the polyline/polygon
points are not the same as the auxiliary points introduced before. The auxiliary points only serve to find potential
contact pairs, the polyline/polygon points characterize the actual contact area between such contact pairs.
: closest point projection
na 1
1 + B
1 B
na 2
ns ns
Distance Check And Direction Check Passed Distance Check Passed, Direction Check Failed
na
n a n s
ns
Main Index
Chapter 9: Contact 433
Segment-to-Segment Contact Theory
: polyline points
Figure 9-38 Common Areas of Contact Segments: Polylines (Left) And Polygons (Right)
During the second pass of the contact detection phase, for each polyline/polygon point it is checked whether or not the
segments are locally in contact. This check is equivalent to the check which has been described above, so a distance
check followed by a direction check. At the end of this second pass, there is a set of polyline/polygon points being in
contact and representing the contact area.
T T T T
g n i = G n i G' n i U ; g ti = G ti G' t i U (9-37)
Here, vector G n i and matrix G ti express the dependency of the normal and tangential displacements of a point on a
contact segment on the total set of degrees of freedom. In a similar way, G' n i and G' ti are used for the closest point
projection of this point on the contacted segment. It should be noted that, if applicable, G n i , G' n i , G ti and G' ti also
include the effect of the shell thickness and shell offset vectors. Care has been taken to define them in such a way that
they are able to represent rigid body modes of structures being in contact without introducing spurious energies.
Equation (9-37) can be used to numerically evaluate equation (9-29). First, the contribution to the force vector is given
by:
N N
g T g T
-------n- d + -------t d =
u n u t G n i G' n i n i i + G t i G' ti ti i (9-38)
i = 1 i = 1
Main Index
434 Nonlinear Users Guide
Segment-to-Segment Contact Theory
in which N is the total number of polyline/polygon points, n i and t i are the estimated contact normal and tangential
stresses in polyline/polygon point i (see equations (9-33) and (9-35)) and i is the area corresponding to this point.
Note that a possible dependency of G n i , G' n i , G ti and G' ti on U has been neglected.
Next, for sticking contact, the contribution to the global stiffness matrix is:
N
g n T g n g T g
-------
- E n -------- d + -------t E t -------t d =
T T
u u u u G n i G' n i E n i G n i G' n i i
i = 1
(9-39)
N
T T
+ G ti G' t i E ti G t i G' ti i
i = 1
2
t1 t1 t 2
T 1 ------------- -------------------
2 2
g T g g T t g t g T t t g t
-------n- E -------n- d + -------n- E --------- -------t ---------n- E
u n u u n t u - ------
- d + u t t 2
------- d =
u
t1 t2 t2
------------------- -------------
2 2
t t
N N T
T T t i T T
= Gn i G' n i E n i G ni G' ni i + G n i G' n i E n i ------------ G ti G' t i i +
t i
(9-40)
i = 1 i = 1
2
t1 i t1 i t2 i
N 1 --------------- -----------------------
2 2
n ti ti T T
+ G t i G' ti ---------- E t
t 2
G ti G' t i i
i = 1 t1 i t2 i t2 i
----------------------- 1 --------------
-
2 2
ti ti
The right choice of the penalty factors is important to get accurate results in an acceptable number of iterations. If the
penalty factor for normal contact, E n , is too small, then there will be a relatively large amount of penetration, which
has to be removed by the augmentation process outlined above. However, if the penalty factor is small, then the first
order correction process illustrated in Figure 9-35 slowly converges. On the other hand, if the penalty factor is too large,
the overall system of equations may be ill-conditioned and a non-smooth contact stress field may be found.
The default value of the penalty factor E n i depends on the body stiffness of the two contact bodies involved and a
characteristic length (note that the dimension of the penalty factor is force per cubic length). The body stiffness is either
defined by the average trace of the initial stress-strain law of the elements defining the two contact bodies or by the
average bulk modulus for (nearly) incompressible rubber materials, whichever of the two is the largest. For continuum
Main Index
Chapter 9: Contact 435
Segment-to-Segment Contact Theory
elements, the characteristic length is given by one half of the average length of all the edges being part of the contact
boundary. For shell elements, the characteristic length is given by half of the average thickness of all the shell elements
being part of a contact body. When there is contact between a solid and a shell element, then the characteristic length
is defined by the shell element. If polyline/polygon point i defines contact between the deformable bodies k and l ,
with body stuffiness S k and S l , and the characteristic length of the model is L , then the default value of E n i is given
by:
0.5 S k + S l
E n i = ------------------------------ (9-41)
L
In case of contact with a rigid body, the equations shown before can be applied with minor modifications. First, the
displacements G' n i U and G' ti U of the closest point projection directly follow from the motion of the rigid body.
Second, the issue of an ill-conditioned system does not easily occur and the default penalty factor can be chosen to be
larger. Since there is no body stiffness associated with a rigid body, the default value of E n i is related to the deformable
body only and given by:
1000S k
E n i = ----------------- (9-42)
L
For tangential contact, the default penalty factor E t i is related to the penalty factor E n i according to:
E t i = E n i (9-43)
3 6
where = 1 10 for deformable contact and = 1 10 for rigid contact.
As mentioned before, you can define the penalty factors. This can be done either globally via the BCPARA bulk data
entry or per body combination via the BCTABLE/BCTABL1 bulk data entry. The penalty factors are by default printed
in the output file.
Main Index
436 Nonlinear Users Guide
Segment-to-Segment Contact Theory
Note that the last term in equation (9-44) results from the augmentation procedure.
Since there are in general more polyline/polygon points on a contact segment than there are nodes on such a segment,
enforcing a zero gap function everywhere would in general result in an over-constrained system. To this end, a
reduction of the penetration field has to be applied. The following options are available:
No augmentation. This results in a pure penalty method and is recommended for most analyses, since it gives
reasonably accurate results in a relative small number of iterations.
Augmentation based on a constant penetration field. This is the recommended procedure for linear finite
elements.
Augmentation based on a (bi-)linear penetration field. This procedure should be used only for quadratic finite
elements.
It is also possible to let the program decide which of the augmentation methods should be applied (constant or linear).
Then the choice is made based on the combination of elements corresponding to the contacting segments. It is noted
that in case of contact with a rigid body, the augmentation procedure is always applied.
In case of augmentation based on a constant penetration field, the average gap function per contact segment is
determined and applied to each polyline/polygon point corresponding to that segment.
In case of augmentation based on a (bi-)linear penetration field, a least squares method is used to determine the values
of the gap function in the corner nodes of the segment. Based on the position of a polyline/polygon point on this
segment, a (bi-)linear interpolation between the corner values then gives the local value in the point.
Since an exact fulfillment of a zero gap function is difficult to achieve, a threshold value is needed to determine whether
or not an extra iteration due to the augmentation procedure is needed. This penetration distance beyond which an
augmentation will be applied can be calculated by the program or entered by you. The default penetration distance
beyond which an augmentation will be applied is defined as:
a u g = 0.001L (9-45)
where the characteristic length L is the same as used in equation (9-41). Similar to the penalty factor, you can define
the augmentation distance either globally for the analysis on the BCPARA bulk data entry or per body pair on the
BCTABLE/BCTABL1 bulk data entry.
If an augmentation procedure is selected and the maximum penetration is beyond a u g , then an extra Newton-Raphson
iteration will take place, as long as the maximum number of iterations has not been exceeded. If the maximum number
of iterations has been reached, then the program will continue with the next increment if the usual convergence criteria
have been fulfilled; so the analysis will not stop because of the augmentation procedure. In case of friction with
augmentation, the same logic is applied to the tangential behavior.
Separation
Since in each polyline/polygon point the contact normal stress is calculated, separation will always be based on
stresses. By default, if the contact normal stress is in tension, then the corresponding polyline/polygon point will
separate, which implies that this point will no longer contribute to the global stiffness matrix and force vector.
Main Index
Chapter 9: Contact 437
Segment-to-Segment Contact Theory
In order to speed up the separation process on a structural level, there is an extra check performed on a contact segment
basis. If there are points on a segment which separate, then the total contact normal force F n t due to points on this
segment being in tension (negative) is compared with the total contact normal force F n c due to points on this segment
being in compression (positive). These forces are obtained by integrating the contact normal pressure over the
corresponding contact area. If F n c 0.75 F n t , then all the points on this segment will separate. This extra check is
only active if the separation stress is default (zero). As soon as you enter a nonzero separation stress, either globally
for the analysis on the BCPARA bulk data entry or per body pair on the BCTABLE/BCTABL1 bulk data entry, this extra
check is skipped.
Beam Contact
Beam contact can be easily analyzed using the segment-to-segment contact procedure. For the element stiffness and
stress recovery, the beam elements are used as is, but for contact purposes, the beams are exactly represented by
internally expanding them into three-dimensional patches. The contact forces and constraints are automatically
transferred from the patches to the beam nodes.
For 3-D beam elements, the line representation of the beam contact body is automatically replaced by an expanded
representation of the beam contact body based off the section properties. All shapes of the PBARL and PBEAML Bulk
Data entries are supported. In addition, PBAR and PBEAM are supported and replaced by rectangular or circular shapes
when the resultant beam values are equivalent to the rectangular shape or the circular shapes.
A sample expanded representation of nine beam contact bodies comprising of various beam sections is shown in
Figure 9-39. This expanded form is automatically created internally from the given section and orientation data.
Figure 9-39 Expanded Representation of Nine Beam Contact Bodies with Different Beam Sections
The following general points should be kept in mind for the expanded beam bodies:
The expanded beam representation is the only possible scheme available for 3-D beams in segment-segment
contact.
Main Index
438 Nonlinear Users Guide
Segment-to-Segment Contact Theory
Applications of beam contact in the expanded mode include tube-tube contact (internal contact between
beams), beam-beam contact (external contact between beams), beam contact with solids, beam contact with
shells, and beam contact with rigid bodies. This form of beam contact can be used for statics and dynamics.
The expanded beam representation takes into account the current position of the beam based on nodal
coordinates, the beam cross-sectional properties, the beam orientation, and beam offsets at the nodes (if any).
Geometrically, beam patches are only used for contact detection, separation and penetration. Internal rigid
links are automatically generated from each beam node to the generated beam patch nodes. No additional
degrees of freedom are introduced and the movement of the beam patch nodes is completely based on the
translations and rotations of the parent beam nodes. The contact stiffness and forces are calculated at the
patches and transferred to the original beam nodes using the patch vectors.
BCPFLG is an optional Bulk Data entry to select faces of the cross section for the expanded beams in segment-
segment contact. Without this Bulk Data entry, all sections of beams are considered for the contact. This
option will help reducing the number of contact segments of beams and improve the simulation time
performance.
For the contact at the tip of beams in the segment to segment contact, the cap of beams must be set (IESCAP
option in BCSCAP Bulk Data entry).
The distance tolerance calculation for the expanded beam sections incorporates two schemes. In the first, the
normal evaluation (similar to node-to-segment contact for beam elements) is done. In addition, a second
evaluation based on the detailed section properties is carried out. The minimum of the node-to-segment
scheme and the section property scheme is the final distance tolerance. The section property based distance
tolerance evaluation is based on 0.25*minimum of patch edge length and section thickness. Note that the
minimum patch edge length is relevant for all sections, but the minimum section thickness is only relevant for
thin-walled sections when both outer and inner patches are flagged to be created.
The local direction cosine matrix at each beam node is used to establish the 3-D patch vector directions. The
direction cosine matrix for each beam element comprises of the local Z axis along the beam element and the
local X-Y axes in the cross-sectional plane of the element. At each beam node, the nodal direction cosine
matrix is computed as the average of the matrices of two adjacent beam elements meeting at the node. This
average direction cosine matrix is established only if the angle between the local Z axes of the two elements
and the angle between the local X axes of the two elements is less than a certain value (160).
Limitations of the expanded beam representation include the following:
The beam cross-section does not change during the simulation. This means that while the beam patches can
undergo finite translations and rotations, the cross section does not change for large strains as well as contact
interference.
Multiply connected beams are not possible. The analysis will continue with a warning but the neighbor
determination for each of the beams is based on the assumption of a singly connected beam with one
neighbor each at the start and end nodes of the beam element.
Interference Fit
This feature can be used to simulate situations where bodies with overlaps or gaps in the geometry are required to fit
together. Some examples of interference fit usage include spur gears that are mounted tightly on a shaft, couplings,
Main Index
Chapter 9: Contact 439
Segment-to-Segment Contact Theory
collar rings, tire mounted on wheel rim, bearing bushings in hubs, valve seats, rubber seals, bushings made of synthetic
resin pressed material, etc.
The interference fit capability is treated in MSC Nastran using the contact algorithm. The penetration associated with
interference fit can be much larger compared to traditional contact analysis. Ideally, interference fit computations are
performed in a separate load case, which might take more than a one increment. The user can control the number of
increments or time required to accomplish the fit. The user is equipped with many methods which can be adopted
depending on the geometry. There are four methods available in Nastran for the interference fit option. These are
applicable to both Node-to-Segment Contact (N2S) and Segment-to-Segment Contact (S2S). These methods are as
follows:
1. Contact normal
2. Translation
3. Scaling
4. Automatic
1. Contact Normal
This method is generally recommended for situations with small values of interference to be resolved along the normal
direction to the touched interface. The other methods are more suitable when the interference is large. In this case, the
nodes (N2S) or auxiliary points (S2S) of the touching body are projected in a direction normal to the segments of the
touched body. The individual touching entities are positioned at a certain distance (overclosure or gap) from the
touched surface along the normal of the touched surface. The user needs to provide this interference distance
(CINTERF option in BCONPRG card) which signifies the maximum overlap or gap between the bodies. This provided
value is specified as a negative value for an overlap and as a positive value for a gap. The detection algorithm works
as follows:
(Overlap/Gap + Interference) at a node (Node to segment contact) or auxiliary point (Segment to segment contact) <
D (1+B) on the inside or < D (1-B) on the outside
where
D = regular distance tolerance and B = bias factor.
To control the number of increments or time during which the interference fit is achieved, the user can optionally
specify a table in which the magnitude of interference distance is ramped down from the possible maximum value of
interference to zero. Interference can be gradually resolved. Note that if a clearance or overlap needs to be maintained
for the entire job, then the interference value needs to be maintained as a constant value in all the load cases of the job.
This method may not perform adequately when the interference distance is large in this case, the touching entities
may find multiple touched segments within the given closures distance.
Methods 2 to 4 are more general and allow larger amounts of interference (overlap or gap) between the touching and
touched bodies. The general scheme for these methods is to internally compute an initial shift vector between the nodes
of the touching body and the segments of the touched body. This initial vector is also referred to as a pseudo
displacement vector. The following points should be noted about the pseudo displacement vector:
Main Index
440 Nonlinear Users Guide
Segment-to-Segment Contact Theory
The magnitude of the pseudo movement is varied based on a table provided by the user, which is normally a
ramp down table from unity to zero over the load case. Note that the projection of the touching entity to a
touched segment is done for each iteration but the position of the touching entity is always based on a scaling
of the pseudo displacement vector computed at the start of the load case. Therefore, large rotations of the
interfering bodies should not be permitted during the interference load case.
The treatment of the interference across two successive load cases should be noted. Note that the interference
can be specified for any load case, though normally, it is specified at the start of the analysis (zeroth increment
and first load case). Three situations are possible:
The interference is specified in the zeroth increment and also in the first load case this is the recommended
option. During the zeroth increment, the contact between the interfering bodies is established based on the
given data but the interference itself is not resolved. The actual interference is resolved in the first load case.
The interference is not specified in the zeroth increment but specified in the first load case this is not
recommended. In this case, it is possible that some contact is detected between the two bodies at increment
0 based on the regular contact algorithm and this may cause a conflict with the pseudo displacement
contact in load case 1.
The interference is specified in the zeroth increment but not specified in the first load case this is not
recommended. In this case, the contact between the interfering bodies is established at increment 0 and the
computed pseudo displacement is immediately reduced to 0.0 at the start of increment 1. This may cause
convergence difficulties. Each interference method provides a different scheme for the computation of this
pseudo vector.
2. Translation
In this method, for each interfering body pair in the contact definition, the user specifies the following:
Magnitude of the vector
Direction cosines of the vector
The coordinate system in which the direction cosines are specified (default is the global Cartesian system)
The body to which to apply the interference vector and
A table specifying the variation of the interference vector during the load case
The algorithm performs an initial pseudo movement of the interfering body along the direction specified by the user
such that the overclosure is removed. This is followed by a removal of the pseudo movement such that the interference
fit is achieved. This method is especially useful for large initial over-closure between contacting bodies in a specific
direction.
3. Scaling
In this method, for each interfering body pair in the contact table, the user specifies the following:
Position of the Centroid about which the scaling is to occur
X, Y and Z Scaling Factors
The coordinate system in which the scaling should occur (default is the global Cartesian system)
The body to which to apply the interference vector and
Main Index
Chapter 9: Contact 441
Segment-to-Segment Contact Theory
A table specifying the variation of the interference resolution during the load case
The pseudo adjustment is done by scaling one of the bodies such that the overlap is removed. The rest of the algorithm
is similar to that of the Translation option.
4. Automatic
In this method, for each interfering body pair in the contact table, the user specifies the following:
A penetration tolerance
The body to which to apply the interference vector and
A table specifying the variation of the interference resolution during the load case
The overlap distance vectors are found between the nodes of the overlapping body (specified by the user) into the other
body. The penetration tolerance should be a little more than the highest penetration between contact body pair. Nodes
within this tolerance are considered for the interference calculation. The pseudo displacement vector at each node is
established by taking the perpendicular distance between the node and the closest touched segment. The rest of the
algorithm is similar to that of the Translation option. The following additional points should be noted for the
AUTOMATIC method:
In single sided contact, touching body has to be selected as the interfering body.
In general, use of analytical boundary description for the touched body is recommended while using this
method.
For shell contact bodies, user needs to specify the top or bottom surface for detecting contact.
Initial Gap
This option can be used to define the initial gap or overlap between contacting bodies.
A uniform gap between the contacting bodies can be specified using BCONPRG Bulk Data entry. The initial gap option
can only be used at the beginning of analysis; hence, it has to be specified in the initial contact definition. Unlike the
stress free projection option, the contacting node coordinates are not modified when using the initial gap option.
Following are the inputs required to define the initial gap for a contact pair.
1. Gap/overlap magnitude (Positive value indicates gap between contact bodies, and negative value indicates
overclosure between contact bodies)
2. Search tolerance
3. Contact body to be adjusted (required for double sided contact)
With the initial gap option, the nodes of the touching body are projected onto the nearest segment of the touched body
to get the projection vectors at the touching nodes. In case of double-sided contact, the user must specify the contact
Main Index
442 Nonlinear Users Guide
Segment-to-Segment Contact Theory
body, the nodes of which will be projected to obtain the projection vector. Search for the contact is limited to the search
tolerance zone. The projection vector at a touching body node is then modified to accommodate the user-specified gap
or overclosure. During the analysis, the projection vector is continuously updated based on the displacement and the
rotation of the associated node. In case of a gap between the contacting bodies, the nodes are not in contact at the
beginning of the analysis. As the body moves towards each other, contact is detected. In case of an overlap between
the contacting bodies, the interference is resolved in the first increment of the load case.
Following points should be noted to avoid unwanted contact detection:
1. Comparatively finer meshed body should be selected for the initial gap adjustment.
2. Use of analytical boundary description is recommended for the faceted touched segments.
3. In case of penetrating shell bodies, a top or bottom surface should be defined using boundary redefinition option
(COPTS/COPTM) available in BCONPRG.
Main Index
Chapter 9: Contact 443
References
References
1. Oden, J. T. and Pires, E. B. Nonlocal and Nonlinear Friction Laws and Variational Principles for Contact
Problems in Elasticity, J. of Applied Mechanics, V. 50, 1983.
2. Ju, J. W. and Taylor, R. L. A perturbed Lagrangian formulation for the finite element solution of nonlinear
frictional contact problems, J. De Mechanique Theorique et Appliquee, Special issue, Supplement, 7, 1988.
3. Simo, J. C. and Laursen, T. A. An Augmented Lagranian treatment of contact problems involving friction,
Computers and Structures, 42, 1002.
4. Peric, D. J. and Owen, D. R. J. Computational Model for 3D contact problems with friction based on the
Penalty Method, Int. J. of Meth. Engg., V. 35, 1992.
5. Taylor, R. L., Carpenter, N. J., and Katona, M. G. Lagrange constraints for transient finite element surface
contact, Int. J. Num. Meth. Engg., 32, 1991.
6. Wertheimer, T. B. Numerical Simulation Metal Sheet Forming Processes, VDI BERICHET, Zurich,
Switzerland, 1991
Main Index
444 Nonlinear Users Guide
References
Main Index
Chapter 10: Materials
Nonlinear Users Guide
10 Materials
Material Model Overview 447
Linear Elastic 450
Linear Elastic Material Overview 451
Isotropic Material (MAT1) 452
Two-Dimensional Anisotropic Material (MAT2) 454
Axisymmetric Solid Orthotropic Material (MAT3) 456
Two-Dimensional Orthotropic Material (MAT8) 457
Three-Dimensional Orthotropic Material (MATORT) 459
Simplified Nonlinear Orthotropic Material (MATSORT) 463
Classical Laminate Theory (PSHELL) 464
Composite Material (PCOMP or PCOMPG) 466
Nonlinear Material Overview 473
Nonlinear Elastic 476
MATS1 Model 508
Inelastic Behavior 512
Failure and Damage Models (MATF) 542
Progressive Composite Failure 551
Main Index
446 Nonlinear Users Guide
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D
Composite Brick Elements 557
Cohesive Zone Modeling (MCOHE) 560
Creep (MATVP, CREEP) 565
Gasket (MATG) 574
Thermo-Mechanical Shape Memory Model (MATSMA) 579
Mechanical Shape Memory Model (MATSMA) 589
Material Orientation 594
Heat Transfer Properties 596
User Defined Material Subroutines 597
Experimental Data Fitting 599
References 614
Main Index
Chapter 10: Materials 447
Material Model Overview
Main Index
448 Nonlinear Users Guide
Material Model Overview
Main Index
Chapter 10: Materials 449
Material Model Overview
Main Index
450 Nonlinear Users Guide
Linear Elastic
Linear Elastic
SOL 400 is capable of handling problems with any combination of isotropic, orthrotropic, or anisotropic linear elastic
material behavior.
The isotropic linear elastic model is the model most commonly used to represent engineering materials. This model,
which has a linear relationship between stresses and strains, is represented by Hookes Law. Figure 10-1 shows that
stress is proportional to strain in a uniaxial tension test. The ratio of stress to strain is the familiar definition of modulus
of elasticity (Youngs modulus) of the material.
E (modulus of elasticity) = (axial stress)/(axial strain) (10-1)
Stress
Strain
Experiments show that axial elongation is always accompanied by lateral contraction of the bar. The ratio for a linear
elastic material is:
Main Index
Chapter 10: Materials 451
Linear Elastic Material Overview
Material Characteristics
Isotropic (MAT1) Isotropic material No Directional Dependency most commonly used
material property
Two Dimensional Anisotropic General anisotropic two-dimensional material used with plate and shell
(MAT2) elements
Axi-symmetric / Solid Orthotropic Orthotropic three-dimensional material for use with CTRIAX6, PSHLN2, or
(MAT3) PLCOMP
Two Dimensional Orthotropic Two-dimensional orthotropic stress-strain used with the plate and shell
(MAT8) elements
Three-Dimensional Anisotropic Orthrotropic material property for the CHEXA, CPENTA, and CTETRA solid
Material (MATORT) elements and solid shell and plane strain element
Main Index
452 Nonlinear Users Guide
Isotropic Material (MAT1)
E
G = --------------------- (10-3)
21 +
If you enter all three constants and they do not satisfy this relationship, all three constants are used, but that may not
be your intention. Therefore, a warning message is printed indicating that the isotropic relationship has been violated.
It is recommended that you only input two of the three constants. For line elements, E is the modulus of elasticity, and
G is the shear modulus for torsion and transverse shear if it is present in the element. For plate and solid elements, E ,
G , and are used to develop a material matrix for the element.
The MAT1 entry may also be used to define the mass density, coefficient of thermal expansion, and stress limits. The
mass properties are only required in static analysis when a gravity loading or rotating force is used; however, they are
useful for model checkout with any loading condition (of course, they are very important for dynamic analysis). The
format of the MAT1 bulk data entry in the QRG is as follows:
MAT1
1 2 3 4 5 6 7 8 9 10
MAT1 MID E G NU RHO A TREF GE
ST SC SS MCSID
Field Contents
MID Material identification number
E Youngs modulus
G Shear modulus
NU Poissons ratio
RHO Mass density
A Thermal expansion coefficient
TREF Reference temperature
ST, SC, SS Stress limits for tension, compression, and shear (It is recommended to use MATF to define
stress limits and failure critera)
MCSID Material coordinate system identification number
Note that GE shown in field 9 was not described. The GE is a material damping that has no significance for static
analysis. Information regarding material damping can be found in the MSC Nastran Dynamic Analysis Users Guide.
Main Index
Chapter 10: Materials 453
Isotropic Material (MAT1)
Main Index
454 Nonlinear Users Guide
Two-Dimensional Anisotropic Material (MAT2)
me
x G 11 G 12 G 13 x
me
y = G 12 G 22 G 23 y (10-4)
me
xy G 13 G 23 G 33 x y
The me superscripts imply this is the elastic-mechanical strain and does not include the thermal or inelastic strains.
E1 21 E 1 0
1
C = ------------------------------- 12 E 2 E2 0 (10-5)
1 12 21
0 0 1 12 21 G
The format of the MAT2 bulk data entry in the QRG is as follows:
MAT2
1 2 3 4 5 6 7 8 9 10
MAT2 MID G11 G12 G13 G22 G23 G33 RHO
A1 A2 A3 TREF GE ST SC SS
MCSID
Field Contents
MID Material identification number
Gij The material property matrix
RHO Mass density
Ai Thermal expansion coefficient vector
TREF Reference temperature
ST, SC, SS Stress limits for tension, compression, and shear (It is recommended to use MATF to define
stress limits and failure critera)
MCSID Material coordinate system identification number
Main Index
Chapter 10: Materials 455
Two-Dimensional Anisotropic Material (MAT2)
When you perform composite analysis using the PCOMP/PCOMPG entries, MAT2 entries are generated automatically.
The PCOMP/PCOMPG entry is discussed in MSC Nastran Linear Static Analysis Users Guide, Chapter 5 Material
Properties and Composites, The Composite Element (PCOMP or PCOMPG).
Main Index
456 Nonlinear Users Guide
Axisymmetric Solid Orthotropic Material (MAT3)
1- x zx
----- ----------- ----------- 0
Ex E Ez
me
x x z x
1
me ----------- ------ ----------- 0
Ex E Ez
me
= (10-6)
z xz z 1 z
----------- ----------- ----- 0
me
zx Ex E Ez z x
1-
--------
0 0 0
G zx
x x xz zx z z
-------- = -------
-; ------- = ------- ; -------- = -------- (10-7)
Ex E Ex Ez E Ez
The format of the MAT3 bulk data entry in the QRG is as follows:
MAT3
1 2 3 4 5 6 7 8 9 10
MAT3 MID EX ETH EZ NUXTH NUTHZ NUZX RHO
GZX AX ATH AZ TREF GE
Field Contents
MID Material identification number
EX, ETH, EZ Youngs moduli in the x, , and z directions, respectively
NUXTH, NUTHZ, NUZX Poissons ratios
RHO Mass density
GZX Shear modulus
AX, ATH, AZ Thermal expansion coefficients
TREF Reference temperature
GE Structural element damping coefficient.
Main Index
Chapter 10: Materials 457
Two-Dimensional Orthotropic Material (MAT8)
1- 12
----- ----------- 0
me E1 E1
1 1
me 12 1
2 = ----------- ------ 0 2 (10-8)
me E1 E2
12 12
1
---------
0 0
G 12
1 z G1 z 0 1 z
= (10-9)
2 z 0 G2 z 2 z
The format of the MAT8 bulk data entry in the QRG is as follows:
MAT8
1 2 3 4 5 6 7 8 9 10
MAT8 MID E1 E2 NU12 G12 G1Z G2Z RHO
A1 A2 TREF Xt Xc Yt Yc S
GE F12 STRN
Field Contents
MID Material identification number.
E1 Modulus of elasticity in the longitudinal direction.
E2 Modulus of elasticity in the lateral direction.
NU12 Poissons ratio ( 2 1 for uniaxial loading in 1-direction).
Main Index
458 Nonlinear Users Guide
Two-Dimensional Orthotropic Material (MAT8)
Field Contents
Xt, Xc Allowable stresses or strains in tension and compression, respectively, in the longitudinal direction.
Yt, Yc Allowable stresses or strains in tension and compression, respectively, in the lateral direction.
S Allowable stress or strain for in-plane shear.
GE Structural damping coefficient.
F12 Interaction term used in the Tsai-Wu theory.
STRN Request for maximum strain theory.
Main Index
Chapter 10: Materials 459
Three-Dimensional Orthotropic Material (MATORT)
x xm e
y ym e
z zm e
= G (10-10)
xy xme
y
y z yme
z
zx me
zx
The compliance matrix is dependent on the element type an is defined in Table 10-2
x x 1 E xx y x E yy xx
=
y y xy E x x 1. E y y yy
yx = xy E y y E xx
Class 3 NDI = 1,NSHEAR = 1
Beam Elements with transverse shear
1 E xx 0
=
0 1 G xy
Main Index
460 Nonlinear Users Guide
Three-Dimensional Orthotropic Material (MATORT)
xx 1 E xx y x E yy 0 xx
yy = xy E x x 1 E y y 0 yy
x y 0 0 1 G xy x y
yx = xy E yy E xx
Class 5 NDI = 2,NSHEAR = 1
Thick Axisymmetric Shells
mm 1 E mm m E 0 mm
= m E m m 1 E 0
T 0 0 1 Gm T
x x 1 E xx y x E yy zx E zz 0 x x
y y xy E x x 1 E y y zy E zz 0 y y
=
zz x z E xx yz E y y 1 E zz 0 zz
x y 0 0 0 1 G x y xy
y x = x y E yy E xx zy = yz E zz E yy xz = zx E xx E z z
Class 7 NDI = 2,NSHEAR = 3
Thick Shell, Elements
x x 1 E x x y x E yy 0 0 0 x x
y y xy E x x 1 E yy 0 0 0 y y
x y = 0 0 1 Gx y 0 0 xy
yz 0 0 0 1 Gy z 0 y z
zx 0 0 0 0 1 G zx z x
yx = xy E y y E xx
Main Index
Chapter 10: Materials 461
Three-Dimensional Orthotropic Material (MATORT)
x x 1 E xx y x E yy z x E zz 0 0 0 xx
yy xy E x x 1 E y y z y E zz 0 0 0 yy
zz x z E xx yz E y y 1. E zz 0 0 0 zz
=
xy 0 0 0 1 Gx y 0 0 xy
yz 0 0 0 0 1 Gy z 0 yz
zx 0 0 0 0 0 1 G zx zx
The format of the MATORT bulk data entry in the QRG is as follows:
MATORT
1 2 3 4 5 6 7 8 9 10
MATORT MID E1 E2 E3 NU12 NU23 NU31 RHO
G12 G23 G31 A1 A2 A3 TREF GE
IYLD IHARD SY Sornl Y1 Y2 Y3 N/A
Yshr1 Yshr2 Yshr3 N/A N/A N/A N/A N/A
Field Contents
MID Identification number of a MATORT entry. (Integer > 0; No Default)
E1 Modulus of elasticity in longitudinal or 1-direction. (Real > 0)
E2 Modulus of elasticity in lateral direction or 2-direction. (Real > 0; no Default; must be entered)
E3 Modulus of elasticity in thickness direction or 3-direction. (Real > 0; no Default; must be entered)
NU12 Poissons ratio 2 1 for uniaxial loading in 1-direction. (Real; no Default; must be entered)
NU23 Poissons ratio 3 2 for uniaxial loading in 2-direction. (Real; no Default; must be entered)
NU31 Poissons ratio 1 3 for uniaxial loading in 3-direction. (Real; Default = NU23)
RHO Mass density. (Real; Default = 0.0)
G12 Shear modulus in plane 1-2. (Real > 0; no Default; must be entered)
G23 Shear modulus in plane 2-3. (Real > 0.; no Default; must be entered)
G31 Shear modulus in plane 3-1. (Real > 0; no Default; must be entered)
Ai Coefficient of thermal expansion in i-direction (Real; Default = 0.0).
Main Index
462 Nonlinear Users Guide
Three-Dimensional Orthotropic Material (MATORT)
Field Contents
TREF Reference temperature at which the thermal expansion coefficient is measured. TREF is used only
if the thermal expansion coefficient is temperature-dependent. (Real; Default = 0.0)
GE Structural damping coefficient. (Real; Default = 0.0).
IYLD Integer pertaining to one of the following yield criterion: (Integer)
-1 = Elastic
1 = von Mises (Default)
7 = Hill (1948) yield
8 = Barlat (1991) yield
IHARD Integer pertaining to one of the following work hardening rules: (Integer)
1 = Isotropic (Default)
2 = Kinematic
3 = Combined Isotropic/Kinematic
SY Equivalent (von Mises) tensile yield stress. (Real > 0.0 or blank; Default = 0.0)
SORNL Not used
Y1 Hills yield stress ratio in direction 1. (Real > 0.0 or blank; Default = 0.0)
Y2 Hills yield stress ratio in direction 2. (Real > 0.0 or blank; Default = 0.0)
Y3 Hills yield stress ratio in direction 3. (Real > 0.0 or blank; Default = 0.0)
Yshr1 Hills yield shear stress ratio in direction 1. (Real > 0.0 or blank; Default = 0.0)
Yshr2 Hills yield shear stress ratio in direction 2. (Real > 0.0 or blank; Default = 0.0)
Yshr3 Hills yield shear stress ratio in direction 3. (Real > 0.0 or blank; Default = 0.0)
When you use MATORT, it is advisable to define a material coordinate system on the PSOLID entry (using field 4). For
solid elements, stresses are output in the material coordinate system, which by default, is the basic coordinate system.
In general, for solid elements, it is not easy to determine the orientation of the element coordinate system.
As an example of the usage of MAT8, see Laminated Strip under Three-point Bending (Ch. 6) in the MSC Nastran
Demonstration Problems Manual.
Main Index
Chapter 10: Materials 463
Simplified Nonlinear Orthotropic Material (MATSORT)
Main Index
464 Nonlinear Users Guide
Classical Laminate Theory (PSHELL)
1 2 3 4 5 6 7 8 9 10
PSHELL PID MID1 T MID2 12I/T3 MID3 TS/T NSM
Z1 Z2 MID4
Field Contents
PID Property identification number.
MIDi Material identification numbers.
T Default membrane thickness for Ti on the connection entry or average thickness if TFLAG = 1
(Field 13 of plate element), then T i = T i T .
On the PSHELL entry, there are four material ID fields available as shown in Table 10-3.
If you wish to include membranes only, use MID1 only. If you wish to include bending only, use MID2 and, optionally,
MID3. When MID3 is used, thick shell theory (which includes the transverse shear flexibility) is used to develop the
element stiffness matrix. In general, for thin structures, it is advisable to not use MID3.
Main Index
Chapter 10: Materials 465
Classical Laminate Theory (PSHELL)
The material type used for any of the material fields (MID1, MID2, MID3, or MID4) may be MAT1, MAT2, or MAT8.
The PCOMP/PCOMPG entry, which is the subject of the next section, provides a convenient way of entering
appropriate material properties for a shell consisting of layers of unidirectional fibers.
The MID4 field on the PSHELL entry is used to input a coupling relationship between the in-plane forces and bending
moments. This coupling only occurs in shells that are nonsymmetric about the neutral plane or where the neutral plane
is offset from the grid points. Typical applications include reinforced skins and aluminum bonded to fiberglass. The
MID4 field should be left blank if the element cross section is symmetric.
Main Index
466 Nonlinear Users Guide
Composite Material (PCOMP or PCOMPG)
Main Index
Chapter 10: Materials 467
Composite Material (PCOMP or PCOMPG)
*
*
*
*
Beams or Shells with
PCOMP or PCOMPG Options
Figure 10-3 Integration Points through the Thickness of Beam and Shell Elements
Integration Points through the Thickness of Continuum Elements shows the location of integration points through the
thickness of continuum elements. SOL 400 forms the element stiffness matrix by performing numerical integration
based on the standard isoparametric concept.
* *
* *
* *
* *
Figure 10-4 Integration Points through the Thickness of Continuum Composite Elements
PCOMP
1 2 3 4 5 6 7 8 9 10
PCOMP PID Z0 NSM SB FT TREF GE LAM
MID1 T1 THETA1 SOUT1 MID2 T2 THETA2 SOUT2
MID3 T3 THETA3 SOUT3 etc.
Field Contents
PID Property identification number.
Z0 Distance from the reference plane to the bottom surface.
NSM Nonstructural mass per unit area.
Main Index
468 Nonlinear Users Guide
Composite Material (PCOMP or PCOMPG)
Field Contents
SB Allowable shear stress of the bonding material.
FT Failure theory.
TREF Reference temperature.
LAM Blank, SYM, MEM, BEND option.
MIDi Material ID of the various plies. The plies are identified by serially numbering them from 1 at the
bottom layer.
Ti Thicknesses of the various plies.
THETAi Orientation angle of the longitudinal direction of each ply with the material axis of the element.
SOUTi Stress or strain output request.
An alternative to the PCOMP entry is the PCOMPG entry. The PCOMPG entry includes a global ply ID, so it is easier
to track the output for the same ply across the model. The format for the PCOMPG bulk data entry in the QRG is as
follows:
PCOMPG
1 2 3 4 5 6 7 8 9 10
PCOMPG PID Z0 NSM SB FT TREF GE LAM
GPLYID1 MID1 T1 THETA1 SOUT1
GPLYID2 MID2 T2 THETA2 SOUT2
Field Contents
PID Property identification number. (0 < Integer < 10000000)
Z0 Distance from the reference plane to the bottom surface. (Real; Default = -0.5 times the element
thickness.)
NSM Nonstructural mass per unit area. (Real)
SB Allowable shear stress of the bonding material (allowable interlaminar shear stress). Required if FT is
also specified. (Real > 0.0)
FT Failure theory. The following theories are allowed (Character or blank. If blank, then no failure
calculation will be performed)
HILL for the Hill theory.
HOFF for the Hoffman theory.
TSAI for the Tsai-Wu theory.
STRN for the Maximum Strain theory.
TREF Reference temperature. (Real; Default = 0.0)
GE Damping coefficient. (Real; Default = 0.0)
Main Index
Chapter 10: Materials 469
Composite Material (PCOMP or PCOMPG)
Field Contents
LAM Laminate Options. (Character or blank, Default = blank).
Blank All plies must be specified and all stiffness terms are developed.
MEM All plies must be specified, but only membrane terms (MID1 on the derived PSHELL
entry) are computed.
BEND All plies must be specified, but only bending terms (MID2 on the derived PSHELL
entry) are computed.
SMEAR All plies must be specified, stacking sequence is ignored MID1=MID2 on the derived
PSHELL entry and MID3, MID4, TS/T, and 12I/T**3 terms are set to zero).
SMCORE All plies must be specified, with the last ply specifying core properties and the previous
plies specifying face sheet properties. The stiffness matrix is computed by placing half
the face sheet thicknesses above the core and the other half below with the result that
the laminate is symmetric about the mid-plane of the core. Stacking sequence is ignored
in calculating the face sheet stiffness.
GPLYIDi User-defined Global (External) Ply ID. The global ply identification number should be unique with
respect to other plies in the entry. The plies are defined in stacking sequence starting with the bottom
layer. (Integer > 0)
MIDi Material ID of the various plies. The plies are identified by serially numbering them from 1 at the
bottom layer. The MIDs must refer to MAT1, MAT2, or MAT8 bulk data entries. (Integer > 0 or blank,
except MID1 must be specified.)
Ti Thicknesses of the various plies. (Real or blank, except T1 must be specified.)
THETAi Orientation angle of the longitudinal direction of each ply with the material axis of the element. (If the
material angle on the element connection entry is 0.0, the material axis and side 1-2 of the element
coincide.) The plies are to be numbered serially starting with 1 at the bottom layer. The bottom layer is
defined as the surface with the largest -Z value in the element coordinate system. (Real; Default = 0.0)
SOUTi Stress or strain output request. (Character: YES or NO; Default = NO)
Using the material properties for each of the lamina with conventional shell elements, MSC Nastran calculates the
equivalent PSHELL and MAT2 entries generated as shown in Figure 10-5. If nonlinear material behavior including
progressive failure occurs, then the advanced elements should be activated using the PSHLN1 bulk data entry.
Main Index
470 Nonlinear Users Guide
Composite Material (PCOMP or PCOMPG)
CQUAD4
PCOMP or
PCOMPG
EQUIV PSHELL
Output if
ECHO = SORT
MID1 MID2 MID3 MID4 (the default)
If you want stress and/or the failure indices for the composite elements, ELSTRESS must be requested in the case
control section for the appropriate elements. Also, if you want the failure index table, you must enter the stress limits
for each lamina on the appropriate material entry, the shear stress limit Sb, and the failure theory method FT on the
PCOMP/PCOMPG entries.
As shown in Figure 10-5, each PCOMP/PCOMPG (together with the material entries for each lamina) is processed to
form an equivalent PSHELL and four MAT2 entries. To print the equivalent PSHELL and MAT2 entries in the output
file, use NASTRAN PRTPCOMP=1 provided that ECHO=NONE is not set. If you wish to use them for future runs, they
may be punched to an ASCII file. The ID numbers of the MAT2s are important. The MID1 material has an ID in the
range of 100000000 to 199999999. The MID2 material has an ID in the range of 200000000 to 299999999, etc. These
ranges are used to inform MSC Nastran that the material is part of a composite analysis. If you are using the equivalent
properties in a future analysis instead of using the PCOMP/PCOMPG entries and you are entering a thermal coefficient
of expansion, do not change the ID numbers. Also, if you use the equivalent PSHELL and MAT2s, you will not be able
Main Index
Chapter 10: Materials 471
Composite Material (PCOMP or PCOMPG)
to obtain the laminae stress or the failure index table. For the failure index, you have a choice of four failure theories:
Hills theory, Hoffmans theory, Tsai-Wus theory, and the maximum strain theory.
As an example of the usage of PCOMP, see Laminated Strip under Three-point Bending (Ch. 6) in the MSC Nastran
Demonstration Problems Manual.
Laminated
This subordinate form appears when the Input Properties button is selected on the Materials form, Composite is the
selected Object, and Laminate is the selected Method. Use this form to define the laminate lay-up data for a composite
material. If the resulting material is referenced in a Revised or Standard Laminate Plate element property region, then
an MSC Nastran PCOMP entry containing the lay-up data is written. If the resulting material is referenced by any other
type of element property region, the equivalent homogeneous properties of the material are used The difference
between the Total option and the Total -%thicknesses option is that the former requires that you give actual thickness
values of each ply and the latter requires each ply thickness to be given as a percentage of the total layup thickness.
This is the preferred method when applying the composite material to solid (CHEXA) elements or 2-D solid element
(axisymmetric, plane strain).
Main Index
472 Nonlinear Users Guide
Composite Material (PCOMP or PCOMPG)
Main Index
Chapter 10: Materials 473
Nonlinear Material Overview
Main Index
474 Nonlinear Users Guide
Nonlinear Material Overview
Main Index
Chapter 10: Materials 475
Nonlinear Material Overview
The following sections describe how to model material behavior in SOL 400. Modeling material behavior consists of
both specifying the constitutive models used to describe the material behavior and defining the actual material data
necessary to represent the material. Directional dependency can be included for materials other than isotropic
materials. Data for the materials can be entered either directly through the input file or by user subroutines, or material
models may be defined in the Patran Materials Application. Each section of this chapter discusses various options for
organizing material data for input. Each section also discusses the constitutive (stress-strain) relation and graphic
representation of the models and includes recommendations and cautions concerning the use of the models.
Main Index
476 Nonlinear Users Guide
Nonlinear Elastic
Nonlinear Elastic
Hypoelastic
The hypoelastic model is able to represent a nonlinear elastic (reversible) material behavior. For this constitutive
theory, SOL 400 assumes that
ij = L ij kl k l + g i j (10-11)
The stress and strains are true stresses and logarithmic strains, respectively, when used in conjunction with the updated
Lagrange (LRGSTRN=1 or 2).
When used in conjunction with the LRGSTRN=0, equation (10-11) is expressed as
S ij = L ijk l E + g ij (10-12)
kl
where E S are the Green-Lagrangian strain and second Piola-Kirchhoff stress, respectively.
The tensors L and g may be defined by the HYPELA2 or UMAT user subroutine. In order to provide an accurate
solution, L should be a tangent stiffness evaluated at the beginning of the iteration. In addition, the total stress should
be defined as its exact value at the end of the increment. This allows the residual load correction to work effectively.
The material model may be either isotropic or anisotropic. For anisotropic material models, the user defines the
behavior with respect to the preferred direction.
Additional information is available regarding the kinematics of deformation. In particular, the deformation gradient
( F ), rotation tensor ( R ), and the eigenvalues ( ) and eigenvectors ( N ) to form the stretch tensor ( U ) are also
provided. This information is available only for the continuum elements namely: plane strain, plane stress,
axisymmetric, and three-dimensional cases.
To activate the HPYPELA2 or UMAT user subroutine, one also must use MATUDS and reference the user subroutine via
the CONNECT file management statements option.
Hyperelastic Isotropic
Hyperelastic models are specified using either the MATHP or MATHE bulk data entries and are used to describe the
behavior of materials that exhibit large strain elastic response, such as rubber, solid propellant, and other elastomeric
materials. These materials are described in terms of a strain energy potential, U, which defines the strain energy
stored in the material per unit of volume in the initial configuration as a function of the strain at that point in the
material.
Main Index
Chapter 10: Materials 477
Nonlinear Elastic
Elastomeric materials are elastic in the classical sense. Upon unloading, the stress-strain curve is retraced and there is
no permanent deformation. Elastomeric materials are initially isotropic. Figure 10-6 shows a typical stress-strain curve
for an elastomeric material.
, Stress
100%
, Strain
Figure 10-6 A Typical Stress-Strain Curve for an Elastomeric Material
In the rectangular block in Figure 10-7, 1 , 2 , and 3 are the principal stretch ratios along the edges of the block
defined by
i = Li + ui Li (10-13)
Main Index
478 Nonlinear Users Guide
Nonlinear Elastic
L3 3L3
1L1 2L2
Undeformed
L2 Deformed
L1
Figure 10-7 Rectangular Rubber Block
In practice, the material behavior is (approximately) incompressible, leading to the constraint equation
1 2 3 = 1
2 2 2
I1 = + 2 + 3
1
2 2 2 2 2 2 (10-14)
I2 = 1 + +
2 2 3 3 1
2 2 2
I3 = 1 2 3
Depending on the choice of configurations, for example, reference (at t = 0 ) or current ( t = n + 1 ), you obtain
total or updated Lagrange formulations for elasticity. The kinematic measures for the two formulations are discussed
next.
1
E i j = --- C i j ij (10-15)
2
C ij = F ki F kj (10-16)
Main Index
Chapter 10: Materials 479
Nonlinear Elastic
x k
F k j = -------- (10-17)
X j
1
---
2
J = 1 2 3 = det C ij (10-18)
I1 = Ci i (implied sum on i)
2
C ij C i j C i i
I 2 = ------------------------------------------ (10-19)
2
1
I 3 = --- e i jk e p q r C i p C j q C k r = det C i j
6
in which e ijk is the permutation tensor. Also, using spectral decomposition theorem,
2 A A
C ij = A N i N j (10-20)
2
in which the stretches A are the eigenvalues of the right Cauchy-Green deformation tensor, C ij and the eigenvectors
A
are N i .
1
ij = --- ln b ij (10-21)
2
b ij = F ik F jk (10-22)
Thus, using the spectral decomposition theorem, the true strains are written as:
1 A A
ij = --- ln A n i n j (10-23)
2
A
where n i is the eigenvectors in the current configuration. It is noted that the true strains can also be approximated
using first Pad approximation, which is a rational expansion of the tensor, as:
1
ij = 2 V i j ij V ij + ij (10-24)
Main Index
480 Nonlinear Users Guide
Nonlinear Elastic
where a polar decomposition of the deformation gradient F i j is done into the left stretch tensor V i j and rotation tensor
R i j as:
F i j = V i k R kj
1
---
2
J = 1 2 3 = det b ij (10-25)
I 1 = b ii
1 2
I 2 = --- b i j b ij b i i (10-26)
2
1
and I 3 = --- e ijk e p q r b i p b jq b k r = det b ij
6
Strain energy expressed by the invariants described by either equations (10-19) or (10-26) are the same. Also, to
account for the incompressibility condition, in both formulations, the strain energy is split into deviatoric and
volumertic parts as:
Mooney-Rivlin Model
The generalized Mooney-Rivlin model for nearly-incompressible elastomeric materials is written as:
N N
gmr m n
Wd e v i a t o r i c = Cm n I1 3 I2 3 (10-28)
m = 1 n = 1
As an example of the usage of MATHE for defining a Mooney material, see Ball Joint Rubber Boot (Ch. 13) in the MSC
Nastran Demonstration Problems Manual.
Jamus-Green-Simpson Model
A particular form of the generalized Mooney-Rivlin model, namely the third order deformation (tod) model, is
implemented in SOL 400. However, the Ogden formulation (below) is usually better for large strain behavior than even
the fifth order Mooney-Rivlin.
tod 3
W devratoric = C 10 I 1 3 + C 01 I 2 3 + C 11 I 1 3 I 2 3 + C 20 I 1 3 2 + C 30 I 1 3 (10-29)
Main Index
Chapter 10: Materials 481
Nonlinear Elastic
tod
where W deviatoric is the deviatoric third order deformation form strain energy function,
Simpler and popular forms of the above strain energy function are obtained as:
nh
W deviatoric = C 10 I 1 3 Neo-Hookean
(10-30)
mr
W deviatoric = C 10 I 1 3 + C 01 I 2 3 Mooney-Rivlin
Ogden Model
The form of strain energy for the Ogden model in SOL 400 is,
N
k k k
-----k- 1 + 2 + 3 3
ogden
Wd e v i a t o r i c = k
(10-31)
k = 1
k
------
k 3 k
where i = i
are the deviatoric stretch ratios while C mn , k , and k are the material constants obtained
J
from the curve fitting of experimental data.
The Ogden model is usually applied to slightly compressible materials. If no bulk modulus is given, it is taken to be
virtually incompressible. This model is different from the Mooney model in several respects. The Mooney material
model is with respect to the invariants of the right or left Cauchy-Green strain tensor and implicitly assumes that the
material is incompressible. The Ogden formulation is with respect to the eigenvalues of the right or left Cauchy-Green
strain, and the presence of the bulk modulus implies some compressibility. Using a two-term series results in identical
behavior as the Mooney mode if:
1 = 2C 10 , 1 = 2 , 2 = 2C 01 , and 2 = 2
As an example of the usage of MATHE for defining a Ogden material, see Ball Joint Rubber Boot (Ch. 13) in the MSC
Nastran Demonstration Problems Manual.
Arruda-Boyce Model
In the Arruda-Boyce strain energy model, the underlying molecular structure of elastomer is represented by an eight-
chain model to simulate the non-Gaussian behavior of individual chains in the network. The two parameters, nk and
N ( n is the chain density, k is the Botzmann constant, is the temperature, and N is the number of statistical links
of length l in the chain between chemical crosslinks) representing initial modules and limiting chain extensibility and
are related to the molecular chain orientation thus representing the physics of network deformation.
As evident in most models describing rubber deformation, the strain energy function constructed by fitting experiment
data obtained from one state of deformation to another fails to accurately describe that deformation mode. The Arruda-
Main Index
482 Nonlinear Users Guide
Nonlinear Elastic
Boyce model ameliorates this defect and is unique since the standard tensile test data provides sufficient accuracy for
multiple modes of deformation.
2 0
C1
3 0
1 0
k
Consider a cube of dimension 0 with an unstretched network including eight chains of length r 0 = Nl , where the
fully extended chain has an approximate length of Nl. A chain vector from the center of the cube to a corner can be
expressed as:
0 0 0
C 1 = ------ 1 i + ------ 2 j + ------ 3 k (10-32)
2 2 2
Using geometrical considerations, the chain vector length can be written as:
1 12
r chain = ------- Nl 12 + 22 + 32 (10-33)
3
and
r chain 1 12
chain = ------------- = ------- I 1 (10-34)
r0 3
Using statistical mechanics considerations, the work of deformation is proportional to the entropy change on stretching
the chains from the unstretched state and may be written in terms of the chain length as:
r chain
W = nkN ------------- + ln -------------- C (10-35)
Nl sinh
where n is the chain density and C is a constant. is an inverse Langevin function correctly accounts for the limiting
chain extensibility and is defined as:
Main Index
Chapter 10: Materials 483
Nonlinear Elastic
r chain
= L 1 ------------- (10-36)
Nl
1
= coth --- (10-37)
With equations (10-34) through (10-37), the Arruda-Boyce model can be written
Arruda-Boyce 1 1 11
W dev = nk --- I 1 3 + ---------- I 12 9 + ------------------- I 13 27
2 20N 1050N
2
(10-38)
19 519
+ ------------------- I 14 81 + ------------------------- I 15 243
3 4
7000N 673750N
Gent Model
Also, using the notion of limiting chain extensibility, Gent proposed the following constitutive relation:
Gent EI m Im
W dev = ------------- log ----------------- (10-39)
6 I m I 1*
where
I 1* = I 1 3 (10-40)
The constant EI m is independent of molecular length and, hence, of degree of crosslinking. The model is attractive
due to its simplicity, but yet captures the main behavior of a network of extensible molecules over the entire range of
possible strains.
The volumetric part of the strain energy is for all the rubber models in SOL 400 is:
1 2
---
9K 3
W volumetric = ------- J 1 (10-41)
2
when K is the bulk modulus. It can be noted that the particular form of volumetric strain energy is chosen such that:
SOL 400 will use the three field formulation for rubber materials (Mooney, Gent, Arruda Boyce, and Ogden) if the
bulk modulus is entered as a -1. A series form of the volumetric strain energy U to capture the nonlinear pressure-
volumetric strain relationship is implemented. The volumetric strain energy is expressed as:
5
U = Di J 1 2 i (10-42)
i = 1
Main Index
484 Nonlinear Users Guide
Nonlinear Elastic
where D i i = 1 5 are material constants obtained from experimental data. If only a single term is used then
K = 2D 1 .
Foam Model
Sometimes elastomeric materials show large volumetric deformations. For this type of behavior, the models discussed
above are not appropriate. Instead, the foam model expressed by:
N N
n
-----n- 1 + 2 + 3 3 + ------ 1 J
n n n n
W = n n
(10-43)
n = 1 n = 1
should be used. In contrast to the Ogden model, the first part of the foam strain energy function is not purely deviatoric.
The material constants n provide additional flexibility to describe the material behavior also for a large amount of
compressibility.
Main Index
Chapter 10: Materials 485
Nonlinear Elastic
1 3
2
1 3
2
1 3
2
Uniaxial Test
Probably the most popular test is the uniaxial test (see Figure 10-10). This test can be used in tension as well as in
compression, both for incompressible and (slightly) compressible elastomeric materials. The shape of the specimen
Main Index
486 Nonlinear Users Guide
Nonlinear Elastic
used in compression will usually be less slender than the shape used in tension. Within the region indicated by the
dashed line, the state of deformation will be homogeneous, where the deformation can be described by:
1 = = 1 + e 11 , 2 = 3 = J (10-44)
F
11 = = ------ , 22 = 33 = 0 (10-45)
A0
in which F is the applied force and A 0 is the cross sectional area of the undeformed specimen in the E 2 - E 3 plane,
within the region indicated by the dashed line.
F F
E2
E3 E1
Figure 10-10 Uniaxial (Tensile) Test
Necessary input for the curve fitting program in Patran consists of at least engineering strain ( e 11 ) versus engineering
stress ( 11 ) data points. In case of (slightly) compressible materials, information about the volume changes is also
needed. This data can be given either in terms of the area ratio or the volume ratio. The area ratio is defined by the
current cross sectional area A over the original cross sectional area A 0 . Similarly, the volume ratio is defined by the
current volume V over the undeformed volume V 0 . Notice that the volume ratio and the area ratio are related by:
V A
------ = J = ------ 1 + e 11
V0 A0
If, for a particular elastomeric material, both a tensile and a compression test have been performed, all the data points
should be collected into one data file. The layout of a data file containing uniaxial test data is given in the figure below.
The columns may be separated by either spaces or commas. For (nearly) incompressible material behavior, the third
column can be omitted.
Main Index
Chapter 10: Materials 487
Nonlinear Elastic
e 11 11 A A0 e 11 11 V V0
or
Equi-Biaxial Test
The equi-biaxial tensile test outlined in Figure 10-12 can be used to obtain, within the region indicated by the dashed
line, a homogeneous state of deformation defined by:
F F
E2
E3 E1
F
Figure 10-12 Equi-biaxial (Tensile) Test
2
1 = 2 = = 1 + e 11 = 1 + e 22 , 3 = J (10-46)
F
11 = 22 = = ------ , 33 = 0 (10-47)
A0
Main Index
488 Nonlinear Users Guide
Nonlinear Elastic
with A 0 being the original cross sectional area of the elastomeric sheet in the direction perpendicular to the applied
forces, which is assumed to be the same in the E 1 - E 3 -plane and the E 2 - E 3 -plane.
For compressible elastomers, volumetric information is needed. For the equi-biaxial test, this can be given in terms of
a thickness ratio or, similar to the uniaxial test, a volume ratio. The thickness ratio is defined as the current sheet
thickness t over the original sheet thickness t 0 . The relation between the thickness ratio and the volume ratio is:
V t 2
------ = J = ---- 1 + e 11 (10-48)
V0 t0
The layout of a data file for an equi-biaxial tensile test is given in Figure 10-12.
F F
E2
E3 E1
Except for the vicinity of the free edges and the clamps, the state of strain can be found to be substantially uniform,
according to:
J
1 = = 1 + e 11 , 2 = 1 , 3 = --- (10-49)
where the known stress components are given by:
F
11 = = ------ , 33 = 0 (10-50)
A0
Main Index
Chapter 10: Materials 489
Nonlinear Elastic
in which A 0 is the cross sectional area of the undeformed specimen in the E 1 - E 3 -plane. Notice that the engineering
strain e 22 is zero, but that the corresponding engineering stress 22 depends on the material behavior.
U = T S S (10-51)
3 U U
TS = U = 2 S S + (10-52)
S I
1 I 2
x 1 = X 1 + X 2 , x 2 = X 2 , x 3 = X 3 (10-53)
1 0
F = 0 1 0 (10-54)
0 0 1
2F
E2 atan
E3 E1
Notice that det F = 1 , irrespective of the value of , from which it can be concluded that a simple shear test is a
constant volume test.
Based on equations (10-53) and (10-54) and Figure 10-14, the engineering strain tensor and the right Cauchy-Green
strain tensor can be evaluated as:
Main Index
490 Nonlinear Users Guide
Nonlinear Elastic
0 2 0
e = 2 0 0 (10-55)
0 0 0
1 0
C = 1+ 0
2 (10-56)
0 0 1
According to equation (10-56), the principal stretch ratios follow from the principal values of C and read:
2 2
1 2 = 1 + ----- 1 + ----- , 3 = 1 (10-57)
2 4
It can easily be verified that 1 2 3 = 1 , which again shows that the simple shear test is a constant volume test. The
relevant engineering stress is given by:
F
12 = ------ (10-58)
A0
with A 0 being the cross sectional area of the undeformed specimen in the E 1 - E 3 -plane.
The layout of a data file containing measurements of a simple shear test is given in Figure 10-15.
2e 12 = 12
Volumetric Test
Although a uniaxial, equi-biaxial and planar shear test can be used to obtain information about the volumetric behavior,
for compressible materials an additional volumetric test may be preferable. This is especially true for slightly
compressible materials, since volumetric data from other tests other than a volumetric one may easily be inaccurate
(because most of the deformation is deviatoric). Two commonly used volumetric tests are outlined in Figure 10-16. In
Figure 10-16a, a cylindrical specimen is compressed in a cylindrical hole. This test can be successfully applied for
slightly compressible materials. In Figure 10-16b, a specimen is deformed by compressing the surrounding fluid. This
volumetric test can also be used for highly compressible materials.
Main Index
Chapter 10: Materials 491
Nonlinear Elastic
E1
(a) (b)
F
E3 E2
For a volumetric test, the direct true stress components are assumed to be equal to the hydrostatic pressure p and given
by:
F
T 11 = T 22 = T 33 = ------ (10-59)
p
A
p
in which A denotes the area of the piston in the E 2 - E 3 -plane. The deformation can be expressed in terms of an
engineering strain e and corresponding stretch ratio , which can be determined from the measured volume change
according to:
V
e = 1 = 3 ------ 1 = 3 J 1 (10-60)
V0
2
= T 11 (10-61)
Notice that only in the case of Figure 10-16b the engineering strain e and the engineering stress are equal to the
direct components of the engineering strain and the engineering stress tensor.
The layout of the data file corresponding to a volumetric test is given in Figure 10-17. Notice that because of
Figure 10-16b, the entries of the first and the third column are not independent.
Main Index
492 Nonlinear Users Guide
Nonlinear Elastic
e V V0
Relaxation Test
The relaxation test is necessary to obtain the rate response for the viscoelastic material model.
The basic feature of a relaxation test is that the force or stress response to a prescribed fixed displacement or
deformation is measured as a function of time. A relaxation test for a large strain elastomeric material is indicated in
Figure 10-18. By measuring the force needed for a displacement u at different time intervals, the decay of the strain
energy as a function of time can be determined. For linear elastic isotropic material, similar tests can be performed to
get information about the shear modulus and/or the bulk modulus as a function of time. In order to properly measure
the instantaneous values, application of the prescribed displacement should occur sufficiently fast. It should be noted,
due to the assumption introduced in equation (10-84), that for large strain visco-elastic materials the magnitude of (the
instantaneous value of) the strain energy is not important, since every energy term in the Prony series expansion is
related to the instantaneous strain energy using a scalar multiplier. The data does not need to be equispaced in time.
Usually, at the beginning of the relaxation experiment the measurements are done at smaller time intervals than at the
end of the experiment.
If, for linear viscoelastic materials, instead of a relaxation test only a creep test can be performed, the creep data must
be transformed into relaxation data. Converting creep data into relaxation data can be done using a numerical
integration scheme, but is not part of SOL 400.
Main Index
Chapter 10: Materials 493
Nonlinear Elastic
Densification
STRESS
STRAIN
Figure 10-19 Typical Compressive Stress-Strain Curve
Main Index
494 Nonlinear Users Guide
Nonlinear Elastic
STRESS
Main Index
Chapter 10: Materials 495
Nonlinear Elastic
Entry Description
MATHP Specifies material properties for use in fully nonlinear (i.e., large strain and large rotation)
hyperelastic analysis of rubber-like materials (elastomers).
MATHE Specifies hyperelastic (rubber-like) material properties for nonlinear (large strain and large
rotation) analysis in Nonlinear Analysis.
Viscoelastic
The material models discussed in previous sections are considered to be time independent. However, rubber materials
often show a rate-dependent behavior and can be modeled as viscoelastic materials. Viscoelasticity can be applied:
To determine the current state of deformation based on the entire time history of loading.
To include temperature dependencies.
To use in conjunction of isotropic, orthotropic, and anisotropic material for small strain problems.
To use in conjunction with rubber or foam material for large strain problems.
t
d kl
ij t = ijkl
G t ------------------- d + G
d ijkl t kl 0 (10-62)
0
The functions G ij kl are called stress relaxation functions. They represent the response to a unit applied strain and have
characteristic relaxation times associated with them. The relaxation functions for materials with a fading memory can
be expressed in terms of Prony or exponential series.
N
n n
G ijkl t = G ijkl + G ijkl exp t (10-63)
n = 1
Main Index
496 Nonlinear Users Guide
Nonlinear Elastic
n n
in which G ij kl is a tensor of amplitudes and is a positive time constant (relaxation time). In the current
implementation, it is assumed that the time constant is isotropic. In equation (10-63), G ij kl represents the long term
modulus of the material.
The short term moduli (describing the instantaneous elastic effect) are then given by
N
0 n
G ijkl = G ijkl 0 = G ijkl + G ijkl (10-64)
n = 1
The stress can now be considered as the summation of the stresses in a generalized Maxwell model (Figure 10-21)
N
n
ij t = ij t + ij t (10-65)
n = 1
where
ij = G ijkl kl t (10-66)
t
n n n d kl
ij = G ijkl exp t ------------------- d
d
(10-67)
0
1 2 i
q1 q2 qi
E E1
E2 Ei
E0
i = i/Ei
Figure 10-21 The Generalized Maxwell or Stress Relaxation Form
For integration of the constitutive equation, the total time interval is subdivided into a number of subintervals
( t m 1 t m ) with time-step h = t m t m 1 . A recursive relation can now be derived expressing the stress increment
in terms of the values of the internal stresses ijn at the start of the interval. With the assumption that the strain varies
linearly during the time interval h, we obtain the increment stress-strain relation as
Main Index
Chapter 10: Materials 497
Nonlinear Elastic
N N
n n n n
ij t m = G ijkl + h G ijkl kl h ij t m h (10-68)
n = 1 n = 1
where
n
n h = 1 exp h (10-69)
and
n n n
h = h h (10-70)
In SOL 400, the incremental equation for the total stress is expressed in terms of the short term moduli (See equation
(10-64)).
N N
0 n n n n
ij t m = G ijkl 1 h G ijkl kl t m h ij t m h (10-71)
n = 1 n = 1
Note that the set of equations given by equation (10-71) can directly be used for both anisotropic and isotropic
materials.
N
n n
Gt = G + G exp t d (10-72)
n = 1
N
n n
Kt = K + K exp t v (10-73)
n = 1
with short term values given by
N
G0 = G + Gn (10-74)
n = 1
Main Index
498 Nonlinear Users Guide
Nonlinear Elastic
N
K0 = K + Kn (10-75)
n = 1
4 3 2 3 2 3 0 0 0
2 3 4 3 2 3 0 0 0
d = 2 3 2 3 4 3 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
1 1 1 0 0 0
1 1 1 0 0 0
v = 1 1 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
Nd
0 n n
t m = G 1 d h G d t m
n = 1
Nv
0 n n
K 1 v h K v t m (10-76)
n = 1
Nd Nv
n n n n
d h d t m h v h v t m h
n = 1 n = 1
and
n n n n n
d t m = d h G d t m d h d t m h
n n n n n (10-77)
v t m = v h K v t m v h v t m h
Main Index
Chapter 10: Materials 499
Nonlinear Elastic
Note that the deviatoric and volumetric response are fully decoupled.
Note that the algorithm is exact for linear variations of the strain during the increment. The algorithm is implicit; hence,
for each change in time-step, a new assembly of the stiffness matrix is required.
n 0
where E i j are the components of the Green-Lagrange strain tensor, Q ij internal variables and the elastic strain
0
energy density for instantaneous deformations. In SOL 400, it is assumed that = W , meaning that the energy
density for instantaneous deformations is given by the third order James Green and Simpson form or the Ogden form.
The components of the second Piola-Kirchhoff stress then follow from
N
0 n
S ij = ---------- = ----------
E i j E ij Q ij (10-79)
n = 1
The energy function can also be written in terms of the long term moduli resulting in a different set of internal variables
n
Ti j
N
n n
E i j T i j = E ij + Ti j Ei j (10-80)
n = 1
where is the elastic strain energy for long term deformations. Using this energy definition, the stresses are obtained
from
N
E n
S ij = -------------------- +
E ij Ti j (10-81)
n = 1
Observing the similarity with the equations for small strain viscoelasticity the internal variables can be obtained from a
convolution expression
n n
0 S ij exp t
t n
Ti j = d (10-82)
Main Index
500 Nonlinear Users Guide
Nonlinear Elastic
n
where S ij are internal stresses obtained from energy functions.
n n
S ij = ---------- (10-83)
E ij
If, in the energy function, each term in the series expansion has a similar form, equation (10-84) can be rewritten
N
= + n 0 exp t n (10-85)
n = 1
n
where is a scalar multiplier for the energy function based on the short term values.
N
0
S ij = ----------- = 1 n ----------
(10-87)
E i j E i j
n = 1
t
n n 0 n
Ti j = S i j t exp t d (10-88)
0
Analogue to the derivation for small strain viscoelasticity, a recursive relation can be derived expressing the stress
increment in terms of values of the internal stresses at the start of the increment.
The equations are reformulated in terms of the short time values of the energy function
N
S ij t m = 1 1 h S ij t m S i j t m h
n n 0 0
n=1
N (10-89)
n
n S ij t m h
n = 1
Main Index
Chapter 10: Materials 501
Nonlinear Elastic
0 n n
S ij t m = n h n S ij t m S i j t m h n h S ij t m h (10-90)
It is assumed that the viscoelastic behavior in SOL 400 acts only on the deviatoric behavior.
As an example of the usage of MATVE for defining a Mooney material, see Ball Joint Rubber Boot (Ch. 13) in the MSC
Nastran Demonstration Problems Manual.
ln t
E T ln t = E T ln t+f T * T 0 T 0 (10-91)
0
Main Index
502 Nonlinear Users Guide
Nonlinear Elastic
where f T is measured relative to some arbitrary temperature T . The modulus curve shifts towards shorter times with
an increase of temperature; f T is a positive increasing function for T T 0 . If G T t denotes the relaxation modulus
as a function of time at uniform temperature T , so that,
G T t = E T ln t (10-92)
then
GT t = GT (10-93)
0
The relaxation modulus (and the other characteristic functions) at an arbitrary uniform temperature is thus expressed
by the base temperature behavior related to a new time scale that depends on that temperature.
There is some mapping of the time coordinate for nonuniform, nonconstant temperature, T x t , which depends on
the position
For a nonuniform, nonconstant temperature, the shift function is a T x t and the rate of change of reduced time can
be written as:
d = a T T x ,t dt
SOL 400 offers two explicit forms for entering the shift function. The first is based on the familiar Williams-Landel-
Ferry (WLF) equation. Rewriting the above expression for reduced time as
t
dt
x t = ------------------------------
T x t
(10-94)
T
0
C1 T T0
log 10aT T = ----------------------------------- = h T (10-95)
C2 + T T0
and
t
h T t
t = 10 dt (10-96)
0
Typically, the glassy transition point is taken as the reference temperature in the above relation. The logarithmic shift
can also be expressed in a polynomial expansion about the arbitrary reference point as
m
i
log 10 A T T = i T T0 (10-97)
i = 0
Enter the shift function parameters associated with Thermo-Rheologically Simple behavior through the MATTVE bulk
data entry.
Main Index
Chapter 10: Materials 503
Nonlinear Elastic
0
In addition to the Thermo-Rheologically Simple material behavior variations of initial stress-strain moduli G ij kl , the
temperature of the other mechanical properties (coefficient of thermal expansion, etc.) due to changes in temperature
can be specified via the MATT1 or MATTORT bulk data entry.
Note, however, that only the instantaneous moduli are entered. Hence, the long term moduli given by
N
0 n
G ij kl = G ij kl T G ijk l (10-98)
n = 1
which can easily become negative if the temperature effects are not defined properly.
Narayanaswamy Model
The annealing of flat glass requires that the residual stresses be of an acceptable magnitude, while the specification for
optical glass components usually includes a homogenous refractive index. The design of heat treated processes (for
example, annealing) can be accomplished using the Narayanaswamy model. This allows you to study the time
dependence of physical properties (for example, volumes) of glass subjected to a change in temperature.
The glass transition is a region of temperature in which molecular rearrangements occur on a scale of minutes or hours,
so that the properties of a liquid change at a rate that is easily observed. Below the glass transition temperature T g , the
material is extremely viscous and a solidus state exists. Above T g , the equilibrium structure is arrived at easily and
the material is in liquidus state. Hence, the glass transition is revealed by a change in the temperature dependence of
some property of a liquid during cooling. If a mechanical stress is applied to a liquid in the transition region, a
time-dependent change in dimensions results due to the phenomenon of visco-elasticity.
If a liquid in the transition region is subjected to a sudden change in temperature, a time-dependent change in volume
occurs as shown in Figure 10-23. The latter process is called structural relaxation. Hence, structural relaxation governs
the time-dependent response of a liquid to a change of temperature.
Main Index
504 Nonlinear Users Guide
Nonlinear Elastic
T1
T(t)
T2
t0 t
(a) Step Input for Temperature
V(0,T1)
g(T2-T1)
V(0,T2) l(T2-T1)
V(,T2)
t0 t
(b) Volume Change as Function of Temperature
Suppose a glass is equilibrated at temperature T 1 , and suddenly cooled to T 2 at t 0 . The instantaneous change in
volume is g T 2 T 1 , followed by relaxation towards the equilibrium value V T 2 . The total change in volume
due to the temperature change is 1 T 2 T 1 as shown in Figure 10-23b. The rate of volume change depends on a
characteristic time called the relaxation time.
The slope of dV dT changes from the high value characteristic of the fluid 1 to the low characteristic of the glass
g as shown in Figure 10-24. The glass transition temperature T g is a point in the center of the transition region. The
low-temperature slope g represents the change in volume V caused by vibration of the atoms in their potential wells.
In the (glassy) temperature range, the atoms are frozen into a particular configuration. As the temperature T increases,
the atoms acquire enough energy to break bonds and rearrange into new structures. That allows the volume to increase
more rapidly, so 1 g . The difference = 1 g represents the structural contribution to the volume.
Main Index
Chapter 10: Materials 505
Nonlinear Elastic
V(T)
V(T0) l
Liquid
State
V(T1)
g Transition Range
Solidus
State
T0
When a liquid is cooled and reheated, a hysteresis is observed as shown in Figure 10-25.
V
Equilibrium
Nonequilibrium
Tg T
Unfortunately, the notion of a glass transition temperature is insufficient as real glassy materials generally exhibit a
temperature regime, called a transition range, across which their bulk properties gradually change from being solid-
like to liquid-like in nature.
As discussed earlier, properties have a time dependence in the transition range. An explanation for the strong time
dependence lies in that the material resides at a nonequilibrium temperature which lags behind the applied temperature
during the heating-cooling cycle. The nonequilibrium temperature is called the fictive temperature, T f , as shown in
Figure 10-24. The fictive temperature at T 1 T f T 1 is found by extrapolating a line from V T 1 with slope g to
intersect a line extrapolated from V T 0 with slope 1 (see Figure 10-24). For T T 2 (well below the glass
transition), T f reaches a limiting value that is called T g . If the material were equilibrated at T f T 1 , then
instantaneously cooled to T 1 , it would change along the line with slope g because no structural rearrangement could
occur. Therefore, it would have the same volume as the continuously cooled sample.
Main Index
506 Nonlinear Users Guide
Nonlinear Elastic
V T 2 ,t = V T 1 , + g T 2 T 1 + l g T f t T 2 (10-99)
where T f t is the current value of the fictive temperature. The response function, M v , which dictates the value of the
fictive temperature is assumed to be linear in its argument and governs both the value of the fictive temperature as well
as the material property of interest.
Vt V Tf t T
-------------------------------- = M v t = -------------------------------
- (10-100)
V0 V T0 T
By virtue of its linearity, Boltzmanns superposition principle can be invoked to calculate the fictive temperature at any
time:
t
d
Tf t = T t M v t t -------(T t ) dt'
dt
(10-101)
The concept of reduced time, t is introduced in the spirit of Thermo-Rheologically Simple materials to capture the
disparate nonlinear response curves on a single master curve.
The reduced time used in Marc is given by the following expression:
t
ref
t = -------------------
T t
- dt (10-102)
Here ref is the reference relaxation time of the material evaluated at a suitable reference temperature, T ref . The
relaxation time at the given time and temperature can be represented as:
H 1 x 1 x
= ref exp ---- --------- --- ----------------- (10-103)
R T re f T Tf
The parameter x allows you to dictate how much of the fictive temperature participates in the prescription of the
relaxation time, and must, therefore, range between 0 and 1.
H is the activation energy for the particular process and R is the gas constant. A typical response function is:
M v = exp -- (10-104)
Multiple structural relaxation times can exist. The response function has, therefore, been implemented as:
n
Mv = W g i exp ----
i
i = 1
Main Index
Chapter 10: Materials 507
Nonlinear Elastic
12. The fraction parameter x and the activation energy-gas constant ratio.
13. The solid and liquid coefficients of thermal of expansion, g and 1 through the MATTVE bulk data entry.
A stable algorithm is employed to calculate the convolution integrals. For improved accuracy it is recommended that
the time steps used during the simulation be sufficiently small.
In Figure 10-26, the volume of cube of material, which is allowed to contract freely and is experiencing a 100oC
quench, is displayed.
T=100
Volume
gT
lT
Temperature
e
Tim
Entry Description
MATVE Specifies isotropic viscoelastic material properties to be used for quasi-static or dynamic
analysis in SOL 400.
MATTVE Specifies temperature-dependent visco-elastic material properties in terms of Thermo-
Rheologically Simple behavior to be used for quasi-static or transient dynamic analysis in
SOL 400.
Main Index
508 Nonlinear Users Guide
MATS1 Model
MATS1 Model
The MATS1 bulk data entry may be used to define elastic-plastic material behavior or a nonlinear elastic model using
the TYPE=NLELAST. When modeling metal materials subjected to high stress, the use of MATEP is recommended and
is discussed in subsequence sections.
The nonlinear elastic capability was developed to predict the multi-axial stress-strain behavior for nonlinear elastic
material, for which only the simple tension stress-strain data is available. The theory and algorithm is adequate to trace
the stress-strain curve accurately for the uniaxial loading cases. However, the theory is not based on the classical theory
of finite elasticity. For large strain, the Mooney, Ogden, Arruda-Boyce, or Gent model should be used. Consequently,
some of the constitutive relations may be violated in the multiaxial stress cases. Nevertheless, this capability can be
justified by stating that the design is not intended for the large-strain deformation.
The current nonlinear elastic capability accommodates bilateral properties (uniaxial tension-compression) accurately
for the uniaxial deformations. This design is considered adquate for the multiaxial deformation if the strain remains in
the small range.
Theoretical Basis
The nonlinear elastic capability in MSC Nastran was designed to satisfy the equivalence of the deformation work per
unit volume in the simple tension to the strain energy per unit volume (conservation of energy), while the work done
for deformation may be defined by a stress-strain curve in simple tension; i.e.,
d = <> d
It is further assumed that the effective strain may be defined by:
1
--- E 2 = 1
--- <> D e d
2 2
Substituting the latter in the first equation above, where stresses may be expressed in terms of total strains; i.e.
= ------ D e
E
The tangential matrix for such material may be obtained by differentiating the latter equation; i.e.
D n e = ------------- = ------ D e + ------ ------- --- D e <> D e T
E E
Main Index
Chapter 10: Materials 509
MATS1 Model
Some materials exhibit appreciably different behavior in compression from that in tension even in the small strain
range; i.e., cast iron shown in Figure 10-27. For uniaxial loading, the magnitude of the strain in that direction becomes
the effective strain; e.g.,
The full stress-strain curve x may be defined in the first and the third quadrant to accommodate different
uniaxial compression data. If the curve is defined only in the first quadrant, then the curve must start at the origin
( X1 = 0.0 , Y = 0.0 ) and the compression properties will be assumed identical to tension properties.
Stress
Strain
In case of this unsymmetric material behavior, we need to be able to distinguish between a state of compression and a
state of tension. There are two known data points for one effective strain , namely the effective stress for uniaxial
tension t and the effective stress for uniaxial compression c . Some method of interpolation or extrapolation is
required to predict the effective stress for the general stress state using two known data points.
The first stress invariant I 1 is adopted to determine if we are in the compression or tension zone
I1 = x + y + z
The instantaneous modulus ------- is interpolated or extrapolated in the same manner.
Main Index
510 Nonlinear Users Guide
MATS1 Model
1 2
--- x y 2 + y z 2 + z x 2 + 3 xy 2 2
= + yz + zx for 3-D
2
2 2 2
= x x y + y + 3 xy for plane stress
1
--- x y 2 + y z 2 + z x 2 + 3 x2y for plane strain
=
2
14. Compute the first invariant of I = x + y + z :
I1
r = ------
e
where r signifies the relative distance from the midpoint of c and t at . It would be implausible to process
a large value of r (such is the case with a hydrostatic load). Therefore, r , will be confined to a plausible range,
1 r 1 . The value will be reset to the limit r = 1 if r lies outside the range.
16. Look up the user specified stress-strain curve in the TABLES1 bulk data entry and determine t and c ; i.e.,
t =
c =
t + c t c
= ------------------ + r ------------------
2 2
For the tangent matrix the instantaneous modulus ------- is determined using the same ratio (r) as follows:
Main Index
Chapter 10: Materials 511
MATS1 Model
where x i y i is the i-th data point in the TABLES1 bulk data entry
yj + 1 yj
------- = ----------------------
- for x i x i + 1
c xj + 1 xj
19. Determine ------- based on ------- -------
t , c and r , i.e.,
------- = --- ------- + ------- + --r-
1
------- -------
2 t c 2 t c
Main Index
512 Nonlinear Users Guide
Inelastic Behavior
Inelastic Behavior
Most materials of engineering interest initially respond elastically. Elastic behavior means that the deformation is fully
recoverable, so that, when the load is removed, the specimen returns to its original shape. If the load exceeds some
limit (the yield load), the deformation is no longer fully recoverable. Some parts of the deformation will remain when
the load is removed as, for example, when a paper clip is bent too much, or when a billet of metal is rolled or forged
in a manufacturing process. Plasticity theories model the materials mechanical response as it undergoes such
nonrecoverable deformation in a ductile fashion. The theories have been developed most intensively for metals, but
they are applied to soils, concrete, rock, and ice. These materials behave in very different ways (for example, even large
values of pure hydrostatic pressure cause very little inelastic deformation in metals, but quite small hydrostatic pressure
may cause a significant, non-recoverable volume change in a soil sample), but the fundamental concepts of plasticity
theories are sufficiently general that models based on these concepts have been successfully developed for a wide range
of materials. A number of these plasticity modes are available in the SOL 400 material library.
In nonlinear material behavior, the material parameters depend on the state of stress. Up to the proportional limit, i.e.,
the point at which linearity in material behavior ceases, the linear elastic formulation for the behavior can be used.
Beyond that point, and especially after the onset of yield, nonlinear formulations are required. In general, two
ingredients are required to ascertain material behavior:
1. an initial yield criterion to determine the state of stress at which yielding is considered to begin
2. mathematical rules to explain the post-yielding behavior
There are two major theories of plastic behavior that address these criterion differently. In the first, called deformation
theory, the plastic strains are uniquely defined by the state of stress. The second theory, called flow or incremental
theory, expresses the increments of plastic strain (irrecoverable strains) as functions of the current stress, the strain
increments, and the stress increments. Incremental theory is more general and can be adapted in its particulars to fit a
variety of material behaviors. The plasticity models in SOL 400 are incremental theories, in which the mechanical
strain rate is decomposed into an elastic part and a plastic (inelastic) part through various assumed flow rules.
The incremental plasticity models are formulated in terms of:
A yield surface, which generalizes the concept of yield load into a test function which can be used to
determine if the material will respond purely elastically at a particular state of stress, temperature, etc.;
A flow rule that defines the inelastic deformation that must occur if the material point is no longer responding
purely elastically;
and some evolution laws that define the hardening the way in which the yield and/or flow definitions change
as inelastic deformation occurs.
The models also need an elasticity definition, to deal with the recoverable part of the strain models divide into those
that are rate-dependent and those that are rate-independent.
SOL 400 includes the following models of inelastic behavior.
Metal Plasticity (von Mises, Hill, or Barlat)
Pressure-Dependent models models the behavior of granular (soil and rock) materials or polymers, in which
the yield behavior depends on the equivalent pressure stress.
Linear Mohr-Coulomb
Parabolic Mohr-Coulomb
Main Index
Chapter 10: Materials 513
Inelastic Behavior
Yield Conditions
The yield stress of a material is a measured stress level that separates the elastic and inelastic behavior of the material.
The magnitude of the yield stress is generally obtained from a uniaxial test. However, the stresses in a structure are
usually multiaxial. A measurement of yielding for the multiaxial state of stress is called the yield condition. Depending
on how the multiaxial state of stress is represented, there can be many forms of yield conditions. For example, the yield
condition can be dependent on all stress components, on shear components only, or on hydrostatic stress. A number of
yield conditions are available in SOL 400, and are discussed in this section.
Metal Plasticity
The von Mises yield surface is widely used for plasticity in isotropic metals. It is assumed that the yield and plastic
flow describe isotropic metals at low temperatures where creep effects can be ignored. Anisotropic metals and
composite materials, can be treated by extensions of von Mises yield function, as described in Hills yield function.
von Mises
The success of the von Mises criterion is due to the continuous nature of the function that defines this criterion and its
agreement with observed behavior for the commonly encountered ductile materials. The von Mises criterion states that
yield occurs when the effective (or equivalent) stress equals the yield stress y as measured in a uniaxial test.
Figure 10-28 shows the von Mises yield surface in 2-D and 3-D stress space.
2 3
Yield
Surface
Yield
Surface
Elastic
Region
1
Elastic 1 2
Region
= 1 2 2 + 2 3 2 + 3 1 2 1 2 2 (10-105)
Main Index
514 Nonlinear Users Guide
Inelastic Behavior
2 2
= x y 2 + y z 2 + z x 2 + 6 xy + yz + z2x 1 2 2 (10-106)
The yield condition can also be expressed in terms of the deviatoric stresses as:
3
= --- (10-107)
2 i j ij
1
= i j --- kk ij (10-108)
ij 3
For isotropic material, the von Mises yield condition is the default condition in SOL 400.
As an example of the usage of MATEP for defining a elastic-plastic material, see Tube Flaring (Ch. 15) in the MSC
Nastran Demonstration Problems Manual.
2 2 2 2 2 2
= F 2 3 + G 3 1 + H 1 2 + 2L 23 + 2M 31 + 2N 12
in which the material parameters can be related to the yield stress ratios by
1 1 1 1
F = --- -------- + -------- --------
2 R2 2
R 33 R 11
2
22
1 1 1 1
G = --- -------- + -------- --------
2 R2 2
R 11 R 22
2
33
1 1 1 1
H = --- -------- + -------- --------
2 R2 2
R 22 R 33
2
11
3 3 3
L = ------------ M = ------------ N = ------------
2 2 2
2R 23 2R 31 2R 12
with
Main Index
Chapter 10: Materials 515
Inelastic Behavior
Y1 Y2 Y3
R 11 = ------ R 22 = ------ R 33 = ------
Ya Ya Ya
3T 12 3T 23 3T 31
R 12 = ---------------- R 23 = ---------------- R 31 = ----------------
Ya Ya Ya
where Y 1 , Y 2 , and Y 3 are the initial tensile yield stresses measured in material directions 1, 2, and 3, respectively;
T 12 , T 23 , and T 31 are the shear yield stresses in pure shear; and Y a is the reference yield stress which should be an
average yield stress in all directions.
Note the following points about Hills surface:
1. It degenerates into von Mises surface when all three direct yield stresses are equal
F = G = H and L = M = N .
2. It is invariant with respect to hydrostatic stress, as is von Mises.
In practical applications, however, the initial yield stress cannot be measured in all directions. The plastic anisotropy
is pronounced in the sheet metal due to prior rolling process, for which the plastic anisotropy is customarily
characterized by r-values defined by strain ratio measured in the uniaxial tension, i.e.,
w ln wo w H + 2N F G 4H sin cos
2 2
r a = ------ = ------------------------- = --------------------------------------------------------------------------------------------
t ln to t 2
F sin + G cos
2
where t and w denote thickness and width, respectively; and denotes the angle of orientation (usually measured
from the rolling direction). Assuming that the anisotropy parameters stay constant throughout the deformation, F, G,
H, and N can be determined by r-values from tensile specimen cut at 0, 45 and 90 degrees to the rolling direction as
show in Figure 10-29.
T (Transverse Direction)
Main Index
516 Nonlinear Users Guide
Inelastic Behavior
H H
---- = r 0 ---- = r 90
G F
r0
---- = r 45 + 1
N
--- 1 + ------
-
G 2 r 90
The orthotropic plasticity parameters should be calculated from the r-values and the initial yield stress either in 0 or
90 direction ( Y 0 or Y 90 ) from the experiment. The yield stress in the thickness direction can be written as
r 90 1 + r 0 r 0 1 + r 90
Y t h = Y 0 ---------------------------
- = Y 90 ---------------------------
-
r 0 + r 90 r 0 + r 90
1
T 12 = Y th ---------------------
2r 45 + 1
and
Ya
T 23 = T 31 = -------
3
in which the transverse direction is assumed isotropic.
Note that the use of the Hill Yield Function with all coefficients set to unity will give results varying from the von
Mises case due to differences in return mapping in both cases. By default, the von Mises case uses the mean normal
return scheme whereas the Hill case uses the multistage radial return scheme.
m m m m
f = S1 S2 + S2 S3 + S3 S1 = 2 (10-109)
where S i = 1 2 3 are principal values of a symmetric matrix S ij defined with respect to the components of the Cauchy
stress as
Main Index
Chapter 10: Materials 517
Inelastic Behavior
C 3 xx yy C 2 z z xx
------------------------------------------------------------------------------ C 6 xy C5 z x
3
S = C 1 yy z z C 3 xx yy (10-110)
C 6 xy -----------------------------------------------------------------------------
- C4 z y
3
C 2 z z xx C 2 yy z z
C5 z x C 4 zy ----------------------------------------------------------------------------
-
3
In equation (10-110), the symmetry axes (x, y, z), which represent the mutually orthogonal axes of anisotropy, are
aligned with the initial rolling, transverse, and normal directions of the sheet. During deformation, the anisotropic yield
surface of each material element rotates so that the symmetry axes are all in different directions during deformation.
Therefore, it is necessary to trace the rotation of the yield surface during deformation in order to calculate the plastic
strain increment properly. The rotation of the anisotropy axes is carried out based on the polar decomposition method.
The material coefficients, C i = 1 6 in equation (10-110) represent anisotropic properties. When C i = 1 6 = 1 , the
material is isotropic and equation (10-109) reduces to the Tresca yield condition for m = 1 or , and the von Mises
yield criterion for m = 2 or 4. The exponent m is mainly associated with the crystal structure of the material. A
higher m value has the effect of decreasing the radius of curvature of rounded vertices near the uniaxial and balanced
biaxial tension ranges of the yield surface, in agreement with polycrystal models. Values of m = 8 for FCC materials
(like aluminum) and m = 6 for BCC materials (like steel) are recommended. The yield surface has been proven to
be convex for m 1 . Figure 10-30 shows the yield surfaces obtained from von-Mises, Hill and Barlat yield functions
for Aluminum 2008-T4 alloy.
1.5
yy
Mises
1.0
Hill
(1948)
0.5
Barlats 6D
xx
0.0
-0.5
-1.0
-1.0 -0.5 0.0 0.5 1.0 1.5
Figure 10-30 Comparison of Yield Surfaces Obtained from von Mises, Hill and Barlat Yield Functions
p
Utilizing the normality rule, the associated plastic strain increment i j is obtained from the yield function f as
f
ipj = ---------- (10-111)
i j
Main Index
518 Nonlinear Users Guide
Inelastic Behavior
f
where is a scalar function. The calculation of ---------- in equation (10-111) is lengthy but straightforward. The stress
i j
p f
integration to obtain i j along with the calculation of ------------- are shown in the works by Chung and Shah, [Ref. 2],
and Yoon et al, [Ref. 3].
In order to obtain four unknown independent coefficients C 1 C 2 C 3 C 6 with the assumption of C 4 = C 5 = 1
(isotropic properties for transverse directions), it is necessary to use four stresses from the experimental data
0 45 90 b , where 0 45 90 are the tensile yield stresses at 0o, 45o, and 90o from the rolling direction, and
b is the balanced biaxial yield stress measured from bulge test. A detailed procedure to calculate the coefficients of
Barlats yield function are summarized in the work of Yoon at el., [Ref. 4]. The Barlats coefficients are automatically
calculated from user inputs for 0 45 90 b in Mentat. If biaxial data is not available, thr program assumes
b = 0 . Barlats yield function can be accessed through MATEP.
f = J 1 + J 21 2 ------- = 0 (10-112)
3
where
J1 = i i (10-113)
1
J 2 = --- (10-114)
2 ij ij
3
c = ---------------------------------------------- ; ---------------------------------- = sin (10-115)
12 1 3 2 1 2
3 1 12 2
Main Index
Chapter 10: Materials 519
Inelastic Behavior
Yield Envelope
R
c
x + y
2
f = 3J 2 + 3J 1 1 2 = 0 (10-116)
2 2
2
= 3 c ------ = --------------------------------------------- (10-117)
3 3 3c 2 1 2
2
R c
x + y c2
2
Figure 10-32 Resultant Yield Condition of Plane Strain (Parabolic Mohr-Coulomb Material
Main Index
520 Nonlinear Users Guide
Inelastic Behavior
Original
Isotropic Hardening
The isotropic workhardening rule assumes that the center of the yield surface remains stationary in the stress space,
but that the size (radius) of the yield surface expands, due to workhardening. The change of the von Mises yield surface
is plotted in Figure 10-33a.
A review of the load path of a uniaxial test that involves both the loading and unloading of a specimen will assist in
describing the isotropic workhardening rule. The specimen is first loaded from stress free (point 0) to initial yield at
point 1, as shown in Figure 10-33b. It is then continuously loaded to point 2. Then, unloading from 2 to 3 following
the elastic slope E (Youngs modulus) and then elastic reloading from 3 to 2 takes place. Finally, the specimen is
plastically loaded again from 2 to 4 and elastically unloaded from 4 to 5. Reverse plastic loading occurs between 5
and 6.
It is obvious that the stress at 1 is equal to the initial yield stress y and stresses at points 2 and 4 are larger than y ,
due to workhardening. During unloading, the stress state can remain elastic (for example, point 3), or it can reach a
subsequent (reversed) yield point (for example, point 5). The isotropic workhardening rule states that the reverse yield
occurs at current stress level in the reversed direction.
Main Index
Chapter 10: Materials 521
Inelastic Behavior
5 6
0
3
1
2
4
1 2
4
2
1
y
E
E +4
E
3
0
6
(b) Loading Path
Figure 10-33 Schematic of Isotropic Hardening Rule (Uniaxial Test)
Let 4 be the stress level at point 4. Then, the reverse yield can only take place at a stress level of 4 (point 5).
Main Index
522 Nonlinear Users Guide
Inelastic Behavior
Kinematic Hardening
Under the kinematic hardening rule, the von Mises yield surface does not change in size or shape, but the center of the
yield surface can move in stress space. Figure 10-33a illustrates this condition. Zieglers law is used to define the
translation of the yield surface in the stress space.
The loading path of a uniaxial test is shown in Figure 10-33b. The specimen is loaded in the following order: from
stress free (point 0) to initial yield (point 1), 2 (loading), 3 (unloading), 2 (reloading), 4 (loading), 5 and 6 (unloading).
As in isotropic hardening, stress at 1 is equal to the initial yield stress y , and stresses at 2 and 4 are higher than y ,
due to workhardening. Point 3 is elastic, and reverse yield takes place at point 5. Under the kinematic hardening rule,
the reverse yield occurs at the level of 5 = 4 2 y , rather than at the stress level of 4 . Similarly, if the
specimen is loaded to a higher stress level 7 (point 7), and then unloaded to the subsequent yield point 8, the stress
at point 8 is 8 = 7 2 y . If the specimen is unloaded from a (tensile) stress state (such as point 4 and 7), the
reverse yield can occur at a stress state in either the reverse (point 5) or the same (point 8) direction.
For many materials, the kinematic hardening model gives a better representation of loading/unloading behavior than
the isotropic hardening model. For cyclic loading, however, the kinematic hardening model can represent neither cyclic
hardening nor cyclic softening.
Note: The data points should be based on a plot of the stress versus plastic strain for a tensile test. The elastic
strain components should not be included.
The yield stress and the workhardening data must be compatible with the procedure used in the analysis. For small
strain analyses, the engineering stress and engineering strain are appropriate. If only PARAM,LGDISP is used, the yield
stress should be entered as the second Piola-Kirchhoff stress, and the workhard data be given with respect to plastic
Green-Lagrange strains. If PARAM,LGDISP,1 or 2 are used, the yield stress must be defined as a true or Cauchy stress
with respect to logarithmic plastic strains.
Flow Rules
Yield stress and workhardening rules are two experimentally related phenomena that characterize plastic material
behavior. The flow rule is also essential in establishing the incremental stress-strain relations for plastic material. The
p
flow rule describes the differential changes in the plastic strain components d as a function of the current stress state.
So long as a material point is elastic, Hookes law provides a relationship between total stress and strain. After a
material becomes plastic, however, there is no longer a unique relationship between total stress and strain. The problem
then is usually solved incrementally, following the exact loading path.
Main Index
Chapter 10: Materials 523
Inelastic Behavior
For points which are plastic, a flow rule is used to relate increments of stress to plastic strain. SOL 400 uses an
associated flow rule, which prescribes that increments of plastic strain are computed as a constant times the gradient
of the yield function.
In other words, considering the yield function as a surface in stress space, the plastic strain increment is a vector in the
direction of the outward normal to the surface at the point where it is touched by the stresses on the loading path.
The equation representing this is:
F
d ijp = ---------- (10-118)
ij
F
p = -----------
d xx -
xx
F
p = -----------
d yy -
yy
F
d zpz = -----------
zz
(10-119)
F
d xpy = -----------
x y
F
p = ----------
d yz
y z
F
p = ----------
d zx
x z
These stress vs. plastic strain equations are analogous to the stress vs. total strain equations of elasticity, where elastic
strains can be computed as the gradient of a strain energy potential function; namely:
U
d ij = ---------- (10-120)
ij
Thus, the yield function F plays the role of a plastic potential. If a theory of plasticity uses something other than the
yield function as a plastic potential, a so-called nonassociated flow rule results. Nonassociated flow rules are not
available in SOL 400.
The Prandtl-Reuss representation of the flow rule is available in SOL 400. In conjunction with the von Mises yield
function, this can be represented as:
p
d ij = d p ------------ (10-121)
ij
where d p and are equivalent plastic strain increment and equivalent stress, respectively.
Main Index
524 Nonlinear Users Guide
Inelastic Behavior
The significance of this representation is illustrated in Figure 10-34. This figure illustrates the stress-space for the
two-dimensional case. The solid curve gives the yield surface (locus of all stress states causing yield) as defined by the
von Mises criterion.
Equation (10-136) expresses the condition that the direction of inelastic straining is normal to the yield surface. This
condition is called either the normality condition or the associated flow rule.
If the von Mises yield surface is used, then the normal is equal to the deviatoric stress.
d p p
d 2
p
d 1
Yield Surface
Figure 10-34 Yield Surface and Normality Criterion 2-D Stress Space
For the von Mises and modified Hill yield functions programmed in SOL 400, the derivatives in the yield function are
obtained simply by differentiating with respect to individual components of stress. For example, for the modified Hill
function, we have:
2 x x y y zz x y
p = -----------
d xx - -----------
- -----------
- d xpy = ---------
Fx Fy Fx Fz 2
F x2 F xy
x x 2 yy z z x z
p = -----------
d yy - + -----------
- -----------
- d y2z = -------- (10-122)
Fx Fy F y2 Fy Fz F y2z
xx yy 2 zz y z
p = -----------
- -----------
- + ----------
- p = --------
d zx
d zz
Fx Fz Fy Fz F z2 F z2x
The constant in these flow rule equations is evaluated automatically by SOL 400 on the basis of material stability
during plastic flow (i.e., by the requirement that the stress state remain on the yield surface during plastic straining).
Main Index
Chapter 10: Materials 525
Inelastic Behavior
become more pronounced for temperatures greater than half the melting temperature ( T m ), but are sometimes present
even at room temperature. The following discussion explains the effect of strain rate on the size of the yield surface.
Using the von Mises yield condition and normality rule, we obtain an expression for the stress rate of the form
p
ij = L ij kl k l + r ij
L i jk l = C i jk l C ij mn ------------- ------------ C p q k l D (10-123)
m n p q
and
2
r i j = C ojmn ------------- --- -------- D (10-124)
m n 3 p
where
4
D = --- 2 -------- + ---------- C ij kl ----------- (10-125)
9 p ij k l
As strain rates increase, many materials show an increase in yield strength. The model provided in SOL 400 for this
purpose is
P
where = D ------ 1 for 0
0
Main Index
526 Nonlinear Users Guide
Inelastic Behavior
3. The yield stress may be given as a function of the plastic strain, strain, and/or the temperature using the TABLD3
bulk data entry.
4. The Power Law, Rate Power Law, Johnson Cook model, or Kumar model.
Note: If multiple material models are used, they must all be expressed as piecewise linear or as Cowper and
Symonds model.
Perfectly Plastic
A material is said to be perfectly plastic if, upon the stress state touching the yield surface, an infinitesimal increase
in stress causes an arbitrarily large plastic strain. The uniaxial stress-strain diagram for an elastic-perfectly plastic
material is shown in Figure 10-35. Some materials, such as mild steel, behave in a manner which is close to perfectly
plastic.
xx
.
YS
xx
where y is the yield stress, 0 is the strain corresponding to initial yield stress, p is equivalent plastic strain, p is
rate of equivalent plastic strain rate and A , B , m , and n are material parameters.
Main Index
Chapter 10: Materials 527
Inelastic Behavior
Johnson-Cook Model
n p T Tr o o m m
----------------------------------
y = A + B p 1 + C ln ----
- 1 T -
0 m el t T r o o m
where y is the yield stress, p is the equivalent plastic strain, p is the current equivalent plastic strain rate, 0 is
strain rate of material characterization and A , B , C , m and n are material parameters. T , T r o o m , T m el t are,
respectively, the current, room, and melting temperatures of the material in absolute scale.
The following conditions should be noted for the Johnson-Cook model.
T should be between T r o o m and T m el t . If T T r o o m , T is set to T r o o m . If T T me lt , T = T m el t 0.01 .
0 0 and 0 . If either condition is violated, the middle term in the above expression is set to 1.0.
Kumar Model
1n
p
y = B0 * sin h 1 ----- eQ nRT
A
where y is the yield stress, p is equivalent plastic strain rate, Q is the activation energy, T is temperature, and A ,
B 0 , and n are material parameters.
Main Index
528 Nonlinear Users Guide
Inelastic Behavior
Figure 10-36 Typical Behavior of Material that can be Simulated with Cyclic Plasticity Model
X is the back stress tensor representing the center of the yield surface in stress space.
Isotropic Hardening/Softening
The isotropic hardening/softening determines the size of the elastic region during the plastic loading. In this model, it
is controlled by parameter R and k . The initial conditions of cyclic hardening are given as k = y and R = 0 ,
Main Index
Chapter 10: Materials 529
Inelastic Behavior
while a cyclic softening is initially described by k = y R 0 and R = R 0 . The evolution equation for the variable
R is described as follows:
R = b R R
where b and R are material constants. R represents the limit of the isotropic hardening/softening. In case of
b e
hardening, then R = R 1 e
p s
.
where C and are two material constants. = 0 stands for linear-kinematic rule.
Plastic-strain-range Memorization
Several experimental observations show that the asymptotic stress value of cyclic hardening can depend on the prior
history. The influence of plastic-strain range on the stabilized cyclic response is evident from the comparison between
the different histories of loading used to obtained the cyclic curve. Therefore, an introduction of new internal variables
that memorize the prior maximum plastic range is introduced by defining a memory surface in the plastic strain space
as follows:
2
F = --- e p
3
The evolution of the state variables are as follows
= H F n n *
= 3 2 1 H F n n * n *
where n and n * are the unit normal to the yield surface f = 0 and to the memory surface F = 0 defined as follows:
2 p p
2--- -----------
---- *
n = ---
3 - and n = 3
-
The coefficient is introduced in order to induce a progressive memory. For = 0.5 then, the memorization is
instantaneous and stabilization occurs after one cycle. A progressive memory is given by 0.5 .
Main Index
530 Nonlinear Users Guide
Inelastic Behavior
The dependency between cyclic plastic flow and the plastic strain range is introduced by considering an asymptotic
isotropic state as follows:
2
R = Q M + Q 0 Q M e
f 3 s
where a = ------ = --- ------------- .
2R + k
Since the process involves nonlinear equation, iteration process using predictor-corrector technique is used. The
predictor is calculated using the trial elastic stresses as follows:
B = A + L (10-127)
and then calculate f B based on B and hardening parameter on A. Using the Taylor expansion at B, then
fB
= ------------------------------------------------------------------------------------------------------------------ (10-128)
C
a T La + ------------- a T X a T X + b R R
R+k
Having the global plasticity iteration converged, then the iteration to satisfy the plastic strain memorization is started.
If both iterations are converged, then the total plasticity iteration is considered completed.
Inserting equation (10-129) into equation (10-128) and using equation (10-126), the elasto-plastic classical tangent
modular matrix can be derived as follows:
T
aa L
= L I ------------------------------------------------------------------------------------------------------------------
EP
L (10-129)
T C T T
a La + ------------- a X a X + b R R
R+k
Geological Materials
Data for geological materials are most commonly available from triaxial compression testing. In such a test, the
specimen is confined by pressure and an additional compression stress is superposed in one direction. Thus, the
principal stresses are all negative, with 0 1 = 2 3 .
Main Index
Chapter 10: Materials 531
Inelastic Behavior
3 1
1 = 2 3 1 2 = 3
1 2 2 3
p = 1 3 2 1 + 3
q = 1 3
r 3 = 1 3 3
so that t = q = 1 3
The triaxial results may thus be plotted in the t-p plane shown above. Fitting the best straight line through the results
then provides and d .
Triaxial tension data are also needed to define the bulk modulus K . Under triaxial tension, the specimen is again
confined by pressure, then the pressure in one direction is reduced. In this case, the principal stresses are
1 2 = 3
p = 1 3 1 + 2 3 ,
q = 1 3
r 3 = 1 3 3 ,
so that t = q K = 1 K 1 3
Main Index
532 Nonlinear Users Guide
Inelastic Behavior
K may thus be found by plotting these test results as q versus p and again fitting the best straight line. The triaxial
compression and tension lines must intercept the p-axis at the same point, and the ratio of values of q for triaxial
tension and compression at the same value of p then gives K as shown in Figure 10-38.
hc ht
d
p
s + m sin c cos = 0 ,
where s = 1 2 1 3
is half of the difference between the maximum and minimum principal stresses (and is, therefore, the maximum shear
stress), and
m = 1 2 1 + 3
Main Index
Chapter 10: Materials 533
Inelastic Behavior
Using the results above (for the stress invariants p, q, and r), in triaxial compression and tension, allows the Drucker-
Prager model to be written for triaxial compression as
1
1 --- tan
tan 3 0
1 3 + -------------------------------- 1 + 3 + -------------------------- c = 0 (10-131)
2 + 1 1
--- tan 1 + --- tan
3 6
1
1 --- tan
tan 3 0
1 3 + ----------------------------- 1 + 3 + ------------------------- c = 0 (10-132)
2- 1
--- 1 1
--- tan ---- --- tan
K 3 K 6
We wish to make the equations for triaxial compression and biaxial tension identical to the general Mohr-Coulomb
equation for all values of 1 3 .
Comparing the equations for triaxial compression and triaxial tension requires that:
1 1 1
1 + --- tan = ---- --- tan (10-133)
6 K 6
so that
1
K = -------------------------- (10-134)
1
1 + --- tan
3
Comparing the coefficients of 1 + 3 in the equation for triaxial compression and that for triaxial tension provides:
6 sin
tan --------------------- (10-135)
3 sin
3 sin
K = ----------------------- (10-136)
3 + sin
Main Index
534 Nonlinear Users Guide
Inelastic Behavior
Finally, comparing the last terms in the general expression for the Mohr-Coulomb model and the equation for triaxial
compression and using the expression for tan provides:
0 2c cos
c = ---------------------- (10-137)
1 sin
The expression for tan , K , and this last expression, and thus, provide Drucker-Prager parameters that match the
Mohr-Coulomb model in triaxial compression and tension.
The value of K in the Drucker-Prager model is restricted to K 0.778 for the yield surface to remain convex.
Rewriting the expression for K as:
1K
sin = 3 -------------- (10-138)
1 + k
shows that this implies 22 . Many real materials have a larger Mohr-Coulomb friction angle than this value. In
such circumstances, one approach is to choose K = 0.778 and then to use the expression for tan to define and
0 0
the expression for c to define c , ignoring the expression for K . This matches the models for triaxial compression
only, while providing the closest approximation that the model can provide to failure being independent of the
intermediate principal stress. If is significantly larger than 22, this approach may provide a poor Drucker-Prager
match of the Mohr-Coulomb parameters. SOL 400 uses K = 1 by default.
pl 1
= d -------------------------- t p tan
pl
d (10-139)
1
1 --- tan
3
pl
where d is the equivalent plastic strain increment.
Since we only wish to match the behavior in one plane we can assume K = 1 , which implies that t = q . Then:
p
= d -------------------------- q tan
pl pl 1
d (10-140)
1
1 --- tan
3
Main Index
Chapter 10: Materials 535
Inelastic Behavior
pl pl 1 1 1
d 1 = d -------------------------- ------ 2 1 2 3 + --- tan (10-141)
1 2q 3
1 --- tan
3
pl pl
with similar expressions for d 2 and d 3 .
pl
Assume plane strain in the 1-direction. Then, at limit load, we must have d 1 = 0 . From the above expression, this
provides the constraint:
1 1
------ 2 + --- tan = 0 (10-142)
2q 1 2 3 3
so that:
1 1
1 = --- 2 + 3 --- tan q (10-143)
2 3
Using this constraint, we can rewrite q and p in terms of the principal stresses in the plane of deformation,
3 3
q = --------------------------------------- 2 3 (10-144)
2
2 9 tan
and
1 tan
p = --- 2 + 3 + ----------------------------------------------- 2 3 (10-145)
2 2
2 3 9 tan
With these expressions, the Drucker-Prager yield surface can be written in terms of 2 and 3 as
9 tan tan - 1
---------------------------------------------- 2 3 + --- tan 2 + 3 d = 0 (10-146)
2 2
2 3 9 tan
By comparison,
2
tan 3 9 tan
sin = -------------------------------------------------------- (10-148)
9 tan tan
Main Index
536 Nonlinear Users Guide
Inelastic Behavior
2
3 9 tan
c cos = -------------------------------------------
-d (10-149)
9 tan tan
Now consider the two extreme cases of flow definition: associated flow, = , and nondilatant flow, when = 0 .
3 sin
tan = -------------------------------------- (10-150)
1 2
1 + --- sin
3
and
d 3 cos
--- = -------------------------------------- (10-151)
c 1 2
1 + --- sin
3
d
while for nondilatant flow they give tan = 3 sin and --- = cos
c
0
In either case, c is immediately available as:
1
c0 = ------------------------- d (10-152)
1
1 --- tan
3
The difference between these two approaches increases with the friction angle but, for typical friction angles, the
results are not very different, as illustrated in the table below.
Mohr-Coulomb
Friction Angle, Associated Flow Nondilatant Flow
Drucker-Prager friction angle, d/c Drucker-Prager friction angle, d/c
10 16.7 1.70 16.7 1.70
Main Index
Chapter 10: Materials 537
Inelastic Behavior
As strain rates increase, many materials show an increase in yield strength. This effect often becomes important when
the strain rates are in the range of -0.1 to 1 per second, and can be very important if the strain rates are in the range of
10 to 100 per second, as commonly occurs in high energy dynamic events or in manufacturing processes.
Temperature-Dependent Behavior
This section discusses the effects of temperature-dependent plasticity on the constitutive relation.
The following constitutive relations for thermo-plasticity were developed by Naghdi. Temperature effects are
discussed using the isotropic hardening model and the von Mises yield condition.
The stress rate can be expressed in the form
ij = L ij kl k l + h i j T
L i jk l = C i jk l C ij mn ------------- ------------ C p q k l D (10-153)
m n p q
L i jk l = C i jk l (10-154)
The term that relates the stress increment to the increment of temperature for elastic-plastic behavior is
2
h ij = X ij C ij kl k l C i jkl ----------- p q X p q --- ------- D (10-155)
k l 3 T
H ij = X ij C ij kl k l (10-156)
where
4
D = --- 2 -------- + ---------- C ij kl ----------- (10-157)
9 p ij kl
and
C ij kl e
X i j = --------------- (10-158)
T kl
Main Index
538 Nonlinear Users Guide
Inelastic Behavior
Main Index
Chapter 10: Materials 539
Inelastic Behavior
Entry Description
MATEP Specifies elasto-plastic material properties to be used for large deformation analysis.
MATTEP Specifies temperature-dependent elasto-plastic material properties to be used for static, quasi-
static, or transient dynamic analysis.
Notes: Some versions of Patran use the same menus to support both SOL 400 and SOL 600.
Tresca is not supported in SOL 400.
Table 10-4 shows the various input options and criteria available to you for defining elastoplastic behavior.
Main Index
540 Nonlinear Users Guide
Inelastic Behavior
Tables 10-5 through 10-7 provide descriptions for the input data for each of the four types of nonlinear input.
Table 10-5 Isotropic - Stress/Strain Curve or Perfectly Plastic: All Yield Functions
Property Name Description
Stress /Strain Curve or Yield Defines the Cauchy stress vs. logarithmic strain (also called equivalent tensile
Stress stress versus total equivalent strain) by reference to a tabular field. The field is
selected from the Field Definition list. The field is created using the Fields
application. For Perfectly Plastic models, only a Yield Stress needs to be entered.
Can also be strain rate dependent if Strain Rate Method is Piecewise Linear.
Accepts field of yield stress vs. strain rate.
Coefficient C Visible if Strain Rate Method is Cowper-Symonds.
Inverse Exponent P Visible if Strain Rate Method is Cowper-Symonds.
Alpha When set to Linear Mohr-Coulomb, defines the slope of the yield surface in square
root J2 versus J1 space. This property is required.
Beta When set to Parabolic Mohr-Coulomb, defines the beta parameter in the equation
that defines the parabolic yield surface in square root J2 versus J1 space. This
property is required.
Note: Perfectly Plastic is identical to Stress/Strain except that no hardening rules apply.
Table 10-6 Anisotropic/Orthotropic - Stress/Strain Curve or Perfectly Plastic: All Yield Functions
Property Name Description
Stress vs. Strain or Tensile Same as description for Isotropic Elastic-Plastic. If Strain Rate Method is Piecewise
Yield Stress Linear, accepts field of yield stress vs. strain rate.
Or defines an isotropic yield stress. It is a required property when the Plasticity Type
is Perfectly Plastic.
Main Index
Chapter 10: Materials 541
Inelastic Behavior
Table 10-6 Anisotropic/Orthotropic - Stress/Strain Curve or Perfectly Plastic: All Yield Functions
Property Name Description
Stress 11/22/33 Yield Ratios Defines the ratios of direct yield stresses to the isotropic yield stress in the elements
coordinate system.
Stress 12/23/31 Yield Ratios Defines the ratios of shear yield stresses to the isotropic shear yield stress (yield
divided by square root three) in the elements coordinate system.
Note: Perfectly Plastic is identical to Elastic-Plastic except that no hardening rules apply.
Main Index
542 Nonlinear Users Guide
Failure and Damage Models (MATF)
-----x- for x 0
V1
F1 =
x
--------- for x 0
V2
-----y- for y 0
V3
F2 =
y
--------- for y 0
V4
-----z- for z 0
V5
F3 =
z
--------- for z 0
V6
x y
F 4 = -------
V7
Main Index
Chapter 10: Materials 543
Failure and Damage Models (MATF)
y z
F 5 = -------
V8
zx
F 6 = -------
V9
------x- for x 0
W1
F1 =
x
-------- for x 0
W2
------y- for y 0
W3
F2 =
y
-------- for y 0
W4
------z- for z 0
W5
F3 =
z
-------- for z 0
W6
Main Index
544 Nonlinear Users Guide
Failure and Damage Models (MATF)
x y
F 4 = -------
W7
y z
F 5 = -------
W8
zx
F 6 = -------
W9
Hill Failure Criterion there is no distinction between tensile and compressive behavior. A single failure index is
calculated as:
2 2 2
x y z 1 1 1 1 1 1
F = [ ------ + ------ + ------ ------ + ------ ------ x y ------ + ------ ------ x z
2 2 2 2 2 2
V 1 V 3 Z 5 V 1 V 3 V 5 V 1 V 5 V 23
2 2
2 2 2
1 1 1 xy yz zx
------ + ------ ------ y z + ------- + ------- + ------- /Find
V 23 V 25 V 21 2
V7 V8 V9
2 2
Main Index
Chapter 10: Materials 545
Failure and Damage Models (MATF)
Hoffman Failure Criterion introduces distinction between tensile and compressive stresses to generalize the Hill
Failure Criterion. A single failure index, F, is calculated as:
1 1
F = C x x y + C y y z + C z z x + ------ ------ x
2 2 2
V 1 V 2
2 2 2
1 1 1 1 x y yz z x
+ ------ ------ y + ------ ------ z + ------- + ------- + ------- /Find
V 3 V 4 V 5 V 6 2 2 2
V7 V8 V9
with
1 1 1 1
C x = --- ------------- + ------------- -------------
2 V 1 V 2 V 3 V 4 V 5 V 6
1 1 1 1
C y = --- ------------- + ------------- -------------
2 V 3 V 4 V 5 V 6 V 1 V 2
1 1 1 1
C z = --- ------------- + ------------- -------------
2 V 5 V 6 V 1 V 2 V 3 V 4
Main Index
546 Nonlinear Users Guide
Failure and Damage Models (MATF)
Tsai-Wu Failure Criterion is another generalization of the Hill Failure Criterion. A single failure index F is
calculated as:
2 2 2
1 1 1 1 1 1 x y z
-----
- ------ x + ------ ------ y + ------ ------ z + ------------
F = [ V1 V2 V 3 V 4 V 5 V 6
- + ------------
V1 V2 V3 V4 V5 V6
- + ------------
-
2 2 2
xy yz zx
+ ------- + ------- + ------- + 2V 10 x y + 2V 11 y z + 2V 12 x z /Find
2 2 2
V7 V7 V7
V 10 Real; no Default
V 11 Real; no Default
V 12 Real; no Default
Main Index
Chapter 10: Materials 547
Failure and Damage Models (MATF)
2 1 1 2 1 1 2 1 1
V 10 ------------- ------------- V 11 ------------- ------------- V 12 ------------- -------------
V1 V2 V3 V4 V3 V4 V5 V6 V1 V2 V5 V6
2 1 2
-----x- + ------- 2
- x y + xz x 0
V 1 2
F1 = V 10
0 x 0
x
--------
- x 0
F2 = V2
0 x 0
1 2 1 2 1 2 2
------ y + z + -------- yz y z + -------- x y + xz y + z 0
2 2 2
F3 = V3 V 11 V 10
0 y + z 0
V4 2
1- ----------- 1 1- 2 1
- 1 y + z + ------------ y + z + -------
2 2 2
----- yz y z + -------- x y + xz y + z 0
2V 11 2 2 2
F4 = V4 4V 11 V 11 V 10
0 y + z 0
Puck Failure Criterion five failure indices are calculated for the plane stress case and four for the general case. The
first two failure indices are related to fiber failure:
-----x- x 0
V
F1 = 1
0 x 0
Main Index
548 Nonlinear Users Guide
Failure and Damage Models (MATF)
x
--------
- x 0
F2 = V2
0 x 0
For plane stress failure, indices 3, 4, and 5 are related to matrix failure. For other cases, plane stress failure indices 3
and 4 are used for the matrix failure, and 5 is not used and is set to zero. The following material parameters are used:
W3 p23c, slope 3 of failure envelope (Real > 0.0 or blank; Default; calculated internally; this is recommended for
plane stress)
W4 p23t, slope 4 of failure envelope (Real > 0.0; Default W 3 , not used for plane stress)
Hashin-Tape Criterion a variant of the Hashin criterion adapted for tape type of materials. Four failure indices are
calculated:
2 xy 2 xz 2
-----x- + -------
- + -------
- x 0
F1 = V 1
V 10
V 12
0 x 0
2 xy 2 xz 2
-----x- + -------
- + -------
- x 0
F 2 = V 2 V 10 V 12
0 x 0
1- 2 y z x y 2 x z 2 y z 2
----- y + z ------------ + -------- + -------- + -------- y + z 0
2 2 V 10 V 12 V 11
F3 = V3 V 11
0 y + z 0
Main Index
Chapter 10: Materials 549
Failure and Damage Models (MATF)
V4 2
----------- y + z 1 y z xy 2 xz 2 yz 2 2
- 1 ----------------------- - + -------- + -------- + -------- + V 5 -----x-
- y + z 2 -----------
+ ----------- y + z 0
2 V 11 V 2 2 V V V V
F4 = 4 4 V 11 V 11 10 12 11 6
0 y + z 0
V2 Maximum tape fiber compressive stress (Real > 0.0 or blank; Default V 1 )
V4 Maximum tape cross-fiber compressive stress (Real > 0.0 or blank; Default V 3 )
V 12 Maximum z-x transverse shear stress. (Real > 0.0 or blank; Default V 11 )
V5 Maximum fiber tensile stress for matrix compression (Real > 0.0 or blank; required if V 6 = 1.0 ; otherwise,
not used)
V6 Contribution factor for V 5 (Real 0.0 or 1.0; Default = 0.0)
The Hashin-Fabric criterion is a variant of the Hashin criterion adapted for fabric type of materials. Six failure indices
are calculated:
2 xy 2 xz 2
-----x- + -------
- + -------
- x 0
F1 = V1 V 10 V 12
0 x 0
2 xy 2 xz 2
-----x- + -------
- + -------
- x 0
F 2 = V 2 V 10 V 12
0 x 0
2 xy 2 xz 2
-----y- + -------
- + -------
- y 0
F 3 = V 3 V 10 V 12
0 y 0
2 xy 2 xz 2
-----y- + -------
- + -------
- y 0
F 4 = V 4 V 10 V 12
0 y 0
Main Index
550 Nonlinear Users Guide
Failure and Damage Models (MATF)
2 xy 2 xz 2 y z 2
-----z- + -------
- + -------
- + -------
- z 0
F 5 = V 5 V 10 V 12 V 11
0 z 0
2 xy 2 xz 2 y z 2
-----z- + -------
- + -------
- + -------
- z 0
F 6 = V 6 V 10 V 12 V 11
0 z 0
V2 Maximum first fiber compressive stress (Real > 0.0 or blank; Default V 1 )
V4 Maximum second cross-fiber compressive stress (Real > 0.0 or blank; Default V 3 )
V 12 Maximum z-x transverse shear stress. (Real > 0.0 or blank; Default V 11 )
A MATTF entry with the same MID as MATF may be used to specify the temperature variation of the failure criterion
values.
User Defined Criterion can be defined by the UFAIL user subroutine.
As an example of the usage of MATF, MATUDS, and UFAIL user subroutine for defining a failure criteria, see User
Defined Service Ease of Use (Ch. 87) in the MSC Nastran Demonstration Problems Manual.
Notes: UFAIL user subroutine is used along with MATUDS bulk data entry.
In SOL 400, MATF is only supported for elements with property extensions. This implies that for such
elements, PBAR/PBARL, PBEAM/PBEAML, PCOMP/PCOMPG, PROD, PSHEAR, PSHELL, PLPLANE,
and PSOLID should be associated with PBARN1, PBEMN1, PSHLN1, PRODN1, PSHEARN, PSHLN1,
PSHLN2, and PSLDN1, respectively.
Stress limits such as ST, SC, SS, X t , X c , Y t , Y c in the MAT1, MAT2, and MAT8 entries are not used in
SOL 400.
Stress and/or strain allowables in all directions should be defined if the default is not appropriate. Be sure
not to set some of these values to low numbers.
Main Index
Chapter 10: Materials 551
Progressive Composite Failure
r i = 1 e 1 F (10-159)
This is done differently for different failure criteria as described below. Six such reduction factors are stored and
updated. They are then used for scaling the respective material modulus according to
new orig
E 11 = r 1 E 11 (10-160)
new orig
E 22 = r 2 E 22 (10-161)
new orig
E 33 = r 3 E 33 (10-162)
new orig
G 12 = r 4 G 12 (10-163)
new orig
G 23 = r 5 G 23 (10-164)
new orig
G 31 = r 6 G 31 (10-165)
The Poissons ratios are scaled in the same way as the corresponding shear modulus.
For the maximum stress and maximum strain criteria the reduction factors are calculated separately from each separate
failure index: r 1 is calculated from the first failure index as given by equation 10-160 above, r 2 is calculated from the
second failure index from equation 10-161 etc. Thus, there is no coupling of the different failure modes for these
criteria.
For the failure criteria which only have one failure index: Tsai-Wu, Hoffman and Hill, all six reduction factors are
decreased in the same way, using the smallest of the r i :s .
Main Index
552 Nonlinear Users Guide
Progressive Composite Failure
For the criteria which distinguish between fiber and matrix failure (Hashin, Hashin-tape and Puck), there is a more
complex coupling between the failure modes. There is a default behavior which can be influenced by a number of input
parameters. The default is as follows.
r 1 depends on fiber failure (first and second failure index)
r 2 depends on matrix failure (third, fourth and fifth failure index)
r 3 behaves the same as r 1 ( r 3 = r 1 )
r 4 behaves the same as r 2 ( r 4 = r 2 )
r 5 and r 6 behave the same as r 4 ( r 5 = r 6 = r 4 )
In the MATF option, there are five parameters available for controlling the way the moduli are reduced. With the
exception of a 1 , they are only used for the Hashin variants and Puck.
a1 Residual stiffness factor: The stiffness is never reduced to less than this factor. The default is 0.01.
a2 Matrix compression factor. With this factor, r 2 can reduce less due to failure in matrix compression.
Experiments show that certain materials show less degradation of the matrix properties in compression than in tension.
See for example [Ref. 6]. For the case that F m c indicates failure in matrix compression, equation 10-159 is modified
into
1F
r 2 = 1 a 2 1 e mc (10-166)
a3 Shear stiffness factor: This factor is used for taking into account the effect that the shear stiffness G 12 can reduce
less than the matrix stiffness E 2 . With F m indicating a matrix failure we have
1 Fm
r 4 = 1 a 3 1 e (10-167)
The combined effect of a 2 and a 3 on the shear stiffness reduction for the case of matrix compression failure is then
1 F mc
r 4 = 1 a 2 1 a 3 1 e
a4 E33 reduction from fiber failure: This factor controls the reduction of E 33 due to fiber and matrix failure. The
default is as mentioned above that E 33 reduces due to fiber failure. With this factor this can be changed to vary linearly
with fiber and matrix failure. With F f indicating a fiber failure and F m a matrix failure we have
1F 1F
r 3 = 1 a 4 1 e f a4 1 e m (10-168)
a5 Shear reduction from fiber failure: With this factor it is possible to control the reduction of the shear stiffness
due to fiber failure. By default it only reduces due to matrix failure. With F f and F m as in the previous we have
Main Index
Chapter 10: Materials 553
Progressive Composite Failure
1F 1F
r 4 = 1 a 5 1 e m a5 1 e f (10-169)
The Hashin-fabric failure criterion reduces the first three reduction factors from the respective failure index. The three
shear reduction factors are taken from the worst of the first three factors. The factors a 2 through a 5 are not used for
this criterion.
In addition, it is also possible to use the UPROGFAIL user subroutine to explicitly define the reduction factors r 1
through r 6 . The MATUSR bulk data entry is required to activate the UPROGFAIL user subroutine.
Interface
Digimat-CAE/MSC Nastran SOL 400 contains the material library containing the Digimat capabilities and the
required interfaces in order to be linked with the MSC Nastran SOL 400 Finite Element (FE) solver. Linking MSC
Nastran SOL 400 libraries with Digimat-CAE/MSC Nastran SOL 400 gives the user access to all linear and nonlinear
small-strain material models available in Digimat for FE small-strain analyses, just like any other MSC Nastran SOL
400 material models. It also enables to take into account fiber orientation computed by an injection molding code. The
interface thus allows to model the impact of the injection process on the structural behavior of composite parts. Note
that not all the available Digimat material models can be used with the Digimat-CAE/MSC Nastran SOL 400 interface.
Main Index
554 Nonlinear Users Guide
Progressive Composite Failure
myAnalysis.mat file, the Digimat material file subsequently used for the structural FE analysis; i.e., the
main input file;
myAnalysis.nas file, the file containing the information related to the MSC Nastran SOL 400 user
material definition, subsequently copied in the MSC Nastran input file as well as the meaning of the state
variables that will be written in the MSC Nastran postprocessing files (.op2 files).
myAnalysis.log file, the file containing the run information including error messages in case of
unsuccessful run.
Main Index
Chapter 10: Materials 555
Progressive Composite Failure
SUBCASE 1
SUBTITLE=DigimatAnalysis
STEP 1
ANALYSIS = NLSTAT
SPC = 2001
LOAD = 1
NLSTEP = 10
NLSTRESS(PLOT,NLOUT=1)=ALL
Insert the user material related entries in the Bulk Data section.
MATDIGI 1 1 1.4E-009
UDNAME 1
myAnalysis
The entry MATDIGI indicates to MSC Nastran that the material behavior will be modeled by a Digimat
material. The first field after the MATDIGI keyword contains the material identification number. The third field
contains the identification for the UDNAME card. The last field contains the density of the composite
(automatically computed by Digimat).
Define state variables.
To couple Digimat to MSC Nastran SOL 400, a set of external state variables (ESV) must be defined in the
bulk data section through the entry UDSESV. The number of ESV depends on the Digimat material model. The
first state variable corresponds to the temperature, even for a purely mechanical analysis.
UDSESV 72
$
SV2 HV2 SV3 HV3 SV4 HV4 SV5 HV5
SV6 HV6 SV7 HV7 SV8 HV8 SV9 HV9
SV10 HV10 SV11 HV11 SV12 HV12 SV13 HV13
SV14 HV14 SV15 HV15 SV16 HV16 SV17 HV17
...
The number 72 indicates that the Digimat material model requires the definition of 72 external state variables.
The strings HVi correspond to the name of state variables. To get the physical meaning of these names, please
refer to the .nas file previously generated.
The above cited lines are automatically generated by Digimat and inserted in the .nas file. You can directly
copy/paste them from the .nas interface file in the .bdf file.
4. Ask for Digimat output results.
The external state variables defined in UDSESV can be output to MSC Nastran .op2, .h5, and .f06 files by
means of the NLOUT entry. This entry referenced in the subcase definition must be declared in the bulk data
section such as:
NLOUT 1 TOTTEMP
ESV SV2 SV3 SV4 SV5 SV6 SV7
SV8 SV9 SV10 SV11 SV12 SV13
...
The first field after the NLOUT keyword must be identical to the one used in the output request NLSTRESS.
The ESV keyword indicates that the external state variables are requested as output. This section is
automatically generated by Digimat and written in the .nas file.
Main Index
556 Nonlinear Users Guide
Progressive Composite Failure
The supported MSC Nastran SOL 400 formulations for shell elements are:
CQUAD4
CQUAD8
CTRIA3
CTRIA6
The supported MSC Nastran SOL 400 formulations for solid elements are:
CPENTA
CHEXA
CTETRA
As an example of the usage of MATDIGI for defining a composite material, see Interfacing Digimat to Nastran SOL400
and SOL700 - Example for solid and shell elements (Ch. 89) in the MSC Nastran Demonstration Problems Manual.
Licensing
The e-Xstream Digimat Material Modeling System model and advanced composites in MSC Nastran require separate
licensing and can be obtained from your local MSC offices.
Main Index
Chapter 10: Materials 557
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D
Composite Brick Elements
Another addition made for composite analysis is the calculation of interlaminar shears. These interlaminar shears are
printed in the local coordinate system above and below each layer selected. These values are also available for
postprocessing. The TSHEAR keyword on the NLMOPTS bulk data entry must be used for activating the parabolic
shear distribution calculations.
In Marc, the distribution of transverse shear strains through the thickness for thick shell and beam elements was
assumed to be constant. From basic strength of materials and the equilibrium of a beam cross section, it is known that
the actual distribution is more parabolic in nature. As an additional option, the formulations for elements 1, 22, 45, 75,
89, 140, 149, 150, and 185 (TSHEAR will be switched off for elements 185, 149 and 150 when the elements are
stacked) have been modified to include a parabolic distribution of transverse shear strain. The formulation is exact for
beam element 45, but is approximate for the other thick shell elements. Nevertheless, the approximation is expected
to give improved results from the previous constant shear distribution. Furthermore, interlaminar shear stresses for
composite beams and shells can now be easily calculated.
The enhancement for interlaminar shear is deactivated when the solid shell or continumm composite elements are
stacked though the thickness.
The generalized stiffness matrix for the complete section excluding transverse shear terms is given by:
F 11 11
F 22 22
F 12
= X Y 12 (10-170)
M 11 Y Z 11
M 22 22
M 12 12
A unique X, Y direction in the plane of the section is defined by a rotation around the element normal which maximizes
the value of X 11 in the above equation.
We then assume that the stresses in the X and Y direction are uncoupled, this gives:
Main Index
558 Nonlinear Users Guide
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements
Fx x X xx Y xx F X yy Y yy
= xx and y y = yy (10-171)
M xx Y x x Z xx xx M yy Y y y Z yy yy
If we assume only bending and transverse shear in the section, all section forces are zero and inverting the above
equation gives:
xx H1 x yy H1 y
= M x x and = My y (10-172)
xx M2 x yy M2 y
For a point in the section, we can now define the stresses as:
xx z = E z x x z = E z x x + xx z = E z H 1 x + H 2 x z M xx and
(10-173)
y y z = E z yy z = E z yy + y y z = E z H 1 y + H 2 y z M y y
Since we assumed that all stresses in the X and Y direction are uncoupled, the equilibrium conditions through the
thickness are given by:
zx xx z zy yy z
---------- + ------------------- = 0 and ---------- + ------------------- = 0 (10-174)
z x z y
M x x M y y
V x + ------------- = 0 and V y + ------------- = 0 (10-175)
x y
zx zy
---------- = E z H 1 x + H 2 x z V x and ---------- = E z H 1y + H 2 y z V y (10-176)
z z
We can integrate this through the thickness giving:
t2 y
zx z = E z H 1 x + H 2 x z V x dz and zy z = E z H 1 y + H 2 y z V y dz (10-177)
t 2 t 2
with the boundary conditions that the shear stresses at the top and the bottom of the shell are zero.
We can now define the transverse shear stiffness by matching the shear strain energy over the section obtained with
the transverse shear stresses given in equation (10-177). This yields the flexibility matrix S:
Main Index
Chapter 10: Materials 559
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements
t2
1 Vx 1 G xx z G xy z z x z
--- V V S = ---
2 x y Vy 2 zx
z zy z dz
G yx z G yy z z y z
(10-178)
t 2
where G is the transverse shear flexibilities of the material through the thickness.
Inversion of the flexibility matrix S gives the transverse shear stiffness of the section.
Main Index
560 Nonlinear Users Guide
Cohesive Zone Modeling (MCOHE)
Considering a 3-D interface element, the relative displacement components are given by one normal and two shear
components, expressed with respect to the local element system:
t op bottom
vn = u1 u1
top bottom (10-179)
vs = u2 u2
top b ot t o m
vt = u3 u3
Based on the relative displacement components, the effective opening displacement is defined as:
2 2 2
v = vn + vs + vt (10-180)
Main Index
Chapter 10: Materials 561
Cohesive Zone Modeling (MCOHE)
2G c v if 0 v vc
t = ---------- -----
vm vc
2G c v m v if vc v vm
t = ---------- ------------------ Bilinear (10-181)
v m v m v c
t = 0 if v vm
v v vc
t = G c ----- e Exponential (10-182)
2
vc
2qG c v if 0 v vc
t = ----------------------- -----
vc q + 2 vc
Linear-exponential (10-183)
2qG c q 1 v v if v vc
t = ----------------------- e c
vc q + 2
in which G c is the energy release rate (cohesive energy), v m is the maximum effective opening displacement (which
is only used by the bilinear model), and q is the exponential decay factor (which is only used by the linear-exponential
model).
t t t
v v v
vc vm vc vc
Figure 10-41 Bilinear (left), Exponential (middle), and Linear-exponential (right) Cohesive Material Model
It can easily be verified that the maximum effective traction t c , corresponding to the critical effective opening
displacement v c is given by:
Main Index
562 Nonlinear Users Guide
Cohesive Zone Modeling (MCOHE)
2G c
t c = ---------- Bilinear (10-184)
vm
Gc
t c = -------- Exponential (10-185)
ev c
2qG c
t c = ----------------------- Linear-exponential (10-186)
vc q + 2
So if the maximum effective traction is known, the critical or maximum effective opening displacement can be
determined by:
2G c
v m = ---------- Bilinear (10-187)
tc
Gc
v c = ------- Exponential (10-188)
et c
2qG c
v c = ---------------------- Linear-exponential (10-189)
tc q + 2
Note that for the bilinear model, the critical effective opening displacement does neither affect the cohesive energy nor
the maximum effective traction.
Until now, the behavior in the normal and shear direction is treated similarly. However, sometimes the behavior of an
interface material may be different in tension and shear. The first method to include such differences is incorporated
by the shear-normal stress ratio 1 , which defines the ratio of the maximum stress in shear and the maximum stress
in tension [Ref. 6]. This ratio is used to redefine the effective opening displacement according to:
2 2 2 2 2
v = vn + 1 vs + 1 vt
The effect of 1 = 0.5 is depicted in Figure 10-42 for the bilinear model.
t t
tension shear
only only
Gc Gc
vc vm v vc vm v
Figure 10-42 Response in Tension and Shear for a Shear-normal Stress Ratio 1 = 0.5 (Bilinear Model)
Main Index
Chapter 10: Materials 563
Cohesive Zone Modeling (MCOHE)
Although the use of the shear-normal stress ratio offers some flexibility, it assumes that the cohesive energy in tension
and shear is the same. If one wants to define a different value of the cohesive energy in shear than in tension, the shear-
normal energy ratio 2 can be used. In a general state of deformation, when 2 1 , the curve defining the effective
traction versus the effective opening displacement is defined as a linear combination of the response in pure tension
and pure shear. Using 1 = 0.5 and 2 = 0.75 , Figure 10-43 shows the response in tension and shear for the
bilinear model.
t t
tension shear
only only
Gc 0.75G c
vc vm v vc vm v
Figure 10-43 Response in Tension and Shear for a Shear-normal Stress Ratio 1 = 0.5 and a
Shear-normal Energy Ratio 2 = 0.75 (Bilinear Model)
In order to avoid convergence problems in a finite element simulation of delamination, one may activate so-called
viscous energy dissipation. The basic idea of the dissipation model is that when delamination starts, the rate of
deformation may suddenly increase. This increase is used to augment the constitutive behavior with a viscous
contribution being equivalent to this rate of deformation:
t c v
t vis = ---------
- (10-190)
v r
in which is the viscous energy factor, v is effective opening displacement rate and v 0 is the reference value of the
effective opening displacement rate. This reference value can either be user-defined or calculated by the program. In
the latter case, the reference value is given by the maximum effective opening displacement rate in any interface
element, as long as the response in all the interface elements is reversible. The viscous energy dissipation model does
not directly have a physical background, but is basically numerical in nature.
In the equations discussed above, no distinction has been made between tensile and compressive loading in the normal
direction. Assuming that in compression the behavior will remain reversible, equation (10-191) will be adapted as:
2 2 2
v = max v n 0 + v s + v t (10-191)
Since, irrespective of the damage level, the interface elements should be able to sustain ongoing loading in
compression (so that inter-penetration is prohibited), it is possible to make the stiffness in compression a function of
the corresponding (negative) opening displacement. By default, the stiffness in compression is constant and given by
Main Index
564 Nonlinear Users Guide
Cohesive Zone Modeling (MCOHE)
the slope of the traction versus opening displacement curve at the origin. If a non-default value of the stiffening factor
in compression F is given, the stiffness at v = v c is given by:
t t n
-------n- = F --------
v n v c v n 0
So far, the constitutive behavior has been discussed in terms of an effective traction versus an effective opening
displacement. The traction components follow from the effective traction according to:
v v
t n = t -------- ; t s ,t = t ----------- (10-192)
v n v s ,t
As an alternative to the above mentioned standard models, the UCOHES user subroutine can be used to enter a
user-defined material behavior. This can be activated by using the MATUSR bulk data entry.
There is a corresponding thermal cohesive model to what was discussed above which uses heat transfer cohesive
elements. In this model, the thermal conductivity can be a function of the displacement opening. The thermal gradient
through the thickness is based upon the temperatures along the top and bottom surfaces.
As an example of the usage of MCHOE for defining a cohesive material, see Composite Fracture and Delamination
(Ch. 20) in the MSC Nastran Demonstration Problems Manual.
Main Index
Chapter 10: Materials 565
Creep (MATVP, CREEP)
Creep Strain
C
Tertiary
Creep
Secondary
Creep
Primary
Creep
Time (t)
Note: Primary Creep: Fast decrease in creep strain rate
Secondary Creep: Slow decrease in creep strain rate
Tertiary Creep: Fast increase in creep strain rate
Figure 10-44 Creep Strain Versus Time (Uniaxial Test at Constant Stress and Temperature)
Main Index
566 Nonlinear Users Guide
Creep (MATVP, CREEP)
c c
where A is a constant; is equivalent creep strain rate; and , , T , and t are equivalent stress, equivalent
creep strain, temperature and time, respectively. The functions f , g , h , and k are piecewise linear. This
representation is shown in Figure 10-45. (Any of the functions ( f , g , h , or k ) can be set to unity by setting the
number of piecewise linear slopes for that relation to zero on the input data.)
F4
F3
S3
Function F (X)
[Such as t , F2 S2
c
g , h (T),
k (t)]
S1
F1
X1 X2 X3 X4
2. The dependence of equivalent creep strain rate on any independent parameter can be given directly in power
law form by the appropriate exponent. The equivalent creep strain rate is
c n
c = A m T p qt q 1 (10-195)
This is often adequate for engineering metals at constant temperature where Nortons rule is a good
approximation.
c = An (10-196)
3. Isotropic creep behavior is based on a von Mises creep potential described by the equivalent creep law
= f c T t (10-197)
Main Index
Chapter 10: Materials 567
Creep (MATVP, CREEP)
c
ij = c ---------- (10-198)
ij
During creep, the creep strain rate usually decreases. This effect is called creep hardening and can be a function of time
or creep strain. The following section discusses the difference between these two types of hardening.
Consider a simple power law that illustrates the difference between time and strain-hardening rules for the calculation
of the creep strain rate.
c
= t n (10-199)
where c is the creep strain, and n are values obtained from experiments and t is time. The creep rate can be
obtained by taking the derivative c with respect to time
c d
c
= -------- = nt n 1
dt
(10-200)
c
c n 1 n
= nt n 1 = n 1 n (10-202)
Equation (10-201) shows that the creep strain rate is a function of time (time hardening). Equation (10-202) indicates
that the creep strain rate is dependent on the creep strain (strain hardening). The creep strain rates calculated from these
two hardening rules generally are different. The selection of a hardening rule in creep analysis must be based on data
obtained from experimental results. Figures 10-46 and 10-47 show time and strain hardening rules in a variable state
of stress. It is assumed that the stress in a structure varies from 1 to 2 to 3 ; depending upon the model chosen,
different creep strain rates are calculated accordingly at points 1, 2, 3, and 4. Obviously, creep strain rates obtained
from the time hardening rule are quite different from those obtained by the strain hardening rule.
Main Index
568 Nonlinear Users Guide
Creep (MATVP, CREEP)
c
1
3
3
1
4
2
0
t
c
1
2
3
3
1 4
2
0
t
Main Index
Chapter 10: Materials 569
Creep (MATVP, CREEP)
It is recommended that you use the implicit formulation described in the following paragraphs to model general
viscoplastic materials.
ee
evp
p vp
= Plastic Element
Inactive if < y
c c n
m P
= A T qt
q1
(10-203)
3 i ij
iij = --- ---------
2
where ij is the deviatoric stress at the end of the increment and y is the back stress. A is a function of temperature,
time, etc. Creep only occurs if sigma is greater y .
One of three tangent matrices may be formed. The first uses an elastic tangent, which requires more iterations, but can
be computationally efficient because re-assembly might not be required. The second uses an algorithmic tangent that
provides the best behavior for small strain power law creep. The third uses a secant (approximate) tangent that gives
the best behavior for general viscoplastic models.
Main Index
570 Nonlinear Users Guide
Creep (MATVP, CREEP)
As an example of the usage of MATVP for defining the creep behavior, see Creep of a Tube (Ch. 51) in the MSC
Nastran Demonstration Problems Manual.
= c s; c1 (10-204)
where c is a material parameter and constant in the constant strain rate test.
m
1 1 p Q R T
c = --- sin h ----- e (10-205)
A
Where
p is the inelastic strain rate,
A is the pre-exponential factor,
Q is the activation energy,
m is the strain rate sensitivity,
is the multiplier of stress,
R is the gas constant, and
T is the absolute temperature,
From equation (10-205), one may obtain that the flow equation as,
Q 1m
p = A exp ------- sin h --- (10-206)
RT s
s is deformation resistance, its evolution is controlled by the following evolution equation as,
d a B d p
----s- = h 0 B ------
- -------- (10-207)
dt B dt
Main Index
Chapter 10: Materials 571
Creep (MATVP, CREEP)
with
S
B = 1 -----
S
1 d p Q n
= S --- -------- exp ------
S
A dt kT
Where
The constants for the Anand Solder model are determined by a steady state creep test followed by use of a nonlinear
fitting method. For the viscoplastic Anand model, total 9 independent material parameters are needed:
A ,Q , ,m ,h 0 ,s ,n ,a ,s 0
Example
Test cases can be found in archives: qa/nastran/tpl/solder/
To activating creep analysis with ANAND creep model, the following are mandatory:
1. In Bulk Data Card NLMOPTS: keyword CREEP and LRGSTRN must be selected with value:
LRGSTRN with value 1 for updated Lagrange formulation, Hypo-elasticity and additive plasticity with mean
normal value.
CREEP with valc3=1, Implicit Creep on; and valc4=0/1/2, for elastic tangent as default (Blank or 0), 1 for
secant tangent, and 2 for radial setting, respectively.
2. In Bulk Data Card MATEP: keyword IMPCREEP is required for ANAND model.
3. In Bulk Data Card NLSTEP: CREEP in GENERAL is required for ANAND creep model.
4. In Bulk Data Card MATVP: material of ANAND creep model are defined in MATVP card with keyword
ANAND in the 3rd field and following 9 fields.
Main Index
572 Nonlinear Users Guide
Creep (MATVP, CREEP)
An example of the mandatory options for using ANAND model is given as follows. MATVP is a secondary material
data that has to refer to a primary material data (MAT1 in this example).
Entry Description
MATVP Specifies viscoplastic or creep material properties to be used for quasi-static analysis in SOL
400.
Main Index
Chapter 10: Materials 573
Creep (MATVP, CREEP)
2. Click Input Properties..., and select Creep from the Constitutive Model pull-down menu and MATVP from the
Creep Data Input pull-down menu.
Creep material models require the following MATVP material data via the Input Properties subform on the Materials
Application form.
Main Index
574 Nonlinear Users Guide
Gasket (MATG)
Gasket (MATG)
Engine gaskets are used to seal the metal parts of the engine to prevent steam or gas from escaping. They are complex
(often multi-layer) components, usually rather thin and typically made of several different materials of varying
thickness. The gaskets are carefully designed to have a specific behavior in the thickness direction. This is to ensure
that the joints remain sealed when the metal parts are loaded by thermal or mechanical loads. The through-thickness
behavior, usually expressed as a relation between the pressure on the gasket and the closure distance of the gasket, is
highly nonlinear, often involves large plastic deformations, and is difficult to capture with a standard material model.
The alternative of modeling the gasket in detail by taking every individual material into account in the finite element
model of the engine is not feasible. It requires a lot of elements which makes the model unacceptably large. Also,
determining the material properties of the individual materials might be cumbersome.
The gasket material model addresses these problems by allowing gaskets to be modeled with only one element through
the thickness, while the experimentally or analytically determined complex pressure-closure relationship in that
direction can be used directly as input for the material model. The material must be used together with 2-D
axisymmetric or 3-D first-order solid composite element types. In that case, these elements consists of one layer and
have only one integration point in the thickness direction of the element.
Constitutive Model
The behavior in the thickness direction, the transverse shear behavior, and the membrane behavior are fully uncoupled
in the gasket material model. In subsequent sections, these three deformation modes are discussed.
2 3
1
1
Midsurface
Integration Point Integration Point Midsurface
Figure 10-49 The Location of the Integration Points and the Local Coordinate Systems in 2- and 3-D
Gasket Elements
Main Index
Chapter 10: Materials 575
Gasket (MATG)
loading path
py1 D
py G
py0 B
unloading path
Gasket Pressure p
A F C
cp0 cp cy0 cp1 cy cy1
Gasket Closure Distance c
The loading and unloading paths of the gasket are usually established experimentally by compressing the gasket,
unloading it again, and repeating this cycle a number of times for increasing pressures. The resulting pressure-closure
data can be used as input for the material model. You must supply the loading path and may specify up to ten unloading
paths. In addition, the initial yield pressure p y 0 must be given. The loading path should consist of both the elastic part
of the loading path and the hardening part, if present. If no unloading paths are supplied or if the yield pressure is not
reached by the loading path, the gasket is assumed to be elastic. In that case, loading and unloading occurs along the
loading path.
The loading and unloading paths must be defined using the TABLES1 bulk data entries and must relate the pressure on
the gasket to the gasket closure. The unloading paths specify the elastic unloading of the gasket at different amounts
of plastic deformation; the closure at zero pressure is taken as the plastic closure on the unloading path. If unloading
Main Index
576 Nonlinear Users Guide
Gasket (MATG)
occurs at an amount of plastic deformation for which no path has been specified, the unloading path is constructed
automatically by linear interpolation between the two nearest user supplied paths. The unloading path, supplied by you,
with the largest amount of plastic deformation is taken as the elastic path at full compression of the gasket.
For example, in Figure 10-50, the loading path is given by the sections AB (elastic part) and BD (hardening part) and
the initial yield pressure is the pressure at point B. The (single) unloading path is curve CDE. The latter is also the
elastic path at full compression of the gasket. The amount of plastic closure on the unloading path is c p 1 . The dashed
curve FG is the unloading path at a certain plastic closure c p that is constructed by interpolation from the elastic part
of the loading path (section AB) and the unloading path CD.
The compressive behavior in the thickness direction is implemented by decomposing the gasket closure rate into an
elastic and a plastic part:
c = c e + c p (10-208)
Of these two parts, only the elastic part contributes to the pressure. The constitutive equation is given by the following
rate equation:
p = D c c e = D c c c p (10-209)
Plastic deformation develops when the pressure p equals the current yield pressure p y . The latter is a function of the
amount of plastic deformation developed so far and is given by the hardening part of the loading path (section BD in
Figure 10-50).
Initial Gap
The thickness of a gasket can vary considerably throughout the sealing region. Since the gasket is modeled with only
one element through the thickness, this can lead to meshing difficulties at the boundaries between thick regions and
thin regions. The initial gap parameter can be used to solve this. The parameter basically shifts the loading and
unloading curves in the positive closure direction. As long as the closure distance of the gasket elements is smaller than
the initial gap, no pressure is built up in the gasket. The sealing region can thus be modeled as a flat sheet of uniform
thickness and the initial gap parameter can be set for those regions where the gasket is actually thinner than the
elements of the finite element mesh used to model it.
Main Index
Chapter 10: Materials 577
Gasket (MATG)
The membrane behavior is defined in the local 1-2 plane (for three-dimensional elements) or the local 3-1 plane (for
two-dimensional elements) and is linear elastic and isotropic. Youngs modulus E m and Poissons ratio m that govern
the membrane behavior are taken from an existing material that must be defined using the MAT1 bulk data entry.
Multiple gasket material can refer to the same isotropic material for their membrane properties.
Thermal Expansion
The thermal expansion of the gasket material is isotropic and the thermal expansion coefficient are taken from the
isotropic material that also describes the membrane behavior.
Constitutive Equations
As mentioned above, the behavior in the thickness direction of the gasket is formulated as a relation between the
pressure p on the gasket and the gasket closure distance c . In order to formulate the constitutive equations of the gasket
material, this relation must first be written in terms of stresses and strains. This depends heavily on the stress and strain
tensor employed in the analysis. For small strain analyses, for example, the engineering stress and strain are used. In
that case, the gasket closure rate and the pressure rate are related to the strain rate and the stress rate by
c = h and p = (10-210)
in which h is the thickness of the gasket.
The resulting constitutive equation for three-dimensional elements, expressed in the local coordinate system of the
integration, now reads
Em m Em
----------------
- ----------------
- 0 0 0 0
2 2
1 m 1 m 11
11
m Em Em 22
22 ----------------
- ----------------
- 0 0 0 0
2 2
33 1 m 1 m p
33 33
= (10-211)
12 0 0 C 0 0 0 12
Em
23 0 0 0 -------------------------- 0 0 23
2 1 + m
31 31
0 0 0 0 Gt 0
0 0 0 0 0 Gt
Main Index
578 Nonlinear Users Guide
Gasket (MATG)
Em m Em
----------------
- 0 ----------------
- 0
11 1 m
2
1 m
2 11
p
22 0 C 0 0 22 22
= (10-212)
33 m Em Em 33
----------------
- 0 ----------------
- 0
2 2
12 1 m 1 m 12
0 0 0 Gt
For large deformations in a total Lagrange formulation, in which the Green-Lagrange strains and the second Piola-
Kirchhoff stresses are employed (as well as in an updated Lagrange environment) in which the logarithmic strains and
Cauchy stresses are being used, similar but more complex relations can be derived.
Entry Description
MATG Specifies gasket material properties to be used in SOL 400.
MATTG Specifies gasket material property temperature variation to be used in SOL 400.
As an example of the usage of MATG for defining gasket material, see Engine Gasket (Ch. 10) or Model Airplane
Engine Analysis (Ch. 28) in the MSC Nastran Demonstration Problems Manual.
Main Index
Chapter 10: Materials 579
Thermo-Mechanical Shape Memory Model (MATSMA)
e q
M s = M s0 + -------- ,
Cm
e q
M f = M f0 + -------- ; and
Cm
e q
A s = A s0 + -------- ,
Ca
e q
A f = A f0 + -------- .
Ca
where e q is the von Mises equivalent stress. At a sufficiently high temperature, often called the M d temperature,
transformation to martensite does not occur at any level of stress.
The transformation characteristics such as the transformation temperatures depend sensitively on alloy composition
and heat treatment.
Main Index
580 Nonlinear Users Guide
Thermo-Mechanical Shape Memory Model (MATSMA)
Mf As
1.0
Austenite to martensite &
0.9 martensite to austentie
decomposition
0.8
Stress = 0
0.7
Martensite Volume Fraction
Note:
0.6 600
After partial
transformation,
0.4
200
0.3
0
0.2 77 150 Ms 200 Af 250 300
Temperature (K)
0.1
MS Af
0.0
0 10 20 30 40 50 60 70 80 90 100
Temperature
Main Index
Chapter 10: Materials 581
Thermo-Mechanical Shape Memory Model (MATSMA)
0 0 0
400
(d) 224K (e) 232K (f) 241K
Tensile Stress (MPa)
300
200
100
0 0 0
(h) 273K (i) 276K
600 (g) 263K
400
200 Ms = 190K
AF = 221K
0 2 4 0 2 4 0 2 4
Strain (%)
In the temperature range where T M f , the microstructures are all martensitic. The stress versus strain curves display
a smooth parabolic type of behavior which is consistent with deformation caused by the movement of defects such as
twin boundaries and the boundaries between variants. Note that unloading occurs nearly elastically and that the
accumulated deformation, caused by the reorientation of the existing martensite and the transformation of any
pre-existing austenite, remains after the specimen is completely unloaded. Note also that the accumulated deformation
is entirely due to oriented martensite and this would be recoverable upon heating to temperatures above the ( A s A f )
range. This would, then, display the shape memory effect.
Pseudoelastic behavior is displayed in the temperature range A f T T c . In this range, the initial microstructures are
essentially all austenitic, and stress induced martensite is formed, along with the associated deformation; upon
unloading, however, the martensite is unstable and reverts to austenite thereby undoing the accumulated deformation.
Note that, as expected, the stress levels rise with increasing temperature. In this range, the transformation induced
deformation is nearly all reversible upon unloading.
At temperatures where T T c , plastic deformation appears to precede the formation of stress induced martensite. The
unloading part of the stress versus strain behavior displays nonlinearity and the unloading is now associated with
permanent (plastic) deformation. Permanent deformation due to plastic deformation of the austenite is nonrecoverable
and if such deformation is large, shape memory behavior is lost.
Main Index
582 Nonlinear Users Guide
Thermo-Mechanical Shape Memory Model (MATSMA)
Constitutive Theory
The model formulated below is based on the kinematics of small strains, although the extension to large strain is
straightforward. Accordingly, the incremental strain, , is simply the sum of the following contributions:
= e l + th + p l + p h (10-213)
In equation (10-213), e l is the incremental elastic, or lattice, strain rate; th is the incremental thermal strain,
p l is the incremental visco-plastic strain, and p h is the incremental strain associated with thermoelastic phase
transformations. The incremental elastic strain is taken to be simply related to a set of elastic moduli, L , and the
incremental Cauchy stress rate, , as
= L e l (10-214)
To calculate the coefficient of thermal expansion of the composite, the rule of mixtures is used as
= 1 f A + f M .
In the above equations, the superscripts A and M refer to the austenite and martensite values, and f is the volume
fraction of martensite.
T 3 '
T RI P = f + g e q eq --- -------- + f + T I + f - P h . (10-215)
2 e q
and
3'
T WI N = fg e q eTq ------------ e q eq ef
g .
f (10-216)
2 e q
1 x+ x
where f = f + + f - and represents McCauleys bracket where x = --- ---------------- , x o .
2 x
part of the transformation, and T is the volumetric part of the transformation strain. The function g eq is
schematically depicted in Figure 10-53, and is a measure of the extent to which the martensite transformation strains
are aligned with the deviatoric stress. e q is the equivalent stress defined as:
Main Index
Chapter 10: Materials 583
Thermo-Mechanical Shape Memory Model (MATSMA)
3 d
e q = --- : d where d is the deviatoric stress.
2
1.1
0.9
0.7
0.5
g
0.3
0.1
-1.0
0.0 0.5 1.0 1.5 2.0
stress/g0
The first two terms in equation (10-215) describe the development of transformation induced strains due to the
formation of stress induced (partially oriented) martensite. f - is the change of formation of austenite; for example,
the rate at which the volume fraction of martensite decreases. The last term in equation (10-215), therefore, represents
the recovery of the accumulated phase transformation strain.
Note that there is no dilatational contribution to T WI N since f is fixed. Note that the twinning strain rate is zero
when e q is less than egff , or when the magnitude of the stress change is negative ( eq 0 ). Hence, egff can be
considered as a stress below which no twinning is possible.
The function g represents the extent to which the transformation strains are coaxial with the applied deviatoric stress.
This function can be calibrated with the experimental data. Note for uniaxial stress-strain curves performed below the
martensite finish temperature, the material starts as 100% martensite, and that other than elastic strains, the
deformation is dominated by the twinning of the randomly oriented martensite.
A functional form that leads to sufficient fit to most experimental data has been implemented in MSC Nastran.
g g g
e q b e q d e q f
g eq = 1 exp g a -------- + g c -------- + g e --------
g0 g0 g0
In most cases, the first term is sufficient, and a value of g a 0 and g b = 2 yields the best results. g 0 is a stress level
used to non-dimensionalize the constants, and can be chosen such that g 1 when the e q g 0 . In some cases, it
is necessary to include the higher powers of equivalent stress for better experimental fits. In these, cases suggested
Main Index
584 Nonlinear Users Guide
Thermo-Mechanical Shape Memory Model (MATSMA)
values for g d = 2.55 or 2.75 and g e = 3 . However, depending on the values of g c and g e , this could lead to
maxima or minima values of g in the range of interest. Note that 0 g 1 and it should be a monotonically
increasing function (an increase in the stress level should lead to an increase in the increment of the phase strains).
Thus, cut off values of g are provided, such that when g reaches its maximum value g = g max at the stress level
g g , it is held constant at the value g
e q = m ax 0 max . For the proper selection of g 0 , see the following section.
The coefficients C M , and C A that provide the stress dependence of the transformation temperatures.
The calibration of the detwinning function, g eq that provides the description of the degree to which
martensite is co-axial with the deviatoric stress state.
The yield stress of the pure martensite and austenite phase ( YM and YA ), and their strain hardening properties.
T and T .
The calibration of the transformation strains, eq v
Main Index
Chapter 10: Materials 585
Thermo-Mechanical Shape Memory Model (MATSMA)
SMA 1)
M s0 : 50C , M f0 : 100C , A s0 : 5C , A f0 : 20C , C M : 6.0Mpa/C , C A : 8.0Mpa/C
SMA 2)
M s0 : 190K , M f0 : 128K , A s0 : 188K , A f0 : 221K , C M : 5.33Mpa/K , C A : 6.25Mpa/K
Stress
CM
CA
M f0 M s0 A s0 A f0
Temperature
Figure 10-54 Typical Stress vs. Temperature Curve Showing the Stress Dependence of Martensite and
Austenite Start and Finish Temperature
Main Index
586 Nonlinear Users Guide
Thermo-Mechanical Shape Memory Model (MATSMA)
Typical pseudoelastic
response
T = T1C
EA
Stress
EM
Strain
Figure 10-55 Typical Stress-strain Curves in the Pseudo-elastic Regime, Depicting the Elastic Moduli
A = 11.0 10 6 C = 6.11 10 6 F
Main Index
Chapter 10: Materials 587
Thermo-Mechanical Shape Memory Model (MATSMA)
g . In addition, it is also found, in practice, that the function g tends to approach unity at a finite equivalent
e q eff
g g
stress level, called 0 . By definition, g 0 = 1 . Also, g a should be chosen to match the general shape of the
function. Since the ratio e q g 0 is less than one, the higher powers take effect later, and thus g c can be added to
lower the middle slope of the curve and g e to fix the final slope of the curve. However, depending on the relative values
of g a , g b and g c , this curve might reach a maximum in the range of interest, and therefore, it should be cut-off at its
g
maximum value g max . The value of g max which is reach at a stress value e q g 0 = m a x are also supplied as
g g
input. Usually g 0 = 2 ef f 10 e ff is a good approximation. But, the selection of g 0 depends on the
experimental measurement.
g0
T WI N = g eq eq
T
Experimental
Model
Stress
g0
Strain
Figure 10-56 Typical Stress-Strain Curve of 100% Martensite Tested Below M f0 Temperature
Others
Yield stresses of the pure martensite and austenite phases for NiTi: YM and YA .
YM = 70-140 Mpa
g0 g g
2 e ff 10 eff
Main Index
588 Nonlinear Users Guide
Thermo-Mechanical Shape Memory Model (MATSMA)
Note: The current model uses a nonsymmetric Jacobian matrix. It is recommended that the nonsymmetric solver
be used to improve convergence.
As an example of the usage of MATSMA for defining the thermal-mechanical model of a shape memory material, see
Shape Memory Analysis of a Stent (Ch. 48) in the MSC Nastran Demonstration Problems Manual.
Main Index
Chapter 10: Materials 589
Mechanical Shape Memory Model (MATSMA)
D
A
At the crystallographic level, if there is no preferred direction for the occurrence of the transformation, the martensite
takes advantage of the existence of different possible habit plates (the contact plane between the austenite and the
martensite during a single-crystal transformation), forming a series of crystallographically equivalent variants. The
product phase is then termed multiple-variant martensite and it is characterized by a twinned structure. However, if
there is a preferred direction for the occurrence of the transformation (often associated with a state of stress), all the
martensite crystals trend to be formed on the most favorable habit plane. The product is then termed single-variant
martensite and is characterized by a detwinned structure. According to the existence of different types of single-variant
martensite species, the conversion of each single-variant martensite into different single variants is possible. Such a
process, known as a reorientation process, can be interpreted as a family of martensite phase transformations and is
associated with changes in the parameters governing the single-variant martensite production (hence, it is often
associated to nonproportional change of stress).
In addition to the thermo-mechanical shape memory model, a superelastic shape memory alloy model is also
implemented in MSC Nastran based on the work of Auricchio [Ref. 7] and [Ref. 8]. This work has been enhanced to
allow different elastic properties for the Austenite and Martensite phases.
The superelastic shape memory model has been implemented in MSC Nastran in the framework of multiplicative
decomposition. We assume the deformation gradient, F as the control variable, and the martensite fraction, S as the
only scalar internal variable. We also introduce a multiplicative decomposition of F in the form:
Main Index
590 Nonlinear Users Guide
Mechanical Shape Memory Model (MATSMA)
F = F e F tr
Assuming an isotropic elastic response, the Kirchhoff stress and the elastic left Cauchy-Green tensor b e , defined
as: b e = F e F e T , share the same principal directions. Accordingly, the following spectral decompositions can be
introduced:
3
= A n A n B
A = 1
3
d
d = A n A n B
A = 1
3
be = Ae 2 n A n B
A = 1
with Ae the elastic principal stretches and d the deviatoric part, according to the relation:
= pI + d (10-217)
where I is the second-order identity tensor, p is the pressure, defined as p = tr 3 , and tr . is the trace
operator. We can write equation (10-217) with the following component form:
d
A = p + A (10-218)
with
d
p = K e , A = 2Ge Ae .
Main Index
Chapter 10: Materials 591
Mechanical Shape Memory Model (MATSMA)
3 12
d 2
d = A .
A = 1
Indicating variants in time with a superposed dot, we assume the following linear forms for the evolution of S :
AS F -
S = H 1 S ------------------- for (A S) (10-220)
F R fA S
F
S = H S A S -------------------- for (S A) (10-221)
F R fS A
where
2 2
R fAS = fA S --- + , R fS A = fS A --- +
3 3
SA
with sA S , fA S , sS A , and fS A material constants. The scalar quantities H AS and H embed the plastic-
transformation activation condition hence, allowing a choice between equations (10-220) and (10-221) and they are
defined by the relations:
H AS = 1 , if R sA S F R fAS , or F 0 . Otherwise, H AS = 0 .
H SA = 1 , if R fS A F R sS A , or F 0 . Otherwise, H SA = 0 .
where
2 2
R sA S = sA S --- + , R sS A = sS A --- + .
3 3
Time-discrete Model
The time-discrete model is obtained by integrating the time-continuous model over the time interval [ t n t ]. In
particular, we use a backward-Euler integration formula for the rate-equations evaluating all the nonrate equations at
time t . Written in residual form and clearing fraction from equations (10-220) and (10-221), the time-discrete
evolutionary equations specialize to:
R A S = F R fA S s H A S 1 S F F n = 0 (10-222)
R S A = F R fS A s H S A S F F n = 0 (10-223)
where
Main Index
592 Nonlinear Users Guide
Mechanical Shape Memory Model (MATSMA)
t
S = dt = S S ,n (10-224)
t
n
The quantity S in equation (10-224) can be computed expressing F as a function of S and requiring the satisfaction
of the discrete equation relative to the corresponding active phase transition.
The detailed solution algorithm for stress update and consistent tangent modulus are given in the work of Auricchio
[Ref. 8].
In the enhanced version of this model, you enter different elastic constants for the two phases. In this case, the effective
elastic moduli are taken as:
A M
E = E 1 S + E S
A M
= 1 S + S
Uniaxial tension test is performed at the same temperature at which the simulation is desired. Here is one
example set for SMA materials.
3. : It is measured from the difference between the response in tension and compression.
Main Index
Chapter 10: Materials 593
Mechanical Shape Memory Model (MATSMA)
Case 1) if the behavior in tension and compression are the same, the value is set to 0.
Case 2) if the behavior in tension and compression have a difference as in the classical case of SMA, the
value is usually set to 0.1 if there is no compression data for the phase transformation. One value
for the phase transformation in compression, say sA S (sigAS_s_compression) is available, is
calculated as follows:
= sqrt 2 3 SA S SA S SA S + SA S
4. L : epsL is a scalar parameter representing the maximum deformation obtainable only by detwinning of the
multiple-variant martensite (or maximum strain obtainable by variant orientation). Classical values for epsL are
in the range 0.005 and 0.10. The default value is 0.07.
Note: The mechanical shape memory model only supports ndi = 3 case (3-D, plane-strain and axisymmetric
elements). It does not support either ndi = 1 or ndi = 2 cases (1-D and plane-stress elements).
CM CM CM
CA CA CA
To To To
Main Index
594 Nonlinear Users Guide
Mechanical Shape Memory Model (MATSMA)
AS 0
f = T o M f C m
0
sS A = T o A s C a
0
fS A = T o A f C a
As an example of the usage of MATSMA for defining the mechanical model of a shape memory material, see Shape
Memory Analysis of a Stent (Ch. 48) in the MSC Nastran Demonstration Problems Manual.
Main Index
Chapter 10: Materials 595
Material Orientation
Material Orientation
For many materials (such as composites, wood, and rolled metals), the material properties are both directionally
dependent and nonhomogenius. A material which is directionally dependent is anisotropic, and a special case of this
is orthotropic. The material behavior is defined through the MAT2, MAT3, MAT8, or MATORT for structural behavior
and MAT5 for thermal behavior.
The material properties defined are by default given with respect to the element coordinate system. This system is often
not conveniend and the definition of the material data is more natrual when defined with respect to the material
preferred system. The material coordinate system can be defined in two ways, depending on the element type.
For shell elements, 2-D planar, or axisymmetric (CTRIA3, CTRIA6, CTRIAR, CTRIAX, CTRIAX6, CQUAD4, CQUAD8,
CQUADR, and CQUADX), the preferred orientation may be given as an angle with respect to an element edge.
yelement
G3
G4
+ xelement
= -------------
2
xmaterial
zelement
THETA
G1 G2
Additionally, when these elements are used as composite shells (PCOMP or PCOMPG), then a ply angle may be used
for each laminate through the thickness of the shell to define the orientation.
For the elements mentioned above and all solid elements such as CTETRA, CHEXA, and CPENTA, the preferred
material orientation may be defined by specifying a coordinate system via CORD1R, CORD2R, CORD1C, CORD2C,
CORD1S, or CORD2S. For shell elements, this coordinate system is projected into the plane of the element as shown
in Figure 10-59.
Main Index
596 Nonlinear Users Guide
Material Orientation
x
MCID
Coordinate
System
z
G2
y G3
ymaterial
xmaterial
G4
G1
Main Index
Chapter 10: Materials 597
Heat Transfer Properties
Thermal Strain
In the additive formulation, the increment of the total strain
= t et me cr
In SOL 400, the thermal strain is based upon the secant approach, such that:
T
t h Tre f T T T r ef
and
T
z
th T = T 2 T 2 T re f T 1 T 1 T r ef
i
if is a constant, then
T2
th T = T 2 T 1
1
F = FFeFp ,
the coefficient of thermal expansion is isotropic.
Main Index
598 Nonlinear Users Guide
User Defined Material Subroutines
It is often useful to define state variables that may be used in the user subroutines. The state variables are allocated for
each layer, each integration point, and for all elements. This is done using the UDSESV bulk data entry.
For most users, one can define the user defined subroutine when executing the NASTRAN command.
Feature Description
Two MSC Nastran command line keywords, uds and udssave, are added in this release. These command line
keywords are used to specify user source file and SCA component build location. In user source file, there may be one
or more predefined user subroutines. These predefined user subroutines will be used to define interfaces in user SCA
service. The build location is where the SCA component will be built.
The command line keyword, uds, is used to specify user source file with predefined user subroutines. It is specified
at MSC Nastran job submittal time.
nast20131 myjob.dat uds=mysource.F
The specified source file will be compiled and implemented. Since only one user source file is allowed, the user must
put all the predefined user subroutines and other related subroutines in one source file. The user source file can be in
FORTRAN or C++.
An alternative method is to incorporate the user subroutine into the model input file. The user subroutine section starts
with a BEGIN UDS and terminates with an ENDDATA or another BEGIN statement. The user can put source code in
this section and set uds command line keyword to model. MSC Nastran extracts the source code in this section by
the following command:
nast20131 myjob.dat uds=model
The user service name in CONNECT SERVICE statement in input file is needed. Only one CONNECT SERVICE
statement is allowed in this release.
Main Index
Chapter 10: Materials 599
User Defined Material Subroutines
Main Index
600 Nonlinear Users Guide
Experimental Data Fitting
Main Index
Chapter 10: Materials 601
Experimental Data Fitting
If for some reason the independent and dependent columns need to be interchanged, you can turn this toggle
on. Check your imported fields before proceeding to ensure they are correct. This is done in the Fields
application.
When you press the Apply button, you will be taken to the second step. If you need to import more than one
file, you will have to reset the Action pull-down.
Main Index
602 Nonlinear Users Guide
Experimental Data Fitting
Main Index
Chapter 10: Materials 603
Experimental Data Fitting
Simple Shear For time independent elastomeric materials (uniaxial, biaxial, planar shear, simple shear
and volumetric tests) the data should be as such in the raw data file:
Volumetric
eng. strain_1, eng. stress_1, volumetric data_1
. . .
If, for uniaxial, biaxial or planar shear data, the third column is left empty, the material is
assumed to be incompressible.
Viscoelastic Turn this toggle on if you wish to do a data fit on viscoelastic relaxation data.
For visco-elastic material behavior (shear relaxation, bulk relaxation and energy
relaxation tests) the data should be as such in the raw data file:
. . .
Main Index
604 Nonlinear Users Guide
Experimental Data Fitting
cycle_1, str_energy_d_1
cycle_2, str_energy_d_2
. .
cycle_n, str_energy_d_n
In addition, before fitting the data, the free energy (which is the strain energy density
corresponding to the undamaged state) is required. Notice that the data points should not
include the range of cycles at which damage did not start to evaluate.
str_energy_d_1, strain_energy_d_1/str_energy_d_1_undamaged
str_energy_d_2, strain_energy_d_2/str_energy_d_2_undamaged
. . .
str_energy_d_n, strain_energy_d_n/str_energy_d_n_undamaged
Notice that the data points should not include the range of cycles at which damage did not
start to evaluate.
Viscoelastic/Damage In this field, select the Viscoelastic or Damage raw test data field.
Select Material Test Data From this list box you select the field corresponding to the Deformation Mode or the
Viscoelastic/Damage data.
Apply Once the test data is associated to the respective modes, the Apply button will take you to
the Calculate Properties action.
Cancel Closes the Experimental Data Fitting tool.
Main Index
Chapter 10: Materials 605
Experimental Data Fitting
Calculate Properties
Once test data has been associated to a test type or mode, the
curve fit is done by following these steps:
1. Select the material Model you wish to do a curve fit for.
The available models will depend on the test data
selected in the previous step. Hyperelastic models will
be available for deformation mode test data.
Viscoelastic models will be available for relaxation test
data.
2. In general, you will leave Use Test Data to All for
hyperelastic models. If however you only want the
curve fit to use one of the deformation modes, you may
set it here.
3. Press the Compute button. The coefficient values will
be displayed in the Coefficients spreadsheet.
4. To visually see the curve fit, press the Plot button.
You may repeat the above four steps for as many
material models as you wish to curve fit.
5. Select an existing material or type in a New Material
Name and press the Apply button to save the material
model as either a Hyperelastic or Viscoelastic
constitutive model for use in a subsequent analysis.
The following notes are made:
The plots are appended to the existing XY Window until you press the Unpost Plot button. You can turn the
Append function on/off under the Plot Parameters... form.
By default, all the deformation modes are plotted along with the raw data even if raw data has not been
supplied for those mode. This is very important. These additional modes are predicted for you. You should
always know your models response to each mode of deformation due to the different types of stress states.
For example, a rule of thumb for natural rubber and some other elastomers is that the tensile tension biaxial
response should be about 1.5 to 2.5 times the uniaxial tension response.
You can turn on/off these additional modes or any of the curves under the Plot Parameters button as well
as change the appearance of plot. More control and formatting of the plot can be done under the XY Plot
application on the Patran application switch on the main form.
Viscoelastic constitutive models are useless without a hyperelastic constitutive model also. Be sure your
model has both defined under the same material name if you use viscoelastic properties.
You may actually change the coefficient values in the Coefficients spread sheet if you wish to see the effect
they have on the curve fit. Select one of the cells with the coefficient you wish to change, then type in a new
coefficient value in the Coefficient Value data box and press the Return or Enter key. Then press the Plot
button again. If you press the Apply button, the new values will be saved in the supplied material name.
Main Index
606 Nonlinear Users Guide
Experimental Data Fitting
For viscoelastic relaxation data, the Number of Terms used in the data fit should, as a rule of thumb, be as
many as there are decades of data.
A number of optional parameters are available to message the data and control the curve fitting. See the
table below for more detailed descriptions.
The following tables more fully describes each widget in the Experimental Data Fitting tool:
Main Index
Chapter 10: Materials 607
Experimental Data Fitting
Main Index
608 Nonlinear Users Guide
Experimental Data Fitting
where lam1, lam2 and lam3 are the principal stretch ratios, J is the determinant of the
deformation gradient, N is the number of terms and mu_n, alpha_n and K are the
material parameters to be determined. The maximum number of terms is 10, but it is
recommended to use no more terms than necessary to get a sufficiently good fit. This
model can be used for incompressible as well as for slightly compressible elastic
materials. Compressibility is included based on a constant bulk modulus. In case of
compressibility, volumetric information is needed, preferably using a volumetric test,
but volumetric data can also be included for uniaxial, biaxial and planar shear tests. In
order to perform a plausible extrapolation for the compressible Ogden model,
dilatational information is needed beyond the data set. This is achieved using linear
extrapolation based on the two start and/or end points of the measured data. This linear
extrapolation may restrict the validity of the response outside the range of the
measured data. For dual mode plotting (except for simple shear), dilatational
information is needed for the compressible Ogden model. For a volumetric test, this
readily follows from the strain, but for uniaxial, biaxial and planar shear tests this must
be calculated. This calculation is based on the requirement that the stress in
perpendicular direction must be zero. If the fitted coefficients do not fulfil this
requirement, zero stresses are returned for such a dual mode.
Main Index
Chapter 10: Materials 609
Experimental Data Fitting
where lam1, lam2 and lam3 are the principal stretch ratios, J is the determinant of the
deformation gradient, N is the number of terms and mu_n, alpha_n and beta_n are the
material parameters to be determined. The maximum number of terms is 10, but it is
recommended to use no more terms than necessary to get a sufficiently good fit. This
model should be used for highly compressible elastic materials. Except for the simple
shear test, volumetric information must be available. In order to perform a plausible
extrapolation for the foam model, dilatational information is needed beyond the data
set. This is achieved using linear extrapolation based on the two start and/or end points
of the measured data. This linear extrapolation may restrict the validity of the response
outside the range of the measured data. For dual mode plotting (except for simple
shear), dilatational information is needed for the foam model. For a volumetric test,
this readily follows from the strain, but for uniaxial, biaxial and planar shear tests this
must be calculated. This calculation is based on the requirement that the stress in
perpendicular direction must be zero. If the fitted coefficients do not fulfil this
requirement, zero stresses are returned for such a dual mode.
Model: Arruda-Boyce Elastomer Free Energy Function; Number of coefficients 1
Ref: "A Three-Dimensional Constitutive Model For the Large Stretch Behavior of
Rubber Elastic Materials" by: Ellen M. Arruda and Mary C. Boyce,
J.Mech.Phys.Solids Vol.41, No.2, pp.389-412
k : Boltzmann constant
T : Temperature
For this model, the calculation of the bulk modulus is not required. A volumetric test
need not be supplied.
Main Index
610 Nonlinear Users Guide
Experimental Data Fitting
Ref: "A new constitutive relation for rubber" by: A.N. Gent, Rubber Chemistry and
Technology, Vol.79, pp.59-61, 1996
N
G(t) = G_infinity W =
{G_n*exp(-t/tau_n)}
n = 1
where t is the time, N is the number of terms and G_infinity, G_n and tau_n are material
parameters to be determined. The data points provided by you must give the value of
the shear modulus at different time stations, which do not need to be equispaced.
Model: Visco Bulk Relax This command is used if experimental bulk relaxation data must be fitted using the
following Prony series expansion for the bulk modulus K:
N
K(t) = K_infinity {K_n*exp(-t/tau_n)}
n = 1
where t is the time, N is the number of terms and K_infinity, K_n and tau_n are material
parameters to be determined. The data points provided by you must give the value of
the bulk modulus at different time stations, which do not need to be equispaced.
Main Index
Chapter 10: Materials 611
Experimental Data Fitting
N
K(t) = K_infinity {delta_n*W0*exp(-t/tau_n)}
n = 1
where t is the time, N is the number of terms, W0 is the instantaneous strain energy and
W_infinity, delta_n and tau_n are material parameters to be determined. The data
points provided by you must give the value of the strain energy at different time
stations, which do not need to be equispaced.
Model: Cont. Damage Not yet supported.
Model: Disc. Damage Not yet supported.
Compute This command starts the data fitting program with the selected data. After fitting, the
measured and fitted curves can be displayed and the corresponding material model
coefficients and the least squares error are reported. For the Mooney-Rivlin, Ogden,
Foam, Arruda-Boyce and Gent models, the response in the modes for which no data is
measured, is predicted. Notice that if volumetric data is relevant in order to predict a
uniaxial, biaxial or planar shear mode, this is calculated using the constraint of a zero
stress component in a direction perpendicular to the direction of the measured stress
component. If this calculation fails, the predicted mode will contain zero stresses.
If the data set contains a large number of entries, or if the model is highly nonlinear
and/or contains many coefficients, then the fitting procedure may take some time.
Since the curve fitting procedure does not use weighting factors per data point, it might
be useful to have many data points near regions where an accurate response is desired.
Coefficient Value This option allows you to manually enter the coefficients of the material model and
updates the response curves. You must select the coefficient to be modified from the
spread sheet and then enter the value by pressing the Return or Enter key after putting
in the new coefficient. Notice that, although no fit is performed, at least one set of test
data must be selected. This feature is available for the Mooney-Rivlin, Ogden, Foam,
Arruda-Boyce and Gent material models.
Plot This plots the current curve fit. The raw data and the calculated curve fits for the
supplied experimental data and the other predicted modes are plotted.
Unpost Plot This clears the plot of all curves. By default additional data fit curves are appended the
XY Window.
New Material Name If a new name is supplied here and this material name does not yet exist, it will be
created with the appropriate constitutive models when the Apply button is pressed.
Main Index
612 Nonlinear Users Guide
Experimental Data Fitting
Main Index
Chapter 10: Materials 613
Experimental Data Fitting
Main Index
614 Nonlinear Users Guide
Experimental Data Fitting
Main Index
Chapter 10: Materials 615
References
References
1. Barlat, F., Lege, D.J. and Brem, J.C., A six-component yield function for anisotropic metals, Int. J. Plasticity,
7, 693-712 (1991).
2. Chung, K. and Shah, K., Finite element simulation of sheet metal forming for planar anisotropic metals, Int.
J. Plasticity, 8, 453-476 (1992).
3. Yoon, J.W., Yang, D.Y. and Chung, K. and Barlat. F., A general elasto-plastic finite element formulation
based on incremental deformation theory for planar anisotropy and its application to sheet metal forming, Int.
J. Plasticity, 15, 35-67 (1999).
4. Yoon, J.W., Barlat, F., Chung, K., Pourboghrat, F. and Yang, D.Y., Earing predictions based on asymmetric
nonquadratic yield function, Int. J. Plasticity, 16, 1075-1104 (2000).
5. Chaboche, J. L., Constitutive Equations for Cyclic Plasticity and Cyclic Viscoplasticity, International
Journal of Plasticity, Vol. 5, pp. 247-302, 1989
6. Camacho, G.T. and Ortiz, M., Computational modelling of impact damage in brittle materials, Int. J. Solids
Struct., Vol. 33, pp 2899-2938, 1996.
7. Auricchio, F. and Taylor, R.L., Shape-memory alloy: modeling and numerical simulations of the finite-strain
superelastic behavior, Comput. Methods Appl. Mech. Engrg., Vol. 143, pp.175-194 (1997).
8. Auricchio, F., A robust integration-algorithm for a finite-strain shape-memory-alloy superelastic model, Int.
J. Plasticity, Vol.17, pp.971-990 (2001).
Main Index
Chapter 11: Element Library
Nonlinear Users Guide
11 Element Library
Introduction 616
Interpolation Functions 617
Element Integration 618
Incompressible Behavior 619
Nonlinear Behavior 620
User Defined Services/Subroutines 621
Property Options 622
Element Classes 623
Automatic Property Mapping 647
Patran Interface 651
Modeling Guidelines 652
References 662
Main Index
616 Nonlinear Users Guide
Introduction
Introduction
Clearly the element is the heart and soul of the FEM. The element is used as a mechanism to integrate a quantity over
a volume, a surface, a curve, or a point. For the simplest simulation, it provides a transfer function (or impedance)
between the degrees of freedom of one grid and the degrees of freedom of another grid. This is a very general definition
which may be applied to any type of physics including structural analysis, thermal analysis, acoustic analysis, and fluid
dynamics among others. For the case of linear structural analysis, this transfer function is called the stiffness matrix.
In most of this chapter, the focus is on the stiffness matrix evaluation of an element. The element definition is used to
obtain other engineering quantities including the mass matrix and the equivalent nodal load.
The selection of the element type and the design of the finite element mesh are important to obtain an accurate solution.
The design of a finite element mesh is done either fully automatically by a mesh generator available from MSC (Patran,
SimXpert) or another mesh generator available on the market. This chapter does not dwell on what is often called the
art of mesh generation. Rather this chapter focuses on the different classes of elements available in MSC Nastran SOL
400 with particular focus on performing nonlinear analysis.
In MSC Nastran, there are two aspects of element definition:
1. The definition of the location of the element, by identifying grip points comprising the element. This is often
called the element topology. This is specified in the C* options discussed below
2. The definition of the characteristics of the element. This in general is provided in the Property options discussed
below.
For nonlinear analysis, the topology of the elements is generally the same for linear analysis (SOL 101) and for
nonlinear analysis (SOL 400), though there are some restrictions as well which are discussed below. Additionally, there
are elements available in the inplicit nonlinear procedure that are not available in the linear solution sequences. For a
detailed description of MSC Nastran element technology, see the MSC Nastran Linear Static Analysis Users Guide,
MSC Nastran Elements (Ch. 4), MSC Nastran Reference Manual, Structural Elements (Ch. 3) and QRG, Bulk Data
Entries.
The definition of the characteristics of the elements is significantly different between linear and nonlinear analysis.
When performing nonlinear simulations, the technology required to represent the topology and interpolation functions
requires alternative formulations from the classical MSC Nastran elements because of either:
1. Material nonlinear behavior
2. Constraints associated with incompressible or nearly incompressible behavior
3. Large deformation, which results in distortion of the element geometry
Because of these conditions, the Classical MSC Nastran elements have been augmented by the element technology in
the Marc product; these elements are often labeled as Advanced elements. The activation of these element technologies
is performed by an associated Property option. Significant effort was made to maintain compatibility between the
different methods.
Notice MSC products (MSC Nastran, Patran, SimXpert and Marc) are designed to work together, but, on occasion, the
technology used to define this technology is inconsistent. When necessary, both terminologies are used.
Main Index
Chapter 11: Element Library 617
Interpolation Functions
Interpolation Functions
Associated with an element are interpolation (shape) functions. These functions describe how the fundamental
quantities such as the coordinate position and displacement (for structural analysis) vary over the element. There are
four types of interpolation functions used in MSC Nastran in either the classic elements or the advanced elements
including:
Linear share functions
Quadratic
Cubic
Assumed
The first three types of shape functions are described in any finite element textbook (Zienkiewicz, Taylor). When using
the first three methods, the shape functions used to describe the coordinate position and the displacements, in which
case, these elements are considered to be isoparametric elements. Such elements are guaranteed to be able to exactly
represent the rigid body modes and the homogeneous modes; a necessary condition for convergence to the exact
solution as the mesh is refined. In such case, the element is said to satisfy the patch test.
The assumed strain shape functions are usually considered to be proprietary and have been tuned to give more accurate
results for a large class of engineering problems. In MSC Nastran, these assumed strain elements are used either to
improve the bending behavior or to satisfy incompressibility requirements.
Main Index
618 Nonlinear Users Guide
Element Integration
Element Integration
Except for a few exceptions, all element are integrated numerically in MSC Nastran, and the default is to use full
integration. Full integration implies that if the displacement vary over the element consistently with the interpolation
function, then the virtual work expression is integrated exactly.
When evaluating integrals, the functions are evaluated at the integration point. The most common integration point
locations and associated weight functions are called
Gauss Points
Barlow Points
Newton-Coates or Lobatto Points
Hourglass Mode
Main Index
Chapter 11: Element Library 619
Incompressible Behavior
Incompressible Behavior
Many materials, such as rubber, are known to be incompressible. Other materials, such as elastic or elastic-plastic
materials, may behave as nearly incompressible when the inelastic strain (plastic or creep) is large relative to the elastic
strain. Such processes may occur at elevated temperatures or during manufacturing.
Lower-order triangular and tetrahedral elements using conventional shape functions and lower-order quadrilateral/
hexahedral elements using conventional integration schemes give incorrect results for plane strain, axisymmetric, and
3-D behavior. This is because the incompressibility acts like a constraint, and the element does not have enough
degrees of freedom to have both good shear behavior and satisfy incompressible behavior.
To overcome this problem for triangular and tetrahedral elements, alternative shape functions may be used. These must
be activated on the Property options.
To overcome this problem for quadrilateral and hexahedral elements, a procedure known as Constant Dilatation of
B-bar is used. This is automatically activated when activating the Large Displacement. For a selection of materials, this
capability is automatically deactivated, including:
Linear elastic with Poissons ratio < 0.45
Mohr Coulomb plasticity
Hypoelastic
Mechanical Shape Memory
Main Index
620 Nonlinear Users Guide
Nonlinear Behavior
Nonlinear Behavior
The advanced elements should be activated in a nonlinear simulation when either large displacement occurs in the
model, large strain, and/or material nonlinearity. In an engineering simulation where either certain regions have
minimal deformation or remain linear elastic, the classical elements may be used. One can always use a mixture of
classical and advanced elements.
Main Index
Chapter 11: Element Library 621
User Defined Services/Subroutines
Main Index
622 Nonlinear Users Guide
Property Options
Property Options
The Property options are used to specify additional element characteristics, geometric properties, and associate
material properties with the element and, in some cases, the material orientation. Material behavior is discussed in
Chapter 10: Materials.
For conventional MSC Nastran solutions, one uses what are called the Primary Property options.
For nonlinear analysis, it is recommended to used the secondary or auxialiary Property options which are discussed
here to activate the advanced elements that fully support geometric and material nonlinearity. This can be done using
the SPROPNAME capability to automatically map the conventional elements to the advanced formulation. See
Automatic Property Mapping section.
Main Index
Chapter 11: Element Library 623
Element Classes
Element Classes
In MSC Nastran, the elements are divided into classes based upon their dimensionality and function. In structural
applications such as automotive body, aerospace, and civil engineering, the use of shells, beams, and rods is prevalent.
In applications such as generators, rockets, and pressure vessels, axisymmetric elements may be used advantageously;
while in automotive engines, housing, etc., 3-D solid elements dominate.
0-D
0-D elements are a single grid; hence, they do not really have a geometry associated with them. Because of this, no
numerical integration is required.
These elements include:
CELAS1, CELAS2, CELAS3, and CELAS4 (if only one grid identified) provides a stiffness matrix
CBUSH (if only one grid identified) provides a stiffness, damping and mass matrix
CDAMP1, CDAMP2, CDAMP3, and CDAMP4 (if only one grid identified) provides a damping matrix
CMASS1, CMASS2, CMASS3, and CMASS4 (if only one grid identified) provides a mass matrix
CONM1 and CONM2 provides a mass/inertia matrix
Note that only CBUSH allows the ability to change the stiffness due to the deformation, and hence, is more powerful
for nonlinear analysis.
The property options used with these elements are
Main Index
624 Nonlinear Users Guide
Element Classes
Note that only CBUSH allows the ability to change the stiffness due to the deformation and, hence, is more powerful
for nonlinear analysis.
The property options used with these elements are:
CROD provides a membrane behavior plus twist, but no bending stiffness. The CROD element is a 2-node
element with linear interpolation. One needs to only define the rod cross-section area.
Main Index
Chapter 11: Element Library 625
Element Classes
CBAR provides membrane, bending and torsion behavior. The CBAR is a 2-grid element that supports
linear or nonlinear material behavior. The type of numerical integration along the length and the
numerical integration across the cross section is determined on the PBARN1.
CBEAM provides general beam behavior. The CEAM is a 2-grid element that will support linear or
nonlinear material behavior. The type of numerical integration along the length and the numerical
integration across the cross section is determined on the PBEMN1.
CBEAM3 provides general beam behavior. The CEAM3 is a 3-grid element with quadratic interpolation
along the length, but only supports linear elastic material.
CBEND is a tube element that may also be used as a curved element. The CBEND is a 2-grid element that
only supports linear elastic material.
Output of Strains
Uniaxial in the truss member.
Output of Stresses
Uniaxial in the truss member.
Main Index
626 Nonlinear Users Guide
Element Classes
Beam section orientation the beam cross section may be either a solid section, an open section, or a hollow
closed section. The definition of the beam section cross section is the same for classical and advanced
elements, but the numberical procedure is different. When using the advanced elements and closed section
cross sections, there is an additional restriction in that the cross section may have only one cavity.
A detailed description of beam cross section and beam orientations is given in the QRG.
Beam cross-section integration this is substantially different for nonlinear analysis when the additional
property option. In the case of linear elastic material behavior, the cross section of the beam is integrated to
obtain the area, moments of inertia, and the torsional moment. When nonlinear material models are present,
the behavior has to be integrated through the cross section.
Beam offsets MSC Nastran has two different methods to apply Beam and Shell offsets:
a. Using rigid elements which is the default.
b. Large rotation method activated by the MDLPRM,OFFDEF,LROFF this is the recommended approach
for a nonlinear analysis.
Beam pin codes this is the same for linear and nonlinear solutions.
Main Index
Chapter 11: Element Library 627
Element Classes
1-D
Number of Interpo- Auxiliary Large Permits
Number Integration lation Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Sect Material Element
CROD 2 1 L ROD L PROD PRODN1* Yes/Yes S Yes 9
CBAR 2 3 LC PROD LC PBAR(L) PBARN1 Yes/No S No 52
CBAR 2 3 LC PROD LC PBAR(L) PBARN1 Yes/No N Yes 52
CBAR 2 1 LC PROD LS PBAR(L) PBARN1 Yes/No S No 98
CBAR 2 1 LC PROD LS PBAR(L) PBARN1 Yes/No N Yes 98
CBEAM 2 3 LC BEAM LC PBEAM(L) PBEMB1 Yes/No S No 52
CBEAM 2 3 LC BEAM LC PBEAM(L) PBEMB1 Yes/No N Yes 52
CBEAM 2 1 LC BEAM LS PBEAM(L) PBEMB1 Yes/No S No 98
CBEAM 2 1 LC BEAM LS PBEAM(L) PBEMB1 Yes/No N Yes 98
CBEAM 2 2 LC BEAM LCC PBEAML PBEMB1 Yes/No N Yes 78
CBEAM 2 2 LC BEAM LCO PBEAML PBEMB1 Yes/No N Yes 79
* When a PRODN1 is used with a CROD to permit nonlinear material behavior, the element behavior changes shuch that it no longer supports torsion.
These elements cannot be used with the Hill, Barlat, Linear Mohr-Coulomb, Parabolic Mohr-Coulomb, or the
IMPLICIT CREEP model specified on the MATEP option.
The user can change the Integration along the length when using PBARN1 or PBEMN1. The choices are:
Linear/cubic means that linear interpolation of the displacement is used along the axis and cubic displacement variation
normal to the beam axis. This results in linear variation of curvature.
The CBAR, PBAR, and PBARN1 does not support LCO which requires element warping.
The cross-section behavior can either be specified by entering an S or N on the SECT option, where N means
numerically integrated and S means smeared. When a nonlinear material behavior needs to be captured, you should
use the N option.
Main Index
628 Nonlinear Users Guide
Element Classes
A*L = A 0 L 0 L
The resultant quantities are always given with respect to an element axis attached to the 1-D element. When large
displacement is included, this is co-rotated with the element.
3 3
6 5
1 2
1 2 1 4 2
4 3
4 3
3 4
1 2
1 2 1 2
4-node quadrilateral, 4-node,
4-point integration point 1-point integration point
Main Index
Chapter 11: Element Library 629
Element Classes
4 7 3 4 7 3
7 8 9
3 4
8 4 5 6 6 8 6
1 2
1 2 3
1 5 2 1 5 2
8-node quadrilateral, 8-node,
9-point integration point 4-point integration point
Plane Stress
Number of Interpola- Auxiliary Large Permits Equivalent
Number Integration tion Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material* Element
CTRIA3 3 1 L PSTRS L PLPLANE PSHLN2 Yes/Yes Yes 201
CQUAD4 4 4 L PSTRS L PLPLANE PSHLN2 Yes/Yes Yes 3
CQUAD4 4 1 L PSTRS LRIH PLPLANE PSHLN2 Yes/Yes Yes 114
CTRIA6 6 7 Q PSTRS Q PLPLANE PSHLN2 Yes/Yes Yes 124
CQUAD8 8 9 Q PSTRS Q PLPLANE PSHLN2 Yes/Yes Yes 26
CQUAD8 8 4 Q PSTRS QRI PLPLANE PSHLN2 Yes/Yes Yes 53
* With exceptions
Main Index
630 Nonlinear Users Guide
Element Classes
Output of Strains
1 = xx
2 = yy
3= zz
4 = xy
Output of Stresses
1 = xx
2 = yy
3= zz = 0
4 = xy
These elements cannot be used with the IMPLICIT CREEP model specified on the MATEP option.
Plane Strain
Number of Interpolat Auxiliary Large Permits Equivalent
Number Integration ion Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material Element
CTRIA3 3 1 L PSTRN L PLPLANE PSHLN2 Yes/Yes Yes 6
CTRIA3 4 3 L & Cubic IPS L PLPLANE PSHLN2 Yes/Yes Yes 155
Bubble
CQUAD4 4 4 L PSTRN L PLPLANE PSHLN2 Yes/Yes Yes 11
CQUAD4 4 1 L PSTRN LRIH PLPLANE PSHLN2 Yes/Yes Yes 115
CTRIA6 6 7 Q PSTRN Q PLPLANE PSHLN2 Yes/Yes Yes 125
CQUAD8 8 9 Q PSTRN Q PLPLANE PSHLN2 Yes/Yes Yes 27
CQUAD8 8 4 Q PSTRN QRI PLPLANE PSHLN2 Yes/Yes Yes 54
Output of Strains
1 = xx
2 = yy
3= zz = 0
4 = xy
Main Index
Chapter 11: Element Library 631
Element Classes
Output of Stresses
1 = xx
2 = yy
3= zz
4 = xy
Caution: The conventional CTRIA element with BEH=PSTRN is known to give very poor results when used with
incompressible or nearly incompressible behavior (including rubber materials, elastic-plastic or creep), the
BEH=IPS should be used.
Axisymmetric
Number of Interpo- Auxiliary Large Permits Equivalent
Number Integration lation Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material Element
CTRIA3 3 1 L AXISOLID L PLPLANE PSHLN2 Yes/Yes Yes 2
CTRIA3 4 4 L & Cubic IAX L PLPLANE PSHLN2 Yes/Yes Yes 156
Bubble
CQUAD4 4 4 L AXISOLID L PLPLANE PSHLN2 Yes/Yes Yes 10
CQUAD4 4 1 L AXISOLID LRIH PLPLANE PSHLN2 Yes/Yes Yes 116
CTRIA6 6 7 Q AXISOLID Q PLPLANE PSHLN2 Yes/Yes Yes 126
CQUAD8 8 9 Q AXISOLID Q PLPLANE PSHLN2 Yes/Yes Yes 28
CQUAD8 8 4 Q AXISOLID QRI PLPLANE PSHLN2 Yes/Yes Yes 55
Output of Strains
1 = rr
2 = zz
3=
4 = rz
Output of Stresses
1 = rr
2 = zz
3 =
4 = rz
To apply a distributed load on this element one needs to apply the pressure using PLOADX1. Note that the distributed
loads are integrated over one radian. Hence, any point force applied should also be prescribed over one radian.
Main Index
632 Nonlinear Users Guide
Element Classes
Caution: The conventional CTRIA element with BEH=AXISOLID is known to give very poor results when used with
incompressible or nearly incompressible behavior (including rubber materials, elastic-plastic or creep), the
BEH=IAX should be used.
Output of Strains
1 = rr
2 = zz
3=
4 = rz
5 = z
6 = z
Output of Stresses
1 = rr
2 = zz
3 =
4 = rz
5 = z
6 = z
Note the displacements for this element are U r , U z , and U which represents the angular displacement about the
symmetry axis measured in radians. This is given as the third degree of freedom, and all torques or single points
constraints on this quantity should be treated as the third degree of freedom.
Main Index
Chapter 11: Element Library 633
Element Classes
Axisymmetric Shell
Number of Interpo- Auxiliary Large Permits Equivalent
Number Integration lation Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material Element
CAXISYM 2 2 L AXISOLID L PLPLANE PAXISYM Yes/Yes Yes 1
CAXISYM 3 1 Q AXISOLID Q PLPLANE PAXISYM Yes/Yes Yes 89
Output of Strains
1 = s = meridional membrane
2 = = circumferential membrane
3 = t = transverse shear strain
Output Of Stresses
1 = s = meridional stress
2 = = circumferential stress
3 = t = transverse shear stress
Ur , Uz , Ut
Main Index
634 Nonlinear Users Guide
Element Classes
plates, and shells are done through the same topological classes. These element satisfy plane-stress conditions as the
normal stresses through the thickness are zero.
These elements exhibit signigicant differences between the classic and the advanced formulation. MSC Nastran has
multiple flavors of these elements:
Homogeneous linear behavior for combined membrane, bending and transverse shear.
Homogeneous linear behavior for pure membrane or bending or transverse shear.
Non-homogeneous linear behavior for combined membrane, bending and transverse shear.
Homogeneous nonlinear behavior for combined membrane, bending and transverse shear.
Homogeneous nonlinear behavior for membranes.
Layered (composite) behavior where material behavior is linear.
Layered (composite) behavior where material behavior is nonlinear.
Here, the word Homogeneous refers to througn the thickness direction. One can use user subroutines to vary the
material behavior over the surface of a membrane or shell element.
Output of Strains
1 = xx
2 = yy
3= xy
Main Index
Chapter 11: Element Library 635
Element Classes
Output of Stresses
1 = xx
2 = yy
3 = xy
Shells
There are two classes of advanced shell elements in MSC Nastran SOL 400.
The first group (marked LDK) is thin shell elements based upon Kirchhoff theory. The second group (marked L, LRIH,
and QRI) is thick shell elements that support transverse shears based upon Mindlin theory. There are the preferred
elements for composite simulation.
3-D Shell
Number of Interpo- Auxiliary Large Permits Equivalent
Number Integration lation Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material* Element
CTRIA3 3 1 L DCTN LDK PSHELL or PSHLN1 Yes/Yes Yes 138
PCOMP or
PCOMPG
CQUAD4 4 4 L DCT L PSHELL or PSHLN1 Yes/Yes Yes 75
PCOMP or
PCOMPG
CQUAD4 4 1 L DCT LRIH PSHELL or PSHLN1 Yes/Yes Yes 140
PCOMP or
PCOMPG
CQUAD4 4 4 L DCTN LDK PSHELL or PSHLN1 Yes/Yes Yes 139
PCOMP or
PCOMPG
CQUAD8 8 4 Q DCT QRI PSHELL or PSHLN1 Yes/Yes Yes 22
PCOMP or
PCOMPG
* With exceptions
The output of strains and stresses are given with respect to a coordinate system attached to the element. This coordinate
system is updated with the deformation if the LGDISP parameter is activated.
Main Index
636 Nonlinear Users Guide
Element Classes
Output of Strains
1 = xx
2 = yy
3 = zz
4 = xy
5 = yz for thick shell only
6 = zx for thick shell only
Output of Stresses
1 = xx
2 = yy
3 = zz = 0
4 = xy
5 = yz for thick shell only
6 = zx for thick shell only
These elements cannot be used with the IMPLICIT CREEP model specified on the MATEP option.
Shear Panel
MSC Nastran supports a 4-node shear panel element for linear analysis. Here, linear analysis means small deformation
and linear isotropic elastic material.
While SOL 400 has an alternative formulation when PSHEARN is entered, it provides a membrane formulation to the
element. It is not recommended that one uses the PSHEARN option in SOL 400.
Main Index
Chapter 11: Element Library 637
Element Classes
The element system has a local coordinate system which is updated with large displacements. In versions prior to the
MSC Nastran 2014 release, the results were given with respect to the basic coordinate system. Currently, the results
are given with respect to the local system.
There is a single integration point per layer, if one specifies only a single layer than the element is under-integrated and
hourglass modes will occur.
8
First Layer (5-6-7-8 face)
Non Composite
Side View
Composite
Side View
5
7
1 1
2
6 3 4
2
4 5
Bending
3 6 7 Strain
Continuous
1 8
4
9 10
3 11
5 12
Integration Point
Last Layer (1-2-3-4 face) 2
Figure 11-1 Connectivity and Integration Points for Composite and Non Composite Materials
Output of Strains
1 = xx
2 = yy
3 = zz
4 = xy
5 = yz
6 = zx
Output of Stresses
1 = xx
2 = yy
3 = zz
4 = xy
5 = yz
6 = zx
Main Index
638 Nonlinear Users Guide
Element Classes
For a layered composite material, use PCOMPLS. Note in the previous figure that for each layer, there are three layer
points, which allows an accurate calculation of the interlaminar shear.
Output of Strains
1 = xx
2 = yy
3 = zz
4 = xy
5 = yz
6 = zx
Output of Stresses
1 = xx
2 = yy
3 = zz
4 = xy
5 = yz
6 = zx
Main Index
Chapter 11: Element Library 639
Element Classes
4 4
10
3
8
7 9
6
1 1
3 2 2
X
4
3
X
3
X
2
X
1
2
4-point integration
6 G6
+6
+4 G15 G14
5 G12
+5
G13
G4 G5
+3
+1 3
G10 G3
2 G11
1 + G9 G8
G7
G1
2 G2
6-node pentahedral, 15-node pentahedral,
6-point integration 21-point integration
Main Index
640 Nonlinear Users Guide
Element Classes
8 6
5
5
7
5 7
8 2
6 6 7
1 1
3 4
1 8
1 4
2
3 3
2 4
G18
G7
6 G6
13
5
14
G19
16 G15
15 7 G17
G14
8 18 G20
17 G8
G2
G10 G5
19 G16 G3
20 2
9 G11 G13 G9
1
10
12 G4 G12
3
11
4 G1
20-node hexahedral, 20-node hexahedral,
27-point integration 8-point integration
Main Index
Chapter 11: Element Library 641
Element Classes
3-D Solid
Number of Interpolat Auxiliary Large Permits Equivalent
Number Integration ion Conventional Property for Rotation/ Nonlinear Marc
Element of Grids Points Scheme BEH INT Property Nonlinear Large Strain Material Element
CTETRA 4 1 L SOLID L PSOLID PSLDN1 Yes/Yes Yes 134
CTETRA 4 4 L& ISOL L PSOLID PSLDN1 Yes/Yes MATEP 157
CUBIC
CPENTA 6 6 L SOLID L PSOLID PSLDN1 Yes/Yes Yes 136
CHEXA 8 8 L SOLID L PSOLID PSLDN1 Yes/Yes Yes 7
CHEXA 8 1 L SOLID LRIH PSOLID PSLDN1 Yes/Yes Yes 117
CTETRA 10 9 Q SOLID Q PSOLID PSLDN1 Yes/Yes Yes 127
CTETRA 10 4 Q SOLID LRIH PSOLID PSLDN1 Yes/Yes Yes 184
CPENTA 15 21 Q SOLID Q PSOLID PSLDN1 Yes/Yes Yes 202
CHEXA 20 27 Q SOLID Q PSOLID PSLDN1 Yes/Yes Yes 21
CHEXA 20 8 Q SOLID QRI PSOLID PSLDN1 Yes/Yes Yes 57
Caution: The conventional CTETRA element with BEH=SOLID is known to give very poor results when used with
incompressible or nearly incompressible behavior (including rubber materials, elastic-plastic or creep), the
BEH=ISO should be used.
8 8 8 Y 7
7 7
5 6 5 6 5 6
ply 3
z
ply 2
ply 3
ply 1
ply 1
ply 3
ply 2
ply 2
ply 1 3 3 3
1 2 1 x 2 1 2
Main Index
642 Nonlinear Users Guide
Element Classes
Large
Auxiliary Rotation/ Permits
Number Integration Conventional Property for Large Nonlinear Marc
Element of Grids INT Property Nonlinear BEH Strain Material Element
Plane Strain
CQUAD4 4 L 2 per layer PLPLANE PLCOMP COMPS Yes/Yes Yes 151
CQUAD8 8 Q 2 per layer PLPLANE PLCOMP COMPS Yes/Yes Yes 153
Axisymmetric
CQUAD4 4 L 2 per layer PLPLANE PLCOMP AXCOMP Yes/Yes Yes 152
CQUAD8 8 Q 2 per layer PLPLANE PLCOMP AXCOMP Yes/Yes Yes 154
Hexahedral
CHEXA 8 L 4 per layer PCOMPLS SLCOMP Yes/Yes Yes 149
CHEXA 20 Q 4 per layer PCOMPLS SLCOMP Yes/Yes Yes 150
Gasket Elements
The lower-order continuum composite elements are also used to model gasket materials. When used in this manner,
the number of layers is one and the material is defined through the MATG option discussed in Chapter 10: Materials.
Note that these elements may be collapsed to a pentahedral so they can model the gasket between two regions modeled
with tetrahedral elements. Care should be exercised to make sure the collasping occurs in the plane of the gasket
material. It is not necessary that the elements match the mesh in the surrounding material. One may utilize the contact
capability to overcome the mesh incompatibility.
Large
Auxiliary Rotation/ Permits
Number Integration Conventional Property for Large Nonlinear Marc
Element of Grids INT Property Nonlinear BEH Strain Material Element
Plane Strain
CQUAD4 4 L 2 per layer PLPLANE PLCOMP COMPS Yes/No MATG 151
Axisymmetric
CQUAD4 4 L 2 per layer PLPLANE PLCOMP AXCOMP Yes/No MATG 152
Hexahedral
CHEXA 8 L 4 per layer PSOLID PSLDN1 SLCOMP Yes/No MATG 149
Interface Elements
There are a series of elements that are used to model the onset and progression of delamination of the bonding materials
using the Cohesive Zone Method. These elements are available for plane strain, axisymmetric, and three-dimensional
behavior. From a meshing perspective, these elements are unique because one can enter a zero thickness. The interface
elements provide two integration schemes. The first uses the conventional Gaussian integration scheme while the other
Main Index
Chapter 11: Element Library 643
Element Classes
uses a nodal lumping scheme (Lobatto-Cotes). The latter scheme may be advantageous when the interface material is
relatively stiff compared to the surrounding material.
Note that the orientation of the element dictates the direction of the interface/delamination. It is not necessary that the
elements match the mesh in the surrounding material. One may utilize the contact capability to overcome the mesh
incompatibility. The material properties are defined using the MCOHE material model which is described in Chapter
10: Materials, Cohesive Zone Modeling (MCOHE). Note that delamination simulations are highly nonlinear and one
must exercise caution in applying the boundary conditions.
The element is written with respect to a local coordinate system, relating the relative displacement and the normal and
shear traction. The output is given with respect to the local coordinate system. This system is updated (rotated) when
large displacement is used. The element does not include a mass matrixor a geometric or initial stress stiffness matrix.
It also does not support application of distribute loads.
The higher-order elements are not fully quadratic; they are quadratic in the plane of the interface but linear through the
thickness.
The elements are shown below.
v 3
1
2
1 3
6
2
7
v 5
2
2
4 4
8
1 1
4-node linear planar interface element 8-node quadratice planar interface element
4 3
1 1 2 2
4 7 3
4 3 8 6
1 2 1 5 2
1 2 Element mid-line
Gaussian integration scheme
Nodal lumping scheme
1 2 3 1 2 3
Gauss Integration Scheme Newton-Cotes/Lobatto Integration Scheme
Main Index
644 Nonlinear Users Guide
Element Classes
8 8
16
4 4 19
20
v v1 11
1 v 7 12 7
3 v3 15
3 3
5 v 18
2 5
v2
13 17
1 1 10
6 9
6
14
2 2
5 3
1
6
2
Element Mid-plane
3
3
4 4
1
2
1
Gauss Integration Scheme
2
Newton-Cotes/Lobatto Integration Scheme
Main Index
Chapter 11: Element Library 645
Element Classes
8
4
7
5 3
1
6
2
Element Mid-plane
7
7 8
8
4 9
5 4 9
1 6
2 5
3
1 6
Gauss Integration Scheme 2
3
Newton-Cotes/Lobatto Integration Scheme
Large Permits
Number of Integration Property for Rotation/ Nonlinear Marc
Element Grids INT Nonlinear Large Strain Material Element
Plane Strain
CIFQUAD 4 2 PCOHE Yes/No MCHOE 186
L
CIFQUAD 8 2 PCOHE Yes/No MCHOE 187
Q/L
Axisymmetric
CIFQUAD 4 2 PCOHE Yes/No MCHOE 190
L
CIFQUAD 8 2 PCOHE Yes/No MCHOE 191
Q/L
Solid
CIFPENT 8 3 PCOHE Yes/No MCHOE 192
L
CIFPENT 15 6 PCOHE Yes/No MCHOE 193
Q/L
CIFHEX 8 4 PCOHE No MCHOE 188
L
CIFHEX 20 8 PCOHE No MCHOE 189
Q/L
Main Index
646 Nonlinear Users Guide
Element Classes
Output of Strain
The three strain components are given at the element integration points. They are determined by the relative
displacements between the top and bottom face and are given in the local element system:
1 = u top u b o t t o m
2 = v top v b o t t o m
3 = w top w b o t t o m for 3-D element only
Output of Stress
1 = n
2 = s1
3 = s2 for 3-D element only
Main Index
Chapter 11: Element Library 647
Automatic Property Mapping
Unsupported
Secondary Primary Features of
Property Property Dimension Material Notes Primary Entry
PBARN1 PBARL 1-D MAT4 Note 4
MATS1 Note 1
MATEP
MATF
MATSMA Note 2
MATVE
MATVP Note 3
PBEMN1 PBEAML 1-D MAT4 Note 4 Tapered Sections
MATS1 Note 1
MATEP
MATF
MATSMA Note 2
MATVE
MATVP Note 3
PRODN1 PROD 1-D MAT4 Note 4 J,C
MATS1 Note 1
MATEP
MATF
MATSMA Note 2
MATVE
MATVP Note 3
Main Index
648 Nonlinear Users Guide
Automatic Property Mapping
Unsupported
Secondary Primary Features of
Property Property Dimension Material Notes Primary Entry
PSHEARN PSHEAR 3-D MAT4 Note 4 F1,F2
MAT8
MATS1 Note 1
MATS8 Note 1
MATEP
MATF
MATORT
MATSMA Note 2
MATVE
MATVP Note 3
PSHLN1 PSHELL 3-D MAT4 Note 4 TS/T, nondefault Z1
MAT5 Note 4 and Z2, 121/T3
MATS1 Note 1
MATS8 Note 1
MATEP
MATF
MATORT
MATSMA Note 2
MATVE
MATVP Note 3
PSHLN1 PCOMP/ 3-D MAT4 Note 4 FT, GE, LAM
PCOMPG options other than
MAT5 Note 4 BLANK and SYM,
MAT8 SOUTi
MATS1 Note 1
MATS2 Note 1
MATS8 Note 1
MATEP
MATF
MATORT
MATSMA Note 2
MATVE
MATVP Note 3
Main Index
Chapter 11: Element Library 649
Automatic Property Mapping
Unsupported
Secondary Primary Features of
Property Property Dimension Material Notes Primary Entry
PSHLN2 PLPLANE 2-D MAT4 Note 4
MAT5 Note 4
MATG Note 5
MATS1 Note 1
MATS3 Note 1
MATS8 Note 1
MATEP
MATF
MATORT
MATSMA Note 6
MATVE
MATVP Note 7
PLCOMP PLPLANE 2-D MAT4 Note 4
MAT5 Note 4
MATEP
MATF
MATORT
MATSMA
MATVE
MATVP
PSLDN1 PSOLID 3-D MAT4 Note 4 IN, ISOP, FCTN
MAT5 Note 4
MATS1 Note 1
MATEP
MATF
MATORT
MATSMA
MATVE
MATVP
Notes: 1. Only when NLMOPTS, LRGSTRN, 1 is flagged or HGENPLAS is provided NLSTEP entry for
coupled analysis.
2. For shape memory materials, define through MATSMA only the thermo-mechanical
model is available.
Main Index
650 Nonlinear Users Guide
Automatic Property Mapping
3. For creep, define through MATVP, VALC=0 must be set on NLMOPTS for explicit
formulation.
4. Only when phase changes are flagged.
5. Gasket materials for BEH4=COMPS or AXCOMP,INT4=L.
6. For shape memory materials, define through MATSMA and BEH=PLSTRS only the
thermo-mechanical model is available.
7. For creep, define through MATVP and BEH=PLSTRS VALC=0 must be set on
NLMOPTS for explicit formulation.
Main Index
Chapter 11: Element Library 651
Patran Interface
Patran Interface
Main Index
652 Nonlinear Users Guide
Modeling Guidelines
Modeling Guidelines
Finite element modeling in many ways is more like an art than a science since the quality of the results is dependent
upon the quality of your model. One of the more common errors that a beginning finite element analyst makes in
modeling is to simply simulate the geometry rather than to simulate both the geometry and the physical behavior of
the real structure. The following modeling guidelines are provided to put a little more science back into the art of finite
element modeling:
Choosing the right element
Mesh density
The above guidelines are by no means complete; however, they do serve as a good starting point. There is no better
substitute for good modeling than experience. It is also good modeling practice to simulate and validate a new
capability or a feature that you have not used before with a small prototype model before applying this feature to your
production model. Model verification techniques are covered in Model Verification (Ch. 10) in the MSC Nastran Linear
Static Analysis Users Guide.
Main Index
Chapter 11: Element Library 653
Modeling Guidelines
General Guidelines
Always experiment with a small test model when using elements that you are not familiar with. This practice is
cheaper than experimenting with a large production model, and it gives you a better understanding of an elements
capabilities and limitations prior to applying it to a large production model.
Zero-Dimensional Elements
When you use CELASi elements to represent concentrated springs between two components of translation, the
directions of the two components must be coaxial. Even small deviations in direction can induce a significant moment
to your model that does not exist in your physical structure. It is recommended that when a CELASi element is used,
the locations of the two end points be coincident in order to avoid this type of problem. If the two end points are not
coincident, you should consider using a CROD or CBUSH element instead.
One-Dimensional Elements
If only an axial and/or torsion load is to be transmitted in an element, then the CROD is the easiest element to use.
A CBAR is easier to use than a CBEAM element. The I1 and/or I2 values can be set to zero.
Use the CBEAM element instead of the CBAR element if any of the following features is important:
The cross-sectional properties are tapered.
The neutral axis and shear center do not coincide.
The effect of cross-sectional warping on torsional stiffness is critical.
The difference in the mass center of gravity and the shear center is significant.
The formulation for the CBEAM element is based on a flexibility approach; the element stiffness matrix is generated
by inverting the flexibility matrix. For this reason, I1 and I2 must not be zero for the CBEAM element.
CBEAM3 elements favor a structure with initial curvatures and with high order shell elements.
Two-Dimensional Elements
In general, quadrilateral elements (CQUAD4 and CQUAD8) are preferred over the triangular elements (CTRIA3 and
CTRIA6). The CTRIA3 element is a constant strain element. It is excessively stiff, and when used alone, it is generally
less accurate than the CQUAD4 element, particularly for membrane strain. Whenever feasible, the CQUAD4 element
should be used instead of the CTRIA3 element. CTRIA3 should only be used when necessary for geometric or
topological reasons, for example, mesh transition between regions of quadrilateral elements with different meshes or
near the polar axis of a spherical shell.
Avoid using CTRIA3 in locations where the membrane stresses are changing rapidly, for example, in the web of an
I-beam. Since CTRIA3 has constant membrane stresses, a large number of them may be needed to obtain acceptable
accuracy. It is better to use quadrilateral elements or CTRIA6 elements, if possible.
Do not use plate or shell elements (CQUADi, CTRIAi) in stiffened shell structures with very thin panels that can buckle.
Shear panels (CSHEAR) should be used in this case or in any situation where direct stresses cannot be supported, such
as in a very thin curved panel.
Main Index
654 Nonlinear Users Guide
Modeling Guidelines
Avoid highly skewed elements (see Figure 11-2). The angle should be as close to 90 as possible.
For the CTRIA3 element, the skew test is based upon the three vertex angles.
Aspect ratio is defined as l (length/width). Very high aspect ratio (see Figure 11-3) should also be avoided, although
it is no longer true that accuracy degrades rapidly with aspect ratios as it once did with some of the obsolete elements.
Warping is a measure of the amount the element deviates from being planar (see Figure 11-4). Element warping should
be minimized.
Element Mid-Plane
For the CQUAD8 elements, if midside nodes are present, they should be located within the middle third of the edge. If
a midside node is located at one-fourth the distance of the edge as measured from either corner node on the edge, the
internal strain field becomes singular at the corners of the element. For best results, it is recommended that the midside
node be located as close to the center of the edge as possible.
Main Index
Chapter 11: Element Library 655
Modeling Guidelines
For single curved structures (e.g., a cylinder), the CQUAD8, in general, yields better results than the CQUAD4 element.
For doubly curved structures (e.g., a spherical dome), the CQUAD4 element, in general, performs better than the
CQUAD8 (see [Ref. 1]).
The shell normal (param,snorm,x) should be turned on when using the CQUAD4 or CTRIA3, element. See Shell
Normals for further details.
Shell Normals
By default, the direction of the normal rotation vector for flat plate elements is assumed to be perpendicular to the plane
of each element. If the model is curved, the shell bending and twist moments must change direction at the element
intersection. If transverse shear flexibility is present, the deformations may be too large. (Because elements using
low-order formulations ignore the edge effect, this rarely causes any problems the default value of 100 on the
parameter K6ROT partially cures the problem.) With the unique normal (SNORM) option, the rotational degrees of
freedom at each corner of an element are measured relative to the specified normal vector direction. Thus, all elements
connected to a grid point will use a consistent direction for defining shell bending and twisting moments.
In CQUAD4 and CTRIA3 elements, the stiffness matrices of the elements are modified to eliminate the undesirable
small stiffness in the rotational motions about the shell normal vector. In effect, the transformation replaces the normal
moments with in-plane forces. No changes were made to the basic element stiffness matrix, and therefore flat plate
models will not be affected. The objective of the new transformation was to remove a potential weakness in curved
shell models and allow the automatic constraint process to remove the true singularity in the assembled stiffness
matrix.
Shell normals are available for CQUAD4 and CTRIA3 elements. Normals are activated if the actual angle between the
local element normal and the unique grid point normal is less than 20, the default value for (see Figure 11-6.) The
default for can be changed by setting PARAM,SNORM, to the desired real value up to 89 degrees. The unique grid
point normal is the average of all local shell element normals at a specific grid point. Generated grid point normals
may be overwritten by user-defined normals.
Main Index
656 Nonlinear Users Guide
Modeling Guidelines
Grid Point
Normal
Shell 1 Shell 2
A shell normal defines a unique direction for the rotational degrees of freedom of all adjacent elements. A shell normal
vector is created by averaging the normal vectors of the attached elements. If the actual angle is greater than the value
defined on PARAM,SNORM, (default = 20.) the edge is assumed to be a corner, and the old method is used. Shell
normals improve the accuracy of the results in curved shells where in-plane shear and twisting moments act together.
If the actual shell normal vector is known, such as in a cylinder or sphere, the automatically calculated values may be
superseded by the actual vectors using the bulk data entry
SNORM,GID,CID,V1,V2,V3
where GID is a unique grid point, CID is the coordinate system for defining the shell normal vector and V1, V2, and
V3 are unscaled components of the vector in the coordinate system.
A second parameter, SNORMPRT, controls the print or punch of values of the internal shell normals. The output format
is the same as the SNORM bulk data so that the individual values may be used and modified on a subsequent job.
Main Index
Chapter 11: Element Library 657
Modeling Guidelines
4. The CQUAD8 and CTRIA6 elements are not included in the shell normal processing. If they are modeled
correctly, they do not require shell normal processing. Connecting these elements to the lower-order flat
elements is not recommended.
5. Curved shell elements have no stiffness in the rotational degree of freedom when the formulation is used
consequently, mechanisms may be introduced. Mechanisms occur when elements, RBEs, or MPCs are
connected to the out-of-plane rotation of the shell grid points. Note that PARAM,AUTOSPC,YES does not
constrain these mechanisms.
6. In linear solution sequences, the values of param,k6rot,100. and param,snorm,20. are the
default.
7. Transverse shear flexibility (MID3 on the PSHELL property entry) should be left on when normals are used.
Three-Dimensional Elements
Although the CHEXA and CPENTA elements are designed to behave reasonably well as thin shell elements, it is
recommended that they not be used in this capacity. The high ratio of extensional stiffness in the direction normal to
the effective transverse shear stiffness can produce significant round-off errors.
As in the case of plate elements, if midside nodes are present for the solid elements, they should be located as close to
the center of the edge as possible. Again, if midside nodes are desired, the general recommendation is to include all of
them.
R-Type Elements
A high degree of precision must be maintained when specifying coefficients for MPCs in order to avoid introduction
of unintentional constraints to rigid body motions. Rigid elements (e.g., RBE2, RBAR, etc.) should be used whenever
possible because their constraint coefficients are internally calculated to a high precision. Furthermore, these R-type
elements require much less user interaction. The weld family of elements offer excellent ways of modeling structural
connections.
Mesh Density
The mesh density in a finite element model is an important topic because of its relationship to accuracy and cost. In
many instances, the minimum number of elements is set by topological considerations; e.g., one element per member
in a space frame or one element per panel in a stiffened shell structure. In the past, when problem size was more
severely limited, it was not uncommon to lump two or more frames or other similar elements in order to reduce the
size of the model. With computers becoming faster and cheaper, the current trend is to represent all major components
individually in the finite element model.
If the minimum topological requirements are easily satisfied, the question remains as to how fine to subdivide the
major components. The question is particularly relevant for elastic continua, such as slabs and unreinforced shells. In
general, as the mesh density increases, you can expect the results to become more accurate. The mesh density required
can be a function of many factors. Among them are the stress gradients, the type of loadings, the boundary conditions,
the element types used, the element shapes, and the degree of accuracy desired.
The grid point spacing should typically be the smallest in regions where stress gradients are expected to be the steepest.
Figure 11-7 shows a typical example of a stress concentration near a circular hole. The model is a circular disk with an
Main Index
658 Nonlinear Users Guide
Modeling Guidelines
inner radius = a and an outer radius = b. A pressure load p i is applied to the inner surface. Due to symmetry, only half
of the disk is modeled. In the example, both the radial stress and the circumferential stress decrease as a function of
2
lr from the center of the hole. The error in the finite element analysis arises from differences between the real stress
distribution and the stress distribution within the finite elements.
In a study of mesh densities, elements and output options, three different mesh densities were used in the example as
shown in Figure 11-7. The first one is a coarse mesh model with the elements evenly distributed. The second model
consists of the same number of elements; however, the mesh is biased toward the center of the hole. The third model
consists of a denser mesh with the elements evenly distributed. These three models are then analyzed with three
different element types CQUAD4, CQUAD8, and CQUAD4 with the corner stress option. The circumferential stress
at the inner radius is always greater than p i , which is the applied pressure load at the inner radius, and approaches this
value as the outer radius becomes larger. The theoretical circumferential stress is given by the following equation
(see [Ref. 2]):
2
2 b
p i a 1 + -----
2
r
= ---------------------------------
2 2
b a
where:
a = inner radius
b = outer radius
The stresses are then plotted as a function of the radius in a nondimensional fashion-stress p i / versus r / a . The results
are summarized in Table 11-1.
Main Index
Chapter 11: Element Library 659
Modeling Guidelines
z x
pi
Fine Even Mesh 2a
2b
pi
Main Index
660 Nonlinear Users Guide
Modeling Guidelines
For this particular case, since the stresses are proportional to 1 r 2 , you expect the highest stress to occur at the inner
radius. In order to take advantage of this piece of information, the obvious thing to do is to create a finer mesh around
the inner radius. Looking at the results for the first two cases in Table 11-1, it is quite obvious that just by biasing the
mesh, the results are 30% closer to the theoretical solution with the same number of degrees of freedom.
A third case is analyzed with a finer but unbiased mesh. It is interesting to note that for case number 3, even though it
has more degrees of freedom, the result is still not as good as that of case number 2. This poor result is due to the fact
that for the CQUAD4 element, the stresses, by default, are calculated at the center of the element and are assumed to
be constant throughout the element. Looking at Figure 11-7, it is also obvious that the centers of the inner row of
elements are actually further away from the center of the circle for case number 3 as compared to case number 2. The
results for case number 3 can, of course, be improved drastically by biasing the mesh.
You can request corner outputs (stress, strain, and force) for CQUAD4 in addition to the center values. Corner results
are extrapolated from the corner displacements and rotations by using a strain rosette analogy with a cubic correction
for bending. The same three models are then rerun with this corner option their results are summarized in cases 7
through 9. Note that the results can improve substantially for the same number degrees of freedom.
Corner output is selected by using a corner output option with the STRESS, STRAIN, and FORCE case control
commands. When one of these options is selected, output is computed at the center and four corners for each CQUAD4
element, in a format similar to that of CQUAD8 elements.
There are four corner output options available: CORNER, CUBIC, SGAGE, and BILIN. The different options provide
for different approaches to the stress calculations. The default option is CORNER, which is equivalent to BILIN.
BILIN has been shown to produce better results for a wider range of problems.
To carry it a step further, the same three models are then rerun with CQUAD8 (cases 4 through 6). In this case, the
results using CQUAD8 are better than those using the CQUAD4. This result is expected since CQUAD8 contains more
degrees of freedom per element than CQUAD4. Looking at column three of Table 11-1, you can see that due to the
existence of midside nodes, the models using CQUAD8 contain several times the number of degrees of freedom as
compared to CQUAD4 for the same number of elements. The results using CQUAD4 can, of course, be improved by
increasing the mesh density to approach that of the CQUAD8 in terms of number of degrees of freedom.
It is important to realize that the stresses are compared at different locations for Cases 1 through 3 versus Cases 4
through 9. This difference occurs because the stresses are available only at the element centers for Cases 1 through 3,
but the stresses are available at the corners as well as the element centers for Cases 4 through 9. When looking at your
results using a stress contour plot, you should be aware of where the stresses are being evaluated.
Main Index
Chapter 11: Element Library 661
Modeling Guidelines
How fine a mesh you want depends on many factors. Among them is the cost you are willing to pay versus the accuracy
you are receiving. The cost increases with the number of degrees of freedom. The definition of cost has changed with
time. In the past, cost is generally associated with computer time. With both hardware and software becoming faster
each day, cost is probably associated more with the time required for you to debug and interpret your results. In general,
the larger the model is, the more time it takes you to debug and interpret your results. As for acceptable accuracy,
proceeding from case 8 to case 6, the error is reduced from 4.7% to 1.5%; however, the size of the problem is also
increased from 194 to 2538 degrees of freedom. In some cases, a 4.7% error may be acceptable. For example, in cases
in which you are certain of the loads to within only a 10% accuracy, a 4.7% error may be acceptable. In other cases, a
1.5% error may not be acceptable.
In general, if you can visualize the form of the solution beforehand, you can then bias the grid point distribution.
However, this type of information is not necessarily available in all cases. If a better assessment of accuracy is required
and resources are available (time and money), you can always establish error bounds for a particular problem by
constructing and analyzing multiple mesh spacings of the same model and observe the convergences. This approach,
however, may not be realistic due to the time constraint.
h-adaptivity
In traditional finite element analysis, as the number of elements increases, the accuracy of the solution improves. The
accuracy of the problem can be measured quantitatively with various entities (such as strain energies, displacements,
and stresses), as well as in various error estimation methods (such as simple mathematical norm or root-mean-square
methods). The goal is to perform an accurate prediction on the behavior of your actual model by using these error
analysis methods. You can modify a series of finite element analyses either manually or automatically by reducing the
size and increasing the number of elements, which is the usual h-adaptivity method. Each element is formulated
mathematically with a certain predetermined order of shape functions. This polynomial order does not change in the
h-adaptivity method.
Main Index
662 Nonlinear Users Guide
References
References
1. A Proposed Standard Set of Problems to Test Finite Element Approach, MSC/NASTRAN Application Notes,
1984.
2. S. P. Timoshenko and J. N. Goodier, Theory of Elasticity, McGraw-Hill Book Company, Third Edition, 1970.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints
Nonlinear Users Guide
Main Index
664 Nonlinear Users Guide
Introduction
Introduction
In engineering problems, it is often useful to apply constraints between different parts or regions on along a surface of
a part. These mathematical constraints are sometimes applied to single nodes of the model, called boundary
constraints, or are sometimes used to connect many nodes or points of the model, called multi-point constraints
(MPCs). Various types of connections such as rigid, limited strength cohesion, moment carrying rigid, etc. can be
simulated by properly using these mathematical constraint equations as part of the model. These connections can even
be used to simulate manufacturing processes such as seam welds or spot welds. The vast flexibility that these
mathematical constraints afford make them extremely useful in FEA for everything from weld, bolt, and fastener
modeling to glued contact to separating contact.
In this chapter, we describe the various types of multi-point constraints, rigid elements, and fasteners available in MSC
Nastran SOL 400 and the input required to incorporate them into a model.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 665
Kinematic Constraints
Kinematic Constraints
SOL 400 allows you to input kinematic constraints through various options that include
MPC Constraints
Rigid Elements Constraint
Bolts
Shell-to-Solid Tying
Support Conditions
Bushings
MSC Nastran RBE2 and RBE3
Beam Shell Offsets
Pin Code for Beam Elements
Main Index
666 Nonlinear Users Guide
MPC Constraint
MPC Constraint
MSC Nastran contains a generalized MPC (constraint) condition bulk data entry. Any constraint involving linear
dependence of nodal degrees of freedom can be included in the stiffness equations.
A MPC constraint involves one tied node and one or more retained nodes, and a MPC (constraint) condition between
the tied and retained nodes. The degrees of freedom (for example, displacements, temperatures) of the tied node are
dependent on the degrees of freedom of the retained nodes through the MPC condition. In some special MPC
conditions, the tied node can also be a retained node. The MPC condition can be represented by a MPC (constraint)
matrix. Note that if the MPC constraint involves only one retained node, the choice of which node is to be tied to
retained is arbitrary.
As a simple example, impose the constraint that the first degree of freedom of node I be equal to that of node J at all
times (see Figure 12-1).
I UI
Constraint Equation:
UI = UJ
2,V
J UJ
As a second example, the simulation of a sliding boundary condition requires the input of both the boundary conditions
and the tying constraints (see Figure 12-2).
Local axes
X
Y (v) i Y
j
k
l
X (u)
The example illustrated in Figure 12-2 enforces rigid sliding on the boundary in the local coordinates defined above.
vi = vj = vk = v1 = 0 (12-1)
ui = uj = uk = u1 (12-2)
The first equation is a set of fixed boundary conditions. The second equation is a constraint equation and can be
rewritten as three constraint equations:
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 667
MPC Constraint
ui = uj (12-3)
uk = uj (12-4)
u1 = uj (12-5)
These equations express all the u displacements in terms of u j . In this example, node j is chosen to be the retained
node; nodes i , k , and l are tied nodes. You can use the MPC option to enter this information.
The MPC option uses the homogeneous linear constraint capability (tying) to input simple constraints of the form
Aj j = 0 (12-6)
j
A j j = 0 (12-7)
j
The total or incremental displacement are always with respect to the local coordinate system.
MPC Types
To create an MPC, first select the type of MPC to be created from the option menu. The MPC types that appear in the
option menu are dependent on the current settings of the Analysis Code and Analysis Type preferences. The following
table describes the MPC types which are supported for MSC Nastran.
Main Index
668 Nonlinear Users Guide
MPC Constraint
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 669
MPC Constraint
Typical applications that use R-type entries are shown in Table 12-1.
Main Index
670 Nonlinear Users Guide
MPC Constraint
Degrees of Freedom
Whenever a list of degrees of freedom is expected for an MPC term, a listbox containing the valid degrees of freedom
is displayed on the form.
The following degrees of freedom are supported by the Patran MSC Nastran MPCs for the various analysis types:
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 671
MPC Constraint
Note: Care must be taken to make sure that a degree of freedom that is selected for an MPC actually exists at the
nodes. For example, a node that is attached only to solid structural elements will not have any rotational
degrees of freedom. However, Patran will allow you to select rotational degrees of freedom at this node
when defining an MPC.
Explicit MPCs
This subordinate MPC form appears when the Define Terms button is selected on the Finite Elements form and Explicit
is the selected type. This form is used to create an MSC Nastran MPC bulk data entry. The difference in explicit MPC
equations between Patran and MSC Nastran will result in the A1 field of the MSC Nastran entry being set to -1.0.
For = B = 0.5 ; SID = 1; T = ID of Tied Node; R = ID of Retained Node. Here is the approximate form of the
MPC bulk data entry:
Main Index
672 Nonlinear Users Guide
MPC Constraint
T Tied Node
T R Retained Node
R
Figure 12-4 Mesh Refinement for 4-Node Quad
For higher-order elements, the same method would apply only care must be taken to select the correct nodes to be tied
and retained. Also, the ratios applied to the coefficients are slightly more complex.
T
R T Tied Node
R Retained Node
T
R
Figure 12-5 Mesh Refinement for 8-Node Quad
For solid elements, there are no rotational degrees of freedom, so only the degrees of freedom 1 to 3 need to be tied.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 673
MPC Constraint
T
R
R
Figure 12-6 Mesh Refinement for 8-Node Brick
For higher-order solid elements, again, care must be taken to select the correct tied (labeled T) vs. retained (labeled R)
nodes.
R
T
R
T
R
T
R
R
Figure 12-7 Mesh Refinement for 20-Node Brick
Main Index
674 Nonlinear Users Guide
Automated Bolt Modeling
Implementation
Assuming that the finite element mesh of the bolt has been split up in two disjoint parts along element boundaries
across the shaft of the bolt, the BOLT bulk data entry imposes the proper multi-point constraints on these parts, such
that the bolt can be pre-stressed, locked, and loosened by applying appropriate boundary conditions to a special
control grid point associated with the bolt. Special treatment of the constraints by the contact algorithm ensures that
the option can be used with contact and that the internally generated contact surface will be continuous across the split.
Grids touching the bolt surface can then slide across the split without problems. The automated bolt modeling method
is based on a mesh split principle and a control node. Figure 12-8 demonstrates how the mesh split principle is applied
to a typical joint.
Figure 12-8 Example of a Bolted Joint using BOLT using the Mesh Split Principle
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 675
Automated Bolt Modeling
You must define the Top Nodes (GT) and the Bottom Nodes (GB) and the control node (GC).
Bolt MPCs have one tied node and two retained nodes. The tied node and the first retained node are usually nodes on
the boundaries of the respective parts (see Figure 12-10). The second retained node is very often a free node and is
usually shared by all bolt MPCs which connect the parts. This node is also called the control node of the MPC, since
it can be used to apply load or control the size of the gap or overlap between the parts.
Main Index
676 Nonlinear Users Guide
Automated Bolt Modeling
undeformed deformed
Figure 12-10 Pre-stressing a Structure by Creating an Overlap Between the top and the Bottom Part Using
BOLT MPCs
in which u tied , u retained and u control are the displacement and rotation degrees of freedom (if any) of, respectively, the
tied node, the first retained node, and the control node of the tying. It immediately follows from this equation that
u control is the displacement difference of the tied and the retained node of the tying and is equal to the size of the
overlap or gap between the parts. Hence, by prescribing this displacement using the SPCD and SPCR options, gaps or
overlaps of a particular size can be created.
Instead of prescribing the size, gaps or overlaps can also be created by prescribing the total force in the model. This
follows immediately from the fact that the work done by a constraint equation is zero. To demonstrate this, consider
the model displayed in Figure 12-8. The model is split into two disjoint parts: the top part and the bottom part. At the
split, corresponding nodes of the respective parts are connected by BOLT MPCs, in which the nodes of the bottom part
are the tied nodes and the nodes of the top part are the first retained nodes of the tyings. Both tyings share a common
(free) control node. If u 1 bot and u 2 bot are the displacements of the bottom nodes and u 1 top and u 2 top are the
displacements of the top nodes, then the two BOLT MPCs impose the following two constraints on this model:
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 677
Automated Bolt Modeling
T
I 0 I 0
T = 0 I 0 I ,
I I 0 0
in which I and 0 are the unit and null matrices, equations (12-9) and (12-10) can be summarized into the matrix
equation:
u = Tu . (12-11)
Let
T T
F = F 1 bot F 2 bot F 1 top F 2 top and F = F 1 top F 2 top F control
be the force vectors which are work conjugate to the displacements u and u , respectively. Then the zero-work
principle states that
T
F u = F u .
T
(12-12)
Substitution of equation (12-11) into equation (12-12) and requiring that the result is valid for all u , it follows that
T F = F , or
T
Equations (12-13) and (12-14) express the equilibrium of forces across the split, in which F 1 top and F 2 top can be
viewed as externally applied forces. While equation (12-15) enables BOLT MPCs to pre-stress a model with a certain
total force, it states that if the constraints described by Equations (12-9) and (12-10) are applied, the force on the control
node is the sum of the forces on the tied nodes of all BOLT constraints which share that control node. Hence, the total
force on the bottom part of the model is prescribed by applying a force to the control node using the FORCE option.
Obviously (by the equilibrium equations (12-13) and (12-14)) if no external force is applied to the nodes on the split,
the total force on the top part is equal but opposite is sign to the force on the control node.
Main Index
678 Nonlinear Users Guide
Automated Bolt Modeling
Note: Both methods to control the size of overlaps or gaps can be combined in one analysis. For example, in the
first part of an analysis, a pre-stress can be defined by prescribing the net force in the structure using the
FORCE option on the control node of the BOLT MPCs. Then in the second part, the size of the gap or the
amount of overlap which is the result of this load, can be fixed by suppressing the displacement changes
of the control node using the SPCD bulk data entry.
The control node of an BOLT MPC has the same displacement and rotation degrees of freedom as the tied and the first
retained node as given by equation (12-8). This allows creation of overlaps or gaps in any particular direction or
combination of directions. Local coordinate systems can be defined at any of the nodes of an BOLT MPC using any of
the CORDnm bulk data entry. Since the constraint equation (12-9) is always applied in the global coordinate system, a
local coordinate system at the control node can be used to pre-stress the model in directions other than the global
coordinate directions.
Sufficient boundary conditions must be applied on the control node to suppress any rigid body modes (e.g. lateral
constraints must be applied to keep the parts of the bolt from sliding tangentially relative to one another), if the two
parts of the structure are not constrained otherwise. Total forces on the split are available for postprocessing as reaction
forces on the control node for all suppressed or prescribed displacements and rotations.
BOLT MPCs can be used in combination with contact; that is, the tied and the first retained node may be nodes on the
boundary of a contact body.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 679
Automated Bolt Modeling
In non-mechanical passes of a coupled analysis, the BOLT MPC reduces to a regular heat-transfer MPC between the
tied node and the retained node. There is no dependence on the control node in that case. This guarantees continuity
of the primary field variable (for example, temperature) across the split.
As an example of the usage of BOLT for defining the constrains, see Engine Gasket (Ch. 10) and Chapter 23: Bolted
Plates in the MSC Nastran Demonstration Problems Manual.
Main Index
680 Nonlinear Users Guide
Support for Bolts
Format:
1 2 3 4 5 6 7 8 9 10
BOLT ID GRIDC
TOP GT1 GT2 GT3 GT4 GT5 GT6 GT7
GT8 GT9 etc.
BOTTOM GB1 BG2 BG3 BG4 GB5 GB6 GB7
GB8 GB9 etc.
Example:
Field Contents
ID Element ID of the bolt. (Integer; Required; no Default)
GRIDC Control grid ID where forces or displacements are applied. (Integer; no Default; Required)
TOP Enter the character string TOP to define the start of the entry that defines all of the grids at
the top of the bolt intersection with the structure. (Integer; no Default)
GT1, GT2, etc. Grid IDs of the grid points at the top of the bolt intersection. (Integer; no Default)
BOTTOM Enter the character string BOTTOM to define the start of the entry that defines all of the grids
at the bottom of the bolt intersection with the structure (do not enter the ID for GRIDC).
(Integer; no Default)
GB1, GB2, etc. Grid IDs of the grid points at the bottom of the bolt intersection. (Integer; no Default)
Remarks:
1. The grids entries of the TOP and BOTTOM keywords are open-ended.
2. GRIDC is the control grid point and usually not connected to any element.
3. (GTi, GBi) are pairs of grid on top and bottom.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 681
Support for Bolts
4. To each pair of (GTi, GBi) and GRIDC, MPCs are created internally to all six degrees of freedom. Since the
GBs always belong to dependent-degrees of freedom, they cannot be applied to any SPC, SPC1, SPCD, and
SPCR.
5. Same number of grid points in TOP and BOTTOM. They should be coincident but it is not required. Users who
do otherwise do so at their own risk since the current design does not consider the initial offset between them.
6. Bolt loads, including enforced motion, are usually prescribed on GRIDC to represent the pre-tension, overlap
or loading of the bolt. BOLT relative displacements are given in the global coordinate system of the control
node.
7. Global Coordinate System may have to be defined at the Control Node if the bolt direction is not a Basic
Coordinate direction and you may want to apply the loads along the shaft of the bolt.
8. Loads in directions other than the shaft of the bolt direction are possible.
9. The internally written MPC relationship is of the form:
u GB = u GT + u GC
10. In 3-D contact analysis, it replaces GBi (bottom bolt segment) by GTi (top bolt segment) on the internally
generated contact surface, which makes contact surface continuous across the mesh split between them.
Force output is obtained through the MPCFORCE case control command.
Figure 12-11 Creating Bolt in Patran using the Automatic Method - Step 1
Main Index
682 Nonlinear Users Guide
Support for Bolts
Figure 12-12 Creating Bolt in Patran using the Automatic Geometric Method - Steps 2-3
4. Select all the elements and press Apply to complete the operation.
Figure 12-13 Creating Bolt in Patran using the Automatic Geometric Method - Step 4
Patran automatically detects the long direction of the set of elements and splits the mesh at the center. A new
coordinate system is created so that the Y is aligned with the long direction. Patran automatically splits the mesh and
creates duplicate nodes at the center of the part. The control node is off-set from the elements and the bolt is graphically
displayed with lines connecting the control node to the split plane. When the model is run, the stresses are a constant
5000 as expected. The following figures demonstrate the steps in Patran.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 683
Support for Bolts
Preliminaries:
1. Create a local coordinate system at the split plane location. Note that Y direction of the local coordinate
system must be aligned with the axial direction of the bolt because Patran will create the split plane in the XY
plane of the local coordinate system. In this example, the Z direction of coordinate system 1 is used as the
axial direction.
Proceed to set up the bolt using the following steps:
1. In the Patran Assemble tab, select the Create 3D Bolt icon to access the 3D Bolt Model form
2. Set the Axial Preload Type to Force and enter a Value of 5000.
3. Set the Method to Vectorial.
4. Select the coordinate system and all elements.
5. Press Apply to complete the operation.
Patran automatically splits the nodes closest to the XY plane of the splitting plane coordinate system. When the model
is run, the stresses are a constant 5000 as expected. The following figures demonstrate the steps in Patran.
Main Index
684 Nonlinear Users Guide
Support for Bolts
Figure 12-15 MSCUG example nug_10.dat BOLT entry (file nug_10-new.dat with BOLT)
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 685
Support for Bolts
Figure 12-16 MSCUG example nug_10-new.dat Stress animation from BOLT preload.
GUI Support
Bolt Preload
The Patran bolt modeling tool allows you to facilitate the creation of bolt preloads and the associated MPCs. This tool
can be used for any MSC Nastran analysis that recognize explicit MPC definitions. In certain cases it will create BOLT
over-closure MPCs.
Main Index
686 Nonlinear Users Guide
Support for Bolts
Regardless of which method is used, the following operations are done to create each bolt pre-load.
1. The mesh of the bolt is split at the designated cutting plane. This cutting plane is either designated by the user
or determined by the tool. Coincident nodes are created for element nodes above and below the plane and the
element connectivities are regenerated.
2. MPCs are created to connect the two sections above and below the determined cutting plane. The coincident
nodes created in the previous step are tied together via MPCs. Either Overclosure type or Explicit MPCs are
created.
3. A control node is created and connected to each set of node pairs. A coordinate system is created at the specified
control node location which can be offset as required for visualization purposes
4. A load or displacement is applied for the pre-load in the local Y direction (axial bolt direction) or the coordinate
system created at the control node location.
5. Each bolt pre-load is named and stored in the database.
6. Undo, Show, and Delete capabilities are available.
Three methods to help you define the bolt pre-loads are available. Each has some specific inputs required. All have
common inputs some specific inputs for each method explained as follows:
Geometric Based
This is by far the easiest and most intuitive method to use. A set of elements is selected that defines the bolt. Based on
the geometric dimensions and Patrans mass property utilities, the most likely axial direction of the bolt is determined.
The mesh is split at the center of the selected element set. If the geometric properties of the selected element set does
not distinctly determine the axial direction of the bolt, the cutting plane may not be determined correctly and a strange
cut will result. At least a two to one axial versus width ratio is recommended.
Vector Based
This method requires that you define a vector, the base of which determines the location of the cutting plane, which is
normal to the defined vector. The element set selected must have a layer or two or three element both above and below
the base of the defined vector plane.
Element Based
This method may be more advantageous for certain element types in that you select element faces (3-D), element edges
(2-D), or elements (1-D) where you want the mesh split and the MPCs applied. For 3-D elements, this can be difficult
as the graphics screen can get messy trying to select element faces, especially TET elements. For 2-D and 1-D elements,
this method is easy. The split occurs only for the selected element entities, which may be problematic if you only select
a portion of the cross section.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 687
Support for Bolts
Action
Create Create an bolt preload. If you need to modify an existing bolt preload, you have to
delete and then re-create the bolt preload.
Delete Delete an existing bolt preload. You can select an existing named bolt preload or if
you are not sure which one you want, you can set the Method to Control Node and
select the control node. It will figure out which bolt preload is associated to the
selected control node. Note that any entities created by the bolt preload are deleted
unless those entities are modified by the user to reference other entities in some way.
Warning message are issued in those cases and some of the entities such as the
coordinates will not be deleted.
Identify Identify allows you to select a control node and the name of the bolt preload is
returned as an informational message.
Show Show is informational only and indicates the total number of bolt pre-loads defined.
Object When creating a bolt preload, you may place either a force or a prescribed
displacement.
Method The three methods are described above as Geometric, Vector, and Element based.
New Bolt Name Enter a bolt preload name.
MPC Type Select the type of MPC required. You may not be given a choice depending on the
Analysis Preference set. Overclosure is the desired method for MSC Nastran SOL
400. Other MSC Nastran solutions and other analysis codes need Explicit MPCs
defined. For MSC Nastran, SOL 400, BOLT entry is created for Overclosure.
Main Index
688 Nonlinear Users Guide
Support for Bolts
Control Node Offset A vector can be supplied (optional) to offset the location of the control node. This is
highly recommended for visualization purposes. A vector is provided, the length of
which is the offset from the center of the cutting plane. You may enter a vector in any
way the is allowable with the select mechanism or you can manually type between
the < > brackets a valid three-component translational vector, e.g., <0, 5, 2>.
Method = Geometric
Element List Select a list of elements that sufficiently defines the bolt geometry. The axial
direction of the bolt should be longer than the lateral directions in order for the
algorithm to determine the proper axial direction of the bolt. The cutting plane will
be created at the center of the bolt as best as possible based on the element
connectivity. The control node offset will be created from the location of the cutting
plane.
Method = Vector
Bolt Axis This is a required vector input to define the axial direction of the bolt. Typically, you
only need to use the select mechanism to define a vector using two nodes on the
outside of the bolt. You can use any method that the select mechanism allows to
define a vector. You may enter a vector in any way the is allowable with the select
mechanism or you can manually type between the < > brackets a valid three-
component translational vector, e.g., <0, 5, 2>. In this method, you can define any
direction to do the cut and apply the preload. Care should be taken in that some
strange cuts may results. This is the method to use if you wish to use a non-axial bolt
preload.
Element List For this method, select enough elements above and below the base of the Bolt Axis
vector such that a proper cut can be made. Two or three layers of elements above and
below is usually sufficient. An error may result or a bad looking cut may be had if
not enough element above and below are given.
Method = Elemental
Target Element Type In this method, you must select the element dimensionality of the bolt: 1-D bars, 2-D
elements, or 3-D solids.
Application Region The application region defines the actual location where the split is made. For 1-D
beam/bar elements, you select a single element and the split is made at the bottom
node. For 2-D (2-D solid) elements, you select element edges that define the cut. For
3-D elements, you select the element faces to define the cut. All the element edges
or faces need to be connected together in order for this method to create a proper cut.
Use the FEM application and do a Verify/Element/Boundary to ensure that the bolt tool properly cut the bolt into two
sections with coincident nodes between.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 689
Introduction to R-Type Elements
Main Index
690 Nonlinear Users Guide
Rigid Elements
Rigid Elements
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 691
Rigid Elements
As an introduction to the R-type elements, consider the RROD rigid element. For the RROD element, you specify a
single component of translation at one of its two end points as a dependent degree of freedom. The equivalent
component at the other end is the independent degree of freedom. Consider the example shown in Figure 12-18.
1 2 3 4
X
Figure 12-18 An RROD Connection
For this example, a rigid connection is made between the X component of grid point 3 to the X component of grid point
2. When you specify the RROD bulk data entry in the QRG as shown, you are placing component 1 of grid point 3 (in
the global system) into the m-set. The remaining five components at grid point 3 and all the components at grid point
2 are placed in the n-set and, hence, are independent. CBAR 1 is not connected to components Y, Z, R x , R y , and R z
of grid point 3 in any manner; therefore, the ends of the bars are free to move in any of these directions.
However, the ends of the two CBAR elements are rigidly attached in the X-direction. Having the connection only in
the X-direction shows that one degree of freedom is placed in the m-set for each RROD element.
For the other six R-type elements, several degrees of freedom may be specified as members of the m-set or the n-set.
(Again, the latter designation may be temporary; they may be removed by additional constraints in your model.) For
the RTRPLT, RBE1, RROD, and RBE3 elements, any unlisted degrees of freedom at the grid points to which the
element is joined are not connected to the element. This lack of connection can be regarded as either a sliding release
or a rotating joint release, or both.
Five of the R-type elements (RBAR, RJOINT, RTRPLT, RBE1, and RBE2) must have exactly six components of motion
in the n-set (i.e., independent degrees of freedom). These six degrees of freedom must be able to represent all of the
rigid body motions of the element.
Main Index
692 Nonlinear Users Guide
Rigid Elements
One word of caution is that the connection of two or more rigid elements to the same grid point should be done
carefully to avoid specifying a degree of freedom as a member of the m-set more than once; otherwise, a fatal message
is issued.
RBAR Element
The RBAR element rigidly connects from one to six dependent degrees of freedom (the m-set) to exactly six
independent degrees of freedom. The six independent degrees of freedom must be capable of describing the rigid body
properties of the element.
The most common approach when using the RBAR element is to define one end of the RBAR with all six independent
degrees of freedom with dependent degrees of freedom at the other end; i.e., the RBAR1 format. (However, placing all
of the independent degrees of freedom at one end is not a requirement.) To determine if the choice you make for the
independent degrees of freedom meets the rigid body requirements, ensure that the element passes the following simple
test.
The format for the RBAR bulk data entry is shown in the QRG.
If you constrain all of the degrees of freedom defined as independent on the RBAR element, is the element prevented
from any possible rigid body motion?
As an example, consider the RBAR configurations shown in Figure 12-19. For configurations (a) and (c), if the six
independent degrees of freedom are held fixed, the element cannot move as a rigid body in any direction. However,
for (b), if all six of the independent degrees of freedom are held fixed, the element can still rotate about the Y-axis.
Configuration (b) does not pass the rigid body test and does not work as an RBAR element.
GA GB
Acceptable
Independent 1 - 6 (a)
GA GB
Not Acceptable
Independent 1,2,3,4,6 (b) Independent 2
GA GB
Acceptable
Z
Independent 1,2,3,4,6 (c) Independent 3
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 693
Rigid Elements
MSC Nastran generates internal MPC equations for the R-type elements. As an example of this, consider the model of
a thick plate with bars attached as shown in Figure 12-20. The interface between the bars and the plate is modeled two
ways, first using MPC entries and second using RBAR elements.
2 Bar
1" Thick Plate .5
QUAD Plate
Bars 1
.5
3 Bar
X
Figure 12-20 Model of a Thick Plate with Bars Attached
Plate theory states that plane sections remain planar. If this is the case, then grid points 2 and 3 are slaves to grid point
1. Therefore, you need to write the equations for the in-plane motion of grid points 2 and 3 as a function of grid point
1. Each RBAR element creates up to six constraint equations.
Looking only at the motion in the x-y plane,
u6 = u6
3 1
u6 = u6
2 1
u2 = u2
2 1
u2 = u2
3 1
Main Index
694 Nonlinear Users Guide
Rigid Elements
MPC
1 2 3 4 5 6 7 8 9 10
MPC 1 2 1 1. 1 1 -1.
1 6 .5
MPC 1 3 1 1. 1 1 -1.
1 6 -.5
MPC 1 3 6 1. 1 6 -1.
MPC 1 2 6 1. 1 6 -1.
MPC 1 2 2 1. 1 2 -1.
MPC 1 3 2 1. 1 2 -1.
Note: MPC = 1 must appear in the case control section to use these entries.
The R-type elements are easier to use than the equivalent MPC entries.
1 2 3 4 5 6 7 8 9 10
$RBAR EID GA GB CNA CNB CMA CMB
RBAR 99 1 2 123456
RBAR 100 1 3 123456
RBAR 99 generates MPC equations for the motion of grid point 2 as a function of grid point 1. Likewise, RBAR 100
generates MPC equations for the motion of grid point 3 as a function of grid point 1. These RBARs generate the MPC
equations for all six degrees of freedom at grid points 2 and 3. If it is desired to have the equations generated only for
the in-plane motion, the field labeled as CMB in the RBAR entries has the values 126 entered.
A particular area of confusion for the new user is when you need to connect R-type elements together. The important
thing to remember is that you can place a degree of freedom into the m-set only once. Consider the two RBAR elements
shown in Figure 12-21 that are acting as a single rigid member.
If you choose grid point 1 for RBAR 1 to be independent (1-6), then grid point 2 for RBAR 1 must be dependent (1-6).
Since grid point 2 is dependent for RBAR 1, it must be made independent for RBAR 2. If you made grid point 2
dependent for RBAR 2 as well as RBAR 1, a fatal error would result. Since grid point 2 is independent (1-6) for
RBAR 2, grid point 3 will be dependent (1-6).
If you chose grid point 1 of RBAR 1 to be dependent, then grid point 2 for RBAR 1 would be independent. Grid point
2 of RBAR 2 would be dependent, and grid point 3 of RBAR 2 would be independent.
The RBAR element is often used to rigidly connect two grid points in your model.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 695
Rigid Elements
Y2 Y2
RBAR 1 RBAR 2
1 2 X2 2 X2 3
Both options for connecting the RBAR elements are shown in Listing 12-1. (See
MSC_DOC_DIR/doc/doc/linstat/rbar1.dat) For clarity, CBAR 3, which is connected to grid point 3, is not
shown in Figure 12-21.
GRID 3 20. 0. 0.
GRID 4 30. 0. 0. 123456
$
$ OPTION 1
$
RBAR 1 1 2 123456
RBAR 2 2 3 123456
$
$ OPTION 2
$
$RBAR 1 1 2 123456
$RBAR 2 2 3 123456
$
CBAR 3 1 3 4 0. 1. 0.
PBAR 1 1 1. 1. 1. 1.
MAT1 1 20.4 .3
$
$ POINT LOAD
$
Main Index
696 Nonlinear Users Guide
Rigid Elements
FORCE 1 1 1. 1. 0. 0.
FORCE 2 1 1. 0. 1. 0.
$
ENDDATA
As a final example of the RBAR element, consider the hinge model shown in Figure 12-22.
CBAR 1 CBAR 2
1 2,3 4
Hinge Free to rotate about YB
Connection the basic Z-axis.
Y2 Y3
XB
1 2 X2
3
X3
4
The simplest way to model the hinge connection with an RBAR is to use coincident grid points at the center of rotation
(grid points 2 and 3 in this example) and define an RBAR between the two grid points. This RBAR has zero length,
which is acceptable for the RBAR. Make all six of the components associated with one grid point independent. Make
only a select number of components of the other grid point dependent, leaving independent the components
representing the hinge. A listing of the input file for this model is shown in Listing 12-2. (See
MSC_DOC_DIR/doc/linstat/rbar2.dat) Note that the components 1 through 6 of grid point 2 are
independent and components 1 through 5 of grid point 3 are dependent. Component 6 of the grid point 3 is left
independent, permitting CBAR 2 to rotate about the Z axis with respect to CBAR 1.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 697
Rigid Elements
FORCE = ALL
BEGIN BULK
$
GRID 1 0. 0. 0. 123456
GRID 2 10. 0. 0.
GRID 3 10. 0. 0.
GRID 4 20. 0. 0. 123456
$
RBAR 99 2 3 123456 12345
$
CBAR 1 1 1 2 0. 1. 0.
CBAR 2 1 3 4 0. 1. 0.
PBAR 1 1 .1 .01 .01 .02
MAT1 1 20.+4 .3
$
$ POINT LOAD
$
FORCE 1 2 100. 0. 1. 0.
$
ENDDATA
A word of caution: if you intentionally use coincident grid points (as in this example), you may run the risk of removing
them accidentally if you later use the equivalence option available in most preprocessors. Equivalencing grid points
causes all duplicate grid points in your model to be deleted; however, this may not be your intention as illustrated in
this example.
Main Index
698 Nonlinear Users Guide
Rigid Elements
Y
1 = 90
X
R1, 1 = 0
Tube Dimensions:
Mean Diameter = 30 mm
T = 3 mm
Length = 90 mm
E = 25 x 104 N/mm2
= 0.3
A partial input file for this model is shown in Listing 12-3. (See /doc/linstat/torque.dat) The goal is to have the end of
the tube rotate uniformly while allowing the tube to expand in the R-direction. The simplest way to accomplish this is
to define a local cylindrical coordinate system with the origin located at the center of the free end as shown in
Figure 12-23. This cylindrical coordinate system is then used as the displacement coordinate system (field 7) for all of
the grid points located at the free end. Now when an RBE2 element is connected to these grid points, the dependent
degrees of freedom are in the local coordinate system. By leaving the R-direction independent, the tube is free to
expand in the radial direction.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 699
Rigid Elements
$
CORD2C 1 0 0.0 0.0 90. 0.0 0.0 91.0
1.0 0.0 91.0
$
$
GRID 101 15. 0.0 0.0
GRID 102 10.6066 10.6066 0.0
GRID 103 7.105-1515. 0.0
GRID 104 15. 0.0 30.
GRID 105 10.6066 10.6066 30.
GRID 106 1.066-1415. 30.
GRID 107 15. 0.0 60.
GRID 108 10.6066 10.6066 60.
GRID 109 1.066-1415. 60.
GRID 110 15. 0.0 90. 1
GRID 111 10.6066 10.6066 90. 1
GRID 112 1.421-1415. 90. 1
GRID 113 -10.606610.6066 0.0
GRID 114 -15. 7.105-150.0
GRID 115 -10.606610.6066 30.
GRID 116 -15. 1.066-1430.
GRID 117 -10.606610.6066 60.
GRID 118 -15. 1.066-1460.
GRID 119 -10.606610.6066 90. 1
GRID 120 -15. 1.421-1490. 1
GRID 121 -10.6066-10.60660.0
GRID 122 0.0 -15. 0.0
GRID 123 -10.6066-10.606630.
GRID 124 -1.78-14-15. 30.
GRID 125 -10.6066-10.606660.
GRID 126 -3.2-14 -15. 60.
GRID 127 -10.6066-10.606690. 1
GRID 128 -4.97-14-15. 90. 1
GRID 129 10.6066 -10.60660.0
GRID 130 10.6066 -10.606630.
GRID 131 10.6066 -10.606660.
GRID 132 10.6066 -10.606690. 1
GRID 999 0.0 0.0 90.
$
RBE2 200 999 23456 110 111 112 119 120 +
+ 127 128 132
$
$QUAD4S REMOVED, SEE THE FILE ON THE DELIVERY MEDIA
$
$ THIS SECTION CONTAINS THE LOADS, CONSTRAINTS, AND CONTROL BULK DATA ENTRIES
$
$
MOMENT 1 999 0 1000. 0.0 0.0 1.
$
SPC1 1 123456 101 102 103 113
SPC1 1 123456 114 121 122 129
SPC1 1 12345 999
$
$
$ THIS SECTION CONTAINS THE PROPERTY AND MATERIAL BULK DATA ENTRIES
$
$
PSHELL 1 1 3. 1
$
Main Index
700 Nonlinear Users Guide
Rigid Elements
MAT1 1 250000. .3
ENDDATA
Grid point 999 is defined at the center of the free end to serve as the independent point. The , Z , R r , R , and R z
components of the grid points on the end of the tube are dependent degrees of freedom. To ensure that the axis of the
tube remains in the same position, an SPC is applied to all components of grid point 999 except in the R z direction,
which is the component about which the torque is applied. It is interesting to note that the CD field of grid point 999,
the independent point, is different than that of the CD field of the dependent point, and this is an acceptable modeling
technique. Furthermore, you should always use a rectangular coordinate system in the CD field for any grid point that
lies on the polar axis. In this example, grid point 999 lies on the Z-axis (see Figure 12-23); therefore, it should not use
coordinate system 1 for its CD field.
The displacement vector of the end grid points is shown in Figure 12-24. The q direction T2 is the same for each of the
end grid points as desired. The R-direction T1 is small but not exactly zero, indicating that the tube is permitted to
expand in the radial direction.
D I S P L A C E M E N T V E C T O R
POINT ID. TYPE T1 T2 T3 R1 R2 R3
110 G 1.487426E-20 2.526745E-04 .0 .0 .0 1.684497E-05
111 G 8.858868E-20 2.526745E-04 .0 .0 .0 1.684497E-05
112 G -4.178414E-21 2.526745E-04 .0 .0 .0 1.684497E-05
119 G -1.985002E-20 2.526745E-04 .0 .0 .0 1.684497E-05
120 G -4.204334E-20 2.526745E-04 .0 .0 .0 1.684497E-05
127 G 4.297897E-20 2.526745E-04 .0 .0 .0 1.684497E-05
128 G 8.858550E-21 2.526745E-04 .0 .0 .0 1.684497E-05
132 G -9.068584E-20 2.526745E-04 .0 .0 .0 1.684497E-05
Nonlinear RBE2
The nonlinear relation of an RBE2 can be expressed as follows:
xt = R r xt xr (12-16)
where x t is the coordinate of the tied node, x r is the coordinate of the reference node, and R r is the rotation matrix
of the retained node. The linearized relation between the incremental displacement of a tied node to a retained node
and its correction term is expressed as follows:
i i
u t = Su r + C n o n (12-17)
where the tying matrix S is derived from the linearized rigid body relation. For linear analysis, C n o n is zero. For large
i
displacement analysis, Su r may not match the nonlinear constraint exactly. Therefore, an error vector, C n o n , is
required to meet the constraint. It is defined as the difference between the expected coordinates ( x te ), using the rigid
body kinematics, and the current coordinates ( x t ) of the tied node.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 701
Rigid Elements
n+i n n+i 0 0
x te = xr + R r xt xr
(12-18)
n+i n+i
C n o n = x te xt
where n is increment number, i is iteration number, and zero is the original value.
When the rotation is finite, then the coordinate system attached to the tied node will be co-rotated according to the
rotation of the retained node. The degrees of freedom of the tied node will be assigned with this co-rotated system. As
an illustration, consider a slider link shown in Figure 12-25. In this case, a local transformation must be defined for the
tied node in which the x-direction is coincident with the line connecting the tied and the retained node. Then, an RBE2
is defined where y-translational degree of freedom is tied but not the x-translational and any of the rotation degrees of
freedom.
.
y x
Y E2 x
RB 2
RBE
As an example of the usage of RBE2 for defining a constraint, see Automated Bolt Modeling (Ch. 72) in the MSC
Nastran Demonstration Problems Manual.
1 2 3 4 5 6 7 8 9 10
RBE3 EID REFGRID REFC WT1 C1 G1,1 G1,2
G1,3 WT2 C2 G2,1 G2,2 -etc.- WT3 C3
G3,1 G3,2 -etc.- WT4 C4 G4,1 G4,2 -etc.-
UM GM1 CM1 GM2 CM2 GM3 CM3
GM4 CM4 GM5 CM5 -etc.-
ALPHA ALPHA
Main Index
702 Nonlinear Users Guide
Rigid Elements
Field Contents
EID Element identification number. Unique with respect to all elements. (0 < Integer < 100,000,000)
REFGRID Reference grid point identification number. (Integer > 0)
REFC Component numbers at the reference grid point. (Any of the integers 1 through 6 with no
embedded blanks.)
WTi Weighting factor for components of motion on the following entry at grid points Gi,j. (Real)
Ci Component numbers with weighting factor WTi at grid points Gi,j. (Any of the integers 1
through 6 with no embedded blanks.)
Gi,j Grid points with components Ci that have weighting factor WTi in the averaging equations.
(Integer > 0)
UM Indicates the start of the degrees of freedom belonging to the dependent degrees of freedom. The
default action is to assign only the components in REFC to the dependent degrees of freedom.
(Character)
GMi Identification numbers of grid points with degrees of freedom in the m-set. (Integer > 0)
CMi Component numbers of GMi to be assigned to the m-set. (Any of the Integers 1 through 6 with no
embedded blanks.)
ALPHA Indicates that the next number is the coefficient of thermal expansion. (Character)
ALPHA Thermal expansion coefficient. (Real > 0.0 or blank)
The manner in which the forces are distributed is analogous to the classical bolt pattern analysis. Consider the bolt
pattern shown in Figure 12-26 with a force and moment M acting at reference point A. The force and moment can be
transferred directly to the weighted center of gravity location along with the moment produced by the force offset.
FA FA
Reference Point
C.G.
MA
e M = MA + FA e
Figure 12-26 RBE3 Equivalent Force and Moment at the Reference Point
The force is distributed to the bolts proportional to the weighting factors. The moment is distributed as forces, which
are proportional to their distance from the center of gravity times their weighting factors, as shown in Figure 12-27.
The total force acting on the bolts is equal to the sum of the two forces. These results apply to both in-plane and
out-of-plane loadings.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 703
Rigid Elements
F1
F2
F5
F3 i
where F i = F A ------------
i
M
F4
F1
r1 F2
r5 r2 F3
M i r i
where F i = ------------------------------------------------------------------------------------------
M 2 2
1 r 1 + 2 r 2 + 3 r 2 + 3 r 3 + 5 r 5 2 2 2
r3
F5 r4
F4
where:
As an example, consider the cantilever plate modeled with a single CQUAD4 element shown in Figure 12-28. The plate
is subjected to nonuniform pressure represented by a resultant force acting at a distance of 10 mm from the center of
gravity location. The simplest way to apply the pressure is to use an RBE3 element to distribute the resultant load to
each of the four corner points.
Main Index
704 Nonlinear Users Guide
Rigid Elements
Pressure Resultant = 10 N
Y 4 3
100 mm
50 mm
1 2
40 mm
100 mm
The input file representing this example is shown in Listing 12-4. (See
MSC_DOC_DIR/doc/linstat/rbe3.dat) Grid point 99 is called the REFGRID and is the location where the
force is applied. This point is connected only to those degrees of freedom listed on the REFC field (the T3 component
in this example). The default action of this element is to place the REFC degrees of freedom in the m-set. The element
has provisions to place other degrees of freedom in the m-set instead. However, this is an advanced feature and is
beyond the scope of this users guide. The groups of connected grid points begin in field 5. For this example, the
connected grid points are the corner points.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 705
Rigid Elements
$
GRID 1 0. 0. 0.
GRID 2 100. 0. 0.
GRID 3 100. 100. 0.
GRID 4 0. 100. 0.
GRID 99 60. 50. 0.
$
PSHELL 1 4 10. 4
$
MAT1 4 4.E6 0.
$
CQUAD4 1 1 1 2 3 4
$
SPC1 1 123456 1 4
ENDDATA
The start of a group is indicated by a real number WTi, which is used as a weighting factor for the grid points in the
group. In this example, a simple distribution based only on the geometry of the RBE3 is desired so that a uniform
weight is applied to all points. The weighting factors are not required to add up to any specific value. For this example,
if the WT1 field is 4.0 instead of 1.0, the results will be the same.
The independent degrees of freedom for the group are listed in the Ci field. Note that all three translational degrees of
freedom are listed even though the REFC field does not include the T1 and T2 direction. All three translational degrees
of freedom in the Ci field are included because the degrees of freedom listed for all points must be adequate to define
the rigid body motion of the RBE3 element even when the element is not intended to carry loads in certain directions.
If any translational degrees of freedom are not included in C1 in this example, a fatal message is issued.
The element described by this RBE3 entry does not transmit forces in the T1 or T2 direction. The two reasons for this
are that the reference grid point is not connected in this direction and all of the connected points are in the same plane.
Note that the rotations are not used for the independent degrees of freedom. In general, it is recommended that only
the translational components be used for the independent degrees of freedom.
A selected portion of the output file produced by this example is shown in Figure 12-29.
Main Index
706 Nonlinear Users Guide
Rigid Elements
D I S P L A C E M E N T V E C T O R
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G .0 .0 .0 .0 .0 .0
2 G .0 .0 6.000000E-04 1.567011E-20 -9.000000E-06 .0
3 G .0 .0 6.000000E-04 3.282253E-21 -9.000000E-06 .0
4 G .0 .0 .0 .0 .0 .0
99 G .0 .0 3.600000E-04 .0 .0 .0
L O A D V E C T O R
POINT ID. TYPE T1 T2 T3 R1 R2 R3
99 G .0 .0 1.000000E+02 .0 .0 .0
F O R C E S O F S I N G L E - P O I N T C O N S T R A I N T
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G .0 .0 -5.000000E+01 7.275958E-12 3.000000E+03 .0
4 G .0 .0 -5.000000E+01 7.275958E-12 3.000000E+03 .0
G R I D P O I N T F O R C E B A L A N C E
POINT-ID ELEMENT-ID SOURCE T1 T2 T3 R1 R2 R3
1 F-OF-SPC .0 .0 -5.000000E+01 7.275958E-12 3.000000E+03 .0
1 1 QUAD4 .0 .0 3.000000E+01 -3.681428E-12 -3.000000E+03 .0
1 *TOTALS* .0 .0 -2.000000E+01 3.594530E-12 4.092726E-12 .0
2 1 QUAD4 .0 .0 -3.000000E+01 8.115214E-12 1.315632E-11 .0
2 *TOTALS* .0 .0 -3.000000E+01 8.115214E-12 1.315632E-11 .0
3 1 QUAD4 .0 .0 -3.000000E+01 7.376250E-12 -6.994877E-12 .0
3 *TOTALS* .0 .0 -3.000000E+01 7.376250E-12 -6.994877E-12 .0
4 F-OF-SPC .0 .0 -5.000000E+01 7.275958E-12 3.000000E+03 .0
4 1 QUAD4 .0 .0 3.000000E+01 -4.079332E-12 -3.000000E+03 .0
4 *TOTALS* .0 .0 -2.000000E+01 3.196626E-12 4.547474E-13 .0
99 APP-LOAD .0 .0 1.000000E+02 .0 .0 .0
99 *TOTALS* .0 .0 1.000000E+02 .0 .0 .0
The displacement of grid points 1 and 4 is zero due to the SPC applied to these points. The sum of the SPC forces at
these two grid points is equal to the load applied to the reference grid point. The load transmitted to the corner points
can be seen by inspecting the GPFORCE output. The force applied to the points due to the R-type elements and MPC
entries is not listed specifically in the GPFORCE output. These forces show up as unbalanced totals (which should
typically be equal to numeric zero). The forces applied to the corner grid points 1 through 4 are -20, -30, -30, and -20
N, respectively.
The most common usage of the RBE3 element is to transfer motion in such a way that all six degrees of freedom of the
reference point are connected. In this case, all six components are placed in the REFC field, and only components 123
are placed in the Ci field.
The load distributing capability of the RBE3 element makes it an ideal element to use to apply loads from a coarse
model (or hand calculation) onto a detailed model of a component. For example, the shear distribution on a cross
section is a function of the properties of that section. This shear loading may be applied to a cross section by performing
a calculation of the shear distribution based on unit loading and using an RBE3 element with appropriate weighting
factors for each grid point. In this manner, only one shear distribution need be calculated by hand. Since there are
usually multiple loading conditions to be considered in an analysis, they may be applied by defining different loads to
the dependent point on the RBE3 element.
The most common user error in RBE3 element specification results from placing 4, 5, or 6 in the Ci (independent
degrees of freedom) field in addition to the translation components. The rotations of the dependent point are fully
defined by the translational motion of the independent points. The ability to input 4, 5, or 6 in the Ci field is only for
special applications, such as when all of the connected points are colinear.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 707
Rigid Elements
Small checkout models are recommended whenever you are specifying elements with nonuniform weight factors,
asymmetric geometry or connected degrees of freedom, or irregular geometry. Using small checkout models is
especially necessary when the reference point is not near the center of the connected points.
The intended use of the RBE3 element is to transmit forces and moments from a reference point to several non-colinear
points. The rotation components 4, 5, and 6 should be placed in the Ci field only for special cases, such as when the
independent points are colinear.
As an example of the usage of RBE3 for defining a constraint, see Large Rotation Analysis of a Riveted Lap Joint
(Ch. 50) in the MSC Nastran Demonstration Problems Manual.
RJOINT Element
The RJOINT element defines a mechanical joint. The format for RJOINT bulk data entry in the QRG is as follows:
RJOINT
1 2 3 4 5 6 7 8 9 10
RJOINT EID GA GB CB
Field Contents
EID Element identification number. (Integer > 0)
GA, GB Grid point identification numbers. (Integer > 0)
CB Component numbers in the global coordinate system at GB. These degrees of freedom are
constrained to move with the same degrees of freedom at GA. (Integers 1 through 6 with no
embedded or blank.)
GA is the independent grid point and all six degrees of freedom are independent. GB is the dependent grid point. The
length between points GA and GB must be zero. Because its length is zero, the thermal load effect is not applicable.
If CB=123456 or is left blank, then the GB grid point is constrained to move with the GA grid point, and two the grid
points move as a single point. If any degree of freedom is released on CB, then RJOINT becomes a mechanical joint.
A mechanical joint is a mechanical system that has two bodies jointed at a point. The two bodies can rotate relatively
about one, two, or three axes of a local coordinate system at that point. For example:
Hinge A hinge is a mechanical joint that rotates freely about one axis about the local coordinate system. It
can be simulated by RJOINT with one rotational degree of freedom released; i.e., CB = 12356, 12346, or
12345.
Universal joint A universal joint is a mechanical joint that rotates freely in two axes. It can be simulated by
RJOINT with two rotational degrees of freedom released; i.e., CB = 1234, 1235, or 1236.
Spherical joint A spherical joint is a mechanical joint that rotates freely about all three axes. It can be
simulated by a zero length RJOINT with all rotational degrees of freedom released; i.e. CB = 123.
Main Index
708 Nonlinear Users Guide
Rigid Elements
Prismatic joint A prismatic joint is a mechanical system with two blocks that are constrained to have the
same rotations, but translate relative with each other along a local axis. It can be simulated by RJOINT with
one translational degree of freedom released; i.e., CB = 23456, 13456, 12456.
Cylindrical joint A cylindrical joint is a mechanical system that allows two grid points to have relative
translation along a moving axis and, at the same, have relative rotation about the same axis. It can be simulated
by RJOINT with one translational degree of freedom and one rotational degree of freedom released; i.e.,
CB = 2356, 1346, 1245.
Guidelines for the RJOINT element:
The theory for the RJOINT is formulated such that a consistent mechanical joint is created even if you request
different global coordinate systems at grid points GA and GB. If different global coordinate systems are used,
the degrees of freedom to be released are determined by the coordinate system at GB.
For linear or nonlinear static analysis, the joints must be constrained by other elements in the structural model.
Otherwise, a singular stiffness matrix may be produced, leading to failure during decomposition.
Lagrange Method
With the Lagrange method, the rigid elements become real finite elements, similar to, for example, a QUAD4
element. Instead of using MPC equations, the element stiffness matrix is computed for each rigid element. All the
limitations for the linear method are removed; i.e., the rigid elements
Include thermal load effect.
Include the differential stiffness.
Support for both the small rotation and the large rotation. Use the large rotation theory in the geometrical
nonlinear analysis (PARAM,LGDISP,1).
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 709
Rigid Elements
Take advantages of the sparse matrix algorithm if the augmented Lagrange multiplier method, defined below,
is used.
For each Lagrange rigid element, a number of Lagrange multiplier degrees of freedom are created internally by MSC
Nastran. For example, one to six Lagrange multiplier degrees of freedom are created for the RBAR and one Lagrange
multiplier degree of freedom is created for the RROD. For each Lagrange rigid element, the independent degrees of
freedom, the dependent degrees of freedom, and the Lagrange multiplier degrees of freedom are left in the solution set
(the l-set).
For the Lagrange Multiplier Method, the solution is obtained with the independent degrees of freedom, the dependent
degrees of freedom, and the Lagrange multipliers degrees of freedom left in the solution set. Thus, the sparse
characteristic of the stiffness matrix is maintained and sparse matrix algorithms can be used.
See Structural Elements (Ch. 3), Rigid Element Enhancements in the MSC Nastran Reference Manual for a theoretical
discussion of these methods and the use of rigid elements for large rotations.
Method Selection
RIDIG control case command in the QRG selects the type of rigid element. It has the following format:
LINEAR
RIGID = LAGR
LGELIM
LINEAR will select the linear rigid elements, LAGR will select the Lagrange rigid element with the Lagrange multiplier
method, and LGELIM will select the Lagrange element with the Lagrange elimination method (not used in Statics). If
the RIGID command does not exist in the user Case Control file, the linear rigid element will be used.
Lagrange Method Bulk Data Entry Differences
The R-type elements; RBAR, RBAR1, RJOINT, RBE1, RBE2, RBE3, RROD, RTRPLT, and RTRPLT1; can be used as
either the linear rigid element or the Lagrange element, as selected by the RIGID case control command. However, the
input rules are not the same for these two types of rigid elements. The following paragraphs discuss the differences.
Except for the RBE3 and RROD, the major difference in the input format between the linear rigid element and the
Lagrange rigid element is the selection of the independent degrees of freedom. This can be illustrated by the RBAR
bulk data entry:
Main Index
710 Nonlinear Users Guide
Rigid Elements
For the RBE3, the REFC degrees of freedom can be any combination of integers from 1 through 6 for the linear rigid
element. For the Lagrange rigid element, RFEC must be 123, 456, or 123456.
For the RROD element, you must select one dependent degree of freedom by inputting either CMA or CMB for the linear
rigid element. However, for the Lagrange rigid element, you can leave both fields blank and let MSC Nastran select
the best component as the dependent degree of freedom. In fact, this is the recommended method.
For all rigid elements, the ALPHA field is the thermal coefficient of expansion. For the Lagrange rigid elements, if
ALPHA is given and the thermal loads are requested by the TEMPERATURE (INITIAL) Case Control command and
TEMPERATURE (LOAD), the thermal load effect will be computed for the rigid elements. The temperature loads are
taken as the average temperature given by the independent grid point and the dependent grid point. For example, the
temperature load for the RBAR element is taken as the average temperature of grid points GA and GB. For the linear
rigid element, no temperature effect is computed and the ALPHA field is ignored.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 711
Patran Support
Patran Support
SOL 400 allows you to input kinematic constraints through various options that include multi-point constraints,
boundary conditions and support conditions.
Multi-Point Constraints
MPCs are special element types which define a rigorous behavior between several specified nodes. The following table
lists the MPC types which are supported for SOL 400.
MPC Types
Explicit RBE1
Rigid (Fixed) RBE2
Cyclic Symmetry RBE3
Sliding Surface RROD
RJOINT
RSPLINE
RSSCON
RBAR RTRPLT
RBAR1 RTRPLT1
Entry Description
MPC Defines a multipoint constraint equation.
Main Index
712 Nonlinear Users Guide
Patran Support
Entry Description
Dependent Terms Dependent terms define the fields for G1 and C1 on the MPC entry. Only one node and
degrees of freedom combination may be defined for any given explicit MPC. The A1 field
on the MPC entry is automatically set to -1.0.
Independent Terms Independent terms define the Gi, Ci, and Ai fields on the MPC entry, where i is greater than
one. As many coefficient, node, and degrees of freedom combinations as desired may be
defined.
Entry Description
RBE2 Defines a rigid body with independent degrees of freedom that are specified at a single grid point
and with dependent degrees of freedom that are specified at an arbitrary number of grid points.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 713
Patran Support
Entry Description
Dependent Terms Dependent terms define the GMi fields on the RBE2 entry. As many nodes as desired may be
selected as dependent terms.
Independent Terms Independent terms define the GN field on the RBE2 entry. Only one node may be selected.
Entry Description
Shell Nodes Dependent terms define the ESi fields on the RSSCON entry. One dependent node must be
selected for every two independent terms.
Solid Nodes Independent terms define the EA and EB field on the RSSCON entry. Two independent terms
are required.
Entry Description
MPC Defines a multipoint constraint equation.
Main Index
714 Nonlinear Users Guide
Patran Support
Entry Description
Dependent Region Specifies the dependent nodes on the sliding surface. The same number of unique nodes must
be specified in both regions.
Independent Region Specifies the independent nodes on the sliding surface. The same number of unique nodes
must be specified in both regions.
Entry Description
RBAR Defines a rigid bar with six degrees of freedom at each end.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 715
Patran Support
Entry Description
Dependent Terms Either one or two nodes may be defined as having dependent terms. The Nodes define the GA
and GB fields on the RBAR entry. The degrees of freedom define the CMA and CMB fields.
Independent Terms Either one or two nodes may be defined as having independent terms.The Nodes define the
GA and GB fields on the RBAR entry.The degrees of freedom define the CNA and CNB fields.
Entry Description
RBE2 Defines a rigid body with independent degrees of freedom that are specified at a single grid
point and with dependent degrees of freedom that are specified at an arbitrary number of grid
points.
Main Index
716 Nonlinear Users Guide
Patran Support
Entry Description
Dependent Terms Dependent terms define the GMi and CM fields on the RBE2 entry. As many nodes as desired
may be selected as dependent terms.
Independent Terms Independent terms define the GN field on the RBE2 entry. Only one node may be selected.
Entry Description
RBE3 Defines the motion at a reference grid point as the weighted average of the motions at a set
of other grid points.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 717
Patran Support
Entry Description
Dependent Terms Dependent terms define the GMi and CMi fields on the RBE3 entry. The first dependent term
will be treated as the reference node, REFGRID and REFC. The rest of the dependent terms
become the GMi and CMi components.
Independent Terms Independent terms define the Gi, j, Ci, and WTi fields on the RBE3 entry.
Entry Description
RROD Defines a pin-ended element that is rigid in translation.
Main Index
718 Nonlinear Users Guide
Patran Support
Entry Description
Dependent Terms Dependent terms define the GB and CMB on the RROD entry. Only one translational degree
of freedom may be referenced for this entry.
Independent Terms Independent terms define the GA field on the RROD entry. The CMA field is left blank.
Entry Description
RTRPLT Defines a rigid triangular plate.
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 719
Patran Support
Entry Description
Dependent Terms Dependent terms define the GA, GB, GC, CMA, CMB, and CMC fields of the RTRPLT entry.
Independent Terms The total number of nodes referenced in both the dependent terms and the independent terms
must equal three. There must be exactly six independent degrees of freedom, and they must
be capable of describing rigid body motion. Defines the GA, GB, GC, CNA, CNB, and CNC
fields of the RTRPLT entry.
Main Index
720 Nonlinear Users Guide
Pin Code for Beam Elements
2
10 20 4
10
20
1 1
3 3
Main Index
Chapter 12: Rigid Elements, Connectors, and Multi-Point Constraints 721
Constraint Conflict
Constraint Conflict
When multiple constraints are associated with a grid, it is possible that the constraints conflict. This may result in either
an overconstrained or a singluar system, in which case, the analysis will terminate. The AUTOMSET parameter
automatically re-writes the constraint equations such that a tied degree of freedom is not used as a retained degree of
freedom in another constraint equation. It also modifies constraint equations where a constrained degree of freedom is
used in a prescribed boundary condition.
Main Index
722 Nonlinear Users Guide
Constraint Conflict
Main Index
Chapter 13: Advanced Topics
Nonlinear Users Guide
13 Advanced Topics
Superelements and Modal Neutral Files 724
Fracture Mechanics 725
Inertia Relief 734
Large Displacement Grid Point Weight Generation (GPWG) 739
BSQUEAL 740
RESTART 745
User Defined Services and Subroutines 750
References 757
Main Index
724 Nonlinear Users Guide
Superelements and Modal Neutral Files
External Superelements
External superelements are not supported.
Main Index
Chapter 13: Advanced Topics 725
Fracture Mechanics
Fracture Mechanics
The fracture mechanics capabilities in MSC Nastran covers the evaluation of energy release rate and J-integral
evaluation.
G = Gc (13-1)
G is defined as
d
G = -------- (13-2)
da
where is the strain energy and a is the crack length. G depends on the geometry of the structure and the current
loading. G c is called the fracture toughness of the material. It is a material property which is determined from
experiments. Note that the energy release rate is not a time derivative but a rate of change in potential energy with crack
length. An important feature of equation (13-1) is that it can be used as a fracture criterion; a crack starts to grow when
G reaches the critical value G c .
The stress and strain fields near the tip of a crack are singular for a linear elastic material model. The stresses and strains
have the principal form
K
= ------ f
r
(13-3)
K
= ------ g
r
in a polar coordinate system centered at the crack tip. Thus, a linear elastic material is said to have a 1 r singularity
near a crack tip. It is easy to demonstrate that in both Griffiths and Irwins considerations, the elastic energy release
rate is determined by a single parameter: the strength of the singularity in the elastic stress field at the crack tip. This
is the so-called stress intensity factor, and is usually denoted by capital K . The magnitude of K depends on the crack
length, the distribution and intensity of applied loads, and the geometry of the structure. Crack propagation occurs
Main Index
726 Nonlinear Users Guide
Fracture Mechanics
when any combination of these factors causes a stress intensity factor K to be equal to or greater than the
experimentally determined material property K c , which is equivalent to equation (13-1). Hence, the objective of linear
fracture mechanics calculations is to determine the value of K .
There are three possible modes of crack extension in linear elastic fracture mechanics: the opening mode, sliding mode,
and tearing mode (see Figure 13-1).
The opening mode (see Figure 13-1a), Mode I, is characterized by the symmetric separation of the crack surfaces with
respect to the plane, prior to extension (symmetric with respect to the X-Y and X-Z planes).
The sliding mode, Mode II, is characterized by displacements in which the crack surfaces slide over one another
perpendicular to the leading edge of the crack (symmetric with respect to the X-Y plane and skew-symmetric with
respect to the X-Z plane).
The tearing mode, Mode III, finds the crack surfaces sliding with respect to one another parallel to the leading edge
(skew-symmetric with respect to the X-Y and X-Z planes).
y
y
x
x
z
z
x
z
It is customary to associate a stress intensity factor with each of these mode: K I , K II , and K II I . There is also an
associated fracture toughness associated with each mode: K Ic , K II c , and K II I c . The most critical mode is usually
mode I and in many cases the other modes are not considered. The connection between the energy release rate and the
stress intensity factors is given by
Main Index
Chapter 13: Advanced Topics 727
Fracture Mechanics
2 2
KI KI I 1 + 2
G = ------- + -------- + ------------- K II I (13-4)
E' E' E
where
E
E' = --------------- for plane strain. (13-6)
1 2
SOL 400 uses the so-called J-integral for evaluating the energy release rate, see below. The J-integral is similar to G
but is more general and is also used for nonlinear applications. J is equivalent to G when a linear elastic material model
is used.
- 1/n
----- = ----- (13-7)
0 0
it can be shown that the singularities in the strain and the stresses at the crack tip are of the form
= f r n n + 1
(13-8)
= g r 1 n + 1
If n approaches infinity, the material behavior becomes perfectly plastic and the singularity in the stresses vanishes.
The singularity in the strains, however, takes the form of
= f r 1 (13-9)
It has not been possible to establish that the strength of the singularity is the only factor that influences initiation of
crack propagation for nonlinear situations. In fact, it is doubted that initiation of crack propagation is dependent on
only a single factor. The J-integral probably offers the best chance to have a single parameter to relate to the initiation
of crack propagation.
Main Index
728 Nonlinear Users Guide
Fracture Mechanics
The J-integral was introduced by Rice as a path-independent contour integral for the analysis of cracks. As previously
mentioned, it is equivalent to the energy release rate for a linear elastic material model. It is defined in two dimensions
as
u j
J = W + T n 1 i j n i -------
x 1
- d (13-10)
where W is the strain energy density, T is the kinetic energy density, ij is the stress tensor and u i is the displacement
vector. The x 1 direction is the same as the x direction in the local crack tip system in Figure 13-2. The integration path
is a curve surrounding the crack tip, see Figure 13-2.The J-integral is independent of the path as long as it starts
and ends at the two sides of the crack face and no other singularities are present within the path. This is an important
feature for the numerical evaluation since the integral can be evaluated using results away from the crack tip.
ni
y
Main Index
Chapter 13: Advanced Topics 729
Fracture Mechanics
a
u
Model A Model B
Here, F is the force (obtained from Model A) that keeps the crack together, and the crack opening, u , is obtained from
Model B. In order to obtain these quantities, we would need to perform two analyses and this method is often referred
to as Crack Closure Technique (CCT). In the virtual crack closure technique, we only do the analysis with a closed
crack (Model A) and use the opening displacement at the closest nodes to the crack tip. Figure 13-4 shows the case of
pure mode I. The other modes are treated similarly and separately. The displacements and reactions are transformed
into the local crack tip system for this evaluation.
a
F
u
With x. y, and z denoting the coordinate directions in the local crack tip system (see Figure 13-1), we obtain:
Fy uy Fx ux Fz uz
G I = ----------- G I I = ----------- G I II = ----------- (13-13)
2a 2a 2a
and the total energy release rate as:
G tot = G I + G I I + G I II (13-14)
Main Index
730 Nonlinear Users Guide
Fracture Mechanics
For higher-order elements, we need to include the contributions from the midside nodes (see Figure 13-5).
F1 u1 + F2 u2
G I = -------------------------------- (13-15)
2a
a
F1
F2
u1
u2
For the case that the midside nodes are not at the middle of the element edges (for example, using the 1/4 point position
for increased accuracy, the displacements for these nodes are interpolated to the appropriate locations.
For 3-D solids, we have a situation as shown in Figure 13-6. The situation is similar to the 2-D case, and the evaluation
is done separately for each node along the crack front. The area is given by the shaded part in Figure 13-6.
For the case of higher-order elements, we obtain the following by using the notation in Figure 13-7.
1
F 1 u 1 + F 2 u 2 + --- F 3 u 3 + F 4 u 4
2
G = -------------------------------------------------------------------------------- (13-16)
2a
u is Crack Opening
Displacement At Crack Face
Crack Front
Main Index
Chapter 13: Advanced Topics 731
Fracture Mechanics
u1 u3
u4
F1 F3
u2
F4 F2
The above figures show a regular mesh of hexahedral elements. While a mesh designed like this is advantageous for
accuracy, it is not strictly necessary, and it is also possible to use a general tetrahedral mesh. The program will find the
appropriate nodes to use for the forces and crack opening displacement and calculate an area a.
The definition of the data involved in the VCCT calculation is done automatically. The user only specifies the crack
tip or crack front. In order to check what the program finds there is debug option available.
Symmetry is automatically detected and accounted for. The symmetry condition can be enforced by boundary
conditions or by rigid contact.
The mesh in the crack region can contain MPCs or contact. The program automatically detects if the crack tip node is
connected to a node, and the node it is connected to is considered part of the crack tip. The supported connections are
glued contact, MPC, RBE2, and RROD. For the case that the connection is done node-to-node, it does not matter if the
tied or retained node of the constraint is selected as the crack node. With glued contact, it is not necessary to connect
the parts node-to-node. If the meshes do not match up, the crack will be treated similarly as a case of symmetry. This
non-matching glued case is automatically detected and for this case, it is necessary to select the touching node as the
crack tip node.
The calculation of VCCT is done in the current geometry in case of large deformations. The updated crack coordinate
system is calculated at each increment and the calculations are done in this system. Thus, arbitrary rotations and
deformations are allowed.
Figure 13-8 shows some examples of supported crack configurations. There is a large flexibility in how parts can be
tied or glued together to form a crack.
Main Index
732 Nonlinear Users Guide
Fracture Mechanics
Crack Tip
Crack Front
Shell Glued to Solid (Line Crack) Shell Glued to Solid (Surface Crack)
As an example of the usage of VCCT for determining the stress intensity factor, see Composite Fracture and
Delamination (Ch. 20) in the MSC Nastran Demonstration Problems Manual.
K I t = R I D a T B a 0 (13-17)
where K I t is the dynamic stress intensity factor for mode I, a is the crack velocity, and R I D is the dynamic crack
propagation toughness, which is assumed to be a material parameter that in general depends on crack velocity a ,
temperature T , and specimen thickness B .
The dynamic stress intensity factor depends on crack length ( a ), applied loading ( ), time ( t ), specimen dimensions
( D ), temperature ( T ), and initial stress fields ( i ) caused by residual stresses or by an initial strain field. The
prediction of the crack propagation history and crack arrest event demands complete knowledge of the R I D vs. a
relation. The Dynamic Fracture Methodology procedure consists of the following two phases:
1. Generation phase in this phase, a crack arrest experiment is performed yielding a crack propagation-versus-
time curve. In addition, a numerical simulation of the experiment is carried out by using the measured crack
propagation curve. This is used as input for the numerical model. This allows the calculation of dynamic stress
intensity factors as a function of time. Combination of the latter relation with the measured crack propagation
curve results in a curve, which can be considered the dynamic crack propagation toughness-versus-crack
velocity relation.
Main Index
Chapter 13: Advanced Topics 733
Fracture Mechanics
2. Application phase in order to predict the crack growth and possible crack arrest point in a structural
component, the inverse problem is solved. Now, the actual stress intensity factors are calculated for the
structural component, that is subjected to a particular loading history, by means of a dynamic finite element
analysis. These calculated values are compared to the fracture toughness curve obtained during the generation
phase, equation (13-17), and from this the crack growth is predicted.
Main Index
734 Nonlinear Users Guide
Inertia Relief
Inertia Relief
Inertia relief is a technique to simulate self-equilibrating quasi dynamic loading in static analyses.
Format:
QLINEAR
IRLOAD =
NONE
Example:
IRLOAD=QLINEAR
QLINEAR is to activate Inertia Load Calculation with small displacement (Quasi-Linear) in SOL 400. NONE means
No Inertia Relief, which is default.
Main Index
Chapter 13: Advanced Topics 735
Inertia Relief
Main Index
736 Nonlinear Users Guide
Inertia Relief
Main Index
Chapter 13: Advanced Topics 737
Example of Inertia Relief
ID MSC, af_ir
SOL 400
CEND
ANALYSIS = NLSTAT
IRLOAD=QLINEAR $ this card is removed in af_noir.dat
SPC=20
LOAD = 400
DISP=all
STRESS=all
SPCF=all
BEGIN BULK
$ Determinate-SPC
SPC1 20 2 1
SPC1 20 23 101
SPC1 20 123 104
Main Index
738 Nonlinear Users Guide
Example of Inertia Relief
The results of DISP, LOAD, SPCF are showed in the following figures. Note that there are 15 orders of magnitude
difference on the spcforces between af_ir and af_noir.
Main Index
Chapter 13: Advanced Topics 739
Large Displacement Grid Point Weight Generation (GPWG)
Main Index
740 Nonlinear Users Guide
BSQUEAL
BSQUEAL
The BSQUEAL bulk data entry is available for brake squeal simulation. It is now possible to determine the unstable
brake squeal roots using MSC Nastrans complex eigenvalue solver and unsymmetric friction stiffness matrices either
for an undeformed structure or after a nonlinear loading. Unstable roots indicate potential brake squeal. They are
designated by positive real roots and negative damping in the .f06 output file.
Input File
ID MSC, NLBSQL01 $
SOL 400
CEND
ECHO=SORT( EXCEPT GRID, CHEXA )
BCONTACT = 0
SUBCASE 1
SUBTITLE=CASE1
STEP 1
LABEL=Nonlinear Static Step, Loading + Contact
NLPARM = 1
BCONTACT = 1
Main Index
Chapter 13: Advanced Topics 741
BSQUEAL
BOUTPUT=ALL
BSQUEAL = 988
SPC = 2
LOAD = 2
CMETHOD=1
METHOD =2 $ Modal Approach
DISP(PLOT)=ALL
AUTOSPC(NOPRINT)=YES
RESVEC=NO
BEGIN BULK
BCPARA 0 NLGLUE 1
PARAM LGDISP 1
NLPARM 1 FNT PV NO
$-------2-------3-------4-------5-------6-------7-------8-------9-------0----
BCTABLE 1 4
SLAVE 9 0. 0. 1. 0. 0 0.
1 2 0
MASTERS 8
SLAVE 10 0. 0. 1. 0. 0 0.
1 2 0
MASTERS 8
SLAVE 11 0. 0. 0. 0. 1 0.
1 2 0
MASTERS 9
SLAVE 12 0. 0. 0. 0. 1 0.
1 2 0
MASTERS 10
EIGC 1 CLAN 20
EIGRL 2 15
$-------2-------3-------4-------5-------6-------7-------8-------9-------0----
$ ID OMETH AVSTIF GLUE ICORD BSONLY
BSQUEAL 988 0.0 1.e+4 YES
0.0 0.0 1.0 0.0 0.0 0.0
PSOLID 1 1 0
$ Pset: "disk" will be imported as: "psolid.1"
CHEXA 1 1 1 2 9 8 1001 1002
1009 1008
CHEXA 2 1 2 3 10 9 1002 1003
1010 1009
CHEXA 3 1 3 4 11 10 1003 1004
1011 1010
.
.
.
$ Elements and Element Properties for region : pad1
PSOLID 2 2 0
$ Pset: "pad1" will be imported as: "psolid.2"
CHEXA 1004 2 2004 2005 2012 2011 3004 3005
3012 3011
CHEXA 1005 2 2005 2006 2013 2012 3005 3006
3013 3012
.
.
.
CHEXA 1030 2 2034 2035 2042 2041 3034 3035
3042 3041
$ Elements and Element Properties for region : pad2
PSOLID 3 2 0
$ Pset: "pad2" will be imported as: "psolid.3"
CHEXA 1031 3 4000 4001 4005 4004 4024 4025
Main Index
742 Nonlinear Users Guide
BSQUEAL
4029 4028
CHEXA 1032 3 4001 4002 4006 4005 4025 4026
4030 4029
.
.
.
CHEXA 1045 3 4018 4019 4023 4022 4042 4043
4047 4046
$ Elements and Element Properties for region : piston
PSOLID 4 2 0
$ Pset: "piston" will be imported as: "psolid.4"
CHEXA 1046 4 5007 5008 5002 5005 5012 5009
5010 5011
CHEXA 1047 4 5004 5001 5008 5007 5014 5013
5009 5012
.
.
.
CHEXA 1053 4 10007 10008 10006 10004 10016 10013
10012 10018
MAT1 1 21000. 8076.92 .3 7.8-6
MAT1 2 2000. 769.231 .3 2.3-6
$ Nodes of the Entire Model
GRID 1 40. 0. 0. 1
GRID 2 50. 0. 0. 1
GRID 3 60. 0. 0. 1
.
.
.
GRID 10018 89.08 31.9591 -10.
$ Loads for Load Case : case1
SPCADD 2 1 3 4
LOAD 2 1. 1. 1
LOAD 4 1.0-8 1. 1
$ Contraints in Cylindrical Coord. 1
$ On one edge
$ SID C G1 G2 .......
SPC1 1 13 1 8 15 22 29 36
43 50 57 64 71 78 85 92
99 106 113 120 127 134 141 148
155 162 169 176 183 190 197 204
211 218 225 232 239 246
$ Displacement Constraints of Load Set : pad_fixed
SPC1 3 12 3004 3005 3006 3007 3039 3040
3041 3042
$ Displacement Constraints of Load Set : pad_fixed2
SPC1 4 12 4000 THRU 4023
$ Deform Body Contact LBC set: disk
BCBODY 8 3D DEFORM 8 0
BSURF 8 1 2 3 4 5 6 7
8 9 10 11 12 13 14 15
16 17 18 19 20 21 22 23
.
.
.
208 209 210 211 212 213 214 215
216
$ Deform Body Contact LBC set: pad1
BCBODY 9 3D DEFORM 9 0
BSURF 9 1004 1005 1006 1010 1011 1012 1016
Main Index
Chapter 13: Advanced Topics 743
BSQUEAL
C O M P L E X E I G E N V A L U E S U M M A R Y
ROOT EXTRACTION EIGENVALUE FREQUENCY DAMPING
NO. ORDER (REAL) (IMAG) (CYCLES) COEFFICIENT
1 1 0.0 0.0 0.0 0.0
2 2 0.0 5.267539E+01 8.383549E+00 0.0
3 3 0.0 5.758587E+01 9.165074E+00 0.0
4 4 0.0 8.833070E+01 1.405827E+01 0.0
5 5 0.0 1.052037E+02 1.674369E+01 0.0
6 6 0.0 1.070820E+02 1.704263E+01 0.0
7 8 -1.901260E+00 1.953634E+02 3.109305E+01 1.946383E-02
8 7 1.901260E+00 1.953634E+02 3.109305E+01 -1.946383E-02
9 10 -1.548788E+00 3.172325E+02 5.048912E+01 9.764370E-03
10 9 1.548788E+00 3.172325E+02 5.048912E+01 -9.764370E-03
11 15 0.0 3.943055E+02 6.275567E+01 0.0
12 14 0.0 4.006665E+02 6.376806E+01 0.0
13 13 0.0 4.113044E+02 6.546112E+01 0.0
14 12 0.0 4.669658E+02 7.431991E+01 0.0
15 11 0.0 4.695969E+02 7.473867E+01 0.0
Main Index
744 Nonlinear Users Guide
BSQUEAL
Brake squeal can also be executed in the linear perturbation by using ANALYSIS=MCEIG. As an example of the usage
of BSQUEAL for calculating the unstable frequencies of a braking system, see Brake Squeal Analysis (Ch. 39) in the
MSC Nastran Demonstration Problems Manual.
Main Index
Chapter 13: Advanced Topics 745
RESTART
RESTART
The purpose of a nonlinear restart is to allow the user to use the material and geometrical properties of a previously
converged solution as a new starting point to continue the analysis. This is useful when the user want to change the
loading sequence, the solution criteria, or to extend the analysis.
For SOL 400, a user-friendly restart procedure has been implemented. For restarts, the following principles are noted:
The restart must be continued at a previous converged solution point in a nonlinear static analysis or nonlinear
transient analysis by specifying a SUBCASE, STEP, and/or TIME (LOADFAC). This is accomplished by using
the NLRESTART case control command. Note that restart is not supported in linear analysis or linear
perturbation analysis.
When the cold start is ANALYSIS=NLSTAT, it can be restarted at any user-specified output load increment
(controlled by NOUT in NLPARM bulk data entry).
When the cold start is ANALYSIS=NLTRAN, it must be restarted from a saved or check-pointed time step.
The checkpoint times are dependent on DT and NO values on the TESTEPNL bulk data entry, and PARAM,
NLPACK of the cold start run. The checkpoint times are integer multiples of (DT x NO) x NLPACK. For
example, if DT=0.001 second, NO=10, and NLPACK=100, the possible times that can be used for restart are
at 1.0, 2.0, etc. If a requested restart time does not match a checkpoint time, the closest checkpoint time is
used. The same logic can also used in the NLSTEP bulk data entry.
The geometry and the initial material properties of the structural model cannot be modified. This is obvious
because any modification to the geometry or the initial material properties would invalidate the previous
analysis and require the nonlinear solution to start from the very beginning. In such cases, it is simpler to
initiate another cold start.
Performing restart is described in the following subsections.
Main Index
746 Nonlinear Users Guide
RESTART
STEP 1
LOAD = 10
STEP 2
LOAD = 20
STEP 3
LOAD = 30
In the above example, the first step through time 0.3 of the second step has been previously executed. The restart
execution begins with time 0.3 of the second step, and continues through the end of the third step. If time 0.3 is not a
restart point saved by NLPACK on the cold start, SOL 400 will search for the nearest restart point on the data base and
use that point to begin the restart. For restart, the case control file structure for SUBCASE and STEP commands must
be the same as the cold start up to the restart point. After the restart point, the user may modify the case control file
structure for the SUBCASE and STEP commands. For example, in the above example, steps 1 and 2 must exist in the
cold start. However, step 3 may or may not exist in the cold start.
The following case control commands may be modified in a nonlinear restart:
Boundary conditions such as MPC and SPC
Nonlinear solution control, NLSTEP, NLPARM, and TSTEPNL
The LOAD requests
Output request such as DISP and NLSTRESS
The analysis type, ANALYSIS
Depending on the option selected with the NLRESTART command, the nonlinear restart may be logically divided into
three types: a case restart, a step restart, or a time restart:
The case restart begins the execution with a SUBCASE. All five types of modification described above are
legal for a case restart.
The step restart begins the execution with a STEP, which may be a new step or a previously executed step.
Although boundary condition and analysis type modifications are allowed, the user has the responsibility to
determine whether they are meaningful. Special attention should be given to the analysis type modification; it
may not be meaningful in many situations and, thus, leads to erroneous results.
The time restart begins execution with a user-specified TIME. For a time restart, the user should not modify the
analysis type, boundary conditions, or load requests. The user needs to exercise discretion when attempting
other types of modification at this level. Also, in order to perform this type of restart, the specified TIME must
be at the NLPACK data group boundary. If it is not, SOL 400 searches for the nearest data boundary and uses
this boundary as the restart point.
Main Index
Chapter 13: Advanced Topics 747
RESTART
As an alternative method using the NLIC case control command, the following demonstrates an initial analysis with a
SUBCASE with three steps followed by a restart analysis where the simulation is started after Step 1.
ID MSC, EX04 $
SOL 400 $
CEND
TITLE= ELASTIC-PLASTIC STATIC & TRANSIENT RESPONSE, EX04
SUBTI= INITIAL ACCELERATION COMPUTED - PARAM,ICOPT,0
SET 1 = 1111
SET 2 = 100
DISP = 1
VELO = 1
ACCE = 1
OLOAD = 1
$ STRESS(PLOT) = 2
SUBCASE 1130
step 1
LABEL=UNIAXIAL TENSION (LOADING)
ANALYSIS=NLSTAT
SPC=100
LOAD=1130
NLPARM = 1
step 2
LABEL=UNIAXIAL TENSION (UNLOADING)
ANALYSIS=NLSTAT
SPC=100
NLPARM = 1
step 10
LABEL=I.C. FROM THE FIRST NLSTAT STEP(50%) - UNBALANCED CASE (NLIC)
ANALYSIS=NLTRAN
NLIC STEP 1 LOADFAC 0.5
SPC=100
DLOAD=2130
TSTEPNL=10
param,icopt,0
BEGIN BULK
PARAm,LGDISP,1
PARAM,W4,1.0
$
NLPARM 1 4 AUTO ALL
1.-6
TSTEPNL 10 2000 0.001 AUTO
$
MAT1 1 30.0+6 11.5+6 0.3 7.332-2 0.01
PSOLID 1 1
$
SPC1 100 123456 1000
SPC1 100 1 1010
SPC1 100 2 1001
Main Index
748 Nonlinear Users Guide
RESTART
ASSIGN RSFILE='DBSDIR:ex04.MASTER' $
RESTART LOGICAL=RSFILE $
$
ID MSC, EX05 $
SOL 400 $
CEND
TITLE= ELASTIC-PLASTIC STATIC & TRANSIENT RESPONSE, EX05
SUBTI= NO INITIAL ACCELERATION COMPUTED - PARAM,ICOPT1
SET 1 = 1111
SET 2 = 100
DISP = 1
VELO = 1
ACCE = 1
OLOAD = 1
$ STRESS = 2
NLRESTART
SUBCASE 1130
step 1
LABEL=UNIAXIAL TENSION (LOADING)
ANALYSIS=NLSTAT
SPC=100
LOAD=1130
NLPARM = 1
step 2
LABEL=UNIAXIAL TENSION (UNLOADING)
ANALYSIS=NLSTAT
SPC=100
NLPARM = 1
step 10
LABEL=I.C. FROM THE FIRST NLSTAT STEP - UNBALANCED CASE (NLIC)
ANALYSIS=NLTRAN
NLIC STEP 1 LOADFAC 0.5
SPC=100
Main Index
Chapter 13: Advanced Topics 749
RESTART
DLOAD=2130
TSTEPNL=10
param,icopt,0
SUBCASE 1131
LABEL=I.C. FROM THE 1st NLSTAT STEP OF PREVIOUS SUBCASE (NLIC)
ANALYSIS=NLTRAN
NLIC SUBCASE 1130 STEP 1
SPC=100
DLOAD=2130
TSTEPNL=10
param,icopt,1
BEGIN BULK
ENDDATA
Main Index
750 Nonlinear Users Guide
User Defined Services and Subroutines
Main Index
Chapter 13: Advanced Topics 751
User Defined Services and Subroutines
The call lists of these subroutines are defined in the MSC Nastran User Defined Services Users Guide.
The options ELMUDS, MATUDS, and BCONUDS allow you to define integer, real, and character data through the bulk
data which is passed into the user subroutines that you are writing.
These options also specify a GROUP ID. This group id is then used to reference the CONNECT SERVICE statement
of the File Management Section (FMS). A CONNECT SERVICE is required for each copy of a user subroutine. For
example, if one had three different material models defined by UMAT, one would have three different CONNECT
SERVICE. It is also possible to have a single UMAT with conditional logic to control which material is used, in which
case a single CONNECT SERVICE is required.
This capability to reference multiple copies of the same use allows this capability to be used in a multi-user
environment, but leads to extra complexity. For additional information refer to the MSC Nastran Quick Reference
Guide, MSC Nastran User Defined Services Users Guide and the SCA Framework Users Guide. For simplification
the remainder of this section will be focused on activating a user subroutine for the most common cases, where a single
file can contain all the user subroutines and will be associated with the current simulation.
Main Index
752 Nonlinear Users Guide
User Defined Services and Subroutines
with each layer point. Note that the first state variable is temperature. Each state variable has its default name as SVi
with the i being the index number of the state variable. For example, the third state variable has its default name SV3.
You can define another name for state variable using the UDSESV entry. These state variables will be passed to material
UDS as arguments when the UDS gets called.
One special note is for the first state variable, it is reserved for temperature. The temperature is passed as the first
element of the state variable array in UDS. User defined state variables are available from the second position in the
state variable array.
Format:
UDSESV NSTATS
SV2 SV2_NAME SV3 SV3_NAME SV4 SV4_NAME SV5 SV5_NAME
SV6 SV6_NAME .etc.
Example:
UDSESV 3
SV2 VAR2 SV3 VAR3
Field Contents
NSTATS The number of user defined state variables. (Integer >= 1)
SVi The default nominal name of state variable (CHARACTER, i >= 1, where i is the index
number of the state variable)
SVi_NAME The state variable name defined by user (CHARACTER, Default = SVi, where i is the
index number of the state variable)
Remarks:
1. This is a global entry that defines user state variables for material user subroutines. The temperature will always
be passed to material use subroutine as the first state variable; its name should not be redefined in this entry.
2. If a state variable is not given a name, SVi will be used as its name. The number i is the index number of the
state variable.
3. For output, either state variable names given in UDSESV or default SVi names can be used in NLOUT entry
in case control. The state variables names will be used as keywords for output selection.
4. The 1st state variable is always temperature. The remaining user defined state variables are defined and used
only by user, MSC Nastran will not use them.
GENUDS
The GENUDS bulk data entry is to specify SCA service that implements the SCAIMDSolverRuntimeInfo interface.
The user supplied input data for the notify method is also defined in this entry. When the notify method is called, the
input data will be passed as arguments to this method.
Main Index
Chapter 13: Advanced Topics 753
User Defined Services and Subroutines
Format:
GENUDS SRV_ID
INT IDATA1 IDATA2 IDATA3 IDATA4 IDATA5 IDATA6 IDATA7
IDATA8 IDATA9 IDATAn
REAL RDATA1 RDATA2 RDATA3 RDATA4 RDATA5 RDATA6 RDATA7
RDATA8 RDATA9 RDATAn
CHAR CDATA1 CDATA2 CDATAn
Example:
GENUDS MY_SRV
INT 1 2 100
Field Contents
SRV_ID The service identifier used in the CONNECT SERVICE statement. (Character, no default)
INT Keyword indicating that the following data is integer. (Character)
IDATAi User supplied integer data. (Integer, no default)
REAL Keyword indicating the following data is real. (Character)
RDATAi User supplied real data. (Real, no default)
CHAR Keyword indicating the following data is character. (Character)
CDATAi User supplied character data. (Character, no default)
Remarks:
1. The SER_ID is the service identifier of SCA service in the CONNECT SERVICE statement. The SCA service
should have implemented the RuntimeInfo interface.
2. A CDATAi entry cannot be the character INT, REAL, or CHAR.
General Parameters
The functions and keywords are used to get model, machine and analysis procedure information, the available
keywords in this category are:
SUBCASE_NUMBER
Main Index
754 Nonlinear Users Guide
User Defined Services and Subroutines
STEP_NUMBER
INCREMENT_NUMBER
SUB_INCREMENT_NUMBER (if applicable)
ITERATION_NUMBER
CURRENT_TIME
INCREMENTAL_TIME
TIME_OF_PREVIOUS_STEP
TIME_OF_PREVIOUS_INCREMENT
FRACTIN_OF_STEP_COMPLETED
LARGE_DISP_FLAG
JOB_NAME
JOB_DIRECTORY
WORKING_DIRECTORY
SCRATCH_DIRECTORY
NUM_PROCS
NUM_CPUS
Element Parameters
The functions and keywords are used to get element related data, the available keywords in this category are:
ELEMENT_TYPE
DIRECT_STRESS_QUANTITIES
SHEAR_STRESS_QUANTITIES
NODES_OF_THE_ELEMENT
INTEGRATION_POINTS_OF_THE_ELEMENT
MATERIAL_ID_FOR_THE_ELEMENT
ELEMENT_CLASS
MAJOR_ENGINEERING_STRAIN
MINOR_ENGINEERING_STRAIN
CURRENT_VOLUME
ORIGINAL_VOLUME
TOTAL_TEMPERATURE
INCREMENTAL_TEMPERATURE
EQUIVALENT_VON_MISES_STRESS
EQUIVALENT_STRESS_YIELD_STRESS_RATIO
Main Index
Chapter 13: Advanced Topics 755
User Defined Services and Subroutines
EQUIVALENT_ELASTIC_STRAIN
EQUIVALENT_CREEP_STRAIN
TOTAL_STRAIN_ENERGY_DENSITY
ELASTIC_STRAIN_ENERGY_DENSITY
PLASTIC_STRAIN_ENERGY_DENSITY
GASKET_PRESSURE
GASKET_CLOSURE
PLASTIC_GASKET_CLOSURE
FAILURE_INDEX
TOTAL_VALUE_OF_FIRST_STATE_VARIABLE
TOTAL_VALUE_OF_SECOND_STATE_VARIABLE
TOTAL_VALUE_OF_THRID_STATE_VARIABLE
VOLUME_FRACTION_OF_MARTENSITE
EQUIVALENT_PHASE_TRANSFORMATION_STRAIN
EQUIVALENT_TWIN_STRAIN
EQUIVALENT_TRIP_STRAIN
COMPONENTS_OF_CAUCHY_STRESS
COMPONENTS_OF_TOTAL_STRAIN
COMPONENTS_OF_ELASTIC_STRAIN
COMPONENTS_OF_PLASTIC_STRAIN
COMPONENTS_OF_CREEP_STRAIN
COMPONENTS_OF_THERMAL_STRAIN
COMPONENTS_OF_STRESS_PREFERRED_SYSTEM
PHASE_TRANSFORMATION_STRAIN_TENSOR
INTERLAMINAR_SHEAR_THICK_ELEMENTS_TXZ
INTERLAMINAR_SHEAR_THICK_ELEMENTS_TYZ
INTERLAMINAR_NORMAL_STRESS
INTERLAMINAR_SHEAR_STRESS
Nodal Parameters
The functions and keywords are used to get nodal data, the available keywords in this category are:
DISPLACEMENT
ROTATION
VELOCITY
Main Index
756 Nonlinear Users Guide
User Defined Services and Subroutines
ROTATIONAL_VELOCITY
ACCELERATION
ROTATIONAL_ACCELERATION
COORDINATE
Example
To use UMAT and UCOHES material subroutines, user should implement the material SCA interface. For users
convenience, template files have been provided and distributed with MSC Nastran installation. These template files
can be found at:
[$InstallationDir]/[MSC Version]/nast/services/Implementations/Materials/src/
Users can write user code in the ext_umat.F or ext_ucohes.F subroutines under the umant and ucohesive
directories for special material behavior. The material user services can be built using MSC SDK tools. The SDK is a
separate installer that provides SCA build tools and environment. For information about SCA services and build,
please see MSC Nastran SCA and User Defined Services documents.
For demonstration, model and template files using UMAT are provided and can be found at:
[$InstallationDir]/[MSC Version]/nast/services/Implementations/Materials/src/umat/
Main Index
Chapter 13: Advanced Topics 757
References
References
1. Krueger, R., Virtual Crack Closure Technique: History, Approach and Applications, Appl. Mech. Rev., Vol.
57:2, pp. 109143, March 2004.
Main Index
758 Nonlinear Users Guide
References
Main Index
Chapter A: Interpreting the Results
Nonlinear Users Guide
Main Index
760 Nonlinear Users Guide
Results Quantities
Results Quantities
Consistent Units
MSC Nastran is unit independent and any system of units may be used. It is essential, however, that all data input
quantities are specified using a consistent unit system. For example, if N, m, and kg are used, then the model di-
mensions must be specified in meters, the elastic modulus in N/m2 and density as kg/m3. With respect to the output of
results if N, m, kg have been used in the data input, the results will provide stresses in N/m2, forces and reaction in
N, displacements in m etc.
Rotations are output in radians. This includes angular velocity and acceleration results.
The units of temperature being used in an analysis are dependent on what is assumed by the input material
properties and loads and constraints.
When choosing a set of units, an appropriate system should be used to avoid problems with numerical round-
off. For instance using the unit of meters to model a component that is only a fraction of a millimeter in overall
size would require that the distance between element nodes would be of the order of 1e-6. Similarly, the use of
Mega-Tonnes would be preferable when analysing a massive civil engineering structure compared to kg.
Using a dimensionally "consistent" set of units is essential in dynamic analyses to obtain meaningful results.
For a static analysis without any body force loading (using density), the requirement for consistency is
relaxed. In general, if units are chosen for mass (m), length (l) and time (t), the measure used for force must
dimensionally correspond with F = m a = ml/t/t.
The standard SI unit system is normally recommended; i.e., Newton, Metres, Kilograms. This complies by
definition with the equation F=Ma (N = kg m/s) and is termed a consistent set of units. If both sides are
divided by 1000, a further set of consistent units becomes apparent as (kN = Tonnes m/s). Similarly,
multiplying both sides gives (N = T mm/s).
Any consistent system can be used; for example:
N, Millimetres, Tonnes
KN, Metres, Tonnes
MN, Metres,
kTonnes Dyne,
Gram, Centimetre
Poundal, Pound
Feet
An example of nonstandard sets of units would be (kN, Millimetres, Tonnes), (N, Millimetres, kg).
Main Index
Chapter A: Interpreting the Results 761
Results Quantities
Deformations
Consider a three dimensional body B in its undeformed and deformed configuration (see Figure A-1). With respect to
a Cartesian coordinate system E 1 E 2 E 3 , the position vector of a material point in the undeformed configuration
is written as:
X = X1 E1 + X2 E2 + X3 E3 (A-1)
In the deformed configuration, the material point has a position vector x , given by:
x = x1 E1 + x2 E2 + x3 E3 (A-2)
The displacement vector u is defined as the difference between the position vector in the deformed and the
undeformed configuration and reads:
u = x X = u1 E1 + u2 E2 + u3 E3 (A-3)
Undeformed
B B
Deformed
u
dA 0
X ^ dA
dF
N
dF
E2 n
x
E1
E3
Figure A-1 Body B in Undeformed and Deformed Configuration
Main Index
762 Nonlinear Users Guide
Results Quantities
It is assumed that there is always a unique relation between the position vector of a point in the deformed and the
position vector of this point in the undeformed configuration. This can formally be expressed as:
x = x X (A-4)
Based on equation (A-4), a fundamental deformation measure can be given, namely the deformation gradient F , which
is defined by:
x x x
--------1- --------1- --------1-
X 1 X 2 X 3
x x x
F = --------2- --------2- --------2- (A-5)
X 1 X 2 X 3
x x x
--------3- --------3- --------3-
X 1 X 2 X 3
Substituting equation (A-5) into equation (A-3) shows that the deformation gradient can also be written as a function of
the coordinates in the undeformed configuration and the displacement components:
u 1 u u u u u
1 + --------- --------1- --------1- --------1- --------1- --------1-
X 1 X 2 X 3 X 1 X 2 X 3
u u 2 u u u u
F = --------2- 1 + --------- --------2- = I+ --------2- --------2- --------2- (A-6)
X 1 X 2 X 3 X 1 X 2 X 3
u u u 3 u u u
--------3- --------3- 1 + --------- --------3- --------3- --------3-
X 1 X 2 X 3 X 1 X 2 X 3
1 0 0
I = 0 1 0 (A-7)
0 0 1
Starting out from the deformation gradient, several well-known symmetric strain tensors can be defined, namely the
engineering strain tensor e , the Green-Lagrange strain tensor E and the right Cauchy-Green strain tensor C :
1 T
e = --- F + F 2I (A-8)
2
Main Index
Chapter A: Interpreting the Results 763
Results Quantities
1 T
E = --- F F I (A-9)
2
T
C = F F (A-10)
T
where F denotes the transpose of F .
Notice that the Green-Lagrange and the right Cauchy-Green strain tensor are related by:
1
E = --- C I (A-11)
2
Example
Suppose that the deformation of a body is described by:
1 1 1
x 1 = 4X 1 cos --- X 2 sin , x 2 = 4X 1 sin + --- X 2 cos , x 3 = --- X 3
2 2 2
This deformation can be obtained by first stretching a block of material in the E 1 direction and then rotating it around
the E 3 axis (see Figure A-2). The deformation gradient can easily be evaluated as:
1
4 cos --- sin 0
2
1
F = 4 sin --- cos 0
2
1---
0 0
2
so that the engineering and the right Cauchy-Green strain tensors are given by:
7---
4 cos 1 sin 0 16 0 0
2
1
e = 7 1 0 --- 0
--- sin --- cos 1 0 , C = 4
2 2
1
1 0 0 ---
0 0 --- 4
2
Main Index
764 Nonlinear Users Guide
Results Quantities
E3
L2
E2
L3
L 2 + L 2
L1 L 3 + L 3
E1 L 1 + L 1
From these expressions, it can be concluded that the engineering strain tensor only provides a useful deformation
measure if the angle remains small, so that cos 1 and sin 0 . On the other hand, the components of the right
Cauchy-Green tensor, and by virtue of equation (A-11) also the components of the Green-Lagrange strain tensor, are
independent of the value of the angle .
4 0 0
cos sin 0 1
F = 0 --- 0 = RU
sin cos 0 2
0 0 1 1
0 0 ---
2
in which R is a rotation tensor and U is a symmetric stretch tensor, where the stretch tensor and the right Cauchy-
Green strain tensor are related by:
12
C = U
It can be proved that in this way any deformation gradient can be uniquely decomposed into a rotation tensor and a
stretch tensor.
If there is no rotation of the material the nonzero components of the right Cauchy-Green strain tensor can be expressed
in terms of the components of the engineering strain tensor as:
C 11 = 1 + e 11 , C 22 = 1 + e 22 , C 33 = 1 + e 33
Instead of 1 + e 11 , 1 + e 22 , and 1 + e 33 , one often uses the principal stretch ratios 1 , 2 , and 3 , respectively.
Main Index
Chapter A: Interpreting the Results 765
Results Quantities
A geometrical interpretation of the principal stretch ratios can be given by indicating the initial edge lengths as L 1 ,
L 2 , L 3 and the changes in edge lengths as L 1 , L 2 , L 3 (see Figure A-2). Now the principal stretch ratios can be
written as:
L 1 + L 1 L 2 + L 2 L 3 + L 3
1 = ----------------------- , 2 = ----------------------- , 3 = -----------------------
L1 L2 L3
In the example discussed above, the right Cauchy-Green strain tensor only has nonzero terms on its main diagonal,
indicating that the deformation consists of a pure stretch. In a general state of deformation, there will also be nonzero
off-diagonal terms. Then the principal stretch ratios must be determined based on the eigenvalues of the right Cauchy-
Green strain tensor. Denoting these eigenvalues as C' 11 , C' 22 , and C' 33 , the principal stretch ratios are generally given
by:
It can be concluded that the principal stretch ratios completely define the stretch of a material, but not the rotation.
Another way to characterize the deformation of a material is based on the invariants of the right Cauchy-Green strain
tensor. These invariants are defined as:
I 1 = C 11 + C 22 + C 33 (A-13)
2 2 2
I 2 = C 11 C 22 + C 22 C 33 + C 33 C 11 C 12 C 23 C 31 (A-14)
2 2 2
I 3 = C 11 C 22 C 33 + 2C 12 C 23 C 31 C 11 C 23 C 22 C 31 C 33 C 12 (A-15)
Because I 1 , I 2 , and I 3 are invariants of the right Cauchy-Green strain tensor, their values can also be determined
based on the eigenvalues of the right Cauchy-Green strain tensor. Using equation (A-12), this yields:
2 2 2
I1 = 1 + 2 + 3 (A-16)
2 2 2 2 2 2
I2 = 1 2 + 2 3 + 3 1 (A-17)
2 2 2
I3 = 1 2 3 (A-18)
1 2 3 = 1 (A-19)
or:
I3 = 1 (A-20)
Main Index
766 Nonlinear Users Guide
Results Quantities
The compressibility can also be expressed in terms of the determinant of the deformation gradient, det F . Since
F = RU , this can be evaluated as:
1
det F = det RU = det R det U = det U = det C --- = 1 2 (A-21)
2 3
det F = 1 (A-22)
Stresses
Consider the deformed configuration of body B , as indicated in Figure A-4. On an elemental area dA with unit normal
vector n , an elemental force vector dF is acting. This force vector is a result of forces being transmitted from one
portion of the body to another. According to the Cauchy stress principle, the stress vector or traction vector t is
defined as:
dF
t = --------- (A-23)
dA
Similar to equation (A-21) to equation (A-23), the components of t , n , and dF are indicated as t 1 , t 2 , t 3 , n 1 , n 2 ,
n 3 , dF 1 , dF 2 and dF 3 . Now the following relation between the components of the stress vector and the components
of the normal vector can be given:
t1 T 11 T 12 T 13 n1
t2 = T 21 T 22 T 23 n2 (A-24)
t3 T 31 T 32 T 33 n3
dF 1 T 11 T 12 T 13 n1
dF 2 = T 21 T 22 T 23 n 2 dA (A-25)
dF 3 T 31 T 32 T 33 n3
In equations (A-24) and (A-25), T 11 to T 33 are the components of the true or Cauchy stress tensor T . The components
T 11 , T 22 and T 33 are called the normal or direct stress components, while the other components are called shear stress
components. The first index of the stress components defines the normal of the plane on which the stress vector acts.
The second index indicates the positive direction of the component (see Figure A-3). It can be shown that the Cauchy
Main Index
Chapter A: Interpreting the Results 767
Results Quantities
stress tensor is symmetric, so T 12 = T 21 , T 13 = T 31 and T 23 = T 32 . The physical meaning of the Cauchy stress
tensor is that it gives the current force per unit deformed area.
Another frequently used stress tensor in a large deformation analysis is the second Piola-Kirchhoff stress tensor. In
order to define this tensor, the force vector dF is transformed using the inverse of the deformation gradient F :
E3
T 23
T 21
E1 E2
T 22
dF 1 dF 1
1
dF 2
= F dF 2 (A-26)
dF 3 dF 3
^
Assuming that the transformed force vector d acts on the elemental area dA 0 with unit normal vector N in the
F
undeformed configuration (see Figure A-4), the components S 11 to S 33 of the symmetric second Piola-Kirchhoff
stress tensor S are defined as:
dF 1 S 11 S 12 S 13 N1
dF 2
= S 21 S 22 S 23 N 2 dA 0 (A-27)
dF 3 S 31 S 32 S 33 N3
The physical meaning of the second Piola-Kirchhoff stress tensor is not so clear. It can be considered to give the
transformed current force per unit undeformed area.
Main Index
768 Nonlinear Users Guide
Results Quantities
Using the deformation gradient, the Cauchy stress tensor and the second Piola-Kirchhoff stress tensor can be related
to another by:
1 1 T
S = det F F T F (A-28)
1 T
T = ----------------- FSF (A-29)
det F
Notice that for small deformations and small rotations, F I , so the differences between the Cauchy stress tensor and
the second Piola-Kirchhoff stress tensor vanish. In that case they reduce to the so-called engineering stress tensor ,
which is known to give the force per unit undeformed area.
Example
Due to a uniaxial tensile load, the state of deformation of a body is assumed to be given by (see also Figure A-4):
1 1
x 1 = 4X 1 , x 2 = --- X 2 , x 3 = --- X 3
2 2
A0
A0
F
A
F
E3
E2
A
E1
Figure A-4 Uniaxially Loaded Body
The force is assumed to be homogeneously distributed over the cross section A in the E 2 - E 3 plane. Evaluating
F T 11 T 12 T 13 1
0 = T 21 T 22 T 23 0 A
0 T 31 T 32 T 33 0
0 T 11 T 12 T 13 0
0 = T 21 T 22 T 23 1 A
0 T 31 T 32 T 33 0
Main Index
Chapter A: Interpreting the Results 769
Results Quantities
0 T 11 T 12 T 13 0
0 = T 21 T 22 T 23 0 A
0 T 31 T 32 T 33 1
so that the only nonzero component of the Cauchy stress tensor is:
F
T 11 = ---
A
Because:
4 0 0
1
F = 0 --- 0 , det F = 1
2
1
0 0 ---
2
it follows from equation (A-28) that the only nonzero component of the second Piola-Kirchhoff stress tensor is:
1F
S 11 = ------ ---
16 A
Upon rewriting the current cross sectional area A in terms of the original cross-sectional area A 0 as A = 1
--- A 0 , the
4
nonzero component of the second Piola-Kirchhoff stress tensor can also be written as:
1F
S 11 = --- ------
4 A0
in which F A 0 is recognized as the engineering stress 11 . The differences between the various stress components
can be summarized as:
1
T 11 = 4 11 , S 11 = --- 11
4
Main Index
770 Nonlinear Users Guide
Results Quantities
t
T
E = dV d (A-30)
0V
stress tensor
strain rate
V element volume
t actual time in the load history
The integration over time leads to the following recursive formula using the trapezoidal rule
1 T
E n + 1 = E n + E n + 1 = E n + --- n + 1 n + n + 1 dV (A-31)
2
1 T
E n + 1 = E n + --- u n + 1 f n + f n + 1 (A-32)
2
For computational convenience, MSC Nastran uses equation (A-32) to calculate the element strain energy. The internal
element forces are readily available in every step because they are needed for the force equilibrium. Note that
temperature effects are included in the internal element forces.
When loads from temperature differences or element deformation are present, the default definition of element strain
energy for linear elements differ from the definition for nonlinear elements. For linear elements, the element strain
energy is defined as
1 T T
E = --- u K e u u P et (A-33)
2
Main Index
Chapter A: Interpreting the Results 771
Results Quantities
where P e t is the element load vector for temperature loads and element deformation. Equation (A-33) assumes that the
temperatures are constant within a subcase. For nonlinear elements, the definition of equation (A-30) is used. In the
case of linear material and geometry, equation (A-30) becomes
1 T 1 T
E = --- u K e u --- u P et (A-34)
2 2
Equation (A-34) assumes that the temperature varies linearly within a subcase. The user may request the definition of
equation (A-34) to be applied to linear elements by adding PARAM,XFLAG,2 to the input file. The default value for
XFLAG is 0, meaning that linear elements will use the definition of equation (A-33).
User Input
The output of grid point force balance and strain energy in nonlinear analysis is requested with the existing GPFORCE
and ESE case control commands, respectively. The case control commands remain unchanged.
Main Index
772 Nonlinear Users Guide
Results Quantities
Stress and strain measures must be considered when displacements become large, for example strain measures used in:
Large Displacement results in Green Lagrange Strain
Updated Lagrangian results in Logarithmic Strain
No Definition results in Engineering Strain
Main Index
Chapter A: Interpreting the Results 773
Grid Point Stresses
5 lb
13 14 15 16
l = 10 in 7 8 9
10 lb
9 10 11 12
l = 10 in 4 5 6
10 lb
5 6 7 8
l = 10 in 2 3
1
5 lb
1 2 3 4
y t = 0.10 in
x
Figure A-5 Uniaxial Load for a 3 x 3 Mesh
Main Index
774 Nonlinear Users Guide
Grid Point Stresses
CQUAD4,4,1,5,6,10,9 x = 10 x = 10 x = 10
CQUAD4,5,1,6,7,11,10
CQUAD4,6,1,7,8,12,11 y = 0 y = 0 y = 0
$
v = 10 v = 10 v = 10
CQUAD4,1,1,1,2,6,5 x = 10 x = 10 x = 10
CQUAD4,2,1,2,3,7,6
CQUAD4,3,1,3,4,8,7 y = 0 y = 0 y = 0
$
v = 10 v = 10 v = 10
x = 10 x = 10 x = 10
CQUAD4,7,1,9,10,14,13
CQUAD4,8,1,10,11,15,14 y = 0 y = 0 y = 0
CQUAD4,9,1,11,12,16,15
$ v = 10 v = 10 v = 10
CQUAD4,4,1,5,6,10,9 x = 10 x = 0 x = 10
CQUAD4,5,1,7,11,10,6
CQUAD4,6,1,7,8,12,11 y = 0 y = 10 y = 0
$
v = 10 v = 10 v = 10
x = 10 x = 10 x = 10
CQUAD4,1,1,1,2,6,5
CQUAD4,2,1,2,3,7,6 y = 0 y = 0 y = 0
CQUAD4,3,1,3,4,8,7
$
v = 10 v = 10 v = 10
To avoid this type of surprise, it may be easier and probably more meaningful to look at the invariant stress quantities,
such as Hencky-von Mises stresses, rather than the component stresses. The grid point stress (GPSTRESS case control
command) option in MSC Nastran offers you another alternative. The grid point stress option calculates the stresses at
the grid points from the adjoining plate and solid elements in a coordinate system defined by you.
This option enables you to request the output of stresses at grid points in surfaces of two-dimensional plate elements-
-namely, the CQUAD4, CQUADR, CQUAD8, CTRIA3, CTRIAR, and CTRIA6 elements and in volumes containing
CHEXA, CPENTA, and CTETRA solid elements.
Main Index
Chapter A: Interpreting the Results 775
Grid Point Stresses
By default, stresses are output at the center of each of these elements and also at the vertices of the CQUADR, CQUAD4
(with corner option), CQUAD8, CTRIAR, and CTRIA6 shell elements as well as the CHEXA, CPENTA, and CTETRA
solid elements. This default output is not universally satisfactory because in the case concerning the surfaces of the
CQUAD4 (without corner output request) and/or CTRIA3 elements, you often have a requirement for more accurate
stress values at the grid points. However, the vertex stresses generated for these elements that connect to a common
grid point are often not identical. These deviations are most pronounced with a crude model and converge as the model
becomes more detailed. The option described in this section provides a rational method that yields reasonable and
unique stresses at each vertex grid point in surfaces or volumes defined by you.
Each stress component (e.g., x ) is treated independently of other stress components in the calculation of average grid
point stress components. The stress invariants at grid points (i.e., principal stresses, the Hencky-von Mises stress, and
the mean pressure) are evaluated from the average values of the stress components at grid points. The grid point stress
components are output in the output coordinate system of the surface.
Two methods are used to calculate grid point stress components for plate and shell elements-namely, topological and
geometric, with topological as the default method. The topological method involves the averaging of stresses at a grid
point due to the contribution of the elements directly attached to this grid point. This contribution does not include the
effect of the element size. On the other hand, the geometric method accounts for the element size based on the
numerical technique of least squares. If the results are substantially different when using the two different methods, it
may be an indication that your model is not fine enough. In general, if your model is fine enough to capture the stress
gradients, the two methods should yield similar results.
A single, simplified, topological method is used to calculate grid point stress components in volumes containing solid
elements. However, prior to a discussion of these methods, it is necessary to define the four categories into which grid
points are divided. These defintions are important for CQUAD4 (without corner output option) and CTRIA3 elements
since the stress averaging method used depends on the type of grid point. However, they have very little meaning for
CQUAD8, CTRIA6, and CQUAD4 (with corner output option) since the stress averaging is performed the same way
regardless of the type of grid points for elements with vertex stresses.
Interior grid points
Grid points that are connected only to interior line segments; i.e., line segments that are coincident with the
edges of two or more elements. If the model in Figure A-5 is used as an example, grid points 6, 7, 10, and 11
are considered as the interior grid points.
Main Index
776 Nonlinear Users Guide
Grid Point Stresses
Main Index
Chapter A: Interpreting the Results 777
Grid Point Stresses
Table A-3 Evaluation of the Average Grid Point Stresses for Plate and
Shell Elements by the Topological Method
Type of
Grid
Point Average Grid Point Stress Component Remarks
Interior Ne If CQUAD4 (without corner
1 output option) and/or CTRIA3
i = ------
Ne e i elements connect to the same
i = 1 grid point as CQUAD8 and/or
CTRIA6 elements, then only
where
CQUAD8 and/or CTRIA6
elements are used in the
i = average stress at interior grid points calculation of average grid
Ne = number of directly connected elements point stresses.
e i = element center stresses if only CQUAD4 and/or CTRIA3
elements are connected; element vertex stresses if
CQUAD8, CTRIA6, and/or CQUAD4 (with corner output
option) elements are connected
Main Index
778 Nonlinear Users Guide
Grid Point Stresses
Table A-3 Evaluation of the Average Grid Point Stresses for Plate and
Shell Elements by the Topological Method (continued)
Type of
Grid
Point Average Grid Point Stress Component Remarks
Corner Corner points connected to CQUAD4 element with the grid point Remarks for interior grid
(CQUAD4 diagonally opposite the corner point, an interior point, or an edge points also apply.
without point.
corner
output c = 2 1 D
option)
where
(If the point diagonally opposite the corner point is not an interior or
edge point, then c = 1 )
Corner Corner points connected to CTRIA3 elements and the other two grid Remarks for interior grid
(CTRIA3) points, namely, E1 and E1 of the connected CTRIA3, are edge points. points also apply.
c = 3 1 E 1 + E2
where
Main Index
Chapter A: Interpreting the Results 779
Grid Point Stresses
Table A-3 Evaluation of the Average Grid Point Stresses for Plate and
Shell Elements by the Topological Method (continued)
Type of
Grid
Point Average Grid Point Stress Component Remarks
Grid points N
c
Remarks for interior grid
connected 1 points also apply.
to c = ------ r i
Nc
CQUAD8, i = 1
CTRIA6,
where
and/or
CQUAD4
(with c = average stress at r-th grid points
corner
Nc = number of CQUAD8, CTRIA6, and CQUAD4 (with corner
output
output option) elements connected to the r-th grid point
option)
ri = stress at the vertex of the i-th element connected to the r-th
grid point
Table A-4 Evaluation of the Average Grid Point Stresses for Solid Elements by the Topological Method
Type of
Grid
Point Average Grid Point Stress Component Remarks
All N
e
Mixtures of
1 CHEXA, CPENTA,
1 = ------
Ne ei and CTETRA
i = 1 elements may
where connect to an
interior or edge grid
point.
1 = average stress at interior or edge grid points
In the geometric method, grid point stresses for points that connect to CQUAD8, CTRIA6, and/or CQUAD4 (with corner
output request) elements use the average of the element corner stresses. If CQUAD4 (without corner output request) or
CTRIA3 elements connect to these points, their influence is not considered. The geometric method for the
determination of average grid point stresses that connect only to CQUAD4/CTRIA3 elements utilizes the numerical
technique of least squares.
Main Index
780 Nonlinear Users Guide
Grid Point Stresses
1
2
3
4
5
Figure A-6 Adjacent Elements with Different Slopes Connected to the Same Line
Note that this situation is geometric rather than topological in nature and is, therefore, only considered under the
geometric interpolation option.
Since exception points are defined as grid points at which stress discontinuities may exist, it is conceivable that a
computed average grid point stress at an exception point may be different for each of the directly connected elements.
However, it is more common to find that a subset of the elements, which are directly connected to an exception point,
forms a subsurface on which stresses may be considered to be continuous. A unique average grid point stress may then
be computed for the elements of the subsurface. For the problem shown in Figure A-6, you may consider creating two
subsurfaces one subsurface consisting of elements within the shaded surface and another subsurface consisting of
elements within the unshaded surface.
The calculation of average grid point stresses for a subsurface that contains two or more elements is performed when
the geometric interpolation method is used for edge points. If a subsurface consists of a single element, the calculation
of an average grid point stress is performed in the same manner as utilized in the calculation of average grid point
stresses at corner points. It must be noted, however, that D , E1 , and E 2 may be interior, exterior, or corner points
(see Table A-4).
As stated previously, each element stress component (independently of all other stress components) is transformed into
the output coordinate system before averaging. It is necessary that this output coordinate system vary slowly over the
surface in order for the interpolated stresses to be accurate. This approach is certainly valid for stress interpolation over
elements that are in a common plane, and it should be adequate for engineering analysis when the difference in slope
between adjacent elements is small. If this slope is not small, it is suggested that you break them into multiple surfaces
or volumes. When the elements do vary substantially from a smooth surface, you must utilize engineering judgment
relative to the accuracy of the resultant stress data.
Main Index
Chapter A: Interpreting the Results 781
Grid Point Stresses
Average grid point stresses are only computed at element vertices. For elements with mid-edge grid points, you can
estimate the stresses at these mid-edge grid points to be the average of the computed stresses at the grid point at the
ends of the edge in question.
In static analysis, place the GPSTRESS command above the subcase level or in individual subcases. If the GPSTRESS
command is above the subcase level, it is utilized by all subcases and subcoms unless overridden by a GPSTRESS
command in a subcase or subcom.
Note that if grid point stress data is desired, an ELSTRESS (or STRESS) request for all elements in the surfaces or
volumes of interest must be included in the Case Control Section. This request is in addition to the GPSTRESS request.
Such requests cause the output of element stress data as well as grid point stress data. If the printed output of element
stress data is not desired, the requests ELSTRESS(PLOT) = x or STRESS(PLOT) = x may be used. In the model
shown in Figure A-5, if grid point stresses are requested for grid point 11, then element stresses must be requested for
a set that includes at least elements 5, 6, 8, and 9. Furthermore, the set of elements in the SURFACE definition must
also include elements 5, 6, 8, and 9. In this example, if one of the elements that is connected to grid point 11 (e.g.,
element 6) is left out in the SET definition, then User Warning Message 4621 is issued. The GPSTRESS values
calculated in this case may be meaningless.
The OUTPUT(POST) section of the Case Control Section contains specific requests for grid point stress data for the
surfaces or volumes of elements. Any number of element surfaces or volumes may be defined, but only those surfaces
or volumes referenced through the GPSTRESS command will have grid point stress data generated and output.
Main Index
782 Nonlinear Users Guide
Grid Point Stresses
Again, the model in Figure A-5 is used for this illustration. In this case, you may want to have two separate surfaces.
The first surface contains elements 1, 2, 4, 5, 7, and 8; the second surface contains all elements. Part of the input file
is shown. (See Install_dir/mscxxxx//doc/linstat/gpstress.dat)
$
$ FILENAME - gpstress.dat
$
STRESS= ALL
SET 10 = 100,200
GPSTRESS = 10
STRFIELD = 10
$
OUTPUT(POST)
SET 1 = 1,2,4,5,7,8
SET 2 = ALL
SURFACE 100 SET 1 NORMAL Z
SURFACE 200 SET 2 NORMAL Z
In the above example, surface 100 contains elements 1, 2, 4, 5, 7, and 8 that were referenced by set 1. Surface 200
contains all of the elements that were referenced by set 2. The element positive normal is in the direction of the z-axis
of the basic coordinate system. By default, the grid point stresses are transformed to the basic coordinate system. The
grid point stresses are called out by the GPSTRESS command, which points to set 10, which, in turn, points to surfaces
100 and 200. Element stresses are requested for all of the elements, which, of course, encompass elements 1, 2, 4, 5,
7, and 8. The STRFIELD command is required to generate postprocessing data for your graphic package. A partial
output of a typical grid point stress is provided as follows.
S T R E S S E S A T G R I D P O I N T S - - S U R F A C E 100
SURFACE X-AXIS X NORMAL(Z-AXIS) Z REFERENCE COORDINATE SYSTEM FOR SURFACE DEFINITION CID 0
GRID ELEMENT STRESSES IN SURFACE SYSTEM PRINCIPAL STRESSES MAX
ID ID FIBER NORMAL-X NORMAL-Y SHEAR-XY ANGLE MAJOR MINOR SHEAR VON MISES
1 0 Z1 1.000E+01 -2.009E-14 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
Z2 1.000E+01 -2.009E-14 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
MID 1.000E+01 -2.009E-14 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
2 0 Z1 1.000E+01 -8.986E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
Z2 1.000E+01 -8.986E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
MID 1.000E+01 -8.986E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
.
.
15 0 Z1 1.000E+01 -1.631E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
Z2 1.000E+01 -1.631E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
MID 1.000E+01 -1.631E-15 -1.093E-07 .0000 1.000E+01 0.000E+00 5.000E+00 1.000E+01
Main Index
Chapter A: Interpreting the Results 783
Grid Point Stresses
Min Patran uses the minimum of the integration point values at each node.
Max Patran uses the maximum of the integration point values at each node.
The accuracy of the extrapolation procedure is dependent on: 1) the presence of a reasonably uniform stress field; 2)
a reasonably regular element shape; and 3) the type of shape function used in the element chosen. For instance, a high
stress gradient across an element would be more likely to extrapolate incorrectly; particularly if a linear shape function
element is being used. Thus, although equilibrium conditions will always be met, the stress distribution can be wholly
inaccurate for an insufficiently refined mesh.
In a typical analysis, each node point will be connected to more than one element and, hence, an averaging process is
required to obtain a single value of stress at this node (simply adding all the component stresses and dividing by the
number of components) this is the methodology selected on Patran when choosing "Averaging Definition" Domain
and Method on the Results> Create: Fringe> Plot Options sub-form. Setting "Domain = Individual" will not perform this
process.
Figure A-7 Patran Averaging and Extrapolation Methods are Controlled on the Fringe Plot Options Form
In areas of interest where stresses will be used as input to the design process, a contour plot of averaged nodal stress
should very similar to one using un-averaged nodal results (a smooth transition across element boundaries). This would
indicate that the stress distribution in the structure is being simulated sufficiently accurately. For other sections of the
model which are not of interest, a coarser mesh would normally be used and a comparison in these areas would
typically give significantly different contour plots the un-averaged contours appearing more like a patchwork quilt.
The difference in contour values at the element boundaries indicating either that the mesh is too coarse or the element
type insufficient to simulate the stress variation adequately.
To turn on/off nodal averaging, select Results> Create: Fringe> Plot Options
By default Patran always displays nodal averaged values, unless the Averaging:None option is used
(Averaging: None on Results> Create: Fringe> Plot Options).
The maximum and minimum values on the legend for a plot are obtained by looping over all the nodes. This
means that values at nodes that do not belong to any element (contact control nodes, for example) do influence
the min/max values.
The maximum and minimum values on the legend for a Contour plot are obtained by looping over all the
nodes of each element. This means that values at nodes that do not belong to any element (contact control
nodes, for example) do not influence the min/max for Contour plots.
Main Index
784 Nonlinear Users Guide
Grid Point Stresses
If all values in a Contour plot are identical (say equal to v), there will only be one.
The fringe range of a contour plot in Patran is dependent on what is currently visible on the screen. This means
that when contouring selected parts of a structure, the elements that are not in one of the posted groups will not
make a contribution to the range calculation process. Depending on the relative size of the stress contributions
from the elements that are missing, the contours around the edge could have will change significantly.
To suppress averaging for specific element combinations, remove the elements from any posted groups. This
means that when contouring selected parts of a structure, the elements that have been removed from the posted
groups will no longer make a contribution to the averaging process. Depending on the relative size of the
stress contributions that are now missing, the contours around the edge could change significantly.
Nodal averaging of element results to the output file is carried out using a weighted averaging scheme based
on the angle spanned at a node of an element. This should not have an effect for a perfectly regular mesh.
The Extrapolation: Centroid on Results> Create: Fringe> Plot Options method in Patran (Results) is a further
way to obtain contour plots. Specifying this command sets the results plotting style to element centroid
contours.
Main Index
Chapter A: Interpreting the Results 785
Shape Function Interpolation
1
N 1 = --- 1
2
1
N 2 = --- 1 +
2
where N 1 and N 2 are the shape functions at nodes 1 and 2 of the element respectively (the order being dependent on
the element node numbering). Diagrammatically, their variation is as follows:
1 1
= 1 = +1 = 1 = +1
Variation of N1 Variation of N2
Main Index
786 Nonlinear Users Guide
Shape Function Interpolation
N2 = 1 + 1
N 3 = --- + 1
2
where N 1 , N 2 , and N 3 are the shape functions at nodes 1, 2 and 3 of the element respectively. Diagrammatically,
their variation is as follows:
1 1
= 1 = 0 = +1 = 1 = 0 = +1
Variation on N1 Variation on N1
= 1 = 0 = +1
Variation on N3
Quadratic variations are also used on eight-noded surface elements. The following diagrams show the variations of
the shape functions at both corner and midside nodes.
Main Index
Chapter A: Interpreting the Results 787
Shape Function Interpolation
N i j j j = 1 (for i = j )
This means that the value of each shape function evaluated at its nodal position must be unity. For example,
N 1 = 1 = --- 1 = 1
2
N2 = 0 = 1 + 1 = 1
N 3 x = 1 = --- + 1 = 1
2
N i j j = 0 (for i j )
Requires that the values of each shape function, evaluated at the other nodes must be zero. That is,
N 1 = 0 = --- 1 = 0
2
N 1 = 1 = --- 1 = 0
2
N i , = 1
i
The sum of all the shape functions, evaluated at any point must be unity. That is
1
N 1 = 1 2 = --- 1 = ---
2 8
6
N 2 = 1 2 = 1 + 1 = ---
8
3
N 3 = 1 2 = --- + 1 = ---
2 8
Furthermore, to ensure that a finite element convergences to the correct result, certain requirements need to be satisfied
by the shape functions, as follows
The displacement function should be such that it does not permit straining of an element to occur when the
nodal displacements are caused by rigid body displacement. This is self evident, since an unsupported
structure in space will be subject to no restraining forces.
The displacement function should be of such a form that if nodal displacements produce a constant strain
condition, such constant strain will be obtained. This is essential since a significant mesh refinement will
cause near-constant strain conditions to occur in elements and they must be able to handle this condition
correctly.
The displacement function should ensure that the strains at the interface between elements are finite (even
though indeterminate). By this, the element boundaries will have no gaps appear between them and, hence,
will show a continuous mesh.
Main Index
788 Nonlinear Users Guide
Shape Function Interpolation
The following sections deal with some of the more frequently encountered practical implications that are related to the
use of shape functions.
u = Nd
where u are the displacements at any point within an element and d are the displacements at the nodes of an
element.
This equation relates the displacements at any point within an element to the nodal displacements according to the
element shape function N . Therefore, the displacement at any point in a two-noded line element can be obtained
from the nodal values using the following equation
1 1
u = N 1 d 1 + N 2 d 2 = --- 1 d 1 + --- 1 + d 2
2 2
If this element is fully fixed at one end d 1 = 0 and sustains a displacement of 2 at the other end d 2 = 2 , the
displacement at the centre of this element (=0) would be given thus
1 1
u 0 = --- 0 + --- 2 = 1
2 2
i.e., half the end displacement as expected.
The same can be done with any quantity that varies across an element, for example, coordinates, strain, stress, and
thickness.
The quadratic terms in thus giving a corresponding quadratic variation of displacement over the element. The strain
variation can be defined as
d 1 1 1 1 1
x = B d = ------ --- 2 + --- d = --- --- 2 + --- d
dx 2 2 J 2 2
where B is that strain-displacement matrix and d are the three element axial displacements. It can be seen from
the terms that the strain is now a linear variation as will be the stress variation. In a similar manner, for a linear
element, the strain and stress variation will be constant.
Main Index
Chapter A: Interpreting the Results 789
Shape Function Interpolation
This has a direct bearing on the type of element to be chosen for an analysis. For instance, consider a bar element under
the action of a constant uniformly distributed load along the length of the element. The resulting axial force variation
will be theoretically linear as in the topmost picture of the following diagram.
Theoretical
Solution
Axial Force
Linear Elements
(constant 'stepped' solution)
Four Element
Mesh
Quadratic Elements
(linear solution)
One Element
Mesh
If this bar is modeled using linear elements (i.e., linear terms in the shape function), the axial force will be
approximated by a constant, stepped response in each element, since the shape function derivatives only contain
constant terms. A quadratic element (i.e., quadratic terms in the shape function) will, however, support a linear
response and provide the correct answer directly, since the shape function derivatives contain linear terms. Thus, the
exact solution can be obtained with a relatively small number of elements (or even with one element only) if the actual
strain field can be matched by the shape functions of the element that is being used. In the above example, the shape
function derivative terms did indeed match the linear strain of the actual analysis.
A frequent observation when inspecting force output at a simply supported section of a structure is to find
(unexpectedly) nonzero values. Depending on the degree of mesh refinement, these values can be significant compared
to the peak values. The reason is directly related to the above discussion. For example, if the force distribution is at
least quadratic in form and linear elements are used (typically supporting a constant force distribution), a stepped
response will be seen hence the nonzero values these constant values represent an average of the force distribution
and, if summed across the structure would be found to be equilibrium. The use of quadratic elements will improve the
situation, but even these will not be able to match third order or higher force distributions without a measure of mesh
refinement performed.
In spite of this sort of discrepancy, it should be noted that, during the solution stage, the equilibrium equation is used
f = K d to ensure that the product of the stiffness matrix and the computed displacements exactly balances
the externally applied forces. This means that, unless there are pertinent warnings or errors output during the solution,
Main Index
790 Nonlinear Users Guide
Shape Function Interpolation
static equilibrium will have been fully achieved. Moreover, the derived quantities of strain and stress will also be found
to be in equilibrium but not necessarily according to an expected distribution as noted in preceding paragraphs.
Similar difficulties can be observed when attempting to compare the reactions at a location in a structure with the
element force output at the same location. The explanation in most cases is, again, related to the order of shape function
that has been used to formulate the element.
The remedies are to either increase the number of linear elements used (and reduce the size of the step change
between each element) or change to quadratic elements (to more closely match the actual variation).
Apart from the consideration of element selection related to the order of shape function, quadratic elements would be
recommended in the presence of high degrees of plastic strain since they are less susceptible to locking. Linear
elements, however, would be recommended when the stress distributions anticipated are constant or linear. Such
elements are computationally cheaper and, in such circumstances, render the use of higher order elements unnecessary.
Gauss Points
0 20 100
Nodal
x x
Temperatures
0 100
x x
0 20 100
T=0 T = 100
Temperature Variation
Across Element
Main Index
Chapter A: Interpreting the Results 791
Shape Function Interpolation
As a result of the quadratic displacement assumption used in higher order elements, the interpolation to the Gauss
points yields the variation of temperature across the length of the element as shown.
This variation will ensure that the applied temperature loading is applied correctly to the structure, but for the Gauss
points nearest to the zero temperature specification this may not be so.
For most cases the negative value is insignificant compared to the temperature loading specified and the variation in
the temperature dependency of the material properties. Mesh refinement in the area of the greatest temperature
variations is the most appropriate remedy.
Gauss Points
1 20 100
x x
Nodal
Thicknesses
1 100
x x
1 20 100
t=1 t = 100
When using the quadratic displacement assumption used in higher order elements, the interpolation to the Gauss points
yields the variation of thickness across the element as shown in the second picture (above). The effect of a significant
variation of thickness over a single element may, thus, cause a zero or negative thickness value at a Gauss point.
The remedy is to check that the thickness variation applied to the specified element is applied correctly. If so, then the
mesh should be refined to reduce the severity of the thickness variation over the element.
Main Index
792 Nonlinear Users Guide
Finite Element Equilibrium
4 3
i
1 2
The following diagram, representing an exploded view of these four elements, shows the forces obtained at the nodal
position (i) and those on element (2).
4 3
1 2
Element forces
are in equilibrium
Main Index
Chapter A: Interpreting the Results 793
Finite Element Equilibrium
Nonlinear analyses may produce out-of-balance residual forces at a node, depending on the degree of convergence
obtained during the solution. For a well-converged solution, however, these are insignificant. See nonlinear iterative
strategy for more information.
Although nodal and element equilibrium is achieved as described above, in a general finite element analysis,
differential equilibrium (e.g. stress continuity) is not necessarily achieved at all points of the continuum considered
most notably at the shared boundaries of elements as discussed below.
0
In the displacement-based FEM, a C continuous approximation for the displacements is assumed within each
element. This means that the displacements at any point in a mesh will be continuous and ensures that no gaps appear
between elements. The element stresses are calculated using derivatives of the displacement, which means that they
will not necessarily be continuous and give rise to inter-element discontinuities or jumps in stress between adjacent
elements. This is particularly the case for coarse element meshes. The discontinuities at adjacent element boundaries
are reduced with mesh refinement and the rate at which mesh refinement reduces such discontinuities is determined
by the order of the elements in the mesh higher order elements converging faster than low order.
For the same reasons that element stresses are not continuous across element boundaries, the element stresses at the
surface of a finite element model are, in general, not in equilibrium with the external applied tractions. Again, this
effect is minimised with mesh refinement.
Experience has shown that the most accurate locations for stress output are the Gauss points. Nodal points, which are
the most accessible, are actually the worst output location for stresses. Reasons have been given above, but include the
fact that shape functions tend to behave badly at element extremities and it is reasonable to expect that the shape
function derivatives (i.e. strains/stresses) sampled in the interior of the element would be more accurate than those
sampled at the periphery of the element.
This evokes the question of how to obtain accurate stress results from a finite element model?
4 3
d c
1 2
a b
Gauss
Points
Main Index
794 Nonlinear Users Guide
Finite Element Equilibrium
This procedure is carried out for the other elements and the nodal stresses at the common node i are obtained as
i 1 , i 2 , etc. As pointed out above, these stresses are not usually equal and a single averaged or smoothed
nodal stress value is obtained using
i 1 + i 2 + i 3 i 4
i = -------------------------------------------------------------------
4
When this procedure is carried out for all nodes in an element assembly, the ensuing averaged stress values provide a
reasonable approximation to a continuous stress field. This is a straightforward and economic solution and works well
on the whole. See later for more details on the circumstances that smoothing should not be used. If smoothed results
are not selected then the extrapolation procedure is still performed, but the averaging process is omitted.
Note that, for shell elements, the local Gauss point stresses and strains are transformed to global stresses and strains
before extrapolation to the nodes. The mean global stresses are then transformed to the local shell system at the nodal
point before evaluation of the nodal stress resultants.
Other methods are available, based on a least squares fit over the integration point stress values of the elements. The
least squares procedure might be applied over the patches of adjacent elements or even globally over a whole mesh.
However, if the domain over which the least squares fit is applied involves many stress points, the solution will be
expensive and, in addition, a large error in one part of the domain may affect rather strongly the least squares prediction
in the other parts.
In general, it is recommended to display unsmoothed stress contours at an early point during the processing of results.
In this way, severe stress discontinuities between elements will be apparent and the possible requirement of mesh
refinement and/or the use of higher order elements may be considered.
In areas of interest where the stress results will be used in the design process, smoothed contours would ideally be
similar to unsmoothed contours. The inference from this being that a smooth stress transition across the element
boundaries indicates that the stress distribution in the structure is being simulated sufficiently accurately. For sections
Main Index
Chapter A: Interpreting the Results 795
Finite Element Equilibrium
of the model that are not of interest, a coarser mesh would normally be used and such a comparison in these areas would
typically give significantly different contours smoothed contours appearing more like a patchwork quilt!
The nodal averaging technique is sufficiently robust that such stress values will tend to be pretty much those that would
be obtained at the same location with mesh refinement as long as the element mesh is reasonably uniform.
At all times, it is imperative to remember that the FEM is an approximate numerical technique (albeit a good one) and
that smoothed stress results can give good results but need careful attention.
x
y
x
y
x
Interconnecting BEAM elements. It is necessary to extract results for longitudinal members and transverse
members separately.
Main Index
796 Nonlinear Users Guide
Finite Element Equilibrium
z
y x x z
z y
y x
x z z
y x x z
y
z 1 y
x z y x
y x z z
y y x
x z z
y x
y
Transverse
Members
Longitudinal
Members
Consider the situation in which M x results have been selected for display, and both transverse and longitudinal
members are active. The averaged value displayed at the central node number (1) will be comprised of the local M x
values from the two longitudinal and two transverse members that connect to this node. It must be noted that the M x
values are local to the elements (as shown in the diagram), so that the M x values for the longitudinal members act at
90 to the M x results in the transverse members. This means that the averaged values will be meaningless since the
M x results from the longitudinal members will be averaged with the M x values of the transverse members that are
acting in a completely different direction.
Continuum Stresses
Also known as physical stresses in MSC Nastran, these are named after the stress definitions used in classical
continuum mechanics theory and are defined in the usual manner as force per unit area and are denoted by the
following symbol types
x , y , z , xy , y z , zx
x , y , z , x y , yz , zx
S x ,S y ,S z ,S xy ,S yz ,S zx
The first subscript for the shear stress component stresses represents the face on which the stress acts, the second is the
normal direction of this face. Hence, the stress x y is acting on the x-face in the global y-direction. The MSC Nastran
definition is given as follows:
Main Index
Chapter A: Interpreting the Results 797
Finite Element Equilibrium
XY
YZ
X
X X XZ
XZ
Z
XY
X
Y
X
Z
y y = y
zz = z
The shear stress in two perpendicular cutting directions at a point have the same value, i.e.
x y = yx
x z = z x
y z = z y
The stress tensor ij comprises the 6 independent stress values from 3 perpendicular cutting planes at a point
xx xy x z
and describes the stress behavior completely: ij = yx yy y z where ij = ji .
zx zy zz
Shell Elements
The shell elements also evaluate stresses at a number of points through their thickness. This is carried out in order to
obtain bending and membrane effects as a result of increasing distance from the neutral axis. These through-thickness
stress evaluation points evaluate what are commonly termed surface, layer, or fibre stress results.
Numerical integration through the shell thickness is performed using Simpson's rule for homogeneous shells and the
trapezoidal rule for a composite shell. For non-homogeneous materials, the number of layers is defined through the
PCOMP or PCOMPG bulk data entry.
Main Index
798 Nonlinear Users Guide
Finite Element Equilibrium
Shell Layer Convention: The layer number convention is such that layer one lies on the side of the positive normal to
the shell, and the last layer is on the side of the negative normal. The normal to the element is based upon both the
coordinates of the nodal positions and upon the connectivity of the element:
Top Layer
(layer 1)
Local
Element Axis
Z
X
Bottom Layer
(layer "n")
Generalized Stresses
Generalized stresses are also known as average membrane stresses and are available for shell elements.
They are dimensionally defined as a stress. They are not to be confused with stress resultant output, defined
as force or moment per unit width
Generalized stresses are obtained at each Gauss point by integrating the shell layer continuum stresses over
the thickness of the shell. In the same way, generalised strain are their work-conjugate counterparts and
examples include curvature and rotation
The equation given for the generalised stress is:
t2
G = dy
t 2
Main Index