(Solid Mechanics and Its Applications 59) Lazarus Teneketzis Tenek, John Argyris (Auth.) - Finite Element Analysis For Composite Structures-Springer Netherlands (1998)

FINITE ELEMENT ANALYSIS FOR COMPOSITE STRUCTURES
SOLID MECHANICS AND ITS APPLICATIONS

Volume 59
Series Editor: G.M.L. GLADWELL

Solid Mechanics Division, Faculty of Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3Gl
Aims and Scope of the Series

The fundamental questions arising in mechanics are: Why?, How?, and How much?
The aim of this series is to provide lucid accounts written by authoritative research-
ers giving vision and insight in answering these questions on the subject of
mechanics as it relates to solids.
The scope of the series covers the entire spectrum of solid mechanics. Thus it
includes the foundation of mechanics; variational formulations; computational
mechanics; statics, kinematics and dynamics of rigid and elastic bodies; vibrations
of solids and structures; dynamical systems and chaos; the theories of elasticity,
plasticity and viscoelasticity; composite materials; rods, beams, shells and
membranes; structural control and stability; soils, rocks and geomechanics;
fracture; tribology; experimental mechanics; biomechanics and machine design.
The median level of presentation is the first year graduate student. Some texts are
monographs defining the current state of the field; others are accessible to final
year undergraduates; but essentially the emphasis is on readability and clarity.
For a list of related mechanics titles, see final pages.

Finite Element
Analysis for
Composite Structures
by
LAZARUS TENEKETZIS TENEK

University of Stuttgart,
Germany
and
JOHN ARGYRIS
University of Stuttgart,
Germany
Springer-Science+Business Media, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.
Printed on acid-free paper
All Rights Reserved

ISBN 978-90-481-4975-9 ISBN 978-94-015-9044-0 (eBook)
DOI 10.1007/978-94-015-9044-0
1998 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1998.
Softcover reprint of the hardcover 1st edition 1998
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner.
Contents
Preface ix
Acknowledgments xi
1 Some results from continuum mechanics 1

1.1 The analysis of stress . 1
1.2 The analysis of strain . . . . 6
1.3 Strain energy . . . . . . . . . 9
1.4 The principle of virtual work 12
1.5 The principle of stationary energy 13
2 A brief history of FEM 17

2.1 The matrix displacement method . . . . . 17
2.2 The finite element method . . . . . . . . . 19
2.3 The natural mode finite element method. 20
2.4 The basic ideas of FEM . . . . . . . . . . 21
3 Natural modes for finite elements 27

3.1 The concept of natural modes . . . 27
3.2 The concept of natural stiffness matrix . . 28
3.3 Natural modes for selected finite elements 31
4 Composites 37
4.1 Fundamental concepts 37
4.2 Basic material unit . . 39
4.3 Laminates....... 41
4.3.1 Special laminates . 42
4.4 Micromechanics and macromechanics . 44
vi CONTENTS
5 Composite beam element 47

5.1 Introduction.... 47
5.2 Kinematics . . . . . . . 48
5.3 The beam element . . . 51
5.4 Natural rigid body modes 52
5.5 Natural straining modes . 56
5.5.1 Mode 1: Extension 57
5.5.2 Mode 2: Symmetrical bending in the z - z plane . 60
5.5.3 Mode 3: Antisymmetrical bending and transverse
shearing in the z - z plane . . . . . . . . . . . . .. 61
5.5.4 Mode 4: Symmetrical bending in the z - y plane.. 64
5.5.5 Mode 5: Antisymmetrical bending in the z - y plane 64
5.5.6 Mode 6: Torsion about the x axis 65
5.6 Natural stiffness matrix . . . . 67
5.6.1 Strain operator matrix . 67
5.6.2 Constitutive relation . 68
5.6.3 Strain energy . . . . . . 70
5.6.4 Evaluation of integrals. 72
5.6.5 Shear correction factor. 78
5.7 Local and global stiffness matrices 81
5.8 Work of external loads . . . . . 86
5.9 Initial load due to temperature 89
5.9.1 Evaluation of integrals. 91
5.10 Postprocessing . . . . . . . . . 95
5.10.1 Computation of forces and moments 95
5.10.2 Natural energies . . . . . . . . 96
5.10.3 Through the thickness stresses 97
5.11 Geometrical stiffness . . . . . . . . . . 99
5.11.1 Elastic buckling . . . . . . . . . 99
5.11.2 Simplified geometrical stiffness 100
5.11.3 The natural geometrical stiffness 106
5.12 Partly simplified geometrical stiffness. 108
5.13 Computational experiments . . . . 114
5.13.1 Isotropic beams and frames 114
5.13.2 Composite beam structures 119
5.14 Problems . . . . . . . . . . . . . . 132
CONTENTS vii
6 Composite plate and shell element 135

6.1 Introduction . . . . . . . . . . . . . . . . . . . 135
6.2 Natural kinematics of the shell element TRIC 137
6.3 Constitutive relation . . . . . . 150
6.4 Stress resultants - equilibrium . . . . . . . . . 158
6.5 Natural modes and stiffness . . . . . . . . . . 163
6.6 Total strain in the natural coordinate system 167
6.7 Axial and symmetrical bending stiffness terms. 170
6.8 Antisymmetrical bending and shearing stiffness terms 175
6.8.1 Antisymmetrical bending terms. 176
6.8.2 Antisymmetrical shearing terms. 185
6.9 Shear correction factors . . . . . . . 189
6.10 Simulative azimuth stiffnesses . . . . 191
6.11 Local and global cartesian stiffnesses 191
6.12 Kinematically equivalent nodal loads 196
6.13 Initial load due to temperature . . . 198
6.14 Computation of stresses and stress resultants 202
6.14.1 Computation of stress resultants . . . 202
6.14.2 Computation of through-the-thickness stresses 204
6.15 The simplified geometrical stiffness 208
6.15.1 Geometrical forces . . . . 208
6.16 Geometrically nonlinear analysis . 217
6.17 Computational experiments . . . . 218
6.17.1 Clamped isotropic plate under central load 218
6.17.2 Thick sandwich plate under uniform pressure 219
6.17.3 Pinched cylinder; Scordelis-Lo roof; pressurized shell 220
6.17.4 Pinched hemispherical shell . . . . . . . . . . . . .. 221
6.17.5 Twisted beam. . . . . . . . . . . . . . . . . . . . .. 224
6.17.6 Eight-layer (0/45/ - 45/90)5 laminate under uniform
load . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
6.17.7 Stresses in a sandwich plate-comparison with the elas-
ticity solution . . . . . . . . . . . . . . . . . 229
6.17.8 Deformation of a (0/90/0) square laminate ... . . . 230
6.17.9 Stresses in a (0/90/0) rectangular laminate . . . .. 233
6.17.10 Large deflections of an isotropic plate-comparison
with experimental results . . . . . . . . . . . . . 233
6.17.11 Buckling of a cross-ply (0/90/90/0) laminate . . 234
6.17.12 Elastic stability of a (0/45/ - 45/90)8 laminate. 239
Vlll CONTENTS
6.17.13 Thermomechanical buckling of a cylindrical compos-

ite panel . . . . . . . . . . . . . . . . . . . . . . . .. 239
6.17.14 Buckling of a composite vessel under hydrostatic pres-
sure . . . . . . . . . . . . . . . . . . . . . . . . . .. 240
6.17.15 Buckling of a rocket-like composite shell under exter-
nal pressure and temperature 250
6.18 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . .. 258
7 Computational statistics 261

7.1 A model problem . . . . 261
7.2 Computational statistics report 261
8 Nonlinear analysis of anisotropic shells 267

8.1 Stable and unstable equilibrium paths 267
8.2 The incremental/iterative scheme. 268
8.3 Numerical examples . . . . . . . . 274
8.3.1 Isotropic cylindrical panels 274
8.3.2 Composite shells . . . . . . 287
9 Programming aspects 301

9.1 Memoryallocation . . . . . . . . 301
9.2 Input data . . . . . . . . . . . . . 302
9.3 Program parameters and storage 309
9.4 Elemental elastic stiffness matrix 311
9.5 Assembly of the global stiffness matrices 313
9.6 Solution of the linear equations 314
9.7 Postprocessing . . . . . . . . . . . . 314
9.8 Geometrical stiffness . . . . . . . . . 315
9.9 Assembly of the geometrical stiffness 319
9.10 Scholium . . . . . . . . . . . . . . . . 319
Appendices 321
A Geometry of the beam element in space 321
B Contents of floppy disk 325
Bibliography 327
Index 335
Preface
This book is an adventure into the computer analysis of three dimensional

composite structures using the finite element method (FEM). It is designed for
Universities, for advanced undergraduates, for graduates, for researchers, and
for practising engineers in industry. The text advances gradually from the
analysis of simple beams to arbitrary anisotropic and composite plates and shells;
it treats both linear and nonlinear behavior. Once the basic philosophy of the
method is understood, the reader may expand its application and modify the
computer programs to suit particular needs.
The book arose from four years research at the University of Stuttgart,
Germany. We present the theory and computer programs concisely and
systematically so that they can be used both for teaching and applications. We
have tried to make the book simple and clear, and to show the underlying
physical and mathematical ideas.
The FEM has been in existence for more than 50 years. One of the
authors, John Argyris, invented this technique in World War II in the course of
the check on the analysis of the swept back wing of the twin engined Meteor Jet
Fighter. In this work, he also consistently applied matrix calculus and introduced
triangular membrane elements in conjunction with two new definitions of
triangular stresses and strains which are now known as the component and total
measures. In fact, he was responsible for the original formulation of the matrix
force and displacement methods, the forerunners of the FEM. A distinct feature
of the present book is the consistent use of the so-called natural modes, first
proposed by Argyris in 1964, here applied to composite shells and structures.
In the present book the technique of the natural modes are described fully in
chapters 3, 5, and 6. In this way, the total displacements of an element are
defined by rigid body modes and the natural or strain modes. This specification
of the kinematics of an element leads automatically to the elimination of parasitic
phenomena like shear locking.
x Preface
An element in space has 6 possible rigid body motions: three

displacements and three rotations. An element with n degrees of freedom will
thus have 6 rigid body modes and n - 6 straining modes. Only the strain modes
involve strain energy. Thus in setting up the stiffness matrix of an element we
can concentrate on the (n - 6)x(n - 6) matrix relating to the strain modes and then
set up the full nxn matrix by using the relations between the n degrees of
freedom and the two sets of modes, 6 rigid body and n - 6 strain modes. We set
up this matrix on the computer itself.
Another distinctive feature of the method is that all the integrations
required in the calculation of the element stiffness matrices are performed in
closedform; no numerical quadratures are needed. This feature leads to more
savings in computation, and is particularly important when the FEM is being
used in each step of an incremental/iterative method, for nonlinear analysis, for
optimization or design, for example.
In the book we aim to show that this systematic use of natural modes and
closed form integrations leads to very general formulations and to considerable
savings in computer time. We show how the method applies to composite
beams, plates and shells, and equip the reader to formulate the method for other
applications.
Chapter I provides some results from continuum mechanics and forms
the Principles of Virtual Work and Potential Energy. Chapter 2 sketches the
history of the FEM, particularly in its displacement form. Chapter 3 introduces
the concepts of natural modes and natural stiffness. Chapter 4 introduces the
basic concepts of the theory of composites. Chapter 5 applies the natural mode
method to the analysis of composite beams in three dimensions. Chapter 6
introduces the composite plate and shell element. Chapter 7 show the
computational advantages of the method on a model problem. Chapter 8 deals
with nonlinear analysis of anisotropic shells. Chapter 9 discusses programming
aspects of the technique.
At this point we apologize for not including crash phenomena and the
inelastic deformations of composite structures. We realize these effects are of
the utmost importance for the behavior of composite systems. Hopefully, we
will be in a position to compose such extensions in a future edition of the present
work.
Acknow ledgments
We thank Lars Olofsson, Per-Olof Marklund and Anders Mattsson who

participated in the development of parts of this text through various re-
search projects. We also acknowledge Dina Apostolopoulou for her input
on Chapter 8.
We extend our gratitude and appreciation to our families for their sup-
port and encouragement during the many long hours of thinking, creating
and typing, and for letting us devote ourselves fully to the research effort
that this project required.
We are indebted to Professor Graham Gladwell of the University of
Waterloo and editor of the solid mechanics series at Kluwer Academic for
his valuable suggestions, and for the fine tuning of the manuscript. The
high standards set by Prof. Gladwell in this series have contributed to its
prestige and to the attention it has received worldwide.
Finally, we acknowledge our publishers at Kluwer Academic, Dr. Karel
Nederveen and Mrs. Catherine Murphy, assistant to Dr. Nederveen, for
making this text a reality and for the professional support offered to the
authors during the completion of this book.
Chapter 1
Some results from

continuum mechanics
1.1 The analysis of stress

A clear, full account of the analysis of stress may be found in Chapter
2 of Sokolnikoff's Mathematical Theory of Elasticity [1]. Here we give an
abbreviated account.
To specify the state of stress at a point P in a continuum we consider
an element f).s of a surface passing through P. Let n be a unit normal to
the surface at P, going from side 1 to side 2. The forces exerted by the
material on side 2 of f).S on the material on side 1 may be reduced to a
force f).T at P and a moment f).G. We shall assume, and this assumption
may be considered as having an experimental basis, that
. f).T
11m
A8-+0
- - T n,
f).S - lim
A8-+0 u
~S = 0, (1.1)
where Tn is a vector, called the traction or the stress vector across the
surface at P. In general Tn will not be in the direction n, but will have
components in each of the three coordinate directions. We write
Tn = O"nx i + O"ny; + O"nz k . (1.2)

In particular, if n = i, then the traction on the surface with normal i is
Ti = O"xxi + O"xy; + O"xzk. (1.3)
L. T. Tenek et al., Finite Element Analysis for Composite Structures

Springer Science+Business Media Dordrecht 1998
2 Some results from continuum mechanics
The quantity (J'xx, and the analogous (J'yy, (J'zz are called normal (or direct)
stresses; (J'xy, (J'xz etc. are called shear stresses.
It is found that the forces acting on a volume ~ V of a continuum may
be divided into surface forces caused by the tractions described above, and
body forces. The most common example of the latter is provided by gravity,
which exerts a downward force (p 9 ~ V) on an element of volume ~ V and
mass density p. We shall assume that if the body forces acting on a small
volume element surrounding a point P are equivalent to a force ~R at P
and a moment ~M, then
~R F
11m . ~M
AV = ,
~ V --+0 L.l. ~V
hm
--+0
AV
L.l.
= O. (1.4)
The vector F is called the body force per unit volume. We shall assume
that F is a continuous function of the coordinates (x, y, z), and write
F = Fxi + Fyj + Fzk. (1.5)
Consider a small rectangular box with sides 2a, 2b, 2c parallel to the
axes and centered at P. The moment in the z-direction, of the forces acting
on the element, is provided by the moment ~Mz of the body forces and
the moments of those surface stresses shown in Fig. 1.1. Thus equilibrium
of the element demands that
~Mz - b(4ac(J'yx) - (-b)(-4ac(J'yx) + (-a) (-4bc(J'xy) + a(4bc(J'xy) = O.

(1.6)
But the volume of the element is ~V = 8abc so that
~Mz
~V - (J'yx + (J'xy = 0, (1.7)
and as ~ V --+ 0, the first term tends to zero on account of assumption

(1.4), giving
(J'yx = (J'xy' (1.8)
The symmetry of the other shear stresses may be proved in a similar way,
giving
1.1 The analysis of stress 3
2a
<1yz ___ _
2b
~~~~~~====;1~::::::::::~<1~~/~/~/5/~----~~ x
__-1______<1YZ /
"y/
2c
/
/
y z
z
nM
oJ [>x
nonnal
y y
~ ________-r__________ ~t>X
Figure 1.1: Surface and volume forces on rectangular and tetrahedron ele-
ments
a zy = a yz , axz = a zx , a yx = axy (1.9)
Now consider a small rectangular tetrahedron element PABC with

three sides parallel to the coordinate planes and the fourth with normal
n = (l, m, n) as shown in Fig. 1.1. Consider the forces acting on the el-
ement in the x-direction. If the area ABC is !:l.S, and the height P N is
h, then the areas PBC, PCA and PAB will be l!:l.S, m!:l.S, n!:l.S, and
the volume of the tetrahedron will be !:l.V = h!:l.S/3. Equilibrium of the
element in the x-direction demands that
!:l.Rx + !:l.S(anx -laxx - ma yx - na zx ) = O. (1.10)
The surface stresses may, when hand !:l.S are small, be approximated by
their values at P. Divide equation (1.10) by !:l.S and consider the limit as
!:l.V -t 0
lim !:l.Rx = lim !:l.Rx.!:l. V = lim !:l.Rx.!: = O.

Ll V -to!:l.S Ll V -to
!:l.V !:l.S Ll V -to
!:l.V 3
(1.11)
Thus we obtain the first of the three equations
a nx= laxx + mayx + nazx ,

any = laxy + mayy + nazy , (1.12)
anz = laxz + mayz + nazz
Often it is convenient to use numerical suffices 1,2,3, instead of x, y, z.

Then the stresses are an, aI2, ... ,a33, and the symmetry relation (1.8) is
aij = aji (1.13)
In this notation, the components Ti of the traction across a surface with

normal VI, V2, V3 is given by (1.12) as
3
Ti = LaijVj. (1.14)
j=1
1.1 The analysis of stress 5
Often we use Einstein's double suffix convention and rewrite this
Ti = O'ijVj. (1.15)
This equation allows us to show that the O'ij are the components of a tensor
of order 2. If we change from one orthonormal triad of axes iI, i 2, i 3, to
another, i~, i~, i~, where, in double suffix notation,
'ti
./ = lij'tj,
' (1.16)
then the stresses O'~j in the new frame are
O'~j = lik1jiO'kl. (1.17)
Now consider the dynamic equilibrium of an arbitrary volume V bounded

by a surface S, in a continuum. The volume will be in dynamic equilibrium
under the action of the body forces and the surface forces; thus
Is T n dS + Iv FdV = Iv p ~:~ dV. (1.18)
Replacing Tn by its components Ti, and using Eq. (1.14) we find
r r r a2Ui
is O'ijVj dS + iv Fi dV = iv P at2 dV. (1.19)
where VI, V2, V3 are the components of the outward drawn normal on the
surface.
Now use Gauss' divergence theorem. In its usual, vectorial form it states
that
Is A .v = Iv divA dV. (1.20)
In double suffix notation this is
1S
Av
] ] dS -1
-
V
A] ,. dV
], (1.21)
where ,j = o/chj. Applying this to Eq. (1.19) we find
Iv (aij,j + Fi - pili) dV = O. (1.22)
But V is arbitrary so that the integrand must be identically zero, i.e.
aij,j + Fi = pili (1.23)
In the important, static case, these equations are
oaxx + oaxy + oaxz + Fx = 0,

ox oy OZ
oayx + oayy + oa yz + Fy = 0, (1.24)
ox oy OZ
oazx + oazy + oazz + Fz = O.
ox oy OZ
These equations are called the equilibrium equations.
1.2 The analysis of strain

Consider a rectangle in the x, y plane with sides dx and dy depicted in Fig.
1.2. Under the application of a horizontal load the element extends in the
x direction an amount dux. The elongation per unit length (axial strain)
is defined as
ux(x + dx, y, z) - ux(x, y, z) oU x

xx = dx ~ ox (1.25)
In the same manner, a load applied in the y direction causes strain yy and
a load applied in the Z direction causes strain zz. We have
OU x OUy oU z
xx -- ---,
ox yy -
-~,
oy zz = oz (1.26)
1.2 The analysis of strain 7
I _~ru, I
L.
I [ I I
I ,hou ,om, I
du
1= x co 1
dy
I_ <=1
I"'
dx
Figure 1.2: Axial and shearing strains
Now suppose that the left vertical side of the rectangle displaces in
the x direction by dux and the bottom horizontal side moves in the y
direction by du y as shown in Fig. 1.2. As a result, the left vertical side
rotates counterclockwise by an angle (h while the vertical displacement du y
induces a clockwise rotation (h. Due to these rotations the original1r/2
angle becomes 7jJ, and the change of angle is
7r
'Y = "2 - 7/J = 01 + O2 (1.27)
In the context of small deformation theory
OU
oy ~ O2 ,
tan02 = OU
tan 01 = -OX
y '" 01 x
- , (1.28)
We define as shear strains the quantities
1 (ou x ouy)
Exy =2 oy + ox '
1 (ou x ou z ) (1.29)
Exz =2 oz + ox '
1 (Ou y ou z )
Eyz =2 oz + oy .
The two shear strains Exz , Eyz are obtained considering angular displace-
ments in the xz and yz planes, respectively. We may now show that the
nine quantities
Ell = Exx , E12 = Exy , E13 = Exz , ... ,E33 = Ezz , (1.30)
form the components of a tensor. By this we mean that (just as we found

with the stresses) when we change from axes iI, i2, i3, to axes i~, i~, i;
related to iI, i 2, i3, by equation (1.16), then the strains are transformed
according to
E~j = likljlEkl. (1.31 )
We note that the strain tensor, like the stress tensor, is symmetrical, i.e.
Eij = Eji (1.32)
Note also that we may write equations (1.27), (1.30) in the succinct form
1
Eij = 2(Ui,j +Uj,i)' (1.33)
1.3 Strain energy 9
1.3 Strain energy

Consider a volume V with surface S in an elastic continuum. Suppose that
the volume is acted on by surface tractions and body forces. The rate at
which these forces do work is
~~ = Is T(Ui dS + Iv FiUi dV. (1.34)
Express Ti in terms of the stresses by using equation (1.15), then
TiUi = (J'ijVjUi. (1.35)
Transform the surface integral in (1.34) into a volume integral by using

Gauss' divergence theorem (1.21). We have
Is TiUi dS = Is (J'ijuil/jdS = Iv ((J'ijUi) ,jdV

(1.36)
= Is (J'ij,jUi dV + Iv (J'ijUi,jdV
so that
:: = Iv ((J'ij,j + Fi)Ui dV + Iv (J'ijUi,jdV. (1.37)
The dynamic equilibrium equation (1.23) allows us to write the first integral
as
Iv ((J'ij,j + Fi)Ui dV = IvpUiUi dV = ! Iv ~pU~dV,

(1.38)
and we notice that this is the rate of change of the kinetic energy
1 { .2
T = "2 }VPUi dV, (1.39)
of the volume. Thus

dE
dt
- -dT -
dt
-1 V
-Ui J- dV
a-ZJ, (1.40)
We can write the integrand on the right as
. 1. -+ 1 . {1- (u-
. - +u--
.)}
ZJ Z,J- -- -a--u-
a--u- 2 ZJ Z,J 2 JZ J,Z- -- a--
-a--u- ZJ 2 Z,J J,Z
(1.41)
= aijEij,
where we have used the symmetry of the stresses (aij = aji) and the relation
(1.33). Thus
dE
dt
-dT
- -
dt
-1 V
a-ZJ-iJ- dV (1.42)
Up to this point we have used the stress tensor with components aij and
strain tensor strain tensor with components Eij; both of these are symmet-
ric, and they have 6 independent components. For practical computations
it is convenient to string these components into vectors:
0' = {al a2 a3 a4 a5 a6} = {an a22 a33 a12 a13 a23}

e = {El E2 E3 4 E5 E6} = {En E22 E33 2E12 2E13 2E23 } (1.43)
Note that the components of e corresponding to the shear strains are dou-
bled, thus E4, E5, E6 are the engineering shear strains, denoted by ')'12, ')'13,
')'23 respectively
8u x 8u y
4 = 2E12 = -
8y
+ -8x = ')'12,
8u x 8u z
E5 = 2E13 = -
8z
+ -8x = ')'13, (1.44)
_ _ 8u y 8u z _
E6 - 2E23 - - +- - ')'23
8z 8y
This renaming has an important advantage: it allows the sum aijij in the
integrand in (1.42) to be written as an inner product of 0' and E
1.3 Strain energy 11
t
q e = 0'11:1.+ . .
0'21:2 + ... + 0'61:6
= 0'11 El1 + 0'22 E22 + 0'33 E33 + 20'12 E12 + 20'13E13 + 20'23E23
= O'ijl:ij (1.45)
There is one drawback to the use of q and e. These quantities are not
physical vectors, i.e. they do not change as a vector should (i.e. according
to O'~ = lijO'j) when the axes are changed.
With the new notation, we may write equation (1.42) as
dE _ dT = { oe dV = ( 0" Ol:i dV.

qt
(1.46)
dt dt 1v at lv' ot
Suppose now that there is a function F(l:l' 1:2, ... ,1:6) such that
of
O'i = Ol:i' (1.47)
then (1.46) may be written
dE _ dT = { of Ot:i dV =.!!.- ( FdV. (1.48)

dT dt lv Ol:i at dt lv
Integrating this equation between t = 0 and t = t, where t = 0 corresponds
to the natural state we obtain
E=T+U, (1.49)
where
U= iFdV. (1.50)
F is called the strain energy density, and U the strain energy of the vol-
ume. In infinitesimal elasticity theory F(l:l' 1:2, ... ,1:6) is a homogeneous
quadratic form in 1:1, 1:2, ... ,1:6 so that Euler's theorem on homogeneous
functions gives
8F
2F = -Ei = (7iEi, (1.51 )
8Ei
thus
1 f 1 f t
(1.52)
U = 2" 1v (7i Ei dV = 2" 1v u edV.
Reversing the steps from (1.52) to (1.34) we may deduce that, in the
static case,
1ft If If
U = 2" 1v u edV = 2" 1v FiUi dv + 2" ls TiUi dS. (1.53)
We prefer to use a different notation for the body forces and surface trac-
tions:
F=pv, T =Ps. (1.54)
With this notation we may write (1.53) as
1ft 1ft 1ft

U = 2" 1v u e dV = 2" 1vPvudV + 2" lsPsudS
(1.55)
1.4 The principle of virtual work

In section 1.3 we started with equation (1.34). This expressed the state at
which work is done by the body forces and surface tractions. The dynamic
equilibrium equations (1.23) allowed us to express the rate of work done by
(1.42). To obtain the Principle of Virtual Work we consider the work done
by the body forces and surface tractions on some virtual displacement~
these are displacements which do not themselves affect the stress distribu-
tion in the body. This virtual work is
6W = Iv p\,r6u dV + Is p~6u dS. (1.56)

An analysis exactly similar to that given in section 1.3 shows that we can
write
6W = Iv u t 6g dV. (1.57)
In addition to body forces and surface tractions, there may be con-

centrated forces R and concentrated moments M; these will do work on
virtual displacements 6u and rotations MJ to give the total virtual work
equation
6U=6W
Iv u t 6g dV = Is p~6u dB + Iv pi.t6u dV + R t 6u + M t 68 a.58)
This equation is the principle of virtual work. In words, this states that
Definition 1 The virtual work done by the body forces and surface trac-
tions on the virtual displacements, is equal to the work done by the stresses
on the virtual strains.
1.5 The principle of stationary energy

In infinitesimal elasticity theory, the strain energy density F is a quadratic
form, i.e.
lit
F = -C""E"E"
2 tJ t J = -g
2 Cg ' (1.59)
where C = (Cij) is a symmetric matrix of elastic coefficients. In this case,

we may identify the work 6W done by the stresses on the virtual strains as
the first order change in the strain energy, i.e.
6W = 6U. (1.60)
For
oU = o(~ Iv utgdV)
=~ Iv (u t + out) (g + og)dV - ~ Iv utgdV (1.61)
= ~ Iv utogdV + ~ Iv outgdV,
where, to find OU, we neglect the second order terms. But equations (1.51)
and (1.59) give
8F
(Ti = 8i = Cijj, (1.62)
so that
u = Cg. (1.63)
Thus
OUtg = OgtCg = gtCOg = UtOg , (1.64)
so that
OU = Iv UtOg, (1.65)
which, by equation (1.57), is oW.

If, as is usually the case, the body forces and surface tractions are
conservative, then they have a potential function Ue and
oW = -oUe . (1.66)
Thus we may rewrite equation (1.66) as
OUi + oUe = 0, (1.67)

where the suffix i denotes the internal strain energy. This states that the
total potential, of external stresses, body forces and surface tractions, is
stationary, i.e.
i5(Ui + Ue ) = O. (1.68)
The quantity
U = Ui + Ue , (1.69)
is called the total potential energy of the body.

In this brief sketch we have ommited many details: the regularity con-
ditions on the surface and on the integrands which ensure that Green's
theorem may be used; the conditions that the virtual displacements must
satisfy the geometric boundary conditions; how other effects, for instance
those due to temperature, may be included; etc. When all these and other
similar considerations are taken into account, we may enunciate the Prin-
ciple of Stationary Energy:
Definition 2 Of all continuous displacements satisfying the geometrical

boundary conditions, those that satisfy the equilibrium conditions make the
potential energy stationary.
For the stable structures considered in this book the potential energy
in this stationary configuration is always a minimum. For further details
on this and other energy theorems, the reader is referred to the books of
Przemieniecki [2], Tauchert [3] and Gould [4].
Chapter 2
A brief history of FEM
An analysis of complex structures and other systems in a matrix formula-

tion is now unthinkable without the finite element method. Our personal
belief is that the origins of such a rich and applicable method cannot be
attributed solely to one person or school of thought but rather to a syn-
ergy of various scientific developments at various research establishments.
The notion of geometrical division can be traced back to the Greek natural
philosopher Archimedes who in order to compute the area of a complex
shape divided it into triangles and quadrilaterals whose area could be eas-
ily computed; the assembly of the individual areas provided the total area
of the complex shape. More recently, Courant used variational and mini-
mization arguments for the solution of physical problems. Courant [5], and
Prager and Synge [6] had both proposed the concept of regional discretiza-
tion which is essentially equivalent to the assumption of constant strain
fields within the elements. The adaptation, however, and development of
these concepts for structural analysis and other physical and technical prob-
lems was not conceptually achieved until during and shortly after World
War II.
2.1 The matrix displacement method

During World War II the demand for more efficient aeronautical structures
and methods for the analysis of complex structural systems provided the
second author, John Argyris, with the incentive for developing the matrix
displacement method, a concise matrix representation of the equilibrium
equations governing a skeletal structure. It was wartime that necessitated
this sudden explosion of knowledge. In fact in 1944, towards the end of

18 A brief history of FEM
World War II, with the advent of jet propulsion, there suddenly arose the
necessity of developing high subsonic speed fighters and fighter bombers.
At the same time there appeared the first electromechanical calculating
devices; the advent of such machines in the United States, Great Britain
and Germany was a revelation to the engineers who were active in aircraft
engineering and were faced with the necessity of analyzing and designing
swept-back wings for the first modern high-speed fighters. In those days,
practically all aircraft calculations were performed by some kind of force
method in which forces or stresses appeared as the major unknowns. When
the swept-back planes were to be designed it became apparent that in
a structure which had practically no two parts at right angles, the force
method was most inconvenient and unpractical.
In those days, John Argyris had heard of the Meteor fighter designed
by Gloucester Aircraft as the answer to the German ME-262. The ME-262
had good aerodynamic performance, but suffered many failures due to the
poor structural design of its swept back wings; this was a problem which
the manufacturers could not overcome at that time. In those far-away days
of 1944 John Argyris was working wit~ HL Cox at the National Physical
Laboratory at Teddington, London. HL Cox was then the prime figure in
aircraft technology in Great Britain. In spite of the strict security precau-
tions, they spoke about the Meteor wings and the difficulty of analyzing
them. So, during the course of three brainstorming days and nights, Ar-
gyris realized that the force method was not suitable for this problem due
to the great difficulty in developing the self-equilibtating systems. Conse-
quently, he toyed with the idea of the displacement method and it suddenly
occurred to him that the triangle was incredibly well suited to these odd
swept-back wing structures -by using a number of triangles it was possi-
ble to discretize such high-speed subsonic structures! As the next step, an
elementary matrix code was designed. The first structure studied using
this new method was a simple wing model described by a set of 64 linear
equations. It was analyzed on the electromechanical calculator at the Na-
tional Physical Laboratory similar to a computing device being developed
at Harvard University in the United States. Needless to say, no software
was available to solve for the unknowns, so that simple, elementary solu-
tion techniques were conceived. In 2-3 days of frenzied work, displacements,
strains and stresses for a loading case were obtained. HL Cox, who was
aware of the on-going work, advised a direct comparison with experimental
tests for a simple wing model. All were surprised when it was discovered
that the maximum stresses did not deviate more than 8 to 9% from the
2.2 The finite element method 19
experimental ones! This astonishing revelation percolated to the offices of

the aircraft industry, and immediately a security clamp down was imposed.
By 1945 the breakdown of the continuum into triangular elements had
been accomplished, and engineers had started to apply the matrix displace-
ment method to the analysis of swept-back wings. It was not immediately
realized that these developments had led to the birth of the finite element
method, but the importance of the matrix displacement method was under-
stood and the notion of ideal design and analysis of such complex structures
by triangular components or elements was being grasped.
During the course of that work, it was realized that normal and shear
stresses were inappropriate for analysis involving triangles; it was simpler
to use three normal stresses in directions parallel to the sides of the triangle.
This came to engineers as no surprise since they were working daily with
strain rosettes that measured normal (direct) strains in three directions.
The knowledge of the data and simple transformation rules sufficed to yield
the classical cartesian strains and stresses. .
After the War, in 1953, permission was granted to publish the basic
idea of matrix methods, but without the major theoretical aspects of the
triangular element. Aspects of the triangular element were put in two inter-
nal reports of the Department of Aeronautics of Imperial College, London,
by Sydney Kelsey and S. Bagat in 1953 [7], [8]. In 1954-55, also with this
restriction in mind, a series of articles was published in Aircraft Engineer-
ing by Argyris and Kelsey [9], [10] presenting in a concise matrix notation,
unknown in those days, the interesting aspects of the unit load and unit
displacement methods, including thermal and other initial strains. Sydney
Kelsey also designed a wing-like structure with honeycomb panels with no
right angle. Tests performed on this structure in the experimental lab-
oratory of Imperial College revealed an astonishing accuracy of the stress
pattern under a number of loads. This secured the acceptance of the matrix
displacement method.
2.2 The finite element method

Essential contributions to the finite element method were made at Boeing
during the summers of 1952 and 1953 under the direction of M.J. Turner
[11] and led to a publication of an original paper [12]. As stated by R.
W. Clough in [11]: "Mr. Turner saw the need for an improved way of
taking account of the contributions of the wing skin to the stiffness of
airplane wings or arbitrary configurations, and he recognized that a Ritz-
type procedure could be used to evaluate the contributions of individual

skin elements if the wing were represented as an assemblage of such discrete
structural components". It was the work conducted at Boeing that provided
R.W. Clough with the inspiration for naming the method as the finite
element method. That name appeared first in a paper presented at the
1960 ASCE Conference [13]. In the 1960s, other researchers helped to
extend, disseminate and apply the finite element method.
There are other notable pioneering names: Olieg C. Zienkiewicz, Samuel
Levy, Harold C. Martin, Borje Langefors, Paul H. Denke, Baudoin Fraejis
De Veubeke, L. Brandeis Wehle Jr., Theodore H. Pian, Warner Lansing,
Bertran Klein, John S. Archer, Robert J. Melosh, John S. Przemieniecki,
Ian C. Taig, Richard H. Gallagher, Bruce Irons, etc. -the list does not
claim to be exhaustive.
2.3 The natural mode finite element method
In 1962 a new approach in the context of the matrix displacement method

was initiated by Argyris and Scharpf [14], [15]. Experimentation with prob-
lems involving large displacements, strains and inelastic behavior led to the
basic concepts of geometrical stiffness and the idea of natural modes. These
natural modes essentially represented invariant fields, including both rigid
body and pure straining measures, that were used to describe the elemental
kinematics. Since then this fundamental idea has evolved significantly, and
has facilitated the computer simulation of large and complex structures in
the linear and nonlinear regimes. In addition, it has provided an insight
into the fundamental behavior of structures undergoing small and large dis-
placements. It is now possible to create finite elements based on rigid body
and straining modes of deformation. The technique is particularly suited
to the creation of simple truss, beam, plate and shell triangular elements,
and tetrahedron volume elements for large scale and fast engineering com-
putations.
Although, over the years, many finite elements have been developed on
these principles, the natural mode method has only recently been applied
by Tenek and Argyris [16], [17] to the analysis of laminated beam and shell
composite structures through a general formulation able to treat, as special
cases, isotropic and sandwich plates and shells. The culmination of four
years of research conducted on this field is presented in this text in a simple
and comprehensible manner for linear and nonlinear statics.
2.4 The basic ideas of FEM 21
2.4 The basic ideas of FEM

In the article "A brief history of the beginning of the finite element method"
Gupta and Meek [18] presented summaries of the works of several authors
associated with the invention of the finite element method and its applica-
tion to structural mechanics. They discerned five groups of papers which
may be considered as the starting inspirations of FEM. They are the papers
by Courant [5], Argyris and Kelsey [9], [10], Turner, Clough, Martin and
Top [12], Clough [13], and Zienkiewicz and Cheung [19].
Courant
Courant [5] developed the idea of the minimization of a functional using
linear approximation over subregions. He specified the values at discrete
points similar to the node points of a finite element mesh. In his paper
he shows a mesh subdivision up to 9 approximate points to solve the St
Venant's torsion of a square hollow box. He considered the potential energy
U of the system and used the condition of stationary potential energy which
he expressed as
c5{V + U} = O. {2.1}
He specified the shear stresses in the shaft as first derivatives of a stress

function cp, and stated that if cp is described in terms of a number of dis-
crete parameters ai, the stationary condition {2.1} leads to a set of linear
equations
8{V + U} = O. {2.2}
8ai
Courant also applied the Rayleigh-Ritz method to create a functional for
cp with two unknowns.
Courant did not fully clarify his piecewise linear approximation to the
cp surface nor did he give any mathematical details of this approximation.
He clearly indicated, however, a procedure which could be used for the
minimization of the total potential energy for the torsion problem.
Argyris and Kelsey

In the series of papers "Energy Theorems and Structural Analysis" Argyris
and Kelsey [9], [10] developed the matrix theory of structures for discrete
elements. They developed the concepts of flexibility and stiffness and pro-
vided equations which have become standard in structural mechanics. They
include
flexibility: [F] = i[b]T[J][b]dV,

(2.3)
stiffness: [K] = i[a]T[k][a]dV,
in which [f], [k] are constitutive relations and [b]' [a] are stress-force and
strain-displacement relationships, respectively. They applied their theory
to a rectangular panel for the case of nodal displacements which vary lin-
early along the edges of the element and calculated its stiffness matrix; the
first element in plane stress using interpolation functions in terms of nodal
displacements was developed! The panel had an 8 x 8 stiffness matrix which
was expressed as a sum of shear and direct strain matrices:
[K] = [Ks] + [Kd]. (2.4)
Thus Argyris and Kelsey developed the rectangular panel stiffness matrix
in plane stress using element interpolation functions in terms of nodal dis-
placements. They also showed that, in the context of aircraft skin models,
the triangular panel behavior may well be approximated by using energy
minimization procedures which were shown later to provide a firm basis for
finite element formulation.
Turner, Clough, Martin and Top

The pioneering paper of Turner et al. [12] discusses the truss member and
derives its stiffness matrix in global coordinates in the form:
>.2 _>.2 >'11- ->.11-]

AE [ _>.2 >.2 _>'11- >'11-
[K] = L >'11- _>'11- 11-2 -11- 2 ' (2.5)
_>'11- >'11- _11- 2 11-2
where >',11- are member direction cosines, L is the member length and A its
area of cross-section. The paper continues with a discussion on rectangular
plate elements and then turns to triangular elements, stating that the latter
will be used as the basic building block for calculating stiffness matrices for
plates of arbitrary shape. It then proceeds to the study of the triangular
element in plane stress. It starts by assuming constant strains which are
integrated to yield the displacements u, v. Then, it expresses the relation
between stresses O"xx, O"yy and O"xy and the nodal displacements in the form
0" = [S]{8}. (2.6)
The nodal forces were obtained from the three stresses via
F = [T]{ O"} = [T][S]{ 8}. (2.7)
The element stiffness matrix was defined as
[K] = [T][S]. (2.8)
Note that equation (2.8) is equivalent to the second of equations (2.3)

developed by Argyris and Kelsey. The Turner paper also addresses the
question of convergence.
Clough
The name finite element method is attributed to Clough [13], [11], who
was a coauthor of the original paper by Turner [12]. In [11], [13], he out-
lined how he first invented the name finite element method because he
wished to show the distinction between the continuum analysis and the
matrix method of structural analysis. In [20] he outlined the research pro-
gram undertaken at Boeing Company in 1952-1953 for the calculation of
the flexibility coefficients for low aspect ratio wing structures for dynamic
analysis. He extended Turner's work from 1957 onwards; continued conver-
gence studies on stress components; and popularized the ideas of the finite
element method.
Zienkiewicz and Cheung

The development of the non-structural applications by means of minimiza-
tion of the total potential energy of a system is developed systematically
for the first time in the paper by Zienkiewicz and Cheung [21], in which
heat transfer and St Venant's torsion of prismatic shafts are analyzed. In
this paper they set up the approximation to the functional in terms of the
nodal values of the triangular domain, into which the region is subdivided.
The function minimization techniques, originally discussed by Courant,
were clarified by Zienkiewicz in 1965; this opened the way to the analysis
of field problems by the FEM. A first book on the FEM was published in
1957 [19] with new editions appearing in recent years [22].
Zienkiewicz expressed his personal views on the origins, milestones and
directions of the finite element method in [23]. He stated that the finite
element method was made possible only by the advent of the electronic
digital computer, and discussed the "variational" approaches via extremum
principles.
Thus if the strain field in an elastic continuum is defined by a suitable
operator S acting on the displacements u as
e=Su, (2.9)
with the corresponding stresses given as
(T = De, (2.10)
where D is a matrix of elastic constants, then the finite element solution
sought could be obtained by the minimization of the potential energy
II = ~ { eTDedn - { tT udI' - { bT udn. (2.11)

2 in irt in
The displacement field is approximated as
uh = N-u, (2.12)
where u are nodal values of u or other parameters satisfying prescribed

displacements on the boundary rt; t are the prescribed tractions on the
boundary rt; and b are the body forces. The functions N are given in
terms of the coordinates and are known as shape or basis functions. The
minimization of (2.11) leads to a set of discrete algebraic equations of the
form
Ku=!, (2.13)
where
K=LKe, Ke = r (SNfD(SN)dn,
Jn e
(2.14)
and
1= Lfe, Ie = r NTbdn + Jrq NT1,df'.

Jn e
(2.15)
The domains ne and r e correspond to elements into which the whole con-
tinuum problem is divided. Equations (2.14), (2.15) are used to generate
element stiffness coefficients and forces providing that the approximation
shape functions of Eq. (2.12) are defined on a local basis. These equa-
tions summarize the basic philosophy of classical finite element methods.
Zienkiewicz points out that such a derivation of the finite element procedure
is a particular case of the approach introduced much earlier by Rayleigh [24]
and Ritz [25]. The main difference lies in the use of the local shape func-
tions N which yield a banded structure of the assembled stiffness matrix
K and preserve the local assembly structure of the matrix equations.
Chapter 3
Natural modes for finite

elements
3.1 The concept of natural modes

In the natural mode method we express the deformation u(x, y, z) at any
point in a finite element as a linear combination of the nodal cartesian
displacements r e via
u (x,y,z) = C (x,y,z) re , (3.1)

(3xl) (3xn) (nxl)
where n represents the nodal degrees of freedom. A finite element e can

deform in n different modes. Then, the total deformation of the element
may be expressed as a linear combination of the imposed n modes so that
U = UI + U2 + ... + Un. (3.2)
If p is the vector containing the natural modes we write
wu2 n
:~];,;.
WI
PI] _
U = [ w! wv2 Wu
[
P2
-
W p.
(3xn)(nxl)
(3.3)
wI
w ww2
The natural displacements p produce nodal cartesian displacements r, viz.

28 Natural modes for finite elements
r = A p =?p=A-1r=ar. {3.4}
(nxl) (nxn)(nxl)
Therefore
u w p = ware, {3.5}
(3xl) (3xn)(nxl) (3xn)(nxn)(nxl)
from which we deduce that
C=wa. {3.6}
The elemental natural modes p will be always split in two parts; the nat-
ural rigid body modes Po and the natural straining modes p N. The latter
describe pure straining deformation.
3.2 The concept of natural stiffness matrix

In the framework of the natural mode method we develop all elemental ma-
trices in a natural coordinate system which is in harmony with the elemental
geometry; for truss and beam elements the natural coordinate is simply a
single direction a spanning the truss or beam axis; for a 3-node plane stress
triangle as well as the triangular plate and shell element, the natural co-
ordinate system, denoted by a(Jy, spans the three triangular edges. The
adoption of a natural coordinate system facilitates the assignment of invari-
ant deformation measures, the natural nodes, to describe the kinematical
field of a finite element.
One of the most important things when developing finite elements is to
ensure that rigid body motions do not produce strain energy. The natural
mode method handles this problem by separating rigid body and strain-
ing displacements -only straining modes enter the elastic stiffness matrix.
Thus, for a generalized nodal displacement vector p' containing rigid body
Po and straining modes p N, we demand a unique and reversible relation to
a Cartesian vector p so that
p' = [;;] = [:;] p = ap =? Po = aop, PN = aNP, {3.7}

3.2 The concept of natural stiffness matrix 29
p = a-Ip' = [Ao AN] [;;] = Ap' => A = a-I. (3.8)
To every set of nodal displacement, Cartesian p, or generalized (natural)

p', there exist load vectors P (cartesian) and Po, PN (natural rigid body
and straining generalized forces, respectively), so that the external work
can always be obtained as a scalar product of the displacements and loads.
From the invariance of virtual work in cartesian and natural coordinate
systems and using equations (3.7-3.8) we write
p t 8p = P't8pl = p'ta8p
-+ P at pI = a~Po + a~PN (3.9)
-+ pI = [;;] = Atp = [it] P, (3.10)
from which we deduce
PN = aNP, Po = Ab P ,
PN A~P, P=a~PN' pI = Atp. (3.11)
From equation (3.11) we see that the connection between natural and carte-
sian displacements as well as natural and global load vectors is provided by
matrices a, A. Usually these matrices are functions only of the geometry of
the element and can be easily constructed. Ignoring the presence of initial
strains, we write in the natural coordinate system
g = aNPN, (3.12)
where g is the elastic strain vector, PN is the vector of natural modes, and
aN is a strain operator matrix. Inserting (3.12) in the first integrand of
the expression of virtual work (1.58) we obtain
Iv u t 8gdV = Iv gtK8gdV = P't8pI. (3.13)
The right hand side includes surface, volume and concentrated loads refer-
ring to the natural coordinate system. Substituting equation (3.12) in the
middle and right integrands above
8P~{l UtaNdV] = 8pk[1v a1v Ka NdV]PN
Pb 8po + pk8PN (3.14)
Collecting terms, we find
Po=o, (3.15)
[Iv a 1vKaN dV]PN = Iv utaNdV * kNPN = PN.

(3.16)
Equation (3.15) states that the rigid body forces acting on an element must
equal zero.
We are now in the position to define the elemental natural elastic stiff-
ness kN as
kN = Iv [a1vKaN] dV, (3.17)
and the resultant natural loads as
PN = Iv [a1vu] dV, (3.18)
so that,
Po=o,
kNPN = PN (3.19)
Matrix k N is the natural stiffness matrix. Once formed, simple congru-

ent transformations transform it first to the local elemental and ultimately
to the global cartesian coordinates. This general methodology, briefly de-
scribed in this section, will be used extensively in the sequel for the deriva-
tion of the natural stiffness matrices for laminated composite beam, plate
and shell elements.
3.3 Natural modes for selected finite elements

The natural modes must satisfy all kinematic compatibility conditions at
the boundaries. They are not unique or complete but are always equal in
number to the difference between the global nodal degrees of freedoms and
the rigid body freedoms. We state now the natural modes of some finite
elements.
Truss finite element in space

The three dimensional truss element has 2 nodes and three cartesian
displacements (u, v, w)pernode, for a total of 6 degrees of freedom.
With respect to local cartesian coordinates x' y' z' attached to the
element center with axis x' directed along the element, the truss can
have 5 rigid body modes. These modes include three rigid body
translations in x', y', z', and two rigid body rotations with respect
to the y' and z' axes, respectively. Subtracting the rigid body modes
from the total degrees of freedom we are left with only one degree of
freedom -the latter is the natural straining mode PN. Therefore for
the space truss finite element the natural stiffness matrix is of order
(lxl)!
Beam finite element in space

The beam element in space has 2 nodes and six cartesian freedoms
(u,v,w,(},,'I/J) per node, for a total of 12 degrees of freedom. With
respect to local cartesian coordinates attached to the beam center
with axis x' directed along the element, it can move in space in 6
possible rigid body modes. These modes include three rigid body
translations in x', y', z', and three rigid body rotations with respect
to the x', y' and z' axes, respectively. Subtracting the number of
rigid body modes from the total degrees of freedom we are left with
only six degrees of freedom. Therefore, we must choose six strain
modes to describe the deformation of the beam element. For the
three dimensional beam finite element the natural stiffness matrix is
of the reduced order (6x6).
Plane stress triangle

The plane stress triangle includes 3 nodes and two cartesian displace-
ments (u, v) per node as global unknowns, for a total of 6 degrees
of freedom. With respect to local cartesian coordinates located at
the triangle barycenter the element can displace as a rigid body in
3 possible modes. These include two rigid body translations in x',

y', and one rigid body rotation with respect to the z' axis which is
perpendicular to the plane of the triangle. Therefore, three natural
strain modes suffice for the description of the element deformation.
These modes include axial straining of the three triangular edges. For
the plane stress triangular finite element, the natural stiffness matrix
is of order (3x3).
Plate and shell triangular finite element

The triangular plate and shell finite element has 3 nodes and six
cartesian displacements (u, v, w, fJ, </J, '1/1) per node, for a total of 18
degrees of freedom. The element can move in space as a rigid body
in 6 possible modes. They include three rigid body translations, and
three rigid body rotations. Subtracting the six rigid body modes
from the total 18 degrees of freedom we are left with only twelve
degrees of freedom. Therefore, twelve strain modes must describe the
deformation of the plate and shell finite element. In this case the
natural stiffness matrix is of the reduced order (12x12).
Tetrahedron finite element

The tetrahedron finite element has 4 nodes and three cartesian dis-
placements (u,v,w) per node, for a total of 12 degrees of freedom.
The element can move in space as a rigid body in 6 possible modes
which include three rigid body translations, and three rigid body ro-
tations. Subtracting the number of rigid body modes from the 12
degrees offreedom we are left with only six degrees offreedom which
must describe the deformation of the tetrahedron finite element. The
six natural modes are simply the extensions of the four edges of the
tetrahedron, and its natural stiffness matrix is of order (6x6).
Figures (3.3-3.3) show the straining modes for the selected finite elements.
We stress that once the natural stiffness matrices are computed, simple
congruent matrix multiplications transform them to the local and global
cartesian coordinates.
3D truss element
("2' v 2 'W 2)
2
2
PN extension
z.w
x, v
plane stress triangle
I ( "I' vI)
x, "
2/
\3
(")' v)
2/ /
) ' ,~ ==--- )3
3~'Y1II
1/2'Yry
Figure 3.1: Strain modes for truss and plane stress triangular finite elements
1I2p
I NlI
~
~
EI
1/2pNl
I extension
r~
1I2p N2
a symmetrical bending in x-z
[ Mode 3
II2P~
'v-I~ ~q _-.. ...."m.'

0
m,;" ,
0 m 0 _.._-..
112Pm
B_-----;Jt..__ y
8
,../ ---j ,/ ' ,
1/2PN4 symmetrical bending in x-y
1I2PN4 --- ,/' (........ --i>x
<
[ ModeS
112PN5
.",/ ,-- _--_
\
1 ;'
,,/
~Y
112PN5 8 antisymmetrical bending in x-y
--'---------
< ,"
-----
\
'-
fI"~ ~I2P
i 8
1/2PN6 Mode 6
~
torsion about x
l/
Figure 3.2: Strain modes for a space beam finite element

1I2Y,a
~ --<t,:
)3'Ytjl ...{')-- { {
i ;
::::;;::000
1I2Yta
1/2YI"(
1I2"';a
Figure 3.3: Strain modes for a plate and shell triangular finite element
Z,W
x, v
2
(u 2,v2 ,w2)
.x.u
1
(ul'vl' WI)
PNI - extension of side 23 PN2 - extension of side 13 PN3 - extension of side 12
2 2 2
PN4 - extension of side 34 PNS - extension of side 24 PN6 - extension of side 14
2
2 2
Figure 3.4: Strain modes for a tetrahedron finite element

Chapter 4
Composites
4.1 Fundamental concepts

Composite materials are one of the great technological advances of the last
half of the twentieth century [26]. By the term composite we usually refer
to materials that are combinations of two or more organic or inorganic com-
ponents, of which one serves as a matrix and the other as fibers. The fibers
provide virtually all the strength and stiffness. The purpose of the matrix
is to bind the fibers together and keep them in proper orientation, transfer
the load to and between them and distribute it evenly, protect the fibers
from hazardous environments and handling, provide resistance to crack
propagation and damage, provide all the interlaminar shear strength of the
composite, and offer resistance to high temperatures and corrosion. Poly-
mer composites are defined as reinforcement fibers supported by a polymer
matrix. Individual fibers are usually referred to as filaments; sometimes the
fibers are called the reinforcement and the matrix the binder. In structural
polymer composites, the fiber is stiffer and stronger than the matrix. The
most powerful concept behind composites is that the fibers and the matrix
can blend into a new material with properties that are better than those
of the constituent parts. In addition, by changing the orientation of the
fibers, we can optimize composites for strength, stiffness, fatigue, heat and
moisture resistance, etc. It is therefore feasible to tailor the material to
meet specific design needs. Furthermore, a preferred fiber orientation may
be used to increase the modulus and strength well above isotropic values;
the resulting lightweight material may have much higher strength/weight
ratios than conventional materials. By changing different matrix and fiber
types, we can construct composite structures for engineering applications

38 Composites
and modern industries. Resistance to certain temperatures can also be

achieved. The reinforcing pattern can range from unidirectional to random
dispersion, to woven shapes, or to various combinations of these.
Composite materials and structures can be found in many natural and
biological structures. Wood is the most common natural composite mate-
rial. Plywood was used by ancient Greek and Egyptian civilizations who
realized that a specific arrangement of wood could yield superior strength as
well as resistance to thermal expansion and moisture. Our skeleton, muscles
and bones all use the principle of reinforcement. Modern composite ma-
terials are used for weight sensitive structures, and for applications where
high (stiffness/weight) ratios are required. Fiberglass was widely used in
the 1950s for boats, and for automobiles such as the Chevrolet Corvette.
The first aircraft to use composites was the Boeing 707 which is about two
percent fiberglass [26]. The Boeing 777 has 10% of its structural weight in
composites; the British Aerospace-McDonnel Douglas AV-8B Harrier has
25%. This trend is expected to continue, with some experts predicting that
21st Century aircraft may contain as much as 65% advanced composites.
High speed flight leads to intense aerodynamic heating; to withstand high
temperatures, the aircraft skin must be made of some form of composite
material.
The applications of composite materials are not restricted to aviation.
They are used by many industries such as sporting, electronics, nuclear,
shipping, automotive, as well as for biomedical applications such as medical
implants and joints. The application areas are expected to widen signifi-
cantly in the future.
Reinforcing fibers
In fiber reinforced composites, the fibers provide virtually all the strength
and stiffness. The most widely used fibers include the following:
Glass: Glass fibers have an effective balance of mechanical, chemical,

and electrical properties at a moderate cost. They are widely used for
aircraft panels, rocket motor cases, helicopter rotor blades, pressure
vessels, sporting equipment, etc.
Carbon: Carbon fibers are the most popular and widely used rein-
forcing fibers in high-performance composite structures. They are
commonly made from rayon and polyacrylonitrile precursors (PAN).
4.2 Basic material unit 39
Aramid: Aramid fibers are used in many applications due to their

high strength. However, they tend to absorb moisture, and display
relatively poor adhesion to matrix materials.
Boron: Boron fibers were first developed in the early 1960s and were
the first reinforcements for advanced composites. They are of high
modulus and high strength, but expensive due to the high raw mate-
rial costs. The United States Air Force F-15 and Navy F-145 fighter
planes both use boron fiber reinforced composites as structural com-
ponents.
Matrix materials
Common matrix materials are as follows:
Epoxy: Epoxy thermoset resins are the most commonly used matrices
in advanced composite materials. Although limited to a maximum
service temperature of 120C, they provide an almost unbeatable
combination of characteristics such as handling, processing flexibility,
mechanical properties and cost.
Bismaleimide: Bismaleimide thermoset resins (BMI) have excellent

composite properties, but unfortunately they are fairly brittle. Their
damage tolerance is comparable to epoxy resins.
Polyimide: Polyimide thermoset resins produce good quality, low

void-content composite parts, but pose processing difficulties, and
are therefore relatively expensive.
4.2 Basic material unit

We proceed now to the definition of the basic forming material block,
and provide some nomenclature that will be used later. Fig. 4.1 shows a
basic unit of composite material in which unidirectional fibers are embed-
ded in a matrix. As depicted in the figure, the resulting material is stiff and
strong along the fiber direction, but soft and weak in the two transverse
directions. For a bidirectional reinforcement, i.e., a woven fiber pattern,
there is stiffness in two plane directions, but the superior directional stiff-
ness is lost. Often designers combine unidirectional layers with woven plies
to tailor the material stiffness, and minimize damage such as delamination,
or to protect the material from high temperatures.
40 Composites
soft and weak basic unit of material

I
I
........
stiff and strong

....
soft and weak
element in material and cartesian coordinate systems
3,z
local coordinate
/2~x +9,))
x
fiber coordinate - e
Figure 4.1: Basic unit of unidirectional composite material
Turn to Fig. 4.1. We shall always choose the coordinate system 123
so that the 1 axis lies along the fiber direction and axes 2 and 3 are di-
rected perpendicular to it. This coordinate system will be called the ma-
terial coordinate system and will differ for each layer, depending upon the
orientation of the fibers. When multiple plies are present -as for a compos-
ite laminate- a second coordinate system is chosen with respect to which
all material transformation and homogenization procedures (through the
thickness integrations etc.) are performed. This coordinate system will be
labeled xyz, and called the local coordinate system (when a global coordi-
nate system is also used the local coordinate system will be x'y' z'). For a
beam element the local coordinate system is placed at its center. Then, as
shown in Fig. 4.1, the material axis 1 forms, for each ply k, an angle Ok
with the local axis x, the latter being common to all layers. This angle is
4.3 Laminates 41
A (-45/ 45/ 0 /90 ) laminate
k~~~~7 9=-45
,/\\\\\7
9=45
.x
y~ 7 9=0
////7 9=90
Figure 4.2: An exploded view of a (-45/45/0/90) composite laminate
taken positive if measured counterclockwise starting from the x axis. The

direction (-(h) is said to be the opposite of ((h).
4.3 Laminates
A structural laminate is created by stacking together many plies of
fiber reinforced layers. The different layers of lamina are permanently
bonded under heat and pressure using a hot press or autoclave. There are
many types of composite laminates; carbon matrix-carbon fiber compos-
ites (CCC), metal-matrix composites (MMC), ceramic-matrix composites
(CMC), titanium-matrix composites (TMC), etc. Some of these laminates
combine the advantages of both metallic and non-metallic technology. For
example, silicon carbide, boron carbide, alumina and graphite fibers can be
combined with a metal matrix made of aluminum or magnesium to form a
metal matrix composite stronger, stiffer and lighter than its constituents.
The stiffness of the laminate is obtained from the properties of the con-
stituent layers; the laminate can resist loads in several directions. Using
various theoretical methods such as the classical lamination theory (CLT)
42 Composites
[27], we may derive the stiffness coefficients of the laminate from the stiff-
nesses of the individual plies. Usually the stress strain relations for a thin
lamina of an orthotropic material are written down, and subsequent trans-
formation and integration procedures yield the stiffness entries of the whole
laminate. Figure 4.2 displays an unbonded (-45/45/0/90) laminate made
from four layers with different fiber orientations.
The intrinsic mechanical behavior of composite laminates offers tremen-
dous possibilities. Figure 4.3 shows two composite laminates, a (0/90/90/0),
or simply (0/90)5, and a (90/0/90/0), or (90/0)s. Both consist of the
same geometry, four layers of the same thickness, and are subjected to the
same axial force N and end moments M. The figure shows that when
the same axial load is applied, both stretch the same amount. However,
when the same moment is applied, the laminates behave differently, with
the (0/90/90/0) exhibiting a stiffer response than the (90/0/0/90) plate.
This illustrative and interesting example shows the great advantages of
structures made from composite materials; by varying the fiber orientation
we can alter and optimize the mechanical response of the structure un-
der certain loadings. For example, the three laminates (+45/ - 45/0/90)5'
(+45/0/ - 45/90)5, and (+45/90/ - 45/0)5, have the same number of lay-
ers and fiber orientations but different lamination schemes; the different
laminations will lead to different behaviors under various loads.
4.3.1 Special laminates

For many composite laminates, there are cases for which the stiffnesses
take on simplified values. These special laminates are particularly popular
in practice. The following special cases are recognized:
Symmetric: A laminate is symmetric if, for every layer on one side

of the mid surface, there is another layer with identical thickness,
material properties and fiber direction, at the same distance from
the mid surface, but on the other side. A (+45/ - 45/0/90/90/0/ -
45/ + 45), or simply (+45/ - 45/0/90)5' is an example of a symmetric
laminate.
Balanced: A laminate is balanced if for every layer with a spe-

cific thickness, specific material properties and specific fiber direc-
tion, there is another layer somewhere in the laminate with the same
thickness and same material properties but opposite fiber orientation.
Both (0/45/90/ - 45) and (0/ - 30/60/30/90/ - 60) are balanced
4.3 Laminates 43
a (0/90/90/0) laminate
M
N~
..... \
-- ... .
.. . .

N~~ ...... :(~:~~ "
.. . . ....
....N
N
identical extensions of the 0/90/90/0) and (90/0/0/90) laminates

f---l f---l
.. . ..
21
....N
N
.. ..
M
M( (0/90/90/0)

)

i-"'i
M
)
differences in bending deformations
.--.
Figure 4.3: Axial and bending deformations of two cross-ply laminates
laminates. (Note that the opposites of () = 0 and () = 90 are () = 0,

() = 90, respectively!)
Angle-ply: If every layer has either angle () or angle -(), the lami-
nate is said to be an angle-ply laminate. An example of an angle-ply
44 Composites
laminate is (30/ - 30/30/ - 30), or simply (30/ - 30h.
Cross-ply: If every layer has either () = 0, or () = 90, the laminate

is said to be a cross-ply laminate. A (0/90/0/90), or (0/90h, is an
example of a cross-ply laminate.
Quasi-isotropic: If every layer has either () = 0, () = 90, () =

+45, or () = -45, symmetrically arranged with respect to the mid-
dle surface, the laminate is said to be a quasi-isotropic laminate.
(+45/ - 45/0/90)8' (+45/0/ - 45/90)8 and (+45/90/ - 45/0)8 are
all quasi-isotropic laminates. Note that a quasi-isotropic laminate is
always balanced and symmetric.
4.4 Micromechanics and macromechanics

A composite material is a solid heterogeneous medium -a mixture of
several media of different physical properties with well defined interfaces.
This heterogeneous medium can be regarded as a medium with piecewise
constant properties. As shown previously, the basic material unit is made
of two or more constituent materials, and so is the lamina. Some queries
arise:
1. Can we predict the macroscopic (i.e. large scale, or overall) physical

properties from those of the constituent materials?
2. Can the proportion of the constituent materials be varied to achieve

a desired stiffness and strength for the composite? If it can, how can
it be done best?
3. Can we analyze the internal fields in the heterogeneous medium?
4. Can we replace the piecewise constant heterogeneous medium by a

homogeneous one by somehow smearing the heterogeneity?
These questions show that there are two basic problems: determine the
lamina properties by the procedures of micromechanics; use these proper-
ties in macromechanics for a global analysis of the continuum. In the first
(micro) problem many difficulties arise due to uncertainties which may ne-
cessitate the adoption of stochastical or statistical methods. The objective
of most approaches is to determine the moduli of a composite material from
the moduli of the constituent materials, or determine the strengths of the
4.4 Micromechanics and macro mechanics 45
unit cell ~
matrill
~ interface
fiber
ply with unidirectional fibers pi)' with woven fibers l

1
1
l--I~=--)
I~
~ supersonic tmnspon
~ ~~'
slruClure l
Figure 4.4: From micro to macro-structures
composite in terms of the strength of its phases. Some basic approaches

include mechanics of materials, and elasticity theories based on the repeti-
tion of a unit cell or some other representative volume, assuming that there
is a perfect bond between fibers and matrix, which mayor may not be true.
46 Composites
Frequently, the micromechanical theories are validated with experimental

work. Figure 4.4 indicates the hierarchy structural evolution from a unit
cell through a lamina to the global composite structure. For more details
concerning micromechanics and macromechanics as well as more techni-
cal aspects of composite materials and structures the interested reader is
referred to the texts of Jones [27], Whitney [28] and Calcote [29].
Chapter 5
Composite beam element
5.1 Introduction
Beams are autonomous or secondary load bearing members of many struc-
tures. They are used extensively in the formation of linkages, shafts and
frames and as reinforcements on plate and shell panels. They are also used
in robotics and high speed machinery. Laminated composite beams are in-
tegral parts of lightweight structures that require high strength over weight
ratios. In general, laminated beams have been studied to a lesser degree
than laminated plates and shells. This may not come as a surprise consid-
ering the fact that some aspects of the theory of beams are more intricate
than aspects of plate and shell theory. For example, the geometrical stiff-
ness of a shell element, required for buckling and large deflection analyses,
is sufficiently developed using membrane stresses. However, for beam ele-
ments other forces or moments may contribute to the geometrical stiffness,
albeit not significantly. In the latter case, the geometrical stiffness is more
difficult to derive than for plate and shell elements.
The analysis of beams is usually based on four approaches; classical
theory of elasticity [30]; theory of mechanics of materials [31], [32]; and
variational methods and strain energy statements [29]. For practical engi-
neering computations of three dimensional beam structures, these models
are usually followed by associated finite element implementations; these
lead to various beam elements with or without transverse shear deforma-
tion. Most of the literature is devoted to the development of isotropic beam
elements. Laminated beams have received less attention. Various straight
and curved laminated beam finite elements were developed by Venkatesh
and Rao [33], Yuan and Miller [34], Oral [35], Chan and Yang [36].

48 Composite beam element
Most isotropic and composite beam finite elements are based on clas-
sical finite element methods in conjunction with classical, first or higher
order lamination theories. These models often lead to parasitic phenomena
such as shear locking, rigid body failure, zero-energy modes, etc. Most
require some sort of stabilization. These phenomena also occur with high
anisotropy and the low ratios of axial to shear moduli encountered in com-
posite materials. When higher order approximations are adopted some of
these problems may vanish, but then the delicate balance between economy
and accuracy is often sacrificed.
In developing a beam finite element we follow a different approach than
classical methods by adopting invariant deformation measures, the so-called
natural modes, to describe the elemental kinematics. The natural modes
include both rigid body and straining components [37]. We assume that
the 12 cartesian freedoms produce both rigid body motion and pure defor-
mation. Only the modes contributing to the pure deformation are used in
the elemental stiffness matrix. Thus, the order of the problem is automati-
cally reduced. Some similar underlying concepts applied to the analysis of
composite structures can be found in [16], [38]. For the composite beam
finite element in space we will select 6 straining modes (equal to the total
cartesian degrees of freedom minus the rigid body freedoms), and using
concise energy and work statements will develop a 6x6 sparse natural stiff-
ness matrix. Subsequent congruent transformations will refer this matrix
to the local and global cartesian coordinates. This approach leads to great
savings in computing time since most CPU time is spent on algebraic op-
erations such as vector and matrix multiplications [39]. The approach has
also significant potential for vectorization/parallelization.
5.2 Kinematics
Figure 5.1 displays a cross section of a beam element under the action
of edge moments. Our basic assumption is that the vertical displacement
w varies only with x. Then, symmetrical considerations lead to v = 0.
Using these assumptions, we express the displacement field within the beam
element as
u = u{x, z),
B
v == 0, By = 0, (5.1)
w = w{x).
5.2 Kinematics 49
For a beam element one may also adopt the usual assumption a yy = azz =
o.
We begin by considering only bending in which the normals remain
straight and normal after deformation, and derive expressions for the axial
displacement u and the axial strain 'Yxx due to an applied moment M. Turn
to Fig. 5.1 and observe that at a distance z from the middle surface the
beam is displaced horizontally by the amount
u= d = -ztan(}. (5.2)
We need to express the angle (} as a function of the displacement w. Ob-

serving the second illustration in Fig. 5.1 we have
!:1w
tan{l = . x + !:1uo (5.3)
Dividing the top and bottom of the right hand side by !:1x we find
8w
tan {I = Ox,
1+ x
10 (5.4)
or
8w
tan {I = Ox 8w (5.5)
1 + xo ::: 8x
If {I is sufficiently small, that is
8w
{I 1, then {I = 8x. (5.6)
In deriving this relation we have assumed that the axial displacement of

the beam Uo due to the applied moment is small and negligible. We derive
now an expression for uo. Using the Pythagorean theorem
(!:1X)2 = (!:1w)2 + (!:1x)2 + 2(!:1x)(!:1uo) + (!:1uO)2. (5.7)
Dividing both sides by (!:1x)2 and rearranging terms, we find

0 P Uo
~ x,U
' w(x)
M
( .....w = W(X)
................. 11-- M
y, v
z, w ~ /1 --....::.:::- r l i - .
)
d = z tane
original length
z=o
o ( P Q x, U
.ix
wi Iw+.iw
U o P'
----- 7/ ~
U o +.iuo
final length
z, w
Figure 5.1: Deformation of a beam cross section
~UO = _~ [(~W)2 (~UO)2] = (5.8)

~x 2 ~x + ~x ExO,
or
5.3 The beam element 51
~UO(I+!~UO) =-!(~Wr. (5.9)

~x ~ 2~x
small
Ignoring the second order term, we obtain
xO -
_ ~uo '"
~x -
_!2 (~W)2
~x
= O((p),
I
(5.10)
Uo 10 -
1
1___
2
J(~W)2
_
~x
dx.
o
We deduce from these equations that the strain xO of the middle plane is
of order (j2 and therefore quite small in the context of linear theory. In
classical theory we assume that Uo ~ 0 (~x* = ~x). Now the expression
for U becomes
dw dw
u( x, z) = Uo - z tan 0 ~ Uo - z dx ~ - z dx . (5.11)
This displacement gives rise to the direct axial strain
ou(i, z) d2w(x)
'Yxx = ox = -z dx 2 (5.12)
5.3 The beam element

A 2-node three dimensional beam finite element is shown in Fig. 5.2. The
element has 12 degrees of freedoms (6 per node) referred to a global coor-
dinate xyZj these are grouped in the vector
p = [Ul VI WI 01 11 'l/Jl U2 V2 w2 O2 12 'l/J2]'

(5.13)
Here u, V, w represent the translational degrees of freedom and 0, 1, 'l/J
denote the rotational degrees of freedom. These freedoms can refer either
to a local or to a global cartesian coordinate system which are related

through submatrices containing direction cosines.
The element is assigned a local cartesian coordinate system x' y' z' placed
at its center to which the following cartesian degrees of freedom belong:
p = [U1 'ih W1 01 1 1/;1 U2 V2 W2 O2 2 1/;2].

{5.14}
We also adopt a natural coordinate a directed along the beam axis. It

coincides with the local cartesian axis x'. The local cartesian degrees of
freedom are transformed into natural invariant rigid body and straining
modes Po, PN, respectively, so that a unique and reversible relation exists
between the natural modes and the local and global degrees of freedom in
the sense
P ~ p ~ Po ,PN PN = P- Po {5.15}
(12x1) (12x1) (6x1) (6x1)
For a multilayer construction, an additional coordinate system is defined

for every layer (k), the 123 material coordinate system with axis 1 along
the principal reinforcement direction and axis 2 perpendicular to it. Note
that material axis 3 is parallel to the local cartesian axis z'. Then, for
every layer k, axis 1 forms an angle (h with the local axis x' (see Fig. 5.2).
This angle is positive if measured counterclockwise from x'. Otherwise is
negative. The laminated beam element BEC comprises 12 cartesian degrees
of freedom but the actual number of straining modes is
12 cartesian d.o.f - 6 rigid body d.o.f = 6 straining modes.

(5.16)
5.4 Natural rigid body modes

Figure 5.3 displays the rigid body modes for the beam element. These
modes include three translations and three rotations in space which are
grouped in the vector
Po = {POl P02 P03 P04 P05 P06 }. (5.17)

laminated beam elemem

loW
'V
y. "
e x. u
z'.3
z', w
2
Z'k+1
"I
xyz global coordinate

x 'y'z lOCAl coordinate
123 material coordinate
u narural coordinale
~ _ _ _ _,----.
~ _
~."..,- ~ eL
~~ " .t'
Figure 5.2: Laminated beam element in space
To the rigid body modes correspond rigid body forces and moments com-
ponents of the vector
Po = {POI P02 P03 P04 P0 5 P06}. {5.18}
The three translational modes are deduced from a perusal of Fig. 5.3 as
follows:
~, EI
, ,/ '
El
z
2
I
X"
2E1
2
~ 2
Figure 5.3: Natural rigid body modes for the beam element
Ul = POI,
U2 = POl,
1
VI = P02 - "2P06,
1 (5.19)
V2 = P02 + "2P06,
1
WI = P03 + "2P05,
1
W2 = P03 - "2P05,
from which we obtain
1
POI = "2(Ul + U2)
1
P02 = "2(Vl + V2)
1
P03 = "2(Wl + W2) (5.20)
1
P05 = y(Wl - W2)
1
P06 = y(V2 - vd
The fourth rigid body mode is a rotation with respect to the x axis (a
twisting movement). It is expressed as
1
P04 = "2(91 + 92 ) (5.21)
From these relations we construct the matrix ao which extracts the natural
rigid body modes from the local cartesian degrees of freedom
1 1
2 2
1 1
2 2
1 1
2 2
1 1
(6xl2)
iio = '". 1
2
-7
1
2 1 (5.22)
1
7 1
-7 7
or using compact matrix notation
Po = iio p, (5.23)
(6xl) (6xl2)(l2xl)
where p is the vector of the local cartesian degrees of freedom. Thus, when
the local nodal degrees of freedom are available (they can be easily deduced
from the global displacements and rotations) we can explicitly define the
rigid body modes. The element rigid body modes can also be extracted
from the global nodal degrees of freedom. If
p = T04 p, (5.24)
(l2xl) (l2xl2)(l2xl)
where p, T04 represent the vector of global nodal displacements and matrix
of direction cosines, respectively, equation (5.23) becomes
Po = iio P = iio T 04 P (5.25)

(6xl) (6xl2)(l2xl) (6xl2)(l2xl2)(l2xl)
It is thus possible to extract the natural rigid body modes from both the
local and global cartesian degrees of freedom.
5.5 Natural straining modes

Figure 5.4 illustrates the selected natural straining modes for the beam
element and Fig. 5.5 depicts the corresponding generalized natural forces
and moments. The natural straining modes are grouped in the vector
5.5 Natural straining modes 57
PN = {PNI PN2 PN3 PN4 PN5 PN6 }, (5.26)

(6xl)
and the natural forces in the vector
PN = {FN M M1 M' M1 MT}. (5.27)

(6xl)
The natural straining modes and forces are related via the natural stiffness
matrix,
I PN = kNPN f (5.28)
We will use a non-dimensional coordinate ( which is placed at the beam

center spanning
(= 2x I I
-1 ~ (~1, --2-
< x <-.
-2
(5.29)
1'
Also
o{) o{) o( 0 0 2
- = --::::} - = - - . (5.30)
ox o( ox ox o( 1
The natural stiffness of the beam element will be based only on defor-
mations and not on rigid body motions. The stiffness kN corresponding to
these deformations is denoted the natural stiffness matrix.
5.5.1 Mode 1: Extension

The first natural mode is a unit extension PNl. Half of it is assigned to the
left beam edge and half to the right as shown in Fig. 5.4. The displacement
along the beam is written as a linear function of the coordinate ( as
1
u{() = "2 (PNI. (5.31)
112pNl r::::7i
~
B
I I 112pNl
I I extension
I
a
Mode2
1!2p N2
symmetrical bending in x-z
i
I Mode3
II2P~
112P~4::\ b ~~
~
antisyrnmetriCalbe~dinginx-z
\:Y
112PN2 _ transversesheanng
+ ~
'l'---~
Il2pJ.n.
1!2PN4
////)
I
_----
Mode4
_----.tly
<",/
I
,/' --....
C",
1I2PN4
a symmetrical bending in x-y
B
1I2PN5
-- )
..,.......
-----\
1 /
,"
,,"
~y
112PN5
'> ....
a antisymmetrical bending in x-y
---
< I _;a
\ ...... _--
a
1I2PN6 Mode 6
fI "JI ~I2PN6
~ t",," ~
torsion about x
Figure 5.4: Natural straining modes for the beam element

[M~
B
FN FN
.q-- --i> extension
I M~21
B
t~3
symmeUicaI bending in x-z
I I M~ 3 g antisymmeUicaI bending in x-z

L.;J transverse sbearing
(~ ;>qt XW." MA
a
M~_=) _____----___('",~
r~~~]
symmeUical bending in x-y
[~-;;;;;J
r: 2
MA
<
........ j . . _----\ MA~
\. . _-----'>_ . . '
I
2
8 antisymmeUicaI bending in x-y
r~6J
8
~\ l
torsion about x
Figure 5.5: Generalized natural forces

The displacement u gives rise to the axial strain
au au a( _ PNI = a f PNl, 1
I~x = ax = a( ax - l a -
- -l
(5.32)
5.5.2 Mode 2: Symmetrical bending in the x - z plane

Figure 5.4 illustrates the second natural straining mode which is a symmet-
rical bending mode in the x-z plane and is denoted PN2. Essentially this
mode is a unit rotation; 1/2pN2 is assigned at the left node and 1/2pN2 at
the right. Then the elastic curve takes the form of the quadratic polynomial
l l
w(x) = ax 2 + bx + c, x: -2' --+ 2" (5.33)
Using the set of boundary conditions
l l
w(-2') =0, w(2') =0,
(5.34)
'( l 1 , l) 1
w -2') = 2'PN2, w (2' = -2' PN2,
we have
1 2 l
w( x) = - 2l x PN2 + 8 PN2 (5.35)
Using
l
x = 2''' (5.36)
we may write (5.35) as
w(()
l
= 8(1- ( 2) PN2 (5.37)
5.5.3 Mode 3: Antisymmetrical bending and transverse shear-

ing in the x - z plane
The third mode is a unit antisymmetrical rotation PN3 which includes both
bending and transverse shearing deformation, i.e.
b + s
PN4 = PN4 PN4' (5.38)
Solely antisymmetrical bending

For pure bending -in the Kirchhoff sense where normals remain straight
and normals- the first part of the third mode depicted in Fig. 5.4 is an an-
tisymmetrical bending mode in the x-zplane denoted by /N3' The equation
of the displacement curve takes the form of a cubic polynomial, namely
w(x) = ax 3 + bx2 + ex + d. (5.39)
Using the set of boundary conditions
1 1
w(-2)=0, w(2) = 0,
(5.40)
'( l) 1 '( 1) 1 b
= PN3,
w -2 = 2 PN3 , w 2 = 2 PN3 , PN3
and proceeding with the algebraic manipulations we arrive at the following

equation for the elastic curve :
w(() = g1 (( 3 - () PN3. (5.41)
Superposing the elastic displacement curves due to the second and third
modes we obtain
1 2 1 3
w(() = g{1- ( )PN2 + g (( - () PN3 (5.42)
We shall use equation (5.12) to estimate the strain arising from the sym-
metrical and antisymmetrical bending modes. We have
{Pw z z
'xx = -Z--2
ax = -1 PN2 - 3- (PN3
1
(5.43)
Antisymmetrical bending and transverse shearing

The transverse shear strains are generated in our theory via the antisym-
metric natural straining mode PN3' In fact, equilibrium arguments sub-
stantiate the existence of the transverse shear force in order to equilibrate
the antisymmetrical bending moments (see Fig. 5.6). Thus
Q= 2MA (5.44)
1 .
We assign two antisymmetrical moments of equal magnitude MA on the
two beam edges as shown in Fig. 5.6. From elementary beam theory we
may express the angle on the left edge due to the action of the moment MA
on the same edge as
III MAL
__ ,
Ub = 3EI (5.45)
where E, I, 1 are the Young's modulus, moment of inertia, and beam's

length, respectively. When the same moment MA is applied on the right
edge the angle on the left edge is (see Fig. 5.6)
MAL
O~ = - 6EI' (5.46)
Superposing the expressions, we find
MAL
Ob = ot + O~ = 6EI' (5.47)
For transverse shearing (significant for moderately thick and thick cross
sections), the angle of straining is
(1 Q 2MA
Os =, = G = (GA) = l(GA)' (5.48)
":>+
Mtt~ El .MA)
I ~Q
MA( k
9
======-=-
2
~ )A M
~ ~
~1>
t 9s
~
ilikkbo= "'1~Q*
Figure 5.6: Bending and shearing as induced by the antisymmetrical mode
The total angle due to antisymmetrical mode components P'N3 and PN3 is
MAl 2MA)
PN3 = P'N3 + PN3 = 2{fh + Os) = 2 ( 6EI + l{GA) .
(5.49)
An expression of MA as a function of PN3 can be deduced from (5.49):
3(E!)
M - PN3 = 1 + \2(EI)
A- 2(I 2)
6(EI) + I(GA)
I
I (GA)
PN3, (5.50)
or
MA = 3{EI) _ 3{EI) >. = 12{EI)

,
l( 1 + i22((g~n PN3 - 1{1 + >.)PN3,
-~, - "
(5.51)
where>. is the beam shearing coefficient. Equation (5.51) shows the series
connection between the antisymmetrical bending and the transverse shear-
ing and naturally leads to the circumvention of the transverse shear locking
phenomenon frequently encountered in classical finite element methods. For
a thin beam
(GA) -7 00 :::} A -7 0, MA = 3(El) (5.52)

1
The strain energy associated with the antisymmetrical mode PN3 is
1 1 3(ElN)
UN3 = 2MA PN3 = 2PN3 ", , ,\ PN3, (5.53)
where
(ED)
(ElN) = (El) - (EC)z5, Zo = (EC) (5.54)
In equations (5.52-5.54), (ElN) is the bending stiffness referred to the beam

neutral axis; Zo gives the z coordinate of the middle axis; and (EC), (ED),
(El) are. the axial, coupled axial-bending, and bending stiffnesses of the
beam, respectively. They will be appropriately defined in the sequel.
From (5.53) we deduce the stiffness coefficient that enters the natural
stiffness matrix, viz.
k 3,3 _ 3(ElN)
(5.55)
N - 1(1 + >.)
5.5.4 Mode 4: Symmetrical bending in the x - y plane

The fourth natural straining mode PN 4 is similar to the symmetrical bend-
ing mode PN2 with the exception that it is assigned to the x - y plane. The
equation of the elastic curve due to this mode is
1 2
v(() = 8(1- ( )PN4. (5.56)
5.5.5 Mode 5: Antisymmetrical bending in the x - y plane

The fifth natural straining mode is similar to mode PN3 except that it occurs
in the x - y plane and includes only bending; transverse shear deformation
Z,W
x,u
........ .........
..- .... ..- ........
..-"-
..- .... ..-
..-"-
y. v
..-"-
..-
..-"-
..-"-
.""...""..".,
Figure 5.7: Displacement field due to torsion
is not accounted for. Note that in the absence oftransverse shearing forces,
the elemental equilibrium may not be completely satisfied since the anti-
symmetrical moment M1 is not equilibrated. This moment, however, for
a laminated beam is small and its contribution to the overall deformation
is negligible. We superpose the displacement curves stemming from the
fourth and fifth modes to obtain
l
v{() = 8(1- ( 2 )PN4 + 8l (( 3 - () PN5 (5.57)
The fourth and fifth natural modes give rise to an axial strain which is
derived by using once again (5.12). We have
a2v y y
'Yxx = -y ax 2 = yPN4 - 3 y (PN5 (5.58)
5.5.6 Mode 6: Torsion about the x axis

The last straining mode is a unit twisting angle PN6 about the x axis
depicted in Fig. 5.4. Half the angle is assigned to the left beam edge and
half to the right. Its linear variation with respect to the non-dimensional
coordinate ( is
1
O(() = 2( PN6 (5.59)
The displacement field due to this twisting angle is (see Fig. 5.7)
U(x, y, z) = O(x),x w(y, z), v(x, z) = -O(x) z, w(x, y) = O(x) y,

(5.60)
where w(y, z) is the so-called warping function. On account of equation

(5.59), equation (5.60) gives
1
u((, y, z) = O,x w(y, z) = 1 w(y, z) PN6,
1 1
v((, z) = -0 z = -2( z PN6, w((, y) = Oy = 2( Y PN6 (5.61)
In order to derive the strain field arising from this mode we need to express
the field as a function of the coordinate x. This is accomplished by using
(5.29) and leads to
1
u(x, y, z) = 1 w(y, z) PN6,
x X
(5.62)
v(x, z) = -y Z PN6, w(x, y) = yYPN6.
As with all previous modes we proceed now to the derivation of the

strain field. We use tensorial notation for all strain components. Thus
1 (QUi
lij =2 QXj
QUj )
+ QXi '
i,j = X,y,z. (5.63)
We need to insert a v'2 in front of all shear strains and stresses and derive
a consistent expression fro the strain energy via
1 1 1 (QU' QU')
U = 2(J"ij/ij = 2[v'2(J"ij][v'2, ij] = 2(J"ij QX; + QX~ .
(5.64)
5.6 Natural stiffness matrix 67
The strain field is as follows:
au av aw
'xx = ax = 0, ,yy = ay = 0, ,zz = az = 0,
1
v'22,xy = v'2 (au 1(1
av) = v'2 l aw
ay + ax ay - lz) PN6,
In
y2,xz
1(auaz + aw)
= v'2 1(1l awaz + ly)
ax = v'2 PN6,
(5.65)
In
y2,yz 1 (av
= v'2 az + aw) 1 (x
ay = v'2 x) PN6 = 0.
-y + Y
Grouping the non-zero strains due to the torsional natural mode we find
'xx = 0, y
(aw(y,
In
ay z) - z )PN6,
2,xy
1 1
= v'2l
(5.66)
In 1 1 (aw(y,z) )
y 2,xz = v'2 l az + y PN6
We put this strain field in the elegant matrix form

[Ji;x
v'2
,xz
y] =!
l
11.
v2 ay z) - z )
(aw(y, PN6, (5.67)
~ (aw~~, z) + y)
or using compact matrix notation
T
IT = a PN6 (5.68)
(3xl) (3xl)(lxl)
5.6 Natural stiffness matrix

5.6.1 Strain operator matrix
The complete strain field arising from the natural straining modes is ar-
ranged in the vector
'Yxx] ['Yix + 'Y~; + 'Y1: + 'Yix]

[ V2'Yxy = V2'Yxy , (5.69)
V2'Yxz V2'Yxz
h
were 'Yxx,
E Sb 'Yxx,
'Yxx, T are the straIns
Ab 'Yxx, . d ue t 0 t h e axIa
. IstraInIng,
. symmet-
rical bending, antisymmetrical bending and torsional modes, respectively.
We point out that the in-plane shear strain 'Yxy arises solely from the tor-
sional mode.
Let us collect the expressions for the axial strain 'Yxx. We have
E
'Yxx = 'Yxx + 'Yxx
Sb
+ 'Yxx
Ab
+ 'Yxx
T
= t [PNI + ZPN2 - 3Z(PN3 + YPN4 - 3Y(PN5] ,

(5.70)
or in matrix form
1 Z -3z( Y -3y( PNI

PN2
[ V2'Yxy
7.. ] 1 C:~W{Y, z) _ Z) PN3
=y1 V2 By PN41{5.71)
V2'Yxz
PN5
_1 (OWhY,z) +Y) N6
V2 Z
Using compact notation we write
I = aN PN (5.72)
(3xl) (3x6)(6xl)
The matrix aN is a strain operator matrix. It connects the cartesian strains

with the natural straining modes.
5.6.2 Constitutive relation

The material constitutive relation referred to the material coordinate 123
for every layer k reads
0'= K12i, (5.73)

O's = Ksis
For an orthotropic material
El 1/12E2
0"11
[ 0"22
1 _1_
2E2
[ 1/12E2 E2 . E ] ['Y11 1
'Y22,
V20"12 k 1- 1/12 El . 2G12 (1 - 1/f2 E~} k V2'Y12 k
(5.74)
[V20"13]
V20"23 k =
[2G13
.
.]
2G23 k
[V2'Y13]
V2'Y23 k
For complete isotropy, 1/12 = 1/21 = 1/, El = E2 = E, G 12 = G, and these
equations reduce to
'Y11
[ 'Y22
1,
V2'Y12
(5.75)
[V20"13] [2G.] [V2'Y13]

v'20"23 k = . 2G k V2'Y23 k
The material stiffnesses are transformed from the material coordinate 123
to the local cartesian coordinate system x' y' z' using the congruent trans-
formations
K'= [A~KI2Adk' (5.76)

G' = [A~GI2As]k
where
A= [
C~
So2
S2
o
..,fisoco
-..,fisoco
C2
o
1
-..,fisoco ..,fisoco (C~ - S~) k
(5.77)
As = [ Co so] Co = cos 0, So = sin O.

-So Co k'
The angle 0 represents the angle formed between the fiber axis 1 and and
local axis x' (see Fig. 5.2). Since we account only for strains "{xx, "{xy and
"{xz, the transformed constitutive relation takes the form
[
1
(7xx [Exx Exy
..,fi(7xy = Exy 2G xy . . 1[..,fi"{Xy
"{xx 1 =} U = K, .
..,fi(7xz . 2G xz ..,fi"{xz (5.78)
5.6.3 Strain energy

In order to derive the natural stiffness matrix for the multilayer beam ele-
ment we shall use the expression for the strain energy:
u= 21 ju t ,dV. (5.79)
v
Using equations (5.71) and (5.78), namely
,= OtNPN, U =K" (5.80)
we may write equation (5.79) as
PNt [j OtN
u = -21 (lx6) t K OtNdV ] PN (5.81)
(6x3)(3x3)(3x6) (6xl)
v
''------....,-----
natural stiffness matrix
Substituting in (5.81) the expressions for K, o.N we obtain the following

expression for the natural stiffness matrix
kN = (5.82)
(6x6)
Exx zExx -3z(Exx yExx -3y(Exx ~(\fI,y-z)

z 2E xx -3z 2(E xx zyExx -3zy(Exx (\fI ,y-Z)
zExy
v'2
9z 2( 2E xx -3zyExx 9zy(2Exx -3z(Exy
v'2
(\fI ,y _z )
y 2E xx -3y2( 2Exx v'2 (\fI ,y-Z)
yExy
~JI
v
symm. 9y2( 2Exx -3y(Exy (\fI _ ) I dV.
\12 ,Y Z
G Xy (\fI,y_Z)2
+G xz (\fI,z+y) 2
This full natural stiffness matrix represents the most general stiffness for-
mulation for an anisotropic beam element.
Simplifications
Some of the entries in this full stiffness matrix represent very small to
negligible contributions and therefore introduce unnecessary complications
in practical engineering computations. In particular, the coupling term Exy
is very small (identically zero for isotropic materials). Therefore, we can
omit some of these terms without a significant loss of accuracy. In addition,
we shall assume that some of the natural straining modes do not interact.
This is illustrated via
Exy ~ 0,
PNI !><l PN3, PNl!><l PN5, PN2!><l PN3,
(5.83)
PN2 !><l PN5, PN3 !><l PN4, PN3!><l PN5,
P N 4 !><l P N5, !><l: no interaction.
With these simplifications the natural stiffness is

1
kN = - (5.84)
(6x6) Z2
-Exx zExx yExx
z 2 E xx zyExx
9Z 2( 2Exx
y 2Exx
! symm. 9y2(2 Exx dV.

v
zr
r
GXY(W,y -
+Gxz ( w,z +y
5.6.4 Evaluation of integrals

We now estimate the integrands entering the natural stiffness matrix. Entry
(1,1) is
l b ~
k~l = Z; / Exx{z)dV = Z; / dx / dy / Exx{z)dz
v 0 0 _l!
2
b N (5.85)
= Y2)Zk - zk+dE;x
k=l
where N denotes the total number of layers. The z-coordinates of the

plies are shown in Fig. 5.2. Note that due to the discontinuity of the
material properties across the beam thickness piecewise summation has to
be used. For isotropic materials, equation (5.85) reduces to the well known
expression
11 b(h h ) EA
kN =y "2-(-"2) E=-Z-' (5.86)
Similarly,
1 b ~
k~2 = l; / zExx{z)dV = l; / dx / dy / zExx{z)dz
v 0 0 _l!
2
N (5.87)
1 2 '"
= ~! L..,.{Zk2 - 2
zk+1)E k
xx
k=1
For an isotropic material or symmetrical lamination k~2 = O. For term

(1,4) we obtain
kN14 = l21 / yExx{z)dV =0 (5.88)

v
Proceeding to term (2,2) we find
1/ z Exx{z)dV = l/l dx
kN22 = l2 2 l2 /b dy /~2z Exx{z)dz
v 0 0 _l!
2
N (5.89)
_bl", 3 3 k
- [3 L..,.{Zk - zk+1)Exx
k=1
For isotropic materials
k N2,2 -_ Ebh 3 EI
-
12Z -
-
-z-,
yy (5.90)
where Iyy is the bending moment of inertia with respect to the y axis. With
the evaluation of the integrals k~l, k~2, we are in the position to define the
quantities (EC), (ED) and (EI) that enter (5.54). They are
N
(Ee) = L(Zk - Zk+1)E;x,
k=l
(ED) = 2"1"
N
2 2) k (5.91)
~(Zk - zk+l Exx,
k=l
N
(EI) = 31"
~(Zk3 - 3) k
Zk+l Exx
k=l
The evaluation process continues with term (2,4) which is
kN24 1
= [2 I yzExx(z)dV = 0 (5.92)
v
and term (4,4)
kN44 = 1
[2
I y2 Exx(z)dV = [21 Il Ibdx Y2 dy I~ Exx(z)dz
v 0 0 _l:!
2
1 b3 N
(5.93)
= 12 T L(Zk - zk+1)E;x
k=l
For isotropic materials the expression for term (4,4) reduces to
k 4 ,4 = E b3 (!!:. _ (_!!:.)) = Elzz (5.94)

N 12 [ 2 2 ['
in which Izz represents the bending moment of inertial with respect to the
z axis.
We are only left with terms (5,5), and (6,6). We write
k~5 = ~ I
v
9y2(2 Exx(z)dV
36
=F II x dx Ib y dy I~ Exx(z)dz
2 2
(5.95)
o 0 _l!
2
1 b3 N
= 4T ~)Zk - zk+dE!x
k=l
For an isotropic material this relation simplifies to
k5,5 = ! b3 (!!. _(-!!.))E = 3Elzz (5.96)

N 4l 2 2 l
The last term (6,6) induced by the torsional mode PN6 yields the integrand
k~6 = z21 I [G XY (OW

oy -
) Z
2
+ Gxz (OW)
OZ + y
2
]dV
v (5.97)
In order to evaluate this integrand, we need the expression for the warping
function. If warping is not considered, W-= 0 and we have
k~6 = l; I [z
v
2 GXY + y 2 Gxz ] dV. (5.98)
Piecewise integration is again used to evaluate the integrals containing Gxy ,

G xz . We have
N
6,6a _ 1 b ""' 3 3 k
kN - 31 ~(Zk - Zk+l)G xy ,
k=l
N
(5.99)
b3
6,6b _ 1 ""'
- 12 T ~(Zk - Zk+l)G xz
k
kN
k=l
For isotropic materials excluding warping (5.98) reduces to
66 G(bh 3 hb3 ) G GJp

kN = T 12 + 12 = T(Iyy + I zz ) = -Z-,
(5.100)
where Jp is the polar moment of inertia.

Following Timoshenko [30], for isotropic materials with rectangular
cross section of width b and height a and warping effects considered, the
torsional constant J can be expressed as
J=kbh 3 , (5.101)
where
k = 0.1406, for b/h = 1,

k = 0.2910, for b/h = 5, (5.102)
k = 0.3333, for b/h = 00.
The first equilibrium equation of elasticity is [30] namely
O'xx,x + O'xy,y + O'xz,z = o. (5.103)
When only torsion is applied we have
1
O'xx = Exx'Yxx = E xxy 'I!PN6 => O'xx,x = 0, (5.104)
f) 2 'I!
O'xy = Gxy'Yxy = G XYy1 (f)'I!
!l - Z
)
PN6 => O'xy,y =
1
yGXY {j2PN6,
uy Y (5.105)
2
O'xz = Gxz'Yxz = G xz -1Z (f)'I! ) 1 f) 'I!
! l + Y PN6 => O'xz,z = -Z Gxzf.l2"PN6.
uZ uZ (5.106)
Substituting equations (5.104-5.106) in equation (5.103) we get

1
( yG 82lJ! 1 82lJ!)
XY 8y2 + yGxz 8z2 PN6 = 0, (5.107)
( Gxy 82lJ! 8 2lJ!)

8y2 + Gxz 8z2 = O. (5.108)
The warping function must satisfy (5.108). In the absence of warping, the
natural stiffness matrix becomes
Exx zExx
z2E xx
9Z 2( 2E xx
kN =~J y 2Exx dV.
(6x6) I v
symm. 9y2(2 Exx (5.109)
z2Gxy+y2Gxz
This form of the stiffness matrix does not include the effect of transverse
shear deformation. In order to account for the latter, we replace term (3,3)
with the previously derived expression (5.55) and write
Exx zExx
z 2 E xx
kN~~ Ii dV,
(6x6) I V y2Exx
(5.110)
symm. 9y2( 2Exx
z2G xy +y2G xz
where
1 z
G xz dz
k=l
N
= 2)Zk - Zk+1)G~z' (5.111)
It is worth mentioning that for isotropic materials this stiffness matrix

reduces to
EA
-[-
Elyy
[
[2
kN = Elzz
(6x6) [ (5.112)
3Elzz
symm. [
GJ
-[-
5.6.5 Shear correction factor

The transverse shear stresses at the top and bottom surfaces of the beam
must be zero. Since our theory does not satisfy this requirement, the trans-
verse shear stiffness needs to be multiplied with a shear correction factor.
Essentially the correction factor adjusts the transverse shear energy with
regard to the exact energy. If we denote by G(z) the actual variable shear
modulus which is a function of the thickness coordinate, and by G the shear
modulus under the assumption of constant shear strain, then
G = Be/G, (5.113)
where Be/ denotes the shear correction factor. To the shear moduli corre-
spond transverse shear energies, viz.
Us = Be/Us. (5.114)
From elementary beam theory it is known that the transverse shear stress
f7 q at position z is given by [40]
f7 q =Q
Rg(z), (5.115)
where
!
h/2
R = Exx(z - zo)2dz, flexural beam stiffness,

-h/2 (5.116)
and
z
g(z) =- / Exx(z - zo)2dz, shear stress shape function.
-h/2 (5.117)
The transverse shear strain energy is given by
h h
2 2
1/ (72 1 Q2 / g2 (z)
Us = "2 G(:) dz = "2 R2 G(z) dz. (5.118)
h h
-2 -2
The energy expression associated with constant shear strain is
/2 iG(z)idz.
h
_ 1
Us = "2 (5.119)
h
-2
Using
-
, =
{7q
G - hG'
-!L (5.120)
we may write equation (5.119) as
h
_ 1 Q2 2
Us = "2 h2G2 / G(z)dz. (5.121)
h
2
By setting
h
2
hG = / G(z)dz, (5.122)
h
-2
equation (5.121) reduces to

_ 1 Q2
US = ---"
2 l!
(5.123)
!
2
G(z)dz
_l!
2
The expression for the shear correction factor reads
- 2
S _ Us _ R
(5.124)
! !
e! - U - h/2 h/2
g2(z)
G(z)dz G(z) dz
-h/2 -h/2
In order to proceed with the elaborate computation of the shear correction
factor Se!, the z coordinate of the neutral plane must be defined. This is
given by
!
h/2
zE!x dz
-h/2 (ED)
ZOa = h/2 - (EC)'
(5.125)
! -h/2
E!x dz
In the computer program, the shear correction factor is automatically com-

puted. It reads
[!h/2 E;x(z - zOa)2dz] 2
h/2
=
[- !
Se! h/2 ]2
(5.126)
E!x(z - zOa)2dz
h/2
Ga
!
-h/2
h/2
Gka
dz
where
! G~dz = L G~zhk'
h/2 N
Ga = (5.127)
-h/2 k=1
The corrected transverse shear modulus is
G xz = ScI G xz (5.128)
5.7 Local and global stiffness matrices

To transform the natural stiffness matrix kN to the local and global carte-
sian coordinates, we need to establish the connection between the natural
straining modes and the local nodal degrees of freedom. The relevant ma-
trix equation is
PN = aN P (5.129)
(6xl) (6xI2)(12xl)
The connection matrix aN contains only zeros, unit numbers and geomet-
rical parameters. The contribution of the vertical nodal displacements to
the antisymmetrical mode is illustrated in Fig. 5.8. We now express the
natural modes as functions of the local cartesian coordinates, viz.
PNI = U2 - UI
PN2 = <P2 - <PI
PN3 = -(<PI + <P2) + 2(WI :- W2)
(5.130)
PN4 = tPl - tP2
PN5 = (tPl + tP2) + 2(VI :- V2)
PN6 = (h - (h
z
3
w I
o~I ~~
~_
~
10
1= "I
112
Figure 5.8: Antisymmetrical rotation due to vertical nodal displacements
from which we deduce the matrix
-1
-1 1
2 2
-1 -7 -1
aN 7
-1
(6x12) (5.131)
2 2
7 -7 1
-1
We are now in the position to provide the complete matrix connection

between the natural rigid body and straining modes and the local cartesian
degrees of freedom. The two matrices that provide these connections are
assembled as follows:
[(:~)l = (5.132)
(6x12)
1 1
2 2
1 1
2 2
1 1
2 2
1 1
2 2
1 1
1 -1
1 1
-1 1
-1
-1 1
2 2 "":1
1 -1 -1
1 -1
2 2
1 1 -1 1
-1 1
Using this expression and equations (5.25), (5.129), we write
[
1 [aD 1 [ aD 1
(6x1)
Po
(6x12)
~; = ~~ (~~) = ~; (1~~i2)(lfx1)'
(6x12)
(5.133)
(6x1) (6x12) (6x12)
Conversely,
P
(1"'1)
= T- 1
(12:2)
AD
[(12 X 6)
I A
(1':;;)
1PO
1 ;~ .
(6x1)
[ (5.134)
(6x1)
Obviously,
[at 1 = A. (5.135)
The components of the matrix A explicitly read

[ Ao : AN ] (5.136)
(12x6) (12x6)
r
1
1 -2
I
1 -2
I
1 -2
1
-2
1 1
1 -2 ~2
1 1
1 2 2
1
1 2
1
1 2
I
1 -2
1
1 2
1 1
2 -2
1 1
1 -2 2
Now the expression for the strain energy takes the form
u= ~/ ut,dV
v
= -I PN t [/t
aN K, aN dV ] PN
2 (lx6) (6x3) (3x3)(3x6) (6xl)
v
....
(5.137)
= -1 pt [ii}y[ / a}y K, aN dV]aN
] P
2 (lx12) (12x6) (6x3)(3x3)(3x6) (6x12) (12xl)
v
....
~------~v~--------~
local cartesian stiffness matrix
The local elemental vector p is related to the global elemental vector P via
p = T04P, (5.138)
where T04 is a matrix containing submatrices of direction cosines, namely
=
[TO. To
T04
To
: ] = rTo To To Toj,
(12x12) : (5.139)
To
with
CX' x Cx' y Cx' z]

To = [ Cy'x Cy'y Cy'z . (5.140)
(3x3) Cz'x Cz'y Cz'z
In equation (5.140), Cx'x, denotes the cosine of the angle formed between
the local cartesian axis x' and global cartesian axis x. The same convention
is used for all other entries (see Appendix A). Using
PN = aN" = aNT04P = aNP [, (5.141)
we rewrite the strain energy for the beam element as
u=-1 PtIt [-
T04 aN t
2 (lx12 (12x12) (6x12) v
[I ON t K, ON dV ]
(6x3)(3x3) (3x6)
- ] T04
aN 1 p.
(6x12) (12x12) 12xl)(5.142)
v ~
natural stiffness matrix kN

'"
local cartesian stiffness matrix Ii:
v
global cartesian stiffness matrix k
The matrix equilibrium equations in the natural, local and global coordi-
nate systems are
PN = kNPN natural coordinate

P = kp local coordinate (5.143)
P=kp global coordinate
Next, we consider transformation rules for the stress resultants, i.e. the
generalized natural and cartesian forces and moments in the various coor-
dinate systems. Using the equivalence of work
- -
P- t_P = ptNPN = ptNaNP, (5.144)
we deduce
- -t - t -
P = aN PN, PN = AN P ,
(12xl) (12x6)(6xl) (6xl) (6x12)(12xl)
PN
(6xl)
= A~ T06 P,
(6x12)(12x12)(12xl)
P
(12xl)
= Tal ak PN.
(12x12)(12x6)(6xl)
(5.145)
5.8 Work of external loads

In the presence of external loads in the form of distributed or concentrated
loads and/or moments, the work produced by the action of these loads is
W=~! p~udS+Ftu+Mt8, (5.146)

s
where Ps, F, M, represent the vectors of distributed surface loads, con-
centrated forces, and concentrated moments, respectively. Concentrated
loads and moments can be directly imposed on the nodes, provided that
their components refer always to the global coordinate system. When dis-
tributed surface loads are present (they are usually defined in the local
coordinate frame) kinematically equivalent nodal loads must be computed.
The local cartesian degrees of freedom are interpolated from the nodal val-
ues via
u w P (5.147)
(6xl) (6x12)(12xl)
Similarly, for the vector of distributed loads

5.8 Work of external loads 87
Ps = W PSI, (5.148)
(12xl) (6x12)(12xl)
where PSI includes the nodal values of the distributed load. Substituting
equations (5.147), (5.148) in equation (5.146) we obtain
Wp =-II
2
S
PsudS
t =- 1[ tit
PSI
2 (lx12)
S
W W
(12x6)(6x12)
dS T04 ] P
(12x12) (12xl)
1 t (5.149)
= 2PsP ,
from which we deduce the global vector of distributed loads
Ps =
(12xl)
Tb4
(12x12)
[I t w w dS] PSI
(12x6)(6x12) (12xl)
(5.150)
S
Note that in order for (5.150) to be valid, the distributed load vector PSI
must always be given with respect to the local elemental cartesian coordi-
nate system x'y'Z'. We embark now on the computation of the term wtw.
For this purpose we use the non-dimensional coordinate 'f/
x
'f/: 0 -+ 1, 'f/ = l' dx = l d'f/. (5.151)
If we denote by Ul, U2 the vectors of the nodal freedoms, we interpolate

the displacements and rotations of a generic point within the beam element
from the nodal values via
u = (1 - 'f/)Ul + 'f/U2 (5.152)
In so doing, we create the interpolation matrix

W
(6x12) (5.153)
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
Then, a straightforward line integration yields
n
(12x12)
=j wtwdx
x
1 1
'3 6
1 1
'3 6
1 1
'3 6
1 1
'3 6
1 1
'3 6
= 11
1
'3
1
11 (5.154)
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
The complete global load vector becomes
P = (bl) Tb4 n PSI (5.155)

(12xl) (12x12)(12x12)(12xl)
This relation provides us with the global nodal loading vector when local
distributed loads are prescribed. For three commonly used loading cases,
the equivalent nodal loads in the local coordinate are
uniform load: load at left and right nodes equal to ~ql

triangular load: load at left node kql; load at right node ~ql
trapezoidal load: load at left node {~ql + kq2)l; load at right node
(kql + ~q2)l
These are line loadings. If the loads are also applied across the width of
the beam b, the above values must be multiplied by b.
5.9 Initial load due to temperature

We consider the natural and cartesian initial load vectors due to tempera-
ture. We assume a linear temperature distribution through the thickness,
namely
T{x, z) = To (x) + zTdx). (5.156)
this temperature field produces a constant temperature gradient. The pa-

rameters To, Tl in (5.156) are related to the temperatures at the top and
bottom surfaces of a laminated beam Tt , n, respectively, via
1 1
To = "2{Tt + n), Tl = -,;,{Tt - n), (5.157)
where h represents the beam thickness. In addition we assume a linear

temperature variation along the beam axis x. Using the non~dimensional
coordinate ( we have
T t {() = 1 -2 ( 1:1t + 1 +2 ( 1:2t,

(5.158)
n{() = 1- ( T,1 + 1 + ( T,2
2 b 2 b'
in which Tl, Tl, Tl, T~, represent the nodal temperatures at the top and
bottom surfaces at nodes 1,2, respectively. As for the elastic stiffness, the
thermal load vector will be computed first in the natural coordinate and
then transformed to the local and global coordinates before assembled.
The natural thermal load vector J N is written as the sum of two vectors,
namely
JN = J NO + J Nl , (5.159)
(12xl) (12xl) (12xl)
where J NO, J Nl are the thermal loads due to uniform and linear through
the thickness temperature distributions, respectively. To this end we define
the thermoelastic coefficients for every layer in both material and local
coordinates via
at
12
[ O:'tll
.
.] ,
O:'t22
at = axx
[ a yy
1 , (5.160)
V2a xy
where, O:'tu, at22 are the coefficients of thermal expansion along and perpen-
dicular to the fiber direction, respectively. The thermoelastic coefficients in
the local coordinate system are obtained from the thermoelastic coefficients
of the material coordinate system using
[ axx V2a xy ] = [A~a}2 AS]k' (5.161)

V2a xy a yy k
The total natural strain is the sum of the elastic and thermal strains, re-
spectively, namely
I = g + TJ, g: elastic strain TJ: thermal strain.

(5.162)
With respect to the local coordinate system the thermoelastic constitutive

relation reads
U=K[,-TJ] =Kg. (5.163)
Substituting equations (5.162) and (5.163) in the strain energy expression

(5.79) we obtain
u=~/ ut,dV = ~ / [It -l1t]~,dV

v v
I ak~aNdV 1
= pk [
= Ue + ut ,
V
natural ...'
, elastic stiffness
PN - [
------
l11t~aNdV 1PN
v
natural thermal vector
(5.164)
from which we deduce the natural thermal load, viz.
Jk = / l1t~aNdV. (5.165)
v
Using the basic definition of thermal strain
l1t = atb..T = at(To + zTI), (5.166)
we find the initial load due to temperature
Jk = Jko +Jkl
(lx6)
I
= a~~aNTo(z,()dV +
V
I za~~aNTl(z,()dV
V
(5.167)
5.9.1 Evaluation of integrals

We substitute all the matrix expressions we have found in (5.167). The
first entry in the natural thermal load vector being induced by the first
straining mode is of the form
11
JtJ11 = 1 T(z,()ExxcxxxdV. (5.168)
v
Proceeding with the tedious algebraic manipulations we are able to dis-

tinguish two integrals representing contributions from uniform and linear
through the thickness temperature distributions, respectively. They are
N
J NO
l,l _
- 4b (T2t +Tb
2
+
TIl) "'(
t +Tb L...J Zk - Zk+l
)Ek k
xxaxx
k=l
N
b (5.169)
J Nl
l,1
= 4h
(T2 fTl2
t -.Lb
Tl
+
fTll) " ' ( 2 2 )Ek k
t -.Lb L...J Zk - Zk+l xxaxx
k=1
The addition of these two entries yields the first component of the natural
thermal vector. For an isotropic beam of area A and Young's modulus E
under equal temperatures on the top and bottom surfaces, respectively, the
addition of the two integrals yields
J NO
1,1 + J Nl
1 ,1
-- (AE) aT, (5.170)
which is a well known expression of basic beam theory. The second entry
in the initial natural thermal load vector produced by the second straining
natural mode is of the form
JJJ21 = l 1! zT(z, ()Exxaxx dV. (5.171)

v
More elaborately
N
J NO
2,1 _
- 8"b (T2t fTl2
+.Lb + Tl fTll) " ' ( 2 2 )Ek k
t +.Lb L...J Zk - Zk+l xxaxx,
k=l
N
J Nl b (T2t 5.172)
2,1
= 6h fTl2
-.Lb + Tl fTll) " ' ( 3 3 )Ek k
t -.Lb L...J Zk - Zk+l xxaxx,
k=l
For an isotropic material or symmetrical lamination (with equal applied

temperatures on the top and bottom surfaces) the addition of the two
contributions yields zero.
We are next concerned with the term (3,1) which is the effect of the
antisymmetrical bending natural mode, namely
31 =
J'; -1 1/ 3( zT(z,()ExxaxxdV. (5.173)
v
After some algebraic manipulations this term gives
N
J NO
3,1 _
- -8"b( - T2t '1"12
-.Lb + Tlt '1"11) " ' ( 2
+.Lb L...J zk
2) k k
- zk+l Exxa xx ,
k=1
N
J 3,1 = _~ (_7'.2 + T,2 + Tl _ T,1) ' " (z3 _ z3 )Ek ak (~.174)
Nl 6h t b t b L...J k k+ 1 xx xx
k=1
For isotropic materials under the same applied temperature, the sum of
these two entries is zero. Entries (4,1) and (5,1) represent the effect of the
symmetrical and antisymmetrical natural straining in the x - y plane and
are
41
J'; =11/ yT(z,()ExxaxxdV=O, (5.175)
v
51
J'; = -1 1/ 3(yT(z,()Exxa xx dV = O. (5.176)
v
We are only left with entry (6,1) -the direct result of the action of the
torsional natural straining mode. The corresponding integral is
1/
J,;61 = 1 [
T(z,() 2G xy (B'I!(Y,z)
By - z ) a xy ] dV.
(5.177)
v
Since \lI,y = 0, equation (5.177) simplifies to
IN
61 = -Y 1! 2zT(z, ()GxylxydV. (5.178)
v
Using piecewise summation we derive the expressions
N
JNO
6,1
= -4b (fTl2
.Lt
fTl2 fTl1
+.Lb +.Lt +.Lb
fTl1) ~(2
~ Zk -
2 )Gk
Zk+1
k
xylxy
k=1
N
J N1
6,1 _ _ ~(fTl2_fTl2 fTl1_fTl1)~(3_ 3 )Gk k (~.179)
- 3h.Lt .Lb +.Lt .Lb ~ Zk Zk+1 xylxy,
k=1
For isotropic materials under the same applied temperature the addition of
these two terms gives zero.
Note that the natural thermal load vector has been completely defined.
Equilibrium in the natural coordinate yields the equation
kNPN = PN+JN I (5.180)
The natural initial load due to temperature is transformed to the local and
global coordinate systems via
- ]t t-
J = [aN T04 JN = T04 J
(12x1) (12x12)(12x12) (12x1) (12x12)(12x1)
-- t
aN J N
(5.181)
(12x12)(12x1)
Global equilibrium now reads

5.10 Postprocessing 95
I KEr~P+J I'
(5.182)
where
K = Ak, P = A.P, J = AJ, A: assembly operator.

(5.183)
The solution of (5.182) yields the global nodal displacements and rotations.
5.10 Postprocessing
5.10.1 Computation of forces and moments
To gain a better insight into the behavior of the beam element, we look
at the forces, moments, stresses and strain energies. In our method they
are all obtained in a postprocessing mode following the computation of the
global displacements.
We use equation (5.141), namely
PN = iiN T 04P, (5.184)
to extract the natural modes PN, from the global cartesian vector P the
natural modes PN. This is performed for every element. Then we multiply
the vector of natural modes by the natural stiffness to obtain the generalized
natural forces and moments
PN =kNPN (5.185)
We use the following sign convention for natural forces and moments:
positive moment (both ends): clockwise
positive axial force (both ends): to the right
positive shear force (both ends): upwards

With this convention, the forces and moments become:
PLANE x-z
1. Left axial force: -FN.

Right axial force: FN.
2. Left edge moment: - M}; - Ml.

Right edge moment: M}; - Ml.
3. Left edge shear force: T

2M!
Right edge shear force: - T

2M!
PLANE x-y
1. Left edge moment: - M' - Ml

Right edge moment: M - Ml
2. Left edge shear force: T
2M2
Right edge shear force: - T

2M2
3. Left-right edge torsional moment: MT.
5.10.2 Natural energies

A unique feature of our theory is the separation of the strain energy into
separate components. This feature provides us with a clear picture re-
garding where the various strain energy components are allocated during
deformation. These invariant strain energy measures can be used to predict
failure, especially in composite structures. For example, the strain energy
induced by the antisymmetrical natural mode may be responsible for the
onset of delamination.
The definition of the natural strain energies is straightforward follow-
ing the computation of the natural straining modes and forces. They are
defined as follows:
5.10 Postprocessing 97
1
Axial straining energy in x - x : -FNPNl,
2
1 1
Symmetrical bending energy in x - z : -MSPNl,
2
1 2
Symmetrical bending energy in x - y : -MSPN4,
2
1 1
Antisymmetrical natural energy in x - z : -MA PN3, (5.186)
2
1 3
Antisymmetrical natural energy in x - y : -MA PN5,
2
1
Torsional natural energy in y - z : -MTPN6.
2
5.10.3 Through the thickness stresses

The through the thickness stresses will be computed for every beam element
in the local cartesian coordinate system x' y' z' . The u xx direct stress is
computed for the kth layer using
k
u xx = Ekxx'Yxx
k (5.187)
where
k PNI Zk Zk
'Yxx = -l- + TPN2 - 3 T (PN3. (5.188)
By substituting ( = -1, ( = 1 in this equation, we can obtain the direct

strains at the left and right nodes of a beam element, respectively. In
our theory we do not account for U yy . It is possible, however, to obtain
approximate values for this stress from the ply constitutive relation
k
U yy = V12 Ekxx'Yxx'
k (5.189)
The in-plane shear stress u xy arises solely from the torsional straining mode
PN6 Therefore
Z
'Yxy = U,y +v'x = -yPN6' (5.190)
and
k
u xy = Gkxy 'Yxy
k (5.191)
In our theory the transverse shear stresses are constant through the beam
thickness. For an accurate computation of the through the thickness trans-
verse shear stresses u xz we shall proceed as follows: we first state equilib-
rium in the x direction via
auxx auxz _ 0
ax + az - , (5.192)
or
auxz auxx
az - ax ' u xz -
__ ! auaxz
xx dz + Cl. (5.193)
h
-2"
Elementary beam theory states that
!
h
2"
zExx(z)
auxx aMz aM ~ (z - zo),
ax ax ax
!
I ~
(5.194)
Exx(z - zo)2dz
h
-2"
where Zo is the z coordinate of the neutral axis. A combination of these

two relations yields
!
h
2"
zExx(z)
! ax
z h
aM -2" (z - zo)dz + Cl.

U xz ~
!
= -
(5.195)
-~ Exx(z - zo)2dz
h
-2"
5.11 Geometrical stiffness 99
The shear force per unit length is defined as
aM 2MA
q=-=--, (5.196)
ax l
On account of (5.196), equation (5.195) yields
!
h
2"
zExx{z)
! 2~A
z
h
CT xz =_ h -2" {z - zo)dz + CI
!
2" (5.197)
_l!:.
2
Exx{z - zo)2dz
h
-2"
Since the transverse shear stresses vanish at the top and bottom beam
surfaces, the constant CI is zero.
5.11 Geometrical stiffness

5.11.1 Elastic buckling
Elastic buckling is an important phenomenon for isotropic and composite
structures, particularly thin wall structures, due to the presence of signif-
icant membrane stressing. These structures include rods, Euler columns,
plates and shells. The problem itself is a difficult one, for instability can be
affected by various factors such as initial geometrical and material imper-
fections, non uniformity in load distribution, and the pattern of loading.
Most studies are concerned with isotropic and orthotropic structures, the
latter including beams, finite strips, plates and some special forms of shells
-arbitrary geometries are much more difficult to analyze. In view of these
factors, it is imperative that we develop efficient techniques and algorithms
for the study of elastic buckling. This is accomplished by formulating the
so-called geometrical stiffness for beam, plate and shell elements. The ge-
ometrical stiffness can also be used for large displacement, small strain,
nonlinear analysis. Elastic stability theory requires that the equilibrium
equations be written for the deformed geometry. This is in contrast to

most elasticity problems where the equilibrium equations are written with
respect to the original unloaded configuration.
Suppose that an axial compressive force of magnitude P is applied at
an end of a beam so that no bending occurs. When this force is small,
there is a unique and stable form of equilibrium with the beam remaining
straight. If a small lateral load is applied to the beam it will deflect, and if
this lateral load is removed the beam will revert to its original undeformed
configuration. However, when the load P reaches a critical value Per, the
structure will remain in the deflected position upon removal of the lateral
load. At this stage the structure passes from a stable form of equilibrium
to an unstable. The latter is characterized by a curvature of the middle
surface.
In a huge displacement but small strain analysis the geometrical stiff-
ness is added to the elastic stiffness to form the tangent stiffness. While
large deflections require incremental/iterative procedures, elastic buckling
is usually reduced to an eigenvalue problem. For engineering applications,
a large displacement theory restricted to small strains must be simple and
economical. The geometrical stiffness of plate and shell elements usually
presents no great difficulty, for the only geometrical stiffness of importance
is that arising from membrane stresses. The derivation of geometrical stiff-
ness is easier for natural modes depending only on nodal displacements
(translations). In this case, the resulting elemental matrix is always sym-
metric. When nodal rotations or higher order parameters are considered,
new definitions of moments and rotations must be introduced, for as soon
as second-order effects are considered the usual definitions of moments and
rotations lead to nonconservative loads and nonunique functions of the ro-
tations [41].
5.11.2 Simplified geometrical stiffness

The basic assumption of our theory is that within each element the small
displacement stress-strain relations are valid. For this reason the natural
mode method is applicable. In the context of small strain large displace-
ment theory the natural modes are small even when the global displace-
ments are large. That is to say, during deformation the element does not
change shape with respect to the natural coordinate system. The natural
modes are assigned to a convective natural coordinate which follows the
structure during deformation within the framework of an Eulerian motion.
We have seen that the generalized natural forces P N produce cartesian
nodal forces via (see (5.145))
-
P = -t
aN
P N. (5.198)
(12xl) (12x6)(6xl)
Essentially matrix aN defines the nodal cartesian loads corresponding to

unit generalized straining forces. For the calculation of the geometrical
stiffness the rigid body resultants do not enter. Thus, we have to consider
only the change in a~PN when PN is not varied. The part due to the
variation of P N corresponds to the tangent elastic stiffness of the element.
A small variation in P yields
dP = da~P N + a~ dP N. (5.199)
The term da~ represents the effect of change of geometry when the natural
forces are not varied and this is the main contributor to the geometrical
stiffness, while a~ dPN stems from a variation of PN, the latter being
usually an initial end load. In our theory we assume that only the rigid body
rotations contribute appreciably to the geometrical stiffness. Therefore the
term da~ in equation (5.199) will be derived based only on this assumption.
We consider the vector of rigid body rotations
POR = {P04 P05 P06}. (5.200)
A small rigid body rotational increment gives
dPOR = {dp04 dP05 dp06}. (5.201)
These rotational variations will induce geometrical forces. First they give
rise to the incremental rigid body moments
dPOR = {dP04 dP05 dP06 }. (5.202)
Now a small change of the cartesian nodal loads produces rigid body mo-
ments
- t - - t -t
dPOR = AORdP = AORdaNPN, (5.203)
where AOR is the submatrix of Ao corresponding to rigid body rotations.

Since
- t -t
AORaN = 0, (5.204)
we have
- t -t - t -t
dAoRaN + AORdaN = o. (5.205)
Using (5.205), we may write equation (5.203) as
dPOR = -dAoRaNPN.
- t -t
(5.206)
The incremental rigid body rotational moments give rise to the geometrical
cartesian forces
- G = aOR
dP -t dP OR -t dA- t -t P
= -aOR ORaN N, (5.207)
(12x3) (3xl)
whereby
-t -t -t
- t OAOR OAOR OAOR
dAoR = -!:}-dP04 + -!:}-dP05 + -!:}-dP06' (5.208)
UP04 UP05 UP06
Note that in the three dimensional case we consider the rotations POR to be
applicable only over a small range, so that they can be added vectorially.
Substituting equation (5.208) in equation (5.207) we get
-
oAt
OR -t P ] - d-
dP G = - aOR OP04 aN N a04 P
-t [
-t [
oAtOR -t P ] - d-
- aOR OP05 aN N a05 P (5.209)
-t [
oAtOR -t P ] - d-
- aOR OP06 aN N a06 P
= [a&RkGRaOR] dp = kGdp
We have now derived the rotational geometrical stiffness, namely
kCR = -[kCR4 kCR5 kCR6], (5.210)

(3x3) (3xl) (3xl) (3xl)
in which
-t
kCR4 = OAOR]
[- - -t
aN PN,
OP04 (12x6)(6xl)
(3x12)
-t
kCR5
OAOR]
= [- - -t
aN PN, (5.211)
OP05 (12x6)(6xl)
(3x12)
-t
kCR6 = OAOR]
[- - -t
aN PN
OP06 (12x6)(6xl)
(3x12)
The rotational geometric stiffness kCR represents the increments of rigid

body moments produced on the element by the initially self equilibrating
cartesian forces P = a~PN when the element rotates as a rigid body.
All terms in (5.21O) are available except the matrix containing the partial
derivatives. To compute it, we first note that if the element undergoes
small rotations P04, P05, P04, the body axes, say X, y, z, which are originally
parallel to the local axes x' y' z' acquire direction cosines with respect to the
local axes given by the rows of the rotation matrix
R =
c xx ,
[ cyx'
Cxy'
Cyy'
c xz']
cyz' =
[1 -P06
P06
1
-P05]
P04 .
Czx' czy' Czz' P05 - P04 1 (5.212)
Let us derive one of these matrix entries. We choose
OX (7r ) .
Cxz = OZ' = cos 2" - P05 = - smp05 = -P05, (5.213)
for P05 small. All other entries are defined in a similar way. We state now
the incremental rigid body rotation matrix
. dP06
-dP05 ]
dR = [ -dp06 dP04 , (5.214)
dP05 -dP04
as well as the submatrices
R4 = [: _ ~] P04, Rs= ~ [. . -1]: Pos ~ = ~1

[
. 1 .]
:
P06
(5.215)
. 1.
Now the global force increments may be written as
dPG = dR4P, (5.216)
with
dE4= [
dR
dR :1= rdR dR dR dR j.
(5.217)
dR
Similarly, the variation of a, A is
- - t - - - t -
da = adR4, dA=dR4 A, dAoR = dR4 AOR
(5.218)
Differentiating the last expression in (5.218) with respect to the rigid body
modes P04, Pos, P06, we obtain
- t t-
oA OR = A OR R 4,
OP04
- t t-
oA OR = AORRs , (5.219)
opos
- t t-
oA OR = AOR~'
OP06
where
R4=r R 4 R4 ~ R4j,
Rs = r Rs Rs Rs Rsj, (5.220)
R6=r~ ~ ~ ~j.
Using equation {5.219}, we can put the rotational geometrical stiffness in

the form:
kGR
(3x3)
= {5.221}
- [A~R ~ a}y PN
(3"'12) (12'" 12) (12"'12) (12'" 1)
A~R Rs a}y PN
(3"'12) (12"'12) (12:1: 12) (12"'1)
A~R ~ a}y PN]
(3:1:12)(12:1:12)(12:1: 12) (12:1:1)
When all natural forces are included in {5.221} the rotational geomet-
rical stiffness kOR is unsymmetric and the resulting cartesian geometrical
stiffness k also displays asymmetry. The asymmetry can be explained by
the fact that higher order rotation terms are involved which are not commu-
tative. Using more mathematical phraseology, we may say that successive
finite rotations referred to fixed axes introduce path dependent work and
as such they are nonconservative. A first approximation would be to sym-
metrize the geometrical stiffness via
- s 1 - -t
kG = 2[kG + kG] {5.222}
When only the axial natural force FN enters the rotational geometrical
stiffness the latter is always symmetric. It is given by
kGR =
(3x3)
- [A~R R4 a}y FN
(3:1:12) (12:1:12) (12:1:1) (l:t 1)
A~R Rs a}y FN A~R ~ a}y FN] {5.223}
(3:1:12) (12:1: 12) (12:tl) (1", 1) (3"'12) (12'" 12) (12'" 1) (l:t 1)
This rotational geometrical stiffness is used to form the simplified geomet-

rical stiffness which owes its name to the fact that only the axial natural
force operating on rigid body rotations induces geometrical forces. The
symmetric simplified geometrical stiffness suffices for most practical engi-
neering computations. The cartesian nodal forces due to three rigid body
rotations PORt::.. are
PGt::.. = atR PORt::.. = [ atR kGR aOR] Pt::.. = kG Pt::..

(12xl) (12x3) (3xl) (12x3)(3x3)(3xI2) (12xl) (12xI2)(12xl) {5.224}
Following the computation of the local cartesian geometrical stiffness
ka, our next task involves its transformation to the global coordinate sys-
tem. This is accomplished by using the matrix of direction cosines T04 and
the congruent transformations
t -
kG = T04kG T o4. (5.225)
5.11.3 The natural geometrical stiffness

In addition to the change of geometry, additional energy is produced in
structural instability phenomena by the stiffening or softening effect, that is
from the coupling between the axial force due to an initial end load and the
symmetrical and antisymmetrical bending rotations. In this way additional
geometrical forces are induced which affect the overall equilibrium. These
additional forces give rise to the so called natural geometrical stiffness kNG.
We consider again a beam element of length I, which under the action
of a symmetrical bending rotation PN2, becomes a circular arc of arclength
s and radius r. With respect to the original undeformed state and for small
PN2 we have
rsin PN2 I . PN2 PN2

2 - 2' Slll-- ~--
22'
rpN2 = I. (5.226)
If, however, the rotation is not sufficiently small, nonlinear terms influence
(5.226). We approximate the sin by
3
. PN2 PN2 PN2
s l l l T = T - 48' (5.227)
so that
3 I 2 I 2
PN2 PN2) I PN2 PN2
r ( T - 48 = 2" ~ rpN2 - = rpN2 24 = 24'
(5.228)
from which we deduce
f:11 = lp'~2 (5.229)

24 .
This equation expresses the change of the beam length f:11 produced by
the natural mode PN2. Now the axial natural force FN operating on this
displacement produces work
12
W=FN PN2 {5.230}
24 .
From this energy expression we immediately observe a coupling between

the natural axial force FN and the symmetrical bending mode PN2. Note
that force FN produces work when operating on the symmetrical bending
mode PN4. Therefore
1 FNl 2 FNl 2
W NG = 24 PN2 + 24 PN4' {5.231}
Proceeding with similar arguments for the antisymmetrical modes PN3,

PN5, the work done by the axial force FN on the change of length due to
these rotations is
2 FNl 2 FNl 2
W NG = 40 PN3 + 40 PN5' {5.232}
Note that the energy W~G is usually very small and can be neglected. From
{5.231-5.232} we deduce the natural geometrical stiffness via
k aWhG aW~G {5.233}

NG - a
(12x12)
a t
PN PN
+ a at'
PN PN
or explicitly
1
12
1
kNG = FNl 20
1
(12x12) 12 {5.234}
1
20
1
= FNlr 12 20 12 20
~ ~ ~ .J
Once the natural geometrical stiffness is established, a transformation to

the global coordinate is needed before assembly. This transformation is
accomplished via
kNG = [aN T 04]tkNG [aN T 04]. (5.235)
Alternatively, one may add the natural geometrical stiffness to the simpli-
fied geometrical stiffness and then proceed to global transformation proce-
dures.
5.12 Partly simplified geometrical stiffness

The simplified geometrical stiffness considers only rigid body rotations on
which only the axial natural force operates. However, when the bending or
torsional mode is important, the natural moments must also enter in the
geometrical stiffness.
We now introduce the partly simplified geometrical stiffness in which
only the rotation increments are considered, but the natural symmetrical
bending moments and the natural torsional moment generate, in addition
to the axial natural force, geometrical forces.
The change of geometry aNt:. induces incremental geometrical forces
PGt:. = a~t:.PN' (5.236)
which may be regarded as the result of rotating the force vector
P=a~PN' (5.237)
through the small angles PORt:.. We require P Gt:. in terms of PORt:.. The
nodal force increments are
P Gt:. = Rt:.e P , (5.238)
where Rt:. is the incremental rigid body rotation matrix
. POM -P05t:.]
Rt:. = [ -POM P04t:. , (5.239)
P05t:. - P04t:.
5.12 Partly simplified geometrical stiffness 109
and
R~
R~
R~
R~e= (5.240)
R~
R~
R~
Noting that
rOR~ = aORP ~, (5.241)
we may rearrange (5.238) as a function of the local nodal cartesian forces
P = {Xl Y I Zl Mxl Myl MzI X2 Y2 Z2 Mx2 My2 Mz2 },

(5.242)
as
P3 -P2
-P3 PI
P2 -H
Pij -P5
-Pij P4
PG~= I
P5 -P4
Pg R
- a I aORP~
(5.243)
-Pg P7
Pa -P7
Pl2 -Pn
-P12 PlO
Pn -PlO
This equation may be written for each natural mode separately by replacing
P r by PNpaNpr in the above matrix. Thus
aNp3 -aNp2
-aNp3 aNpl
aNp2 -aNpl
aNp6 -aNp5
-aNp6 aNp4
PGt:. = PNp aNp5 -aNp4

aNp9
. IaORPt:. = kGPt:., (5.244)
-aNp8
-aNp9 aNp7
aNp8 -aNp7
aNpl2 -aNpll
-aNpl2 aNplO
aNpll -aNplO
where aNpr = rth member of row p of aN. In (5.244), kG is the partly

simplified geometrical stiffness which includes the contribution of all natural
forces but takes into account only the rigid body rotations of the element.
Geometrical stiffness for PNl
To form kG we substitute in (5.244) the elements ofrow of aNI in (5.131)

and get
-1
1
kG! = PNl aOR (5.245)

1
-1
Carrying out the matrix multiplications we are led to

1 1
I -I
1 1
I -I
kGl = PNl
1 1 (5.246)
-I I
1 1
-I I
This symmetric matrix is identical to the simplified geometrical stiffness of

the previous section.
Geometrical stiffnesses for PN21 PN4
To form kG for PN2 we substitute in (5.244) the elements of row of aN2

(5.131) to obtain
1-1
= PN2 (5.247)
A
kG2 aOR
-1
Carrying out the matrix multiplications

1 1
-I 1
1 1
-2" -2"
kG2 = PN21
(5.248)
1 1
:I
1 -I
1 1
2" 2"
Proceeding in a similar manner for the symmetrical moment PN 4 we get
1 1
1 -I
1 1
-2" -2"
I
kG4 = PN4 .
(5.249)
1 1
:I
-I 1
1 1
2" 2"
Geometrical stiffness for PN6
To form kG for PN6 we substitute in (5.244) the elements of row of O,N6 in

(5.131) and get
-1
1
kG6 = PN6 aOR (5.250)
1
-1
Carrying out the multiplication with aOR
1
Y -y1
1
-y1
kG6 = PN6 Y
(5.251)
-y1 1
1 Y 1
-y Y
we observe that the geometrical stiffnesses due to PN2, PN4, PN6 are
unsymmetric. We examine now whether symmetrization is feasible. By
adding the terms
Yl PN2
= +-1-(()2 - )
()l ,
PN2
Y 2 = --1-(()2 - ()l ) ,
(5.252)
PN2 )
Mxl = +-2-('l/J2 - 'l/Jl ,
PN2 )
Mx2 = +-2-('l/J2 - 'l/Jl ,
matrix kC2 becomes symmetrical. The same holds for PN 4. These terms
are proportional to the deformations (fh - (h) and {'l/J2 - 'l/Jd which are
small compared to deformations {V2 - vd, (lh + ( 2) which are proportional
to rigid body movements according to equation (5.20). Likewise, we may
symmetrize kC6 by adding the terms
PN6 )
YI = +-l-{<PI - <P2 ,
PN6
Y2 = --l-{<PI - <P2 ) ,
(5.253)
}JN6 )
ZI = +-l-{'l/JI - 'l/J2 ,
}JN6 )
Z2 = --{'l/JI-'l/J2 .
1
These terms are proportional to the deformations (<PI - <P2) and ('l/JI - 'l/J2)
and are small compared to rigid body deformations -(V2 - VI), (WI - W2).
Therefore, we may use the symmetrical parts of kC2' kC4 and kC6.
The contributions of the antisymmetrical moments PN3, PN4 to the
partly simplified geometrical stiffness are negligible and they will be ig-
nored.
5.13 Computational experiments

We now validate the theory we have developed with numerical examples.
5.13.1 Isotropic beams and frames

We consider thin isotropic beams of simple geometry with various boundary
conditions and loadings for which exact analytic solutions are available in
the text of Beer and Johnson [32]. We select six test problems shown in
Figs. 5.9, 5.lD, 5.11 and 5.12. All geometrical and material parameters
are stated in the figures. We compare displacements and slopes at selected
points using our beam element with and without shear deformation with
the analytic solutions of [32]. In addition, shear force and moment diagrams
are compared with their exact values for the beam illustrated in Fig. 5.11.
A perusal of the results reveals that convergence is achieved with only a
small number of degrees of freedoms in the finite element model. Minor
differences are observed when shear deformation is included. In the latter
5.13 Computational experiments 115
-Ite
p
1 . d A
~
elements dof Kirchhoff shear deformation
1 6 0.0120 0.0120104
2 12 0.0120 0.0120104
WA analytic 0.0120 0.0120104
~::J
shear correction factor 0.8333
E= 90.109
[ J V =0.3
A m I = 0.30
~ r>J b =h =0.01
elements dof Kirchhoff shear deformation
2 11 3.28125 x 10-4 3.31090 x 10-4
4 23 3.28125 x 10-4 3.31090 x 10-4
wA analytic 3.28125 x 10-4
Figure 5.9: Deflection of isotropic beams; comparison with the analytic

solution
case, a shear correction factor equal to 5/6 is automatically computed and

multiplies the transverse shear stiffness.
Next we consider isotropic frames. We choose a model problem from
Wo
E = 90.109
v= 0.3
I = 0.30
b = h = 0.01
Wo= ION/m
slope atA deflection at C
-6
Kirchhoff 2.9883 x 10-5 2.5268 x 10
4 elements
shear def. -5 -6
23 dof 2.9979 x 10 2.5499 x 10
-6
analytic 3.1000 x 10-5 2.5312 x 10
A
P
1
~
1 E= 90.109
v =0.3
I = 0.30
b = h =0.01
/<J I>j p= lOON
a = 1/3
I<l I>j.:::J I>j.:::J [>j
slope atA deflection at C
-3
Kirchhoff 1.1111 x 10-3 1.0000 x 10
3 elements
18 dof shear def. -3 -3

1.0938 x 10 1.0052 x 10
-3
analytic UIII x 10-3 1.0000 x 10
Figure 5.10: Deflection and slope comparison for two isotropic beams
text of Meek [42]. It examines a plane frame under the action of a horizontal
force as shown in Fig. 5.13. For all structural members, values for the
axial forces, moments and shear forces were given in [42]. They are shown
together with the present results in Fig. 5.14. The convention assigned
E= 90.109
1
w w
C-]
v =0.3
A BCD IE / = 0.20
im ,Id, b = h =0.01
w= 10N/m
a =/12
dof slope at E deflection at E
-5 -6
Kirchhoff 9.0122 x 10 8.3946 x 10
48
shear def. -5 -6
9.0123 x 10 8.4124 x 10
-5 -6
Kirchhoff 8.9331 x 10 8.3553 x 10
84
shear def. -5 -6
8.9335 x 10 8.3730 x 10
-5 -6
analytic 8.8889 x 10 8.3333 x 10
Q analytic -- wa = - 1.0 shear force diagram
B D
moment diagram
M _ 2
analytic - IV II 12 = - 0.005
Figure 5.11: Deflection, slope and moment comparison with the analytic
solution for an isotropic beam
by us for the stress resultants is also shown therein. Excellent agreement

is achieved. In Fig. 5.15 we show the breakdown of the various natural
w= 30kN/m
.Vt III I I I I I C " D

II 1
e= 200. 109
v -0.3
1= l.20m
b =0.036
eI 'I F
h -O.IOm
w=30N/m
0.036 m
a =112
H
DI,,
141<N 141<N
0.4 m
<l 0.8 m c-j

1.2 m
dof deflection at C
Kirchhoff 3.3578 x 10-3

36
sheardef. 3.4093 x 10-3
Kirchhoff 3.2178 x 10-3

54
shear def. 3.2655 x 10-3
analytic 3.2400 x 10-3
Figure 5.12: Deflection of an L-shaped isotropic beam structure
strain energies in the individual finite elements. They are all expressed on a
percentage basis. From these bar diagrams one can immediately see how the
various components of the strain energy are allocated during deformation.
Figure 5.16 illustrates a four storey frame which was studied also in [42]
where reference values for the moments were reported (see Fig. 5.16). The
values for the axial, bending and shearing forces as obtained by our beam
element are shown in Fig. 5.17. The moment values agree very well with
the solution of Meek [42]. Figures 5.18, 5.19 display the deformation of
the isotropic frame under the action of the horizontal force and a vertical
applied displacement at the right support, respectively. In the latter case

we notice that the foundation settlement induces rigid body motion.
5.13.2 Composite beam structures

Having evaluated our methodology with reference analytical solutions for
isotropic beams and frames, we proceed now to the study of laminated com-
posite beams. We select a (45/ - 45)8 cantilever composite beam shown in
Fig. 5.20 with all the material, geometrical and loading parameters. An
analytic solution for this model problem (using strip theory and classical
lamination theory) was given by Calcote [29]. This test example was also
studied by Venkatesh and Rao [33] and Oral [35]. The tip displacements
as obtained with only one element is compared with the analytical solution
of Calcote [29] and the finite element solution of Venkatesh and Rao [33]
in Fig. 5.20. The present results are in very good agreement with respect
to both analytical and numerical solutions. Note that our element reports
identical u, v. This was also reported in [33], [35]. Both finite element
solutions differ marginally from the analytical value of Calcote [29]. The
reason for this discrepancy may be that strip and beam theories use dif-
ferent assumptions for v. The convergence of the tip displacements with
mesh refinement is shown in Fig. 5.20. The simple beam element displays
excellent convergence characteristics. The through the thickness stresses at
the left clamped edge are shown in Fig. 5.21. We observe that the trans-
verse shear stress a xz is much smaller than the other stresses, with stress
a xx dominating.
Next we consider buckling of isotropic and composite beam columns.
For the former, analytical solutions can be extracted from the well known
Euler formula. For composite beams, analytical solutions (using strip the-
ory) was provided by Calcote [29]. Figure 5.22 illustrates a beam column on
which an axial compressive force is applied at the right edge. Of interest is
the estimation of the critical buckling load. We study an isotropic column
with both ends simply supported and two composite columns with lamina-
tions (45/ - 45/0/90)8 and (30/ - 30k The convergence of the buckling
loads for both material types is shown in Fig. 5.22. The agreement with
the analytical solutions is evidently very good.
We conclude the computational experiments by studying the buckling
of a three dimensional composite satellite beam space structure. The com-
posite satellite is depicted in Fig. 5.23. Also shown is the finite element
mesh. All members comprise 8 layers of the quasi-isotropic type with lam-
ination (45/ - 45/0/90)s. In order to obtain the critical buckling load we

10
21
1 12 elements 1 10
66 degrees of freedom
~ ~
tr
20
4, f4
21 .60 ' 2.14
ft4~~r4
10.91 10.91
t1
0.23
4.93{ f ' 4
0.23
t7
10.45
4.93
10.45
4
I 2.14
fit
21.13
21._~ 21.13
3.51
3.511~
d.
15.84 t
2.14
IS.
211.17
ll4
,----=--:...J -==------------.
29. 10~t 2~. 17~

ll4
Figure 5.13: Isotropic frame under horizontal force; elemental forces and
moments
I ---.
- 4.93
-
+Qt 5M
----t:>+N
0
+ 2.14 - 2.14
M Q
~t !t-N
,... M
L.........c=
_ 2.14 I reference sol. -21.40
- 21.4
o
reference sol. - 28.50 reference sol. 28.50
-2.14
+5.07
(!] +4.93
Figure 5.14: Reference force and moment diagrams for the isotropic frame
100
energy (%)
80
\0
21 60
10
40
20
k. t>l
o
Elem I Elem2 Elem3 Elem4
energy (%)
100
80
o axial straining mode

60
o symmeu-ical bending mode

40
;mlisymmelrical mode
20
o VI ~i
Elem5 Elem6 Elem7 Elem8

energy (%)
elem.5 elem.6 elem.7 elem.S 100 ~~------------------------------------,
==
~ elem.4 elem.9
80
~ elem.3 elem. \0
60
~ elem.2 elem. II 40
~ ~
-
elem. 12 20
element I
o VL..d ~I
Elem 9 Elem 10 Elem II Elcm 12
Figure 5.15: Percentage of the various elemental natural energies

10 kips
32 elements
132 in.
168 dof
1,0 kips ......
132 iii.
10 kips
132 in.
1,0 kips
E= 3,0 x I,oin.- kip
132 in.
1= lOOOin4
f<l .
,olD.
r>j
Ifoundation1,0 insettlement
V W=
reference solution for memeber moments (Meek)
513.6
1
1940.9 1940.9
Figure 5.16: Four storey isotropic frame; reference moment diagram

. ..
- 5.00
3.40 - 3 40
o
+ 10.00 - 5.00 -J( 00
+ 20.13 - 5.00
J
- 2( 13
l
+ 31.00
- 5.00
......
- 31 00
513.7 5137
J
o
1945 1941.5
- 3.42
+ 5.00 + 5.00
+ ).00 - 6.66 + 10.00
~
+ 15.00 - 10.05 + 15.00
+2O.Q3 + 20.03
- 10.93
Figure 5.17: Present force and moment diagrams for the 4 storey isotropic
frame
...,.--___l
==-=t
I
--'
/
/
1/----~11
I I~ I
Figure 5.18: Displacement of the 4 storey isotropic frame under the action
of a horizontal force
----
--
---
Figure 5.19: Displacement of the 4 storey isotropic frame induced by a

vertically applied displacement at the right support
a O. llb
EI - 3.107 psi
2 - 3. 106 psi
G12=G n =Gn = 106 p'i
v I2 =vJ3=vn =0.3
45
_ 45 y
:r _ 45
45
present 16-dof curved beam

tip analyticat(in) I eiement
I element
u 0.5102 x 10-7 0.5113 x 10-7 0.51 \3 x 10-7
v 0.\399 x 10-4 0.5113x 10-5 0.5113 x 10-5
w - 0.1274 x 10-3 - 0.1278 x 10-3 - 0.1291 x 10-3
+ Calcote Venkatesh and Rao
tip I element 2 elements 3 elements 5 elements
-7 -7
u 0.5113 x 10-7 0.51 \3 x 10 0.5113 x 10 0.5113 x 10-7
-5
v 0.51 \3 x 10-5 0.5113 x 10-5 0.51 \3 x 10 0.51 \3 x 10-5
w - 0.12782 x 10-3 - 0.12842 x 10-3 - 0.12783 x 10- - 0.12782x 10-3
Figure 5.20: Comparison of the tip displacements for a (45/ -45) s composite
beam with analytical and numerical solutions
restrain the bottom boundary while applying compressive loads at the top.
An eigenvalue analysis is conducted using the simplified geometrical stiff-
ness. The satellite structure bifurcates at a critical load of Per = 19079.6
Pa. The critical buckling mode is shown in Fig. 5.24. This example clearly
illustrates the applicability of our composite beam finite element to the
study of three dimensional beam structures.
l!h l!h
05 0.5
0.25
1
CJ
xx
7
Z4~1 0.25
_ 45
/
~~I / -4~1
0
-0.25
7
45
-0.51/1 1 -0.5
-3 -2 -1 0 2 3 4 -1 -0.5 0 0.5 (pSI.) 1
(psi)
l!h
r-------~:;:>'..::.i 0.5
;/' 10.25
I { I0
""" 1-0.25
1- =::"'" -0.5
-10 -8 -6 -4 -2 0
Figure 5.21: Through-the-thickness stresses for the {45/ - 45)8 composite

beam at the clamped end
~ J1~ p
i<' ~ ~
hI~
I" b .. I
buckling load both ends simply supponed

isotropic
simply supported Iclamped
isocropic Li.OIrn~~~
2 elements 82.8400 172.5730
3 clements 82.3767 169.2890 E=90. 109

v =0.3
4 elements 82.2889 168.6020 I =3.oom
b=O.OI m
9 elements 82.2480 168.2700
h=O.OI m
analytical* 82 .2467 N 167.8504 N
* Euler formula Lcompo;;;'

both ends .imply supported both ends simply $Upponed
buckling load
(451-45I0I90), (JOIJO)2
2 elements 13941.9 23878.0

EI = I. 109
4 elements 13844.8 23711.9 E2 = I. 106
G 12= G12= G12= 105

lOelemenlS 13837.5 23700.1
v I2 = v13= v23= 0.3
anolytical* 13837.8 23699.7 1= 100
- h=O.1
Calcote
Figure 5.22: Critical buckling loads for isotropic and (45/ - 45/0/90)8'
(30/ - 30h composite beam columns
I:)
,.0
-s
.....
"'""
~
@
o
00
Q)
Q)
Q)
~
+
+
Figure 5.23: A quasi-isotropic {45/ - 45/0/90)8 composite satellite struc-

ture; finite element mesh
Figure 5.24: First buckling mode of the composite satellite at a critical top
compressive load of Per = 19079.6 Pa; the bottom boundary is clamped
5.14 Problems
1. What is a natural rigid body and a natural straining mode ? How do
they relate to the various coordinate systems defined in this chapter?
2. Under which displacement assumptions can rigid body motion occur
in a beam element? Provide an example.
3. Write the expression for the elastic curve due to symmetrical and
antisymmetrical bending modes as a function of the coordinate x,
where x varies from 0 to l. Write also this expression when x varies
from -l/2 to 1/2. Using these coordinates, what would the equation
for the elastic curve be due to the fourth and fifth natural straining
modes?
4. Using elementary beam theory, prove expressions (5.45) and (5.46).
5. Explicitly evaluate the integrals k~l_k~5.
6. A beam element has length 10 cm. Under the action of external
loadings it displaces by Ul = 2, U2 = 3, VI = 3, V2 = 2, WI = 0.2,
W2 = -0.05, (h = 1.5, (h = -2, </>1 = 0.2, </>2 = -0.05, and "pI = 0.5,
"p2 = -0.001 in the local coordinate x'y'z'.
compute the natural rigid body and straining modes

estimate matrix ii. N
determine matrix Ao
compute matrix AN
7. Explicitly perform the integration in (5.169)
8. Evaluate the integrands for J~I_J~1 analytically.
9. Consider a 4-layer composite beam AB with global coordinates A (0,2,1),
B(1O,0,-3). Display the natural, local and global coordinates. After
orienting the beam in space, compute the direction cosine matrix To.
10. A laminated beam of thickness 0.01 m is composed of8 equal thickness
layers. Determine the Zk coordinates of the individual plies. What
are the latter when the beam comprises 9 layers of equal thickness?
11. What is the prerequisite for the beam to carry transverse shear defor-
mation? State the equilibrium of forces. In the absence of transverse
shear deformation, is equilibrium satisfied?
5.14 Problems 133
12. Express the equation of the elastic curve as a function of the beam
coordinate x when the latter varies between 0 and t. What is the
strain expression 'Yxx for this case?
13. What is the physical meaning of the strain operator matrix? Define
this matrix for a rod element under the action of the natural mode
PNl
14. Write the strain energy expressions using compact matrix expressions
for the beam element in the natural, local and global cartesian coor-
dinates. Show the dimensions of the matrices used.
15. In which loading cases do you expect the antisymmetrical bending

and shearing modes to affect the deformation significantly? what
about the influence of boundary conditions?
16. In which areas do you expect the linear through the thickness tem-
perature variation of (4.171) to hold more accurately?
17. Express equation (5.165) in all coordinate systems.
18. What is the physical meaning of the geometrical forces? Why are
rigid body rotations important in elastic buckling?
19. Follow the methodology of section 5.10.3 and derive an expression for
the through-the-thickness stress (1zz.
20. Derive all entries of the rotational matrix {5.213}.
21. Starting from the matrix equation A = a-I, provide cross relations
for the submatrices ao, aN, Ao, AN, ao, aN, Ao, AN' Prove that
AORa N = o.
- t -t
22. Show in a small example that successive finite rotations produce path
dependent work.
23. Provide examples in which you expect the partly simplified geomet-
rical stiffness to be more accurate than the simplified geometrical
stiffness.
24. Suppose that a two node beam finite element is under a sinusoidal
load of half period. Provide the kinematically equivalent nodal loads.
Chapter 6
Composite plate and shell

element
6.1 Introduction
Shells are used in many modern structures because of their efficiency and
economy, their ability to retain their form, and because of other features
stemming from their reaction to certain loads. Their shape allows certain
membrane stress systems to develop parallel to their tangential plane and
become prime carriers of the deformation. Indeed the analysis of many
thin shells is solely based on the membrane theory of shells which neglects
their bending rigidity. On the other hand, bending becomes important in
the presence of rapidly changing loads (e.g. concentrated loads, line loads
etc.), near edge constraints, near discontinuities in the shell geometry and
in nonlinear deformations. The development of a general membrane and
bending theory as well as related numerical implementations are subjects
of intense research efforts which aim at providing deeper understanding of
the mechanics of load carrying. The literature on this subject has been
growing at a rapid pace over the last decades.
When the shell thickness is comparable to its radii of curvature, trans-
verse shear deformation cannot be ignored (especially in multilayer com-
posite shells) and must be treated in the context of a rather general and
rigorous theory. Ultimately, finite element models must be developed for
the analysis of shell structures, whatever their complexity. Furthermore,
the introduction of composite materials and their ever expanding use in
high-performance lightweight structures has called for a unification of con-
cepts involved in the structural analysis of isotropic and anisotropic shells

136 Composite plate and shell element
as well as a full exploration of their tailoring capabilities.

The theory of shell structures has been the subject of many books and
numerous monographs. In cases in which the structural geometry assumes
a special shape (e.g. cylindrical, spherical, surface of revolution etc.), the
basic kinematical assumptions are usually simple. Related simplified theo-
ries may involve the following: direct stresses in shells of revolution; direct
stresses in cylindrical shells; direct stresses in shells of arbitrary shape;
bending of circular cylindrical shells; bending stresses in shells of revolu-
tion; and elastic stability [43], [44]. The complexity of the problem can be
reduced further for shallow shells, that is, shells whose middle surface does
not deviate much from a plane.
Despite significant advances in this field, it is still the objective of re-
searchers to develop reliable computer methods and procedures for the
analysis of general and arbitrarily shaped shell structures. Considerable
effort has been and is being extended to-tke construction of simple but
accurate finite elements that can be used in both industrial applications
and academic research. Many shell finite elements are available as part
of general purpose finite element programs or in codes developed in aca-
demic institutions. In spite of the plethora of numerical tools, the efficient
and economical analysis of isotropic and especially laminated composite
shells of arbitrary geometry still remains a challenge, and presents a task
which requires a skillful blend of physical, geometrical and mathematical
concepts and ideas. What is really needed are lower order, inexpensive but
accurate, anisotropic triangular elements free from the usual deficiencies
of isoparametric displacement elements (i.e. membrane and shear locking,
spurious modes, reduced quadrature etc.) which can be used for arbitrary
shell geometries. It is imperative that these elements ensure zero strains
under rigid body motion, a condition which, if not satisfied, prohibits the
extension of the formulation into the realm of large displacements.
The starting point of our work was the isotropic triangular shell ele-
ment TRUMP [45] that was developed in the 70's and applied to numerous
linear and nonlinear computations. This element included transverse shear
deformation but had constraints: all triangles had to have acute angles.
In addition, physical lumping techniques and beam idealizations were used
to formulate some of its stiffness entries relating to transverse shear de-
formation. Using the original theory of the element TRUMP, we refined
and expanded the basic concepts of the Natural-Mode method [16] and
conceived the flat and shallow shell composite triangular elements LACOT
and LACOS, respectively [46], [47], [48], [49], [50], [51], [52], [53], [54],
6.2 Natural kinematics of the shell element TRIe 137
[55], [56]. These elements were applied to numerous linear and nonlinear
computations of isotropic and composite structures and have shown superb
behavior with respect to both accuracy and economy.
Nevertheless these composite shell elements did not bring perfection to
the theory of natural modes. The method of physical lumping was still
in use and explicit constraints were needed: when angles greater than 90
were involved the elements yielded negative stiffnesses. The authors then
decided to pursue further research to circumvent this and other constraints.
This was followed with diligence and persistence and led to a new theory
with no constraints. With the evolution of new concepts a unification of
the natural mode method was accomplished.
6.2 Natural kinematics of the shell element TRIC

We will denote the Euclidean triangle shown in Fig. 6.1 by 6123. Numbers
1,2, and 3 represent the three vertices. The three angles are defined as
L. a = 213, L. f3 = 123, L. 'Y = 231. (6.1)
We focus on a multilayered construction shown in Fig. 6.1 which will

degenerate -as a special case- to a sandwich or a single-layer configuration.
For this stack of layers (plies) we define the following coordinate systems
(see Fig. 6.1):
Definition 3 The material coordinate system 123 is defined for each

ply with axis 1 along the fiber (reinforcement direction) and axes 2, 3 per-
pendicular to it.
Definition 4 The x'y' z' coordinate system is referred to as the local ele-
mental cartesian coordinate system and is placed at the barycenter of the
triangle.
Definition 5 The xyz coordinate system is referred to as the global carte-

sian coordinate system. Ultimately, global equilibrium must be referred to
this system.
The z' coordinates of the plies are shown in Fig. 6.2. The cartesian strain
tensor is defined as
composite laminate
r
I ~I .. X'
~ a. x'
L
y'
ry
P~-- -x' r x'
.x
p natural coord. material coord.
x'
a
~ p
'~l
global coord. local coord.
x'
l:' x
~ x'
sandwich p~e-I--I isotropic panel I

r
y'
Figure 6.1: The multilayer triangular element; coordinate systems; angles

k=1
lk 1
k=2
recursion formula zk+1 2
IW2
ZI =+h12 k=1 t h
zk =zk_l- hk k=2 ...... .N+ 1
N
k= N+l
Figure 6.2: z-coordinates of the plies
_l(i i) .. _ (6.2)
"tiJ' - -2 u.'3 +u.,
"
~,J - X,Y,Z,
where u = {U X u Y u Y } is the 3 x 1 displacement vector. We represent

the shear strains "t, u by
"t =bxx ,yy ,zz J'i'Xy J'i,xz J'i,yz},

(6xl)
(6.3)
u
(6xl)
= {O'xx O'yy O'zz J'iO'XY J'iO'xz J'iO'yz}'
With these strain and stress definitions, a consistent expression for the
strain energy is (assuming O'zz = 0)
1
E: U =-,tu
2
1
=2" (O"XXIXX + O"YYIYY + v'2O"xyv'21XY + v'2O"xzv'21xz + v'2O"yzv'21YZ)
1 1
=2" (O"XXIXX + O"YYIYY + O"Xyv'2v'22" (u,y + v,x)
1 1
+ O"xzv'2v'22" (u,z + w,x) + O"yzv'2v'22" (v,z + w,y))
(6.4)
1 e
=2"(O"XXIXX + O"YYIYY + O"xy Ixy
~
engineering shear strain
e
+ O"xz I~Z +O"yz Iyz ).
~ ~
engineering shear strain engineering shear strain
The three direct total strains will be measured parallel to the edges of
the triangular element. We measure these strains if we place a triangular
strain rossete on an aircraft panel (see Fig. 6.3) and measure three strains
, namely Ita, It(3 and It'Y along the rossete edges. These direct total strains
implicitly contain all the plane cartesian strains.
Definition 6 The three strains measured parallel to the edges of a triangle

are called the total natural strains.
Remark 1 The total natural strains are equivalent to the three in-plane
cartesian strains in the sense that they contain the same information.
We group the three total strains in the vector
'ta]
It = [It(3 , (6.5)
It'Y
and the three cartesian strains in the vector
IXX ]
I' = [ 'Vy , bzz = 0). (6.6)
V2,xY
o
strain rosette
"fry
"fla
++
++
Figure 6.3: An aircraft panel; triangular strain rossete; total strains
The concept of total strain allows us to use the triangle (alternatively the
tetrahedron in three dimensions) as the fundamental block on which equi-
librium can be prescribed (see Fig. 6.4).
Turn to Fig. 6.4. Shown is a square element S in a state of plain strain
i.e. TZZ = TXZ = TYZ = O. During the course of an elastic deformation,
the element strain state can be fully described by the three cartesian strain
measures, namely TXX, TYY and TXY If we ignore any rigid body motion
occurring during deformation, the square element deforms into a parallelo-
gram with sides of length ~s(l + TXX) and ~s(l + TYY), respectively. Our
objective is to derive an expression for the strain Tt(a) along an arbitrary
direction forming an angle a with axis x. For this purpose we consider the
right triangle t::.ABC and the oblique triangle t::.A' B'G' into which the first
is deformed. Applying the law of cosines to t::.A' B'G'
(A' B')2 = (A'G,)2 + (B'GY - 2(A'C')(B'G') cos( ~ + 2Txy ).

2
(6.7)
Using the geometrical relations

Y&[J
oL
As
6= 90-2yxy
Yxy =112 (6 1+6 2 )
x
B B'
oLA x
lu C A' Ax (1+ Yxx)
90 + 2 Yxy
l)
E
~
Figure 6.4: Deformation of a square element; fundamental blocks in two

and three-dimensional elasticity
6.2 Natural kinematics of the shell element TRIe 143
(A'e') = ~x{l + ')'xx) ,

(B'e') = ~y{l + ')'yy) , (6.8)
(A'B') = ~s{l + ')'t(a)) ,
equation (6.7) becomes
~s2{1 + ')'t{a))2 =~x2{1 + ')'xx)2 + ~y2{1 + ')'yy)2

7r
- 2~x~y{1 + ')'xx)(l + ')'yy) cos{ 2" + 2,),xy). (6.9)
Taking into account the equalities
~x = ~scosa, ~y = ~ssina, (6.1O)
we reduce equation (6.9) to
~s2{1 + ')'t{a))2 = ~s2 cos2 a{l + ')'xx)2 + ~s2 sin2 a{l + ')'yy)2
- 2~s2 sina cos a{l + ')'xx)(l + ')'yy)cos{ i+2')'xY~6.11)
Following the assumptions of linear elasticity, we consider ')'xy to be small.

Therefore
cos{i + 2,),xy) = - sin{2,),xy) ~ -2')'xy. (6.12)
A combination of (6.11-6.12) yields
{1 + ')'t{a))2 = cos2 a{l + ')'xx)2 + sin2 a{l + ')'yy)2

(6.13)
+ 4sinacos a{l +')'xx)(l +')'yyhxy.
Expanding terms, we find
(1 + 2')'t{a) + ')';{a)) + 2,),xx + ')'~x) + sin2 a{l + 2')'yy + ')'~y)

= cos 2 a{l
+ 4sinacosa{1 + ')'xx + ')'yy + ')'xx')'yyhxy. (6.14)
Neglecting second and third-order terms, namely
r;(a) = r;x = r;y = rxxrxy = ryyrxy = rxxryyrxy ~ 0,

(6.15)
we rewrite (6.14) as follows:
1 + 2rt (a) = sin2 a + cos 2 a + 2rxx cos2 a

(6.16)
+ 2ryy sin2 a + 4 sin a cos a rxy .
Remembering that cos 2 a + sin2 a = 1, we reduce equation(6.16) to
1 + 2rt (a) = 1 + 2rxx cos 2 a + 2ryy sin2 a + 4sina cos a rxy ,

(6.17)
rt(a) = rxx cos 2 a + ryy sin2 a + V2 sinacos a(V2rxy ).

(6.18)
Equation (6.18) expresses the total natural strain rt in any direction a as

a function of the cartesian strains rxx, ryy and rxy as well as the angle a
that this direction forms with the x axis. The two strain measures It, I'
are connected via matrix B as follows:
[ [
It = Bt,' ~ rt(31 C~""
rta
C2
a",'
s2
a y'
S~y, V2s (3"" c(3y'
1
V2 s a '" ,Ca Y ,] ['"II/xx
ryy .
rt"( c"("" s"(y' V2s"(",,c'Yy' V2r xy (6.19)
The angles ax" a y', etc. are shown in Fig.6.1. As long as one of the
triangular angles is not zero, equation (6.19) can be inverted to read
It = B t r , => I '=
B - t It, (6.20)
where we have used the notation
B- t = [Bttl. (6.21)
Remark 2 Three total natural strains suffice for the definition of the in-
plane cartesian strain field.
The notion of total strain leads us to more general concepts regarding

representation of a vector. Turning to Fig. 6.5 we observe that a vector
can be decomposed in a non-unique manner in components using a paral-
lelogram law; from this concept arises the notion of component strain. Al-
ternatively, a vector can be decomposed in a unique manner by orthogonal
projections in three axes giving rise to the notion of total strain. Finally, we
note the familiar cartesian definition according to which the vector projects
orthogonally to two perpendicular axes. We stress that all these definitions
are equivalent.
A physical idealization of the total natural strains is depicted in Fig.
6.6. We notice that straining of one side leaves the other triangular sides
unstrained. However, due to this strain, component stresses develop on all
triangular edges.
To the total natural strains correspond component natural stresses
grouped in the vector
ca
Uc = [uu c{3
] (6.22)
Uc-y
To the natural component stresses correspond cartesian stresses, viz.
Uxx ]
u' = [ U yy . (6.23)
../2uxy
The square root is introduced here in order to define the shear stress using
tensorial notation [51]. Using this notation, we can write the strain energy
expression for plane stress as
: ~,tu = ~(O"xx'Yxx + O"yy'Yyy + V2O"xy V2'Yxy)

1 1
= 2"{O"xx'Yxx + O"yy'Yyy + O"xyV2V22" {u,y + v,x))
1 e (6.24)
= 2" {O"xx'Yxx + O"yy'Yyy + O"xy 'Yxy ).
"'-..I"""
engineering shear strain
natural and cartesian
p
x
p
0.
I
0.
component definition total definition

Cartesian definition
( non unique decomposition of a vector) ( unique decomposition of a vector)
n'y
r=rc~+ 'q + rca r="p + '/"(+ ',a
r=,x+ ry
,x
0. p p
0.
component strain cartesian strain I 1- total strain I

Yn
Figure 6.5: Component, cartesian and total representation of a vector r

2
Ym =l!Jalla
l!Jp= l!J, =0
Po.
CJ,u
Ik~I-L..J....-.......- I
'!-It.....J'-.1-L...L...J.wl aen
CJcb
Po.
Po.
P,
Po.
Pp
Figure 6.6: Total strain idealization; component stresses; nodal loads

We express the equivalence of the strain energy density in the natural and
local cartesian coordinates, viz.
c. 1 t _ 1 't ,(6.20) 1 tB- 1 ,

" . 21tUc - 21 U - 21t U, (6.25)
--......-.-
natural
~
local
where e is an energy operator symbol. Note that there are three pairs of
stresses and strains to operate with, these being, U Clt , Utlc' U I , where
I c are the component strains which we did not apply here. As a result, the
expression for the strain energy may be expressed in the alternative forms
1 tIt 1 t
e: 21tUc = 2/cUt = 21 u. (6.26)
From (6.25) we deduce that
U e = B -1 U , {:::::::} U' =
B U c. (6.27)
To this end we regroup the derived expressions for strains and stresses as
follows:
[;:] = [~t B~1] [;:] (6.28)

[;:] = [B.-t B] [;:]
As for the total axial natural strains we define total natural, transverse
shear strains shown in Fig. 6.7. According to our physical idealization,
transverse shearing of one side leaves all other side angles orthogonal. The
total transverse shear strains are related to the local cartesian transverse
shear strains via
[~J. :ox,] :>/s=Ts/~.

COx'
[
cf3x ' [rxz] (6.29)
f3 x' ryz k
C-yx' s'Y x ' k
/~
'Yqa b /. -' '
\ G
tr/, ,tt,
qa ..........
Qa
'.2
q~
Qr
Itt[;~=fJ
a,
Figure 6.7: Total transverse shear strain idealization; component shear

stresses; nodal shearing forces
for the transverse shear stresses
USOl.]
[Us /3 COl. X'
[ c/3x'
SOI.X']
s/3x'
[UXZ] {:} Us
u yz k
= Tsu~. (6.30)
U s'"( k Cux' Sux' k
The two previous equations which are derived from pure geometrical argu-
ments can be grouped together as follows:
[,s] = [Ts .] [,:] . (6.31)

Us . Ts Us
Recalling equation (6.28) we write
[~~l ~ [Bt B- 1
Ts
J[~l (6.32)
6.3 Constitutive relation

Definition 7 A coordinate system with axes parallel to the edges of a tri-
angle is called the natural coordinate system and is denoted by af3'Y.
In the natural coordinate, the constitutive relation for each layer (k) reads
UCOI. /'i,0I.0I. /'i,0I./3 /'i,0I.'"( 'YtOi.

u c/3 /'i,0I./3 /'i,/3/3 /'i,/3'Y 'Yt/3
uC'"f 'Yt'"(
=
/'i,0I.'Y /'i,/3'Y /'i,n
UqOl. XOI.OI. XOI./3 XOI.'Y 'Yo. I 1(6.33)

uq/3 XOI./3 X/3/3 X/3'Y 'Y/3
uq'Y k XOI.'Y X/3'Y X'Y'Y k 'Y'Y .J k
6.3 Constitutive relation 151
in which Xij are the transverse shear stiffness coefficients. Using compact
notation we write
[O'C] [K.ct .] ['t] (6.34)

0' 5 k . Xs k 15k
We observe from equations (6.33-6.34) that the axial stresses remain un-
coupled from the transverse shearing strains.
With respect to the material coordinate system, the constitutive rela-
tions read
E1 E2
1- V12V21 V 12 1 _ V12 V 21
[
0"11
0"22
1 - E2 E2
[
1
1'I'll
22 ,
V20"12 k
V12 1- V12 V 21 1- V12V21 V21'12 k (6.35)
2G12.J k
[V20"13]
V20"23 k =
[2G 13
. 2G 23
.] [V21'13]
k V21'23 k
(6.36)
Equations (6.35-6.36) can be put in the condensed form
[ 0'12] = [K.12 .] ['12] (6.37)

Us k . Gs k IS k
In general all material properties are temperature dependent. Using the

relations [27]
V12 V21
E1 = E2
E2 E1
(6.38)
V12
(1 - V12v21)
= V21 (1 - V12V21
)'
we can write K.12 alternatively as

El V12E2
1 V12 E 2 E2
~12 = 2 E2
1- v 12 El (6.39)
2 E2)
2(i12(1 - V 12 El
For complete isotropy, V12 = V21 = v, El = E2 = E, (i12 = (ij then this

equation reduces to
1
~12 = - - vE
[E vE
E
1- v 2 2G(1:- V2J
E E
v-- (6.40)
1- v2 I- v2
E E
v--
1- v 2 I- v2
2(i
We initiate now the following sequence of material transformations.
material coord. 1---+ local coord. 1---+ natural coord. OP
where OP stands for a sequence operation.

It is our objective to create the natural elemental stiffness matrix. We
begin by contemplating once again the equivalence of the strain energy in
the various coordinate systems via
c: 21'12t 0"12 = 1 It ,
-"V 1
2 ,t
t
(6.41)
------- -----
2 0"
I O"c
~
material coord. local coord. natural coord.
The following relation holds between strains and stresses, in the material
and local coordinate systems [51]
['12]
IS
[A. As.] ["]
IS
(6.42)
[0"12]
O"s
= [~12
.
.] ['IS12] = [~12A.
GS
.] [,:]
GsAs IS
.
(6.43)
To derive matrices A, As, we use the following notation:
U, v E xyz material coordinate,

u ' , v' E x' y' z' local coordinate.
In two dimensions we have
[u'v' ] - [-sinO
cos 0 0] [u]v
sin
cosO -
[C(J S(J]
-S(J C(J
[u]v ' (6.44)
where 0 is the angle between x' and x. Also
[~] = [~: ~:(J] [~;] . (6.45)
Similarly,
[:] = [~: ~:(J] [:;] , (6.46)
I
which implies
-ax -ay
ax' ax '
[ ax ay = [~!(J ~:]. (6.47)
ay' ay'
Using the chain rule of differentiation we have
a]
ax' = [aax ax'
ax + oya ax'
oy]
[
-
a -
a -ax+ a-oy -
. oy' ax oy' oy oy'
:][:]
ax
[ (6.48)
= ::
oy' oy' oy
= [co so]
-so Co
[:x].
a
oy
Now the strain expressions in the local coordinates read
au' = -(cou
'YX'X' = -
a + sov)
ax' ax'
= (co! +so~)(cou+sov)
20U 20V (OV au (6.49)
= co- oy + coso -ax +-
ax +so- oy
2"xy
= chxx + shyy + v'2soco(v'2'YXy),
(6.50)
2"xy
= chyy + shxx - v'2soco( v'2'Yxy),

V2'YXlyl = V2'21 (OU' OV' ) = V2'21 ( oy'

oy' + ox' 0 (cou + sov) + ox'
0 (cov - sou) )
~ v'2*CO ~ - SD :X) (CO u + SD') + (CO :X +so ~) (CO' - SDU)]

20U + Co-
1n1 [(Co-
= v2- 20V + -So-
20V - 20U + 2coso (OV
So- - -OU)]
-
2 oy ox ox oy oy ox
2- [ Co2 (OU
= v1n1 - + -OV) - (OV + -OU) + 2soco (OV
So2 - - - -OU)] (6.51)
2 oy ox ox oy oy ox
~ v'2~ [v'2c~ (v'21.") - v'2s~ (v'21'",,) + 2s,co b.. - 1)]

= V2~V2{c~ - s~)(V2'Yxy) + V2~2Socobyy - 'Yxx)
= (c~ - s~)(V2'Yxy) + V2socobyy - 'Yxx).
We can replace indices xyz with 123 in equations (6.49-6.51) and obtain the
local cartesian strains from the material strains 'Yn, 'Y22 and 'Y12. Matrices
A, As are easily deduced as
c2
o s~ V2S0CO]
A= [
So2
c~ -V2s oco
-V2soco V2soco (c~ - s~)
(6.52)
Co so]
[-so
As = Co
Using A, As, we find the expression for the strain energy in th~ material
coordinates 123
"
c- :
1 t
2'120'12 2' It[At K12,
{6.42},{6.35} 1
=
A]'
1 't
= -,
2
[K'lT'
(6.53)
= ![Bt,,]tKct[Bt,']
2
= ~,/t[BKctBtlT"
Thus, for each layer (k)
Kk = [AtKI2A]k = [BKctBt]k
(6.54)
K~t = [B-l[AtKI2A]B-tL
Proceeding in a similar manner for the transverse shear stiffness we find
1 t 1 't ,
& : 2'sO's = 2'sO's
1 't [ t ] ,
= 2's AsGsAs,s (6.55)
1 't [ '] ,
= 2's Gs 'S,
G's = [A~GsAs]k' (6.56)
where
G's = [2Gxx
2G xy
2Gxy]
2G yy . (6.57)
The next step is to compute the natural transverse shear stiffness. Ob-
serving (6.29) and aiming at solving for the transverse shear stresses as
functions of the natural shearing strains we realize that we have at our
disposal three equations containing only two unknowns. It appears appro-

priate to solve (6.29) in pairs and produce a shear stiffness in a weighted
sense. The following relations hold for the three natural strains
= cJ.lXI'Yxz + cJ.lyl'Yyz,
'YJ.I
(6.58)
'Yv = CVX''Yxz + cvy''Yyz, jL, v = Ct', {3, 'Y.
Solving the above system of equations for 'Yxz, 'Yyz we find
['Yxz] = lJ.llv [ cvy -CJ.ly] ['YJ.I] . (6.59)

'Yyz 20 -Cvx cJ.lX 'Yv
Using the trigonometric definitions
Xo: = lo:co: x = X3 - X2, X{3 = l{3c{3x = Xl - X3,

X"( = l"(C-yx = X2 - Xl, Yo: = lo:co:y = Y3 - Y2, (6.60)
Y{3 = l{3c{3y = YI - Xy, Y"( = l"(C-yy = Y2 - YI,
we obtain equation (6.59) as
IS = ['Yxz] = ~ [ lo:Y{3 -l{3YO:] ['Yo:] = W 0:{31 0:f3'

'Yyz 20 -lo:x{3 l{3xo: 'Y{3
, _ ['Yxz] _ ~ [ l{3Y"( -l"(Y{3] ['Yf3] = W {3"(1{3"('

IS - 'Yyz - 20 -l{3x"( l"(x{3 'Y"(
(6.61)
, _ ['Yxz] -_ ~ [ l"(yo:
IS- -lo:Y"(] ['Y"(] = W "(0:1"(0:'
'Yyz 20 -l"(xo: lo:x"( 'Yo:
We use the principle of virtual work in order to generate the natural trans-
verse shear stiffnesses. We have
8Us = ! u 'st , dV =
8,s ! 't,
IS , dV.
G s 8,s (6.62)
v v
On account of equation equation (6.61) we may write equation (6.62) as
6Us = J ;~.B[W~.BGsWQ.B]h/Q.BdV=;~.B J[W~.BGSWQ.B]dV6;Q.B.

v v (6.63)
Proceeding in a similar manner for the other two shear strains we obtain
the following three expressions for the transverse shear stiffnesses:
Xll X12]
Xl = W~.BGsWQ.B = [ b ~2'
Xl Xl
Xll X12]
X2 = W~'YGs W.B'Y = [ I2
X2
~2'
X2
(6.64)
X3 = W~QGsW'YQ = [ Xll
~2
X12]
~2
X3 X3
To retain symmetry, it appears natural to create an average of Xl' X2, X3

in the sense
1 [XP + X~2 xt2

+ Xl22 X~2 1
Xs =3 xF X ll
2
Xl2
X~2 X~2 X~l ~ X~2 (6.65)
Both axial and transverse shear stiffnesses K,ct, Xs have now been defined
in the natural coordinate system.
6.4 Stress resultants - equilibrium
The axial stresses produce the nodal forces and moments

6.4 Stress resultants - equilibrium 159
= n!
h/2
Pa = 1
2CJca h ah la k
CJcadz,
-h/2
1.
P {3 = 2CJc{3h{3h = l{3n! h/2
CJck{3dz, (6.66)
-h/2
P -y = 21 CJe-yh-yh
k n!
= l-y
h/2
k
CJc-ydz,
-h/2
1
MSa = 2mSa h a h = la n! h/2
k
zCJcadz,
-h/2
MS{3 1
= 2ms{3h{3h = l{3 n! h/2
k
zCJc{3dz, (6.67)
-h/2
Ms-y 1
= 2ms-yh-yh = l-y n! h/2
k
zCJe-ydz.
-h/2
where mSa, mS{3 and ms-y, are the moments per unit length. The nodal
force and moment resultants are shown in Fig. 6.8.
Remark 3 The nodal forces and moments constitute self-equilibrating stress

systems. However, these systems which are based on the constant stress as-
sumption cannot carry the transverse shearing forces.
The generation of transverse shearing forces requires variable moments.

For this purpose we augment our system of moments with antisymmetri-
cal moments assigned as pairs along the edges of the triangle. To satisfy
equilibrium, shearing forces are needed. They are shown in Fig. 6.9. The
equilibrium of forces and moments in the natural coordinates demands
po. W ,p.
MsfI.
( 2
)Sa
M
cb
MS')'~
Py
Figure 6.8: Component and nodal stresses and moments

6.4 Stress resultants - equilibrium 161
11" "Y"II" ~"=Z,,Jha

11~ =Y~/I~ ~ =~ / h~
n =Y II r,.., "Z1 I h1
'1 1 1
2 11" .Y" 3 I" 10. "I
N"
MS'( Q.,
M"~ ND
... ( :::4'1
pi
M
< IN"
i4N
"( M:
"
a: M"
\'..i,/
:I
N"
p.4 M((lt?>2?!))
.10. MStl M
. p
MStl A"
Figure 6.9: The 12 generalized natural forces and moments

Pa:::: P a
P,8::::P,8
P')'::::P')'
MSa:::: MSa
Ms,8:::: Ms,8
Ms')' ::::Ms')'
(6.68)
Qa- 2MAa =0
la
Q,8 - 2MA,8 = 0
l,8
Q,),_ 2MA')' =0
l')'
Ma +M,8 +M')' =0
We augment the system of forces and moments by three simulative azimuth

(or drilling) moments M a , M,8 and M')' shown in Fig. 6.9. The complete
vector of generalized natural forces is
PN ={PNa PN,8 PN,), MSa MAa Ms,8 MA,8 Ms')' MA,), Ma M,8 M')'}.
(12xl) , " (6 6~)
generalized natural forces .
Next, we define pairs of natural and cartesian stress resultants as follows:
PN = [PN']
PN,8
PN,),
= r. l
P,8l,8.]
P')'l')'
Fe = [ N
xx
Nyy ] ,
../2Nxy
(6.70)
Ms =
[MS,]
Ms,8 Me= [ M .. ] ,
Myy (6.71)
Ms')' ../2Mxy
QN ~ [~~] Q e-
_ [Qxz]
Qyz . (6.72)
6.5 Natural modes and stiffness 163
ZI', WI
.'v'
~
<2', .W2'
'V'
'V '2
loW
x', u'
0'
'V
y. v
x, u 0'3
o
Figure 6.10: Multilayer triangular shell element with local cartesian free-
doms; local and global coordinates
The relation between the natural and cartesian stress resultants reads
[~~] = [~ B] C~:] (6.73)

Qc = [~(lX'
(ly, ~~X'
~y'
~X']
~y'
QN
6.5 Natural modes and stiffness

In Fig. 6.10 we show the multilayer triangular shell element TRIC. As
for the beam element, its natural stiffness is based only on deformations
and not on associated rigid body motions. The element has 18 degrees of
freedom but the actual number of straining modes is
18 cartesian d.o.f - 6 rigid body d.oJ = 12 straining modes.

(6.74)
We proceed as in Fig. 6.11: we project the nodal displacements and
rotations and corresponding forces and moments on a triangular edge, say
Q'. Then, we decompose the rotations and moments into symmetrical and
antisymmetrical components. From this decomposition the axial as well as

symmetrical and antisymmetrical modes of deformation arise. As for the
beam element, the latter carry -in a series connection- both antisymmet-
rical bending and transverse shear deformations. We proceed in a similar
manner for the other two triangular sides.
Now we have generated invariant deformation measures, the natural
modes, to which correspond generalized natural forces P N. The element
TRlC includes 6 rigid body and 12 straining natural modes illustrated in
Figs. 6.12, 6.13, and grouped in the vector
, PO]
(6xl)
(l~:l) = [ PN
(6.75)
.
(12xl)
Po, PN represent the rigid body and straining modes, respectively, with
corresponding entries
Po = {POI P02 P03 P04 P05 P06},

(6xl)
PN = h~ 'Y$ 'Y~ 'l/Jso: 'l/JAo: 'l/Js/3 'l/JA/3 'l/Js'Y 'l/JA'Y 'l/Jo: 'I/J/3 'I/J'Y}. (6.76)
(12xl)
The following subvectors are contained in (6.76):
1~ = h~ 'Y$ 'Y~} axial straining mode (6.77)

1/JS = {'l/Jso: 'l/Js/3 'l/Js'Y} symmetrical bending mode (6.78)
1/J A = {'l/JAo: 'l/JA/3 'l/JA'Y} antisymmetrical bend.+shearing (6.79)
tPo: = {'l/Jo: 'I/J/3 'I/J'Y} azimuth rotational mode. (6.80)
The antisymmetric mode is the sum of the antisymmetrical bending mode

plus the antisymmetrical shearing mode,
6.5 Natural modes and stiffness 165
l::
a1= 112 ('!'Sa + '!'Aa)
decomposition of rotatioos momepts
symmetric mode
aI s =1I2'!'sa
1I2Msa
antisymmetric mode
aI A =1I2'!'Aa
1I2MAa
'Z.JC'\ series connection

,t~, \~
-- ,
Figure 6.11: Projection and decomposition of the cartesian rotations and

moments on triangular side a
1/Ji = 1/J~ + 1/J:, i = 0, /3". (6.81)
Remark 4 The antisymmetrical mode is the key to shear locking elimina-

tion.
To the rigid body and straining modes Po, PN correspond generalized

forces and moments, viz.
P6 = {POI P02 P03} Po = {POI P02 P03} (6.82)

p~ = {P04 P05 P06} Mo = {MOl M02 M03} (6.83)
i~=h~ ,$ ,~} PN = {Pala P{3l{3 PyLy} (6.84)
1/Js = {'l/Jsa 'l/JS{3 'l/Js'Y } Ms = {Msa MS{3 Ms'Y} (6.85)
1/J A = {'l/JAa 'l/JA{3 'l/JA'Y} MA = {MAa MA{3 MA'Y} (6.86)
1/Ja = {'l/Ja 'I/J{3 'I/J'Y} Mz = {Ma M{3 M'Y}' (6.87)
All matrices will be referred first to a local elemental cartesian coordinate

x'y' z' and ultimately to a global cartesian coordinate system xyz. We now
define the following elemental displacement vectors including the nodal
degrees of freedom in the local and global coordinate systems:
P = {u' v' w' 8' / 'I/J'}n

(IBxl)
n = 1,2,3.
P = {u v w 8 'I/J}n (6.88)
(IBxl)
The natural straining vector PN is related to the elemental vector p via
PN = o,NP (6.89)
Matrix 0, N is a function of the current geometry of the element and will be

given in the sequel. The local elemental vector p is related to the global
elemental vector P via
p = T 06 P, (6.90)
where T06 is a matrix containing direction cosines [51]. Using equation

(6.90) we may write equation (6.89) as
PN = aNP = aNT06P f (6.91)
6.6 Total strain in the natural coordinate system

Along a natural coordinate i = a, {3, 'Y we define the vertical displacement w
as a cubic function of the symmetrical and antisymmetrical bending modes.
Observing Fig. 6.13 we write
Wi = '2
1 (,2
-t + Yi )'l/JSi 1(
+ '2 Yi -
3,2 + 2,3)
z:- II 'l/JAi,
, ~ , y ,
symmetrical bending mode antisymmetrical bending mode (6.92)

i=a,{3,'Y, li --+ 0 -;- Yi.
The horizontal displacement along a natural direction is
_ .8Wi
Ui = Uio -z 8Yi (6.93)
On account of (6.92-6.93), the total strain becomes
8Ui 8uo 82 w i
'Yti = 8Yi = 8Yi - z 8Yl
o 'l/JSi
='Yi +zT+ z
(3Z;- q
6Yi) .
'l/JA~
(6.94)
The terms coupling the total axial strain with the antisymmetrical bending
mode are very small and will be neglected. With this assumption, equation
(6.94) yields
natural rigid-body modes

z'
'1"
z'
z
x'
z' z'
@
~
[9 z'
o
Wo
Wo
x'
Figure 6.12: The 6 natural rigid body modes

~ ~/;
)}1/fJ ....{'J' 0')
I ==- I
1I21,u
11211'(
12V~
,,
IV;yl
,,
112V;u
Figure 6.13: The 12 natural straining modes

Ita = I~ + z 7/JSa
la
It(J = 1$ + z 7/JS(J {6.95}
l(J
It-y = I~ + z 7/JS-y
l-y
We shall need relations connecting the total strains ft' the natural dis-
placement vector PN, the local elemental cartesian vector p, and the global
cartesian vector p. They are
(6.91) _ _ (6.91) _
ft = ONPN = ONaNP = oNaNT o6 P {6.96}
An examination of {6.95} gives rise to the following matrix connecting the

total axial strains with the natural modes:
z
1 10
= [. z
J
ON 1 1{3
(3xI2) . 1 z {6.97}
r:;
6.7 Axial and symmetrical bending stiffness terms

The principle of virtual work in the natural coordinate states
8U = ! 0'~8f
v
tdV. {6.98}
A variation of the total strain can be expressed as

6.7 Axial and symmetrical bending stiffness terms 171
fJft = Ot.NfJPN = Ot.No,NfJp = Ot.No,N T o6fJp. (6.99)
Substituting (6.99) in equation (6.98) we find
fJU = ! t
u cfJft dV =
(6.34)! t
ft""ctfJftdV
v v
= P'" [ f a'".."aNdV ] /jPN, (6.100)
V
from which we deduce an expression for the natural stiffness matrix con-
taining contributions from the axial and symmetrical bending modes
kN([~' 1/JS) = !
V
Ot.hr""ctOt.NdV (6.101)
We now transform the natural stiffness matrix first to the local coordinate
and then to the global coordinate systems. Substituting equation (6.91)
in equation (6.100) we generate the following sequence of congruent oper-
ations: :
/jU = f
P'" [T~ [a'" [ a'" .."aNdV JaN 1T06] PN (6.102)
v
natural coord. (12x12)
v
local coord. (18x18)
, '
global coord. (18x18)
The (12xI2) natural stiffness matrix kN is

kN =0 (6.103)
(12x12)
-kQQ kQP kQ'Y l:kQQ r kQf3 ~kQ'Y
f3 1"1
kf3f3 kf3'Y I: kQf3 fpkf3f3 r "I k f3'Y
k'Y'Y l:kQ'Y fri kf3'Y tyk'Y'Y

%2 %2 k %2 k
tr
01
kQQ IQlf3 Qf3 101 1"1 01"1
kQQ k Qf3 kQ'Y
A A A
%2 %2 k
tr kf3f3 1f31'Y f3'Y
f3
f3f3
kA k f3 'Y
A
symm. %2 k
tr "1"1
"I
k'Y'Y
A
kQQ k,;f3 k';'Y
%
k~f3 k~'Y
k'p
For an isotropic plate or symmetric composite laminate the axial-bending

coupling terms disappear and the natural stiffness matrix reduces to
kN =0 (6.104)
(12x12)
- kQQ
kQf3 kQ'Y
kf3f3 kf3'Y
k'Y'Y
%2 %2 k %2
tr01 kQQ IQlf3 Qf3 IQI'YkQ'Y
kQQ k Qf3 kQ'Y
A A A
%2 %2
tr kf3f3 1f3 1'Y kf3'Y
f3
k f3f3 k f3 'Y
A A
%2
symm. trk'Y'Y
"I
k'Y'Y
A
k';Q k,;f3 k';'Y
k~f3 k~'Y
k'p
For isotropic materials only 18 entries need to be computed since the coef-
ficients corresponding to the azimuth rotations are arbitrarily assigned.
We now define the entries referring to the axial and symmetrical bend-
ing terms. Piecewise summation is used to account for the fact that the
6.7 Axial and symmetrical bending stiffness terms 173
material properties may be discontinuous between adjacent layers. The

following groups of stiffness entries are defined:
J
h
2" N
kaa = fi.aadz = L fi.~a(Zk - Zk+1),
h k=l
-2"
J
h
2" N
ka(3 = fi.a(3dz = L fi.~(3(Zk - Zk+l),
h k=l
-2"
h
J
2" N
ka"{ = fi.a"{dz =L fi.~"{(Zk - Zk+1),
h k=l
-2"
(6.105)
J
h
2" N
k(3(3 = fi.(3(3dz = L fi.~(3(Zk - Zk+1),

h k=l
-2"
J
h
2" N
k(3"{ = fi.(3"{dz = L fi.~"{(Zk - Zk+1),
h k=l
-2"
h
J
2" N
k"{"{ = fi."{"{dz =L fi.~"{(Zk - Zk+1),
h k=l
-2"
J
2" N
zkaa = Zfi.aadZ = "2 L....t fi.aa(Zk2 -
1", k 2)
Zk+l ,
h k=l
-2"
J
h
2" N
zka(3 = Zfi.a(3dz = ~ L fi.~(3(Zf - Zf+1)' (6.106)
h k=l
-2"
h
J
2" N
1", k
zka"{ = Zfi.a"{dz = "2 L....t fi.a,,{(Zk2 - 2)
Zk+1 ,
h k=l
-2"
J
"2 N
1~
zkf3f3 = ZK.f3f3dz = 2 L- K.f3f3(Zk
k 2
-
2
zk+l)'
h k=l
-"2
h
J
"2 N
1~
zkf3'Y = ZK.f3'Y dz = 2 L- K.f3'Y(Zk
k 2
-
2)
Zk+1 , (6.107)
h k=l
-"2
h
J
"2 N
1~
zk'Y'Y = ZK.'Y'Y dz = 2 L- K.'Y'Y(Zk
k 2
-
2)
Zk+1 ,
h k=l
-"2
J
h
"2 N
z 2kaa = z 2K.aadz = ~ ~ K.~a(z~ - Z~+1)'
h k=l
-"2
J
h
"2 N
z 2kaf3 = Z2K.af3 dz = ~ ~ K.~f3(z~ - z~+d, (6.108)
h k=l
-"2
h
J
"2 N
z2 ka'Y = z2 K.a'Ydz = ~ ~ K.~'Y (z~ - z~+1)'
h k=l
-"2
J
h
"2 N
z 2kf3f3 = z 2K.f3f3dz = ~ ~ K.~f3(z~ - z~+1)'
h k=l
-"2
J
h
"2 N
z 2kf3'Y = Z2K.f3'Y dz = ~ ~ K.~'Y(z~ - z~+1)' (6.109)

h k=l
-"2
h
J
"2 N
z2k'Y'Y = Z2K.'Y'Y dz = ~ ~ K.~'Y(z~ - z~+1).
h k=l
-"2
Now all terms due to axial and symmetrical bending modes have been
defined.
6.8 Antisymmetrical bending and shearing stiff-

ness terms
The antisymmetrical modes are split into two components for each trian-
gular side, namely the antisymmetrical bending modes 1/J~, 1/J~, 1/J~ and the
antisymmetrical shearing modes 1/J~, 1/Jp and 1/J~. We write
1/J a = 1/J~ + 1/J~,

1/J /3 = 1/J~ + 1/Jp, (6.110)
1/J'Y = 1/J~ + 1/J~.
The antisymmetrical rotations give rise to antisymmetrical moments
MA = KA1/JA' (6.111)
In matrix form
Kll
A
K12
A
K13
A
[MAO]
MA/3
K12
A
K22
A
K23
A
[~AO]
'l/JA/3 . (6.112)
MA'Y 'l/JA'Y
K13 K23 K33
A A A
If we suppress all shear deformations we may write
MA = K~1/J~, (6.113)
in which K~ is the stiffness matrix when only bending is allowed. Likewise,

when only shear deformation is permitted
MA = K~1/J~. (6.114)
The following two relations are evident

~/.b
'f'A b ]-1 MA,
= [KA
(6.115)
1/J~ = [K~rlMA.
Substituting equation (6.115) in equation (6.110) we get
1PA = [KAb]-1 MA + [8]-1

KA MA => MA = [[ KA
b]-1 + [KA8]-1]-1 1PA,
(6.116)
from which we deduce the antisymmetrical stiffness matrix, viz.
KA = [[K~] -1 + [K~] _1]-1 (6.117)

(3x3)
By setting the inverse antisymmetrical shearing stiffness to zero (which

implies an infinite transverse shearing modulus) Le., [K~rl = 0, we
recover the Kirchhoff solution as the thin plate limit.
6.8.1 Antisymmetrical bending terms

Figure 6.13 shows the selected antisymmetrical bending modes. They are
surfaces of the third order. Following this modelization, the vertical dis-
placements in the the natural coordinates are represented by the following
cubic polynomials:
1
WAa = "2la({3(,Y(({3 - (-y)'l/JAa
1
WA{3 = "2l{3(-y(a((-y - (a)'l/JA{3 (6.118)
1
WA-y = "2l-y(a({3((a - ({3)'l/JA-y
These displacements are functions of the homogeneous coordinates

6.S Antisymmetrical bending and shearing stiffness terms 177
WA = 1((00 (p, (-y), (a + (p + (-y = 1. (6.119)
In the process of deriving the elastic stiffness matrix, the following deriva-
tive matrix is needed:
8WAa 8WAa 8WAa

8(a 8(p 8(-y
8WA
8t;
= I8WAP
8(a
8WAP
8(p
8WAP
8(-y
8WA-y 8WA-y 8WA-y

8(a 8(p 8(-y _
(6.120)
(210(/3(-y-lo(~)'I/J Ao ( - 210 (/3 (-y +10 (3 )'I/J Ao -

=~ I(-21/3(-y(0+1/3(~)'l/JA/3 (21/3 (-y (0 -1/3(~)'I/J A/3
(21-y(0 (/3 -1-y(3)'I/J A-y (-21-Y(-Y(/3-1-y(~ )'l/JA-y
In order to develop the strains arising from the antisymmetrical bending

modes we need the following derivative expressions:
8(a 8(a 8(a

8"'a 8",p 8",-y
8e
8."
= I8(p
8"'a
8(p
8",p
8(p
8",-y = [~l 1 -1]
-1
1
.
. (6.121)
8(-y 8(-y 8(-y

8"'a 8",p 8",-y
The horizontal displacements and strains in the three natural directions,

are
00 8WAa 0(3 8WAa a'Y _ 8WAa

u =-z-- u =-z-- u - -z 8Y; , {6.122}
8Ya 8Y(3 'Y
(30 8WA(3 (3(3 8WA(3 (3'Y _ 8WA(3
u =-z-- u =-z-- u - -z 8Y; , {6.123}
8Ya 8Y(3 'Y
8WA'Y 8WA'Y
u 'Y(3 = - z 'Y'Y _ 8WA'Y
u 'YO = - z -- -- u - -z 8Y; . {6.124}
8Ya 8Y(3 'Y
With respect to the non-dimensional parameters 11
18wAa 18wAa
uaa=-zla 8"'0 u a(3 = -z 1(3 8"'(3 ua'Y = -z ~
1'Y 8WAa
8.,,'Y l6.125)
1 8WA(3 1 8WA(3 ~ 8WA(3 l6.126}
u(3a = -z 10 8"'0 u(3(3 = -z 1(3 8"'(3 u(3'Y = - Z 1'Y 8.,,'Y
1 8WA'Y 1 8WA'Y ~ 8WA'Y .{6.127}
u'Ya = -z 10 8"'0 u'Y(3=-zl(3 8"'(3 un = -z 1'Y 8.,,'Y
One more differentiation yields the expressions for the total strains, namely
b _ 1 8u aa 1 8u(3a 1 8u'Ya
Ita - -zf2a;;- - zf2a;;- - zf2a;;-
a '/0 a '/0 a '/0
b 1 8u a(3 1 8u(3(3 1 8u'Y(3
It(3 = -z 1~ 8"'(3 - z 1~ 8"'(3 - z 1~ 8"'(3
{6,128}
b 1 8u a'Y 1 8u(3'Y 1 8u n
It'Y = -z [2 a;;-
'Y '/'Y
- z 12'Y a;;-
'/'Y
- z 12'Y a;;-
'/'Y
We define now the above terms, Using the chain rule of differentiation we
write
8w Aa 8w Aa 8(0 8w Aa 8((3 8w Aa 8('Y

- -- - --- + ---- + - ---
8"'0 - 8(0 8"'0 8((3 8"'0 8('Y 8"'0'
8w Aa 8w Aa 8(0 8w Aa 8((3 8w Aa 8('Y
- - =- - - - + - - - - + ----,
8"'(3 8(0 8"'(3 8((3 8"'(3 8('Y 8"'(3 {6,129}
8w Aa 8w Aa 8(0 8w Aa 8((3 8w Aa 8('Y
- -- - --- + ---- + ----
8.,,'Y - 8(0 8.,,'Y 8((3 8.,,'Y 8('Y 8.,,'Y'
8w A{3 8w A{3 8(a 8w A{3 8({3 8wA{3 8('Y

-- = ---- + ---- + ----
8"'a 8(a 8"'a 8({3 8"'a 8('Y 8"'a'
8WA{3 = 8WA{38(a + 8WA{38({3 + 8WA{3 8('Y (6 130)
8"'{3 8(a 8"'{3 8({3 8"'{3 8('Y 8"'{3' .
8WA{3 8WA{38(a 8WA{38({3 8WA{38('Y
-- - ---- + ---- + ----
8",'Y - 8(a 8",'Y 8({3 8",'Y 8('Y 8",'Y'
8wA'Y 8w A'Y 8(a 8w A'Y 8({3 8wA'Y 8(.y

- -- - --- + - --- + - ---
8"'a - 8(a 8"'a 8({3 8"'a 8('Y 8"'a'
8wkr 8w A'Y 8(a 8w A'Y 8({3 8wA'Y 8('Y ()
-- = ---- + ---- + ---- 6131
8"'{3 8(a 8"'{3 8({3 8"'{3 8('Y 8"'{3' .
8WA'Y 8WA'Y 8(a 8WA'Y 8({3 8WA'Y 8('Y
- - = ---- + ---- + ----.
8",'Y 8(a 8",'Y 8({3 8",'Y 8('Y 8",'Y
With the aid of equations (6.120), (6.121), we perform the tedious algebraic
evaluations of the partial derivatives above and write for the horizontal
displacements
u aa = - z1 -8w-Aa 1 [1 1 2
- = -z-l -2la({3 + "2la('Y -
2 1 r r ] .J,b
2 a'>{3'>'Y 'f'Aa'
la 8"'a a
u a{3 = - z1- 8w-Aa- = -z- [1
1 --la({32 + la({3('Y ] .J,b
'f'Aa'
l{3 8"'{3 l{3 2
(6.132)
1 8w Aa 1 [1 2
ua'Y = -z--8-- = -zZ -"2la ('Y + la({3('Y 'f'Aa'
] .J,b
l'Y "''Y 'Y
u{3a = - z1-8w-A{3 1 [1
- = -z- -"2l{3(a2 + 1{3'>'Y,>a
r r ] .J,b
'f'A{3'
la 8"'a la
u{3{3 = - z1 -8w-A{3
- = -z-1 [1-l{3(a2 1 2 - 2l{3,>a,>'Y
+ "2l{3('Y r r ] .J,b
'f'A{3'
l{3 8"'a l{3 2
(6.133)
u{3'Y = - z1- 8w-A{3
- = -z-1 [1
-"2l{3('Y2 + 1{3'>'Y,>a
r r ] .J,b
'f'A{3'
l'Y 8",'Y l'Y
U'Ya = - z1-8WA'Y
- - = -z- 1 [1
--l'Y({32 + l'Y'>'Y,>a
r r ] .J,b
'f'A'Y'
la 8"'a la 2
u'Y{3 = - z1-8w -A'Y
- = -z-1 [--l'Y({3
1 2 + l'Y,>a,>{3
r r ] .J,b
'f'A'Y'
l{3 8"'{3 l{3 2
(6.134)
U'Y'Y = 1 8w A 'Y
_Z _ _ _ = -z-1 [1-l'Y({32 + "2l'Y(a
1 2 - 21'Y'>{3,>a
r r ] .J,b
'f'A'Y.
l'Y 8",'Y l'Y 2
One more differentiation with respect to the coordinate 1] provides us with

the total strain expressions
b _ 1 ouo.o. 1 ou{Jo. 1 ou'Yo.

Ito. - - z 1~ 01J0. - z 1~ 01J0. - z 1~ 01J0.
= _ z~ [OUo.o. 0(0. + ouo.o. o({J + ou M O('Y]
1~ 0(0. 01J0. o({J 01J0. o('Y 01J0.
1 [OU{Jo. 0(0. ou{Jo. o({J ou{Jo. O('Y]
- z 1~ 0(0. 01J0. + o({J 01J0. + o('Y 01J0. (6.135)
_ z~ [OU'Yo. 0(0. + ou'Yo. o({J + ou'Yo. O('Y]
1~ 0(0. 0"10. o({J 0"10. o('Y 0"10.
1 [ 310. (('Y - ({J)"p Ao.
= - z 1~ b + (0.1{J"p A{J
b - b ]'
(0.1'Y"p A'Y
b 1 ouo.{J 1 ou{J{J 1 ou'Y{J

It{J =- z 1~ O"1{J - z 1~ O"1{J - z 1~ O"1{J
= _ z~ [OUo.{J 0(0. + ouo.{J o({J + ouo.{J O('Y]

1~ 0(0. O"1{J o({J O"1{J o('Y O"1{J
1 [OU{J{J 0(0. ou{J{J o({J ou{J{J O('Y]
- z 1~ 0(0. O"1{J + o({J O"1{J + o('Y O"1{J (6.136)
_ z~ [OU'Y{J 0(0. + ou'Yo. o({J + ou'Y{J O('Y]
1~ 0(0. O"1{J o({J O"1{J o('Y O"1{J
= - z 1~1 [ bA{J + 31{J ((0. - ('Y)"p Ao.

-({J10."p b + ({J10."p A'Y
b ]'
1 au 0.'Y 1 ou{J'Y 1 au'Y'Y

---a-
b
It'Y =- 12 1J - z [2 a:;;- - z 12 a:;;-
Z
'Y 'Y 'Y '/'Y 'Y '/'Y
__ z~ [ouo.'Y 0(0. + ouo.'Y O({J + ouo.'Y O('Y]
- 1~ 0(0. O"1'Y O({J O"1'Y O('Y O"1'Y
1 [ou{J'Y 0(0. ou{J'Y o({J ou{J'Y O('Y]
-zl~ 0(0. O"1'Y + o({J O"1'Y + o('Y O"1'Y (6.137)
_ z~ [ou'Y'Y 0(0. + ou'Y'Y o({J + ou'Y'Y O('Y]
q 0(0. O"1'Y o({J O"1'Y o('Y O"1'Y
=- 1 [
z 1~ b + 31'Y(({J - (o.)"p A'Y
bAo. - ('Y1{J"p A{J
('Y10."p b ].
The matrix connecting the total strains and the antisymmetrical bending
modes is
Itb = CtNb'YA,
"I.b
3((7 -(fJ) (a -(a ]

CtNb = -zl-2 [ -(fJ 3((a - (7) (fJ l (6.138)
(7 -(y 3((fJ -(a)
where l denotes the diagonal matrix
~ ~1= fl-
l ..
1 1 I-I 1-IJ (6.139)
1= [ : a fJ 7'
and,
3((7 - (fJ) (a -(a ]

Ah = [ -(fJ 3((a - (7) (fJ . (6.140)
(7 -(y 3((fJ -(a)
Recalling once again the principle of virtual work we have
8U = ! 0'~8,~dV ! b~tKct8b~]dV
v
=
v
[.p~]t [{ f
h
=
__ n
1[.p~],
[z'IAhr'""I-' Ahl] dndz (6.141)
2
v
natural antisymmetric bending stiffness
from which we deduce the stiffness terms due to the antisymmetrical bend-
ing mode, viz.
h
2"
K~ = / / [z2IAhl-2Kctr2Ahl] dndz
n _l!
2
K11 KI2 (6.142)

= [K12
A
K13
K22 K23 ]
A A
K AI 3 K23 K33
A A
All terms included in K~ are integrated explicitly using the symbolic sys-
tem MACSYMA [57] with the aid of the integration formula
1/ 2!p!q!r!
0, (~($(~dn= (2+p+q+r)!" (6.143)
n
Explicitly, the antisymmetrical bending terms are
h h h
2 2" 2 2" 2 2"
K A11 = OlQ
6[4
/ z 2 1'i.'Y'Y dz - 6[2[2
OlQ / z 2 1'i.{3'Y dz + OlQ
6[4
/ z 2 1'i.{3{3dz
'Y h {3'Y h {3 h
-2" -2" -2"
h h h
0, /2" 0, /2" 30, /2" (6.~44)
+~2 ~~~+~2 ~~~+~ ~~~
'Yh {3h h
-2" -2" -2"
h h h
2 2 2 2
22
~=~ 0,[{3 / 2 0 / z~~+~
z~~+m2 2 30 / z~~
2
-Yh -Yh {3h

-2 -2 -2
h h h
0,[2 /2 0 2 0,[2 /2 (6.145)
- 61 2 :2 z2/'i,a-ydz + 2[2 / z2/'i,a{3dz + 6l: Z2/'i,aa dz
a-y h a h a h
-2 -2 -2
h h h
2 2 2 2
33 30 / Z~~+~2
~=+~2 2 0 / z~~+~
2 O,[-y / 2
z~~
-Yh {3h {3h
-2 -2 -2
h h h
o /2 0,[2 /2 0,[2 /2 (6.146)
+ 2[2 z2/'i,a-ydz - 6l 2 ;2 z2/'i,a{3dz + 6l J Z2/'i,aa dz
a h a{3 h a h
-2 -2 -2
!
h h
2 2
O,[a l {3!
K A12 =- ~
2
z /'i,-y-ydz -
O,[a
6l{3[2
2
z /'i,{3-ydz
-y h -y h
-2 -2
h h
2 2 (6.147)
0,[{3 /
- 6l l2
2
Z /'i,a-ydz -
50 /
6fT 2
z /'i,a{3dz
a-y a{3
h h
-2 -2
J J
h h
2 2
13
K A = - 6l 2 l
nIa 2
z K,/1,,"(dz -
nIcJy
"6l4 2
z K,/1/1dz
/1""( h /1 h
-2 -2
J J
h h
2 2 (6.148)
50 2 nI""( 2
- 6la l""( Z K,a""(dz - 6lal~ Z K,a/1dz
h h
-2 -2
J J
h h
2 2
K A23 =- 501
-6l z 2 K,a""(dz - nI/1
[2"l z 2 K,a""(dz
/1""( 6 a , "(
h h
-2 -2
J J
h h
2 2 (6.149)
- 6lnI""(
2l
2 nI/1 l""(
z K,a/1dz - """"6z4 2
z K,aa dZ
a/1 a
h h
-2 -2
These integrands are evaluated using piecewise integration as follows:

h
2" N
/ Z
2
/'i,OI.OI.dz = 'l~k{3
3 .l....J /'i,0I.0I. Zk 3)
- Zk+l ,
h k=l
-2"
h
2" N
/ Z
2
/'i,{3{3dz = 'l~k{3
3 .l....J /'i,{3{3 Zk 3)
- Zk+1 ,
h k=l
-2"
h
2" N
/ Z
2
/'i,n dz = 'l~k{3
3 .l....J /'i,'Y'Y Zk 3)
- Zk+1 ,
h k=l
-2"
h
(6.150)
2" N
/ Z
2
/'i,OI.{3dz = 'l~k{3
3 .l....J /'i,0I.{3 Zk - 3)
Zk+l ,
h k=l
-2"
h
2" N
/ Z
2
/'i,OI.-ydz = 'l~k{3
3 .l....J /'i,0I.-y Zk 3)
- Zk+1 '
h k=l
-2"
h
2" N
/ Z
2
/'i,{3-ydz = 'l~k{3
3 .l....J /'i,{3-y Zk 3)
- Zk+1 ,
h k=l
-2"
where N denotes the number of layers.
6.8.2 Antisymmetrical shearing terms

The creation of the antisymmetrical shearing stiffness requires the integra-
tion of (6.64) and the adoption of averaging procedures. These include
Kll K12]
Kl = i[W~{3G'sW OI.{3]dV = [K[2 K?2 ,
K2 = i[W~-yG'sW{3-yldV = [Kll K12]

KJ2 K~2 , (6.151)
K3
Kll
= i[W~OI.G'sW'YOI.ldV = [K12
K12]
K~2 .
In line with previous arguments, an average of a combination of K 1, K 2,

K 3 is created as follows:
1
Kll1
+ K22
3
K121 K12
3 ]
Ks-
- - [' K12
1 KF +Kr2 K12
3 K12
3
K122 KJ1 ~ Ki 2
Kll K12
s s K;3] (6.152)
= K12
[
s
K22
s
K23
s .
K13
s
K23
2
K33
s
This matrix is pivotal in our theory. It will enable us to proceed with

the computation of the transverse shear stiffness. Note that all elements
of the matrices have been explicitly derived using MACSYMA [57]. The
antisymmetrical shearing terms read
1~nl~ [! Gxxdz(y~+y~) ! Gyydz(x~+x~)

~ ~
K;1 = +
h h
-2 -2
,
-21
h
(~.153)
Gxydz(x,y, + Xpyp) 1
-2
1~n l~ [! Gxxdz(y~ ! Gyydz(x~

~ ~
K;2 = + y;) + + x~)
h h
-2 -2
,
-21
h
(~.154)
Gxydz(xaYa + X,y,) 1
-2
1~nl; [JGxxdz(y~ +y~) JGyydz(x~ +x~)

~ ~
K;3 = +
h h
-2 -2
-2{ (~.155)
h
G""dz(xm +X.Y.) 1
-2
-1~nlaZ.B [ J J
~ ~
K;2 = (YaYp) Gxxdz + (XaXp) Gyydz
h h
-2" -2
h
6.156}
- (x.YP + xpV.) { Gxvdz 1
-2
-1~nlal" [
~ ~
K;3 = (YaY,,) / Gxxdz + (XaX,,) / Gyydz
h h
-2 -2
h I'
1(6.157)
- (x.Y, + x,V.) { Gx.dz 1
-2"
= -1~nl{3l'Y [ ! !
~ ~
K;3 (Y{3Y'Y) Gxxdz + (x{3x'Y) Gyydz
h h
-"2 -"2
h
(6.158)
- (XPY7 + X7YP) { G",dz 1
-"2
where
!
h
"2 N
Gxxdz = L G~x(Zk - Zk+1), etc. (6.159)
h k=l
-"2
For a right triangle with la = l{3 = 1 and isotropic material of shear modulus
G, thickness h, and area A, the shear stiffness K s simplifies to
K _
s -
Gh
12A
[3 . 3
v'v'22] . (6.160)
V2V24
The expression for the transverse shear strain energy is
1 t
Us = 2'sKs's (6.161)
A quick examination of Fig. 6.13 reveals
1 s
Is = 2'I/JA- (6.162)
Us = ~,;Ks,s = ~['I/J~]t~Ks['I/J~], (6.163)

6.9 Shear correction factors 189
which provides us with a new expression for the transverse shearing matrix,
namely
.ii"n [(12
= [K12 [(13]
- 1
Ks = -Ks
4
[(22 .ii"23 . (6.164)
[(13 [(23 K33
6.9 Shear correction factors

The methodology to obtain the shear correction factors is similar to that
used for the beam element. The same expression, derived previously for
a single beam direction, is applied here to the three triangular sides. We
recall the expression for the shear correction factor Sc!, namely
R2
Us = !
h/2 2 )
(6.165)
!
Sci = U h/2 9 (z dz
G(z)dz G(z)
-h/2 -h/2
In order to proceed with the computation of the three shear correction fac-
tors, the coordinates of the neutral planes along the three natural directions
o!, /3, "I must be defined. They are
! zK~adz ! zK~(3dz ! zK~'"(dz

h/2 h/2 h/2
-h/2 -h/2 -h/2

ZOa = , zO(3 = , zo'"( =
! K~adz ! K~(3dz ! K~'"(dz

h/2 h/2 h/2
(6.166)
-h/2 -h/2 -h/2
Definition 8 Three shear correction factors, automatically computed on

the natural coordinates, adjust the transverse shear stiffnesses so that shear
stress-free top and bottom surfaces exist.
The shear correction factors for edge a is
h/2
[! K~a(Z _ zoa)2dZ] 2
Aa = ___-_h..:.../_2- ; : J . ; - - - - - - -
! K~a(Z
h/2
(6.167)
h/2 [- - Zoa)2dZf
Ga !
-h/2
-h/2
Gk
a
dz
where
Gxx GXY] [ca",,]

G~ = [ca"" Sa"" 1[Gxy G yy k Sa"" , (6.168)
and
! G~dz = L G~hk.
h/2 N
Ga = (6.169)
-h/2 k=l
The shear corrections factors for the other two natural directions are ob-
tained by replacing a with {3, 'Y. The new corrected transverse shear stiff-
ness is
6.10 Simulative azimuth stiffnesses 191
6.10 Simulative azimuth stiffnesses

Arbitrary stiffnesses pertaining to elastic springs assigned at the vertices
are used to simulate the in-plane rotations about the z-axis, viz.
k OlOl k~{3 -0.5

kz = [ k~{3 k~{3 kg,
Ol
k '] = kz [ -0.5
1 1
-0.5]
-0.5
k~' k~' ki' -0.5 -0.5 1 (6.171)
The parameter kz is taken as the maximum of the three edge bending

stiffnesses, namely
kz 1 / z 2Yi.OlOldz,
= n max [[2 1/2
l2 1 / z 2Yi."dz ] .
z Yi.{3{3, dz, l2
01 {3 , (6.172)
Note that in most numerical tests performed this somewhat arbitrary stiff-
ness posed no problem; in most cases fast convergence was obtained. When
much higher modeling accuracy is required, the useful concept of "vertex
rotation" (or drilling freedoms) can be adopted following Allman [58], [59],
[60].
6.11 Local and global cartesian stiffnesses

The elemental natural stiffness matrix is now established. The next step
is to transform it first to the local elemental cartesian coordinate and then
to the global cartesian coordinate before initiating the assembly procedure.
The matrix equilibrium equations in the natural coordinate system are
kNPN =PN. (6.173)
The following sequence of congruent transformations are now performed

{see also (6.102)):
KE = [Ti6[iilv[ ~ ]iiNl TOO] (6.174)

natural stiff. (12xI2)
, y #
localsiff. k
(18x18)
global stiff. K E
(18x18)
Our next task is to establish matrix aN.
Definition 9 Matrix aN is solely a function of the geometry of the ele-

ment. It establishes the relation between the natural straining modes PN
and the unit local cartesian freedoms p.
Using pure geometrical arguments and equation (6.60), we write
'Yto: = Xo:
p(u3 - U2)
Yo:
+ p(V3 - V2)
0: 0:
Xf3 Yf3
'Ytf3 = r(UI - U3) + r(VI - V3) (6.175)
f3 f3
X-y Yf3
'Yt-y = p(U2 - ud + r(V2 - vd
-y -y
Xo: Yo:
"pso: = -I (fh - ( 3) + p(<P2 - <P3)
0: 0:
Xf3 Yf3
"pSf3 = -1 (03 - ( 1) + r(<P3 - <PI) (6.176)
f3 {3
X-y (
"ps-y = -1 01 - O2) + r(<Pl
Yf3
- <P2)
-y -y
6.11 Local and global cartesian stiffnesses 193
,pAo = Xlo (02 + (3 ) + Yo (2 + 3) _ 2(W3 - W2)

o Po l0
,pA(3 = 7(3 (03 + OI) + Y: (3 + d _ 2(Wl - W3)
(3 l(3 l(3
(6.177)
,pA-y = Xl'Y (0 1 + (2 ) + Y(3 (1 + ,1.2) _ 2(W2 - wI)
'Y l2
'Y
0/ l 'Y
,po = ,pI - P06

,p(3 = ,p2 - P06 (6.178)
,p'Y = ,p3 - P06
Matrix aN is partitioned as follows:
-11 -12
aN aN aN
aN =
[
(6x6) (6x6)
_
13
(6x6)
-23
1 .
(6.179)
(12x18) a21 a22
N N aN
(6x6) (6x6) (6x6)
The sub matrices contained in (6.179) are
~
l{3
?1{3
_'!IT
-11 _
aN -
(6x6)
I_~
P
'1
rr'1
(6.180)
_'MP.. ~
l{3 1{3
-f-f
I", -~
I",
~
1"'( t
1"'(
-12
aN = lh..
I",
-~
I",
.1 , (6.181)
(6x6)
2 lh.. -~
I", I", I",
f-f
I", ~
I",
-~
1{3
-l-
1{3
-13 _ _lh..
aN - ~ . I , (6.182)
I", I",
(6x6)
2 lh.. -~
-I", I", I",
1I.i! -~
1{3 1{3
2 1I.i! -~
-lfj 1{3 1{3
1b:. -~
1"'( 1"'(
2 1b:. -~
-21 r:; 1"'( 1"'(
I~
aN (6.183)
(6x6) lh..
4!l 4!l 1
~ lh..
4!l 4!l
~ lh..
4!l 4!l
6.11 Local and global cartesian stifInesses 195
(6.184)
(6.185)
(6.186)
is computed, the following transformation refers it to the global coordinate

system:
t -
KE = T06KTo6. (6.187)
Matrix T06 comprises the block diagonal matrix
T06 = rTo To To To To ToJ, (6.188)

(18xI8)
with
cx, X c x' y c x ' z]

To = [ cY'x Cy'y Cy'z . (6.189)
(3x3) CZ'x CZ'y CZ'z
In equation (6.189), Cx'x etc., denote the cosine ofthe angle formed between
the local cartesian axis x' and global cartesian axis x. The same convention
is adopted for all other entries.
6.12 Kinematically equivalent nodal loads

If the structure is subjected to distributed surface load, for example uniform
pressure, sinusoidal loading etc., kinematically equivalent nodal forces must
be computed. According to the principle of virtual work
oU=oV, (6.190)
where OV is the variation of work performed by external forces given by
oV = I
S
pto'U dV + F~ori + M~o8i' (6.191)
in which P are the applied distributed surface loads and F, M are applied
concentrated forces and moments, respectively. Then, at a point within the
triangle the displacement 'U is interpolated from the nodal values via
U = (iUi, i = 1,2,3. (6.192)
Using matrix notation we have
'U = W p, o'U = wop. (6.193)

(6xl) (6x18)(18xl)
The modal matrix w is given by the block diagonal matrix
(6x18)
w = [(016 (/316 (,y I 6]' (6.194)
6.12 Kinematically equivalent nodal loads 197
in which 16 denotes the diagonal unit matrix
16=fl 1 1 1 1 IJ. (6.195)

In a similar manner, a distributed load is interpolated via
p = W PSI. (6.196)
(6xl) (6x18)(18xl)
Substituting equations (6.193) and (6.196) in equation (6.191) we obtain
8V = P~I [I wtw dS] 8p + F} 8r i + M}MJi' (6.197)

S
from which we deduce the vector of kinematically equivalent nodal loads
(in the local elemental coordinate) through
pI
(lx18)
= ptSI IwtwdS + F~z + M~.z (6.198)
S
The integration is carried out explicitly to yield
pI = P~I 12
(lx18)
n [216 16
16 216 16 +F~+M~
16 16 216
h] (6.199)
We stress that pI includes the kinematically equivalent nodal loads with

respect to the local coordinate system. Prior to assembly this vector is
transformed to the global coordinate via
R = Tb6 Pl . (6.200)
Following assembly of both elastic stiffnesses and applied loads, the matrix
global equilibrium is
Kr=R. (6.201)
The solution of this system of linear equations provides us with the global
nodal unknowns.
6.13 Initial load due to temperature

We now formulate the cartesian thermal load vector. For this purpose we
adopt a linear through-the-thickness spatial temperature variation, namely
T(x, y, z) = To(x, y) + ZTI (x, y). (6.202)
The parameters To, TI in (6.202) are related to the temperatures at the top
and bottom surfaces of a laminate T t , n, respectively, via
1 1
To = 2(Tt + n), TI = X(Tt - n), (6.203)
where h represents the plate or shell thickness. For non-uniform tempera-

ture distribution, T t , n are computed in an average sense from the nodal
points n using
1 3 1 3
Tt=a LTt , n=a LT:. (6.204)
n=I n=I
Alternatively, if greater accuracy is required, the top and bottom temper-

atures can be interpolated from the nodal values via
Tt = Tl(i, n = Tt(i, i = 1,2,3. (6.205)
The thermal load vector will be computed first in the natural coordinate
and then transformed to the local and global coordinate systems. The
natural thermal load vector J N is written as the sum of the two vectors
JN = J NO + J NI, (6.206)
(12xI)
where J NO, J NI, are the thermal loads due to uniform and linear through
the thickness temperature distributions, respectively. To this end we define
for every layer the thermoelastic coefficients in the three coordinate sys-
tems, namely the material, the local cartesian and the natural coordinate
as
Ctt
123 _
-
[atll
.
.] ,
at22 Ct~ = [
aXX
a yy
V2a xy
]
, Ctt = l
ata]
at{3 .
at'Y (6.207)
In (6.207), atll' at22' are the coefficients of thermal expansion along and
perpendicular to the fiber direction, respectively. These coefficients are
transformed (for every layer k) to the local coordinates using matrix As
{see (6.52)), viz.
[ axx V2a xy ] = [A~Ct:23 As] k. (6.208)

V2a xy a yy
The entries of Ctt are obtained from Ct~ using matrix B defined in (6.19) as
follows:
[ca""
l ata]
at{3
at'Y
C{3""
C'Y""
Say'
S(3y'
S'Yy'
V2 S a",'Ca y ,] [ a xx ]
V2S{3""C{3y'
V2s'Y",' C-yy'
a yy .
V2a xy
(6.209)
Using matrix notation we have
Ctt = B t Ct t, (6.21O)
The total natural strain is the sum of the elastic and thermal strains
It = et + 'fJt, et ~ elastic strain 'fJt ~ thermal strain.

(6.211)
The natural thermoelastic constitutive relation now reads
Uct = Kcd/t - 'fJt] = Kctet (6.212)
Substituting equations (6.211) and (6.212) into the principle of virtual work
we obtain
8U = ! (T~81'tdV = ! b~ - '1~]
V V
K.ct 81't dV
~ P'" [! a'""",aNdV ]"PN - [ ! ":",,aNdV ] 8PN

(6.213)
v v ,
'----.,'" ' ...
natural elastic stiffness natural thermal vector
= 8Ue + BUt
From this equation we deduce the natural thermal load, viz.
J}y = ! '1~K.ctoNdV.
v
(6.214)
Using the basic definition of thermal strain, i.e.
'1t = ot!:J.T = ot{To + zTd, (6.215)
we find that the initial load due to temperature becomes
J tN -Jt Jt
NO + Nl
! [ToO~K.ct]ONdV + ! [ZTIO~K.ct]ONdV
-
=
(6.216)
v v
We now define all integrals included in the natural thermal load vector. We
begin with
[(~)] = n !
h
2"
[Too}YK.ctot]dz
h
-2"
/'i,o{3 (6.217)
N [/'i,OO
= no}y LTo{Zk - Zk+1) /'i,o{3 /'i,{3{3 /'i,07l
/'i,{37 [atol
at{3 ,
k=l /'i,07 /'i,{37 /'i,77 k at7 k
[~]
(~:) = n ~ [zToal.""at]dz
1 N [Kaa
= na}.r"2 LTo(z~ - z~+l)l-l "'af3
"'af3
"'f3f3
K...,]
"'f3'Y
[aatf3tat218)
.
k=l "'a'Y "'f3'Y "''Y'Y k at'Y k
Vector J NO is
JNO [ J1] .
(3xl)
= Jg (6.219)
(12xl) (::)
Similarly,
[J:::
11
(3xl) ]
=n 2"
, t
[zT1aN""at]dz
N
= na}.r"21 LTo(z~ r~
- z~+l) "'af3
"'af3 K...,]
"'f3'Y
[aatf3ta ] (6.220)
"'f3f3
k=l "'a'Y "'f3'Y "''Y'Y k a t'Y k
[03 ] ~
r
(~) = n [z'Tlal.",,,a,Jdz
1 N
= na}.r3 LTl(ZZ -
["aa
ZZ+l)Z-l "'af3
"'af3
"'f3f3
7
"'f3'Y ]
"a [aatf3
ta
,
221
)
, k=l "'a'Y "'f3'Y "''Y'Y k at'Y k
and
Jl]
(3xl)
JNl
(12xl)
= [(::)
Ji . (6.222)
Now the natural thermal load vector is completely defined from its vec-
tor components, and we may write down the equilibrium equation in the
natural coordinates via
kNPN=PN+JN. (6.223)
The natural initial load due to temperature is transformed to the global

coordinate system via
J
(18xl)
= [aN T06
(12x18)(18x18)
t JN
(12xl)
= akr JN
(18x12)(12xl) (6.224)
Global equilibrium now reads
KEr=R+J (6.225)
6.14 Computation of stresses and stress resultants

6.14.1 Computation of stress resultants
After solution of the global equilibrium, and estimation of the global un-
knowns r, the elemental natural straining modes and generalized natural
forces are obtained by
PN = aNP, PN = kNPN' (6.226)
Alternatively, the natural stress resultants which are attributed to an ele-

ment's center are defined as
! !
h h
"2 "2
PN = ucdz, MN = zucdz. (6.227)

h h
-"2 -"2
PN, MN were defined in (6.70). Substituting the expression for total

strain (6.95) in (6.227), the relations for the natural axial forces, bending
moments and transverse shear forces become
h h
2 "2
PN = [J ~ctdzl'Y~+ [J z~ctdz]l-l1/Js
h h
-"2 -"2
h h
2 2
MN = [J z~ctdz]"Y~ + [J z2~ctdz]I-11/Js (6.228)
h h
-2 -2
QN = 2MA MA = [MAa MA{3 MA'Y]

lz '
If the transverse shear forces per unit length are required, the last of equa-
tions (6.228) becomes
2MA
qN=n- (6.229)
The stress resultants are transformed to the local cartesian coordinate x'y' z'
using
C~~] = [~ B] [~:] (6.230)

c
Qc = [cax1 c{3x l C-rXI] QN
ayl C{3yl C-ryl
Remark 5 The local cartesian forces and moments can be easily referred
to the three triangular nodes. This requires only the separate projection
of the the various natural nodal axial forces as well as symmetrical and
antisymmetrical nodal moments to the three local coordinate systems at the
vertices, and superposition of all components in the local directions.
6.14.2 Computation of through-the-thickness stresses

The natural component stresses for each layer are
k k k
U e = Ketlt (6.231)
They are transformed to local cartesian stresses using
Uk=
, B U ek , (6.232)
where u' = {u xx U yy J2u xy }.

In order to compute the through-the-thickness transverse shear stresses
we postulate equilibrium in a natural direction, say 0, viz.
{)Uea + {)Usa _ 0
(6.233)
{)X a {)z - ,
where U ea is the direct stress corresponding to the antisymmetrical bending

mode M~a' and Usa is the transverse shear stress from the antisymmetrical
shear mode M Aa . Then, equation (6.233) yields
--J
z
{)U sa {)U ea
--=---,
{)z {)x a Usa - {)u ea
{)X a dz + Cl (6.234)
h
-"2
If we assume a fictitious beam along natural direction 0, elementary beam

theory gives
{)U ea {)MA Z _ {)MA kNB(Z) (z _ ZOa),

---- h
{)xa {)xa Ia {)x a
J
"2
(6.235)
kNB(z - zOa)2dz
h
-2"
where ZOa is the Z coordinate of the neutral axis (see also (6.166)) and
kNB is the bending stiffness in this direction. A combination of these two
relations yields
a so. =- Jz aMA
axo. %
kNB{Z) (z - zoa,}dz + CI.
-% J h
k%B{z - zOo)2dz
(6.236)
-"2
We have already provided the definition for the shear force per unit length,
namely
aMA 2MAo.
qo = axo = -0-' (6.237)
in which case (6.236) becomes
I
Z
kNB{Z)
a so =- qo % (z-zOo.)dZ+CI
-% Ih
k%B(Z - zOo)2dz
{6.238}
-"2
Since the transverse shear stiffnesses vanish at the top and bottom sur-
faces, constant CI also vanishes. Now appropriate stiffnesses kNB must be
assigned. For this purpose we use the natural stiffness K.ct and invoke the
physical lumping method [16], [51]. We assign the following values to the
k~B
Ko.o. + Ko.{3 + Ko.'Y]

[ Ko.o. - Ko.{3 - Ko.'Y
ko. - max ,
NB - Ko.o. - Ko.{3 + Ko.'Y
Ko.o. + Ko.{3 - Ko.'Y
Ko.{3+ K{3{3 + K{3'Y]

a
-max [
-Ko.{3 + K{3{3 - K{3'Y , (6.239)
-Ko.{3 + K{3{3 + K{3'Y
I-'
kNB -
Ko.{3 + K{3{3 - K{3'Y
Ko.'Y K{3'Y + + K'Y'Y]

k 'Y _ -Ko.'Y - K{3'Y + K'Y'Y
NB- max [ .
- Ko.'Y + K{3'Y + K'Y'Y
Ko.'Y - K{3'Y + K'Y'Y
Equation (6.238) enable us to compute the through-the-thickness transverse
shear stresses in the three natural coordinates. Once this is accomplished,
the stresses are transferred to the local coordinates (always referring to
an element center) using equation (6.73). For the analysis of composite
structures the normal stress a zz is also of interest since it may contribute
to damage initiation such as delamination. Therefore an estimation of this
stress component is desirable. Equilibrium in a natural direction a reads
oaso. + oazz = 0 (6.240)

oxo. oz '
or
z
oazz
--=---,
oz
oa so.
oxo. azz -
--j oaso. dz
oxo.
+ CIZ + C2,
h (6.241)
-2"
and taking into account (6.238)
=
jz jZ oxo.
oqo.
~
kNB{Z) {z _ zOo.)dz + CIZ + C2.
(6.242)
j
azz
_l!
2
_l!2 NB {Z - ZOo. )2dz
ko.
h
-2"
If an applied traction of magnitude p is present at the top plate surface we

have
8qo.
- ----,
p- 8xo. (6.243)
and equation (6.242) becomes
JJ
z z
kNB{Z)
!
a zz =- P ~2 ) dz + Cl Z + C
( -zZOo.
h
-2" -2"
h
k):,B(Z _ z(}a)'dz ,. (6.244)
-~
2
Equation (6.244) indicates that the stress a zz is a cubic function of z. Note

that the two constants of integration may be determined using the two
available boundary conditions, i.e.,
h h
at Z = --2 -+ a zz = 0, at Z = + -2 -+ a zz = -po
(6.245)
With these boundary conditions, the final expression for a zz becomes
JJ
z z
a zz = - P kNB ()
Z (z - ZOo. )dZ - !!'z-'!!.
h 2
J
h
h h 2" 2 (~.246)
-2" -2"
kNB(z - ZOo.) dz
-~
It is also possible to construct an average stress a- zz attributed to an element

center, viz.
a- zz -_1(0.
3" a zz + a f3zz + a 'Y)
zz . (6.247)
6.15 The simplified geometrical stiffness

6.15.1 Geometrical forces
Geometrical stiffness is the basis for any attempt to study the behavior
of shells under conditions in which large deflections may occur with small
strains. Most cartesian theories and their finite element implementations
usually derive the geometrical stiffness based on a second variation of the
strain energy, using the so-called Von Karman strains [61], [62]. As for the
beam element, and in the context of the natural-mode method, we follow
a different path for the construction of the geometrical stiffness for the
triangular shell element. Our physical assumption is that since the strain
remains small and the element does not change shape, the rigid body rota-
tions become noticeable. The natural forces will operate on these rotations
and produce rigid body moments. In this respect, stress systems which pro-
duce rigid body moments due to rigid body rotations are generally most
important for the formation of the geometrical stiffness. Since the geomet-
rical stiffness will be derived here based solely on rigid-body rotations [63],
[64] we need to define matrix 0.0.
Definition 10 Matrix 0.0 relates the natural rigid body modes Po to unit
cartesian nodal displacements and rotations.
In order to estimate the translational modes Po as functions of the dis-

placements and rotations at vertices 1, 2 and 3, we shall assume a linear
displacement field with respect to the local elemental coordinates, namely
u = Po + PIX + P2Y, V = qo + qIX + q2Y, W = roo

(6.248)
If the origin of the local coordinate is placed at the element's barycenter,

equation (6.248) may be written as
UI + U2 + U3 = 3po, VI + V2 + V3 = 3qo, WI + W2 + W3 = 3r o
(6.249)
The translational modes Po are expressed as the average of the displace-

ments at the vertices, that is
1
POI = Po = "3(Ul + U2 + U3)
1
P02 = qo = "3 (VI + V2 + V3) (6.250)
1
P03 = TO = "3 (WI + W2 + W3)
The rigid body rotation P06 will be estimated using
1 (oV aU) = 1 (6.251)

P06=-
2 ax - -oY
- -(ql - P2).
2
If we write (6.248) for every vertex, the result is
1
P06 = - 40 (XaUl + YaVl + Xf3 U 2 + Yf3 V 2 + X-y U 3 + Y-y V 3)
6.252)
Turn now to Fig. 6.14 which shows a rotation (h. The idea here is to
project this rotation on the x, Y axes. The angle (h shown in Fig. 6.14 is
o WI wlla
(6.253)
1 = ha = 20 .
It projects on the local cartesian coordinates via
= 01 cos ax = 01 z:: = 20 WI,

Xa Xa
Olx
(6.254)
ax = 1 z:: = 20 WI
0 Ya Ya
O
ly = O1 sm
Similarly,
Xf3 X-y
02x = 20 W2, O3x -- -W3,
20 (6.255)
Yf3 Y-y
O2y -- -W2,
20 03y = 20 W3 (6.256)
*-
X
/ a: ' IC>
2 p(J4
2
(xt Y2)
3
(x)' Y3 ) ~3
Figure 6.14: Rotation (h
Hence the rigid-body rotations P04, P05 are
Xo xf3 X"(
P04 = ()lx + ()2x + ()3x = 20 WI + 20 W2 + 20 W3
Yo Yf3 Y"(
P05 = () ly + ()2y + ()3y = 20 WI + 20 W2 + 20 W3 (6.257)
Matrix i:io partitions as follows:
ao
[-11
(3x6)
a-12
o
(3x6) -13]
ao
(3x6) ,
i:io (6.258)
(6xI8) -21 -22 -23
ao ao ao
(3x6) (3x6) (3x6)
with
-11
ao
(3x6)
-12
ao
(3x6)
-13
ao
(3x6) [~ 3"
1
1
3" J (6.259)
~
211
-21
ao = 1k
211 . I , (6.260)
(3x6)
-~ _1k
411 411
~
211
-22
ao = ~
211 . I , (6.261)
(3x6)
-~
411
-~
411
!1..
211
-23
ao = h
211 (6.262)
(3x6)
_!1.. _h
411 411
The complete transformation matrix a has now been defined via
_ ao ]
(6x18)
(18~18) = [ aN .
(6.263)
(12x18)
Its inverse matrix A is
A = a-I , A = [Ao AN ], (6.264)

(18x18) (18x6) (18xI2)
in which Ao is defined as
-t
Ao =
[-1
Ao -2 A-3]o ,
Ao (6.265)
(6xI8) (6x6) (6x6) (6x6)
and
- 1
AO = Yl 1
., , {6.266}
(6x6)
-Xl 1
-Yl Xl 1
., ,
1
AO
- 2
(6x6)
= I 1
Y2 1
{6.267}
-X2 1
-Y2 X2 1
1
- 1
AO = Y3 1
{6.268}
(6x6)
-X3 1
-Y3 X3 1
Incidentally,
a..4. = 1 18, {6.269}
where 1 18 is a unit matrix of dimensions 18 x 18.

In the course of a rigid body motion the triangular element remains
unchanged with respect to the natural coordinate system. Consequently,
within the following system of reference {a convective system} we can use
matrices ao, aN and ..4. 0 .
Definition 11 Nearly all geometrical stiffness is generated by the rigid

body movements of the element [63j, [64j. Therefore the geometrical stiff-
ness includes only those natural forces which produce rigid body moments
when the element receives rigid body rotations. The geometrical forces PG
due to PN arise primarily from
o
W.Z Wp=W -XPOS+YP04
'r :' =," +< PM - ' p.

P06 v. Y
P~
01
I
Up=U
o
'YP06 +ZPOS
WI
I
------...J
o
U I
/
/
/
/
/
/
o
/ 0
/ v
/
/
/
X. U
Figure 6.15: Translation and rotation of a point P in space
To construct the geometrical stiffness we will focus on small rigid body

rotation increments about axes x'y' z' combined in the vector (note dp =
p!:J.)
POR!:J. = aORP!:J. = {P04!:J. P05!:J. P05!:J.}. (6.270)
In concert with the principles of our theory, rigid body motion should not
change the shape of an element and therefore we can use the linear matrix
aN to estimate the natural deformations which arise from a small addi-
tional nodal displacement 8p. A coordinate transformation T{po) yields
the position of the element after rigid body motion. Initially for small
displacements (see Fig. 6.15)
[P~6
-P06
+ P05]
-~04
[X]
;
=UO+TR'X,
[:] = [::] -P05 (6.271)
P04
where
pos]
[P~6
-P06
TR = -~04 , (6.272)
-Pos P04
and
TCJt = [: : ~1]'
. 1 .
TCJf = [.. .. 1]. , TC: = [. -1 .]
-1 . .
1
.
.
.
.
.
.
(6.273)
We have defined in equation (5.212) the rotational matrix RR as
-pos]
[-~06
P06
RR = P04 , (6.274)
Pos -P04
as well as the submatrices
RCJt = [: ~ ~], RCJf = [~. . -1]: ' [. 1

RC: = ~1 :] (6.275)
. 1.
We have already seen that
aN = aN T 06 (6.276)
Now a small rigid body rotation increment leads to analogous expressions

for all, All, namely
all = {aOLl aNLl} = aR6,

(6.277)
All = {AOLl ANLl} = R6t A- ,
with
R6 = rRR RR RR RR RR RRJ. (6.278)

(18xI8)
We effect a rigid body rotation increment POR' The global forces rotating
with the element, originally P, become P + P A and a rigid body rota-
tion initially A~R becomes A~R + A~RA' Since the resultants of all forces
produced by rigid body motion must vanish, we write for the rigid body
moments
-t -t-t
POR = AOR P = [AOR + AORA][P + P A] = O.
(3xl) (3x18)(18xl)
(6.279)
Neglecting second order magnitudes we find
-t -t -t -t
AORP 06.+ AORAP = 0 =} AORP A = -AORAP = PORA'
(6.280)
Hence
- t - t -t
PORA = -AORAP = -AORA aN PN, (6.281)
(3x18) (3x18) (18x12)(12xl)
which is a measure of the effect of the natural forces due to the change of
geometry. Equations (6.281), (6.277) can be combined to yield
- t -t
PORA = -AORAaNPN
(3xl)
-- - [A- OR
t
6 aN p
R(4)-t 1
N P04/). - [A- OR
t
6 aN p
R(5)-t 1
N P05/). - [A- OR
t
6 aN p
R(6)-t 1
N P06/).
-- kGRPORA' (6.282)
(3x3) (3xl)
whereupon
R6 - fR Oi
(i) - R R Oi
R R Oi
R R Oi
R R Oi
R ROiJ - 4"
R, i - 5.
6
(18x18) (6.283)
It is clear that the forces PORA depend on the values of PN; this is an im-
portant point since the nonlinearity arises from this dependence. Equation
(6.282) estimates the nodal cartesian moments PORA due to rotation in-
crement PORA and yields an appropriate expression for the local rigid body
rotational geometrical stiffness kGR, namely
(3x3)
- [A-OR
k G R-- t R(4)
6
-t
aNN
P
(3",18)(18:>:18)(18",3)(3:>:1)
A- OR
t R(5)
6
-t
aNN
P
(3:>:18)(18:>:18)(18"'3)(3"'1)
A- OR
t R(6)
6
-t
aNN
P J.284)
(3"'18)(18"'18)(18"'3)(3"'~
We employ symbolic computation and carry out the matrix multiplications

in (6.284) explicitly. In so doing we get
kGR = (6.285)
(3x3)
. ]
~
~+~+
I", 1{3 I"{ -(~+~+~) I", I~ I"{
( ~+~+~)
I", 1{3 I"{
P"'2~
I",
+P{32"'1{3 $ +P,,{;;
I"{
.
. ~+~+~
The cartesian nodal forces due to three rigid body rotations PORL1 are
- -t
PGL1 = aOR P ORL1
(lSxl) (lSx3) (3xl)
= [ abR kGR aOR] PL1 kG PL1 (6.286)

(lSx3)(3x3)(3xlS) (lSxl) (lSxlS)(lSxl)
where, kG is the so-called simplified geometrical stiffness with respect to

axes x' y' z'. The term simplified refers to the fact that only the middle plane
axial natural forces Pa , P{3 and P-y are included in kG. They fully represent
the prestress state within the material. In this case the geometrical stiffness
is symmetric.
As mentioned before, nearly all geometrical stiffness arises from the
rigid body movements of the element. In buckling phenomena, quite often
the membrane forces are relatively large, and in this case it may be worth
considering an additional approximate natural geometrical stiffness arising
from the coupling between the axial forces and the symmetrical bending
mode (stiffening or softening effect). This natural geometrical stiffness is
similar to the one used for the beam element and comprises the following
diagonal matrix (here only the coupling between the axial force and the
symmetrical bending mode is considered)
6.16 Geometrically nonlinear analysis 217
1
kNG= 12r. . Pa . P{3 . P'"( .J.
(6.287)
Once the simplified geometrical stiffness is formed, it is transformed to

the global coordinate p. Respectively, the natural geometrical stiffness is
transformed first to the local and ultimately to the global coordinates. This
is accomplished via
(1~?s) = [Tb6 kGT06] + [[aNTo6fkNG[aNT06]]

(6.288)
where KG is the global elemental geometrical stiffness. For some other

applications of the natural approach to the estimation of the geometrical
stiffness for shell elements see the papers of Zhu [65], [66].
For initial buckling analysis, the membrane stresses are supposed pro-
portional to a load factor AL. The determination of the buckling loads and
the corresponding buckling modes is then reduced to the classical eigenvalue
problem
KE+ALKG=o. (6.289)
Matrices K E, KG denote the assembled global elastic and geometrical

stiffnesses of the structure, respectively.
6.16 Geometrically nonlinear analysis

In the absence of bifurcations the computational strategy for geometri-
cally nonlinear analysis is based on a simple incremental/iterative approach
which evolves around the pivotal relation of the accumulation of natural
forces. The algorithm evolves mainly around three loops; an outer loop
over the desired load and temperature increments; a first inner loop over
iteration numbers; and a second loop over finite elements. The latter pro-
cedure involves a few critical steps with the major task being the update
of the natural modes from the increments of the global displacements. The
anchor equation here is
1
PNA = "2 [aN{K b) + aN{Ke)]rA, (6.290)
where K b , K e represent the structural geometry at the beginning and at

the end of the effected deformation step, respectively. At the end of the
current step the accumulation of the natural forces is performed (note that
the natural forces follow the structure during deformation and are additive
therefore), viz.
PN = P'N + PNA = P'N + kNPNA + J NA, (6.291)

where P'N denotes the natural forces at the beginning of the step. Subse-
quently, the accumulated natural forces are converted to global nodal loads
RL, assembled and subtracted from the applied cartesian elastic REA and
thermal ReA global forces to provide the unbalanced forces
Ru = REA - ReA - RL (6.292)

At the end of the current step the elastic, geometrical and tangent ele-
mental stiffness matrices are formed. The unbalanced forces Ru are then
distributed in one or more cycles (using the global tangent stiffness K T )
until a predefined convergence criterion is met. This solution phase can
be considered as the partial application of the Newton-Raphson technique.
Following convergence, the global displacements are updated and the next
loading step is effected.

6.17.1 Clamped isotropic plate under central load
We will validate the triangular plate and shell element with numerical ex-
amples. The first example is a fully clamped isotropic plate under both
uniform pressure and applied concentrated load for which analytical Kirch-
hoff solutions for the central displacement and moment were given by Tim-
oshenko [44]. They are
w~ef = 0.00126 ~ , Mx = 0.0231ql2, uniform pressure,
We n,
ref = 0.00560 Pl 2 centra11oad .
Table 6.1: Central displacement error for a fully clamped isotropic plate
under uniform pressure and central load
I mesh unknowns ~c{p~essure) . Mx{pressure) We {central load) I
3x3 16 -7.68% 16.41 1.03%
5x5 80 -1.58% 21.00 0.89%
7x7 192 -0.38% 21.96 0.80%
9x9 352 0.04% 22.33 0.80%
exact 23.10
Due to symmetry we discretize a quarter-plate with a set of triangular

elements. Table 6.1 shows the obtained percentage error for the central
displacement with successive mesh refinements for both loading cases. It
also shows the moment values for the plate subjected to uniform pressure.
Note that the moment is computed at the centroid of a nearby element
and not at a node, and therefore some slight deviations are expected with
the reference solution. The classical value of ScI = 0.83333 (= 5/6) is
automatically computed and used as the shear correction factor. For a
visual inspection the results are also presented in Fig. (6.16).
6.17.2 Thick sandwich plate under uniform pressure
We consider a simply supported sandwich plate subjected to uniform pres-

sure load for which an analytical Reissner solution for the central displace-
ment and moment taking into account transverse shear deformation is avail-
able. Figure 6.17 shows the geometrical and material data for this problem.
The sandwich plate comprises a fairly thick panel (l/h = 8.14) and is a chal-
lenging test for simple shear deformable elements. The panel includes two
thin faces of thickness 1 mm each and a 15 mm core. For accurate bend-
ing estimation this structure requires shear correction factors. They are
automatically computed as ScI = 0.1101. Values for the central deflection
and moment and associated percentage errors as obtained with the element
TRIC using three layers and a quarter plate mesh discretization are shown
in Table 6.2. Also shown is the analytical Reissner solution. The agreement
is very good. By using only 8 degrees of freedom we achieve an error of
-1.19 %.
2 error %
1
0 I 01"\ ::I .f. unknowns
: 16 ov
352
-1 192
-2
-3
-4 ----ft- w uniform pressure
-5
-6
--e-- w central load
-7
-8
Figure 6.16: Convergence of central displacement for a fully clamped

isotropic plate
Table 6.2: Central displacement and moment for a thick (i;=8.14) sandwich
plate
mesh unknowns - - We - M~I
3x3 8 31.08{-1.19%) 3058
5x5 104 31.22{-0.74%) 3646
7x7 228 31.22{-0.74%) 3749
9x9 400 31.22{-0.74%) 3767
exact 31.4537 3668.3
6.17.3 Pinched cylinder; Scordelis-Lo roof; pressurized shell

Shown in Fig. 6.18 are three cylindrical shells, namely a pinched cylin-
der supported by end rigid diaphragms, a cylindrical roof supported also
by rigid diaphragms and loaded by its own weight, and a pressurized or-
thotropic shell with clamped edges. Due to symmetry, only one octant of
the pinched cylinder and the glass-epoxy pressurized shells, and a quarter
of the cylindrical shell roof are discretized with a set of three-node trian-
error %
2
L5
-e-- wc f
0.5 unknowns
18 104 228 400 f
0
-0.5
-I
-1.5
Ef =6.8e9 Pa E.,= 4.8 e8 Pa
-2 analytical sol. we = 31.4537 mm
Gf = 2.615 e9 Pa =
Gw 38 e6 Pa
Me = 3668.3
Vf =0.3 V., = 0.3
",=0.001 m hw=0.0I5 m
Figure 6.17: Convergence of central displacement and moment for a sand-

wich plate under uniform pressure
gular elements. Appropriate symmetry boundary conditions are imposed.

Tables 6.3, 6.4 present the normalized vertical displacement under the load
for the pinched cylinder, and at mid-side of the free edge for the cylin-
drical roof. Comparison is attempted with normalized solutions extracted
from the literature. The latter were obtained using 4-node quadrilateral
elements. For all three shells, our simple triangular element performs very
well and converges to the exact solution. Table 6.5 lists our finite element
solution for the maximum radial displacement, central hoop stress, and
central bending moment for the glass-epoxy pressurized shell. These are
compared with the analytic solution. Good agreement is achieved. Figure
6.19 displays the convergence of the normalized displacements for all three
shells.
6.17.4 Pinched hemispherical shell

Figure 6.20 displays a quarter of a pinched hemispherical shell which is
under the action of two concentrated forces. Large portions of this shell un-
dergo rigid body rotations. This challenging problem tests the ability of a
shell element to represent inextensional bending modes since the membrane
strain developing in the thin shell is rather insignificant. Note that accu-
rate representation of the shell motion requires the ability to model rigid
pinched cylinder
E=3.0x 106
v =0.3
h=3.0
L= 600.0
R =300.0
p= 1.0
rigid diaphragm
E= 3.42 x 106
v=O
h =0.25
q = 90.0/ unit area
L=50.0
R=25.0
9= 80
Internal pressure p
z
=
2.0372 psi
y
1<1 t>i
L == 20 in
E
xx
= 2.0 x 106 psi E == 7.5 x 106 psi
yy
Gxy = 1.25 x 106 psi Gxy = 0.625 x 106 psi vxy = 0.25
Figure 6.18: Pinched cylinder; Scordelis-Lo roof; glass-epoxy pressurized

cylinder
Table 6.3: Normalized displacement (in parentheses) for pinched cylinder;

comparison with 4-node quadrilateral elements; w re / = 1.8245 X 10- 5
1_no~es_ SRI[67] __RSDS[68]_ Mixed[69] QPH[70] Present I
5x5 (0.373) (0.469) (0.399) (0.370) (0.394)
9x9 (0.747) (0.791) (0.763) (0.740) (0.778)
13x13 (0.905)
17x17 (0.935) (0.946) (0.935) (0.930) (0.953)
25x25 (0.988)
29x29 (0.996)
Table 6.4: Normalized displacement (in parentheses) for the Scordelis-Lo

roof; comparison with 4-node quadrilateral elements
I nodes SRI[67] MITC4[71] Mixed[69] QPH[70] Present I
i 5x5 (0.964) (0.940) (1.083) (0.940) (0.697) I
9x9 (0.984) (0.970) (1.015) (0.980) (0.902)

13x13 (0.981)
15x15 (1.001)
17x17 (0.999) (1.000) (1.00) (1.010)
exact 0.3024
Table 6.5: Maximum displacement, central hoop stress and central moment
for a glass-epoxy pressurized cylinder (R/h = 20)
I nodes unknowns w~ax x 10 4 O'~Mc I
3x3 22 2.44 79( -33.30%) 16.12 2.080
5x5 94 3.4076(-7.15%) 29.93 4.427
7x7 214 3.6300(-1.09%) 35.15 6.041
9x9 382 3.6558(-0.39%) 37.74 6.98
exact 3.67xl0 [72] 36.70 7.92
w/w
1.1 ~ ref
21 25 29
1
nodes x
0.9
0.8
0.7 w pinched cylinder
0.6 w Scordelis-Lo roof
w pressurized shell
0.5
0.4"
0.3
Figure 6.19: Convergence of normalized displacement for the pinched cylin-

der, the Scordelis-Lo roof, and the glass-epoxy pressurized cylinder
body rotations. In addition to modeling inextensional bending modes, the

problem is a test for warping behavior of shell elements. Table 6.6 shows
the normalized values (in parentheses) with mesh refinement as obtained
with the present formulation and other 4-node quadrilateral elements. The
element converges rapidly and compares very favorably with all other shell
elements. Figure 6.21 shows the undeformed and deformed structures.
6.17.5 Twisted beam

Figure 6.20 displays a cantilever beam along with its geometrical and ma-
terial properties. It is subjected to a concentrated load applied at the tip
in the in-plane direction. The undeformed cantilever has a 90 twist. This
particular example tests the effect of warping on the response of shell ele-
ments. The exact solution for the static displacement in the load direction
is 0.005424 as quoted in [73]. The convergence characteristics of our shell
- 6.825010 '
v - OJ
R - 10.0
h O,04
,
F - I
1
a
;" "J
E - 2,9 10'
v _ 0.22
h -0.32
~ u - 12.0
b- U
F, m 1

F _ I
Figure 6.20: Pinched hemispherical shell and twisted beam

~
N
l
M
II)
N
~
~
'<I;
;t
~
~
~
::!:
~
Ii
en
w
N~
a:
~

Q.
Z
o
~
::!:
a:
12w
Cl
Figure 6.21: Undeformed and deformed plots for the pinched hemispherical
shell
Table 6.6: Normalized displacement (in parentheses) for pinched hemi-

sphere; comparison with 4-node quadrilateral elements; maximum deflec-
tion u x 10- 1
I nodes SRI[67] MITC4[71] Mixed [69] QPH[70] Present
5x5 (0.412) (0.390) (0.651) (0.280) 0.945(1.022)
9x9 (0.927) (0.910) (0.968) (0.860) 0.936(1.013)
13x13 0.927(1.003)
17x17 (0.984) (0.989) (0.993) (0.990) 0.924(1.000)
Exact 0.924
Table 6.7: In-plane normalized tip displacement for the twisted beam ;
comparison with 4-node quadrilateral elements; tip displacement x 10- 2
I nodes RSDS[68] QUAD[73] URI[68] URI [74] Present
, 13x3 (0.993) (1.009) (1.004) 0.5347(0.986)
15x3 0.5466(1.007)
25x5 (1.411) (1.001) (0.999)
Exact 0.5424
element are presented in Table 6.7 where it is also compared with other
available solutions. Very good agreement is again obtained. Figure 6.22
provides us with plots of the undeformed and deformed beams.
6.17.6 Eight-layer (0/45/ - 45/90)5 laminate under uniform

load
We consider a simply supported 8 layer quasi-isotropic laminate subjected
to uniform pressure load with the following material properties:
El = 40, G12 = 0.5, G23 = 0.2, 1I12 = 0.5, G 13 = G12, 1I13 = 1I12
E2 E2 E2 (6.293)
The central normalized deflection (as obtained with a quarter-plate mesh

of 9 x 9 nodes resulting in 400 unknowns)
E~OOO"
9<:LlOOO"
6SL Loo"
StSLoo"
~
-
Q)
S
Q)
c:E6Loo"
Q)
.-
.- 8LEC:OO"
]
CIJ t>OLc:oo"
't:l
~ 060E00"
....
Q)
co;l
.- LltEoo"
0-
....
Q)
..... E~oo"
CIJ
0
S'
0
61><:1>00"
0 91:91>00"
<:lOSoo"
8OI>SOO"
S6LSOO"
tS:SS:SL 96-6n V-6<: Soc: N\I\I1Vd n-ti N\I\I1Vdr.>S~ n=S3t1 1"<:=::>1 l01d NOI1V~tI
00 tS:SS:SL 96-00 '1""6<: S"<: N\I\I1Vd 1>" L-ti N\I\I1Vdr.>S~ (9V~-o3AlI" L=S3t1 1"<:=::>1 l01d 39NltI
C"I
C"I
Table 6.8: Central normalized displacement (w = wE2/f

qo
3 .104 ) for a (0/45/-
45/90)8 laminate for various aspect ratios s

I s = l/h higher-order [75] TRIC I
4 1.6340 1.6802
10 0.5904 0.5956
20 0.4336 0.4327
50 0.3857 0.3845
100 0.3769 0.3770
- 104 ,
W = WE2h3.
qo
.A
is compared with the a higher-order theory by Phan and Reddy [75] in

Table 6.8 for various aspect ratios s = 1/ h. The shear correction factors
computed on the three triangular edges are Se/a = 0.8545, Sel{3 = 0.5434
and Sel"! = 0.6394. Note that the higher-order theory does not require
shear correction factors. An examination of Table 6.8 reveals favorable
agreement.
6.17.7 Stresses in a sandwich plate-comparison with the

elasticity solution
Pagano [76] considered the response of a square (a = b) sandwich plate
under distributed sinusoidal loading. The material properties of the face
sheets were
El = 25 X 106 E2 = 106
G 12 = 0.5 X 106 G 13 = G 23 = 0.2 X 106 (6.294)
1/12 = 1/13 = 0.25
and of the core were (transverse isotropy with respect to the z axis was
assumed)
Table 6.9: Normalized through-the-thickness stresses axx and ayy , for a

square sandwich plate for various aspect ratios s (CPT is an abbreviation
for Classical Plate Theory)
s = l/h - (a b 1) - (a b 1)
U xx 2' 2' 2 U yy 2' 2' 2
4 present 0.7769 0.1960

elasticity 1.5560 0.2595
10 present 0.9715 0.0970
20 present 1.0312 0.0662
50 present 1.0550 0.0553
100 present 1.0610 0.0535
CPT 1.0970 0.0543
Exx = Eyy = 0.04 X 106 Ezz = 0.5 X 106

Gxz = Gyz = 0.06 X 106 Gxy = 0.016 X 106 (6.295)
Vzx = V zy = v xy = 0.25
The thickness of each face sheet was h/10. Normalized stresses at selected
through the thickness locations as obtained from the present analysis and
the three dimensional solution of [76] are presented in Tables 6.9, 6.10.
6.17.8 Deformation of a (0/90/0) square laminate

A symmetric 3-ply square composite laminate (a = b) composed of or-
thotropic layers of equal thickness with lamination (0/90/0) is considered.
Simply supported boundary conditions are imposed. The cross-ply com-
posite plate is subjected to a double sinusoidal load. For this laminate
three-dimensional solutions were reported by Pagano [76]. The following
material properties are used:
Table 6.10: Normalized through-the-thickness stresses O'xy, O'xz, O'yz, for a

square sandwich plate for various aspect ratios s
s = l/h O'xy(O,O,!) O'xz(O, !, 0) O'yz(~,O,O)
4 present 0.1065 0.2567 0.0940 !
elasticity 0.1437 0.2390 0.1072

10 present 0.0613 0.2987 0.0492
elasticity 0.0707 0.3000 0.0527
20 present 0.0251 0.3100 0.0348
elasticity 0.0511 0.3170 0.0361
50 present 0.0431 0.3120 0.0283
elasticity 0.0446 0.3230 0.0306
100 present 0.0424 0.3122 0.0271
elasticity 0.0437 0.3240 0.0297
--
CPT - -- ----_
0.0433
_-_
.. .. _ --
0.3240 0.0295
El = 25 X 106 E2 = 106
G 12 = 0.5 X 106 G 13 = G23 = 0.2 X 106 (6.296)
1I12 = 1I13 = 0.25
All displacements and stresses are presented in the following normalized

form:
100E2 w 1 . 7rX . 7ry

ill = qo h S 4 ' S = h' q = qo sm --;; sm b'
1
(O'xx, O'yy, O'xy) = --2
qos
(o-xx, O-YY' o-xy) , (6.297)
1
(O'xz, O'yz, O'xy) = -(o-xz,
qos
o-yz).
Due to biaxial symmetry, all computations are conducted using a quarter

plate discretization. Figure 6.23 shows the convergence of the normalized
central displacement with mesh refinement as well as the exact elasticity
/ / quarter model (294 dot)
~ol ~I/.
dof wlwexact error (%) 112
24 0.9164 -8.35
54 0.9653 -3.47
96 0.9834 -1.66
c III
150 0.9917 -0.83
216 0.9962 -0.38 0.q (/ 12,0)
294 0.9988 -0.12
IIh=\o
zlh
zlh
-0.5 -0.5 , - - - - elasticity
-0.4 -0.4167 \?",
V shell element
-0.333 'r7 '
V
....
..........
-0.3
-0.25 V ...... - ........................
-0.2 ...............
sbell element -0.167 V "" .... ,
-0.1 elasticity -0.0832 V \
o o V \
0.1 0.0832 V ./
0.167 V .... .... '"
------
........
0.2
0.25 V
0.3 ....
0.333 V./"' ....
./
0.4 (j (/12,0) 0.417 ~./ iiy/O,V2)
.q ./
0.5''''''''''''- , 0.5'
..~-.l.--...L.-.......- - . l . - -......- .....- - - '
o 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Figure 6.23: Displacements and stresses for a square (0/90/0) composite

laminate
solution. Also shown are the through-the-thickness distributions of the

normalized transverse shear stresses o-xz, o-yz for l/h = 10 along with the
three-dimensional elasticity solution of [76]. We observe that the conver-
gence of the central displacement to the exact solution is very good, and
the distribution of the transverse shearing stresses given by our lower order
(0190/0)
simply supported
z
K ~/ t::>-x
Figure 6.24: A rectangular (0/90/0) composite laminate
element is very satisfactory.
6.17.9 Stresses in a (0/90/0) rectangular laminate

Figure 6.24 displays a simply supported bidirectional rectangular laminate
(b = 3a) subjected to a double sinusoidal load. This composite plate was
also considered by Pagano [76]. The layer material properties are similar
to those used for the square laminate of the previous example. All dis-
placements and stresses are computed in normalized form. Due to biaxial
symmetry, only a quarter plate is discretized using 9 x 9 nodes in the x and y
directions, respectively. The three shear correction factors are computed as
Se/a = 0.5827, Se/(3 = 0.8786 and Sef'y = 0.7630. Tables 6.11, 6.12 present
the finite element results for the maximum displacements and stresses for
the rectangular cross-ply laminate and their favorable agreement with the
three-dimensional solution. We notice in particular the accurate results
obtained for the transverse shearing stresses. For aspect ratio s = 4, the
transverse shear stress is plotted in Fig. 6.25 and compared with the elas-
ticity and classical plate solutions [76].
6.17.10 Large deflections of an isotropic plate-comparison

with experimental results
We consider large deflections of an isotropic square plate shown on top of
Fig. 6.26 with all its geometrical and material properties. Two edges of
the plate are free and two fully clamped. A point load is applied near the
Table 6.11: Central normalized displacement ill and through the thickness
stresses (jxx, (jyy, for a (0/90/0) rectangular laminate for various aspect
ratios s. CPT is an abbreviation for classical plate theory
s = l/h ill - (a b 1) - (a b 1)
f7 xx 2' 2' 2 f7 yy 2' 2' 6"
2 present 9.8156
elasticity 8.1700
4 present 3.0320 0.6000 0.0972
elasticity 2.8200 0.7260 0.1190
10 present 0.9982 0.6012 0.0420
elasticity 0.9190 0.7250 0.0435
20 present 0.6384 0.6012 0.0285
elasticity 0.6100 0.6500 0.0299
50 present 0.5191 0.6020 0.0233
elasticity 0.5200 0.6280 0.0259
100 present 0.5009 0.6021 0.0224
elasticity 0.5080 0.6240 0.0253
CPT 0.5030 0.6230 0.0252
left boundary at point N. Due to symmetry, a half-plate finite element

mesh utilizing 6 by 11 nodes in the x and y directions, respectively, is used.
The load-displacement curves for points A, B are presented in Fig. 6.26
along with the experimental results of Kawai and Yoshimura [77]. It is seen
that close agreement with the experiment is obtained. Figure 6.27 shows a
deformation plot at P = 20kg.
6.17.11 Buckling of a cross-ply (0/90/90/0) laminate
We consider buckling of a symmetric four layer (0/90/90/0) simply sup-

ported square laminate subjected to in-plane uniaxial compressive loads.
Due to biaxial symmetry, only one-quarter of the plate is discretized with
a set of triangular elements. The following material and geometrical prop-
erties are used:
Table 6.12: Normalized through the thickness stresses O'xy, O'xz, O'yz, for a
(0/90/0) rectangular laminate for various aspect ratios s
s = l/h O'xy(O, 0, k) O'xz(O, ~,O) O'yz(~, 0, 0)
4 present 0.0212 0.4318 0.0366

elasticity 0.0281 0.4200 0.0334 I
10 present 0.0112 0.4241 0.0169
elasticity 0.0123 0.4200 0.0152
20 present 0.0090 0.4218 0.0110
elasticity 0.0093 0.4340 0.0119
50 present 0.0083 0.4226 0.0090
elasticity 0.0084 0.4390 0.0110
100 present 0.0082 0.4249 0.0099
elasticity 0.0083 0.4390 0.0108
CPT 0.0083 0.4400 0.0108
El = 40, G 12 = 0.6, G23 = 0.5,

E2 E2 E2
G12 = G13, 1/12 = 0.25, 1/13 = 1/12, 8 = l/h = 10. (6.298)
The ratio El / E2 is usually referred to as the material anisotropy ratio.

The elastic stability of this laminate was examined by Phan and Reddy
[75] and Noor [78]. They reported normalized buckling loads by exact
implementations of a classical plate theory (CPT), a higher-order shear
deformation theory (HSDT) [75], and theory of elasticity [78]. Table 6.13
presents the comparison of the buckling loads. Our result is obtained by
using a mesh of 5x5 nodes in the X,y directions resulting in 97 unknowns.
The Table reveals that our solution conforms very well with the reference
solutions. The agreement is very good even for high material anisotropy
ratios. It is seen that with increase of the anisotropy ratio, the ability of
the classical plate theory to accurately predict the buckling loads decreases.
0.5 Z ~ TRIC
0.4 0 3-D
\
Ll CPT
0.2 [ Ll
oI
0.1 Ll _ _
I I I I I I
~tyz (a/2,0,zJ
I
o 0.005 0.01 0.015 0.02 0.025 0.03 0.0 0.04
-0.1 I- Ll
o
-0.2 ~ Ll 0
-0.3
Ll
-0.4
-0.5
Figure 6.25: Normalized transverse shear stress distribution for the

(0/90/0) composite laminate. Comparison with the elasticity and classical
plate solutions
Table 6.13: Normalized buckling loads (. ~:i~) for a (0/90)8 square laminate
(l/h = 10). CPT, HSDT, are abbreviations for classical plate theory and
higher-order shear deformation theory, respectively
E 1 /E2 CPT[75] HSDT[75] present elasticity
3 5.7538 5.1143 5.3638 5.2944

10 11.492 9.7740 9.8724 9.7621
20 19.712 15.298 15.1041 15.0191
30 27.936 19.957 19.3093 19.3040
40 36.160 23.340 22.7810 22.8807
400mm
I<::l l>-I
clamped edge
E =2.15 X 104 kg/mm2

v =0.3
P=20kg
somm T h = 1.9Smm
't
(7""" '. '.e 400mm
free edge I SO mm free edg<
clamped edge
w(~m~m~)~~~~~=j____~/:::: __________~
21~
I.S point A
o experiment
1.6
point B
1.4
-------
1.2
/"
//
// _-...0-
O.S // //...0-
// //0
0.6
/ ..-0"/
/~/
~'l/
~'/'
0.4
/d'
O.H- I //
//
. // P (kg)
Olk/
o-t-T~-1-
2 4 6 S 10 12 14 16 IS 20
Figure 6.26: Free-free-clamped-clamped plate under point load; load-

displacement curves; experimental results
...
co
~ lii'"
a>
co
~ ~
~
"!
...
'" a '"
CD
O!
'"
~
'"
~
<D
~
III
co
~
l!iJ
~
;;;
~
...'"
'"
o
M
~
i.,
(/)
~
II>
N
~
~
~
ci:.
z
~
~
~
::1:
~
W
N~
II:
'"
<If
()
...J
b
...J
0..
Figure 6.27: Magnified deflection of the point-loaded plate at P = 20 kg

Table 6.14: Normalized buckling loads (. :;i!) for a (45/ -45/0/90)8 square
laminate (l/h = 10) for various aspect ratios s
s = l/h present HSDT[75]
10 27.0966 26.799
20 37.4311 37.115
50 42.1436 41.877
100 42.9401 42.819
6.17.12 Elastic stability of a (0/45/ - 45/90)8 laminate

We turn now our attention to quasi-isotropic laminates which are com-
posed from a combination of four layers (in any order) with fiber orientation
+45, -45, 0, 90. These special laminates are used extensively in practical
applications. For this class of composite plates close form solutions can-
not be obtained. We study in particular an eight-layer simply supported
(0/45/ - 45/90)8 square laminate for which we seek the lowest buckling
load. We compare our solution with the finite element solution of a higher-
order theory by Phan and reddy [75]. One-quarter of the plate is considered.
The material properties are identical to those used in the previous example
for the (0/90/90/0) cross-ply plate. Furthermore, we vary the length over
thickness aspect ratio s, and monitor the effect on the corresponding lowest
buckling mode. Table 6.14 shows the finite element solution and its com-
parison with the results of [75]. Table 6.14 reveals an excellent agreement
between our natural theory and the higher-order theory. Note that further
refinement of the mesh did not modify the critical load significantly.
6.17.13 Thermomechanical buckling of a cylindrical com-

posite panel
Next we consider larger composite panels for which our element is con-
ceived. Fig. 6.28 shows a quasi-isotropic cylindrical panel subjected to
end compressive load and temperature increase. The finite element mesh
is shown in Fig. 6.29. It includes 441 nodes and 2398 unknowns. Our
objective is to study the thermomechanical buckling using the developed
element and methodology. This is successfully accomplished. The first
z.w
IV
y. v
u=w=, ='11 = 0
x,u
. - ~90
2 '
T 45
w=9=O
441 nodes
-45
2398 unknowns
EI = 150GPa
E2 = 10 GPa
G I2 =6GPa=GI3 hL =1.35 x 104 m
G23 =4GPa Rlh = 92.59
VI2=VI3=v23=O.25
a l = 2.5 x 10.8 C I
a2 = 30 x 10-6 C l
Figure 6.28: Composite cylindrical panel under edge load and temperature
increase
four buckling modes due to applied edge load are shown in Figs. 6.30-6.33,
while the first four buckling modes due to temperature are depicted in Figs.
6.34-6.37.
6.17.14 Buckling of a composite vessel under hydrostatic

pressure
Figure 6.38 illustrates a {45/ - 45/0/90)8 composite cylinder closed with
hemispherical caps. The cylinder is a rather thick shell with a radius-to-
thickness ratio of R/ h = 12.5. This can be verified from the geometrical
and material data provided in the figure. The shell is subjected to external
'v
+
Figure 6.29: Finite element mesh for the {45/ -45/0/90)8 cylindrical panel
c;:?
f'-
M
~
6.
Ql
en
~
10
N
~
~
'<l:
ri:
z
~
~
U
en
::2;
'v
+
Ii
en
w
a:
N
II
o
...J
b
...J
Il..
Z
o
~
::2;
a:
ou..
W
o
Figure 6.30: First elastic buckling mode (Per = 65914Pa) for the composite
cylindrical panel
~
I"-
M
i
~
II)
N
~
~
~
~
Ii:
~
!(
D..
<3
en
>v
::::E
~
;g +
w
II:
"!
~
9
D..
Z
o
~
~
f(
w
o
Figure 6.31: Second elastic buckling mode at Per = 71117Pa

g
t.:.
M
~
~CD
en
~
Il)
N
~
~
.q;
d:
I
en
::i!
'v
Ii +
en
w
II:
'"
N
g"
b
--'
D..
Z
o
~
::i!
II:
ou.
W
Cl
Figure 6.32: Third elastic buckling mode (PeT = 71700 Pa) for the compos-
ite cylindrical panel
It)
<0
M
tOJ
(J)
~
It)
N
~
~
~
d:
z
~
~
(J)
::!!
+
~w
a:
~
~
b
'\/
....I
a.
z
o
~
::!!
a:
1(
w
c
Figure 6.33: Fourth elastic buckling mode at Per = 77627 Pa

:;:
co
c.;
~
~
til-
C/l
.,j.
l/)
N
~
'<I:
~
z
~
C3
C/l
:i:
'v
~
Ii +
faII:
en
~

Q.
Z
o
~
:i:
II:
f2w
o
Figure 6.34: First thermal buckling mode (Tcr = 168.74C) for the com-
posite cylindrical panel
8
OJ
M
~
5r
~
Ii)
N
~
~
'<I:
r:i:
z
~
~
~
~
Ii
en
+
'\/
w
a:
o
N
~
Il..
Z
o
~
~
a:
~
w
Cl
Figure 6.35: Second thermal buckling mode at Tcr = 199.34C

~
Oi
(0)
~
10
N
~
~
~
ci:
~
~
::;:
~
'v
Ii
en +
w
a:
CJi
g
ga.
z
o
~
a:
2
w
c
Figure 6.36: Third thermal buckling mode (Tcr = 294.71 DC) for the com-
posite cylindrical panel
;:Ii
0;
~
c.;
~
.,
6.
en
..j.
~
on
N
~
~
~
~
ri:
z
~
~
U
en
::E
'v
Ii
en +
w
II:
N
~
a.
z
o
~
::E
II:
f(
W
C
Figure 6.37: Fourth thermal buckling mode at Tcr = 355.14C

hydrostatic pressure and supported at three points. It is discretized with a

set of triangular shell elements resulting in 735 nodes and 3954 degrees of
freedom. Figure 6.39 displays the finite element mesh. Figures 6.40, 6.41,
display the first two buckling modes. The first computed critical pressure
is qcr = 0.7060 X 107 Pa while the second buckling mode corresponds to a
critical pressure of qcr = 0.7340 X 107 Pa.
6.17.15 Buckling of a rocket-like composite shell under ex-

ternal pressure and temperature
The top of Fig. 6.42 shows a schematic of an unstiffened rocket-like com-
posite shell. All geometrical and material data are shown in the same figure.
The composite shell is supported by rigid diaphragms. No displacements
are allowed at the supports. It is subjected to external pressure and tem-
perature increase. The mesh includes 919 nodes and 1960 elements for a
total of 4894 degrees of freedom. Using this mesh a convergent solution
is obtained. The thin rocket-like shell comprises 8 layers. Eight differ-
ent lamination schemes are considered, namely (0)8, (90)8, (0/90/0/90)5'
(45/ -45/0/90)5' (45/0/ -45/90)5, (30/ - 30)5, (60/ - 60)5, (75/ -15/0/0)5'
Figure 6.42 displays the critical pressures and temperatures obtained
for the different lamination schemes. We observe first that the shells with
the highest buckling pressures do not display equally high buckling temper-
atures. Indeed the (0/90/0/90)5 shell has the highest critical pressure load
but the third largest critical temperature. Interestingly enough, the shell
with the highest critical temperature is the one with lamination (30/ - 30)4
-it has only the fifth higher critical pressure load. The shells with lamina-
tions (60/ - 60)4 and (90)8, respectively, possess the two lowest buckling
pressures and temperatures. Figure 6.43 illustrates the critical buckling
mode for the (75/ - 15/0/0)5 laminated shell. A similar buckling mode is
obtained for pressure loading.
From the above results we deduce the high potential and tailoring capa-
bility of fiber composites. To further illustrate this point, we combine some
of the above fiber orientations and create the eight layer shell with lami-
nation (0/30/ - 30/90)5' This lamination scheme combines the 0 and 90
fiber orientations of the cross-ply shell which exhibited the highest buckling
pressure, and the 30, -30 fiber orientations of the shell with the highest
critical temperature. An elastic stability analysis of this structure yields a
critical pressure load of 29145.91 Pa, and a critical temperature of 41.23 C.
--
c:::;:::::::.. \-- --
-- -- c:::;:::::::..
--
--
c:::;:::::::..
c::::::>
-- -- --
-- ~w
--
--
(45/-45/0/90)s 735 nodes
3954 unknowns
EI =20GPa
R = 250 mm
E2 =! GPa
h=20mm
G I2 =0.6 GPa =G I3
1= 1400mm
G23 =0.5 GPa
Rlh= 12.5
V 12 =V 13 =V 2J =0.25
Figure 6.38: Schematic of a thick composite submarine vessel subjected to

external pressure
Figure 6.39: Pressurized submersible composite shell; finite element mesh

253
Figure 6.40: First buckling mode for the composite pressurized vessel
l
CS:Cl>:l>L 96-fin'l-9Z S"Z N'II::U'Id n -l:l N'Il:ll'ldr.>S~ L" L=S3l:l L " z=~:n .10ld NOll"'1~Il:lOlj3IJI
CS:Cl>:l>L 96-00'l-9Z S"ZN'Il:l.1'1d n-l:l N'Il:l.1'1d/:>Sr-I (')'1r-1~3NL"L=S3l:l L"Z=:>1 .10ld
FR NGE
I
t,;)
c;,
NGE
FRINGE PLOT LC=2.2 RES=1.1(VEC-MAG) MSC/PATRAN R-l.4 PATRAN 2.5 26-Aug-96 14:45:57 >I:>--
EFORMATION PLOT LC=2.2 RES=1.1 MSC/PATRAN R-1.4 PATRAN 2.5 26-Aug-9614:45:57
"'rj
......
aq oo
~
(1) .
~
~
......
~ ....
...... (1)
en '"C
(1) [
C")
o (1)
::l
0-
c"" 0-
00
~ ~
(1)
-
Jg'
.2971
[
g. .23n a
.1783
.1188
j-' .05942
These values are the highest of all previously obtained. Of course, a more
optimum orientation of fibers should be obtained by employing formal op-
timization procedures.
Cf 0>
diaphragm
EI = 150GPa
c;~~. E2 = 10 GPa
GI2 =6GPa=G I3
G23 =4GPa
fc::!
0.250 mcf::J 1.40 m
t>j<I
0.30 m
t>j v 12 =v 13 =v23 =0.25
a l = 2.5 x 10-8 C- I
a =30 x 1O.6 c- 1
2
-1 10-4 m
-V (751-151010), hL ' 1.35 X
8
RI h = 231.48
7 V (601-60)4
R= 0.250m
6 iii (301-30)4
5 -V (45101-45190),
4 II (45/-4510190),
3 v (0I901019j I),
2 v (90)g
" (O)g
o 5000 10000 15000 20000 25000 30000
critical pressure (Nlm 2)
t1-
8'l..- -::::::J (751-15I0I0
lL
7 J (601-60)4
b::
6 - . (3 01-30)4
~
5
C:::
'-- -:II '=:J (45101-45190)
4 --. (451-45I0I90) ,

3 -::J (019010190),
IJ.
2 -:1 (90)g
J,..
... (0),
o 5 10 15 20 25 30 35
critical temperature (0 C)
Figure &.42: Rocket-like shell; critical pressures and temperatures for eight
lamination schemes
Figure 6.43: First thermal buckling mode for the (75/ -15/0/0)8 composite
shell at Tcr = 29.4193C
6.18 Problems
1. A strain rosette is placed on a plane so that it forms the following
three angles with respect to the x axis: a) 30, b) -30, c) 90. The
strains determined by the rosette during a static test are El = 800,
E2 = 200, E3 = -50.
what are the total natural strains?

place the natural coordinate system in xy
derive the expression for matrix B
determine the three strains with respect to the xy axis
sketch the natural forces on the rosette
place a triangular element so that it conforms with the measured
strains
illustrate the component stresses
2. Suppose that we apply the triangular element to study a flat plate

and a cylindrical panel. How would geometry affect the following: 1)
the natural stiffness matrix 2) the local elemental stiffness matrix 3)
the global stiffness matrix of the element
3. State the equilibrium of forces in the natural coordinate system for a

triangular finite element
4. Do the natural straining modes satisfy the compatibility conditions

along the edges of a triangular element? Substantiate your assertions.
5. Under which displacement assumptions can rigid body motion occur

in a triangular element? Provide an example.
6. In section 6.2 we defined total strains and component stresses. Is it

also possible to define component strains and total stresses? Provide
the physical idealization of the latter.
7. On the edges of a triangular sides are imposed natural symmetrical

and antisymmetrical moments Ms a , MAa, respectively. What are
the cartesian moments acting on the triangular edges? What are the
corresponding cartesian rotations?
8. Sketch a triangle with all generalized natural forces and moments.

Express these forces in the local cartesian coordinate system. Also
6.18 Problems 259
express the stress resultants in the three nodal local cartesian coor-
dinate systems.
9. On the edges of a triangular side are acting cartesian moments M 1 ,
M 2 , respectively. What are the edge symmetrical and antisymmetri-
cal moments and rotations?
10. Provide a definition for the shear correction factor. Under which
displacement or strain assumptions is the shear correction factor not
required?
11. A triangular element of area n is subjected to a uniform pressure
load of intensity q. What is the local and global vectors of kinemati-
cally equivalent loads? How would the presence of concentrated nodal
forces and moments affect the formation of this vector?
12. Derive explicitly the expressions for the natural thermal load vector
if the temperature interpolation relations (6.205) are used. Proceed
with the evaluation of all integrals.
13. How does the geometrical stiffness of the shell element relate to the
geometrical stiffness of the beam element?
14. Prove expression (5.261).
15. For the pinched cylinder and the Scordelis-Lo roof shown in Fig. 6.18
provide the symmetry boundary conditions
16. Explain the rational for the creation of a weighted antisymmetrical
shear stiffness for the triangular element TRIC.
17. Why are rigid body rotations important for the formation of the ge-
ometrical stiffness?
18. Comment on the possibility of computing the through the thickness
stresses at the triangular nodes
19. Why are the natural modes and forces additive during a geometrically
nonlinear analysis?
20. In which cases do you expect the shear correction factor to signifi-
cantly affect the deformation?
21. What is the physical meaning of matrices a and A? Starting from
aA = 118, provide cross relations for the sub matrices of a and A
22. Under the action of temperature increase can a composite laminate

shrink?
23. Provide several examples of cross-ply and quasi-isotropic laminates
24. We considered a composite cylinder discretized with triangular ele-

ments. Make a coarse mesh and select in random 4 elements. For a
(45/90/ - 35/20/ - 10) lamination sketch the fibers for every layer
and show the forming angles with respect to the natural and local
cartesian coordinate systems
25. Suppose that a laminate is under the action of a hygrothermal strain

h. Provide the constituent relation and state the strain energy. Then,
derive the initial natural loads and estimate the local and global initial
loads.
Chapter 7
Computational statistics
7.1 A model problem

Following the theoretical formulation and computational validation of our
finite element methodology we now address some computational aspects
of our method. the cylindrical composite shell of Fig. 6.28 is selected as a
model problem to assess the computational advantages of the methodology
and obtain an indication about the efficiency of the computer program. The
cylindrical panel of Fig. 6.28 (the mesh is shown in Fig. 6.29) includes 800
elements for a total of 2398 degrees of freedom. The elastic and geometrical
stiffnesses contain 295969 elements which are stored in skyline form as one
dimensional arrays. Of interest is the estimation of the first four elastic
buckling modes of the composite shell structure. In order to compute the
critical loads both static and eigenvalue analyses are required. The code
is executed on a CRAY-C94 supercomputer, and special compilation direc-
tives are issued in order to measure the performance of all routines used
in the estimation of the buckling loads. Following execution, a statistics
report is issued by the computer in which the breakdown of the computing
time per routine is shown. This report is provided in the following section.
7.2 Computational statistics report

Following completion of the computational experiment, the following statis-
tics report is issued by the CRAY-C94 computer.
Flowtrace Statistics Report

262 Computational statistics
Showing Routines Sorted by CPU Time (Descending)

(CPU Times are Shown in Seconds)
Routine Name Tot Time Avg Time Percentage Accum%

---------------- -------- -------- ---------- -------
MIMA 2.21E+00 2.11E-05 19.14 19.14 ****
MATINV 1.38E+00 1.92E-04 11.94 31.08 **
CCLAYER 1.19E+00 1.86E-04 10.30 41.37 **
ZEROV 7.68E-01 2.85E-06 6.63 48.01 *
DECOMP 7.32E-01 7.32E-01 6.33 54.34 *
REDBAK 7.14E-01 1.05E-02 6.17 60.51 *
COLSOL 6.57E-01 3.29E-01 5.68 66.19 *
MANMESH 6.36E-01 6.36E-01 5.50 71.69 *
MULT 6.05E-01 8.90E-03 5.23 76.92 *
NATURAL3 4.54E-01 5.68E-04 3.93 80.85
OUTROUT1 4.20E-01 4.20E-01 3.63 84.47
OUTROUT 2.63E-01 2.63E-01 2.28 86.75
GEOMS 2.03E-01 2.53E-04 1. 75 88.50
GEOMSTIFHEATBUCK 1.78E-01 2.22E-04 1.54 90.04
SSP ACE 1.65E-01 1.65E-01 1.43 91.46
MIV 1.53E-01 7.63E-06 1.32 92.78
TRANSP 1.33E-01 4.63E-06 1.15 93.93
MATASS1 1.30E-01 1.62E-04 1.12 95.05
MATASS 1.29E-01 1.61E-04 1.11 96.17
FINDSKY 1.26E-01 1.26E-01 1.09 97.26
FORMRHS2 9.46E-02 1.18E-04 0.82 98.08
MINV 5.41E-027.52E-06 0.47 98.54
SANI 3.99E-02 3.99E-02 0.34 98.89
TRANSF3 3.09E-02 3.86E-05 0.27 99.16
FORMCAR 2.79E-023.48E-05 0.24 99.40
TRANSF2 1.41E-02 1.76E-05 0.12 99.52
FORMLOC 1.15E-02 1.15E-02 0.10 99.62
READ 1.11E-02 1.11E-02 0.10 99.71
TRANSF1 8.26E-03 1.03E-05 0.07 99.79
INV3B3 8.16E-03 3.40E-06 0.07 99.86
DETER 4.54E-03 2.84E-06 0.04 99.90
DRIVENAT 4.17E-03 4.17E-03 0.04 99.93
DRIVECAR 3.85E-03 3.85E-03 0.03 99.96
JACOBI 3.08E-03 1.81E-04 0.03 99.99
COPYV 1.04E-03 1.04E-03 0.01 100.00

FORMRHS1 1.35E-05 1.35E-05 0.00 100.00
SETUP 3.33E-06 3.33E-06 0.00 100.00
=========================================================
Totals 1. 16E+01 456174
The above CPU time list shows that most CPU time is primarily consumed
on algebraic operations such as matrix multiplications, and secondarily on
the core finite element computational procedures such as formulation of
stiffness matrices, assembly procedures, solution of the linear system of
equations, etc. This is in direct contrast with classical finite element meth-
ods in which most of the computing time is consumed on the computation
of the stiffness matrices, the assembly procedure, the solution of the linear
system of equations etc. More specifically the following observations are
made:
Routine MXMA : Performs matrix multiplications. It is seen from

the list that most of the CPU time (19.14%) is spent on matrix multi-
plications. To obtain the solution, 104800 matrix multiplications are
performed in 2.21 seconds. It is evident that vectorization or paral-
lelization can significantly speed up the computation (this is suggested
also by the CRAY computer with four star symbols).
Routine MATINV : This routine performs matrix inversions. A
total of 11.94% (1.38 seconds) time is spent on this procedure. Most
of the matrix inversions, however, are performed on 3x3 matrices; this
can be done analytically and not numerically as presently in the code
so that the computing time spent on this routine can be substantially
reduced.
Routine CCLAYER : This routine performs material transforma-
tions for each ply and element. It is called 6400 times and consumes
10.30% of the total CPU time. It is a perfectly parallel procedure.
Routine ZEROV : Zeros out all vectors and matrices. The routine
is not essential and can be eliminated (with caution).
Routine COLSOL : Solves the linear system of equations formu-
lated for the static problem in order to obtain the prestress state
within the structure. For a vector containing roughly 300000 ele-
ments, 5.50% of CPU time is considered a very good performance.
264 Computational statistics
Routines DECOMP, REDBAK, JACOBI, SSPACE : Used for

the estimation of the eigenvalues (buckling loads) using the Subspace
Iteration Method of [79]. All consume 13.96% of the total CPU time.
Routines MXV, MULT : Perform matrix-times-vector multipli-

cations. They are called a total of 2068 times and consume 6.55% of
total CPU time. Again, vectorization and/or parallelization can save
time.
Routine NATURAL3 : Computes the natural stiffness matrix of

800 triangular finite elements in 0.454 seconds (3.93% of CPU time).
This operation can be also performed in parallel.
Routine FORMCAR : Thansforms the elemental natural stiffness

matrices to the global cartesian coordinate system. It consumes only
0.0279 seconds (0.24% of CPU) for 800 elements.
Routine GEOMS : Computes the geometrical stiffness matrix of

800 triangular finite elements in 0.203 seconds (1.75% of CPU time).
Similarly to the natural stiffness matrix this operation can be per-
formed in parallel.
Routine MATASS : Performs the assembly operation of 800 elastic

stiffnesses. It requires 0.129 seconds or 1.11% of total CPU time. A
perfectly parallel procedure.
Routine MATASSl : Performs the assembly operation for 800

geometrical stiffnesses. It requires 0.130 seconds or 1.12% of total
CPU time.
Routines TRANSF1, TRANSF2, TRANSF3 : These routines

are called in the beginning of the computation and provide elemen-
tal data such as local-global direction cosine matrices, geometrical
transformation matrices, triangular areas, heights, etc. They require
0.46% of CPU time.
The rest of the routines, as revealed by the CRAY statistics report,

consume an insignificant amount of computing time.
It is evident from these observations that a high degree of computational
efficiency has been achieved and that the method shows great potential for
vectorization/parallelization. It is also clear that the theory presented
is sufficiently shaped and tailored for the computer. One important note:
the model problem which we have chosen here represents a medium size
problem. That is, with the increase of the unknowns it is expected that
procedures such as solution of the linear system of equations will consume
more computing time. This, however, represents another modern research
area in computational mechnics which deals with efficient solution algo-
rithms in association with fast finite element technology.
Chapter 8
Nonlinear analysis of
anisotropic shells
8.1 Stable and unstable equilibrium paths

In the presence of large deflections, bifurcations, and load and displace-
ment limit points, the analysis of arbitrary anisotropic shells requires the
adoption of incremental and iterative procedures. In many cases the
load-displacement curves may exhibit unstable branches followed by stable
equilibrium paths. The true response is dynamic in nature. However, a full
dynamic analysis is impractical and expensive. Thus in most cases a fully
static solution or a combined static and dynamic solution is performed.
The latter must be able to predict and pass critical limit points and predict
collapse loads. The state of the art in current solution algorithms is given
by Papadrakakis [80] and Crisfield [81].
The nonlinear analysis of shells requires the efficient blend of finite
element technology and path following techniques. Due to the iterative
nature of the solution process it is imperative that the structural response
is obtained by simple and inexpensive finite elements. Isoparametric fi-
nite elements based on higher order interpolation functions and multiple
quadrature loops can prove very expensive and cumbersome when applied
to large and complex multilayer shells.
In this Chapter we develop a powerful algorithm for nonlinear analy-
sis of isotropic and composite shells based on the natural mode method
and the cylindrical arc-length method for the determination of the loading
path. The blend of the natural mode method, on which the formulation
of the triangular shell element TRIC is based upon, and the arc-length

268 Nonlinear analysis of anisotropic shells
method, refined to work with TRIC, shows many advantages over classical
formulations: analytic and elegant expression for all elemental matrices; a
series of vector and matrix multiplications that can be easily optimized for
maximum speed; material generality; a choice of Newton-Raphson or modi-
fied Newton-Raphson solution algorithms; accurate location of bifurcation,
limit and displacement points; computational efficiency and economy; and
full potential for vectorization/parallelization. The arc-length method and
its variations have been extensively used by many authors for the study of
nonlinear shell behavior -we quote here the works by Riks [82], Crisfield
[83], [81] and Papadrakakis [84]. However, most of the studies are conducted
using classical finite element methods and confined to isotropic panels of
simple geometry. Thus we aim at developing a computational procedure for
large-scale nonlinear analysis of isotropic and composite shells of arbitrary
geometry by blending our finite element methodology with the arc-length
method. The latter is refined to work in concert with the natural mode
method.
8.2 The incremental/iterative scheme

Suppose that a structure is subjected to a load P and deforms in space by
u. During the course of a large displacement but small strain motion, the
equilibrium of the structure is expressed as
KTU=P, (8.1)
where K T is the global tangent stiffness matrix. We have seen that the
tangent stiffness matrix is given by the expression
KT ~ [T:6[ii~[ ~ JiiN]T06]
natural stiff. (I2xI2)
,
'"
local siff. k
(18x18)
global stiff. K E
(18x18) (8.2)
+ [Tb6 kGT06] + ([aNTo6fkNG[aNTo6]] .

---........-..-
simplified geometrical stiff. (I8xI8)
, . '
natural geometrical stiff. (I8xI8)
8.2 The incremental/iterative scheme 269
We note the first advantage of our formulation: The local elastic stiffness
matrix a~K NaN along with k-G and a~K NGaN can be expressed analyt-
ically. These elemental matrices are all added together to form the local
tangent stiffness matrix K LT in an element level. Thus only matrix mul-
tiplications Tb6 K LTT06, with T06 being a hyper diagonal matrix, suffice
for the formation of the global tangent stiffness. This action will signifi-
cantly speed up the computation process. Naturally, vector and parallel
computers can perform these operations very fast.
Now during nonlinear analysis with automatic load adjustments we will
use two measures to guide us in identifying stable, neutral, strongly and
weakly unstable regions. These measures include the energy of the struc-
ture (or stiffness parameter) using a predicted displacement vector, and
the nature of eigenvalues of the tangential stiffness matrix including posi-
tive eigenvalues, zero eigenvalues, number of negative eigenvalues etc. The
eigenvalue information is used to determine the sign of the determinant of
the tangent stiffness matrix. Symbolically, these parameters are expressed
as
predictor energy = U~KTUp,

Ep
or current stiffness parameter Sp = xbP, (8.3)
eigenvalues, determinant: An, sign(1 KT I).
We now list various possible equilibrium states.
1. When both the energy and first eigenvalue are positive, the motion
of the structure is said to be stable.
2. When the energy is positive but the first eigenvalue is zero a bifurca-
tion point has been encountered. Depending on the number of zero
eigenvalues, the possibility exists for other equilibrium branches be-
sides the primary path. Various bifurcations may include the follow-
ing: stable symmetric bifurcation; unstable symmetric bifurcation;
and unsymmetric bifurcation.
3. When energy and first eigenvalue are both zero the structure is at
neutral equilibrium; other equilibrium branches may exist. In general,
the existence of zero eigenvalues indicate the presence of bifurcations
and the possibility of secondary paths.
4. When one or more negative eigenvalues of the tangent stiffness are
found, the structure operates in an unstable region:
If both energy and eigenvalues are negative, the structure is said

to operate in a strongly unstable region.
If the energy is positive but one or more eigenvalues negative,
the structure has come into a weakly unstable region and tends
to gain positive energy. Frequently, this path is associated with
snap-back phenomena.
In the arc-length method, the structure is subjected to a loading AP,

where vector P represents a fixed external loading and A is a load-level
parameter. If m is the load step number and i the nonlinear iteration we
write
PH-I = P"t + OAi+1 P . (8.4)
In the course of an incremental process we seek to determine the param-

eter OAi+1 in order to advance the loading. The fundamental equilibrium
equation is
K"tXi = PH-I - F"t, (8.5)
where F"t is the vector of internal loads. If we denote the unbalanced force
vector by
gz' = Ff!tz - p.z, (8.6)
the solution vector is expressed as
Xi = OAi+IXf + xi (8.7)
The predictor displacement vector is
P
Xi = K-1p
Ti , (8.8)
and the corrector displacement vector
Xi
C
= - K-1
Tigi (8.9)
The new displacement increment at iteration i + 1 is
~Ui+1 = ~Ui + 8>'i+1 Xf + xL (8.10)
while the new displacement vector is
Ui+1 = Ui + 8>'i+1 X f + xi (8.11)

Following Crisfield [81], the cylindrical arc-length method gives a measure
of the arc-length radius ~l via
~U~+1 ~Ui+1 = ~l. (8.12)
On account of (8.10), equation (8.12) leads to the quadratic equation [80]
a18r+1 + a28i+1 + a3 = 0, (8.13)
where
P)t P
a1 = ( xi xi'
a2 = 2 [~Ui + xi]t xf, (8.14)
a3 = [2~Ui + xi]t xi.
For the first increment, 8>'1 is given as an input data and is used to de-
termine the arc-length radius ~l which remains constant throughout the
computational procedure. For subsequent increments, the arc-length is de-
termined by the roots of the quadratic equation.
Now before initiating the incremental/iterative procedure the following
computing steps are undertaken:
1. Formation of the local-global cosine matrix for each finite element.
2. Determination of the local coordinates of the nodal points.
3. Computation of the natural-local cartesian direction cosine matrix for

each triangular element.
4. Formation of the natural stiffness matrix K N for each distinct trian-

gle.
5. Conduct the elemental operation a~KNaN.
6. Formation of the loading vector.
Following these preliminary steps we initiate the incremental nonlin-

ear analysis. The computational procedure is described by the following
algorithm:
Loop over desired load increments inc
- Loop over maximum iteration numbers ite

* If ite=1 or Newton-Raphson solution method is used then
1. Compute and assemble the elemental geometrical and
tangent stiffnesses KG, K T, respectively.
2. Decompose the global tangent stiffness matrix KT.
* If ite=1 check for the presence of zero or negative eigen-
values and determine the sign of the determinant II K T II.
In case of a single or multiple zero eigenvalues decide about
branch switching and exploration of secondary paths.
1. Solve for KTXO = P.
2. Compute II Xo II = [X6 xo] 1/2.
If inc=1, then determine the arc-length radius from
fll = 8),1 II Xo II. fll remains constant through the
incremental process.
If inc;f 1 compute the current stiffness parameter (en-
ergy) Sp = x~P, and use it together with the eigenvalue
information to determine the sign of the first incremen-
tal loading parameter 8),1 = :=h
[
Ll 2 ] 1/2
:1:0:1:0
.
Update displacement flUI = 8),IXO. Update the load
parameter ),1 = ),0 + 8),1. GOTO 1
* If ite;f 1
1. If Newton-Raphson method is used, compute estimate
predictor displacement vector xP by solving K TXP =
P and corrector vector by solving KTX c = -g, where
g is the unbalanced load vector.
2. If the modified Newton-Raphson method is used com-
pute only corrector by solving KTX c = -g.
Calculate coefficients of quadratic equation, namely al 0[+1 +

a20i+l +a3 = 0, where al = (xf)txf, a2 = 2 [~Ui + xilt xf,
a3 = [2~Ui + xilt xi.
Solve quadratic equation.
Update A, U as follows: Ai+1 = Ai + OAi+1, ~Ui+1 =
~Ui + OAi+1 X f + ui
1. Update global displacements Ui+1 = Ui + OAi+1 X f + xi.
2. Update geometry.
3. For every element, compute the natural displacement incre-
ment PNI1 = 1/2aN [T06b + T06el r 11, where r 11 = OAi+1 X f +
xi. T 06b is the matrix of direction cosines at the beginning
of the iteration, and T06e the matrix of direction cosines at
the end.
4. Compute the natural force increment: PNI1 = KNPNI1.
5. Superpose the natural modes and forces as follows:
i i+l i
piJl =
PN+PNI1,P N =P N +PNI1.
6. Convert the natural forces to global cartesian forces using:
P i+l= aNt pi+l Tt -t pi+1
N = 06e a N N .
7. Update load Pi+1 = Ai+lP.
8. Compute residual forces gi+1 = P i+1 - Fi+1
9. Perform convergence check II gi+1 11/ II Pi+l II::; E, where E
is an input tolerance number.
10. If convergence is achieved
* Write output information.
* Create visualization data.
* Decompose the total natural energy into its invariant
measures, namely the axial straining energy, the sym-
metrical bending energy, the antisymmetrical bending
energy, the antisymmetrical transverse shearing energy,
and the natural energy associated with the azimuth ro-
tations. Express all energies on a percentage basis.
* Go to next increment.
11. If convergence fails proceed to the next iteration.
- Next iteration ite
Next increment inc
The complete procedure has been programmed into our finite element pro-
gram SANI (Structural Analysis and Information).
8.3 Numerical examples

8.3.1 Isotropic cylindrical panels
We now validate our algorithm with computational experiments. Figures

8.1, 8.3 show two isotropic shallow cylindrical panels along with all ge-
ometrical and material data. The right and left edges are hinged while
the top and bottom edges are free. Both panels are subjected to a cen-
tral concentrated load. The first shell has a radius/thickness ratio of 200.
For the second this ratio is 400. For both structures, references solution
were given by Crisfield [83]. Due to symmetry, one quarter of the panel
is discretized with 9 nodes in the x direction and 9 nodes in the y direc-
tion. The modified Newton-Raphson method is used for both paradigms.
Central load-displacement curves for the first shell are shown in Fig. 8.1.
The agreement with the results of [83] is excellent. In Fig. 8.2 we show
the variation of the separate natural strain energy components with iter-
ations (note that we have used the term iteration instead of increment).
We observe that the panel has insignificant antisymmetrical bending and
transverse shearing energies while the natural symmetrical bending and
axial energies dominate. At the end, the natural axial straining energy
diminishes to zero and the panel is in a state of pure bending.
Turn now to Fig. 8.3. We show the comparison of our method for
the thinner cylindrical panel with the results of [83], and results obtained
by a 8-node serendipity isoparametric shell element [85]. The triangular
mesh (shown in Fig. 8.5) comprises of 8 x 8 x 2 = 128 elements with 407
total unknowns, while the mesh used for the isoparametric shell element
is 5 x 5 resulting in 394 total unknowns. The load-displacement curve of
the triangular element TRIC is in very good agreement with the curve of
Crisfield, while the isoparametric shell element agrees fairly well until the
first snap-back is encountered. After this point, the isoparametric degener-
ated element follows a different path from the other two elements. Central
load and displacement curves at the various increments are given on the
top of Fig. 8.4. On the bottom of the same figure we show the decompo-
sition of the natural strain energies into their invariant components. It is
observed that when snap-back occurs, an interchange of the natural axial
straining and symmetrical bending energies take place. Then, at the next
stable region, a second interchange occurs and as the end of the incremen-
tal procedure is reached the panel is in a state of pure bending. Figures
8.6-8.9 show the deformation of the cylindrical isotropic panel at various
load levels during nonlinear deformation.
8.3 Numerical examples 275
Pc 9 x 9 nodes in quarter (TRIC)

free
E= 3105 N/mm2
v=O.3
L=504mm
R=2540mm
t= 12.7mm
9 = 5.6773 0
Rlt= 200
TRIC (triangular element)

o Crisfield
Pc (N)
2500
2000
1500
1000
500
wc(mm)
o
o 5 10 15 20 25 30
Figure 8.1: Hinged isotropic shell; comparison with the reference solution
Pc (N)
2500
2000
1500
1000
500
0
1/ iteration NO
0 5 10 15 20 25
wc(mm)
30
25
20
15
10
0
0 5 10 15 20 25
natural energy (%)

100
axial strain energy
90
symmetrical bending energy
80 antisymmetrical bending energy
anti symmetrical transverse shearing energy
70
60
50
40
30
20
10
iteration NO
o --:--~
o 5 10 15 20 25
Figure 8.2: Decomposition of natural straining energies

P
c 9 x 9 nodes in quarter (TRIC)
free
E= 3105 N/mm2
v=0.3
L=504mm
R=2540mm
t= 6.35 mm
9= 5.6773 0
Rlt=400
TRIC (triangular element)

o Crisfield
Pc(N) x 8-node degenerated shell el.
700
500
300
100
-100
-300
wc(mm)
-500
o 5 10 15 20 25 30
Figure 8.3: Thin hinged shell; comparison with reference solutions

PeeN)
600
400
200
-200
P 4 8
-400
30 r We (mm)
25
20
15
10
5
iteration NO
4 8 12 16 20 24 28 32
natural energy (%)1-,_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _--,
100 axial strain energy

....
antisymmetrical transverse shearing energy
80
70
60
". ............
50
40
30
20
10
4
- 8 12 16 20 24 28 32
Figure 8.4: Decomposition of natural straining energies for the thinner

panel
C'C1lD t'lj
"lj
(;)
ro Oti
S .:
~
C'C1lD "'1
"1
~ C'C1lD
...-
CI.l
00
' - ' 00
.
...- 01
g"t'lj
t:r~
'v..... 71
;7/ ~ 71 0 /7 ,,.,,, /7 .~~ 71
.-._. tlU
IlU 81
I/ A~ 11a :7J 1?n :?J :.?7 A
-o S
..... .. "" 77
Il' ...-
.ft. "" '"'"
p..C1l
p..C'D
CI.l _
.....
00 C'C1l
_D
C1l
C'D
.g
~ S
'"C:j
"i::j C1l
C'D
00
00
e:::a
:::-:a
C1l
C'D !;.Ij
w
p..S
o C1l
C'D
Z
~ CI.l
00
.:
~
::r' S
S
~ C'C1lD
o 0 "1
"'1
p..-,
p......., .....
(")
(")
C1l
C'D .: ~ Il'
....... '"C:j
f-'''i::j
'"C:j
"i::j CC'1lD
-e.
C1l
C'D
><
><
"'1
"1
Il'
Il'
"'1 S
S
"1 "i::j
..... '"C:j
()'q
C'D
C1l
00
CI.l
~
-
..c
.0
~
.:
Il'
"'1
"1
...-
C1l
C'D
"'1
"1
..-.,
..-.,
00
.......
f-'
~
0
5
p..
p..
C1l
C'D
~CI.l
_00
.......
f-'
t-;)
t-;)
00
00 I~
I~
Figure 8.6: Deformation of upper right quarter at P = 550.5480 N (Top

edge hinged; left edge free; load applied at bottom right node)
Figure 8.7: Deformation at P = 260.4401 N following snap-through

Figure 8.8: Deformation at P = -327.2101 N following snap-back

Figure 8.9: Deformation at P = -239.0802 N

To show the computational advantages of our algorithm we provide

now the CPU times for the natural and isoparametric finite elements (we
remind the reader that both meshes included the same number of degrees
of freedom). The computations were conducted on a SUN Station 10 using
standard FORTRAN without issuing any special optimization directives.
In these lists we show' four columns, namely the number of increments,
the iterations per increment, the time for the formulation of the tangent
stiffness matrix at every increment, and the time required for convergence
in the individual increments. It is evident from the lists that our element
TRIC is the clear winner. It completes the computation 6 times faster
than the isoparametric shell element; is 14 times faster in computing the
tangent stiffness matrix; and achieves convergence 6 times faster during all
increments. These lists demonstrate the great advantages of the natural
mode finite element method and the accuracy of the element TRIC. We
comment that further reduction of the computational time is possible for
TRIC. One may attempt to use vectorization and/or parallelization, or
revert to special solution techniques for the solution of the linear system of
equations. On the contrary, little can be done for the isoparametric shell
element since the formation of its tangent stiffness is expensive and tedious
and multiple quadrature loops cannot be avoided.
*************************************************************
TRIANGULAR COMPOSITE ELEMENT T RIC
*************************************************************
Increment Iteration Time/Tangent Stiff. Time/increment
1 4 1.39 4.24
2 4 1.38 4.26
3 4 1.39 4.28
4 4 1.40 4.28
5 4 1.40 4.26
6 3 1.41 3.69
7 3 1.38 3.65
8 3 1.40 3.66
9 3 1.39 3.65
10 3 1.39 3.65
11 4 1.39 4.24
12 4 1.39 4.23
13 6 1.40 5.43
14 6 1.39 5.45
15 6 1.39 5.46
16 6 1.39 5.42
17 6 1.39 5.41
18 6 1.40 5.44
19 8 1.38 6.59
20 6 1.39 5.43
21 8 1.40 6.69
22 8 1.41 6.68
23 10 1.39 7.78
24 12 1.39 8.98
25 11 1.39 8.35
26 47 1.39 29.62
27 28 1.39 18.39
28 9 1.39 7.18
29 5 1.39 4.82
30 4 1.38 4.25
31 5 1.39 4.84
32 7 1.39 6.02
33 4 1.39 4.26
34 3 1.40 3.71
TOTAL TIME 47.32 214.29
TOTAL CPU TIME (sec): 222.97

*************************************************************
=============================================================
8-NODE ISOPARAMETRIC SHELL
=============================================================
Increment Iteration Time/Tangent Stiff. Time/increment
1 5 12.54 29.35
2 5 19.81 36.20
3 4 19.82 33.28
4 4 19.79 33.29
5 4 19.78 33.25
6 4 19.79 33.25
7 4 19.82 33.30
8 4 19.79 33.27
9 4 19.82 33.28
10 4 19.88 33.39
11 4 20.01 33.58
12 6 19.95 39.35
13 7 19.81 42.10
14 7 19.85 42.11
15 7 19.87 42.17
16 6 19.80 39.09
17 7 19.89 42.18
18 8 19.90 45.26
19 8 19.86 45.04
20 8 20.21 46.47
21 8 19.96 45.20
22 9 19.82 47.92
23 9 19.83 47.94
24 8 19.97 45.25
25 8 19.87 45.63
26 8 19.81 44.99
27 8 19.87 45.47
28 8 20.43 46.24
29 6 19.84 39.19
30 4 19.92 33.54
31 4 19.99 33.57
32 3 19.99 30.61
33 3 19.90 30.51
34 2 19.87 27.55
TOTAL TIME 669.04 1312.82
TOTAL CPU TIME (sec): 1325.84

=============================================================
8.3.2 Composite shells

Next we consider laminated composite shells. The shallow shell shown in
Fig. 8.3 is considered again but with four different laminations. These lami-
nations include 48 layers and are: (06/906/06/906)s, (452/ -452/02/902hs,
(452/ - 452)6s, (45/ - 45/02/902/60/ - 30hs. Our purpose is to com-
pute the nonlinear response of the composite panels and to asses the im-
pact of the fiber orientation. The Newton-Raphson method is used. The
(06/906/06/906)s panel is shown in Fig. 8.10 with all its geometrical and
material properties. Due to biaxial symmetry, only one quarter of the panel
is considered and discretized with a set of triangular multilayer elements
TRIC. The central load-displacement curve is shown in Fig. 8.10, while
the increments for convergence at each step, the natural energies and the
v - w curve for point A are shown in Fig. 8.11. The response of the cross-
ply shell shows some similarities with the thin isotropic panel of Fig. 8.3,
however a much higher load is needed before snap-through occurs. We ob-
serve also the intricate interchange of the natural symmetrical bending and
axial straining energies during snap-back. This seems to be a very inter-
esting phenomenon which calls for a further investigation. The nonlinear
responses of the other three composite panels are shown in Fig. 8.12, while
the decomposition of the total natural energy into its invariant components
is shown in Fig. 8.13. The effect of fiber orientation is evident from the
plots. Indeed, much more complicated instability phenomena arise in com-
posite panels. This may be attribute to anisotropy. We discern multiple
snap-through and snap-back paths with interchanges of the natural ener-
gies according to various patterns shown in Fig. 8.13. It appears that as
we are heading for snap-through, the natural symmetrical bending which
dominates the initial response becomes smaller, while the natural axial
straining energy increases. At snap-back, an interchange of the two nat-
ural energies takes place while during following stable or weakly unstable
regions the natural symmetrical bending energy dominates again. This ap-
peared to be the case in all examples considered here. We stress that these
are only preliminary observations and that much more research is needed
before more definitive statements can be made. The deformation of the
{452/ - 452)68 panel at various load levels is illustrated in Figs. 8.14-8.17.
The plots reveal a double snap-through and snap-back.
We conclude the numerical examples with a postbuckling analysis of
a 4-layer (45/ - 45/ - 45/45) laminated composite cylinder shown in Fig.
8.18. One edge of the cylinder is clamped while the free edge is subjected
to a compressive force P which is distributed to the nodal points. An initial
imperfection in the form of the first buckling mode scaled down to 5% of
the shell thickness is given before the nonlinear analysis is initiated. For
points A, B, we provide the load-displacement curves in Figs. 8.18, 8.19.
Deformation plots at two distinct loads are shown in Figs. in Figs. 8.20,
8.21. We observe that following buckling, the composite cylinder reaches
a value after which only local edge movements occur while the cylinder
body close to the clamped support seems to be relatively unaffected. From
this numerical example we deduce the intricate load and energy absorption
mechanisms of composite cylinders that makes them attractive for crash
prone vehicles.
Pe 9 x 9 nodes in quarter (TRIC)

free
EI = 150GPa
E2 = 10 GPa
L=504mm GI2 =6GPa=G I3
R=2540mm G23 =4GPa
h =6.48mm v 12 =v 13 =v23 =0.25
8 = 5.6773 0 a.
I
=2.5 x 10-8 C-I
Rlh=400 a. = 30 x 10-6 C-I
2
hL =1.35 X 10-4 m
(0/90/0/906)s
PeeN)
15000
10000
5000

-5000
We (mm)
-10000
5 10 15 20 25 30
Figure 8.10: Nonlinear deformation of a cross-ply (06/906/06/906)8 com-

posite panel
12 r increments for convergence
10
2
iteration NO
o
o 5 10 15 20 25 30 35 40 45 50 55 60
natural energy (%)

100
axial strain energy

90
70
60
50
40
30
20
10
o
--
... -...
o 5 10 15 20 25 30 35 40 45 50 55 60
o r wA
-0.5
-I
-1.5
-2
-2.5
-3
-3.5
-4
-4.5
==r: ,vA
-5
-0.012 -0.01 -0.008 -0.006 -0.004 -0.002 o 0.002 0.004
Figure 8.11: Decomposition of natural straining energies for the

(06/906/06/906)8 composite panel
12000 f Pc (N)
10000
8000
6000
4000
2000
0
2000 f I (45/-45/0/90 2)3S

-4000
-6000 ~ ~ wc(mm)
-8000
0 5 10 15 20 25 30
8000 fPc (N)
6000
4000
2000
-2000 r
(45/-45 2)6S
,
\. /'
wc(mm)
-4000 ' . I ,
0 5 10 15 20 25 30
12000 f Pc (N)
10000
8000
6000
4000
2000
=t
0
(451-4510/90/60/-30)3S
~ wc(mm)
-6000
0 2 4 6 8 10 12 14 16 18 20
Figure 8.12: Load-displacement curves for three laminated composite pan-

els
natural energy (%)

axial strain energy
100
.: ......
90
antisymmetrical bending energy
80
70
60
50
40
30
20 t I (45l4VO/ 90 2)3S
lO
0
0 20 40 60 80 100 120 140 160 180
100
.'
90
80
70
60
50 ..... ' .. / ........... .. - .... _-" ... -.... -.
40
II " ....... -
30
20 WI (45/-45 2)6S
lO
0
0 20 40 60 80 100 120 140
80
70
60 t. .---. .....
:"
# - ..
- ..........
50
40
30
20
lO ~
iteration NO
I === I 1 _I
20 40 60 80 100 120 140 160 180 200 220 240
Figure 8.13: Decomposition of the strain energies for the three composite
panels
Figure 8.14: Deformation of the (452/ - 452hs panel at P = 6382.40 N

Figure 8.15: Deformation of the (452/ - 452hs panel at P = -2007.52 N

Figure 8.16: Deformation of the (452/ - 452 hs panel at P = 4262.65 N

Figure 8.17: Deformation of the (452/ - 452)35 panel at P = -2758.91 N

30000 total edge load (Pa)
25000
20000 (451-451-45/45)
p
15000
R =0.250m
10000
a =1.40m
5000
0 uA (m)
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003
30000
25000
20000
15000
10000
5000
vim)
01/
0 0.0005 0.001 0.0015 0.002 0.0025
30000
25000
20000
15000
10000
5000
II
wim)
0
0 0.0005 0.001 0.0015 0.002 0.0025
Figure 8.18: Load-displacement curves for point A

total edge load (Pa)

30000
25000
20000
15000
10000
5000
uim)
0
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003
30000
25000
20000
15000
10000
5000
0
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012
30000
25000
20000
15000
10000
5000
0
-0.00054 -0.00045 -0.00036 -0.00027 -0.00018 -0.00009 0
Figure 8.19: Load-displacement curves for point B

0')
0') ~
C"I
I-<
Q.l
-
96~WOO'
"0
......
$:l
x~ :>.
--
6ZOOO' u
~
68SOOO' u:I
"<t'
I
S8L~' -.....
u:I
"<t'
-
'-'
Z86S000'
<n
Q.l ]...,
~
S
8L~LOOO'
.....
0
ce
-
~ vL8000' $:l
Q.l
...,
.9
ce ce
......U
I-<
~LS6000'
S
I-<
S
Q.l
LLO~OO'
.2Q.l
;:; "0
Z "0
96~ ~OO' Q.l
M
q::l
00 ......
$:l
9~LOO' bO
ce
S
9vLOO'
~
......
...,
......
SSSLOO' $:l
~ Z
M
SL9LOO' 0 0')
C"I
t:-..=
00 0
S6LLOO' Q.lO
Zv:OS:v L L6-lnr-8 L S'Z NVI:IJ.V d V"L -1:1 NIfI:l.lV d10Sv/ ~. ~-S31:1 L' Z-Ol .lOld NOUVV'jI:lO'.::I3'OI I-<M
ZV:OS:V~ L6-lnr-8~ S'Z NlfI:l.lVd V"L-I:j NIfI:l.lVd/OSv/ (OVV'j-03J\)L ' L-$31:1 L' Z-Ol .lOld 30N
FR
;:;
"
bO
1:0..
Z
Figure 8.21: Deformation of the {45/ - 45)8 cylinder at P = 27909 N

0)
90&1>000' 0
.. 98000'
x~ 0)
t-
c-:I
1\
~
-
l6l ~OO'
00 ii3
Nonlinear analysis of anisotropic shells
lU~OO '
'"'
-....
Q) Q)
...r::: "'0
00
.::
.... Co)
0..
ESLlOO'

Co)
0
.....'"' E9SlQO'
..........
tI)
....
.::ce
0
00
I>LOEOO'
l!";)
-.:::I'
I
...... ...........
0
....
I>'I>I>EOO' l!";)
-.:::I'
00 ..........
-
00
Q)
SLSEOO' ...r:::
ce .....
.::
ce ......
9OC1>OO' 0
~
.....::
-....
Q) 0
.:: 9ELI>OO'
ii3
0
.:: S
Z L9~SOO'
'"'
.
Q)
L6SSOO' 0
..-i
c-:I
9l09OO'
00
Q)
9SI>9OO' '"'
....be
;::l
DEFORMAnON PLOT
9O:0S:I>~ L6-lnr-8~ S'l NV\:ilVd I>' L-I:I NVI:I1Vd/OSr-l L'L=S3I:1l'l=01 lOld
FRINGE
0 9O:0S:I>L L6-lnr-8L S'G NV\:ilVd n-I:I NVI:I1Vd/OSV\I (~VV\I-03"h ' L=S3I:1l' l=01 lOld
300
~
0
M
Chapter 9
Programming aspects
In this chapter we introduce some programming aspects of our method and

explane key elements of the computer program included with the book.
We will describe the general structure of the computer program which is
based on generic finite element programming concepts. We stress that the
basic code structure is similar for beam and shell elements, or for any other
finite element type per se. Only key parts of the code will be highlighted
here. We believe that the reader, using basic knowledge of Fortran 77 [86]
and having understood the principles of our method can easily follow the
structure of the computer program.
9.1 Memory allocation

Every routine of the computer program reads the short file parcb.h in
which the user specifies the maximum size of his problem. Thus, if the
maximum number on nodes needs to be increased, this is done only once in
this file and not repeated on each separate routine. Note that these param-
eters depend on both the discretization of the problem and the available
memory on the user's computer. The file parcb.h is
* paracb.h
integer znn,zne,znde,zndf,zu,zf
integer zmaxa,zlb,znl,znne,znmt,znumna,
znnped2,znna,znumk,
zndh,znumh,zheat
parameter(znn=500,zne=1000,znne=2,zndf=6,
znmt=zne,znde=zne,znnped2=znne*zndf,znna=znnped2-6,

302 Programming aspects
znumk=znnped2*znnped2,znumna=znna*znna,
zndh=2,znumh=znne*zndh,zheat=znumh*znumh,
zu=1000000,zf=znn*zndf,znl=30,
zmaxa=zf+1,zlb=zf)
The most important variables set here are
znn: maximum number of nodes (here 500)
zne: maximum number of elements (here 1000)
znne: maximum number of node per element (2 for the beam element)
znne: maximum number of of degrees offreedom per node (6 for both

beam and shell elements)
zu: maximum number of elements in skyline (set here equal to 1000000)

znl: maximum number of layers (set here equal to 30)
The user must modify these parameters according to his problem and
available core memory. For example, he can set the maximum number of
nodes equal to 200 (znn=200).
9.2 Input data

First we shall select a model problem and explain all input parameters.
We choose the problem of Fig. 5.20 concerned with the static analysis of
a cantilever composite beam. For this particular example, the basic input
data is read from file meshb.dat using the Fortran commands
c ..... Read NO nodes,NO elements,NO unknowns,No mater. type

read(12,*) nn,ne,nu,nmt
c ..... Read Global Coordinates
do ip=1,nn
read(12,*) idum,xn(ip),yn(ip),zn(ip)
enddo
c ..... Read Index Table
do ip=1,nn
read(12,*) idum,(indx(ip,j),j=1,6)
enddo
c ..... Element, Element ID, nne, distinct el.,material type,
9.2 Input data 303
c extra point coord., width

do it=l,ne
read(12,*) idum,idel(it),noel(it),iesmno(it),iemat(it),
exx(it),exy(it),exz(it),bw(it)
enddo
c ..... Element, Element id, Connectivity, Temperatures
do iel=l,ne
read(12,*) idum,idel(iel),(node(iel,k),k=1,2),
(gtemp(iel,j),j=1,4)
enddo
c ... Read material data
do imt=l,nmt
read(12,*) nl(imt)
read(12,*) (theta(i,imt),th12(i,imt),hl(i,imt),
youngl(i,imt),young2(i,imt),young3(i,imt),
g12(i,imt),g13(i,imt),g23(i,imt),
poi12(i,imt),poi13(i,imt),poi23(i,imt),
tal(i,imt),ta2(i,imt),fal(i,imt),fa2(i,imt),
iact(i,imt),piezdl(i,imt),piezd2(i,imt),
cden(i,imt),i=l,nl(imt
enddo
c ... Read loading data
read(12,*) ipress,press
do i=l,nn
read(12,*) ido,(bload(i,j),j=1,6),(buvw(i,j),j=1,6)
enddo
c ... Read options
read(12,*) ifunc,igeoms,ishear,iprint,ipost,ipatran,
iwhichel
The parameters used in the above Fortran module are:
nn: number of nodes
ne: number of elements
nu: number of unknowns
nmt: number of material types
idum: node counting index

xn(ip), yn(ip), zn(ip): x, y, z global coordinates of node ip
idum: node counting index
indx(node,dof): global degree of freedom for node. The parameter

dof ranges from 1 to 6. If the global degrees of freedom for this node
are suppressed, then indx{node,dof}=O
idum: element counting index
idel{it): identification number for element it. For the beam element,
idel{it} =2
noel(it): number of nodes per element. For the beam element,

noel{it} =2
iesmno(it): distinct number for element it
iemat(it): material type for element it
exx(it), exy(it), exz(it): x, y, z global coordinates of an extra

point fixing the orientation of the middle surface (see Appendix A)
bw(it): width of element it
idum: element counting index
idel(iel): identification number for element iel. For the beam element
idel{iel} =2
node(iel,k): connectivity array for element iel with nodes k=l and
k=2. If node{iel,1}=2, then the first node of element iel connects to
global node 2. The same convention holds for the second node
gtemp(ieIJ): top and bottom nodal temperatures for element iel.

The convention used is as follows:
gtemp{iel,l}: temperature at top of first node for beam iel
gtemp{iel,2}: temperature at bottom of first node for beam iel
gtemp{iel,3}: temperature at top of second node for beam iel
gtemp{iel,4}: temperature at bottom of second node for beam iel
nl{imt): number of layers for material type imt

9.2 Input data 305
theta(i,imt): fiber angle for layer i and material type imt formed by
the local cartesian axis x' and fiber angle 1
th12(i,imt): always equal to 90.0

hl(i,imt): thickness of layer i for material type imt
youngl(i,imt): Young's modulus En (along fiber direction)
young2(i,imt): Young's modulus E22 (perpendicular to fiber direc-

tion)
young3(i,imt): Young's modulus E33
g12(i,imt): in-plane shear modulus G12
g13(i,imt): transverse shear modulus G 13

g23(i,imt): transverse shear modulus G23
poi12(i,imt): in-plane Poisson's ratio 1I12
poi13(i,imt): out of plane Poisson's ratio 1I13
poi23(i,imt): out of plane Poisson's ratio 1I23
tal(i,imt): thermoelastic coefficient am for layer i and material

type imt
ta2(i,imt): thermoelastic coefficient at22 for layer i and material

type imt
fal(i,imt): always equall.

fa2(i,imt): always equall.
iact(i,imt): always equal O.
piezdl(i,imt): always equall.
piezd2(i,imt): always equall.

cden(i,imt): density of layer i
ipress: 'unifp' for uniform pressure on all beam elements.
press: Pressure magnitude with sign opposite to all local z' axes.
ido: node counting index

bload{iJ): applied global loads at node i. Set as follows:
bload{i,l}: Fx force at node i
bload{i,2}: Fy force at node i
bload{i,3}: Fz force at node i
bload{i,4}: M(J moment at node i
bload{i,5}: MIjJ moment at node i
bload{i, 6}: M'Ij! moment at node i
buvw{iJ): applied global displacements and rotations at node i. Set

as follows:
buvw{i,l}: x displacement at node i
buvw{i,2}: y displacement at node i
buvw{i, 3}: z displacement at node i
buvw{i,4}: () rotation at node i
buvw{i, 5}: rotation at node i
buvw{i, 6}: 'lj; rotation at node i
ifunc: type of analysis desired

ifunc= 'static': static analysis
ifunc='buck': buckling analysis
igeoms: type of geometrical stiffness

igeoms = 'simpgs': simplified geometrical stiffness
igeoms= 'psimpgs': partly simplified geometrical stiffness
ishear: shear deformation option

ishear=.true.: shear deformation included
ishear=.false.: Kirchhoff solution
iprint: printout
iprint=. true.: extensive printout
iprint=.Jalse.: limited printout
The output information is written on file outb.dat
9.2 Input data 307
ipost
ipost=.true.: postprocessing is conducted with the computation of
forces, moments, natural energies for all elements and the estimation
of the through the thickness stresses for a selected element
All this information is written on output file nmq.dat
ipatran
ipatran=.true.: a neutral interface to Patran [87] is provided. The
finite element mesh is written on file beam.dat, while the nodal
displacements on file beam.dis
iwhichel: element for which the through the thickness stresses are
required
The input data for the problem of Fig. 5.20 comprising 6 nodes, 5 beam
elements and 30 unknowns is given as
6, 5, 30, 1
1,0.,0., O.
2,5.,0., O.
3,10.,0., O.
4, 15., 0., O.
5, 20., 0., O.
6, 25., 0., O.
1, 0, 0, 0, 0, 0, 0
2, 1, 2, 3, 4, 5, 6
3,7,8,9, 10, 11, 12
4, 13, 14, 15, 16, 17, 18
5, 19, 20, 21, 22, 23, 24
6, 25, 26, 27, 28, 29, 30
1, 2, 2, 1, 1, 2.5, 10., 0., 5.
2, 2, 2, 2, 1, 7.5, 10., 0., 5.
3, 2, 2, 3, 1, 12.5, 10., 0., 5.
4, 2, 2, 4, 1, 17.5, 10.,0., 5.
5, 2, 2, 5, 1, 22.5, 10., 0., 5.
1, 2, 1,2,0.,0.,0., O.
2,2,2,3,0.,0.,0., O.
3,2,3,4,0.,0.,0., O.
4,2,4,5,0.,0.,0., O.
5,2,5,6,0.,0.,0., O.
4
45.,90. ,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
-45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
-45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
'unifp' , -0 .
1, O. , O. , 0., 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
2, O. , O. , O. , O. , O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
3, O. , O. , O. , 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
4, O. , O. , 0.,0.,0.,0.,0.,0.,0.,0.,0.,0.
5, O. , O. , 0., 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
6, 0.1, 0.1, -0.1, -0.1, O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
'static','isimpgs',.true.,true.,.true.,tr~e.,l
Following the reading of the input data, the routine constructs the
global load vector, computes the total thickness for each material type as
well as the z coordinates for all plies, estimates the length of each element,
and depending on user requests prints out all parameters and prepares the
neutral Patran file. Some of these tasks are accomplished with the Fortran
modules
c ... Construct load vector

do 11 i=l,un
do 12 j=1,6
ix=indxCi,j)
if(ix.eq.O) goto 12
v(ix)=bload(i,j)
12 enddo
11 enddo
c ..... TOTAL SHELL THICKNESS
do imt=l,nmt
thic(imt)=O.
enddo
do imt=l,nmt
do j=l,nl(imt)
thic(imt)=thic(imt)+hl(j,imt)
9.3 Program parameters and storage 309
enddo
enddo
c ..... CONVERT DEGREES TO RADS
do imt=l,nmt
do j=l,nl(imt)
theta(j,imt)=theta(j,imt)*acos(-1.)/180.
th12(j,imt) =th12(j,imt)*acos(-1.)/180.
enddo
enddo
c ..... Z-COORDINATES
do imt=l,nmt
do i=O,nl(imt)
if(i.eq.O) hz(i+l,imt)=thic(imt)/2.
if(i.ne.O) hz(i+l,imt)=hz(i,imt)-hl(i,imt)
enddo
enddo
c ... Compute beam length
do k=l,ne
xl=xn(node(k,l
yl=yn(node(k,l
zl=zn(node(k,l
x2=xn (node (k, 2
y2=yn(node(k,2
z2=zn(node(k,2
bl(k)=sqrtx2-xl)**2+(y2-yl)**2+(z2-z1)**2)
enddo
9.3 Program parameters and storage

After all input data is read, the program issues a call to routines setupb
and findsky in order to define some problem parameters and reserve the
storage places for the global arrays. . Important parameters in routine
setupb are
nne: number of nodes per element
ndf: number of cartesian degrees of freedom per node
nna: number of natural degrees of freedom for element
numk: total entries in local and global elemental stiffness matrices

numna: total entries in natural stiffness matrix
Routine findsky defines the skyline storage, the positions of the diag-
onal terms and computes the total number of elements under the skyline.
For more information on this storage scheme see the text of Bathe [79].
The storage scheme is declared in the following loops
c ..... SKYLINE STORAGE

do i=l,nu
IbCi)=O
enddo
do 4550 it=l,ne
do 4560 in=l,noel(it)
ip=node Cit, in)
do 4570 iu=l,ndf
i=indxCip, iu)
if(i.eq.O) goto 4570
do 4580 jn=l,noel(it)
ipp=node Cit, jn)
do 4590 iuk=l,ndf
j=indx Cipp, iuk)
if(j.eq.O) goto 4590
ij=j-i
if(ij.gt.lb(j Ib(j)=ij
4590 continue
4580 continue
4570 continue
4560 continue
4550 continue
c ..... ADDRESSES OF DIAGONAL TERMS
maxa(1)=l
do 4600 j=2,nu
4600 maxa(j)=maxa(j-l)+lb(j-l)+l
c ..... NUMBER OF ELEMENTS UNDER SKYLINE
nvk=maxa(nu)+lb(nu)
nvm=nvk
maxa(nu+l)=nvk+l
9.4 Elemental elastic stiffness matrix 311
9.4 Elemental elastic stiffness matrix

Routine beamstif computes the natural, local and global elemental stiff-
ness matrices. It begins by computing the direction cosines for the beam
element as well as its area O. It then proceeds to compute the transformed
stiffnesses for each ply. The later are stored into the array storec(i,3,3)
for every layer i. For example storec(i,l,l)=Exx , storec(i, 2, 2)=Eyy , and
storec(i,3,3)=G xy for layer i. The transformed transverse shear coefficients
are stored in array shear(i, 2, 2), and the transformed thermoelastic coef-
ficients in array stherm (i, 2, 2). Following these transformations, piecewise
summations are invoked in order to integrate these coefficients through the
thickness. This is accomplished with the loop
c----------------------------------------------------------
c Piecewise integration through the layers.
c----------------------------------------------------------
do i=l,nl(imt)
rknll=rknll+storec(i,l,l)*(hz(i,imt)-hz(i+l,imt))
rkn12=rkn12+storec(i,1,1)*(1./2.)*(hz(i,imt)**2-hz(i+l,imt)**2)
rkn22=rkn22+storec(i,1,1)*(1./3.)*(hz(i,imt)**3-hz(i+l,imt)**3)
rkn44=rkn44+storec(i,1,1)*(1./12.)*(hz(i,imt)-hz(i+l,imt))
rkn55=rkn55+storec(i,1,1)*(1./4.)*(hz(i,imt)-hz(i+l,imt))
rkn661=rkn661+storec(i,3,3)*(1./3.)*(hz(i,imt)**3-hz(i+1,imt)**3)
rkn662=rkn662+shear(i,1,1)*(1./12.)*(hz(i,imt)-hz(i+l,imt))
rlp=rlp+shear(i,l,l)*(hz(i,imt)-hz(i+l,imt))
enddo
Following the above summations, the shear correction factor is com-

puted, and once this is accomplished the beam's natural stiffness matrix is
written down via
c----------------------------------------------------------
c The beams natural stiffness matrix
c----------------------------------------------------------
snb (1 , 1) =rkn11
snb(1,2)=rkn12
snb (1 ,3) =0 .
snb(1 ,4)=0.
snb (1 , 5) =0 .
snb (1 ,6) =0 .
snb(2,2)=rkn22
snb(2,3)=O.
snb(2,4)=O.
snb(2,5)=O.
snb(2,6)=O.
snb(3,3)=rkn33
snb(3,4)=O.
snb(3,5)=O.
snb(3,6)=O.
snb(4,4)=rkn44
snb(4,5)=O.
snb(4,6)=O.
snb(5,5)=rkn55
snb(5,6)=O.
snb(6,6)=rkn661+rkn662
The natural stiffness matrix is then transformed to the local coordinate

using matrix aN
an (1 , 1)=-1.
an (1 ,7)=1.
an(2,5)=-1.
an(2,11)=1.
an(3,3)=2./bl(id)
an(3,5)=-1.
an(3,9)=-2./bl(id)
an(3,11)=-1.
an(4,6)=1.
an(4,12)=-1.
an(5,2)=2./bl(id)
an(5,6)=1.
an(5,8)=-2./bl(id)
an(5,12)=1.
an(6,4)=-1.
an(6,10)=1.
In the above array, bl (id) represents the length of element id. A series of
matrix multilplications then yield the local and global elemental cartesian
stiffnesses. The same computing procedures are also followed in setting
up the natural thermal load and transforming it to the local and global
9.5 Assembly of the global stiffness matrices 313
coordinates. Then, the global thermal load vector is assembled into the
global loading vector via
c
c ASSEMBLE THE THERMAL LOAD VECTOR
c
do 2040 in=l,nne
ip=node {id, in)
do 2060 iu=l,ndf
i=indx{ip,iu)
if(i.eq.O) goto 2060
v(i)=v(i)+sttin-l)*ndf+iu)
2060 continue
2040 continue
At this point, the global load vector including the applied loads and
temperatures is defined and stored into the one dimensional array v{nu),
where nu denotes the total number of unknowns.
9.5 Assembly of the global stiffness matrices

Routine matassbe performs the assembly procedure for the global elemen-
tal cartesian stiffness matrices. In addition, it imposes on the right hand
side loading vector any given applied displacements or rotations. The main
assembly loops are
c ..... M A I N ASS E M B L Y L 0 0 P
do 140 in = 1,noel(id)
ip = node(id,in)
do 160 iu = 1,ndf
i = indx(ip,iu)
if(i.eq.O) goto 160
do 170 inn = 1,noel(id)
ipp = node(id,inn)
do 180 ju = 1,ndf
j = indx(ipp,ju)
if(i.gt.j) goto 180
ij=maxa(j)+(j-i)
a(ij)=a(ij)+k(iu,in,ju,inn)
180 continue
170 continue
160 continue
140 continue
All elemental cartesian stiffnesses are assembled into the one dimen-
sional array a(nwk), where nwk is the total number of entries in the sky-
line.
9.6 Solution of the linear equations

Since we have defined and assembled the elemental stiffness matrices and
global load vector we can proceed to the solution of the linear system of
equations that describe the structural equilibrium. This is accomplished by
using the Fortran routine coisol of [79]. But before solving the system we
thought appropriate to check the eigenvalue of the global stiffness matrix
for any indications of ill conditioning. This is accomplished by using routine
eigen which computes the first eigenvalue of the global stiffness matrix. In
case this eigenvalue returns a zero or negative eigenvalue the reader must
carefully examine the accuracy and meaningfulness of his input data and
make sure that he has set up a well posed physical problem. Following so-
lution of the equilibrium equations, the code prints out the obtained global
unknowns and prepares the Patran output files for graphical visualization
of the deformation results.
9.7 Postprocessing
Following computation of the global nodal unknowns, and if desired by
the user, the program proceeds to postprocessing of the results by calling
routine bepost. The routine first computes the generalized natural forces
and moments and subsequently estimates the cartesian stress resultants as
they refer the local elemental coordinate system. Then, it computes all
natural energies. These operations are performed for all elements via the
commands (follow also the comments before each task)
c-----------------------------------------------------------
c ..... COMPUTE R = a R ELEMENT NATURAL MODES
c N N
c 6x1 6x12 12x1
c-----------------------------------------------------------
call zerov(rhoen,6*1)
call mxv(dan,6,rhoel,12,rhoen)
c-----------------------------------------------------------
c COMPUTE P = k R ELEMENT NATURAL FORCES
c N NE N
c 6x1 6x6 6x1
c-----------------------------------------------------------
call zerov(fnat,6*1)
call mxv(dd,6,rhoen,6,fnat)
c-----------------------------------------------------------
c GET NATURAL ENERGIES 1/2 Pn rn
c--------------------------------------------------~--------
en1=(1./2.)*fnat(1)*rhoen(1)
en2=(1./2.)*fnat(2)*rhoen(2)
en3=(1./2.)*fnat(3)*rhoen(3)
en4=(1./2.)*fnat(4)*rhoen(4)
en5=(1./2.)*fnat(5)*rhoen(5)
en6=(1./2.)*fnat(6)*rhoen(6)
all=en1+en2+en3+en4+en5+en6
All elemental natural energies are also computed on a percentage basis.

Following the computation of the stress resultants, the routine bepost
computes the through the thickness stresses for the element ielwhich defined
by the user in routine meshb.
9.8 Geometrical stiffness

The simplified and partly simplified geometrical stiffnesses are computed
for every element using the subroutine pgeoms. The routine first extracts
the elemntal natural modes and forces and compute the natural geometrical
stiffness kNG as follows:
c-----------------------------------------------------------
c ..... COMPUTE R = a R ELEMENT NATURAL MODES
c N N
c 6x1 6x12 12x1
c-----------------------------------------------------------
call zerov(rhoen,6*1)
call mxv(dan,6,rhoel,12,rhoen)
c-----------------------------------------------------------
c COMPUTE P = k R ELEMENT NATURAL FORCES

c N NE N
c 6xl 6x6 6xl
c-----------------------------------------------------------
call zerov(fnat,6*1)
call zerov(pn,6*1)
call mxv(dd,6,rhoen,6,fnat)
call copyv(fnat,pn,6)
c-------------------------------------------
c k
c NG
c-------------------------------------------
call zerov(kng,6*6)
kng(2,2)=(bl(id)*pn(1/12.
kng(3,3)=(bl(id)*pn(1/20
kng(4,4)=(bl(id)*pn(1/12.
kng(5,5)=(bl(id)*pn(1/20
c
c t
c k = a k a
c NG N NG N
c (12x12) (12x6) (6x6) (6x12)
c
call zerov(hl,6*12)
call zerov(kngl,12*12)
call multtt(kng,dan,hl,6,6,12)
call multtt(danp,hl,kngl,12,6,12)
Then it proceeds in computing explicitly the simplified (isimp='simpgs') or

partly simplified (isimp='psimpgs') geometrical stiffnesses as
c
c EFFECTS of PNl
c
if(igeoms.eq.'simpgs'.or.igeoms.eq.'psimpgs') then
call zerov(copnl,12*12)
copnl(2,2)=(pn(1)/bl(id
copn1(2,8)=-(pn(1)/bl(id))
copn1(3,9)=-(pn(1)/bl(id))
copn1(8,2)=-(pn(1)/bl(id))
copn1(9,3)=-(pn(1)/bl(id))
endif
c
c EFFECTS of PN2
c
if(igeoms.eq.'psimpgs') then
call zerov(copn2,12*12)
copn2(4,2)=-(pn(2)/bl(id))
copn2(4,8)=(pn(2)/bl(id))
copn2(6,4)=-(pn(2)/2.)
copn2(6,10)=-(pn(2)/2.)
copn2(10,2)=(pn(2)/bl(id))
copn2(10,8)=-(pn(2)/bl(id))
copn2(12,4)=(pn(2)/2.)
copn2(12,10)=(pn(2)/2.)
copn2(2,4)=-(pn(2)/bl(id))
copn2(4,6)=-(pn(2)/2.)
copn2(10,6)=-(pn(2)/2.)
copn2(2,10)=(pn(2)/bl(id))
copn2(8,10)=-(pn(2)/bl(id))
copn2(4,12)=(pn(2)/2.)
copn2(10,12)=(pn(2)/2.)
c
c EFFECTS of PN4
c
copn4(4,9)=-(pn(4)/bl(id))
copn4(9,4)=-(pn(4)/bl(id))
copn4(5,4)=-(pn(4)/2.)
copn4(4,5)=-(pn(4)/2.)
copn4(5,10)=-(pn(4)/2.)
copn4(10,5)=-(pn(4)/2.)
copn4(10,3)=-(pn(4)/bl(id))
copn4(3,10)=-(pn(4)/bl(id
copn4(10,9)=(pn(4)/bl(id
copn4(11,4)=(pn(4)/2.)
copn4(4,ll)=(pn(4)/2.)
c
c EFFECTS of PN6
c
copn6(2,ll)=-(pn(6)/bl(id
copn6(8,ll)=(pn(6)/bl(id
endif
Finally, it adds all separate contributions and transforms the local geomet-
rical stiffness to the global coordinate system. This is accomplished via
commands
do i=l,12
do j=l,12
kgeom(i,j)=kngl(i,j)+copnl(i,j)+
copn2(i,j)+copn4(i,j)+copn6(i,j)
enddo
enddo
c
c T
c FORM k = T k T
9.9 Assembly of the geometrical stiffness 319
c G 06 G 06
c (12x12) (12x12)(12x12)(12x12)
c
call zerov(multll,12*12)
call zerov(gg,12*12)
call multtt(kgeom,t04,multl1,12,12,12)
call multtt(t04t,multll,gg,12,12,12)
9.9 Assembly of the geometrical stiffness

Before proceeding to the solution of the eigenvalue problem we assemble
the geometrical stiffness matrix into the one dimensionmal array b(nwk).
This is accomplished in subroutine assba using the main loop
c..... MA I N ASS EM B L Y L0 0P
do 140 in = 1,noel(id)
ip = node(id,in)
do 160 iu = 1,ndf
i = indxCip,iu)
if(i.eq.O) goto 160
do 170 inn = 1,noel(id)
ipp = node(id,inn)
do 180 ju = 1,ndf
j = indx(ipp,ju)
if(i.gt.j) goto 180
ij=maxa(j)+(j-i)
b(ij)=b(ij)+k(iu,in,ju,inn)
180 continue
170 continue
160 continue
140 continue
Following assembly of both elastic and geometrical stiffnesses, we can

proceed to the solution of our static or eigenvalue problem.
9.10 Scholium
We emphasize that only the general logic of the computer program is given
in this chapter. As stressed in the beginning, we believe that if the reader
comprehends the theory presented he can easily follow the computer pro-
gram and in addition refine it and expand it according to his own flavor
and technique.
Appendix A
Geometry of the beam

element in space
We consider a 2-node beam element in space equiped with a local coordinate

system x'Y' z' placed at its center. We seek to derive the matrix of direction
cosines To beetwen the local coordinate x'y'z' and the global coordinate
xyz.
The beam element comprises two nodes with global coordinates Xl, YI,
Zl and X2, Y2, Z2, respectively. In order to orient the middle surface we
must define an extra point P* with global coordinates x*, y*, z* so that
points P*, 1, 2, form a plane that passes through the middle plane of the
beam element. We define the differences of the global coordinates for nodes
1,2 as
X2 -
[ Y2 -
Xl]
YI = [XI2]
Yl2 , (A.l)
z2 - zl zl2
and the differences of the global coordinates between point P* and node 1
as
Xl]
Y: - YI
X* -
Xle]
[ = [Yle . (A.2)
Z - Zl Zle
In addition
322 Geometry of the beam element in space
Xl -
[ Yl - Y:X*] =
[xel]
Yel . (A.3)
Zl - Z Zel
The entries of the first row in To comprise components of the unit vector
PX' = CXIX]
[ Cx' Y = l1 [X12]
Y12 , (A.4)
CX' Z Z12
where
1= VX~2 + Y~2 + z~2' (A.5)
The unit vector perpendicular to the beam's surface is defined as
.... CZIX] 0.. .12 X 0.. .Ie 1 [Y12 Zle - Z12Yle]

PZI = [ Czl y = 20 20 Z12 X le - X12 Zle , (A.6)
CZI z X12Yle - Y12 X le
where 0 is the area of the triangle formed by the two nodes and the extra
point defined as
1
0= -JC12
2
+ C2 2 + C3 2 , (A.7)
in which
Cl] [Y12Zel - Yel Z12]

[ C2 = Z12Xel - Zel X12 . (A.8)
C3 X12Yel - XelY12
Finally, the components of the unit vector Pyl are simply given by
CyIX] [CZIYCXIZ - CZIZCX1Y]

Pyl [ Cyl y = PZI X Pxl = CZ'ZCX' X - CZ'XCX' Z (A.9)
CyIZ CzlxCxly - CzlyCxlx
323
We are now in the position to define the matrix of direction cosines via
cx, X c x' Y zl
c x'
To = [ cy'x Cy'y cy'z . (A.lO)
Cz'x Cz'y Cz'z
Note that matrix To is derived in an identical manner for the 3-node shell
element by simply replacing the extra point P* with node 1, and the beam
nodes 1, 2 with the triangular nodes 2, 3, respectively.
Appendix B
Contents of floppy disk
Included with in the floopy disk accompanying the book is a model com-
puter program for static and buckling analyses of isotropic and laminated
composite beams, frames and large three-dimensional beam assemblies.
The name of the program is beam1.f and is written in standard Fortran
77. It can be compiled and run on any computer with Fortran 77. The pro-
gram was originally written and compiled on a Sun Sparcstation 10 using
the command 17 beam1.f.
In addition to the source code, four example problems are provided to
familiarize the user with the input data and the execution of the computer
program. All axamples were described in the text.
The following files are included in the floopy disk:
1. beam1.f: Main computer program.
2. parcb.h: A short file setting the problem parameters.
3. meshb.dat: A file containing the four test problems, namely
static analysis of an L-shaped isotropic beam {example in Fig.

5.12}
static analysis of an isotropic frame {example in Fig. 5.13}
static analysis of a {45/ -45/ -45/45} cantilver composite beam
{example in Fig. 5.20}
buckling analysis of a {45/ -45/0/90}8 composite beam {example
in Fig. 5.22}
Bibliography
[1] I.S. Sokolnikoff. Mathematical Theory of Elasticity. McGraw-Hill,

1956.
[2] J.S. Przemieniecki. Theory of Matrix Structural Analysis. Dover Pub-
lications. Inc., New York, 1985.
[3] T.R. Tauchert. Energy Principles in Structural Mechanics. McGraw-
Hill Inc., New York, 1974.
[4] P.L. Gould. Introduction to Linear Elasticity. 2nd ed.,Springer-Verlag
New York, Inc., 1994.
[5] R. Courant. Variational methods for the solution of problems of equi-
librium and vibration. Bull. Am. Math. Soc., 49:1-43, 1943.
[6] W. Prager and J.L. Synge. Approximation in elasticity based on the
concept of function space. Q.J. Appl. Math., 5:241-269, 1947.
[7] J. Argyris, S. Kelsey and S. Bagat. A constant strain triangular ele-
ment for the matrix displacement method. Imperial College Internal
Report, 1955.
[8] J. Argyris, S. Kelsey and S. Bagat. The triangular membrane element
for the matrix displacement method and the analysis of complex wings.
Imperial College Internal Report, August 1955.
[9] J. Argyris. Energy theorems and structural analysis. Aircraft En-
gineering, pages reprinted by Butterworth's Scientific Publications,
1960, 5th reprint 1977, 26 (1954) 347-356, 383-387; 27 (1955) 42-58,
80-94, 125-134,145-158.
[10] J. Argyris and S. Kelsey. Energy theorems and structural analysis.
Aircraft Engineering, pages reprinted by Butterworth's Scientific Pub-
lications 1960, 5th reprint 1977, 26 (1954) 410-422.
328 BIBLIOGRAPHY
[11] RW. Clough. The finite element method: A personal view of its
original formulation. in From Finite elements to the Troll platform,
Ivar Holland 70th anniversary, pages 89-97, 1994.
[12] M.J. Turner, RW. Clough, RW. Martin and L. Topp. Stiffness and
deflection analysis of complex structures. J. Aero. Sci., 25:805-823,
1956.
[13] RW. Clough. The finite element method in plane stress analysis.
Proc. 2nd ASCE Conf. on Electronic Computation, Pitsburg, Pa.,
Sept. 1960.
[14] J. Argyris. Continua and discontinua. Oppening address to the Interna-

tional conference on matrix methods of structural mechanics, Dayton,
Ohio, Wright-Patterson U.S.A.F. Base, pages 1-198, 1965.
[15] J. Argyris and D.W. Scharpf. Some general considerations on the

natural mode technique, part i, small displacements, part ii, large dis-
placements. Aero. J. Roy. Aero. Soc., 73:219-226, 361-368, 1969.
[16] J. Argyris and L. Tenek. Natural mode method: A practicable and

novel approach to the global analysis of laminated composite plates
and shells. Appl. Mech. Rev., 49:381-400, 1996.
[17] J. Argyris and L. Tenek. Recent advances in computational ther-

mostructural analysis of composite plates and shells with strong non-
linearities. Appl. Mech. Rev., 50:285-306, 1997.
[18] KK Gupta and J.L. Meek. A brief history of the beginning of the
finite element method. Int. J. Numer. Meth. Engrg., 39:3761-3774,
1996.
[19] O.C. Zienkiewicz and Y.K Cheung. The Finite Element Method in
Continuum and Structural Mechanics, volume 1. McGraw-Hill, New
York, 1965.
[20] RW. Clough. Original formulation of the finite eleemnt method.

ASCE Structures Congress Session on Computer Utilization in Struc-
tural Eng., pages 1-10, 1989.
[21] O.C. Zienkiewicz and Y.K Cheung. Finite elements in the solution of
field problems. The Engineer, 220:507-510, 1965.
BIBLIOGRAPHY 329
[22] O.C. Zienkiewicz and RL. Taylor. The Finite Element Method, vol-
ume 1. McGraw-Hill, fourth edition, 1991.
[23] O.C. Zienkiewicz. Origins, milestones and directions of the finite ele-
ment method - a personal view. Archives of Comput. Meth. Engnrg.,
2:1-48, 1995.
[24] Lord Rayleigh. On the theory of resonance. Trans. Roy. Soc., A161:77-
118, London, 1870.
[25] W. Ritz. Uber einer neue methode zur losung gewissen variations-
probleme der mathematischen physik. J. Reine angew. Math., 135:1-
61, 1909.
[26] B. Siuru and J.D. Busick. Future Flight: The next generation of air-
craft technology. TAB AERO, Blue Ridge Summit, PA, 1994.
[27] RM. Jones. Mechanics of Composite Materials. Hemisphere, New

York, 1975.
[28] J.M. Whitney. Structural Analysis of Laminated Anisotropic Plates.

Technomic, Lancaster, 1987.
[29] L.R Calcote. The Analysis of Laminated Composite Structures. Van

Nostrand Reinhold Co., NY, 1969.
[30] S.P. Timoshenko and J.N. Goodier. Theory of Elasticity. 3rd Edn.,
McGraw-Hill, New York, 1970.
[31] S. Timoshenko. Strength of materials, Part II. 2nd Edn., Van Nos-
trand, Amsterdam, 1974.
[32] F.P. Beer and E.R Johnston Jr. Mechanics of Materials. Mc Graw-
Hill, 1981.
[33] A. Venkatesh and K.P. Rao. A laminated anisotropic curved beam and
shells stiffening element. Comput. Struct., 15:197-200, 1982.
[34] F. Yuan and RE. Miller. A new finite element for laminated composite
beams. Comput. Struct., 31:737-745, 1989.
[35] S. Oral. A shear flexible finite element for nonuniform, laminated

composite beams. Comput. Struct., 38:353-360, 1991.
330 BIBLIOGRAPHY
[36] A.T. Chen and T.Y. Yang. Static and dynamic formulation of a sym-
metrically laminated beam finite element for a microcomputer. J.
Compos. Mater., 19:459-475, 1985.
[37] A. Mattsson. A 2-node composite beam element. Master's thesis,

University of Stuttgart, 1997.
[38] J. Argyris, L. Tenek and L. Olofsson. Tric: a simple but sophisti-
cated 3-node triangular element based on 6 rigid-body and 12 strain-
ing modes for fast computational simulations of arbitrary isotropic
and laminated composite shells. Comput. Meth. Appl. Meeh. Engrg.,
145:11-85, 1997.
[39] L. Tenek and J. Argyris. Computational aspects of the natural-mode
finite element method. Comm. Num. Meth. Eng., 13, 1997.
[40] D.R.J. Owen and J.A. Figueiras. Anisotropic elasto-plastic finite ele-
ment analysis ofthick and thin plates and shells. Int. J. Numer. Meth.
Engrg., 19:541-566, 1983.
[41] J.H. Argyris, H. Balmer, J. St. Doltsinis, P.C. Dunne, M. Haase,

M. Kleiber, G.A. Malejannakis, H.-P. Mlejnek, M. Muller and D.W.
Scharf. Finite element method - the natural approach. Comput. Meth.
Appl. Meeh. Engrg., 17/18:1-106, 1979.
[42] J.L. Meek. Matrix Structural Analysis. Mc Graw-Hill, 1971.

[43] W. Flugge. Stresses in shells, 2nd ed. Springer-Verlag, Berlin, 1973.
[44] S. Timoshenko and S. Woinowski-Krieger. Theory of Plates and Shells.
McGraw-Hill, New York, 1959.
[45] J.H. Argyris, P.C. Dunne, G.A. Malejannakis and E. Schelkle. A simple
triangular facet shell element with applications to linear and nonlinear
equilibrium and inelastic stability problems. Comput. Meth. Appl.
Meeh. Engrg., 10:371-403, 1977.
[46] J.H. Argyris and L. Tenek. A natural triangular layered element for
bending analysis of isotropic, sandwich, laminated composite and hy-
bride plates. Comput. Methods Appl. Meeh Engrg, 109:197-218, 1993.
[47] J.H. Argyris and L. Tenek. Buckling of multilayered composite plates
by natural shear deformation matrix theory. Comput. Methods Appl.
Meeh Engrg, 111:37-59, 1994.
BIBLIOGRAPHY 331
[48] J.H. Argyris and L. Tenek. Linear and geometrically nonlinear bending
of isotropic and multilayered composite plates by the natural mode
method. Comput. Meth. Appl. Meeh Engrg., 113:207-251, 1994.
[49] J.H. Argyris and L. Tenek. Nonlinear and chaotic oscillations of com-
posite plates and shells under periodic heat load. Comput. Methods
Appl. Meeh Engrg, 122:351-377, 1995.
[50] J.H. Argyris and L. Tenek. Postbuckling of composite laminates un-

der compressive load and temperature. Comput. Methods Appl. Meeh
Engrg, 128:49-80, 1995.
[51] J.H. Argyris and L. Tenek. An efficient and locking-free flat anisotropic
plate and shell triangular element. Comput. Meth. Appl. Meeh. Engrg.,
118:63-119, 1994.
[52] J .H. Argyris and L. Tenek. A practicable and locking-free laminated

shallow shell triangular element of varying and adaptable curvature.
Comput. Meth. Appl. Meeh. Engrg., 119:215-282, 1994.
[53] J.H. Argyris, L. Tenek and L. Olofsson. Nonlinear free vibrations of

composite plates. Comput. Methods Appl. Meeh Engrg, 115:1-51, 1994.
[54] J .H. Argyris and L. Tenek. High-temperature bending, buckling and

postbuckling of laminated composite plates using the natural mode
method. Comput. Methods Appl. Meeh Engrg, 117:105-142, 1994.
[55] L. Olofsson. On the natural mode method. Master's thesis, University

of Stuttgart, 1994.
[56] P.O. Marklund. A study of thick cylindrical laminates with pentahedra

elements using the sub element technique. Master's thesis, University
of Stuttgart, 1996.
[57] MACSYMA. Reference manual. Macsyma Inc., Arlington, MA, 1992.
[58] D.J. Allman. A compatible triangular element including vertex rota-

tions for plane elasticity analysis. Computers and Structures, 19:1-8,
1984.
[59] D.J. Allman. A quadrilateral finite element including vertex rotations

for plane elasticity analysis. Int. J. Num. Meth. Eng., 26:717-730,
1988.
332 BIBLIOGRAPHY
[60] D.J. Allman. Evaluation of the constant strain triangle with drilling
rotations. Int. J. Num. Meth. Eng., 26:2645-2655, 1988.
[61] C.Y. Chiao Nonlinear Analysis of Plates. McGraw-Hill, NY, 1980.
[62] PK Sinha B Chattopadhyay and M Mukhopadhyay. Geometrically

nonlinear analysis of composite stiffened plates using finite elements.
Compos. Struet., 31:107-118, 1995.
[63] J.H Argyris and D.W. Scharpf. Large deflection analysis of prestressed
networks. ASCE J. Struet. Div., 98:633-654, 1972.
[64] J.H Argyris and P.C. Dunne. A simple theory of geometrical stiffness
with applications to beam and shell problems. lSD-Report No 183,
Stuttgart, 1975.
[65] J.F. Zhu. Application of natural approach to nonlinear analysis of

sandwich and composite plates and shells. Comput. Meth. Appl. Meeh.
Engrg., 120:355-388, 1995.
[66] J.F. Zhu. A new consideration on the derivation of the geometrical

stiffness matrix with the natural approach. Comput. Meth. Appl. Meeh.
Engrg., 123:141-160, 1995.
[67] T.J.R. Hughes and W.K Liu. Nonlinear finite lement analysis of shells
part ii: two-dimensional shells. Comput. Meth. Appl. Meeh. Engrg.,
27:167-182, 1981.
[68] D. Lam KK Liu, E.S. Law and T. Belytschko. Resultant-stress

degenerated-shell element. Comput. Meth. Appl. Meeh. Engrg.,
55:259-300, 1986.
[69] J.C. Simo and D.D. Fox. On a stress resultant geometrically exact
shell model. part ii: the linear theory; computational aspects. Comput.
Meth. Appl. Meeh. Engrg., 73:53-92, 1989.
[70] T. Belytschko and I. Leviathan. Physical stabilization of the 4-node

shell element with one-point quadrature. Comput. Meth. Appl. Meeh.
Engrg., 113:321-350, 1994.
[71] KJ. Bathe and E.N. Dvorkin. A formulation of general shell elements-
the use of mixed interpolation of tensorial components. Int. J. Numer.
Meth. Eng., 22:697-722, 1986.
BIBLIOGRAPHY 333
[72] K.P. Rao. A rectangular laminated anisotropic shallow thin shell finite
element. Comput. Meth. Appl. Mech. Engrg., 15:13-33, 1978.
[73] R.H. MacNeal and R.L. Harder. A proposed standard set of problems
to test finite element accuracy. Finite Elements Anal. Des., 1:3-20,
1985.
[74] B.L. Wong T. Belytschko and H. Stolarski. Assumed strain stabiliza-

tion procedure for the 9-node lagrange shell element. Int. J. Numer.
Meth. Engrg., 28:385-414, 1989.
[75] N.D. Phan and J.N. Reddy. Analysis of laminated composite plates
using a higher-order shear deformation theory. Int. J. Numer. Meth.
Eng., 21:2201-2219, 1985.
[76] N.J. Pagano. Exact solutions for rectangular bidirectional composites

and sandwich plates. J. Compos. Mater., 4:20-34, 1970.
[77] T. Kawai and N. Yoshimura. Analysis of large deflection of plates by

the finite element method. Int. J. Numer. Meth. Engrg., 1:123-133,
1969.
[78] A.K Noor. Stability of multilayered composite plates. Fibre Sci. Tech-
nol., 8:81-89, 1975.
[79] K-J Bathe. Finite Element Procedures in Engineering Analysis.

Prentice-Hall, Englewood Cliffs, New Jersey, 1982.
[80] M. Papadrakakis. Solving Large-Scale Problems in Computational Me-

chanics, volume 1. John Wiley and Sons, Ltd, 1993.
[81] M.A. Crisfield. Non-Linear Finite Element Analysis of Solids and

Structures, volume 2: Advanced Topics. John Wiley and Sons, Ltd.,
West Sussex, England, 1993.
[82] E. Riks. An incremental approach to the solution of snapping and

buckling problems. Int. J. Solids Structures, 15:524-551, 1979.
[83] M. A. Crisfield. A fast incremental/iterative solution procedure that

handles "snap-through". Computers and Structures, 13:55-62, 1981.
[84] M. Papadrakakis. A truncated newton-Ianczos method for overcoming

limit and bifurcation points. Int. J. Numer. Meth. Engrg., 29:1065-
1077, 1990.
334 BIBLIOGRAPHY
[85] E. Hinton and D.R.J. Owen. Finite Element Software for Plates and
Shells. Pineridge Press, Swansea, U.K., 1984.
[86] User's guide. FORTRAN 3.0.1. Sun Microsystems, Mountain View,
California, 1994.
[87] User's guide. PATRAN - Vol. 1. PDA Engineering, Santa Ana, CA,
1984.
Index
accumulation, 217 body force, 2

algebraic operations, 263 body forces, 2
angle-ply, 43 boundary conditions, 221, 230
angle-ply laminate, 44 buckling, 99, 119, 217, 234
anisotropy ratios, 235 buckling loads, 217, 261
antisymmetrical bending mode, 61, buckling modes, 217, 240, 261
164
antisymmetrical bending terms, 182 cantilever beam, 224
antisymmetrical moments, 175 cantilever composite beam, 119,
antisymmetrical rotations, 175 302
antisymmetrical shearing mode, 164 cartesian moments, 215
antisymmetrical shearing stiffness, cartesian nodal forces, 105, 216
185 cartesian strains, 140
antisymmetrical shearing terms, 186 central displacement, 218
aspect ratio, 233 clamped isotropic plate, 218
aspect ratios, 229 classical finite element methods,
aspect ratios , 234 25, 263
assembly loops, 313 classical lamination theory, 41
assembly operation, 264 classical plate of solution, 233
averaging procedures, 185 close form solutions, 239
axial direct strain, 51 CLT, 42
axial strain, 60 code structure, 301
azimuth moments, 162 collapse loads, 267
component stresses, 145, 204
balanced, 42 composite, 37
balanced laminates, 43 composite beam columns, 119
beam element, 31 composite beams, 119
beam shearing coefficient, 63 composite cylinder, 240
biaxial symmetry, 231, 233 composite laminate, 230
bidirectional laminate, 233 composite laminates, 42
bifurcations, 267 composite material, 44
binder, 37 composite satellite, 119
336 INDEX
composite shell, 261 Euler formula, 119

composite shells, 136 exact solution, 221
compressive loads, 119, 234 experimental results, 234
computational aspects, 261 external loads, 86
computational efficiency, 264
computer program, 261, 320 FEM, vii
computing time, 264 fiber reinforced composites, 38
congruent operations, 171 fibers, 37
congruent transformations, 191 filaments, 37
convergence, 221, 232 finite element method, vii, 19, 20
convergence characteristics, 119,224 finite element programming, 301
coordinate transformation, 213 force method, 18
CPU time list, 263 Fortran commands, 302
CRAY-C94, 261 generalized forces, 166
critical buckling load, 119 geometrical forces, 212
critical load, 127, 239 geometrical stiffness, 20, 99, 208,
critical pressure, 250 319
cross-ply laminate, 44, 233 global arrays, 309
cross-ply plate, 239 global cartesian coordinate, 166
cylindrical arc-length, 267 global elemental vector, 84
cylindrical panel, 239 Global equilibrium, 202
cylindrical roof, 220 global force increments, 104
global load vector, 88
deformation plot, 234
graphical visualization, 314
deformed beams, 227
deformed geometry, 100 hemispherical shell, 221
design, viii heterogeneous medium, 44
direct, 2 High speed flight, 38
direction cosines, 85, 167 higher-order shear deformation the-
double sinusoidal load, 230, 233 ory, 235
higher-order theory, 229
efficiency, 261 homogeneous coordinates, 176
eigenvalue analyses, 261 hydrostatic pressure, 240
eigenvalue problem, 217
elastic curve, 61 ill conditioning, 314
elasticity solution, 232 incremental, 267
elements, 25 inextensional bending modes, 221
engineering shear strains, 10 initial load vectors, 89
equilibrium, 6, 85 input parameters, 302
error, 219 integration formula, 182
INDEX 337
interpolation matrix, 87 model problem, 261, 302

isoparametric, 267
isotropic beams, 114 natural coordinate system, 28, 191,
isotropic column, 119 212
isotropic frames, 115 natural directions, 177
isotropy, 152 natural forces, 212
iterative, 267 natural geometrical stiffness, 106,
216
jet propulsion, 18 natural modes, vii, 20, 48
natural stiffness matrix, 30, 57
Kirchhoff, 176 natural strain energies, 96
natural straining modes, 56
lamina, 44
natural thermal load, 91, 200
laminated beam element, 52
natural thermal load vector, 89,
lamination schemes, 250
198
large deflections, 233
Newton-Raphson, 268
limit points, 267
normal, 2
load factor, 217
normalized central displacement,
load-displacement curves, 234
231
loading vector, 313
normalized deflection, 227
local coordinate system, 40
normalized form, 231
local coordinates, 154
normalized solutions, 221
locking, 63
normalized stresses, 230
lower order theory, 233
normalized values, 224
macromechanics, 44 numerical quadratures, viii
material anisotropy ratio, 235 opposite, 41
material coordinate system, 40, 52 optimization, viii
material properties, 229 optimization procedures, 250
material transformations, 263 orthotropic layers, 230
matrix, 37 orthotropic shell, 220
matrix displacement method, 19
matrix materials, 39 parallelization, 263, 264
matrix multiplications, 263 partly simplified geometrical stiff-
membrane forces, 216 ness, 108, 110
membrane stresses, 100, 217 path, 267
membrane theory of shells, 135 physical lumping, 136
micromechanics, 44 physical problem, 314
middle plane, 321 physical vectors, 11
middle surface, 321 Piecewise summation, 172
minimum, 15 pinched cylinder, 220
338 INDEX
plane stress triangle, 31 sinusoidal loading, 229

Polymer composites, 37 skyline, 261
postprocessing, 95 skyline storage, 310
prestress state, 263 square laminate, 234
principal of virtual work, 13 stationary, 15
principle of stationary energy, 15 statistics report, 261
stiffening or softening effect, 106
quasi-isotropic, 119 stiffness matrix, vii
quasi-isotropic laminate, 227
storage scheme, 310
quasi-isotropic laminates, 44 strain energy, 84, 145
rectangular laminate, 234 strain energy density, 11
reference solution, 219 strain modes, vii
regional discretization, 17 strain operator matrix, 29
reinforcement, 37 strain rossete, 140
representative volume, 45 straining modes, vii, 163
rigid body modes, vii, 52 stress tensor, 10
rigid body moments, 101, 208, 212, stress vector, 1
215 structural laminate, 41
rigid body rotation, 215 structural mechanics, 21
rigid body rotations, 101, 208, 212, surface forces, 2
221 surface load, 196
rigid diaphragms, 250 symbolic computation, 216
rocket-like composite shell, 250 symmetric, 42
rosettes, 19 symmetric laminate, 42
rotational geometrical stiffness, 103, symmetrical, 8
215 symmetrical bending mode, 60
symmetrical lamination, 93
sandwich plate, 219, 229 symmetry, 220
shear, 2
shear correction factor, 78, 219 tangent stiffness, 100
shear correction factors, 189, 219, temperature, 89, 240
233 temperature increase, 239
shear deformations, 175 tensor, 5, 8
shear locking, 48, 166 tetrahedron element, 32
shear strains, 8 thermal load vector, 198
shell element, 32, 163 thermal strain, 200
shell elements, 224 thermoelastic coefficients, 90, 198
shell structures, 136 thermomechanical buckling, 239
simplified geometrical stiffness, 105, three dimensional solution, 230
216 total natural strain, 144
INDEX 339
total natural strains, 140

total potential energy, 15
total strain, 167
traction, 1
transformation matrix, 211
transverse shear deformation, 219
transverse shear strains,. 148
transverse shear stresses, 204
triangle, 18
truss element, 31
uniform pressure, 219

uniform pressure load, 227
unit extension, 57
unit matrix, 212
unit vector, 322
unstable branches, 267
vectorization, 263, 264

vertex rotation, 191
vertical displacement, 221
virtual displacements, 12
virtual work, 12, 13, 181, 196
warping, 224
weight sensitive structures, 38
Mechanics
SOUD MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
44. D.A. Hills, P.A. Kelly, D.N. Dai and A.M. Korsunsky: Solution of Crack Problems. The
Distributed Dislocation Technique. 1996 ISBN 0-7923-3848-0
45. V.A. Squire, R.J. Hosking, A.D. Kerr and P.J. Langhorne: Moving Loads on Ice Plates. 1996
ISBN 0-7923-3953-3
46. A. Pineau and A. Zaoui (eds.): IUTAM Symposium on Micromechanics of Plasticity and
Damage of Multiphase Materials. Proceedings of the IUTAM Symposium held in Sevres,
Paris, France. 1996 ISBN 0-7923-4188-0
47. A. Naess and S. Krenk (eds.): IUTAM Symposium on Advances in Nonlinear Stochastic
Mechanics. Proceedings of the IUTAM Symposium held in Trondheim, Norway. 1996
ISBN 0-7923-4193-7
48. D. Iean and A. Scalia: Thermoelastic Deformations. 1996 ISBN 0-7923-4230-5
49. J. R. Willis (ed.): IUTAM Symposium on Nonlinear Analysis of Fracture. Proceedings of the
IUTAM Symposium held in Cambridge, UK 1997 ISBN 0-7923-4378-6
50. A. Preumont: Vibration Control ofActive Structures. An Introduction. 1997
ISBN 0-7923-4392-1
51. G.P. Cherepanov: Methods of Fracture Mechanics: Solid Matter Physics. 1997
ISBN 0-7923-4408-1
52. D.H. van Campen (ed.): IUTAM Symposium on Interaction between Dynamics and Control in
Advanced Mechanical Systems. Proceedings of the IUTAM Symposium held in Eindhoven,
The Netherlands. 1997 ISBN 0-7923-4429-4
53. N.A. Fleck and A.C.F. Cocks (eds.): IUTAM Symposium on Mechanics of Granular and
Porous Materials. Proceedings of the IUTAM Symposium held in Cambridge, U.K. 1997
ISBN 0-7923-4553-3
54. J. Roorda and N.K. Srivastava (eds.): Trends in Structural Mechanics. Theory, Practice,
Education. 1997 ISBN 0-7923-4603-3
55. Yu. A. Mitropolskii and N. Van Dao: Applied Asymptotic Methods in Nonlinear Oscillations.
1997 ISBN 0-7923-4605-X
56. C. Guedes Soares (ed.): Probabilistic Methodsfor Structural Design. 1997
ISBN 0-7923-4670-X
57. D. Franyois, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume I:
Elasticity and Plasticity. 1998 ISBN 0-7923-4894-X
58. D. Franyois, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume II:
Viscoplasticity, Damage, Fracture and Contact Mechanics. 1998 ISBN 0-7923-4895-8
59. L. T. Tenek and J. Argyris: Finite Element Analysis for Composite Structures. 1998
ISBN 0-7923-4899-0
Kluwer Academic Publishers - Dordrecht / Boston / London

(Solid Mechanics and Its Applications 59) Lazarus Teneketzis Tenek, John Argyris (Auth.) - Finite Element Analysis For Composite Structures-Springer Netherlands (1998)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Solid Mechanics and Its Applications 59) Lazarus Teneketzis Tenek, John Argyris (Auth.) - Finite Element Analysis For Composite Structures-Springer Netherlands (1998)

Uploaded by

Copyright:

Available Formats

FINITE ELEMENT ANALYSIS FOR COMPOSITE STRUCTURES

SOLID MECHANICS AND ITS APPLICATIONS

Series Editor: G.M.L. GLADWELL

Aims and Scope of the Series

For a list of related mechanics titles, see final pages.

LAZARUS TENEKETZIS TENEK

Springer-Science+Business Media, B.V.

Printed on acid-free paper

All Rights Reserved

1 Some results from continuum mechanics 1

2 A brief history of FEM 17

3 Natural modes for finite elements 27

5 Composite beam element 47

6 Composite plate and shell element 135

6.17.13 Thermomechanical buckling of a cylindrical compos-

7 Computational statistics 261

8 Nonlinear analysis of anisotropic shells 267

9 Programming aspects 301

A Geometry of the beam element in space 321

B Contents of floppy disk 325

This book is an adventure into the computer analysis of three dimensional

An element in space has 6 possible rigid body motions: three

We thank Lars Olofsson, Per-Olof Marklund and Anders Mattsson who

Some results from

1.1 The analysis of stress

Tn = O"nx i + O"ny; + O"nz k . (1.2)

Ti = O"xxi + O"xy; + O"xzk. (1.3)

L. T. Tenek et al., Finite Element Analysis for Composite Structures

F = Fxi + Fyj + Fzk. (1.5)

~Mz - b(4ac(J'yx) - (-b)(-4ac(J'yx) + (-a) (-4bc(J'xy) + a(4bc(J'xy) = O.

But the volume of the element is ~V = 8abc so that

and as ~ V --+ 0, the first term tends to zero on account of assumption

(J'yx = (J'xy' (1.8)

a zy = a yz , axz = a zx , a yx = axy (1.9)

Now consider a small rectangular tetrahedron element PABC with

!:l.Rx + !:l.S(anx -laxx - ma yx - na zx ) = O. (1.10)

lim !:l.Rx = lim !:l.Rx.!:l. V = lim !:l.Rx.!: = O.

Thus we obtain the first of the three equations

a nx= laxx + mayx + nazx ,

Often it is convenient to use numerical suffices 1,2,3, instead of x, y, z.

aij = aji (1.13)

In this notation, the components Ti of the traction across a surface with

Often we use Einstein's double suffix convention and rewrite this

then the stresses O'~j in the new frame are

O'~j = lik1jiO'kl. (1.17)

Now consider the dynamic equilibrium of an arbitrary volume V bounded

Is T n dS + Iv FdV = Iv p ~:~ dV. (1.18)

Replacing Tn by its components Ti, and using Eq. (1.14) we find

Is A .v = Iv divA dV. (1.20)

In double suffix notation this is

where ,j = o/chj. Applying this to Eq. (1.19) we find

Iv (aij,j + Fi - pili) dV = O. (1.22)

But V is arbitrary so that the integrand must be identically zero, i.e.

aij,j + Fi = pili (1.23)

In the important, static case, these equations are

oaxx + oaxy + oaxz + Fx = 0,

These equations are called the equilibrium equations.

1.2 The analysis of strain

ux(x + dx, y, z) - ux(x, y, z) oU x

Figure 1.2: Axial and shearing strains

In the context of small deformation theory

We define as shear strains the quantities

form the components of a tensor. By this we mean that (just as we found