You are on page 1of 6

ELEMENT FREE GALERKIN METHOD AKG

Significance of Dimensionless Size of Support Domain


In Element Free Galerkin Method
J.S. Kushawaha
Indus Institute of Technology and Management, Bilhaur, Kanpur 209002
k.jitendrasingh@yahoo.com
----------------------------------------------------------------------------------------------------------------------------------------
Abstract--The aim of this paper is to give a practical view of
meshfree methods based on the study performed with the help of
MATLAB code. Various selectable parameters have been
brought together and presented for quick reference. The focus
has been made for selection of the dimensionless size of support
domain, S in Element Free Galerkin (EFG) method. A
comparison of finite element solution has been also performed to
compare the accuracy of the solutions. The EFG method was
selected for the study for the reason of being easy to implement
and versatile. Effect of variation of S is studied and the
significance of intuition, knowledge and experience of the analyst
is brought to front with the help of nodal parametric solutions of
a tapered elastic bar.
Keywords: Galerkin method, Meshfree methods, Finite element
solutions
I. INTRODUCTION
Meshfree, Meshless or element free methods are the state-of-
the art methods for analysis of various physical problems
which can be modeled and represented by differential
equations. These are called so because these do not require a
mesh of elements, which is mandatory for the mesh based
methods like Finite Element Methods.
The finite element methods are well established and powerful
computational / simulation techniques which are used for
modeling and analysis of physical phenomena in different
fields of engineering and applied sciences. It has successfully
been applied for a large number of engineering applications,
for example solid mechanics, structure mechanics, electro
magnetism, geo-mechanics, bio mechanics, aerodynamics and
so on, but it is with some shortcomings as mentioned below:
1. The use of mesh in modeling these problems creates
difficulties in the treatment of discontinuities which do not
coincide with the original mesh lines.
2. In stress calculations, the stresses obtained using FEM
packages are discontinuous and less accurate.
3. For modeling large deformation problems, considerable
loss in accuracy arises when the elements in mesh become
extremely skewed or compressed.
4. It is very difficult to simulate the breakage of material into
a large number of fragments as FEM is essentially based on
continuum mechanics, in which the elements formulated
cannot be broken.
5. Adaptive processors require mappings of field variables
between meshes in successive stages in solving the problem.
The frequent re-meshing and mapping introduces numerical
error.
6. Creation of a mesh for the problem domain consumes major
time of analyst using FEM packages. It becomes a major
component of the cost of a simulation project. The concern is
more the manpower time, and less the computer time.
To overcome the above difficulties meshfree methods have
been developed which do not require a mesh to discretize the
problem and any connectivity between the nodes. The
advantages of mesh free methods are summarized as follows:
1. The approximate solution is constructed entirely in terms of
a set of nodes.
2. There is no need to provide in advance any information
about the relationship of the nodes, so it provides flexibility in
adding and deleting nodes whenever and wherever needed.
3. Since there is no need to create a mesh, and the nodes can
be created by a computer in a fully automated manner, it saves
a lot of human effort.
4. They can easily handle very large deformation because
connectivity can change with time.
Though there are so many advantages, mesh free methods too
are not without disadvantages. Since mesh free shape
functions are rational functions, therefore, it requires highly
accurate integration scheme to be applied. It is to be noted that
treatment of essential boundary conditions is not straight
forward as in mesh based methods since the shape functions
of mesh free methods are not interpolants but approximants,
which implies that they do not satisfy the Kronecker delta
property.
II. ELEMENT FREE GALERKIN METHOD
This meshfree method was developed in by the group of Prof.
Ted Belytschko [2] based on the idea of Lancaster and
Salkauskas and probably motivated by the purpose to model
arbitrary crack propagation without computational expensive
re-meshing. The method has been selected for the study of the
variation of meshfree solution with the variation in the
dimensionless size of the support domain.
Boundary Representation: The Boundary representation in
meshfree methods is done only by the arbitrary distribution of
AKGEC JOURNAL OF TECHNOLOGY, vol., no.1
the nodes, which may or may not be uniform. Fig. 1 shows the
boundary representation for meshfree methods only by the
nodes and for FEM using the nodes & elements. The field
variables of interest are specified on these nodes.
Figure 1. Boundary representations in Meshfree Methods

Dimension of support domain: The dimensions of domain of
influence affects the accuracy of the interpolation at the point
of interest, therefore the selection of suitable dimension of
support domain is very important.[2,7] To define the support
domain for a point xQ, the dimension of the support domain ds
is determined by
c s s
d d (1)
Where s is the dimensionless size of the support domain. dc is
the characteristic length that relates to the nodal spacing near
the point xQ. If the nodes are uniformly distributed dc is simply
the distance between the two neighboring nodes. The uniform
distribution has been considered for the ease of programming
and understanding the methodology; however the non-
uniform distribution is also possible.
Meshfree solution Procedure: The solution procedure for the
displacement parameters at the end of the bar using meshfree
methodology is as described by the following steps:
Step-1 Discretize the bar into number of nodes distributed
uniformly over the domain = (0, 1). The selection of the
number of nodes is random and the convergence study has to
be performed to arrive at the optimum number of nodes,
depending upon the accuracy and tolerance requirement of the
solution. The discretized bar is represented by fig. 2.
Step-2 The integration cells are constructed for the one-point
gauss-quadrature integration of the nodal discrete equations.
The number of integration cells is equal to (Number of nodes
-1). The one-point gauss-quadrature integration is performed
using equation (2):
Figure 2. Meshfree representation of tapered elastic bar.
I
r
I
I
b
a
w F d F dx x F


1
) (

) (

) (
(2)
The integration points, weights and the Jacobeans for the each
cell are set. WI = 2 for the one-point gauss quadrature.
Figure 3. One-point Gauss quadrature
Step-3: The shape function is constructed by the MLS
approximation method. In this method the linear basis
functions in one-dimension:
[ ] x x p
T
1 ) ( (3)
and the quartic spline weight function:

'
+

0
, 3 8 6 1
) (
4 3 2
r r r
x x W
i
1
1
>

r
r
(4)
Is used. In matrix form the shape function is obtained as:
40
Element
) ( ) ( ) (
1
x B x A p x
i
T
i

(5)

Step-4: The approximate solution is obtained using the shape
function as:
i
n
i
i i
u x u

1
) (
(6)
Step-5 The nodal discrete equations are obtained using the
constrained Galerkin weak form using the Lagrange
multiplier[2]. The weak form is developed from the
governing differential [2, 3, 5, and 9] equation:
0 ) (
,
_

dx
du
x EA
dx
d
, 0<x<L (7)
,
subject to the boundary conditions of the problem at hand are
given by:
u(0)=0,
P
dx
du
x EA
x

1
]
1

1
) (
(8)
where

,
_


L
x
A x A
2
1 ) (
0
(9)
The weak form of governing equation is solved for the
discrete nodal equations as
dx
dx
d
dx
d
EA K
j
L T
i
ij

0
(10)
i is the trial function, j is the test function and in the EFG
method the trial and test functions are same as the shape
function. A good elaboration on this can be found in the books
and papers [1, 2, 3, 4, and 6].
ui
L
k ik
d G
u

(11)

t i i
d P f
t

(12)
u i
d u q
u


(13)

Gik ,Kij , fi and qi are the matrices for the boundary conditions,
stiffness, force and displacement respectively.
The above equations are solved using the one-point gauss
quadrature integration referred in the step-2 [4].
Step-6 The above discrete nodal equations are assembled into
global matrix:
ELEMENT FREE GALERKIN METHOD

1
]
1

1
]
1

1
]
1

q
f u
G
G K
T
0
(14)

is the Lagrange multiplier
Step-7 The essential boundary condition is imposed. The
specified displacement at fixed end is equal to zero, hence
q=0.
Step-8 The global matrix is solved to obtain the nodal
displacement parameters.
Step-9 The curves are plotted and the result is obtained.
The solution code using the MATLAB language is developed
and the resulting data and plots are recorded for further
analysis.
III. TAPERED STEEL BAR
The Element Free Galerkin method is used for obtaining the
displacement parameter at the end of the bar applying the
developed MATLAB code. The problem has been considered
from Fundamentals of finite element analysis by David V
Hutton [1].
Fig. 4 depicts the tapered elastic bar subjected to an applied
tensile load P at the free end and attached to a fixed support at
the other end. The cross-sectional area varies linearly from A0
at the fixed support at x = 0 to A0/2 at x = L. Calculation of the
displacement parameter at the end of the bar was performed
considering the following:
(a) Exact analytical solution
(b) Using the meshfree methodology
The material property and load data considered are given by:
Length of the bar, L = 1 m
41
Area of cross-section at fixed end, A0 = 1 m
2
Area of cross-section at free end, A1= 0.5 m
2
Youngs modulus of Elasticity, E =
200*10
9
N/m
2
Point Load, P =
1000*10
3
N
The material properties considered in the referred original
problem are taken as unity and the plot obtained are given by
the Fig. 5; the figure has been taken directly from the source.
The plot obtained by considering the similar data of unit force
and material property and using the MATLAB code for
meshfree method is represented by Fig. 6 and three curve for
AKGEC JOURNAL OF TECHNOLOGY, vol., no.1
21 and 111 nodes along with the exact solution are plotted to
indicate the effect of number of nodes on the accuracy of the
solution.
Selection of dimensionless size of support domain: The
selection of the dimension of support domain affects the
accuracy of the approximate solution. This is one of the
selectable parameters which need to be selected appropriately
by the analyst to arrive at the correct and accurate solution.
This has to be different for the different classes or
complexities of the problems, from simple geometry to
Fig.4 . The tapered elastic bar subjected to an applied
tensile load P

Fig 5. Displacement solutions plot from the original problem
using FEM
Figure 6. Displacement solutions using meshfree method
complex shapes. The selection depends on the intuition and
the experience of the analyst otherwise fresher has to adopt
the trial and error methodology and verification from other
methods to arrive at the optimum choice. The other selectable
variables for the Element Free Galerkin method and present
study are given in table-1.
TABLE 1 SELECTABLE VARIABLES
S
N
Selectable
variable
Value
1 Weighting
Function
Quartic Spline:

'
+

0
, 3 8 6 1
) (
4 3 2
r r r
x x W
i

1
1
>

r
r
2 Basis Function One Dimensional Linear:
42
[ ] x x p
T
1 ) (
3 Integration One point Gauss-Quadrature
4 Prevention of
Singularity of
Moment matrix
Shifting of nodes
5 Construction of
shape function
Moving Least Square Method
6 Number of Nodes 21 and 111
To study the effect of variation of S, on the solution for the
nodal displacement parameters of the bar, with the above data,
the different values for the S were chosen and the plots for
the meshfree solution of the bar obtained which is shown by
the Fig. 5 along with the exact solution. It is clearly noticeable
that the value of S=2.5 gives a result in the form of rough
approximating curve whereas decrease in its value provides
better approximation to the curve of exact solution. The
various displacement values of the end of the bar are tabulated
in Table-2. The data obtained is represented by the fig. 6.
Fi
g 6. Effect of dimensionless size of support domain, S
IV. CONCLUSION
From variation of solution for displacement parameter at the
end of the bar with respect to S, presented by fig.7, it is clear
that there is an optimum value for the S where we can get the
best approximation to the exact solution.
In present case, it approximates to 2.1. Hence, the judgment,
knowledge, experience and intuition of the analyst plays vital
role in the selection of the variable parameters of the meshfree
methods for getting the significantly correct solution for the
problem in hand depending upon the complexity and its
nature.
Fig 7. Effect of dimensionless size of support domain, S

ELEMENT FREE GALERKIN METHOD
TABLE 2 DISPLACEMENT PARAMETER
S.
no.
Value of
Alpha
Displacement Parameter at
the end of bar
1 1.4 6.80700 microns
2 1.5 6.80700 microns
3 1.6 6.81733 microns
5 1.7 6.83977 microns
6 1.8 6.86420 microns
7 1.9 6.88305 microns
8 2.0 6.89325 microns
9 2.1 6.89540 microns
10 2.2 6.89181 microns
11 2.3 6.88476 microns
12 2.4 6.87596 microns
13 2.5 6.86658 microns
14
Exact
Solution
6.93147 microns
V. REFERENCES
[1] David V. Hutton, Fundamentals of finite element analysis,
Mc-Graw Hill, 2004.
[2] G.R. Liu, Mesh Free Methods: Moving beyond the finite
element method, CRC Press, 2003.
[3] J. Dolbow and T. Belytschko, An introduction to programming
the meshless element- free galerkin method, Archives in
Computational Mechanics, 5 (1998) 207-241.
43
[4] J.N. Reddy, An introduction to the finite element methods ,
McGraw-Hill, 2003
[5] J.S. Kushawaha, Mesh free analysis of elastic bar, Master
Thesis, Department of mechanical engineering, HBTI, Kanpur,
2009.
[6] T. Belytscho, Y.Y. Lu, L.Gu, Element-free galerkin methods,
International Journal for Numerical Methods in Engineering,
37(1994) 229-256.
[7] T. Belytscho, Y. Krongauz, D. Organ, M. Fleming, P. Krysl,
Meshless methods: An overview and recent developments,
Computer Methods in Applied Mechanics and Engineering, 139
(1996) 3-47.
[8] T.P. Fries, H.G. Matthies, Classification and overview of
meshfree methods, Scientific Computing, 2004.
[9] Vivek Varshney, Element free galerkin method for analysis of
reinforced granular beds on soft soil, Master Thesis,
Department of civil engineering, IIT, Kanpur, 2008.
Jeetender Singh Kushawaha is a
lecturer of mechanical engineering in the
Indus Institute of Technology and
Management, Kanpur. He completed his
AMIE from Institution of Engineers,
Kolkata and M-Tech from Harcourt
Butler Technological Institute, Kanpur.
He has eighteen years industrial
experience of aviation maintenance in
Indian Air Force. He has published a
number of research papers.
44

You might also like