You are on page 1of 8

Chin. Phys. B Vol. 22, No.

12 (2013) 120203

An improved interpolating element-free


Galerkin method for elasticity∗
Sun Feng-Xin(孙凤欣)a)c) , Wang Ju-Feng(王聚丰)a)b) , and Cheng Yu-Min(程玉民)a)†
a) Shanghai Institute of Applied Mathematics and Mechanics, Shanghai Key Laboratory of Mechanics in Energy Engineering,
Shanghai University, Shanghai 200072, China
b) Ningbo Institute of Technology, Zhejiang University, Ningbo 315100, China
c) Faculty of Science, Ningbo University of Technology, Ningbo 315016, China

(Received 30 March 2013; revised manuscript received 2 May 2013)

Based on the improved interpolating moving least-squares (IIMLS) method and the Galerkin weak form, an improved
interpolating element-free Galerkin (IIEFG) method is presented for two-dimensional elasticity problems in this paper.
Compared with the interpolating moving least-squares (IMLS) method presented by Lancaster, the IIMLS method uses the
nonsingular weight function. The number of unknown coefficients in the trial function of the IIMLS method is less than
that of the MLS approximation and the shape function of the IIMLS method satisfies the property of Kronecker δ function.
Thus in the IIEFG method, the essential boundary conditions can be applied directly and easily, then the numerical solutions
can be obtained with higher precision than those obtained by the interpolating element-free Galerkin (IEFG) method. For
the purposes of demonstration, four numerical examples are solved using the IIEFG method.

Keywords: meshless method, improved interpolating moving least-squares (IIMLS) method, improved inter-
polating element-free Galerkin (IIEFG) method, elasticity
PACS: 02.60.Cb, 02.60.Lj, 02.30.Em DOI: 10.1088/1674-1056/22/12/120203

1. Introduction meshless methods based on the MLS approximation, the es-


The finite element method (FEM) [1,2] and the boundary sential boundary conditions cannot be applied directly. [7–9,15]
element method (BEM) [3] are very successful numerical meth- Lancaster and Salkauskas [17] presented an interpolating
ods for science and engineering problems. However, they can- moving least-square (IMLS) method by using the singular
not be effectively applied to some complicated problems, such weight function. The shape function of the IMLS method
as extremely large deformations and crack growth problems. satisfies the property of the Kronecker δ function, and then
In recent years, more and more attention has been paid to the the meshless methods based on the IMLS method can used
research on the meshless (or meshfree) method. The most im- as the essential boundary condition directly without any ad-
portant common feature of meshless methods is that the trial ditional numerical effort. Kaljevic and Saigal [18] presented
function is constructed from a set of nodes with a minimum of an improved formulation of the element-free Galerkin (EFG)
meshing or no meshing at all. They show some advantages method based on the IMLS method. Ren et al. [19–22] revised
over FEM and BEM when solving the problems which are the formulae of the IMLS method, and presented an interpolat-
not well suited to traditional computational methods. Based ing element-free Galerkin (IEFG) method and an interpolating
on different approximation functions, many kinds of meshless boundary element-free (IBEF) method for potential and elas-
methods have been proposed. [4–15] ticity problems. In the IEFG and IBEF methods, the essential
The moving least-squares (MLS) approximation is one boundary conditions are applied directly.
of the most widely used methods to form the shape function A disadvantage of the IMLS method is that the weight
in the meshless method. The MLS approximation was intro- function used in it must be singular at nodes. The singular
duced first by Shepard [16] for constructing smooth approxima- weight function makes the matrix in the system equation sin-
tions to fit a specified cloud of points. It was then extended by gular, which causes many difficulties in obtaining the deriva-
Lancaster and Salkauskas [17] to surface generation problems. tives of the approximation function in the IMLS method. To
The shape function of the MLS approximation is simple, and be free from the requirement of the singularity, Netuzhylov [23]
the meshless method based on the MLS approximation can presented the perturbation technique in the IMLS method by
obtain a solution with high precision. [6–9] using a small positive number ε within the regularized weight
The shape function of the MLS approximation does not function matrix. Thomas and Christian [24] also presented a
satisfy the property of the Kronecker δ function. Then in special non-singular interpolating weighting function, then the
∗ Projectsupported by the National Natural Science Foundation of China (Grant No. 11171208) and the Shanghai Leading Academic Discipline Project, China
(Grant No. S30106).
† Corresponding author. E-mail: ymcheng@shu.edu.cn

© 2013 Chinese Physical Society and IOP Publishing Ltd http://iopscience.iop.org/cpb http://cpb.iphy.ac.cn

120203-1
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
MLS shape functions satisfy the interpolation condition ex- The function v(𝑥, 𝑥I ) obviously satisfies
actly.
To be free from the requirement of the singularity, Wang v(𝑥I , 𝑥J ) = δIJ , (5)
et al. [25] presented an improved interpolating moving least-
and
squares (IIMLS) method. The shape function of the IIMLS
n
method satisfies the property of the Kronecker δ function.
∑ v(𝑥, 𝑥I ) = 1, ∀x ∈ Ω . (6)
Thus the essential boundary condition can be applied directly I=1
and easily in meshless methods based on the IIMLS method.
Next, we will obtain the approximation function of the
Compared with the IMLS method presented by Lancaster and
new function ũ(𝑥)
¯ x by the MLS approximation with the new
Salkauskas, the weight function used in the IIMLS method
basis functions.
is not singular at any point. Any weight function used in
Define a local approximation function as
the MLS approximation can be applied in the IIMLS method.
Then the IIMLS method can overcome the difficulties caused m

by the singular weight function in the IMLS method. And the ũL (𝑥, 𝑥)
¯ = ∑ p̃i (𝑥)
¯ x 𝑎˜ i (𝑥), (7)
i=1
number of the unknown coefficients in the trial function of the
IIMLS method is less than that of the MLS approximation. where 𝑎˜ i (𝑥) are the unknown coefficients of the new basis.
Then we can select fewer nodes in the local influence domain From Eqs. (1) and (6), we have
in the IIMLS method, and high computational precision can n
be obtained. ¯ x = 1 − ∑ v(𝑥, 𝑥I ) = 0.
p̃1 (𝑥) (8)
I=1
In this paper, combining the IIMLS method with the
Galerkin weak form of elasticity problems, an improved in- Then
terpolating element-free Galerkin (IIEFG) method is presented m
for two-dimensional elasticity problems. In the IIEFG method, ũL (𝑥, 𝑥)
¯ = ∑ p̃i (𝑥)
¯ x 𝑎˜ i (𝑥). (9)
i=2
the essential boundary conditions are applied naturally and di-
rectly. Thus the IIEFG method gives a high computational The m − 1 coefficients 𝑎˜ i (𝑥), i = 2, 3, . . . , m, can be ob-
precision. For the purposes of demonstration, some selected tained by using the weighted least-square method.
numerical examples are solved using the IIEFG method. Define a functional
n
2. Improved interpolating moving least-squares J˜ = ∑ w(𝑥 − 𝑥I )[ũL (𝑥, 𝑥I ) − ũ(𝑥I )x ]2
I=1
method n  n
Let p1 (𝑥), p2 (𝑥), . . . , pm (𝑥) be the given basis functions, = ∑ w(𝑥 − 𝑥I ) ∑ v(𝑥, 𝑥I )u(𝑥I )
I=1 I=1
and p1 (𝑥) ≡ 1. In order to let the shape function of the IIMLS m 2
method, in which the nonsingular weight function is used, sat- + ∑ p̃i (𝑥I )x 𝑎˜ i (𝑥) − u(𝑥I ) , (10)
isfy the property of the Kronecker δ function, we first generate i=2

a new set of basis functions. then equation (10) can be written as


For a given point 𝑥 ∈ Ω , let
n J˜ = (𝑉 (𝑥)𝑢 + 𝑃˜ (𝑥)𝑎(𝑥)
˜ − 𝑢)T 𝑊 (𝑥)(𝑉 (𝑥)𝑢
p̃i (𝑥) ¯ − ∑ v(𝑥, 𝑥I )pi (𝑥I ),
¯ x = pi (𝑥) i = 1, 2, . . . , m, (1)
I=1 + 𝑃˜ (𝑥)𝑎(𝑥)
˜ − 𝑢), (11)
where 𝑥¯ is the point in the local approximation of 𝑥, 𝑥I
where
(I = 1, 2, . . . , n) are the nodes with domains of influence that
cover the point 𝑥, and 𝑎˜ T (𝑥) = (𝑎˜ 2 (𝑥), 𝑎˜ 3 (𝑥), . . . , 𝑎˜ m (𝑥)), (12)
ζ (𝑥, 𝑥I ) T
𝑢 = (u(𝑥1 ), u(𝑥2 ), . . . , u(𝑥n )), (13)
v(𝑥, 𝑥I ) = n , (2)  
∑ ζ (𝑥, 𝑥J ) w(𝑥 − 𝑥1 ) 0 ··· 0
J=1  0 w(𝑥 − 𝑥2 ) ··· 0 
∏ ||𝑥 − 𝑥J ||2 𝑊 (𝑥) =   , (14)
 
.. .. .. ..
J6=I  . . . . 
ζ (𝑥, 𝑥I ) = . (3)
∏ ||𝑥I − 𝑥J ||2 0 0 · · · w(𝑥 − 𝑥n )
J6=I  
p̃2 (𝑥1 )x p̃3 (𝑥1 )x ··· p̃m (𝑥1 )x
Applying the same transformation to u(𝑥), we have  p̃2 (𝑥2 )x p̃3 (𝑥2 )x ··· p̃m (𝑥2 )x 
𝑃˜ (𝑥) =  , (15)
 
n .. .. .. ..
. . . .
¯ − ∑ v(𝑥, 𝑥I )u(𝑥I ).
 
ũ(𝑥)
¯ x = u(𝑥) (4)
I=1 p̃2 (𝑥n )x p̃3 (𝑥n )x ··· p̃m (𝑥n )x
120203-2
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
and 𝑉 (𝑥) is an n × n matrix in which each row is 3. The improved interpolating element-free
Galerkin method for elasticity
𝑣(𝑥) = (v(𝑥, 𝑥1 ), v(𝑥, 𝑥2 ), . . . , v(𝑥, 𝑥n )). (16)
Based on the IIMLS method and the Galerkin weak form,
To obtain 𝑎˜ i (𝑥), we let an interpolating element-free Galerkin (IIEFG) method is pro-
∂ J˜ posed for two-dimensional linear elasticity problems.
˜ 𝑎(𝑥)
= 𝐴(x) ˜ ˜
− 𝐵(𝑥)𝑢 = 0, (17) We employ nt nodes in the domain Ω , and the union of
∂ 𝑎˜
its compact support domains must cover the whole domain Ω .
where
For elasticity problems, from the IIMLS method, the displace-
˜
𝐴(𝑥) = 𝑃˜ T (𝑥)𝑊 (𝑥)𝑃˜ (𝑥), (18) ment at an arbitrary point x in the domain can be expressed
˜
𝐵(𝑥) = 𝑃˜ T (𝑥)𝑊 (𝑥)(𝐸 − 𝑉 (𝑥)), (19) as

and 𝐸 is an n × n identity matrix. 𝑢(𝑥) = (u1 (𝑥), u2 (𝑥))T , (27)


From Eq. (17) we have
and
𝑎(𝑥)
˜ = 𝐴˜ −1 (𝑥)𝐵(𝑥)𝑢.
˜ (20) n
u1 (𝑥) = ∑ Φ̃I (𝑥)u1 (𝑥I ), (28)
Then the local approximation function can be obtained as I=1
n
m u2 (𝑥) = ∑ Φ̃I (𝑥)u2 (𝑥I ), (29)
ũL (𝑥, 𝑥)
¯ = ∑ p̃i (𝑥)
¯ x 𝑎˜ i (𝑥) = 𝑝( ¯ x 𝐴˜ −1 (𝑥)𝐵(𝑥)𝑢,
ˆ 𝑥) ˜ (21) I=1
i=2
where n is the number of nodes whose compact support do-
where mains cover the point 𝑥.
𝑝(
ˆ 𝑥)
¯ x = ( p̃2 (𝑥)
¯ x , p̃3 (𝑥)
¯ x , . . . , p̃m (𝑥)
¯ x ). (22) Then it follows from Eq. (27) that

From Eqs. (4) and (21), the global interpolating approxi- 𝑢 = 𝑁 (𝑥)𝑈 , (30)
mation function of u(𝑥) is obtained as
where
n
˜
uh (𝑥) = 𝛷(𝑥)𝑢 = ∑ Φ̃I (𝑥)u(𝑥I ), (23) 𝑈 = (u1 (𝑥1 ), u2 (𝑥1 ), u1 (𝑥2 ), u2 (𝑥2 ), . . . ,
I=1
u1 (𝑥n ), u2 (𝑥n ))T , (31)
˜
where 𝛷(x) is the shape function vector,
𝑁 (𝑥) = (N1 (𝑥), N2 (𝑥), . . . , Nn (𝑥)), (32)
 
˜
𝛷(𝑥) = (Φ̃1 (𝑥), Φ̃2 (𝑥), . . . , Φ̃n (𝑥)) Φ̃I (𝑥) 0
𝑁I (𝑥) = . (33)
0 Φ̃I (𝑥)
= 𝑣(𝑥) + 𝑝˜ T (𝑥)𝐴˜ −1 (𝑥)𝐵(𝑥), ˜ (24)
The strain can be expressed as
and  
u1,1
𝑝˜ T (𝑥) = (g2 (𝑥), g3 (𝑥), . . . , gm (𝑥)), (25) ε(𝑥) =  u2,2 
n u1,2 + u2,1
gi (𝑥) = pi (𝑥) − ∑ v(𝑥, 𝑥I )pi (𝑥I ). (26)  n 
I=1 Φ̃I,1 (𝑥)u1 (𝑥I )

 I=1 
Equation (24) is the shape function of the IIMLS method,  n 
= ∑ Φ̃I,2 (𝑥)u2 (𝑥I )
 
and then the IIMLS method is presented.  I=1


 n n
The shape function of the IIMLS method can satisfy the

∑ Φ̃I,2 (𝑥)u1 (𝑥I ) + ∑ Φ̃I,1 (𝑥)u2 (𝑥I )
property of the Kronecker δ function. Compared with the I=1 I=1

IMLS method presented by Lancaster and Salkauskas, the ≡ 𝐵(𝑥)𝑈 , (34)


nonsingular weight function is used in the IIMLS method.
where
Then any weight function used in the MLS approximation can
also be applied in the IIMLS method. 𝐵(𝑥) = (𝐵1 (𝑥), 𝐵2 (𝑥), . . . , 𝐵n (𝑥)), (35)
 
From Eq. (9), it can be seen that the number of the un- Φ̃I,1 (𝑥) 0
known coefficients in the trial function of the IIMLS method 𝐵I (𝑥) =  0 Φ̃I,2 (𝑥)  . (36)
is less than that in the trial function of the MLS approximation. Φ̃I,2 (𝑥) Φ̃I,1 (𝑥)
Therefore we can select fewer nodes in the meshless method Then the stress-strain relation is
based on the IIMLS method than those based on the MLS ap-
proximation. 𝜎(𝑥) = 𝐷 · 𝜀(𝑥) = 𝐷𝐵(𝑥)𝑈 , (37)
120203-3
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
where Since the shape function of the IIMLS method satisfies
  the property of the Kronecker δ function, the essential bound-
1 ν 0
E  ary conditions can be applied directly. And then, substitut-
𝐷= ν 1 0  (38)
1 − ν2 ing the boundary conditions of Eq. (41) into Eq. (46), we
0 0 (1 − ν)/2
can obtain the unknowns at nodes by solving the linear equa-
for a plane stress problem, and tions (46).

1−ν ν 0
 Thus the IIEFG method is presented for two-dimensional
E
𝐷=  ν 1−ν 0  (39) elasticity problems.
(1 + ν)(1 − 2ν)
0 0 (1 − 2ν)/2

for a plane strain problem, E is the Young’s modulus and ν is 4. Numerical examples
the Poisson’s ratio. Four example problems are presented to demonstrate the
We consider the following two-dimensional problem on applicability of the IIEFG method for two-dimensional elas-
the domain Ω bounded by Γ : ticity problems. The linear basis function and the cubic spline
weight function are used in these examples. The results of the
∇ · 𝜎 + 𝑏 = 0, in Ω , (40) IIEFG method are compared with the existing analytical re-
sults and the ones of the IEFG method or ABAQUS software.
where 𝜎 is the stress tensor and 𝑏 is the body force vector.
The first example considered is a Timoshenko beam sub-
The corresponding boundary conditions are
jected to a parabolic traction at the free end as shown in Fig. 1.
𝑢 = 𝑢0 , on Γu , (41) The beam is of length L and height D, and has a unit thickness.
The beam is assumed to be in a state of plane stress. This is a
𝑡 = 𝜎 · 𝑛 = 𝑡0 , on Γσ , (42)
benchmark problem and the corresponding analytical solution
where 𝑡 is the traction vector, 𝑢0 and 𝑡0 are respectively the is available. [26]
prescribed displacement and traction vector on the displace-
x
ment boundary Γu and on the traction boundary Γσ , and 𝑛 is
the unit outward normal to the boundary Γ .
The Galerkin weak form of Eqs. (40)–(42) is O D x
Z Z Z p
𝜎 T δ ε dΩ − 𝑏T δ 𝑢dΩ − 𝑡T δ 𝑢dΓ = 0, (43) L
Ω Ω Γσ

where δ is the variational operator. Fig. 1. Cantilever beam.


Substituting Eqs. (30), (34), and (37) into Eq. (43) yields
Z Z The problem is solved with E = 3.0 × 107 Pa, v = 0.3,
T T
(𝐷𝐵(𝑥)𝑈 ) δ (𝐵(𝑥)𝑈 )dΩ − 𝑏 δ (𝑁 (𝑥)𝑈 )dΩ L = 48 m, D = 12 m, and p = 1000 N.
Ω Ω
Z Regular nodal distribution as shown in Fig. 2 is employed.
− 𝑡T δ (𝑁 (𝑥)𝑈 )dΓ = 0, (44) To evaluate the stiffness matrix, 20 × 8 background integration
Γσ
cell structure is used, and 4 × 4 Gauss quadrature is employed
i.e.,
in each integration cell.
Z Z
T
(δ 𝑈 ) 𝐵 T (𝑥)𝐷𝐵(𝑥)𝑈 dΩ − 𝑁 T (𝑥)𝑏dΩ x
Ω Ω
Z 
T x
− 𝑁 (𝑥)𝑡dΓ = 0. (45)
Γσ O

Because the nodal test function δ 𝑈 is arbitrary, the final


discretized equation is obtained as Fig. 2. Node distribution of the cantilever beam.

𝐾𝑈 = 𝑓 , (46)
The numerical solutions of the deflection of the beam
where along the x1 axis by the IIEFG and IEFG methods are shown
Z in Fig. 3. It is evident that the numerical results of this pa-
𝐾IJ = 𝐵IT 𝐷𝐵J dΩ , (47) per are in excellent agreement with the analytical results. The

Z Z stresses σ11 and σ12 using the IIEFG and IEFG methods at the
𝑓I = 𝑁IT 𝑏dΩ + 𝑁IT 𝑡dΓ . (48) section of x1 = L/2 are shown in Figs. 4 and 5. The results of
Ω Γt
120203-4
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
this paper also accord well with the analytical results and the to be the horizontal distance between the nodes in the model,
IIEFG method has higher accuracy than the IEFG method. and in each case dmax = 2. The present results show that the
For the error analysis, define the following energy norm: convergence of the IIEFG method is good.
r Z
1
eε = (ε num − ε exact )T D(ε num − ε exact )dΩ , (49)
2 Ω
-0.2
where ε num and ε exact are respectively the numerical and ana-
lytical solutions of the strain tensor. -0.3
IIEFG
-0.4

lg(eε)
IEFG
-0.5
0

-2 -0.6
u/10-3 m

-4 -0.7

analytical 0.5 0.6 0.7 0.8


-6
IIEFG lg(h)
-8 IEFG
Fig. 6. Convergence of energy error norm.
-10
0 10 20 30 40 50
x/m The second example is a perforated plate with a central
Fig. 3. Displacement u2 along the x1 axis. circular hole as shown in Fig. 7. The plate is subjected to a
distributed load p = 1000 N/m in the x1 -axis direction. Sup-
pose that the plate is in the plane stress state, and ν = 0.25,
1000 E = 2.0 × 107 Pa. When the condition b/a ≥ 5 is satisfied,
analytical the solution of this finite plate is very close to that of the
500
IIEFG infinite plate. The analytical solution of an infinite plate is
σ/Pa

IEFG
0 available. [26]

-500

-1000

-6 -4 -2 0 2 4 6
x/m
p 10 m p
Fig. 4. Stress σ11 at x1 = L/2.

a=2 m

0
b=10 m
-20

-40 Fig. 7. A rectangular plate with a hole under a distributed load.


analytical
σ/Pa

-60
IIEFG
-80 IEFG Due to the symmetry of the model, only a quarter of the
-100 model needs to be considered in the analysis (see Fig. 8). The
-120 nodal distribution shown in Fig. 9 is employed.
-140
The values of displacement ur at θ = 0 and π/2 by the
-6 -4 -2 0 2 4 6 IIEFG and IEFG methods are respectively shown in Figs. 10
x/m
and 11. And the stresses σθ at θ = π/2 are shown in Fig. 12.
Fig. 5. Stress σ12 at x1 = L/2. Again, the IIEFG method has higher accuracy than the IEFG
method.
To study the convergence, the regular 13 × 7, 17 × 9, The third example considered is a circular ring with the
21 × 11, and 25 × 13 nodes are employed. Then the rates of radii b = 5 m and a = 1 m (see Fig. 13). The distributed inner
convergence in energy norm for this problem using the IIEFG pressure is p = 1000 N/m. The ring is in the plane stress state,
and IEFG methods are shown in Fig. 6. The value h is chosen and ν = 0.25, E = 106 Pa.
120203-5
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
x

3000

2500
q analytical

σ/Pa
2000 IEFG
IIEFG
1500
x
O
1000

Fig. 8. A quarter of the rectangular plate and the boundary conditions. 1 2 3 4 5


x/m
x
Fig. 12. Stress σ11 at θ = π/2.

a b

O   x

Fig. 9. The nodes in the domain of the rectangular plate with a hole. Fig. 13. A circular ring under a distributed inner pressure.

3.0
2.8 Due to the symmetry, only a quarter of the model needs
2.6 to be considered (see Fig. 14), and the node distribution shown
ur/10-4 m

2.4 analytical in Fig. 15 is used.


IIEFG
2.2
IEFG The values of displacement ur at θ = 0 by the IIEFG and
2.0
IEFG methods are all shown in Fig. 16. And the values of
1.8
stress σθ at θ = 0 are shown in Fig. 17. It can also be seen
1.6
1.4
that the results obtained using the IIEFG method show excel-
1 2 3 4 5 lent agreement with the analytical results.
x/m

Fig. 10. Displacement ur at θ = 0.


x

-5.0

-5.5
ur/10-5 m

-6.0

-6.5
analytical
-7.0 IIEFG
IEFG
-7.5

-8.0 O x
1 2 3 4 5
x/m

Fig. 11. Displacement ur at θ = π/2. Fig. 14. The numerical model of the circular ring.

120203-6
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
x
p

20 m

30 m


10 m
O   x

Fig. 15. Node distribution. 10 m 20 m 10 m

Fig. 18. A bridge pier.


14
x
12
analytical B C
ur/10-4 m

10 IIEFG
IEFG
8

4
1 2 3 4 5
r/m A
Fig. 16. Displacement ur at θ = 0.
x
1200
O E D
1000
Fig. 19. Node distribution.
800
analytical
σθ/Pa

600 IIEFG
IEFG
400 0

200 -1 IIEFG
u/10-5 m

0 ABAQUS
1 2 3 4 5 -2
r/m
-3
Fig. 17. Stress σθ at θ = 0.

The fourth example considered is a bridge pier as shown -4

in Fig. 18. [27] Suppose that it is subjected to a distributed load


-5
p = 50 kN/m on the top of the pier. The bridge is in the plane 0 5 10 15 20 25 30
x/m
strain state with ν = 0.15, E = 40 GPa. Considering the sym-
metry of the bridge, only the right half of the pier is modeled Fig. 20. Displacement u2 on curve EAB.
as shown in Fig. 19.
The node arrangement used by the IIEFG method is
5. Conclusions
shown in Fig. 19. The problem is also analyzed by FEM soft-
ware ABAQUS. The displacements of the IIEFG method and Based on the IIMLS method and the Galerkin weak form,
ABAQUS on boundary curves EAB are shown in Fig. 20. The the IIEFG method for two-dimensional plasticity problems is
results obtained by the present method are in very good agree- presented in this paper. The shape function of the IIMLS
ment with those obtained by using ABAQUS. method satisfies the property of the Kronecker δ function.
120203-7
Chin. Phys. B Vol. 22, No. 12 (2013) 120203
Then in the IIEFG method, the essential boundary condition [6] Cheng Y M and Peng M J 2005 Sci. China Ser. G: Phys. Mech. Astron.
48 641
can be applied easily and directly without any additional nu-
[7] Yang X L, Dai B D and Zhang W W 2012 Chin. Phys. B 21 100208
merical effort. Compared with IMLS method presented by [8] Cheng Y M, Li R X and Peng M J 2012 Chin. Phys. B 21 090205
Lancaster and Salkauskas, the IIMLS method uses the nonsin- [9] Belytschko T, Lu Y Y and Gu L 1994 Int. J. Numer. Meth. Eng. 37 229
[10] Liu W K, Jun S and Zhang Y F 1995 Int. J. Numer. Meth. Eng. 20 1081
gular weight function. Then the IIMLS method can overcome
[11] Liu G R and Gu Y T 2004 Eng. Anal. Bound. Elem. 28 475
the difficulty caused by the singularity of the weight function [12] Zhang J M and Tanaka M 2004 Int. J. Numer. Meth. Eng. 41 1147
in the IMLS method. Any weight function used in the MLS [13] Zhang J M and Tanaka M 2008 Comput. Mech. 41 777
[14] Bai F N, Li D M, Wang J F and Cheng Y M 2012 Chin. Phys. B 21
approximation can be chosen as the weight function of the 020204
IIMLS method. And the number of unknown coefficients in [15] Liew K M, Ren J and Reddy J N 2005 Int. J. Numer. Meth. Eng. 63
1014
the trial function of the IIMLS method is less than that of the
[16] Shepard D 1968 Proceeding of the 23rd ACM National Conference
MLS approximation. Hence, the IIEFG method gives high (New York) p. 517
computational precision. [17] Lancaster P and Salkauskas K 2007 Math. Comput. 37 141
[18] Kaljevic I and Saigal S 1997 Int. J. Numer. Meth. Eng. 40 2953
[19] Ren H P, Cheng Y M and Zhang W 2009 Chin. Phys. B 18 4065
References [20] Ren H P, Cheng Y M and Zhang W 2010 Sci. China Ser. G: Phys. Mech.
Astron. 53 758
[1] Liu Z S, Harsono E and Swaddiwudhipong S 2009 Int. J. Appl. Mech. [21] Ren H Pand Cheng Y M 2011 Int. J. Appl. Mech. 3 735
1 61
[22] Ren H P and Cheng Y M 2012 Eng. Anal. Bound. Elem. 36 873
[2] Liu Z S, Hong W, Suo Z G, Swaddiwudhipong S and Zhang Y W 2010 [23] Netuzhylov H 2008 Eng. Anal. Bound. Elem. 32 512
Comput. Mater. Sci. 49 60
[24] Thomas M and Christian B 2008 Eng. Anal. Bound. Elem. 32 461
[3] Brebbia C A and Wrobel L C 1984 Boundary Element Techniques: [25] Wang J F, Sun F X and Cheng Y M 2012 Chin. Phys. B 21 090204
Theory and Applications in Engineering (Berlin: Springer-Verlag) [26] Timoshenko S P and Goodier J N 1970 Theory of Elasticity, 3rd edn.
[4] Chen S S, Li Q and Liu Y 2012 Chin. Phys. B 21 110207 (New York: McGraw-Hill)
[5] Cheng R J and Ge H X 2012 Chin. Phys. B 21 100209 [27] Liu G R and Gu Y T 2001 Int. J. Numer. Meth. Eng. 50 937

120203-8

You might also like