Académique Documents
Professionnel Documents
Culture Documents
Department of Civil Engineering, University of Siegen, Paul-Bonatz-Strae 9-11, 57076 Siegen, Germany
Piping Department, Petrovietnam Engineering Company, Ho Chi Minh, Vietnam
c
Department of Civil Engineering, Lac Hong University, Dong Nai Province, Vietnam
b
art ic l e i nf o
a b s t r a c t
Article history:
Received 3 April 2013
Received in revised form
13 February 2014
Accepted 19 February 2014
Available online 12 March 2014
An efcient, smooth and accurate quadrilateral element with four-node based on the consecutiveinterpolation procedure (CIP) is formulated. The CIP is developed recently by Zheng et al. (Acta Mech Sin
26 (2010) 265278) for triangular element with three-node. In this setting the approximation functions
handle both nodal values and averaged nodal gradients as interpolation conditions. Two stages of the
interpolation are required; the primary stage is carried out using the same procedure of the standard
nite element method (FEM), and the interpolation is further reproduced in the secondary step
according to both nodal values and average nodal gradients derived from the previous interpolation.
The new consecutive-interpolation quadrilateral element with four-node (CQ4) deserves many desirable
characteristics of an efcient numerical method, which involves continuous nodal gradients, continuous
nodal stresses without smoothing operation, higher-order polynomial basis, without increasing the
degree of freedom of the system, straightforward to implement in an existing FEM computer code, etc.
Four benchmark and two practical examples are considered for the stress analysis of elastic structures in
two-dimension to show the accuracy and the efciency of the new element. Detailed comparison and
some other aspects including the convergence rate, volumetric locking, computational efciency,
insensitivity to the mesh, etc. are investigated. Numerical results substantially indicate that the
consecutive-interpolation nite element method (CFEM) with notable features pertains to high accuracy,
convergence rate, and efciency as compared with the standard FEM.
& 2014 Elsevier B.V. All rights reserved.
Keywords:
FEM
Consecutive-interpolation nite element
Stress analysis
Numerical methods
Quadrilateral element
1. Introduction
Design procedures of improving and enhancing the performance of engineering structures through stress analysis are often
time-consuming and expensive. Nowadays, simulation technologies using advanced numerical methods in engineering and
science are popular and have been emerged rapidly. The motivations are to accurately model practical problems as exact as the
techniques can. The nite element method (FEM) [13] and the
boundary element method (BEM) [4] have become very powerful
and versatile numerical methods, which are the most common
and extensively used methods in a broad range of engineering
applications. Owing to the simplicity, the three-node triangular
and four-node quadrilateral nite elements are often introduced
and applied to solve engineering problems in two-dimensions
(2D). Because of the linear approximations, the spatial derivatives
of the eld variables are constant within each element [5]. Such
http://dx.doi.org/10.1016/j.nel.2014.02.004
0168-874X & 2014 Elsevier B.V. All rights reserved.
constant-strain nite elements are easily formulated and implemented but their performance in practical applications is often
unsatisfactory and, frequently low accuracy is obtained due to
their low-order trial functions [5,6]. Moreover, the gradients on
element-edges in both constant elements and mapped elements
are discontinuous, and demanding smoothing operation in postprocessing step is rigorous [7]. Other relevant issues involving
volumetric locking and sensitivity to mesh, etc. for such elements
can be found in Refs. [13,58] for instance.
A number of advanced numerical methods have been developed in order for improving the accuracy and efciency of the
conventional FEM methods. For instance, Hansbo proposed a nonconforming rotated Q1 tetrahedral element for linear elastic [9]
and elastodynamic problems [10]. By containing the bilinear
terms, the Q1 element performs substantially better than the
standard constant-strain one in bending and allows for underintegration in nearly incompressible situations. Papanicolopulos and Zervos [11] presented a means for creating a class of
triangular C1 nite element particularly suitable for modeling problems where the underlying partial differential equation
is of fourth-order (e.g., beam and plate bending, deformation of
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
15
sij;j bi 0 in
where bi denotes the components of the body force vector and sij
is the stress tensor. The balance equations, Eq. (1), satisfy the
following boundary conditions:
sij nj t i on t and ui ui on u
where Dijkl is the elasticity tensor and s uij denotes the symmetric part of the displacement gradients i.e., s uij ui;j uj;i =2.
In the FEM we approximate solutions to Eq. (3) by dividing
and the boundary into small elements, and the interpolation is
then determined by approximating the displacement eld in each
element. The element stiffness matrix is derived and it is then
assembled into the global stiffness matrix [13]. Generally, this
step is accomplished almost identically for the FEM and the CFEM.
Now, we start describing the CIP procedure in a general and
brief way. In the subsequent sections the formulation of the novel
CQ4 element with four-node is detailed. For the sake of brevity,
the following presentation will be focused on the displacement
component u1 u only. The function ux with x fx; yg in 2D in
the FEM can be approximated by
n
ux N i xdi Nxd
i1
ui
;x N;x d
where
i
N;x
i
5
are the averaged derivative of Ni , and calculated by
N;x we U Nie
;x
e A Si
16
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
i1
i1
i
~
ux
i ui ix ui
;x iy u;y
i
i
i Ni ix N;x iy N;y d
or
~
~
ux
Nxd
" x
9
i1
r
x
s
y #
r
y
s
Li
r
4 L
Lj
r
Lj
s
Lk
r
Lk
s
8
x
>
3>
> i
Lm >
< xj
r
5
Lm > x
>
s
> k
>
: xm
9
yi >
>
>
>
yj =
yk >
>
>
>
ym ;
13
14
where
12
with
J1
Ni Li ;
N k Lk ;
N m Lm
16
11
N j Lj ;
e
;
e~ A Si e~
with eA Si
17
Fig. 1. A quadrilateral element in (a) global coordinate and (b) natural coordinate systems.
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
17
ux N l xdl
18
l1
l1
ns
ie
ui
;y we u;y
e A Si
l1
e A Si
"
we N ie
l;y
e A Si
#
dl ;
19
or in compact form
ns
ui
;y
l1
i
20
e A Si
ix;y xl 0;
iy xl 0;
iy;x xl 0;
iy;y xl il ;
i;y xl 0;
23
24
i
N l;x we N ie
;
l;x
ix;x xl il ;
where
i
N l;y
i;x xl 0;
ix xl 0;
i
N l;y dl
e A Si
i xl il ;
ui
;x N l;x dl ;
l1
ns
condition)
we N ie
l;y
21
25
i
ix ui
;x iy u;y
j
j
jx uj
;x jy u;y
~
ux
i u
j u
|{z} |{z}
related to nodei
k
k
kx uk
;x ky u;y
m
mx um
;x my u;y
k u
m u
|{z} |{z}
node k
26
l1
node j
m
~
~
ux
N~ l xdl Nxd
22
node m
node j
k
m
m
k N k
kx N l;x ky Nl;y m Nm
mx N l;x my N l;y
l
l
|{z}
|
{z}
node k
27
node m
18
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
In Eq. (28), the domain e is different between the FEM and the
CFEM, which fully depends on the supporting nodes and neighboring elements determined by the CFEM procedure. The matrix of
the derivative of the shape functions Be is also different between
the FEM and the CFEM. Obviously, the difference can be depicted
explicitly, for instance, let us consider a quadrilateral element with
Fig. 3. Comparison of the shape functions (a) and their rst-order derivatives (b) in 1D classical and consecutive FEM. Visualization of the shape functions: Q4 (c) and CQ4
(d) elements.
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
N j
y
N k
y
N j
x
Ni
x
N m
y
N k
x
N m
x
N~ 1
6 x
6
BCFEM
6
e
6 0
4 ~
N 1
y
N~ 2
x
N~ l
x
4. Numerical examples
N~ ns
x
N~ 2
y
N~ l
y
N~ 2
x
N~ l
x
N~ 1
y
N~ 2
y
N~ l
y
N~ ns
y
N~ 1
x
0
N~ ns
y
N~ n
x
7
7
7
7
5
32ns
30
In Eq. (30), ns is the number of the supporting nodes, 1 ol ons
and ns 4 4. It is because that the displacements in the CFEM are
not only interpolated from the nodal displacements of the considered element, but also interpolated from the supporting nodes
determined by other neighboring elements, as sketched in Fig. 2.
The size of the matrix BFEM
is of 3 8 and that is much smaller
e
than the size of the matrix BCFEM
, 3 2ns , which results in an
e
increased bandwidth of the stiffness matrix of the CFEM.
In the above equations the derivative of the CFEM shape
functions is calculated by
N~ l i i ix i iy i j j jx j jy j
N
N
N
N
N
N
z
z l
z l;x z l;y |{z}
z l
z l;x z l;y
|{z}
related to node i
In this section four benchmark numerical examples are examined to show the accuracy, the efciency and the convergence of
the present CQ4 element. In addition, other two examples are
considered as practical applications to illustrate the applicability of
the proposed CQ4 element in dealing with complex geometries.
The materials used for all the four benchmark examples are
assumed to be linear elastic with Young's modulus E 1000 and
Poisson's ratio 0:3, whereas they are specied thereafter for
the two practical examples. The units used in the examples can be
any consistent unit based on the international standard unit
system, if not specied otherwise. To accomplish the convergence
study, two error indicators with respect to the displacement and
energy norms are dened as follows: [19,20]
v
u n
u
uexact unumer
2
i
ed t i 1 i
;
32
2
n
exact
i 1 ui
ee
node j
k k kx k ky k m m mx m my m
N
N
Nl;x
N
N
N
z l
z{z
z l;y} |{z}
z l
z l;x
z l;y
|
node k
accomplished by considering the rst six two-dimensional Gaussian quadrature rules from 1 1 to 6 6 respectively, and the two
error indicators in the displacement and energy norms, as well as
the maximum and minimum values of the von Mises stresses will
be investigated accordingly.
38
19
node m
31
with z x; y.
In order to perform the numerical integration of the stiffness
matrix of the CFEM method, we adopt a set of 3 3 Gaussian
quadrature points for all the implementations throughout the
study. It is because, in general, the implementation of the
numerical integration in the CFEM is realized similar to that in
the conventional FEM. None of any special techniques is required
for the numerical integrations of the CQ4 element. Any quadrature
rules used for the FEM can be applied the same for the CFEM, but it
may obey the relation of the number of Gaussian points ngp with
respect to the polynomial of order p as indicated in Ref. [2], i.e.,
p r 2ngp 1. Nonetheless, the variation and the inuence of the
Gauss quadrature rules on the accuracy of the solutions shall be
studied numerically in the numerical example part. This is
1
A
Z
1
exact numer T Dexact numer d;
2
33
Fig. 4. Domain discretization of a unit square patch test using the CQ4 elements: regular (a) and irregular (b) meshes.
20
uy y
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
34
sx PL I xy;
sy 0;
xy
"
#
P D2
y2
2I 4
36
the
the
the
the
Fig. 6. Three discretized regular (a) and distorted (b) meshes using quadrilateral elements for a cantilever beam.
Fig. 7. Comparison of the convergence rates for a cantilever beam using the regular meshes obtained by the CQ4, CT3, Q4 and T3 elements: relative errors in displacement
norm (a) and energy norm (b).
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
21
Fig. 8. Comparison of the convergence rates for a cantilever beam using the regular and distorted meshes obtained by the present CQ4 element: relative errors in
displacement norm (a) and energy norm (b).
Fig. 9. Comparison of the deections along the neutral line of a cantilever beam
obtained by the analytical, regular and irregular CQ4, CT3, Q4 and T3 solutions.
Fig. 10. Comparison of the shear stress distributions along the line (x L/2) of a
cantilever beam obtained by the analytical, regular and irregular CQ4, CT3, Q4 and
T3 solutions.
components obtained by the CQ4 and CT3 are all in good agreement with the analytical ones, whereas less accuracy can be found
for the Q4 and T3 elements. For a better representation, the
normal and shear stress distributions (e.g., sx , xy ) obtained by
the CQ4, Q4, T3 and SQ4 elements are depicted in Fig. 11.
Obviously, the stresses achieved by the Q4 and T3 elements are
discontinuous and non-smooth whilst the developed CQ4 element
works well, i.e., the stresses are continuous and smooth. It should
be recalled that the developed CFEM is very smooth though no
post-processing is performed.
In Fig. 11, the SQ4 element is the standard Q4 but its nal
results (see Fig. 11d and h) are smoothed out by further applying a
smoothing stress recovery technique in the post-processing step.
These SQ4 results are motivated since it may be interesting to see
how smooth on the stresses obtained by the present CFEM method
and the one using the stress recovery technique as usually done in
practice. By accomplishing that, we merely adopt one of the
simplest stress recovery techniques that have been found most
useful in practice, the averaged nodal stresses [22]. The realization
of this so-called unweighted averaging is carried out by assigning
the same weight to all elements that meet at a node. It is obvious
that the SQ4 results are as smooth as the CQ4 ones.
To check the ability of the present CFEM in treating the volume
locking phenomenon in the incompressible materials (i.e.,
Poisson's ratio tends toward 0.5), the same problem but the beam
22
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
Fig. 11. Comparison of the stress distributions in the cantilever beam obtained by the CQ4 (a, e); Q4 (b, f); T3 (c, g) and SQ4 (d, h) elements: normal stress (a)(d) and shear
stress (e)(h).
Fig. 12. Comparison of the deections along the neutral line of a cantilever beam
obtained by the analytical, regular and irregular CQ4, CT3, Q4 and T3 solutions with
0:49 for nearly incompressible materials.
Fig. 13. Comparison of the computational time for a cantilever beam subjected to a
parabolic traction at the free-end obtained by different approaches.
convergence, or issues related to pressure modes, infsup satisfaction, etc. are not studied and do not cover in this manuscript
because of simplicity. However, they are scheduled as our future
research works. We see in the picture that the CFEM performs well
as compared with the exact deection, and it shows higher
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
accuracy than the FEM. More importantly, based on these preliminary results it may be stated that the nearly incompressible
materials can be treated by the CFEM with high accuracy, and the
volumetric locking is alleviated.
Another important issue in relevance to the computational
efciency of the developed CQ4 element is also studied. The
computational time needed for the T3, Q4, CT3 and CQ4 elements
tested on three different meshes of 9 5, 16 10 and 30 20
elements, is investigated. Here, only the time required for the
computation of the global stiffness matrix is measured and
estimated. It is because that the most difference among the
aforementioned methods is induced by the implementation of
the stiffness matrix, which substantially pertains to the difference
of establishing the shape functions and their derivatives. The
comparison is performed on the same PC of Intel(R) Pentium(R)
Dual-Core 2.6 GHz, 2.GB RAM. 5 calculations are carried out for
each mesh and the averaged computational time is then reported.
Fig. 13 shows the required CPU time of different approaches using
the same direct solver. Because of an extra task of the consecutiveinterpolation implementation of the shape functions and their
derivatives, the CQ4 and CT3 elements obviously require more
time than the standard Q4 and T3 ones. On the other hand, the
computational efciency in terms of the relative errors in both the
displacement and energy norms against the computational time
(in seconds) is compared and depicted in Fig. 14 under a loglog
plot. It is evident that with respect to the computational efciency
(computational time for the same accuracy) the CFEM is more
efcient as clearly shown in the relative error results in comparison with the FEM. From the practical point of view, it should be
23
Fig. 14. Comparison of the computational efciency in terms of displacement (a) and energy (b) error norms for a cantilever beam subjected to a parabolic traction at the
free-end obtained by different approaches.
Table 1
Variation and the effect of the Gaussian quadrature rules on the numerical results of the cantilever beam.
Mesh
16 8
smax
smin
ed
ee
32 16
smax
smin
ed
ee
11
22
33
44
55
66
24.0883
1.0232
0.0169
0.0188
23.2061
0.8472
0.0021
0.0446
23.1883
0.8562
0.0019
0.0434
23.1863
0.8563
0.0019
0.0442
23.1863
0.8563
0.0019
0.0442
23.1863
0.8563
0.0019
0.0442
23.818
0.5771
0.0042
0.0051
23.5251
0.4586
0.00027
0.0161
23.5174
0.4543
0.00025
0.0155
23.5165
0.4543
0.00026
0.0158
23.5165
0.4543
0.00026
0.0158
23.5165
0.4543
0.00026
0.0158
24
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
Tx
a2
a4
1 r 3 sin 2;
4
r
r
sx T x 1
37
a2 3
3a4
cos 2 cos 4 4 cos 4 ;
2
r 2
2r
Fig. 15. Geometry of an innite plate with a central circular hole (a) and its quarter model (b).
Fig. 16. Comparison of the convergence rates for an innite plate with a circular hole obtained by the CQ4, CT3, Q4 and T3 elements: relative errors in displacement norm
(a) and energy norm (b).
Fig. 17. Comparison of the stress distributions along the left boundary (a) and the bottom boundary (b) of the quarter plate with a circular hole subjected to a unidirectional
tension.
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
25
Fig. 18. Comparison of the normal stress (e.g., sx ) distributions in an innite plate with a central circular hole obtained by the CQ4 (a), CT3 (b), Q4 (c) and T3 (d) elements.
Fig. 19. Comparison of the displacement solutions of an innite plate with a central circular hole for plane-strain condition with 0:4999 obtained by the CQ4, Q4 and
analytical methods. Displacement distribution along the bottom boundary (a) and the left boundary (b) of the quarter plate.
a2 1
3a4
cos
2
cos
4
cos
4
;
r2 2
2r 4
2
a 1
3a4
sin 2 sin 4 4 sin 4 ;
xy T x
2
r 2
2r
sy T x
38
26
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
Fig. 20. Geometry of internally pressurized hollow cylinder (a) and its quarter model.
Fig. 21. Regular (a) and irregular (b) meshes of 12 12 quadrilateral elements.
Fig. 22. Comparison of the convergence rates of an internally pressurized hollow cylinder hole obtained by the CQ4, CT3, Q4 and T3 elements: relative errors in displacement
norm (a) and energy norm (b).
uy y 0 0; 1 r x r 5
39
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
27
Another benchmark example considers a hollow cylinder subjected to an internal pressure as depicted in Fig. 20 to further show
the accuracy of the CFEM. The cylinder is designed with an inner
radius of a, an outer radius of b and a unit thickness. A uniform
pressure of p is applied to the inner surface at r a, whilst
traction-free boundary condition is assigned at the outer surface
r b. Only one-quarter of the cylinder is modeled due to the
geometrical symmetry of the structure. The analytical solutions of
the displacement and stress elds of this internally pressurized
hollow cylinder are available and given by [21].
"
#
2
a2 pr
b
ur r
1
;
2
r2
Eb a2
u 0;
40
#
2
b
1 2 ;
sr r 2
r
b a2
"
#
2
a2 p
b
1 2 ;
s r 2
r
b a2
a2 p
r 0
"
41
28
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
Fig. 25. Comparison of the radial stress (e.g., sr ) of an internally pressurized hollow cylinder obtained by the CQ4 (a), CT3 (b), Q4 (c) and T3 (d) elements.
Fig. 26. Geometry of a corner angle bracket (a) and its nite element mesh (b).
1
Note that we use the ANSYS's command such as (PLESOL, S, EQV) to plot the
stresses.
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
Fig. 27. Comparison of the von Mises stress distributions in a corner bracket obtained by the CQ4 (a), Q4 (b), ANSYS with coarse (c) and ne (d) meshes.
Fig. 28. Geometry of a quarter micro-motor rotor (a) and nite element mesh with 96 quadrilateral elements.
Fig. 29. Comparison of the von Mises stresses in a micro-motor rotor obtained by the CQ4 (a) and Q4 (b) elements.
29
30
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
1 r
4 1 r 1 r 1 r
s
A1
The Jacobian matrix and its inverse are explicitly
8
> xi
>
"
#>
>
1 s
1 s 1 s < xj
1 1 s
J
xk
1 r >
4 1 r 1 r 1 r
>
>
>
: xm
5. Conclusions
J
An efcient and accurate quadrilateral element with four-node
based on the CIP is formulated. The developed CQ4 element is
applied to the stress analysis of 2D elastic structures. The advantages of the CFEM are due to the fact that the stresses and strains
at nodes are continuous, smooth and it requires no smoothing
operation in the post-processing step. From the numerical results
of the four benchmark examples, it conrms the high accuracy of
the CFEM that pertains to high convergence rate as compared with
the standard FEM (Q4, T3) and even the CT3. Highly accurate
results in the practical examples further conrm the applicability
of the present CFEM in the stress analysis of 2D elastic structures
with complex conguration. The CIP algorithm is general and has
no limitations on its extension to other elements such as tetrahedral and brick elements in 3D. The extension of the method is
also attractive to the incompressible media, in which some
relevant issues pertaining to the pressure modes, convergence,
the infsup satisfaction, etc. should be investigated and addressed
in detail. From the present research work, the following conclusions can be drawn:
Acknowledgment
The support of the German Academic Exchange Service (DAAD,
Project-ID: 54368781) is gratefully acknowledged.
1
"
J1
1
4detJ J3
J2
J4
with J1 yi 1 r yj 1 r yk 1 r ym 1 r;
J2 yi 1 s yj 1 s yk 1 s ym 1 s;
J3 xi 1 r xj 1 r xk 1 r xm 1 r;
J4 xi 1 s xj 1 s xk 1 s xm 1 s
A3
i
L2i 2Li Lj ;
Lj
i
i
L2i 2Li Lk ;
L2i 2Li Lm
Lk
Lm
ix
xi xj 2Li Lj pLj Lk pLj Lm
Li
xi xk 2Li Lk pLk Lm pLk Lj
xi xm 2Li Lm pLm Lj pLm Lk ;
ix
xi xj L2i pLi Lk pLi Lm
Lj
xi xk pLi Lk xi xm pLi Lm ;
ix
xi xj pLi Lj xi xk L2i pLi Lm pLi Lj
Lk
xi xm pLi Lm ;
ix
xi xj pLi Lj xi xk pLi Lk
Lm
xi xm L2i pLi Lj pLi Lk ;
iy
yi yj 2Li Lj pLj Lk pLj Lm
Li
yi yk 2Li Lk pLk Lm pLk Lj
yi ym 2Li Lm pLm Lj pLm Lk ;
iy
yi yj L2i pLi Lk pLi Lm
Lj
yi yk pLi Lk yi ym pLi Lm ;
iy
yi yj pLi Lj yi yk L2i pLi Lm pLi Lj
Lk
yi ym pLi Lm ;
iy
yi yj pLi Lj yi yk pLi Lk
Lm
yi ym L2i pLi Lj pLi Lk
Appendix A
expressed as
9
yi >
>
>
>
yj =
A2
yk >
>
>
>
;
ym
A4
T.Q. Bui et al. / Finite Elements in Analysis and Design 84 (2014) 1431
4 2 0 0 2
s
"
#
2yi 2ym
2yi 2yj
1
;
J1
2xi 2xj
4detJ 2xi 2xm
()
()
h
i
h
i
x
Li Lj Lk Lm J 1 r
Li Lj Lk Lm
"
#
2yi 2ym
2yi 2yj 1 2
1
2xi 2xj 4 2
4detJ 2xi 2xm
"
#
yj ym ym yi 0 yi yj
1
4detJ xm xj xi xm 0 xj xi
2
0
0
0
0
2
A5
1;
1;
1 and
1
Li
Lj
Lk
Lm
A6
1 6 ym yi 7
6
7
1 1 1 1
6
4detJ4 0 7
5
yi yj
1
y ym ym yi yi yj 0
4detJ j
8 L 9
i
>
y >
>
>
>
>
>
>
>
Lj >
>
>
=
<
h
i
i
i
i
i
i
y
Li Lj Lk Lm
i;y xi
Lk >
>
y
>
>
> y >
>
>
>
>
>
;
: Lm >
y
2
3
xm xj
6
7
1 6 xi xm 7
6
7
1 1 1 1
7
4detJ6
4 0 5
xj xi
1
xm xj xi xm xj xi 0
4detJ
A7
31
References
[1] Hughes TJR, The Finite Element Method: Linear Static and Dynamic Finite
Element Analysis, Prentice-Hall, Englewood Cliffs, 1987.
[2] J. Fish, T. Belytschko, A First Course in Finite Elements, John Wiley & Sons Ltd,
England, 2007.
[3] G.R. Liu, S.S. Quek, The Finite Element Method: A Practical Course, Butterworth-Heinemann, Elsevier Science, Burlington, MA, 2003.
[4] G. Beer, I. Smith, C. Duenser, The Boundary Element Method with Programming For Engineers and Scientists, Springer-Verlag, Wien, Germany, 2008.
[5] C.R. Dohrmann, M.W. Heinstein, J. Jung, S.W. Key, W.R. Witkowski, Node-based
uniform strain elements for three-node triangular and four-node tetrahedral
meshes, Int. J. Numer. Methods Eng. 47 (2000) 15491568.
[6] C.R. Dohrmann, S.W. Key, M.W. Heinstein, J. Jung, A least-square approach for
uniform strain triangular and tetrahedral nite elements, Int. J. Numer.
Methods Eng. 42 (1998) 11811197.
[7] C. Zheng, S.C. Wu, X.H. Tang, J.H. Zhang, A novel twice-interpolation nite
element method for solid mechanics problems, Acta Mech. Sin. 26 (2010)
265278.
[8] J. Bonet, A.J. Burton, A simple average nodal pressure tetrahedral element for
incompressible and nearly incompressible dynamic explicit applications,
Commun. Numer. Methods Eng. 14 (1998) 437449.
[9] P. Hansbo, A nonconforming rotated Q1 approximation on tetrahedral,
Comput. Methods Appl. Mech. Eng. 200 (2011) 13111316.
[10] P. Hansbo, Nonconforming rotated Q1 tetrahedral element with explicit time
stepping for elastodynamics, Int. J. Numer. Methods Eng. 91 (2012) 11051114.
[11] S.A. Papanicolopulos, A. Zervos, A method for creating a class of triangular C1
nite element, Int. J. Numer. Methods Eng. 89 (2012) 14371450.
[12] G.R. Liu, A generalized gradient smoothing technique and the smoothed
bilinear form for Galerkin formulation of a wide class of computational
methods, Int. J. Comput. Methods 05 (2008) 136199.
[13] G.R. Liu, N. Nourbakhshnia, Y.W. Zhang, A novel singular ES-FEM method for
simulating singular stress eld near the crack tips for linear fracture problems,
Eng. Fract. Mech. 78 (2011) 863876.
[14] P. Liu, Q.T. Bui, Yu TT Zhang Ch, G.R. Liu, M.V. Golub, The singular edge-based
smoothed nite element method for stationary dynamic crack problems in 2D
elastic solids, Comput. Methods Appl. Mech. Eng. 233236 (2012) 6880.
[15] T.J.R. Hughes, J.A. Cottrell, Y. Bazilevs, Isogeometric analysis: CAD, nite
elements, NURBS, exact geometry and mesh renement, Comput. Methods
Appl. Mech. Eng. 194 (2004) 41354195.
[16] S.N. Atluri, T. Zhu, A new meshless local Petrov-Galerkin (MLPG) approach in
computational mechanics, Comput. Mech. 22 (1998) 117127.
[17] Q.T. Bui, N.M. Nguyen, Ch. Zhang, An efcient meshfree method for vibration
analysis of laminated composite plates, Comput. Mech. 48 (2011) 175193.
[18] Q.T. Bui, N.M. Nguyen, Zhang Ch, D.A.K. Pham, An efcient meshfree method
for analysis of two-dimensional piezoelectric structures, Smart Mater. Struct.
20 (2011) 065016.
[19] G.R. Liu, G.Y. Zhang, Y.Y. Wang, Z.H. Zhong, G.T. Li, X. Han, A nodal integration
technique for meshfree radial point interpolation method (NI-RPIM), Int. J.
Solids Struct. 44 (2007) 38403860.
[20] G.R. Liu, G.Y. Zhang, A novel scheme of strain-constructed point interpolation
method for static and dynamic mechanics problems, Int. J. Appl. Mech. 1
(2009) 233258.
[21] S.P. Timoshenko, J.N. Goodier, Theory of Elasticity, 3rd ed., McGraw-Hill, New
York, 1970.
[22] C.A. Felippa, Introduction to Finite Element Methods. University of Colorado,
Boulder. Available from: http://www.colorado.edu/engineering/CAS/courses.
d/IFEM.d/Home.html (accessed 15.11.13).
[23] ASNYS User's Manual, Release 12.1, ANSYS, Inc, Technology Drive Connosburg,
PA, 2009.