The Modal Behavior of The MITC3+ Triangular Shell Element
The Modal Behavior of The MITC3+ Triangular Shell Element
The Modal Behavior of The MITC3+ Triangular Shell Element
a r t i c l e i n f o a b s t r a c t
Article history: In this paper, we investigate the static and dynamic modal behavior of the MITC3+ triangular shell
Received 9 December 2014 element (Lee and Bathe, 2004; Lee et al., 2014). We focus on bending-dominated situations because such
Accepted 24 February 2015 shell problems are particularly difficult to solve when using low-order elements. For comparison, the
Available online 29 March 2015
pure displacement-based (DISP3), MITC3 and MITC4 shell elements are also studied. First, static mode
solutions are performed for a single right-angled shell element and an assemblage of two right-angled
Keywords: shell elements. The detailed strain fields are established in the bending modes. This study provides
Shell structures
insight into how shear locking occurs on the mode level. We then analytically show how the MITC3+ shell
Shell finite elements
3-node element
element properly represents the pure bending conditions in a two-sided clamped plate problem.
MITC method Considering free plate and free hyperboloid shell problems, we finally present the excellent performance
Mode analysis of the MITC3+ shell element in dynamic mode solutions.
Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction To significantly reduce the locking, the MITC method was devel-
oped and first applied to develop a 4-node shell element, the
Shell structures are ubiquitous in various engineering and MITC4 element [11,12]. Then the method was used to develop
scientific fields. In the past decades, the finite element method higher-order quadrilateral elements [13,14] and a 3-node triangu-
has been dominantly used to analyze shell structures. In these lar shell element, the MITC3 element, as well as higher-order
analyses, it is crucial to use reliable and effective shell finite triangular shell elements [1,15]. However, the performance of the
elements. However, due to the complex and highly sensitive MITC3 shell element is not as good as of the MITC4 shell element,
behaviors of shells, it is challenging to develop shell finite elements even though the MITC3 shell element yields much more accurate
and, indeed, a great challenge to reach ‘‘ideal’’ (that is, optimal) ele- solutions than the displacement-based 3-node shell element, the
ments [1–6]. In particular, it is important to obtain an ideal 3-node DISP3 element [16].
shell element. Due to its simple geometry defined by only three Recently, based on the concept of the MITC3 shell element, a
corner nodes, such an element can be very effectively used in the new 3-node triangular MITC3+ shell finite element was developed
automatic mesh generation of complex shells. [2]. The MITC3+ shell element uses a cubic bubble function for the
An ideal shell finite element should uniformly converge with interpolation of the rotations to enrich the bending displacement
optimal rate in all problems for all asymptotic behaviors (mem- fields. The corresponding rotation degrees of freedom can be
brane dominated, bending dominated and mixed behaviors) statically condensed out on the element level. A new assumed
irrespective of the shell geometry, boundary conditions and transverse shear strain field was developed with a new tying
applied loadings [3–10]. Of course, the shell element should satisfy scheme to alleviate shear locking while satisfying the consistency
the ellipticity and consistency conditions. A main difficulty in shell and ellipticity conditions. The MITC3+ shell element passes the
finite element analysis is that the solution accuracy may become three basic tests (the patch, zero energy mode and isotropy tests)
drastically worse as the shell thickness decreases. This and shows excellent convergence behaviors in various shell prob-
phenomenon is called locking, which occurs when the element lems even when highly distorted meshes are used [2]. An excellent
discretization of a shell structure cannot properly represent pure performance of the MITC3+ shell element in geometric nonlinear
bending displacement fields [4]. analysis was also measured [17]. With such encouraging results
at hand, it is clearly of interest to further study this element and
identify more deeply its fundamental features.
⇑ Corresponding author. Our objective in this paper is to present a deeper understanding
E-mail address: phillseung@kaist.edu (P.S. Lee). of the MITC3+ shell element through detailed static and dynamic
http://dx.doi.org/10.1016/j.compstruc.2015.02.033
0045-7949/Ó 2015 Elsevier Ltd. All rights reserved.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 149
mode solutions. The DISP3, MITC3 and MITC4 shell elements are
also considered for comparisons. In the static mode solutions of a
single right-angled triangular element and an assemblage of two
right-angled triangular elements, we investigate the transverse
shear strain fields in the bending modes. This study shows how
the MITC3+ shell element performs on the mode level. The
transverse shear strain fields of the MITC3+ shell element are also
studied analytically in a two-sided clamped plate problem, to see
whether and how locking occurs. In the dynamic mode solutions,
the frequencies and mode shapes are evaluated considering a free
square plate problem and a hyperboloid shell problem and the
convergence of the frequencies is presented.
In the following sections, we first review the formulations of the
3-node triangular shell finite elements that we consider in this
paper and then we present the results of our investigations.
Fig. 2. Tying positions for the assumed transverse shear strain field of the MITC3
2. Formulations of 3-node triangular shell elements shell element.
The constant
ptransverse
ffiffiffi shear strain conditions are imposed along its
ð3Þ ð3Þ ð3Þ
edges eqt ¼ est ert = 2 .
In this section, the formulations of three 3-node continuum
mechanics based triangular shell finite elements are briefly
reviewed: the DISP3, MITC3 and MITC3+ shell elements. and to each other, and ai and bi are the rotations of the director
! ! !
vector V in about V i1 and V i2 , respectively, at node i.
2.1. The DISP3 shell element The linear part of the displacement-based covariant strain
components is given by
The geometry of a standard 3-node continuum mechanics based 1
triangular shell finite element is interpolated by [1,3] eij ¼ ð~
gi ~
u;j þ ~
gj ~
u;i Þ; ð3Þ
2
X3
tX 3
! in which
~
xðr; s; tÞ ¼ hi ðr; sÞ~
xi þ ai hi ðr; sÞ V in
i¼1
2 i¼1 @~
x @~
u
~
gi ¼ ; ~
u;i ¼ with r 1 ¼ r; r 2 ¼ s; r 3 ¼ t: ð4Þ
with h1 ¼ 1 r s; h2 ¼ r; h3 ¼ s; ð1Þ @ri @r i
The displacement-based 3-node shell finite element has no spurious
where hi ðr; sÞ is the two-dimensional interpolation function of the
zero energy mode and satisfies the isotropy condition. However, in
standard isoparametric procedure corresponding to node i; ~ xi is
bending-dominated problems, this shell finite element is extremely
the position vector of node i in the global Cartesian coordinate
! stiff due to shear locking [16].
system, and ai and V in denote the shell thickness and the director
vector at the node, see Fig. 1. 2.2. The MITC3 shell element
The corresponding displacement interpolation of the element is
given by The geometry and displacement interpolations are identical to
! the DISP3 shell element, but the MITC method is applied [1].
X3
tX 3
!
~
uðr; s; tÞ ¼ hi ðr; sÞ~
ui þ ai hi ðr; sÞ V i2 ai þ V i1 bi ; ð2Þ Since the geometry of the 3-node triangular shell finite element
i¼1
2 i¼1 is flat, the MITC method is only applied to construct the covariant
transverse shear strain field with constant covariant transverse
in which ~
ui is the nodal displacement vector in the global Cartesian shear strain conditions along the edges. The assumed transverse
! ! !
coordinate system, V i1 and V i2 are unit vectors orthogonal to V in shear strain field of the MITC3 shell element is given by [1]
Fig. 1. A standard 3-node triangular continuum mechanics based shell element. Fig. 3. Geometry of the MITC3+ shell element with an additional bubble node.
150 Y. Lee et al. / Computers and Structures 153 (2015) 148–164
Fig. 4. Tying positions (A)–(F) for the assumed transverse shear strain field of the MITC3+ shell finite element.
X3
tX4
!
Table 1 ~
xðr; s; tÞ ¼ hi ðr; sÞ~
xi þ ai f ðr; sÞ V in
Tying positions for the assumed transverse shear strain field for the MITC3+ shell i¼1
2 i¼1 i
element. The distance d is defined in Fig. 4(b). ! 1 !1 ! !
with a4 V 4n ¼ a1 V n þ a2 V 2n þ a3 V 3n ; ð6Þ
Tying positions r s 3
Fig. 4(a) (A) 1/6 2/3 in which f i ðr; sÞ are the two-dimensional interpolation functions
(B) 2/3 1/6
with the cubic bubble function f 4 corresponding to the internal
(C) 1/6 1/6
node 4
Fig. 4(b) (D) 1=3 þ d 1=3 2d
(E) 1=3 2d 1=3 þ d 1 1 1
(F) 1=3 þ d 1=3 þ d f 1 ¼ h1 f 4 ; f 2 ¼ h2 f 4 ; f 3 ¼ h3 f 4 ; f 4 ¼ 27rsð1 r sÞ:
3 3 3
ð7Þ
ð1Þ ð2Þ
~eMITC3 ¼ ert þ cs; ~eMITC3 ¼ est cr; ð5Þ
rt st From Eq. (6), the displacement interpolation is described by
ð3Þ ð1Þ ð3Þ ð2Þ
in which c ¼ ert
ert
þ est est
and the tying points are shown in X3
tX 4 ! !
Fig. 2. ~
uðr; s; tÞ ¼ hi ðr; sÞ~
ui þ ai f i ðr; sÞ V i2 ai þ V i1 bi ; ð8Þ
i¼1
2 i¼1
The MITC3 shell element passes the basic tests: zero energy
mode, isotropy and patch tests. The MITC3 shell element shows a where a4 and b4 are the rotation degrees of freedom at the internal
quite reasonable convergence behavior. However, as the thickness node [2]. These additional rotation degrees of freedom can be
decreases, some locking occurs in bending-dominated problems condensed out on the element level.
[1]. Unlike the standard 3-node shell elements, the MITC3+ shell
element has an internal node and the corresponding cubic bubble
2.3. The MITC3+ shell element function in the geometry and displacement interpolations. The
internal node with only rotation degrees of freedom is positioned
The geometry of the MITC3+ shell element, see Fig. 3, is on the flat surface defined by the three corner nodes of the
interpolated by element. Hence, the geometry of the MITC3+ shell element is
Fig. 5. A single right-angled triangular shell element in plate bending (L ¼ 1:0; a=L ¼ 1=10; 000; E ¼ 1:7472 107 and m ¼ 0:3Þ. (a) A single 3-node triangular shell element. (b)
A single 3-node triangular shell element with a bubble node on rotations.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 151
Table 2
Eigenvalues of the stiffness matrix of the single triangular shell element for the element geometry shown in Fig. 5. Note that modes 1 to 6 produce zero eigenvalues corresponding
to rigid body modes. The order of the zero eigenvalues calculated is smaller than 1012 .
B: symmetric bending modes, T: in-plane twisting mode, S: transverse shearing modes, M: membrane modes, B+: anti-symmetric bending modes due to the bubble function
enrichment.
Fig. 6. Shapes of the bending modes for the single right-angled MITC3 shell element shown in Fig. 5(a). The dotted and solid lines correspond to the top surfaces before and
after deformation, respectively, on the xy-plane. Plus and minus signs denote small out-of-plane displacements in the z and –z directions, respectively.
always flat and only the transverse shear strain components are 2 ðBÞ 1 ðBÞ 1 ðCÞ ðCÞ
1
^ertMITC3þ ¼ ert est þ e þ est þ ^cð3s 1Þ;
established using the MITC method. 3 2 3 rt 3
The assumed transverse shear strain field of the MITC3+ shell 2 ðAÞ 1 ðAÞ 1 ðCÞ ðCÞ
1
finite element is given by [2]
^estMITC3þ ¼ e ert þ e þ est þ ^cð1 3rÞ; ð9Þ
3 st 2 3 rt 3
152 Y. Lee et al. / Computers and Structures 153 (2015) 148–164
Fig. 7. Mode shapes of the single right-angled MITC3+ shell element shown in Fig. 5(b). The dotted and solid lines correspond to the top surfaces before and after deformation,
respectively, on the xy-plane. Plus and minus signs denote small out-of-plane displacements in the z and –z directions, respectively.
Table 3
Strain fields of the single right-angled DISP3 and MITC3 shell elements in the bending modes B1, B2 and B3. a=L ¼ 1=10; 000; z ¼ aðt=2Þ.
Table 4 Table 5
Strain fields of the single right-angled MITC3+ shell element in the bending modes Eigenvalues of the stiffness matrix of the assemblage of two right-angled triangular
B1+ and B2+, and in-plane twisting mode T. a=L ¼ 1=10; 000; z ¼ aðt=2Þ. shell elements shown in Fig. 8 when a=L ¼ 1=10; 000. Note that modes 1 to 6 produce
zero eigenvalues corresponding to rigid body modes. The order of the zero
Bending mode Eigenvalue 8.3107E06 eigenvalues calculated is smaller than 1012 .
B1+ Bending exx ¼ ½0:019 21:3ðs 2rs s2 Þz
strains eyy ¼ ½0:019 þ 21:3ðr 2rs r 2 Þz Mode MITC4 MITC3 MITC3+
cxy ¼ 21:3ðr s r 2 þ s2 Þz 7 7.2000E07 BL1 9.9556E07 BC1 9.3805E07 BC1
Transverse cxz ¼ 0:0 8 7.2000E07 BL2 1.1200E06 BC2 1.0608E06 BC2
shear cyz ¼ 0:0 9 BC1 BC3 BC3
9.9556E07 2.0800E06 1.9629E06
strains
10 1.1200E06 BC2 3.2000E06 BL2 3.0544E06 BL2
Bending mode Eigenvalue 1.4128E05 11 BC3 BL1 BQ1+
2.0800E06 3.4167E+01 8.9316E06
B2+ Bending exx ¼ ½0:001 21:8ðs 2rs s2 Þz 12 5.6000E+01 TQ 5.6000E+01 TQ 1.1912E05 BQ2+
strains eyy ¼ ½0:001 21:8ðr 2rs r2 Þz
2 2 13 5.0400E+02 BC4 8.4000E+02 SQ1 1.4173E05 TQ
cxy ¼ ½0:064 21:8ðr þ s 4rs r s Þz
Transverse cxz ¼ 0:0 14 8.4000E+02 SQ1 9.1783E+02 SQ2 1.5159E05 BL1
shear cyz ¼ 0:0 15 8.4000E+02 SQ2 1.3440E+03 MQ 1.6660E05 BQ3+
strains 16 8.6400E+02 MQ 1.3440E+03 MQ 9.3333E+01 BQ4+
In-plane twisting Eigenvalue 1.3599E05 17 8.6400E+02 MQ 1.3440E+03 MQ 8.0267E+02 SQ1
mode T Bending exx ¼ ½0:514 þ 1:49ðs 2rs s2 Þz 18 1.3440E+03 MQ 1.5120E+03 BC4 8.4000E+02 SQ2
(d ¼ 1=10; 000Þ strains eyy ¼ ½0:514 1:49ðr 2rs r2 Þz 19 1.3440E+03 MQ 2.4960E+03 MQ 1.3440E+03 MQ
cxy ¼ 1:49ðr s r 2 þ s2 Þz 20 2.4960E+03 MQ 3.8400E+03 MQ 1.3440E+03 MQ
21 - - 1.3440E+03 MQ
Transverse cxz ¼ 1:99dð1 3sÞ
22 - - 1.5493E+03 BC4
shear cyz ¼ 1:99dð1 þ 3rÞ
strains 23 - - 2.4960E+03 MQ
24 - - 3.8400E+03 MQ
BC: bending modes with constant bending strain fields, BL: bending modes with
linear bending strain fields, TQ: in-plane twisting mode, SQ: transverse shearing
modes, MQ: membrane modes, BQ+: bending modes due to the bubble function
enrichment.
/si ¼ ksi ~
K~ /si with i ¼ 1; 2; . . . ; N; ð10Þ
Fig. 10. In-plane twisting of two 3-node triangular shell elements. The dotted and
solid lines correspond to the top surfaces before and after deformation, respectively,
in the xy-plane.
Table 6
Eigenvalues of the stiffness matrix of the assemblage of two MITC3+ shell elements shown in Fig. 8(b) according to the tying distance d in the plate bending problem when
a=L ¼ 1=10; 000. Note that modes 1 to 6 produce zero eigenvalues corresponding to rigid body modes. The order of the zero eigenvalues calculated is smaller than 1012 . Modes
19–21, 23 and 24 are membrane modes. For d ¼ 1=10; 000, the eigenvalues are given in Table 5.
Mode d
1/100 1/1,000 1/100,000 0.0
7 9.3805E07 BC1 9.3805E07 BC1 9.3805E07 BC1 9.3805E07 BC1
8 1.0611E06 BC2 1.0611E06 BC2 1.0545E06 BC2 1.0099E06 BL1
9 1.9629E06 BC3 1.9629E06 BC3 1.1303E06 BL1 1.0522E06 BC2
10 3.0544E06 BL2 3.0544E06 BL2 1.5394E06 TQ 1.4140E06 TQ
11 9.7882E06 BQ1+ 9.7846E06 BQ1+ 1.9629E06 BC3 1.9629E06 BC3
12 1.1912E05 BQ2+ 1.1912E05 BQ2+ 3.0544E06 BL2 3.0544E06 BL2
13 1.6660E05 BQ3+ 1.6660E05 BQ3+ 1.0320E05 BQ1+ 1.0312E05 BQ1+
14 1.2767E01 TQ 1.2782E03 TQ 1.1912E05 BQ2+ 1.1912E05 BQ2+
15 1.2768E01 BL1 1.2783E03 BL1 1.6660E05 BQ3+ 1.6660E05 BQ3+
16 9.3340E+01 BQ4+ 9.3333E+01 BQ4+ 9.3333E+01 BQ4+ 9.3333E+01 BQ4+
17 8.0267E+02 SQ1 8.0267E+02 SQ1 8.0267E+02 SQ1 8.0267E+02 SQ1
18 8.4001E+02 SQ2 8.4000E+02 SQ2 8.4000E+02 SQ2 8.4000E+02 SQ2
22 1.5493E+03 BC4 1.5493E+03 BC4 1.5493E+03 BC4 1.5493E+03 BC4
BC: bending modes with constant bending strain fields, BL: bending modes with linear bending strain fields, TQ: in-plane twisting mode, SQ: transverse shearing modes, BQ+:
bending modes due to the bubble function enrichment.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 155
Table 7
Strain fields of the MITC4, MITC3 and MITC3+ shell elements in the bending mode BL1. a=L ¼ 1=10; 000; z ¼ aðt=2Þ.
Table 9
Strain energies for the two-sided clamped plate problem shown in Fig. 11 using the
Mesh A with the MITC3+ shell element according to the tying distance d.
a=L d
d = 1/100 d = 1/1,000 d = 1/10,000 d = 1/100,000
1/100 1.5589E01 4.7837E01 4.8848E01 4.8858E01
1/1,000 2.2787E+00 1.5587E+02 4.7820E+02 4.8830E+02
1/10,000 2.2893E+01 2.2787E+03 1.5587E+05 4.7820E+05
Table 8
Strain energies for the two-sided clamped plate problem shown in Fig. 11. The tying distance d ¼ 0 is used for the MITC3+ shell element.
Table 10
Frequencies of the modes 7–11 for the free square plate problem shown in Fig. 13
with N N meshes when a=L ¼ 1=1; 000.
Fig. 13. Free square plate problem for dynamic mode analysis with 2 different 5 5 mesh patterns (L ¼ 1:0; E ¼ 2:07 1011 , m ¼ 0:3 and q ¼ 7:8 103 Þ. (a) Mesh pattern used
for the MITC4 shell element. (b) Mesh pattern used for the MITC3 and MITC3+ shell elements.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 157
Fig. 14. Shapes of the modes 7–11 for the free square plate problem shown in Fig. 13 with 5 5 element meshes when a=L ¼ 1=1; 000.
shell elements, this phenomenon should be overcome. Hence, we No anticlastic curvature exists because Poisson’s ratio m ¼ 0. If a
investigate the effectiveness of the MITC3+ shell element in the triangular shell element does not show shear locking, Eqs. (14)
analysis of this two-sided clamped plate problem. and (15) should be satisfied.
For element-I, we derive the transverse shear strain field in
4.1. Transverse shear strain fields Appendix, see Eq. (A.3). Using the conditions given in Eqs. (12)
and (13) with the symmetry condition a2 ¼ b2 , the transverse
Using the two MITC3+ shell elements, the two-sided clamped shear strain fields for the two elements are for element-I
plate is modeled as shown by element-I and element-II in
Fig. 12. The geometry of the elements is given by the following a 1 1 1
eIrt ¼ w2 a2 a4 þ ^cI ð3s 1Þ;
nodal positions 4 2 2 3
x1 ¼ 1;y1 ¼ 0; x2 ¼ 1;y2 ¼ 1; x3 ¼ 0;y3 ¼ 0; x4 ¼ 2=3;y4 ¼ 1=3 for element-I; I a 1 1 I 3
est ¼ b4 þ ^c ð1 3rÞ with ^cI ¼ ada2 ð16Þ
x1 ¼ 0;y1 ¼ 1; x2 ¼ 1;y2 ¼ 1; x3 ¼ 0;y3 ¼ 0; x4 ¼ 1=3;y4 ¼ 2=3 for element-II; ð12Þ 4 2 3 4
where the subscripts are the node numbers on the element level,
and for element-II
see Fig. 12.
The boundary condition of the elements is given by
a 1 1 1
eIIrt ¼ w2 þ b2 þ b4 þ ^cII ð3s 1Þ;
u1 ¼ u3 ¼ v 1 ¼ v 3 ¼ w1 ¼ w3 ¼ a1 ¼ a3 ¼ b1 ¼ b3 ¼ 0 for element-I; 4 2 2 3
u1 ¼ u3 ¼ v 1 ¼ v 3 ¼ w1 ¼ w3 ¼ a1 ¼ a3 ¼ b1 ¼ b3 ¼ 0 for element-II: ð13Þ a 1 1 3
eIIst ¼ a4 þ ^cII ð1 3rÞ with ^cII ¼ adb2 : ð17Þ
4 2 3 4
For this pure bending problem, the exact transverse shear strains
are
With the tying distance d ¼ 0, the transverse shear strains vanish
ert ¼ est ¼ 0 in elements-I and II; ð14Þ with
1 1
and the theoretical relationship among the deflection w2 ð¼ wtip Þ, w2 a2 a4 ¼ 0; b4 ¼ 0 for element-I; ð18Þ
rotation a2 ð¼ atip Þ and rotation b2 ð¼ btip Þ is given, like for beams
2 2
1 1
in the x- and y-directions [3], as w2 þ b2 þ b4 ¼ 0; a4 ¼ 0 for element-II: ð19Þ
2 2
1 1
w2 ¼ a2 ¼ b2 : ð15Þ When a4 ¼ 0 in Eq. (18) and b4 ¼ 0 in Eq. (19), the pure bending
2 2
condition is satisfied as follows
158 Y. Lee et al. / Computers and Structures 153 (2015) 148–164
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(a) Mode 10
MITC4 MITC3 MITC3+
0 0 0
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(b) Mode 11
Fig. 15. Convergence curves of the 10th and 11th frequencies for the free square plate problem. The solutions of the shell elements are obtained with uniform N N element
meshes (N ¼ 5, 10, 15 and 20). The bold line represents the optimal convergence rate.
Fig. 16. Distorted mesh pattern for N ¼ 5. (a) Mesh pattern used for the MITC4 shell element. (b) Mesh pattern used for the MITC3 and MITC3+ shell elements.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 159
Table 11 1 1
Frequencies of the modes 7–11 for the free square plate problem with N N distorted w2 a2 ¼ 0; w2 þ b2 ¼ 0: ð20Þ
2 2
meshes shown in Fig. 16 when a=L ¼ 1=1; 000.
This result shows that the MITC3+ shell element can correctly
Shell elements N Mode number
represent the pure bending conditions in Eqs. (14) and (15) when
7 8 9 10 11
d ¼ 0. Similarly, we tested the MITC3+ shell element in Mesh B
MITC4 5 21.575 33.071 42.004 57.191 61.420 and observed that the elements in Mesh B also satisfy the pure
10 21.189 31.311 39.109 55.541 56.437
bending conditions. Of course, the MITC3 shell element does not
15 21.087 30.908 38.436 54.892 55.286
20 21.049 30.757 38.183 54.628 54.851
satisfy the pure bending conditions in Mesh A and thus shear
locking occurs [16].
MITC3 5 22.184 33.780 43.522 62.781 105.68
10 21.812 32.397 40.587 60.633 93.456
15 21.692 32.053 39.838 59.615 89.827 4.2. Strain energies
20 21.630 31.829 39.445 58.909 87.021
MITC3+ 5 21.817 33.177 43.060 59.889 69.344 Considering three different thicknesses (a=L ¼ 1=100, 1/1,000
10 21.318 31.546 39.651 56.842 60.559 and 1/10,000), we next solve the two-sided clamped plate problem
15 21.148 31.052 38.734 55.571 57.419
using the MITC3, MITC3+ and MITC4 shell elements with the
20 21.079 30.848 38.366 55.026 56.096
applied moments ma ¼ mb ¼ 2=L, see Fig. 11.
Ref. 21.000 30.564 37.864 54.284 54.284
Table 8 presents the strain energies calculated using the MITC3,
⁄
The reference solutions are obtained by 50 50 element meshes of the MITC4 shell MITC3+ (d ¼ 0Þ and MITC4 shell elements, respectively.
element. Analytically, the calculated strain energy should change in the
order of ða=LÞ3 for this pure bending problem [4]. The MITC3 shell
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(a) Mode 10
MITC4 MITC3 MITC3+
0 0 0
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(b) Mode 11
Fig. 17. Convergence curves of the 10th and 11th frequencies for the free plate problem with the distorted meshes shown in Fig. 16. The bold line represents the optimal
convergence rate.
160 Y. Lee et al. / Computers and Structures 153 (2015) 148–164
element shows correct results only in Mesh B. The MITC3+ shell 5. Dynamic mode solutions
element shows correct behaviors in both Meshes A and B, although
the solutions depend on the mesh patterns. The MITC4 shell In this section, we perform dynamic mode solutions to
element obviously shows excellent results. investigate the predictive capability of the MITC3 and MITC3+ shell
Changing the tying distance d in the assumed transverse shear elements in dynamic analysis. Frequencies and modes are calcu-
strain field of the MITC3+ shell element, we investigate the strain lated and the results are compared with those of the MITC4 shell
energy stored in the two-sided clamped plate problem with Mesh element. We consider two benchmark problems: a free square
A. Four different tying distances are considered: d ¼ 1=100; 1=1; plate problem and a free hyperboloid shell problem [1,2,21–27].
000; 1=10; 000 and 1/100,000. Table 9 shows that, as the distance d For dynamic mode solutions, the following eigenvalue problem
approaches zero, the behavior of the MITC3+ shell element become is solved
better.
We note that the bending mode BL1 of the assemblage of two /di ¼ kdi M~
K~ /di with i ¼ 1; 2; . . . ; N; ð21Þ
right-angled triangular elements is significantly present in the
pure bending displacement field of the problem with Mesh A. where K and M is the stiffness and consistent mass matrices,
Thus, the improved behavior of the MITC3+ shell element is respectively, kdi is the eigenvalue (the square of the free vibration
directly due to the alleviation of shear locking in this mode. frequency xi ; kd ¼ x2 Þ; ~
i /d is the corresponding M-orthonormal
i i
Fig. 18. Free hyperboloid shell problem (L ¼ 1:0; E ¼ 2:0 1011 , m ¼ 1=3 and q ¼ 7:8 103 Þ. (a) Problem description with a uniform mesh. (b) Distorted mesh pattern.
Table 12
Frequencies of the modes 7–12 for the free hyperboloid shell problem with the 4N 2N uniform meshes shown in Fig. 18(a) when a=L ¼ 1=1; 000.
eigenvector (dynamic mode), and N is the number of the degrees of problems. Hence, the reference frequency xref is obtained using
freedom [3]. Note that, unlike in a static mode solution, both, the very fine meshes of the MITC4 shell element.
stiffness and mass matrices are considered in dynamics. For the dynamic analysis of the shell problems, the optimal
To measure the convergence of the finite element solutions, we convergence behavior of the frequencies is given by
use the relative error in the reciprocal frequency as follows
Fig. 19. Shapes of the modes 7–12 for the free hyperboloid shell problem shown in Fig. 18(a) with 20 10 element meshes when a=L ¼ 1=1; 000.
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
Fig. 20. Convergence curves of the 7th and 8th frequencies for the free hyperboloid shell problem with the uniform meshes shown in Fig. 18(a). The solutions of the shell
elements are obtained with uniform 4N 2N element meshes (N ¼ 5, 10, 15 and 20). The bold line represents the optimal convergence rate.
162 Y. Lee et al. / Computers and Structures 153 (2015) 148–164
Table 13
Frequencies of the modes 7–12 for the free hyperboloid shell problem with the 4N 2N distorted meshes shown in Fig. 18(b) when a=L ¼ 1=1; 000.
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(a) Mode 7
MITC4 MITC3 MITC3+
0 0 0
-1 -1 -1
-2 -2 -2
-3 -3 -3
-1.5 -1 -0.5 -1.5 -1 -0.5 -1.5 -1 -0.5
log h log h log h
(b) Mode 8
Fig. 21. Convergence curves of the 7th and 8th frequencies for the free hyperboloid shell problem with the distorted meshes shown in Fig. 18(b). The bold line represents the
optimal convergence rate.
Y. Lee et al. / Computers and Structures 153 (2015) 148–164 163
5.1. Free square plate problem Fig. 16. Table 13 presents the frequencies with distorted 4N 2N
element meshes (N ¼ 5, 10, 15 and 20) when a=L ¼ 1=1; 000.
A free square plate of dimensions L L is considered as shown Fig. 21 shows the convergence curves of the 7th and 8th frequencies.
in Fig. 13. We use L ¼ 1:0; E ¼ 2:07 1011 ,
m ¼ 0:3 and density In both results, we observe that, while the MITC4 and MITC3 shell
q ¼ 7:8 103 . No boundary condition is imposed. We use the elements lock, the MITC3+ shell element still shows good results.
reference solutions obtained using a 50 50 element mesh of the
6. Concluding remarks
MITC4 shell element and verified by Refs. [25,26].
Table 10 presents the frequencies calculated with N N
In this paper, we presented new insight into the static and
element meshes (N ¼ 5, 10, 15 and 20) when a=L ¼ 1=1; 000. The
dynamic modal behaviors of the MITC3+ triangular shell element.
frequencies of the modes 7 to 9 are close to the reference solutions
For comparison, also the DISP3, MITC3 and MITC4 shell elements
for all tested elements. The frequencies corresponding to modes 10
were considered.
and 11 should be identical, but this can hardly be the case using
Detailed static mode solutions were studied for the single
triangular shell elements due to the non-symmetry in the mesh
right-angled triangular shell elements and the assemblage of two
patterns used. While the MITC3 shell element produces con-
right-angled triangular shell elements. We also investigated the
siderably different frequencies for modes 10 and 11, the MITC3+
transverse shear strain fields of the MITC3+ shell element in the
shell element shows very similar frequencies for the modes.
two-sided clamped plate problem. Through these studies, we
Using the MITC4 and MITC3+ shell elements produces much more
found that the MITC3+ shell element formulation contains
accurate frequencies than obtained with the MITC3 shell element.
additional anti-symmetric bending modes (B1+ and B2+) due to
In Fig. 14, we see that the MITC3 shell element produces a wrong
the bubble function enrichment, and shear locking is alleviated
mode shape for the 11th frequency.
in these modes using the MITC method. Consequently, unlike the
The convergence curves of the 10th and 11th frequencies are
DISP3 and MITC3 shell elements, the MITC3+ shell element
plotted in Fig. 15. Three different thicknesses (a=L ¼ 1=100;
performed well in the solution of all cases.
1=1; 000 and 1=10; 000Þ are considered. The shell finite element
We also performed dynamic mode solutions of an unsupported
solutions are calculated using four different N N element meshes
square plate and a free hyperboloid shell, both representing
(N ¼ 5, 10, 15 and 20). We observe that the performance of the
bending-dominated problems in which locking would be seen, if
MITC3 shell element deteriorates as the thickness decreases, that
present. We observed that the MITC3+ shell element shows much
is, locking occurs. However, the MITC3+ and MITC4 shell elements
better convergence behaviors than the MITC3 shell element.
show, similarly, excellent convergence.
Indeed, the MITC3+ shell element gives as accurate results as the
The distorted mesh shown in Fig. 16 is also considered. When
MITC4 shell element in uniform meshes and even significantly
an N N element mesh is used, each edge is discretized by the
better results when distorted meshes are used.
following ratio: L1 : L2 : L3 : . . . LN = 1: 2: 3: . . . N. Table 11 presents
the frequencies with distorted N N element meshes (N ¼ 5, 10, Acknowledgments
15 and 20) when a=L ¼ 1=1; 000. Fig. 17 shows the convergence
curves of the 10th and 11th frequencies. In the both results, the This work was supported by the Human Resources
MITC3+ and MITC4 shell elements still show good results even Development Program (No. 20134030200300) of the Korea
though the predicted modes 10 and 11 are slightly different. Institute of Energy Technology Evaluation and Planning (KETEP)
funded by the Ministry of Trade, Industry and Energy, and the
5.2. Free hyperboloid shell problem Basic Science Research Program through the National Research
Foundation of Korea (NRF) funded by the Ministry of Education,
We consider the free hyperboloid shell shown in Fig. 18. The Science and Technology (No. 2014R1A1A1A05007219).
mid-surface of the shell structure is given by [1,2]
Using Eqs. (9) and (A.2), the transverse shear strain field of the [10] Bathe KJ, Chapelle D, Lee PS. A shell problem ‘highly sensitive’ to thickness
changes. Int J Numer Methods Eng 2003;57:1039–52.
MITC3+ element is given by
[11] Dvorkin EN, Bathe KJ. A continuum mechanics based four-node shell element
a 1 1 1 for general nonlinear analysis. Eng Comput 1984;1:77–88.
^eMITC3þ
rt ¼ w2 w1 þ ðx2 2x1 þ x3 Þb1 þ ð2x2 x1 x3 Þb2 þ ðx2 x1 Þb4 [12] Bathe KJ, Dvorkin EN. A formulation of general shell elements – the use of
4 6 6 2
mixed interpolation of tensorial components. Int J Numer Methods Eng
1 1 1 1
þ ð2y1 y2 y3 Þa1 þ ðy1 2y2 þ y3 Þa2 þ ðy1 y2 Þa4 þ ^cð3s 1Þ; 1986;22:697–722.
6 6 2 3 [13] Bucalem ML, Bathe KJ. Higher-order MITC general shell elements. Int J Numer
a 1 1 1 Methods Eng 1993;36:3729–54.
^eMITC3þ
st ¼ w3 w1 þ ðx3 þ x2 2x1 Þb1 þ ð2x3 x2 x1 Þb3 þ ðx3 x1 Þb4
4 6 6 2 [14] Bathe KJ, Lee PS, Hiller JF. Towards improving the MITC9 shell element.
1 1 1 1 Comput Struct 2003;81:477–89.
þ ð2y1 y2 y3 Þa1 þ ðy1 þ y2 2y3 Þa3 þ ðy1 y3 Þa4 þ ^cð1 3rÞ ðA:3Þ [15] Kim DN, Bathe KJ. A triangular six-node shell element. Comput Struct
6 6 2 3
2009;87:1451–60.
with [16] Lee PS, Noh HC, Bathe KJ. Insight into 3-node triangular shell finite elements:
the effects of element isotropy and mesh patterns. Comput Struct
3 2007;85:404–18.
^c ¼ ad½ðy2 y3 Þa1 þ ðy3 y1 Þa2 þ ðy1 y2 Þa3 þ ðx3 x2 Þb1 [17] Jeon HM, Lee Y, Lee PS, Bathe KJ. The MITC3+ shell element in geometric
4
nonlinear analysis. Comput Struct 2015;146:91–104.
þ ðx1 x3 Þb2 þ ðx2 x1 Þb3 : ðA:4Þ [18] Argyris JH, Papadrakakis M, Apostolopoulou C, Koutsourelakis S. The TRIC shell
element: theoretical and numerical investigation. Comput Methods Appl Mech
Eng 2000;182:217–45.
References [19] Lee Y, Yoon K, Lee PS. Improving the MITC3 shell finite element by using the
Hellinger–Reissner principle. Comput Struct 2012;110–111:93–106.
[20] Kim JH, Kim YH. A three-node C0 ANS element for geometrically non-linear
[1] Lee PS, Bathe KJ. Development of MITC isotropic triangular shell finite
structural analysis. Comput Methods Appl Mech Eng 2002;191:4035–59.
elements. Comput Struct 2004;82:945–62.
[21] Jeon HM, Lee PS, Bathe KJ. The MITC3 shell finite element enriched by
[2] Lee Y, Lee PS, Bathe KJ. The MITC3+ shell element and its performance. Comput
interpolation covers. Comput Struct 2013;134:128–42.
Struct 2014;138:12–23.
[22] Lee PS, Bathe KJ. The quadratic MITC plate and MITC shell elements in plate
[3] Bathe KJ. Finite element procedures. Watertown (MA): KJ Bathe; 2014.
bending. Adv Eng Softw 2010;41:712–28.
[4] Chapelle D, Bathe KJ. The finite element analysis of shells – fundamentals.
[23] Bathe KJ, Lee PS. Measuring the convergence behavior of shell analysis
Second edition. Berlin: Springer-Verlag; 2011.
schemes. Comput Struct 2011;89:285–301.
[5] Chapelle D, Bathe KJ. Fundamental considerations for finite element analysis of
[24] Hiller JF, Bathe KJ. Measuring convergence of mixed finite element
shell structures. Comput Struct 1998;66:19–36. 711–712.
discretizations: an application to shell structures. Comput Struct
[6] Lee PS, Bathe KJ. Insight into finite element shell discretizations by use of the
2003;81:639–54.
‘‘basic shell mathematical model’’. Comput Struct 2005;83:69–90.
[25] Leissa AW. The free vibration of rectangular plates. J Sound Vib
[7] Lee PS, Bathe KJ. On the asymptotic behavior of shell structures and the
1973;31:257–93.
evaluation in finite element solutions. Comput Struct 2002;80:235–55.
[26] Mochida Y, Ilanko S. Bounded natural frequencies of completely free
[8] Beirão da Veiga L. Asymptotic energy behavior of two classical intermediate
rectangular plates. J Sound Vib 2008;311:1–8.
benchmark shell problems. Math Models Methods Appl Sci
[27] Hernandez E, Hervella-Nieto L, Rodriguez R. Computation of the vibration
2003;13:1279–302.
modes of plates and shells by low-order MITC quadrilateral finite elements.
[9] Baiocchi C, Lovadina C. A shell classification by interpolation. Math Models
Comput Struct 2003;81:615–28.
Methods Appl Sci 2002;12(10):1359–80.