Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

www.elsevier.com/locate/!nel

Popular benchmark problems for

geometric nonlinear analysis of shells
K.Y. Szea;∗ , X.H. Liua;1 , S.H. Lob
a
Departments of Mechanical Engineering, The University of Hong Kong, Pokfulam Road,
Hong Kong SAR, PR China
b
The Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong SAR, PR China
Received 15 June 2003; received in revised form 8 November 2003; accepted 8 November 2003

Abstract

In most, if not all, of the previous work on !nite element formulation and nonlinear solution procedures,
results of geometric nonlinear benchmark problems of shells are presented in the form of load–de8ection
curves. In this paper, eight sets of popularly employed benchmark problems are identi!ed and their detailed
reference solutions are obtained and tabulated. It is hoped that these solutions will form a convenient basis
for subsequent comparison and that the tedious yet inaccurate task of reconstructing data points by graphical
measurement of previously reported load–de8ection curves can be avoided. Moreover, the relative convergent
di;culty of the problems are revealed by the number of load increments and the total number of iterations
required by an automatic load incrementation scheme for attaining the converged solutions under the maximum
loads.
? 2003 Elsevier B.V. All rights reserved.

Keywords: Finite element; Geometric nonlinear; Benchmark; Shell

1. Introduction

To examine or demonstrate the accuracy of new !nite element models or the eAectiveness of new
nonlinear solution procedures, popular benchmark problems are often exercised and the predictions
are compared to some reference solutions. Since analytical solutions of shell problems are very
limited, most of reference solutions are previously reported numerical solutions. For linear benchmark
problems, these solutions can be conveniently and concisely expressed in terms of numerical !gures.
∗
Corresponding author. Tel.: +852-2859-2637; fax: +852-2858-5415.
E-mail address: kysze@hku.hk (K.Y. Sze).
1
On leave from Department of Mechanics, Huazhong University of Science & Technology, Wuhan 430074, PR China.

0168-874X/$ - see front matter ? 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.!nel.2003.11.001
1552 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

To this end, the problem sets stipulated by MacNeal and Harder [1] and Hitchings et al. [2] include
some of the most widely attempted tests. Ref. [1] has been well-received in the academic community.
Ref. [2] is a publication of UK’s National Agency for Finite Element Methods and Standards
(NAFEMS) and is mainly adopted by software developers for quality assurance. Noticeably, there
are a number of cases common to both problem sets.
For geometric nonlinear analysis of shells, the most recent and relevant NAFEMS publication is
probably the one by Prinja and Clegg [3]. Besides Ref. [3], more than 40 research papers [4–49]
on geometric nonlinear !nite element formulation have also been surveyed. For conciseness, the
reported predictions are mostly presented in the form of load–de8ection curves. Occasionally, the
results are reported numerically at a few selected load levels. In order to compare the predictions
of new !nite element models or procedures with their precedents, reconstructing the previously
reported load–de8ection curves by extracting data points using graphical measurement is the obvious
and, perhaps, the only choice. The practice is not only inaccurate but also time consuming. In
this context, eight sets of popular benchmark problems for geometric nonlinear analysis of shell
are selected from those considered in Refs. [3–49]. They are attempted by using ABAQUS’s S4R
four-node shell element models [50]. This paper will provide su;cient data points in numerical
format so that the relevant load–de8ection curves can be accurately and e;ciently reconstructed.
To reveal the relative convergent di;culty of the problems, the number of load increments and the
total number of iterations required by ABAQUS’s default automatic load incrementation scheme for
attaining the converged solutions under the maximum loads are reported.

2. Load incrementation scheme

In the nonlinear solution procedure, the full Newton–Raphson method is used. The default con-
vergence criteria are always employed and they are the simultaneous 0.5% force tolerance and 1%
displacement tolerance. The default automatic load incrementation scheme in ABAQUS is adopted
and the procedure is portrayed in Fig. 1. Throughout the scheme, the maximum load Pmax will be
automatically subdivided into NINC load increments which are not necessarily uniform. At the end
of each load increment, a converged intermediate solution is obtained. This reduces the degree of
nonlinearity from an intermediate solution state to another and enhances the chance of obtaining
the ultimate solution. The latter is the one under the maximum load Pmax . The scheme starts with
the load increment SP set to the maximum load Pmax . If the solution cannot converge within 16
iterations (counted by m) or if the solution diverges, the scheme abandons the increment and starts
again with the load increment reduced to one-quarter of the previous value. If the solution still fails
to converge, the scheme further reduces the increment size again. If the solution fails after 5 attempts
(counted by n) of load increment reduction, the analysis will be stopped or aborted. On the other
hand, it automatically increases the load increment by 50% if the last two converged solutions are
both obtained within 5 iterations. If the scheme is not aborted, ABAQUS outputs NINC and NITER.
The latter is the total number of iterations required to obtained the NINC converged intermediate so-
lutions. In this paper, both NINC and NITER are reported to reveal the relative convergent di;culty
of the considered problems.
For the sake of benchmarking, intermediate solutions given at uniform load intervals are de-
sired. However, these solutions cannot be yielded by the afore-discussed default automatic load
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1553

Fig. 1. The automatic load incrementation scheme.

incrementation scheme. To this end, the solutions to be reported are computed by dividing the max-
imum load into a number of equal load increments NINC∗ , where further subdivision of the load
increment is suppressed. In order that the ultimate solution can be successfully obtained, NINC∗ is
often much larger than the NINC required by the default automatic load incrementation scheme.

3. Benchmark tests

In this section, eight sets of popularly employed geometric nonlinear benchmark problems on
beams, plates, cylindrical shells and spherical shell are selected. A small portion of the considered
structures are laminated. In the subsequent description, the following nomenclature is employed:

 Poisson’s ratio
b width
E elastic modulus
G shear modulus
h thickness of beam, plate or shell
I second moment of area
1554 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

Table 1
Features of S4R, S4R5 and S9R5 curved shell element models

S4R S4R5 S9R5

No. of nodes per element 4 4 9

No. of d.o.f.s per node 6 5 5
Hourglass treatment Default stabilization Default stabilization Default stabilization
Drilling rotation treatment Default stabilization Not applicable Not applicable
Lagrangian framework Updated Total Total
Applicable strain Finite/large Small Small
Intended thickness Thin and thick Thin only Thin only

L length or longitudinal length

M; Mmax applied moment and maximum applied moment, respectively
NINC the total number of load increments that yield the ultimate solution (determined by
the default load incrementation scheme),
NINC∗ the number of equal load increments used to obtain the plotted and tabulated data
NITER the total number of iterations that lead to the NINC convergent solutions
P; Pmax applied force and maximum applied force, respectively
R mean radius
U , V and W displacements along the Cartesian coordinates X , Y and Z, respectively.

All benchmark problems to be presented have been attempted by ABAQUS’s S4R, S4R5 and
S9R5 curved shell element models. All of them are Reduced-integrated elements with hourglass
control and their features are summarized in Table 1. In particular, the formulation of S4R changes
when the shell thickness increases from that of the discrete KirchhoA shell to that of the thick shell.
On the other hand, both S4R5 and S9R5 are only recommended for thin shell analysis [50]. In
nearly all problems, the deformed structures will be portrayed and the displacement ampli!cation
factor is always taken to be unity.
When the default automatic load incrementation scheme is adopted, the solution procedures of
S4R5 are aborted before the load reaches its maximum in many benchmark problems. The situation of
S9R5 is even worse. When equal load increments are imposed, S4R5 and S9R5 require considerably
larger number of load increments than that of S4R for securing the ultimate solution. In this light,
only the predictions of S4R and the related NINC and NITER output at the end of the automatic
load incrementation scheme will be reported.
To ensure that the reported solutions have been su;ciently converged with respect to the mesh
density, the mesh is re!ned until the solutions yielded by two successively re!ned meshes are prac-
tically identical. To illustrate that the mesh density is adequate, results predicted by two diAerent
meshes will be shown in all load–de8ection curves. The two sets of results are graphically indistin-
guishable and the discrepancy is typically around 0.1%. The precise relations between the applied
loads and the selected de8ections in numerical format are mostly reported at a load interval given
as 5% of the maximum load. If deemed to be necessary, additional data points are employed to
enhance the quality of the so-constructed load–de8ection curves. Lastly, all problems are geometric
nonlinear in the narrow sense and material nonlinearity is not considered.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1555

E=1.2×106
ν = 0, L = 10
b = 1, h = 0.1
Po = EI/L2 = 1
Pmax = 4Po = 4
(a)

4
8x1 elements
16x1 elements P=4
3.5

2.5
End force

sh
U tip W tip
2

me
d
me
for
1.5

de
P
1

0.5 esh
initial m

0
0 1 2 3 4 5 6 7
(b) Tip deflections (c)

Fig. 2. (a) Cantilever subjected to end shear force. (b) Load–de8ection curves for cantilever subjected to end shear force.
(c) The deformed 16 × 1 mesh under the maximum force.

3.1. Cantilever subjected to end shear force

Fig. 2a shows a cantilever subjected to the end shear force P. The problem has been considered
in Refs. [5,8,19,21,29,32,35,36,45,47,48], among others. A commonly employed mesh for four-node
shell elements is 8 × 1 which is also adequate for the S4R element. Fig. 2b plots the end shear force
against the vertical and horizontal tip de8ections. Table 2a lists the same de8ections whereas Table 2b
lists the NINC and NITER. Fig. 2c portrays the deformed cantilever under the maximum load.

3.2. Cantilever subjected to end moment

Fig. 3a shows a cantilever subjected to end moment M . A commonly employed mesh for four-node
shell elements is 12 × 1. The problem has been considered in Refs. [5,10,13,17,18,21,29,30,32,34,
36,40,46], among others. The cantilever forms a circular arc with its radius R given by the classical
8exural formula R = EI=M . Using the formula, the analytical normalized de8ections can be derived
to be


U Mo M W Mo M
= sin − 1; = 1 − cos
L M Mo L M Mo
where Mo = EI=L. The maximum end moment Mmax is taken to be Mo at which the beam will
be bent into a circle. In this problem, accurate predictions can be yielded by 8 × 1 S4R elements.
1556 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

Table 2

P=Pmax −Utip Wtip P=Po −Utip Wtip P=Po −Utip Wtip

(a) Horizontal and vertical tip de8ections for the cantilever loaded with end shear forcea
0.05 0.026 0.663 0.40 1.184 4.292 0.75 2.541 6.031
0.10 0.103 1.309 0.45 1.396 4.631 0.80 2.705 6.190
0.15 0.224 1.922 0.50 1.604 4.933 0.85 2.861 6.335
0.20 0.381 2.493 0.55 1.807 5.202 0.90 3.010 6.467
0.25 0.563 3.015 0.60 2.002 5.444 0.95 3.151 6.588
0.30 0.763 3.488 0.65 2.190 5.660 1.00 3.286 6.698
0.35 0.971 3.912 0.70 2.370 5.855

8 × 1 S4R elements 16 × 1 S4R elements

(b) NINCs and NITERs required to obtained the ultimate solution for the cantilever loaded with end shear force
(See Fig. 2a.)
NINC 15 15
NITER 78 80
a
Computed by using 16 × 1 S4R elements and NINC∗ = 40, see Fig. 2a.

E=1.2×106
ν = 0, L = 12
b = 1, h = 0.1
Mo = EI/L = 25/3
Mmax = 2πMo = 50π/3
(a)
1
M=0.35M max
0.9

0.8

0.7
Endmoment (x50/3 )

W tip -U tip
0.6

0.5

0.4

0.3 M
M=0.7M max
sh
l me
0.2 exact initia
8x1 elements
0.1 16x1 elements
M=M max
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(b) Tipdeflection (x10) (c)

Fig. 3. (a) Cantilever subjected to end bending moment. (b) Load–de8ection curves for cantilever subjected to end bending
moment. (c) The deformed 16 × 1 mesh under the maximum bending moment.

Fig. 3b plots the end moment against the vertical and horizontal tip de8ections. Fig. 3c portrays the
deformed cantilevers at M =0:35Mmax ; 0:7Mmax and Mmax . Table 3a lists the analytical and computed
tip de8ections, which are highly consistent with each other. Table 3b lists the NINC and NITER.
Compared to the last cantilever problem, the present one converges less readily.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1557

Table 3

M=Mmax −Utip Wtip M=Mmax −Utip Wtip

S4R Exact S4R Exact S4R Exact S4R Exact

(a) Horizontal and vertical tip de8ections for the cantilever loaded with end bending momenta
0.05 0.196 0.196 1.870 1.870 0.55 13.075 13.073 6.788 6.775
0.10 0.773 0.774 3.648 3.648 0.60 13.875 13.871 5.772 5.758
0.15 1.698 1.699 5.249 5.248 0.65 14.384 14.377 4.678 4.665
0.20 2.916 2.918 6.600 6.598 0.70 14.603 14.595 3.583 3.571
0.25 4.357 4.361 7.643 7.639 0.75 14.556 14.546 2.556 2.546
0.30 5.942 5.945 8.338 8.333 0.80 14.280 14.270 1.656 1.650
0.35 7.582 7.585 8.671 8.664 0.85 13.826 13.818 0.931 0.926
0.40 9.191 9.194 8.646 8.637 0.90 13.254 13.247 0.407 0.405
0.45 10.687 10.688 8.291 8.281 0.95 12.625 12.621 0.099 0.098
0.50 12.000 12.000 7.652 7.639 1.00 12.000 12.000 0.000 0.000

8 × 1 S4R elements 16 × 1 S4R elements

(b) NINCs and NITERs required to obtain the ultimate solution for the cantilever loaded with end bending moment
(See Fig. 3a.)
NINC 125 125
NITER 715 714
a
Computed by using 16 × 1 S4R elements and NINC∗ = 80, see Fig. 3a.

3.3. Slit annular plate subjected to lifting line force

The slit annular plate is shown in Fig. 4a. The problem has been considered in Refs. [23,25,27,33,
36,39,40,44,46,48,49], among others. The line force P is applied at one end of the slit while the
other end of the slit is fully clamped. A commonly employed mesh for four-node shell elements
is 6 × 30 which is also adequate for the S4R element. Fig. 4b plots the load against the vertical
de8ections at the tips of the slit A and B. Table 4a lists the de8ections whereas Table 4b lists the
NINC and NITER. Fig. 4c shows the deformed slit plate under the maximum load.

3.4. Hemispherical shell subjected to alternating radial forces

Fig. 5a shows hemispherical shell with an 18◦ circular cutout at its pole. The shell is loaded
by alternating radial point forces Ps at 90◦ intervals. The problem has been considered in Refs.
[18,19,21,23,25,26,28,31,32,34,35,37,40,41,44–49], among others. Owing to symmetry, one-quarter
of the shell is modeled and a commonly employed mesh for four-node shell elements is 16 × 16.
Fig. 5b plots the load against the radial de8ections at the points of loading A and B. Table 5a lists
the same de8ections whereas Table 5b lists the NINC and NITER. Fig. 5a also shows the deformed
hemispherical shell under the maximum load. In this problem, reasonably accurate predictions can
be yielded by using 12 × 12 S4R elements.
1558 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

B E = 21×106,
ν =0
A
Ro Ri Ri = 6, Ro = 10
h = 0.03
Pmax = 0.8 (force/length)

(a)

8
P=0.8
6x30 elements
7 10x80 elements

sh
me
6

d
me
Force/unit length (x0.1)

for
de
5

WA WB
4
initial shape

0
0 0. 2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
(b) Vertical deflections at points A and B (x10) (c)

Fig. 4. (a) The slit annular plate loaded with the line force P. (b) Load–de8ection curves for the slit annular plate lifted
by line force P. (c) The deformed 10 × 80 mesh at P = Pmax .

Table 4

P=Pmax WA WB P=Pmax WA WB P=Pmax WA WB

(a) Vertical de8ections at points A and B of the slit annular plate lifted by a line forcea
0.025 1.305 1.789 0.30 8.340 11.213 0.70 11.970 15.469
0.05 2.455 3.370 0.35 8.974 11.992 0.75 12.310 15.842
0.075 3.435 4.720 0.40 9.529 12.661 0.80 12.642 16.202
0.10 4.277 5.876 0.45 10.023 13.247 0.85 12.966 16.550
0.125 5.007 6.872 0.50 10.468 13.768 0.90 13.282 16.886
0.15 5.649 7.736 0.55 10.876 14.240 0.95 13.590 17.212
0.20 6.725 9.160 0.60 11.257 14.674 1.00 13.891 17.528
0.25 7.602 10.288 0.65 11.620 15.081

6 × 30 S4R elements 10 × 80 S4R elements

(b) NINCs and NITERs required to obtain the ultimate solution for slit annular plate lifted by line force
(See Fig. 4a.)
NINC 61 67
NITER 327 346
a
10 × 80 S4R elements and NINC∗ = 640, see Fig. 4a.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1559

P P 400 E = 6.825×107
400
ν = 0.3
R = 10
h = 0.04
Pmax = 400

P A B 400 400
P
(a)
4
12x12 elements
16x16 elements
3.5

3
Radialforce, P (x100)

2.5

VA -U B
2

1.5

0.5

0
0 1 2 3 4 5 6 7 8 9
(b) Radial displacements at points A and B

Fig. 5. (a) The initial geometry and deformed geometry at P = Pmax for the hemispherical shell subjected to inward and
outward radial forces. (b) Load–de8ection curves for the hemisphere shell subjected to radial forces.

Table 5

P=Pmax VA −UB P=Pmax VA −UB P=Pmax VA −UB

(a) Radial de8ections at points A and B of the hemispherical shell problema

0.05 0.855 0.955 0.40 3.158 5.196 0.75 3.816 7.234
0.10 1.499 1.840 0.45 3.291 5.565 0.80 3.875 7.448
0.15 1.969 2.604 0.50 3.406 5.902 0.85 3.929 7.647
0.20 2.321 3.261 0.55 3.508 6.212 0.90 3.979 7.835
0.25 2.596 3.833 0.60 3.598 6.497 0.95 4.025 8.011
0.30 2.819 4.339 0.65 3.678 6.761 1.00 4.067 8.178
0.35 3.002 4.790 0.70 3.750 7.006

12 × 12 S4R elements 16 × 16 S4R elements

(b) NINCs and NITERs required to obtain the ultimate solution for the hemisphere shell problem
(See Fig. 5a.)
NINC 27 27
NITER 140 136
a
Computed by 16 × 16 S4R elements and NINC∗ = 40, see Fig. 5a.

3.5. Pullout of an open-ended cylindrical shell

Fig. 6a shows an open-ended cylinder being pulled by a pair of radial forces Ps. The problem
has been considered in Refs. [14,24,26,28,32,33,40,44,46,48,49], among others. Owing to symmetry,
1560 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

A
E = 10.5×106

free
ν = 0.3125

e
dge
R = 4.953
Z
L = 10.35
Y h = 0.094
Pmax = 40,000
R X B
C
fre
e

L
ed
ge

(a) P P=40000

4
16x24 elements
24x36 elements
3.5
Pulling force at point A (x10000)

-U C -U B
2.5

2 WA

1.5

-U B -U C
1

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(b) Displacements at points A, B and C (c) P=40000

Fig. 6. (a) The open-end cylindrical shell subjected to radial pulling forces. (b) Load–de8ection curves of the open-end
cylinder subjected to pulling forces. (c) The deformed 24×36 mesh for the open-ended cylindrical under the maximum load.

Table 6

P=Pmax WA −UB −UC P=Pmax WA −UB −UC P=Pmax WA −UB −UC

a
(a) Radial de8ections at points A, B and C of the open-ended cylindrical shell
0.025 0.819 0.864 0.872 0.35 2.321 3.342 3.556 0.70 2.672 4.385 3.378
0.05 1.260 1.471 1.493 0.40 2.376 3.443 3.632 0.75 2.692 4.423 3.353
0.075 1.527 1.901 1.946 0.45 2.425 3.539 3.688 0.80 2.710 4.455 3.332
0.10 1.707 2.217 2.293 0.50 2.473 3.653 3.718 0.85 2.726 4.483 3.313
0.15 1.936 2.641 2.792 0.525 2.543 4.061 3.580 0.90 2.741 4.508 3.297
0.20 2.079 2.904 3.106 0.55 2.577 4.171 3.518 0.95 2.755 4.530 3.283
0.25 2.180 3.087 3.310 0.60 2.618 4.274 3.452 1.00 2.768 4.551 3.269
0.30 2.257 3.227 3.452 0.65 2.648 4.338 3.410
16 × 24 S4R elements 24 × 36 S4R elements
(b) NINCs and NITERs required to obtain the ultimate solution for the open-ended cylindrical shell
(See Fig. 6a.)
NINC 18 18
NITER 91 94
a
Computed by using 24 × 36 S4R elements and NINC∗ = 400, see Fig. 6a.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1561

rigid diaphragm

E = 30×103
ν = 0.3
R = 100
Z
L = 200
Y
h=1
R X Pmax = 12,000
B

(a)
rigid diaphragm
P

1.2
40x40 elements P=12000
1.1 48x48 elements

0.9
Pinched force, P (x10000)

0.8

0.7
UB -W A
0.6

0.5

0.4

0.3

0.2

0.1
P=12000
0 1 2
-2 -1 0 3 4 5 6 7 8 9
(b) Displacements at points A and B (x10) (c)

Fig. 7. (a) Pinched cylindrical shell mounted on rigid end diaphragms. (b) Load–de8ection curves of the pinched cylinder
mounted over rigid end diaphragms. (c) The deformed 48 × 48 mesh of the pinched cylinder under maximum load.

Table 7

P=Pmax −WA UB P=Pmax −WA UB P=Pmax −WA UB

a
(a) Radial de8ections at points A and B of the pinched cylindrical shell
0.05 9.561 −0.233 0.30 65.498 17.979 0.70 78.451 29.772
0.075 15.648 −0.922 0.35 68.229 20.365 0.75 79.339 30.604
0.10 23.164 −2.391 0.40 70.424 22.321 0.80 80.218 31.471
0.125 29.375 −3.872 0.45 72.204 23.916 0.85 81.045 32.299
0.15 36.208 −2.154 0.50 73.790 25.381 0.90 81.766 32.989
0.175 51.499 6.792 0.55 75.139 26.631 0.95 82.435 33.619
0.20 56.373 10.448 0.60 76.331 27.735 1.00 83.102 34.272
0.25 61.877 14.905 0.65 77.472 28.843
40 × 40 S4R elements 48 × 48 S4R elements
(b) NINCs and NITERs required to obtain the ultimate solution for the pinched cylindrical shell
(See Fig. 7a.)
NINC 69 70
NITER 391 406
a
Computed by 48 × 48 S4R elements and NINC∗ = 1280, see Fig. 7a.
1562 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

2
32x32 elements
1.8 40x40 elements
P cla Isotropic Shell
m 1.6
pe
d
E=2.0685×107 , ν =0.3 1.4

Load at point A (x1000)

e A
fre
1.2
Laminated Shell
R 1
E L = 2068.5, ET = 517.125
G LT = 795.6, νLT = νTT = 0.3
0.8
Z
=0 =0
W =θ y W=θ y
Y isotropic
L 0.6
X lamination: [00/900/00] and [90/0/90]
[900/00/900] 0.4 [0/90/0]

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
Downward deflection at point A (x100)
(a) (b)

isotropic [0/90/0]
[90/0/90]

P=2000
P=2000

(c) (d) P=2000

Fig. 8. (a) The semi-cylindrical shell subjected to an end pinching force. (b) Load–de8ection curves of the semi-cylindrical
shell subjected to end pinching force. (c) The deformed 40×40 mesh of the semi-cylindrical isotopic shell under maximum
load. (d) The deformed 40 × 40 meshes of the semi-cylindrical [0◦ =90◦ =0◦ ] and [90◦ =0◦ =90◦ ].

one-eighth of the shell is modeled and a commonly employed mesh for four-node shell elements
is 8 × 12. Fig. 6b plots the load against the radial de8ections at points A, B and C. Table 6a
lists the data points whereas Table 6b lists the NINC and NITER. The deformed geometry under
the maximum load is portrayed in Fig. 6c. In this problem, reasonably accurate predictions can be
yielded by 16 × 24 S4R elements.

3.6. Pinched cylindrical shell mounted over rigid diaphragms

Fig. 7a shows a pinched cylindrical shell mounted on rigid end diaphragms over which the in-plane
displacements U and W are restrained. The problem and its variations have been considered in Refs.
[21,26,37,43–45,48], among others. Owing to symmetry, one-eighth of the shell is modeled and a
commonly employed mesh for four-node shell elements is 40 × 40 which is also adequate for the
S4R element. Fig. 7b plots the load against radial de8ections at points A and B. Table 7a lists the
same de8ections whereas Table 7b lists the NINC and NITER. The deformed geometry under the
maximum load is portrayed in Fig. 7c.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1563

Table 8

P=Pmax Isotropic [0◦ =90◦ =0◦ ] [90◦ =0◦ =90◦ ] P=Pmax Isotropic [0◦ =90◦ =0◦ ] [90◦ =0◦ =90◦ ]

(a) The downward de8ection at A of the pinched semi-cylindrical shellsa

0.05 5.421 15.340 7.558 0.45 124.751 177.404 132.488
0.10 16.100 37.920 22.722 0.50 132.653 180.680 138.740
0.125 22.195 55.145 30.594 0.55 138.920 183.544 144.238
0.15 27.657 93.433 37.897 0.60 144.185 186.099 149.191
0.175 32.700 129.575 45.427 0.65 148.770 188.415 153.728
0.20 37.582 141.562 54.455 0.70 152.863 190.543 157.930
0.225 42.633 149.034 65.814 0.75 156.584 192.520 161.854
0.25 48.537 154.634 79.512 0.80 160.015 194.376 165.540
0.275 56.355 159.141 92.524 0.85 163.211 196.132 169.017
0.30 66.410 162.896 102.357 0.90 166.200 197.808 172.308
0.325 79.810 166.099 109.793 0.95 168.973 199.420 175.430
0.35 94.669 168.884 115.746 1.00 171.505 200.983 178.386
0.40 113.704 173.560 125.094
32 × 32 S4R elements 40 × 40 S4R elements
◦ ◦ ◦ ◦ ◦ ◦
Isotropic [0 =90 =0 ] [90 =0 =90 ] Isotropic [0◦ =90◦ =0◦ ] [90◦ =0◦ =90◦ ]

(b) NINCs and NITERs required to obtain the ultimate solution for the pinched semi cylindrical
(See Fig. 8a.)
NINC 28 36 32 28 36 33
NITER 136 171 184 136 170 184
a
Computed by using 40 × 40 S4R elements and NINC∗ = 320, see Fig. 8a.

3.7. Pinched semi-cylindrical isotropic and laminated shells

Fig. 8a shows the semi-cylindrical shell subjected to an end pinching force at the middle of
the free-hanging circumferential periphery. The other circumferential periphery is fully clamped.
Along its longitudinal edges, the vertical de8ection and the rotation about the Y -axis are restrained.
Besides the isotropic material, laminates with stacking sequences [0◦ =90◦ =0◦ ] and [90◦ =0◦ =90◦ ] are
also considered. In the laminates, all plies are equal in thickness. A ply is of 0◦ if its !bres
are parallel to the longitudinal direction of the shell. The present problems have been considered in
Refs. [16,22,33,40,42,47,48], among others. Owing to symmetry, half of shell is modeled and a
commonly employed mesh for four-node shell elements is 16 × 16. Fig. 8b plots the applied force
against the downward de8ections at A. Table 8a lists the same de8ections whereas Table 8b lists
the NINC and NITER. The deformed shells under the maximum load are portrayed in Figs. 8c and
d. In this problem, reasonably accurate predictions can be yielded by 32 × 32 S4R elements.

3.8. Hinged cylindrical isotropic and laminated roofs

Fig. 9a shows the hinged semi-cylindrical roof subjected to a central pinching force. An isotropic
material, [0◦ =90◦ =0◦ ] laminate and [90◦ =0◦ =90◦ ] laminate at two diAerent thicknesses are considered.
Again, a ply is of 0◦ if its !bres are parallel to the longitudinal direction of the shell. All plies in the
1564 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

free R = 2540, L = 254

P
θ = 0.1 radian
d
ge h = 12.7 or 6.35
hin
C Pmax = 3000

Isotropic Shell
free
E = 3102.75 , ν = 0.3
θ ge
d
hin L Laminated Shell
EL = 3300, ET = 1100
L
GLT = 660, νLT = νTT = 0.25
lamination: [00/900/00] and
Z [900/00/900]
Y

(a) X

3.5
8x8 elements
16x16 elements
3

2.5
isotropic
Load, P (x1000)

2
[90/0/90]
1.5

1 [0/90/0]

0.5

-0.5
0 0.5 1 1.5 2 2.5 3 3.5
(b) Central deflection (x10)

3 3
16x16 elements 16x16 elements
24x24 elements 24x24 elements
2. 5 2.5

2 2
Load, P (x1000)
Load, P (x1000)

[90/0/90]
1. 5 1.5
[0/90/0]
1 isotropic 1

0. 5 0.5

0 0

-0. 5 -0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
(c) Central deflection (x10) Central deflection (x10)

Fig. 9. (a) Hinged cylindrical roof subjected to a central pinching force. (b) Load–de8ection curves of the 12.7 unit thick
hinged cylindrical roof. (c) Load–de8ection curves of the 6.35 unit thick hinged cylindrical roof.

same laminate are equal in thickness. Along the hinged edges, all nodal translations are restrained.
These problems have been considered in Refs. [4–12,15,17,18,20,21,24,26,28–30,33,36,38,43,47,48],
among others, and are particularly popular due to the snapping behavior. At some intermediate
state, the tangential global stiAness matrices become singular. The problems are often, if not always,
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1565

Table 9
De8ections for diAerent thickness isotropic hinged cylindrical shell

P=Pmax −WC P=Pmax −WC P=Pmax −WC P=Pmax −WC

(a) 12.7 Unit thicka

0.0877 0.693 0.7421 11.293 0.3245 16.590 0.3871 24.049
0.1980 1.638 0.7286 12.141 0.2717 17.094 0.4443 24.663
0.3473 3.087 0.7023 12.903 0.2272 17.657 0.5093 25.293
0.4686 4.477 0.6649 13.583 0.1940 18.299 0.5826 25.940
0.5647 5.802 0.6182 14.188 0.1750 19.028 0.6644 26.601
0.6381 7.057 0.5643 14.728 0.1729 19.852 0.7551 27.276
0.6908 8.237 0.5055 15.217 0.1905 20.771 0.8549 27.964
0.7246 9.339 0.4442 15.676 0.2303 21.780 0.9643 28.663
0.7412 10.358 0.3830 16.125 0.2950 22.875 1.0835 29.374

(b) 12.7 unit thick [0◦ =90◦ =0◦ ]a

0.0546 0.807 0.3577 10.735 0.0638 16.199 0.0194 22.707
0.1245 1.956 0.3464 11.516 0.0303 16.648 0.0697 23.824
0.1938 3.281 0.3288 12.227 0.0007 17.141 0.1354 25.005
0.2494 4.548 0.3057 12.872 −0.0236 17.692 0.2178 26.244
0.2925 5.753 0.2778 13.458 −0.0411 18.317 0.3181 27.533
0.3244 6.892 0.2461 13.990 −0.0504 19.024 0.4374 28.864
0.3459 7.961 0.2115 14.477 −0.0503 19.817 0.5769 30.229
0.3582 8.959 0.1749 14.928 −0.0394 20.698 0.7374 31.622
0.3618 9.884 0.0999 15.774 −0.0165 21.663 1.0192 33.748

(c) 12.7 unit thick [90◦ =0◦ =90◦ ]a

0.0556 0.649 0.5336 9.545 0.5801 16.019 0.2834 20.881
0.1299 1.581 0.5550 10.404 0.5685 16.545 0.2789 21.318
0.2090 2.673 0.5716 11.231 0.5542 17.028 0.2812 21.823
0.2784 3.740 0.5838 12.024 0.5374 17.465 0.3099 23.030
0.3389 4.781 0.5919 12.782 0.4971 18.206 0.3755 24.476
0.3914 5.794 0.5963 13.505 0.4498 18.785 0.4832 26.120
0.4365 6.777 0.5970 14.192 0.3991 19.249 0.6380 27.921
0.4748 7.731 0.5945 14.841 0.3501 19.682 0.8444 29.844
0.5070 8.654 0.5888 15.450 0.3090 20.191 1.0356 31.349

(d) 6.35 unit thickb

0.0517 1.846 0.1671 15.501 −0.1001 14.520 −0.0006 24.824
0.1182 5.271 0.1323 16.145 −0.1142 14.451 0.0626 26.565
0.1583 8.257 0.0923 16.602 −0.1247 14.862 0.1427 28.302
0.1837 10.799 0.0504 16.915 −0.1288 15.778 0.2403 30.023
0.1914 11.904 0.0083 17.008 −0.1271 16.961 0.3559 31.720
0.1953 12.892 −0.0312 16.697 −0.1196 18.320 0.4898 33.388
0.1950 13.752 −0.0622 15.780 −0.1055 19.817 0.6417 35.024
0.1901 14.472 −0.0739 15.206 −0.0825 21.420 0.8114 36.626
0.1806 15.050 −0.0861 14.767 −0.0484 23.100 1.0313 38.450

(e) 6.35 unit thick [0◦ =90◦ =0◦ ]b

0.0423 3.414 −0.0817 15.574 0.0593 11.860 −0.0731 17.712
0.0765 8.834 −0.0779 12.857 0.0428 10.506 −0.0593 20.694
1566 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

Table 9 (Continued)

P=Pmax −WC P=Pmax −WC P=Pmax −WC P=Pmax −WC

0.0782 12.280 −0.0650 10.991 0.0210 10.406 −0.0214 24.230

0.0564 14.492 −0.0442 10.126 −0.0006 10.745 0.0556 27.809
0.0271 16.397 −0.0189 10.494 −0.0215 11.263 0.1700 31.174
−0.0059 18.017 0.0007 11.931 −0.0419 11.861 0.3213 34.362
−0.0240 18.602 0.0142 13.623 −0.0607 12.555 0.5085 37.394
−0.0436 18.875 0.0296 14.751 −0.0742 13.602 0.7292 40.274
−0.0657 18.365 0.0499 14.425 −0.0780 15.332 1.0243 43.444

(f) 6.35 unit thick [90◦ =0◦ =90◦ ]b

0.0490 2.699 0.1499 18.140 −0.0117 14.369 −0.0402 22.448
0.0822 5.205 0.1385 18.954 −0.0168 13.519 0.0036 24.905
0.1063 7.479 0.1222 19.416 −0.0297 13.456 0.0693 27.427
0.1249 9.527 0.1019 19.429 −0.0458 13.873 0.1598 29.941
0.1393 11.374 0.0792 19.167 −0.0619 14.469 0.2761 32.401
0.1498 13.043 0.0331 18.249 −0.0752 15.258 0.4180 34.786
0.1562 14.550 0.0134 17.555 −0.0818 16.461 0.5847 37.088
0.1585 15.905 −0.0011 16.641 −0.0792 18.129 0.7747 39.301
0.1565 17.107 −0.0088 15.532 −0.0663 20.157 1.0234 41.773
a
Computed by using 16 × 16 S4R elements and Riks method, see Fig. 9a.
b
Computed by using 24 × 24 S4R elements and Riks method, see Fig. 9a.

solved by Riks solution method and such an option in ABAQUS is adopted. Owing to symmetry,
one-quarter of the roof is modeled and a commonly employed mesh for four-node shell elements is
4 × 4. For the 12.7 unit thick shells, reasonably accurate predictions can be yielded by using 8 × 8
S4R elements as shown in Fig. 9b. For the 6.35 unit thick shells, the same mesh is inadequate.
However, reasonably accurate predictions can be obtained by using 16 × 16 S4R elements as shown
in Fig. 9c. Tables 9a–f list the data points. With the Riks solution method, it is not possible to
yield a solution at a preset load level. This explains why the load levels in the tables are diAerent
whereas the load–de8ections curves produced by the coarser meshes do not reach the last solution
points yielded by the !ner meshes in Figs. 9b and c. Despite the high nonlinearity, the maximum
de8ections are much smaller than the overall dimensions of the roof. Unlike the previous problems,
the deformed meshes are not shown as they can hardly be distinguished from the undeformed ones.
As the default automatic load incrementation scheme does not work for the present problems, there
are no NINC and NITER.

4. Closure

From more than 40 publications on geometric nonlinear analysis of shells, eight sets of popu-
larly employed benchmark problems are identi!ed and the detailed reference solutions are tabulated.
It is hoped that the solutions will form a convenient basis for subsequent comparison and that the
inaccurate and time consuming task of reconstructing data points by graphical measurement of previ-
ously reported load–de8ection curves can be avoided. To reveal the relative convergent di;culty, the
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1567

number of load increments (NINC) and the number iterations (NITER) required by an automatic
load incrementation scheme to attain the maximum loads are also reported. Care has been exer-
cised to ensure that the reported solutions have been highly converged with respect to the mesh
density. It is interesting to note that once the employed mesh becomes su;ciently !ne, the NINC
and the NITER are not sensitive to further mesh re!nement. In view of converging di;culty, the
most demanding problems are the cantilever subjected to end moment and the hinged cylindrical
roofs. While the hinged roof problems must be solved by Riks method, NINC and NITER of the
cantilever problem are signi!cantly higher than that of the remaining benchmark cases.

Acknowledgements

The !nancial support of the William Mong Engineering Research Fund in form of a young
researcher award to the second author is gratefully acknowledged. Mr.W.K.Chan is thanked for
preparing some of the preliminary ABAQUS input date !les.

References

[1] R.H. MacNeal, R.L. Harder, A proposed standard set of problems to test !nite element accuracy, Finite Elem. Anal.
Des. 1 (1985) 3–20.
[2] D. Hitchings, A. Kamoulakos, G.A.O. Davies, Linear Statics Benchmarks. National Agency for Finite Element
Methods and Standards, Glasgow, UK, 1987.
[3] N.K. Prinja, R.A. Clegg, Assembly Benchmark Tests for 3-D Beams and Shells Exhibiting Geometric Non-Linear
Behaviour, NAFEMS, Glasgow, UK, 1993.
[4] A.B. Sabir, A.C. Lock, The application of !nite elements to the large de8ection geometrically nonlinear behaviour
of cylindrical shells, in: C.A. Brebbia, H. Tottenham (Eds.), Variational Methods in Engineering, Southampton
University Press, Southampton, UK, 1972, 7/54-7/65.
[5] G. Horrigmoe, P.G. Bergan, Nonlinear analysis of free-from shells by 8at !nite elements, Comput. Methods Appl.
Mech. Eng. 16 (1978) 11–35.
[6] T.J.R. Hughes, W.K. Liu, Nonlinear !nite element analysis of shells, Part I: Three-dimensional shells, Comput.
Methods Appl. Mech. Eng. 26 (1981) 331–362.
[7] M.A. Cris!eld, A fast incremental/iterative solution procedure that handles “snap-through”, Comput. Struct. 13 (1981)
55–62.
[8] H. Parisch, Large displacements of shells including material nonlinearities, Comput. Methods Appl. Mech. Eng. 27
(1981) 183–214.
[9] K.S. Surana, Geometrically non-linear formulation for the three dimensional solid-shell transition !nite elements,
Comput. Struct. 15 (1982) 549–566.
[10] K.S. Surana, Geometrically nonlinear formulation for curved shell elements, Int. J. Numer. Methods Eng. 19 (1983)
581–615.
[11] J. Oliver, E. Onate, A total Lagrangian formulation for the geometrically nonlinear analysis of structures using !nite
elements. Part I. Two-dimensional problems: shell and plate structures, Int. J. Numer. Methods Eng. 20 (1984)
2253–2281.
[12] S. Saigal, R. Kapania, Y. Yang, Geometric nonlinear !nite element analysis of imperfect laminated shells, J. Compos.
Mater. 20 (1986) 197–214.
[13] J.C. Simo, L. Vu-Quoc, A three-dimensional !nite strain rod model, Part II: computational aspects, Comput. Methods
Appl. Mech. Eng. 58 (1986) 79–116.
[14] F. Gruttmann, E. Stein, P. Wriggers, Theory and numerics of thin elastic shells with !nite rotations, Ing.-Arch. 59
(1989) 54–67.
1568 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569

[15] C.H. Yeom, S.W. Lee, An assumed strain !nite element model for large de8ection composite Shells, Int. J. Numer.
Methods Eng. 28 (1989) 1749–1768.
[16] N. Stander, A. Matzenmiller, E. Ramm, An assessment of assumed strain method in !nite rotation shell analysis,
Eng.Comput. 6 (1989) 58–66.
[17] K.M. Hsiao, Y.R. Chen, Nonlinear analysis of shell structures by degenerated isoparametric shell element, Comput.
Struct. 31 (1989) 427–438.
[18] J.C. Simo, D.D. Fox, M.S. Rifai, On a stress resultant geometrically exact shell model, Part III: computational
aspects of the nonlinear theory, Comput. Methods Appl. Mech. Eng. 79 (1990) 21–70.
[19] J.C. Simo, M.S. Rifai, D.D. Fox, On a stress resultant geometrically exact shell model, Part IV: variable thickness
shells with through-the-thickness stretching, Comput. Methods Appl. Mech. Eng. 81 (1990) 91–126.
[20] G. Laschet, J.P. Jeusette, Postbuckling !nite element analysis of composite panels, Compos. Struct. 14 (1990) 35–48.
[21] A.F. Saleeb, T.Y. Chang, W. Graf, S. Yingyeunyong, A hybrid/mixed model for non-linear shell analysis and its
applications to large-rotation problems, Int. J. Numer. Methods Eng. 29 (1990) 407–446.
[22] H. Parisch, An investigation of a !nite rotation four node assumed strain shell element, Int. J. Numer. Methods Eng.
31 (1991) 127–150.
[23] N. Buechter, E. Ramm, Shell theory versus degeneration-a comparison in large rotation !nite element analysis, Int.
J. Numer. Methods Eng. 34 (1992) 39–59.
[24] X. Peng, M.A. Cris!eld, A consistent corotational formulation for shells using the constant stress/ constant moment
triangle, Int. J. Numer. Methods Eng. 35 (1992) 1829–1847.
[25] Y. Basar, Y. Ding, Finite-rotation shell elements for the analysis of !nite-rotation shell problems, Int. J. Numer.
Methods Eng. 34 (1992) 165–169.
[26] C. Sansour, H. Bu8er, An exact !nite rotation shell theory, its mixed variational formulation and its !nite element
implementation, Int. J. Numer. Methods Eng. 34 (1992) 73–115.
[27] P. Wriggers, F. Gruttmann, Thin shells with !nite rotations formulated in Biot stress: theory and !nite element
formulation, Int. J. Numer. Methods Eng. 36 (1993) 2049–2071.
[28] L. Jiang, M.W. Chernuka, A simple four-noded corotational shell element for arbitrarily large rotations, Comput.
Struct. 53 (1994) 1123–1132.
[29] C.W.S. To, M.L. Liu, Hybrid strain based three node 8at triangular shell elements II: numerical investigation of
nonlinear problems, Comput. Struct. 54 (1995) 1057–1076.
[30] F.G. Flores, E. Onate, F. Zarate, New assumed strain triangles for non linear shell analysis, Comput. Mech. 17
(1995) 107–114.
[31] H. Parisch, A continuum-based shell theory for non-linear applications, Int. J. Numer. Methods Eng. 38 (1995)
1855–1883.
[32] H.C. Park, C. Cho, S.W. Lee, An e;cient assumed strain element model with six dof per node for geometrically
nonlinear shells, Int. J. Numer. Methods Eng. 38 (1995) 4101–4122.
[33] B. Brank, D. Peric, F.B. Damjanic, On implementation of a nonlinear four node shell !nite element for thin
multilayered elastic shells, Comput. Mech. 16 (1995) 341–359.
[34] P. Betsch, E. Stein, An assumed strain approach avoiding arti!cial thickness straining for a non-linear 4-node shell
element, Commun. Numer. Methods Eng. 11 (1995) 899–909.
[35] P. Betsch, F. Gruttmann, E. Stein, A 4-node !nite shell element for the implementation of general hyperelastic
3D-elasticity at !nite strains, Comput. Methods Appl. Mech. Eng. 130 (1996) 57–79.
[36] A. Barut, E. Madenci, A. Tessler, Nonlinear analysis of laminates through a Mindlin-type shear deformable shallow
shell element, Comput. Methods Appl. Mech. Eng. 143 (1997) 155–173.
[37] R. Hauptmann, K. Schweizerhof, A systematic development of solid-shell element formulations for linear and non-
linear analyses employing only displacement degrees of freedom, Int. J. Numer. Methods Eng. 42 (1998) 49–69.
[38] S.J. Lee, Kanok-Numkulchai, A nine-node assumed strain !nite element for large-deformation analysis of laminated
shells, Int. J. Numer. Methods Eng. 42 (1998) 777–798.
[39] C. Sansour, J. Bocko, On hybrid stress, hybrid strain and enhanced strain !nite element formulations for a geo-
metrically exact shell theory with drilling degrees of freedom, Int. J. Numer. Methods Eng. 43 (1998) 175–192.
[40] P. Mohan, R.K. Kapania, Updated Lagrangian formulation of a 8at triangular element for thin laminated shells,
AIAA J. 36 (1998) 273–281.
 K.Y. Sze et al. / Finite Elements in Analysis and Design 40 (2004) 1551 – 1569 1569

[41] E. Providas, M.A. Kattis, A simple !nite element model for the geometrically nonlinear analysis of thin shells,
Comput. Mech. 24 (1999) 127–137.
[42] S. Klinkel, F. Gruttmann, W. Wagner, A continuum based three-dimensional shell element for laminated structures,
Comput. Struct. 71 (1999) 43–62.
[43] K.Y. Sze, S.J. Zheng, A hybrid stress nine-node degenerated shell element for geometric nonlinear analysis, Comput.
Mech. 23 (1999) 448–456.
[44] C. Sansour, F.G. Kollmann, Families of 4-node and 9-node !nite elements for a !nite deformation shell theory. An
assessment of hybrid stress, hybrid strain and enhanced strain elements, Comput. Mech. 24 (2000) 435–447.
[45] N. El-Abbasi, S.A. Meguid, A new shell element accounting for through-thickness deformation, Comput. Methods
Appl. Mech. Eng. 189 (2000) 841–862.
[46] W.I. Hong, J.H. Kim, Y.H. Kim, S.W. Lee, An assumed strain triangular curved solid shell element formulation for
analysis for plates and shells undergoing !nite rotations, Int. J. Numer. Methods Eng. 52 (2001) 747–761.
[47] K.Y. Sze, S.-J. Zheng, A stabilized hybrid-stress solid element for geometrically nonlinear homogeneous and
laminated shell analyses, Comput. Methods Appl. Mech. Eng 191 (2002) 1945–1966.
[48] K.Y. Sze, W.K. Chan, T.H.H. Pian, An eight-node hybrid-stress solid-shell element for geometric nonlinear analysis
of elastic shells, Int. J. Numer. Methods Eng. 55 (2002) 853–878.
[49] C.H. Kim, K.Y. Sze, Y.H. Kim, Curved quadratic triangular degenerated- and solid-shell elements for geometric
nonlinear analysis, Int. J. Numer. Methods Eng. 57 (2003) 2077–2097.
[50] ABAQUS, ABAQUS Theory and User’s Manuals, Version 5.8, Hibbitt, Karlsson & Sorensen, Inc., Pawtucket, Rhode
Island, USA, 1998.