Chapter 2

Shell Element Formulations for General

Nonlinear Analysis. Modeling Techniques

2.1 Introduction

In 1970, Ahmad, Irons and Zienkiewicz [1] presented a shell element formulation
that after many years still constitutes the basis for modern finite element analysis
of shell structures. The original formulation was afterwards extended to material
and geometric nonlinear analysis under the constraint of the infinitesimal strains
assumption [2–4].
The fundamental features of the A-I-Z shell element are,
• using isoparametric interpolation functions the displacements inside the shell
element are interpolated from three displacement-d.o.f. and two rotation-d.o.f. at
each node,
• the interpolated generalized displacement fields present C 0 continuity,
• the element is not based on any plate/shell theory but it is a continuum element
incorporating several assumptions that we list below (degenerated solid
element).
The kinematic and constitutive assumptions are,
• a straight line that is initially normal to the mid-surface remains straight after the
deformation,
• a straight line that is initially normal to the mid-surface is not stretched during
the deformation,
• the through-the-thickness stresses are zero.
It is important to remark that the second assumption precludes the consideration
of finite strain kinematics.
Although the A-I-Z shell element was a breakthrough in the field of finite
element analysis of shell structures, it suffers from the locking phenomenon and
much research effort has been devoted to the development of A-I-Z type elements
that do not incorporate this problem [5, 6].
The MITC4 shell element [7–9] which was developed to overcome the locking
problem of the A-I-Z shell elements has become, since its development in the early

E. N. Dvorkin and R. G. Toscano, Finite Element Analysis of the Collapse 9

and Post-Collapse Behavior of Steel Pipes: Applications to the Oil Industry,
SpringerBriefs in Computational Mechanics, DOI: 10.1007/978-3-642-37361-9_2,
 The Author(s) 2013
10 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

1980s, the standard shell element for many finite element codes. However, the
limitation of infinitesimal strains is still present in the MITC4 formulation.
Many researchers have contributed to the development of shell elements that
can model finite strain situations, among them,
• an early contribution by Rodal and Witmer for elasto-viscoplastic material
models (J2 ) where, at each iteration, after going through the displacements
calculation, the shell element thickness is updated neglecting the elastic strains
and invoking the incompressibility of the viscoplastic flow (J2 ) [10],
• in 1983 Hughes and Carnoy [11] developed a finite strain shell element for the
Mooney-Rivlin material model which uses a plane-stress constitutive relation
for the laminae and updates afterwards the thickness via a staggered iterative
formulation,
• Simo and co-workers in the period 1988–1992 developed a complete 3D non-
linear shell element formulation [12–16],
• Ramm and co-workers developed 3D shell elements considering also through-
the-thickness stretching [17, 18].
In 1995 Dvorkin et al. developed the MITC4-TLH element, that based on the
original MITC4 formulation can model finite strain elasto-plastic (J2 ) deforma-
tions. This element imposes the condition of zero transversal stresses and its
computational cost was rather high [19, 20].
Later, Toscano and Dvorkin developed an element that is also based on the
MITC4 formulation and can efficiently model finite strain deformations using a
general 3D material model: the MITC4-3D element [21, 22].
The most relevant differences with the original MITC4 formulation are:
• for each quadrilateral element there are 22 d.o.f.: 5 generalized displacements
per node plus 2 extra d.o.f. to incorporate the through-the-thickness stretching,
these extra d.o.f. are condensed at the element level;
• a general 3D constitutive relation is used, instead of the original laminae plane
stress constitutive relation.

2.2 The Standard A-I-Z Quadrilateral Shell Element

for Linear Analysis

2.2.1 Linear Analysis Kinematics

When modeling a shell we define, on its mid-surface, nodes and at those nodes we
define director vectors which are the best approximation to the shell mid-surface
normal at the corresponding nodes. The A-I-Z quadrilateral element is defined
using four nodes which are not necessarily coplanar.
2.2 The Standard A-I-Z Quadrilateral Shell Element for Linear Analysis 11

Fig. 2.1 Kinematics of the

A-I-Z shell element 0 1 v n v 1n
0 2
v v 2n
n
1 1
2 2

0 v 3n 0 v 4n v 3n v 4n
4 4
3 3
0 z; z

0 y; y

0 x; x

Under the assumption of infinitesimal strains, for a configuration at a time s, a

point inside the shell element, with natural coordinates [5] ðr; s; tÞ (see Fig. 2.1), is
defined by the Cartesian coordinates,1

s t
xðr; s; tÞ ¼ hk ðr; sÞ s xk þ hk ðr; sÞ½as V n k : ð2:1Þ
2
In the above equation,
hk 2D isoparametric interpolation function corresponding to the k-node [5];
s
xk position vector of the mid-surface k-node at time s;
ajk shell thickness at the k-node (assumed as invariant during the
deformation);   
V jk director vector at the k-node at time s s V n jk  ¼ 1 ;
s n

while the natural coordinates ðr; sÞ are defined on the element mid-surface
ðt ¼ 0Þ the natural coordinate t is measured at any point along the corresponding
director vector direction. The second term on the r.h.s. in Eq. (2.1) shows that at
any point on the element mid-surface the unit director vector times the thickness is
interpolated from the nodal values.
The geometry interpolation in Eq. (2.1) presents C0 continuity.
For describing the kinematics of the A-I-Z element the two main assumptions
are:
• the element thickness remains constant due to the assumed infinitesimal strains
deformation;
• the director vectors remain straight during the deformation.

The covariant base vectors of the ðr; s; tÞ system are determined deriving Eq. (2.1),

1 P
We use Einstein’s notation: ak bk  k ak bk , that is to say repeated indices indicate a
summation.
12 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

s o sx
gi ¼ ð2:2Þ
ori
and the contravariant base vectors need to fulfill the relation,
s i
g  s gj ¼ dij ð2:3Þ

For linear kinematics we consider the s-configuration to be coincident with the

0-(reference) configuration.
The displacements for a time s are,
s
u ¼ s x  0 x: ð2:4Þ
Using the above kinematic assumptions we get,

s t   
u ¼ hk s uk þ hk a s V n  0 V n k : ð2:5Þ
2
Let us now define in the 0-configuration at the node k two vectors that with the
nodal director vector form the ortho-normal basis ð0 V 1 ; 0 V 2 ; 0 V n Þ. We can write,
for infinitesimal rotations [23, 24],
s
V kn ¼ 0 V kn þ s hk  0 V kn
s
hk ¼ ak 0 V k1 þ bk 0 V k2 ð2:6Þ
s
V kn ¼ 0
V kn þ bk V k1
0
 0
ak V k2 :
Therefore,

s t  
u ¼ hk s uk þ hk a ða 0 V 2 þ b 0 V 1 Þ k : ð2:7Þ
2
It is apparent from Eq. (2.7) that this element formulation introduces 5 d.o.f. per
node.
At any point inside the shell we can write the infinitesimal strain tensor in terms
of its covariant components (~elm ) in the ðr; s; tÞ curvilinear system and the corre-
sponding contravariant base vectors,

e ¼ ee ij o gi o g j ð2:8Þ

where we use the notation (o gi o g j ), to indicate the dyadic (tensorial) product

between the two contravariant base vectors.2 In Eq. (2.8), ee tt ¼ 0 because the
thickness is constant.
From the kinematic relations between strain components and displacements
[24] we get,

2
Some authors use the notation o gi  o g j .
2.2 The Standard A-I-Z Quadrilateral Shell Element for Linear Analysis 13

 
1 0 ou 0 ou
ee ij ¼ gi  þ gj  : ð2:9Þ
2 ori ori
Hence, using Voigt notation we can write
~U
ee ¼ B ð2:10Þ
where ee is the (5 9 1) column vector formed with the non-zero curvilinear
components of the strain tensor, U is the (20 9 1) column vector with the element
nodal generalized displacements and B ~ is the (5 9 20) strain-displacement matrix,
formed using Eq. (2.9) [7].

2.2.2 Stress-Strain Relations

The assumption of zero stresses through the thickness is equivalent to consider that
each surface parallel to the mid-surface is in a plane stress condition. In the A-I-Z
finite element discretization, with only C 0 continuity, there are two alternative
ways for imposing through the shell thickness the plane stress condition,
• imposing it to the different laminae with constant t;
• imposing it at every point to the surfaces normal to the director vector.

We have chosen the second alternative.

At each point inside the shell element we define the local Cartesian system
ðbe 1 ; b
e2; b
e 3 Þ with,
0
g2  0 g3
e1 ¼ 
b 0


 g2  0 g3 
0
g3 ð2:11Þ
e3 ¼ 
b 0 

 g3 
b
e2 ¼ b
e 3  eb1 :

In this local Cartesian system we formulate the different plane stress consti-
tutive relations in the plane ðb
e1; b
e 2 Þ.
There is an obvious contradiction since the above defined plane stress state, due
to the imposed kinematic constraint, is also a plane strain state; this can only be
possible in a very specific orthotropic material model. We overlook this contra-
diction as the price that we pay for degenerating the solid into a shell element.
The constitutive tensor can be described as,

C ¼ Cijkl b
ei b
ej b
ek b e pqrs o g o g o g o g
el ¼ C ð2:12Þ
p q r s
14 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

and from the above we can get the curvilinear components C e pqrs .
Then the element stiffness matrix can be calculated as [7],
Z
K¼ ~T C
B ~B~ dV: ð2:13Þ
V

~ collects the curvilinear components of the constitutive

The (5 9 5) matrix C
tensor and it is symmetric for hyperelastic materials models and elasto-plastic
material models when an associated plasticity model is used [24].

2.2.3 The Locking Problem

The locking problem has been very much analyzed in the literature [5, 6]; in the
present section we just present a couple of very simple examples to illustrate it.

2.2.3.1 Shear Locking

Using a 4-node element to model the cantilever under constant moment in Fig. 2.2
we notice that,
• the u2 displacement interpolation is linear along the coordinate x1 with a zero at
node 1,
• the h rotation interpolation is linear with a zero at node 1.

The shear deformation c ¼ dudx1  h has to be zero everywhere and the condition
2

is imposed more strongly when the thickness tends to zero [20, 25].
It is evident that considering the order of the interpolation functions and the
boundary condition, the only solution is u2 ¼ const ¼ 0.

Fig. 2.2 Cantilever under constant moment modeled with one 4-node element: shear locking
2.2 The Standard A-I-Z Quadrilateral Shell Element for Linear Analysis 15

2.2.3.2 Shear and Membrane Locking

When using a parabolic interpolation for the generalized displacements of a curved

cantilever we find the impossibility of imposing together the zero shear and zero
membrane elongation conditions.

2.2.4 Solving the Locking Problem

The first remedy that was proposed for the locking problem was the use of reduced
or selective integration schemes [6]; however, those schemes, even though they are
very simple and produce inexpensive elements, incorporate the difficulty of the
spurious rigid body modes and the oscillation in the stress predictions [5, 6].
The element MITC4 was developed by Dvorkin and Bathe as a solution for the
shear locking problem that does not incorporate numerical drawbacks.

2.3 The MITC4 Quadrilateral Shell Element for Linear

Analysis

This element incorporates the displacement/rotation interpolations used in the A-I-

Z element; the curvilinear covariant strain components ðee rr ; ee ss ; ee rs Þ are directly
calculated from the displacement/rotation interpolations using Eq. (2.9).
For the out-of-surface shear components we use the interpolations in Fig. 2.3,
which can be written as,
1 1
ee rt ¼ ð1 þ sÞ ee rt jDI e rt jDI
A þ ð1  sÞ e C ;
2 2 ð2:14Þ
1 1
ee st ¼ ð1 þ rÞ ee st jDI e st jDI
D þ ð1  rÞ e B :
2 2
In Eq. (2.14) we use the notation,
DI covariant strain component calculated from the displacement interpola-
ee ij P
tions at the sampling point P

The element, defined as we describe in this section, satisfies the Patch Test,
does not present spurious rigid body modes and does not lock. In the literature
there is abundant numerical evidence on the element robustness and accuracy.
16 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

Fig. 2.3 The MITC4 out-of surface shear interpolations

2.4 Nonlinear Analysis Using the MITC4 Element

2.4.1 Infinitesimal Strains Problems: Total Lagrangean

Formulation

There are a number of nonlinear structural and mechanical problems for which the
infinitesimal strains approach provides acceptable results. For these cases we
developed for the MITC4 element a Total Lagrangean Formulation (TLF) [5].
In an incremental analysis we know the s-configuration and we seek the
(s þ Ds)-configuration. Using the Principle of Virtual Work we can state for the
equilibrium at the configuration at (s þ Ds) [24],
2.4 Nonlinear Analysis Using the MITC4 Element 17

Z
sþDs
0S : dsþDs 0 e 0 dV ¼ sþDs dW ext : ð2:15Þ
0V

In the above equation 0 V is the volume of the reference configuration (time = 0);
and for the configuration at (s þ Ds)sþDs 0 S is the second Piola-Kirchhoff stress
tensor; sþDs 0 e is the Green-Lagrange strain tensor (both tensors referred to the
reference configuration) and sþDs dW ext is the virtual work of the external loads.
We can write for the incremental step,
sþDs
0S ¼ s0 S þ 0 S;
sþDs
ð2:16Þ
0e ¼s e þ 0 e:

Where the tensors 0 S; 0 e are increments referred to (time = 0).

We decompose the strain increment into a linear (0 e) and nonlinear part (0 g) in
terms of the incremental displacements and relate the incremental strain tensor
with the incremental stress tensor using and incremental fourth order constitutive
tensor 0 C.
After linearizing we get for the incremental step [24],
Z Z Z
0 s 0 sþDs ext s
d0 e : 0 C : 0 e dV þ 0S : d0 g dV ¼ dW  0S : d0 e 0 dV: ð2:17Þ
0V 0V oV

As it is well known, the ðs þ DsÞ-configuration is determined iterating on Eq.

(2.15) until equilibrium is fulfilled [5].
In the nonlinear problem, we need to handle finite rotations; hence, we can
write [23],
tþDs
V kn ¼ sþDs
0 R  0 V kn : ð2:18Þ

Any rotation matrix can be written as [23],

0 12
sin h 1 sin hk=2
k
s
0R ¼ I3 þ Hk þ @ A H2
k
ð2:19Þ
hk 2 hk=
2
where,
 1=2
hk ¼ a2k þ b2k
2 3
0 0 bk ð2:20Þ
6 7
½Hk  ¼ 4 0 0 ak 5:
bk ak 0
18 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

As in the infinitesimal rotations case, we have 5 d.o.f./node.

In the case of finite rotations the linearized equilibrium equations present extra
terms that were discussed in Refs. [20, 23].

2.4.2 Finite Strains

For the analysis of finite strain problems we use the following interpolation for the
reference configuration geometry [21, 22],

0 t0
xðr; s; tÞ ¼ hk ðr; sÞ 0 xk þ da ð2:21Þ
2
where,

0
hk ðr; sÞ 0 V n k
d¼ 
hk ðr; sÞ 0 V n j  : ð2:22Þ
k

It is important to remark that we considered in Eq. (2.21) elements with uniform

thickness.
For the displacement field we use,

s t  
uðr; s; tÞ ¼ hk ðr; sÞ s uk þ ðs ko þ s k1 tÞ s d  0 d a ð2:23Þ
2
where,

s hk ðr; sÞs V n jk
d¼ 
hk ðr; sÞs V n j  : ð2:24Þ
k

Equations (2.22) and (2.24) are used to avoid spurious director vector stretching
[26] that in this finite strains case, in which we do not neglect the stretching
through the thickness, may affect the results.
In Eq. (2.23) s ko is the constant thickness stretching and s k1 is the through-the-
thickness stretching gradient. In our formulation both stretching d.o.f. are con-
densed at the element level.
The strain interpolations are the same as the ones we used in the infinitesimal
strains case. However, for hyperelastic material models we interpolate the Green-
Lagrange covariant strain tensor components and for the elasto-plastic material we
interpolate the covariant components of the Hencky (logarithmic) strain tensor [24].
We use 3D constitutive relations; hence, the through-the-thickness stress
component is not neglected. In [22] we developed the 3D constitutive equations
for the elasto-plastic element based on Lee’s multiplicative decomposition of the
deformation gradient tensor and maximum energy dissipation [24, 27].
2.5 Modeling Considerations 19

2.5 Modeling Considerations

In this section we discuss several considerations that need to be taken into account
when modeling shell structures.

2.5.1 The Nodal Director Vectors

The nodal director vectors may be either defined by the analyst or calculated by the
finite element code.
When the analyst introduces the director vectors together with the mid-surface
nodes, she/he selects them at each node so as to be the best approximation to the
actual normal to the shell mid-surface.
When the finite element code calculates at a given node the director vectors, the
normals to the interpolated mid-surface are calculated for all the elements sharing
the node; hence, the code defines at the node as many director vectors as the
number of elements sharing the node (the interpolated mid-surfaces only have C0
continuity). All the element normals sharing a node rotate together.

2.5.2 Number of d.o.f. per Node

As discussed above, in a shell model there may be nodes at which only one
director vector is defined and nodes at which multiple director vectors are defined.
For the case of nodes with only one director vector, the analyst has to consider 5
d.o.f. (3 displacements and 2 rotations around the local axes sV1 and sV2).
For the case of nodes with multiple director vectors, the analyst has to consider
6 d.o.f. (3 displacements and 3 rotations around the global Cartesian axes).
The case of a node with multiple but very close director vectors has to be
treated as a case with 5 d.o.f. collapsing the very close director vectors.
It is important to be aware of the fact that when using 6 d.o.f. the rotational
boundary conditions need to be defined along the global Cartesian axes and when
using 5 d.o.f. the rotational boundary conditions need to be defined along the local
axes. In geometrically nonlinear analyses these local axes change for each
incremental step.
When modeling a stiffened shell using shell elements and iso-beam (Timo-
shenko beam) elements [5],3 at the nodes shared by a shell and a beam element 6
d.o.f. have to be used.

3
Please notice that shell elements are not compatible with Bernoulli beam elements.
20 2 Shell Element Formulations for General Nonlinear Analysis. Modeling Techniques

References

1. Ahmad S, Irons B, Zienkiewicz O (1970) Analysis of thick and thin shell structures by curved
finite elements. Int J Numer Methods Eng 2:419–451
2. Ramm E (1977) A plate/shell element for large deflections and rotations. In: Bathe et al (ed)
Formulations and computational algorithms in finite element analysis. MIT Press, Cambridge
3. Kråkeland B (1978) Nonlinear analysis of shells using degenerate isoparametric elements. In:
Bergan et al (ed), Finite elements in nonlinear mechanics. Tapir Publishers, Norwegian
Institute of Technology, Trondheim
4. Bathe K-J, Bolourchi S (1980) A geometric and material nonlinear plate and shell element.
Comput Struct 11:23–48
5. Bathe K-J (1996) Finite element procedures. Prentice Hall, Saddle River
6. Zienkiewicz O, Taylor R (2000) The finite element method. Butterworth-Heinemann, Oxford
7. Dvorkin EN, Bathe K-J (1984) A continuum mechanics based four-node shell element for
general nonlinear analysis. Eng Comput 1:77–88
8. Bathe K-J, Dvorkin EN (1985) A four-node plate bending element based on Mindlin/Reissner
plate theory and a mixed interpolation. Int J Numer Methods Eng 21:367–383
9. Bathe K-J, Dvorkin EN (1986) A formulation of general shell elements—the use of mixed
interpolation of tensorial components. Int J Numer Methods Eng 22:697–722
10. Rodal J, Witmer E (1979) Finite-strain large-deflection elastic-viscoplastic finite-element
transient analysis of structure. NASA CR 159874
11. Hughes T, Carnoy E (1983) Nonlinear finite element shell formulation accounting for large
membrane strains. Comput Methods Appl Mech Eng 39:69–82
12. Simo J, Fox D (1989) On a stress resultant geometrically exact shell model. Part I:
Formulation and optimal parametrization. Comput Methods Appl Mech Eng 72:267–304
13. Simo J, Fox D, Rifai M (1989) On a stress resultant geometrically exact shell model. Part II:
The linear theory; computational aspects. Comput Methods Appl Mech Eng 72:53–92
14. Simo J, Fox D, Rifai M (1990) On a stress resultant geometrically exact shell model. Part III:
Computational aspects of the nonlinear theory. Comput Methods Appl Mechs Eng 79:21–70
15. Simo J, Fox D, Rifai M (1992) On a stress resultant geometrically exact shell model. Part IV:
Variable thickness shells with through-the-thickness stretching. Comput Methods Appl Mech
Eng 81:91–126
16. Simo J, Kennedy J (1992) On a stress resultant geometrically exact shell model. Part V:
Nonlinear plasticity formulation and integration algorithms. Comput Methods Appl Mech
Eng 96:133–171
17. Büchter M, Ramm E, Roehl D (1994) Three-dimensional extension of non-linear shell
formulation based on the enhanced assumed strain concept. Int J Numer Methods Eng
37:2551–2568
18. Bischoff M, Ramm E (1997) Shear deformable shell elements for large strains and rotations.
Int J Numer Methods Eng 40:4427–4449
19. Dvorkin EN, Pantuso D, Repetto E (1995) A formulation of the MITC4 shell element for
finite strain elasto-plastic analysis. Comput Methods Appl Mech Eng 125:17–40
20. Dvorkin EN (1995) Nonlinear analysis of shells using the MITC formulation. Arch Comput
Methods En 2:1–50
21. Toscano RG, Dvorkin EN (2007) A shell element for finite strain analyses. Hyperelastic
material models. Eng Comput 24:514–535
22. Toscano RG, Dvorkin EN (2008) A new shell element for elasto-plastic finite strain analyzes.
Application to the collapse and post-collapse analysis of marine pipelines. In: Abel J, Cooke J
(eds), Proceedings 6th international conference on computation of shell & spatial structures,
Spanning Nano to Mega. Ithaca
23. Dvorkin EN, Oñate E, Oliver X (1988) On a nonlinear formulation for curved Timoshenko
beam elements considering large displacement/rotation increments. Int J Numer Methods
Eng 26:1597–1613
References 21

24. Dvorkin EN, Goldschmit MB (2005) Nonlinear continua. Springer, Berlin

25. Dvorkin EN (1992) On nonlinear analysis of shells using finite elements based on mixed
interpolation of tensorial components. In: Rammerstorfer F (ed) Nonlinear analysis of shells
by finite elements. Springer, New York
26. Gebhardt H, Schweizerhof K (1993) Interpolation of curved shell geometries by low order
finite elements—errors and modifications. Int J Numer Methods Eng 36:287–302
27. Simo J, Hughes T (1998) Computational inelasticity. Springer, New York
http://www.springer.com/978-3-642-37360-2