Facult des Sciences Appliques

Anne Acadmique 2008-2009

Development of solid-shell elements
for large deformation simulation
and springback prediction












Travail prsent par
Nhu Huynh NGUYEN
Ingnieur mcanicien

pour lobtention du grade de
Docteur en Sciences de lIngnieur



Septembre 2009



To my mother for her sacrifices to sons
To my wife for her understanding and encouragement
To my sincere friends for their precious helps


Acknowledgements


All acknowledgements to the help I have received while completing this
thesis are worthless. First of all I would like to thank the financial support from
322 Scholarship Foundation of Vietnamese Government; without that financial
support for 1 year in Vietnam and 2.5 years in Belgium I would not have a chance
of doing this thesis.

I would like to express my deep emotion to my family in Vietnam and to my
Vietnamese friends in Belgium. Thanks to their cares I always felt the familial air
nearby. Special thanks are devoted to Dr. Quoc Viet BUI.

I deeply thank Prof. M. HOGGE and Prof. J.P. PONTHOT for their
precious supports and their patience. I also thank all the members of the Research
Team in MC&T Department University of Lige for their direct or indirect
supports.

Finally, I would like to thank the jury members of my thesis for their
participation in reading the thesis.








i
NOTATIONS AND SYMBOLS


Operator Meaning
D
Differential operator
X
Scalar
x

Vector with components
i
x
[ ] X Matrix
{ } X Column vector

X
Time derivative of X ( dt dX / )
T
X
Transpose of X
( ) x Sgn + or sign of scalar x
x

x x = if 0 > x ; 0 = x if 0 < x
Y X : Double contracted product of X with Y
I
X

First invariant of X : ( ) X Tr X
I
=
II
X
Second invariant of X :
2 2
1
[ ( ) ( )]
2
II
X Tr Tr = X X
III
X

Third invariant of X : det( )
III
X = X
X'
Deviator of X : I - X X'
I
X
3
1
=

Kronecker delta 1 =
ij
if i=j, 0 =
ij
if ij
X
~


A quantity with tilt bar is assumed in natural (isoparametric)
coordinate system or physical coordinate system


Scalar Meaning
u
A
Boundary along which displacements are specified
A

Boundary along which surface tractions are specified
L Contour line
Functional
,
Lam coefficients

Mass density
Poisson ratio
G Shear modulus
ij

Shear strain
A Surface of a body
t Time
V Volume of a body
W Work
) ( E S
W
Stored energy function of a hyperelastic material



ii
1st order
Tensor
Meaning
i
e
Basic unit vector
b Body force per unit mass
u Current displacement vector
ext
F External force vector
int
F Internal force vector
( )
3 2 1
, , X X X X
Material coordinate vector
( )
3 2 1
, , Natural (isoparametric) coordinate vector
h
U
Nodal displacement vector, [ ]
1 1 1
, , ,... , ,
h
n n n
u v w u v w = U
n Outward normal vector to the surface A
( )
3 2 1
, , x x x x
Spatial coordinate vector
* t
Specific tractions along A

* u
Specific displacements along
u
A
t Traction vector, stress vector


2nd and higher
order Tensor
Meaning
Cauchys strain tensor
Cauchys stress tensor
F
Deformation gradient tensor
U Grad
Material displacement gradient tensor, U U
0
= Grad
u grad Spatial displacement gradient tensor, u u = grad
P First Piola-Kirchhoff stress tensor
E Green-Lagrange tensor
E
Green-Lagrange vector (Voigt notation):

T
E E E E E E ] 2 2 2 [
31 23 12 33 22 11
= E
J Jacobian matrix
b
Left Cauchy-Green tensor (Finger deformation tensor),
T
FF b =
v Left stretch deformation tensor,
-1
= v FR
C
4

Constitutive moduli tensor
C
4

Constitutive moduli matrix, dimension (6x6)
C
Right Cauchy-Green tensor, F F C
T
=
U Right stretch tensor,
-1
= U R F
R

Rotation tensor,
-1
FU R =
S Second Piola-Kirchhoff stress (PK2) tensor
S Second Piola-Kirchhoff stress vector (Voigt notation):

T
S S S S S S ] [
31 23 12 33 22 11
= S
K
Tangent stiffness matrix
I
Unity second order tensor

iii
FIGURE LIST

1.1.1

Connection of solid-shell element (white color) with standard solid
element (grey color).................................................................................. 1
2.1.1 Hexahedral solid element ......................................................................... 8
2.1.2 Pure bending of a rectangle....................................................................... 12
2.1.3 Errors in stresses in pure bending.......................................................... 14
2.2.1 Degeneration of a 3D solid element into a degenerated shell element .... 20
2.5.1 ANS method illustration, special case: X-Y aligns
1
-
2
....................... 39
2.5.2 Mid-surface of element in isoparametric coordinates ................................ 39
2.6.1 Membrane patch test ....................................................................................... 45
2.6.2

Eigenvalue analyses of element stiffness matrices, in-plane bending
mode................................................................................................................... 47
2.6.3 Circular cantilever ........................................................................................... 47
2.6.4 Curved cantilever beam: displacements ....................................................... 48
2.6.5 Scordelis-Lo roof ............................................................................................. 49
2.6.6 Scordelis-Lo roof: convergence of finite element solution ........................ 49
2.6.7 Regular block ................................................................................................... 50
A1 Cubic patch test ................................................................................................ 53
A2 Stress in patch elements .................................................................................. 54
3.1.1 Four node element ........................................................................................... 59
3.1.2 Curved beam element ...................................................................................... 60
3.2.1

Initial configuration
0
, current configuration
t
and isoparametric
configuration ................................................................................................. 63
3.2.2 Configuration of low-order solid-shell element ........................................... 64
3.3.1 Degeneration from 3D to 2D of ANS-solid element ................................... 65
3.3.2 Solid element, special case (X,Y,Z) (
1
,
2
,
3
) ........................................ 66
3.3.3 Pure bending in 0X direction ......................................................................... 67
3.3.4 Sampling points for ANSn element .............................................................. 68
3.3.5 Sampling points for 2D.ANSn element ........................................................ 69
3.3.6 Distorted element - sampling points for strains ........................................... 70
3.4.1 Solid-shell algorithm ................................................................................... 76
3.5.1 Static condensation algorithm [SIM93] ........................................................ 78
3.6.1 Regular and distorted cubes ........................................................................... 79
3.6.2 Two deformation modes ................................................................................. 80
3.6.3 Square plate ...................................................................................................... 82
3.6.4 Two bending modes ........................................................................................ 82
3.6.5 Two warping modes of EAS3v6s & Q1 ....................................................... 84
3.6.6 Square plate geometry ..................................................................................... 84
3.6.7 Displacement versus applied force for fully clamped plate ....................... 85
3.6.8 Displacement versus applied force for simply supported plate ................. 85
4.1.1 Cantilever under pure bending ....................................................................... 88
4.1.2 Twisted beam ................................................................................................... 88
4.1.3 Twisted beam, case of load along 0Z ............................................................ 89
4.1.4 Twisted beam, case of load along 0Y ........................................................... 89
4.1.5 Square plate under uniformed pressure ......................................................... 90
4.1.6 A quarter of the plate - distorted mesh .......................................................... 94
4.1.7 Pinched cylinder with two rigid end diaphragms ........................................ 95
iv
4.1.8 Convergence investigation for the pinched cylinder ................................... 96
4.1.9 Morley spherical shell ..................................................................................... 97
4.1.10 Convergence of finite element solution ........................................................ 97
4.1.11 Thick-walled cylinder ................................................................................. 98
4.2.1 Cantilever beam ............................................................................................... 99
4.2.2 Displacement of test point .............................................................................. 100
4.2.3 Load-deflection curves for displacements at points A(u) and B(v) .......... 100
4.2.4 Deformed hemisphere at F = 5 (without any magnification) ..................... 101
4.2.5 Slit annular plate - initial configuration ........................................................ 102
4.2.6

The deformed configuration at maximum load (without any
magnification) .................................................................................................. 102
4.2.7 Load-deflection curves for displacements at points A and B .................... 103
5.1.1 Schematic diagram of multiplicative decomposition .................................. 105
5.1.2 Finite strain theory ........................................................................................... 107
5.2.1 Material stress update algorithm .................................................................... 112
5.3.1 Cantilever at large elasto-plastic deformation .............................................. 113
5.3.2 EAS9 (METAFOR) results with various elements along thickness ........... 114
5.3.3 EAS21 (METAFOR) results with various elements along thickness ......... 115
5.3.4 ANS results with various elements along thickness .................................... 115
5.3.5 SS7n results with various elements along thickness ................................... 115
5.3.6 ANS element -
xz
(mesh 40101) ............................................................. 117
5.3.7 ANSn element -
xz
(mesh 40101) .......................................................... 117
5.3.8 SS7 element -
xz
stress (mesh 40101) ..................................................... 118
5.3.9 SS7n element -
xz
stress (mesh 40101) ................................................... 118
5.3.10 EAS9 element -
xz
stress (METAFOR, mesh 40101) .......................... 119
5.3.11 Pinched cylinder .............................................................................................. 120
5.3.12 EAS12 element (METAFOR) ........................................................................ 120
5.3.13 Pinched cylinder: Force-Displacement ......................................................... 121
5.3.14 SS7 element -
xz
(mesh 16161) ............................................................... 122
5.3.15 SS7n element -
xz
(mesh 16161) ............................................................. 122
5.3.16

EAS12 element -
xz
(EAS12,METAFOR, 16161)
(max. value has been imposed to 26.0) ........................................................ 123
5.3.17

EAS12 element -
yz
(EAS12,METAFOR, 16161)
(min. value has been imposed to -51.7) ........................................................ 123
5.3.18 SS7 element -
yz
(mesh 16161) .............................................................. 124
5.3.19 SS7n element -
yz
(mesh 16161) ............................................................ 124
5.3.20a Reference -
xz
(EAS12,METAFOR, 40401)....................................... 125
5.3.20b

Reference -
xz
(EAS12,METAFOR, 40401) (maximum value has
been imposed to 43.3) ..................................................................................... 125
5.3.21a Reference -
yz
(EAS12,METAFOR, 40401) ........................................ 126
5.3.21b

Reference -
yz
(EAS12,METAFOR, 40401) (minimal value has
been imposed to -49.6) ................................................................................... 126
5.3.22 SS7 element - von Mises stress (mesh 16161) ....................................... 127
5.3.23 SS7n element - von Mises stress (mesh 16161) ..................................... 127
5.3.24 Pinched cylinder: solid-shell elements .......................................................... 129
5.3.25 Pinched cylinder: solid-shell and EAS21 ..................................................... 128
5.3.26 Initial position (left) and final position (right) for stamping ...................... 130
5.3.27 Definition of angles (point E and point F are the farthest contact points 131
v
from the centerline) ............................................................................................
5.3.28 Discretization of the model : definition of the 3 zones ............................... 132
5.3.29

Punch force vs. punch displacement EAS element (quasi-static; Table
5.3.3/cases 1 to 5; penalty =1.010
3
) ............................................................... 133
5.3.30

Punch force vs. punch displacement EAS element (Table 5.3.3/cases 6
to 10) ................................................................................................................. 133
5.3.31

Reference solution of SRI element (Table 5.3.4 /case 7 equivalent plastic
strain, simulation with METAFOR) ................................................................. 134
5.3.32

Punch force vs. punch displacement SRI element (Penalty=1.010
3
;
Table5.3.4/cases 1 to 4) ..................................................................................... 135
5.3.33

Punch force vs. punch displacement SRI element (Penalty=1.010
3
;
Table5.3.4/cases 5 to 8) .................................................................................... 136
5.3.34

Chosen solution of 2D.ANSn element (Table 5.3.5 /case 6 von Mises
stress, simulation with FEAP) ........................................................................... 137
5.3.35

Punch force vs. punch displacement ANSn element (Table5.3.5/cases 1
to 4) .................................................................................................................... 137
5.3.36

Punch force vs. punch displacement ANSn element (Table5.3.5/cases 5
to 7) ................................................................................................................... 138
5.3.37

Reference solution of 2D.SS4n element (Table 5.3.6/case 6 von Mises
stress, simulation with FEAP) ........................................................................... 139
5.3.38

Punch force vs. punch displacement 2D.SS4n element
(Table5.3.6/cases 1 to 3) .................................................................................... 139
5.3.39

Punch force vs. punch displacement 2D.SS4n element
(Table5.3.6/cases 4 and 5) ................................................................................. 140
5.3.40

Punch force vs. punch displacement [NUM02] (BS = Benchmark
simulation result) .............................................................................................. 140
5.3.41

Punch force vs. punch displacement [NUM02] (BE = Benchmark
Experiment result) .............................................................................................. 141
5.3.42

Definition of the angle between 2 contact points which are the farthest
from the centreline ........................................................................................... 142
5.3.43

Angle between 2 contact points which are the farthest from the
centerline [NUM02] ........................................................................................ 142
5.3.44

Open angle between the lines AB and CD before spring back (Table
5.3.3/Case 6 - equivalent plastic strain) ............................................................ 144
5.3.45

Open angle between the lines AB and CD after spring back (Table
5.3.6/Case 3 - equivalent stress - FEAP) ........................................................... 144
5.3.46 Angle before and after springback at the final punch stroke of 28.5 mm. 145













vi
TABLE LIST

1.1.1 Features of low-order solid elements ............................................................. 4
2.1.1 Deformation modes of bilinear element 2D.Q1 ........................................... 12
2.4.1 Index transformation ........................................................................................ 27
2.4.2 EAS elements .................................................................................................... 32
2.4.3 Equivalent bilinear EAS - HR elements ........................................................ 36
2.6.1 Interior nodal coordinates ................................................................................ 44
2.6.2 Displacements of the interior nodes ............................................................... 46
2.6.3 Displacements at interior nodes of elements ................................................ 46
2.6.4 Vertical displacement at the blocks center .................................................. 50
A1 Location of inner nodes ................................................................................... 53
A2 Results of EAS9 ............................................................................................... 54
3.0.1 Dominant features of degenerated shell and solid-shell .............................. 57
3.6.1 Eigenvalues of regular cube ............................................................................ 80
3.6.2 Distorted cube - location of nodes ................................................................ 81
3.6.3 Eigenvalues of distorted cube ......................................................................... 81
3.6.4 Eigenvalues of the square plate ...................................................................... 83
3.6.5 Eigenvalues of the square plate Summary ................................................. 83
4.0.1 Summary of employed elements .................................................................... 87
4.1.1 Numerical results ............................................................................................. 88
4.1.2 Non-dimensional deflection ........................................................................ 91
4.1.3 Normalization of dimensionless deflection - Clamped plate ( = 0.3) .. 91
4.1.4

Normalization of dimensionless deflection - Simply supported plate
( = 0.3) ............................................................................................................. 92
4.1.5 Normalization of dimensionless deflection - Clamped plate ( = 0.0) .. 92
4.1.6

Normalization of dimensionless deflection - Simply supported plate
( = 0.0) .............................................................................................................
92
4.1.7 Center deflection .............................................................................................. 92
4.1.8 Regular mesh - Normalization of displacement ........................................... 94
4.1.9 Irregular mesh - Normalization of displacement .......................................... 94
4.1.10 Morley spherical shell - Normalized displacements at test point ............... 96
4.1.11 Normalized radial displacement at R
i
............................................................ 99
4.2.1 Displacements due to pinched force F = 5 .................................................... 101
5.3.1 Convergence of
xz
(MPa) .............................................................................. 116
5.3.2 Transverse shear stresses ................................................................................. 121
5.3.3 Calculation with 2D.EAS7 element ............................................................... 132
5.3.4 Calculation with 2D.SRI element .................................................................. 135
5.3.5 Calculation with 2D.ANSn element .............................................................. 136
5.3.6 Calculation with solid-shell element ............................................................. 138
5.3.7 Springback angles - calculation with 2D.EAS7 element ............................ 141
5.3.8 Springback angles - calculation with 2D.SRI element ................................ 143
5.3.9 Springback angles - calculation with 2D.ANSn element ............................ 143
5.3.10 Springback angles - calculation with 2D.SS4n element .............................. 143


1
Chapter 1. INTRODUCTION


INTRODUCTION

Since the sixties of the 20
th
century, the development of numerical methods has
been the base for developing various advanced engineering simulation tools. Still
nowadays, at the beginning of the 21
st
century, applications of numerical methods in
simulation and prediction of industrial problems and/or technological processes become
more and more important The most popular numerical methods is the finite element
method with applications for simulation of biomechanic problems [HOL96],
elastoplasticity problems [SIM88a], thermalmechanic problems [HOG76] and contact
problems [PON99], etc.
Thin shell structures, whose numerical analysis is the target of this thesis, appear
in many products, such as the outer-body of a car, the fuselage and wings of an airplane,
etc. Modeling these parts with standard solid elements would require a huge number of
elements and leads to prohibitive computational costs. For instance, to prevent locking
effects, i.e. artificial stiffness in the model, modeling a beam with hexahedral solid
elements requires a minimum of about 3 - 5 elements through the thickness. In such
cases, a low-order shell element can replace 3 - 5 or more solid elements, which improves
computational efficiency immensely. Furthermore, modeling thin structures with
standard solid elements often leads to elements with high aspect ratios, which degrades
the conditioning of the equations and the accuracy of the solution. However, for certain
problems in structural analysis displacement degrees of freedom (DOF) at the nodes of
the element are more advantageous for analysis than displacement and rotational DOFs.
For example, consider complex structures which consist of both thin and thick walls. For
the sake of effectiveness, shell elements should be used for thin-walled parts and solid
elements should be used for thick-walled parts. If both solid and shell elements have the
same DOFs (e.g. only displacements at nodes), the analysis process exhibits one type of
DOF only, no requirement on transition elements exhibiting displacement and rotational
DOF, e.g. Figure 1.0.1.









Figure 1.0.1: Connection of solid-shell element (white
color) with standard solid element (grey color)
Solid-shell
element
Standard solid
element
2
The deformation processes also with contact and friction of shell elements, such as
happening in metal forming, are easier to consider if shell elements have those
configuration displacement DOF. Motivated by these arguments, the thesis concentrates
on developing an element that has simple kinematics (only displacement DOFs at nodes)
as solid elements but is as effective in computation as shell elements. A class of those
finite elements called solid-shell has been recently investigated by many researchers
([HAU98], [HAU00], [QUO03a,b], [TAN05], [JET08]) because that element is not only
capable of modeling complicated structures but it can also be used to simulate metal
forming problems. Literature shows indeed that solid-shell element is the most suitable
choice for the above mentioned tasks.


1.1 OBJECTIVES OF THE THESIS
The use of low-order elements in finite element computations remains a popular
feature in solid mechanics because of the following reasons. First, they require a simpler
manipulation for meshing, especially, for a distorted configuration. Second, these
elements facilitate more convenient manipulations in the adaptive h-type of mesh
refinement. Finally, using low-order elements will remarkably save computation time,
especially, for simulation with large number of DOFs such as in nonlinear problems. In
this thesis, high-order elements are therefore not considered. They would, in addition,
exhibit more difficulties to deal with at contact surface interfaces. From now on, for the
sake of briefness, lets call the low-order standard solid element in three-dimension (3D)
is the eight-node hexahedral element and the low-order standard solid element in two-
dimension (2D) is the four-node quadrilateral element.
Although the structure of the low-order standard displacement elements is
straightforward, they should not be used directly in the following situations:
The elements tend to be too stiff in bending, e.g. for slender beams or thin plates
under bending.
The elements are too stiff in nearly incompressible or incompressible behavior.
In other words, four-node quadrilateral elements and eight-node hexahedral
element in two and three dimensions, respectively, have a major drawback since they
lead to locking in the incompressible limit. It means they do not possess the property of
being uniformly convergent. In addition, even in compressible problems the use of these
standard elements leads to poor accuracy particularly in bending-dominated problems,
when coarse meshes are used.
The linear and nonlinear finite element analysis of plates and shells has attracted
much attention in recent decades. It is necessary to capture the bending-stretching
coupling of thin shell behavior. Hence, one of the motivations for designing new
elements is their potential ability to capture the membrane-bending coupling correctly. So
far, there are two ways in which this could be done. One is to use elements based on
specific shell theories (e.g. the Vlasov [VLA49], Flugge [FLU73] theories, etc.). There
are considerable controversies regarding the relative merits and drawbacks of these
theories. Each theory has been obtained by carrying out approximations to different
degrees when the 3D field equations are reduced to the particular class of shell equations.
3
The second approach is called degenerated shell approach 3D solid elements can be
reduced (degenerated) into shell elements having only mid-surface nodal variables - these
are no longer dependent on the various forms of shell theories proposed and should be
simple to use. They are in fact equivalent to a Mindlin type curved shell element
(equivalent to a shear deformable theory), see [PRA01], [STO95]. With the standard
procedure, a plate or shell theory is used as a basis for the finite element formulation. It
begins with the field equations of the 3D theory and makes various assumptions, which
lead to the plate or shell theory. Meanwhile, in the reduction from the three to two
dimensions, an analytical integration over the thickness was included. The mid-surface
geometry (in the case of shells) and the field variables are approximated using discretized
nodal values and suitable interpolation functions. Integration of various element stiffness
and force terms is carried out over the reference surface. Examples of such an approach
include the simple facet element and many elements derived from the classical thin plate
theory: the Mindlin-Reissner plate theory, shallow shell theory or even high-order shell
theories. Currently, we can introduce the solid-shell theory as the third approach for
capturing the bending-stretching coupling of thin shell behavior. The low-order solid-
shell element has two nodes along vertical sides. Naturally, the low-order solid-shell
element obeys the straight normal assumption of Mindlin-Reissner theories. Without any
assumptions because of possessing solid configuration, integration of the solid-shell
element stiffness is carried out over the elements physical volume.
The solid-shell elements are combinations of solid elements with shell elements.
The elements are generally used for nonlinear problems (finite strain, contact, etc.) so
they have to satisfy some requirements, e.g. free from all locking types, simple
kinematics, etc. The following features of the solid elements and shell elements are found
in the solid-shell elements:
Features from the solid elements:
- Same DOFs as solid;
- Integrating directly 3D material model (versus plane stress assumption);
- In contrast to the degenerated shell concept the complete 3D strain
tensor and stress tensor are used (strains and stresses in thickness
direction are also included).
Features from the shell elements:
- Use of a shell element method to remove transverse shear locking. The
solid-shell elements, hence, are applicable to thin-walled structures.
- The normal to the mid-surface remains straight.
Through investigating the literatures, there are two assumed strain methods that
have been exploited so far to develop solid-shell elements. They are the Assumed Natural
Strain (ANS) of Dvorkin and Bathe [DVO84] and the Enhanced Assumed Strain (EAS)
proposed by Simo and Rifai [SIM90]. The strain field of the EAS element is additionally
modified to be a complete polynomial field. The strain field of the ANS element is
replaced by an incompatible strain field that satisfies the pure bending requirement.
Otherwise, there is a sound variational method: Mixed Enhanced Strain (MES) method
proposed by Kasper and Taylor [KAS00b] where stresses are independent from the strain
field. The MES method requires more variables (apart from the displacement field and
strain field as required by the EAS and ANS methods, the MES further considers the
stress field as variables) than the EAS and the ANS, thus is not an attractive approach.
4
Objective of the thesis is to develop a finite element that is effective for simulation
of thin-walled behavior in metal forming processes. It means the element gives precise
results while proposing a low computational cost. Application of that element is mainly
exploited in springback simulation. Springback relates to the change in shape between the
fully loaded and subsequent unloaded configurations. The springback effect is
encountered during a stamping operation. This can result in the formed component being
out of tolerance and thus creates major problems in assembly or installation. Accurate
description of the contact is one of the main factors which renders metal forming
simulation predictive or not. Lets consider the solid-shell element. Contact algorithms
are more easily applied for solid like elements, thanks to the geometrical description of
the lower and upper surfaces than for shell elements where nodes lie in the mid-plane.
However, due to limitation of a thesis, only performance of the solid-shell for springback
simulation is mainly exploited in this thesis.
Table 1.1.1: Features of low-order solid elements
Element Features Application domain
Standard The standard element exhibits deficiencies as:
Volumetric locking;
Shear locking;
Poisson locking.
Being applicable to all
of problems but require
very fine mesh to
converge
EAS The standard element, which is enhanced by
EAS parameters, is:
locking free;
high in computational cost.
Metal forming
Incompressible
material, etc
ANS The standard element, which adopts the ANS
method, is:
only shear locking free;
cheap in computational cost.
Thin-walled structures
Collapse of shells, etc
Solid-shell The standard element, which adopts both the
ANS and the EAS methods, is:
locking free;
cheaper in computational cost than the EAS
element.
Thin-walled structures
Metal forming
Incompressible
material, etc.


1.2 APPLICATION DOMAIN

Nowadays, aims of the new 3D solid elements are:
1. No locking for incompressible materials;
2. Good bending behavior;
3. No locking in the limit of very thin elements;
5
4. Distortion insensitivity;
5. Good coarse mesh accuracy;
6. Simple implementation of nonlinear constitutive laws.
The first two aims are essential requirements of solid structured analyses. The
third aim is usually required for structural elements, such as plate, shell and beam
elements. The fourth aim is important because in discretizing an arbitrary geometry the
existence of distorted elements is inevitable. In addition, elements can get highly
distorted during nonlinear simulations including finite deformations. The fifth aim results
from the fact that many engineering problems have to be modeled as 3D problems. Due
to computer limitations, quite coarse meshes have to be used often to solve these
problems. Thus, an element which provides good coarse mesh accuracy is valuable in
these situations. The sixth aim is associated with the fact that more and more nonlinear
computations involving nonlinear constitutive models have to be performed to design
engineering structures. Thus, an element formulation which allows a straightforward
implementation of such constitutive equations is desirable.
The third aim becomes increasingly important since it enables the solid elements
to simulate shell problems. This makes the simulation work become more effective,
especially, for simulation of complicated structures. This spares the need for introducing
finite rotations as variables in thin shell elements, results in simpler contact detection on
upper and lower surfaces and provides the possibility to apply 3D constitutive equations
straight away.
The EAS elements can satisfy all requirements, except the third one. The solid-
shell elements are the ones who could satisfy all of that 6 six requirements. Due to their
dominant performance, the solid-shell elements are applicable to various applications in
structural analyses. The solid-shell elements could be applied for both solid and structural
engineering problems and in both linear and nonlinear applications (finite strain, contact,
etc.). Concretely, the solid-shell elements are suitable choice for metal forming
simulations, civil-engineering structures, impact/crash analysis, etc.


SUMMARY OF THE THESIS

Nowadays, in computational mechanics, there is a trend to treat plates and shells
as a 3D continuum, using solid finite elements or 3D-like plate and shell elements, taking
into account thickness changes throughout deformation and using 3D material laws. The
EAS elements are very suitable to that trend; they are applicable to almost any
engineering problems. However, the EAS elements are sensitive to distorted mesh (e.g.
bending patch test), and they exhibit poor performances in bending for very thin-walled
structures. Most important is the fact that the EAS elements are very time consuming for
calculation. In contrary, Reduced Integration (RI) elements are computational time
saving elements. They could be free from volumetric locking and shear locking.
However, in some situations they are not stable (due to hourglass modes), see [HAN98].
The solid-shell elements are attractive ones. Currently, they attract much
consideration of researchers. They possess performance of the EAS elements while they
are insensitive to distorted mesh. Furthermore, they are stable and time saving elements.
6
Obviously, because of having the solids configuration, the solid-shell elements are
suitable for handling contact in metal forming simulation, particularly for simulation of
sheet metal products whose ratio between length and thickness is large. They have only
translational DOFs of solid elements so they can be easily combined with the standard
solid elements in problems dealing with complex structures meanwhile they can also
work as shell elements. For transverse shear locking removal, the elements employ the
ANS techniques because the ANS method is cheaper (higher performance but cheaper
computational cost) than the EAS method in removing transverse shear locking. To get
rid of volumetric locking and membrane locking, the elements adopt techniques of the
EAS method. Due to the use of ANS techniques, that were originally applied for plate
and shell elements, the solid-shell elements are able to simulate thin and moderately
thick-walled structures.
The thesis includes six chapters and is structured as follows. The first chapter
presents the objects for the research. The second chapter introduces background methods
which will be incorporated in the solid-shell elements. The third chapter develops an
alternative ANS technique and applies it to the solid-shell elements. As a result, in that
chapter a new solid-shell element based on the alternative ANS technique is proposed.
Elastic applications of the just developed solid-shell element are illustrated in Chapter 4.
In Chapter 5, plasticity theory and numerical problems in plasticity deformation are
presented. The thesis specially concentrates on treating a current industrial problem:
spring back prediction. Results in Chapter 4 and Chapter 5 demonstrate the capabilities of
the proposed solid-shell element. Chapter 6 withdraws conclusions and then makes some
remarkable future developments.
























7
Chapter 2. BACKGROUND ON THE
DEVELOPMENT OF SOLID-SHELL
ELEMENTS


INTRODUCTION
The solid-shell is a solid element that has incorporated shell features, e.g. ANS
technique for shear locking and curvature thickness locking removals, and EAS or RI
techniques for other locking effects. In order to better understand the solid-shell concept,
lets start considering some features of the low-order standard solid element and the
obstacles that the solid element is facing. The difficulties when using the well-known
degenerated shell elements are also investigated. Later in the chapter, all the methods that
concern the solid-shell elements: the EAS (formulated in Green-Lagrange strain) [KLI97]
and the classical ANS methods [DVO84] (applied for finite deformation solid elements)
are introduced. These methods have been implemented in a MATLAB code. In this
chapter separated performances of the ANS and EAS elements are presented. Details for
their combination and co-operated performances are presented later in Chapter 3, where
we also present how the solid-shell element remedies all the obstacles that the low-order
standard solid elements have to overcome.

2.1 THREE DIMENSIONAL STANDARD ELEMENT
In this section we investigate the low-order standard solid element. Lets consider
the following trilinear displacement field, which is conventionally employed for the
eight-node standard element, Figure 2.1.1. To facilitate understanding, the analysis is
restricted to a rectangular prismatic geometry element so that the physical system
(X,Y,Z) and the isoparametric (natural) system (
1
,
2
,
3
) can be used interchangeably.
The displacement fields u, v and w are linearly interpolated with the help of coefficients
a
i
, b
i
and c
i
(i = 0, ..., 7) stemming from a trilinear assumption:

0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7
v
u a a X a Y a Z a XY a YZ a XZ a XYZ
b b X b Y b Z b XY b YZ b XZ b XYZ
w c c X c Y c Z c XY c YZ c XZ c XYZ
= + + + + + + +
= + + + + + + +
= + + + + + + +

(2.1.1a)
or under the form:

1 2 1 2 1 2
* * * * * *
1 2 1 2 1 2
** ** ** ** ** **
1 2 1 2 1 2
( )( )( )
v ( )( )( )
( )( )( )
u X Y Z
X Y Z
w X Y Z



= + + +
= + + +
= + + +

(2.1.1b)
where
**
1 2 2
, ,... are constants.
The Green-Lagrange strain components used in large deformation theories are
given by:
8

, , , , , , ,
, , , , , , ,
, , , , , , ,
( )
, , , ,
, ,
, ,
( )
0.5( v v )
v 0.5( v v )
0.5( v v )
2 (
2
2
v
X X X X X X X
XX
Y Y Y Y Y Y Y
YY
Z Z Z Z Z Z Z
ZZ com
u
XY Y X X Y
ZX
Z X
YZ
Z Y
com
u
u u u w w
E
u u w w
E
w u u w w
E
E u v u u
E
u w
E
w
+ +



+ +



+ +

= = +
` `
+ +


+

+ )
)

, , , ,
, , , , , ,
, , , , , ,
v v )
( v v )
( v v )
X Y X Y
X Z X Z X Z
Y Z Y Z Y Z
w w
u u w w
u u w w
nonlinear terms





`
+


+ +

+ +
)

(2.1.2)
where:
- The superscript com and the subscript (u) mean that the strain field is
compatible with the displacement field.
-
,
; , v, ; , ,
i A i
u
u u u w A X Y Z
A

= =

;
- The infinitesimal strains are
( )
{ , , , , , }
com T
u X Y Z XY XZ YZ
= .
For the sake of simplicity, locking effects are simply considered with infinitesimal
strains. It means we consider infinitesimal strains instead of Green-Lagrange strain E.
All types of locking and their remedies for low-order solid elements will be mentioned in
details in the following section.
















2.1.1 Difficulties with low-order standard elements
Low-order elements are preferred in nonlinear structural mechanics because of
their low computational cost and simplicity in dealing with the geometry. However, in
many cases, especially in pure bending problems, the low-order solid element exhibits a
low-precision result due to stiffening effects known as locking.
Locking terminology
A number of different concepts to define, explain and quantify the locking effects
have been discussed in the past. In the sequel it is tried to classify some of them in two
different groups.

222 Gauss integration points
Figure 2.1.1: Hexahedral solid element
nodal points
X
Y
Z
1 2
4 3
5
6
8 7
1
2
4
3
5
6
8 7
9
Mathematical point of view:
In the mathematical literature, the term locking is not as popular as in
engineering literature. From a mathematical point of view, it is rather an ill-conditioning
of the underlying mechanical problem, or the system of partial differential equations, to
be more precise. The crucial property is the presence of a certain small scale parameter
within the equations. This parameter leads to a high ratio of the coefficients in the
discretized system of equations (e.g. the stiffness). Thus, the parasitic terms, evolving
from unbalanced shape functions, are overly enlarged. The element locks if there is no
uniform convergence with respect to this parameter (i.e. the rate of convergence in the
range of coarse meshes depends on this parameter). According to Wilson et al. [WIL73],
effect of the ill-conditioning will be minimized by the use of a computer with high
(double, for example) precision or by restricting the application of the element to thick-
walled structures.

Mechanical point of view:
The simplest way to explain locking is to associate the effect with the presence of
parasitic (or spurious) strains or stresses. With parasitic we mean such strains (stresses)
that do not show up in the exact solution of a certain problem. These are, for instance,
transverse shear strains in the case of pure bending of a plate element (transverse shear
locking) or membrane strains in the case of inextensional bending of shells (membrane
locking) or volumetric strain in the case of incompressible behavior (volumetric locking).
In fact, the well-known locking phenomenon of displacement based finite elements for
thin-walled beams, plates, shells and solids is caused by an unbalance of the trial
functions.
It should also be mentioned, that a natural strategy to remedy locking effects is the
design of higher order finite elements. However, they are not attractive for nonlinear
structural analysis because of expensive computation and complicated configuration.
Locking of certain low-order standard displacement based finite elements comes along in
different ways, namely as volumetric locking if incompressible or nearly incompressible
materials are used or as membrane, shear and curvature thickness locking if the stress and
strain space is not compatible due to the spatial discretization. The phenomena presented
hereafter are the severe locking effects that happen with the low-order standard solid
element. Other severe locking effects which happen to the solid-shell elements are
mentioned in the next chapter.

2.1.1.1 Volumetric locking
If nearly-incompressible or incompressible material behavior is concerned, the
low-order standard solid elements suffer from volumetric locking. In dealing with this
locking, the following condition on the volumetric strain
v
is required during the
deformation process:
( ) 0
v
u v w
tr
X Y Z

= = + +

(2.1.3)
With the above tri-linear displacement field (2.1.1), the volumetric strain
com
v
is
correspondingly calculated as:
10
1 2 3 4 6 4 5 6 5 7 7 7
( ) ( ) ( ) ( ) ( )
com
v
a b c b c X a c Y a b Z c XY a YZ b ZX = + + + + + + + + + + + (2.1.4)
The volumetric strain (2.1.4) can be constrained to be zero by imposing that the
coefficients of each term in (2.1.4) vanish as:

1 2 3
4 6
4 5
0
0
0
a b c
b c
a c
+ + =
+ =
+ =

(2.1.5a)
(2.1.5b)
(2.1.5c)

6 5
7 7 7
0
0 ; 0 ; 0
a b
a b c
+ =
= = =

(2.1.5d)
(2.1.5e)
But the volumetric strain
com
v
is generally non-zero, since the presence of the
terms a
7
, b
7
and c
7
in the sense that they come from isolated terms of u or v or w, see
(2.1.1). These isolated terms are different from zero in order to assure the completeness
of interpolation functions. In other words, the terms a
7
, b
7
and c
7
make
com
v
usually
different from zero. Forcing the incompressibility condition (e.g. for = 0.5 in elasticity
or for incompressible plasticity such as J2 von Mises plasticity) will impose
7 7 7
0 a b c = = = and thus an excessive stiffness is generated by this condition hence the
name locking.
Consequently, the constraint of an incompressible material, (2.1.3), cannot
generally be fulfilled by the normal strains of the pure displacement element. The effect
of this deficiency on the deformation behavior of an element can be explained by using
the internal energy. The internal energy
int
of an element consists of a deviatoric term
d
and a volumetric term
v
. It is defined by
e e e
u K u
T
int
) (
2
1
= and the relative
contribution of deviatoric and volumetric deformation to the element stiffness can be
shown to be:
2 : 3 : 2 : ( )
e e e e
2
int d v d d e v v e d d e e
V V V V
G dV dV G dV tr dV = + = + = +


(2.1.6)
with: - bulk modulus G
) 2 1 (
) 1 (
3
2

+
= ;
- volumetric strains
1

3
ii
tr( ) ; i 1,2,3 = =
v
I = ;
- deviatoric strains =
d v
.
In contrast to all other known locking effects, which are primarily kinematic or
geometric effects, volumetric locking depends on a material parameter, Poissons ratio .
Therefore, also the term Poisson locking is sometimes used in the literature. It is
straightforward to consider the bulk modulus as the critical parameter. For = 0 there is
no spurious volumetric locking at all; the effect becomes more and more pronounced as
0.5 because lim
( 0.5)
= . In solid mechanics this effect can occur, e.g. for rubber
materials, but also for metals in the range of plastic deformations (yielding).
The bulk modulus becomes very large for
2
1
. If ) ( tr is not vanishing, the
stiffness of one element or a group of elements will thus be much larger than the stiffness
of the real incompressible continuum, for which the term
v
is vanishing. For the
standard solid element, since the incompressible condition (2.1.3) cannot be
accomplished, consequently, an undesirable stiffness is added to the rigidity of the
11
element and it makes the element stiffer than the real continuum. In other words, the non-
zero volumetric strain
com
v
leads to volumetric locking.
2.1.1.2 Shear locking
Transverse shear locking can occur in shear deformable beam, plate and shell
elements. In principle, it is also present in solid elements if these are applied to the
analysis of thin-walled structures. However, with solid elements, it is simply called
shear locking, because there is no distinct transverse direction in a solid element. On
the contrary, the solid-shell elements take the shell behavior in thickness direction hence
the transverse direction is distinct. Shear locking in the solid elements is one of the most
severe locking effects because it does not only slows down the convergence but also can
essentially preclude an analysis with a reasonable amount of numerical effort in practical
applications.
From the classical theory of elasticity when the element is subjected to a pure
bending situation, such as bending in 0X direction around 0Z axis (Figure 2.1.1), the
shear strain,
XY
in this case, must vanish. However, with the above trilinear field (2.1.1),
the shear strain
com
XY
in (2.1.2) is calculated as:

2 1 5 6 4 4 7 7
( ) ( )
com
XY
u v
a b a b Z a X b Y a XZ b YZ
Y X

= + = + + + + + + +

. (2.1.7)
That shear strain is generally non-zero. It is equal to zero only when the
coefficients of each term vanish as:

2 1 5 6
4 4 7 7
0; 0;
0; 0; 0; 0;
a b a b
a b a b
+ = + =
= = = =

(2.1.8a)
(2.1.8b)
Equations (2.1.8a) contain coefficients from both the contributing interpolation
functions ( u and v ) which are relevant to the description of the shear strain field
com
XY
.
Hence, these coefficients can correctly represent a true zero condition on shear strain
com
XY

when
1 2
b a = and
6 5
b a = . Each equation (2.1.8b) contains only an isolated term from
u or v , (
4
a or
7
a ) and (
4
b or
7
b ). In general, these isolated terms are different from zero
in order to assure the completeness of interpolation functions, see (2.1.1). As a
consequence, these coefficients make
com
XY
also different from zero. In other words, the
presence of coefficients
4
a ,
4
b ,
7
a and
7
b in
com
XY
cause shear locking, hence, they are
called inconsistent terms [CHA89].
The same arguments are applied for the other shear strain components
com
YZ
,
com
XZ
:

3 2 6 4 5 5 7 7
( ) ( )
com
YZ
w v
b c b c X b Y c Z b XY c XZ
y z

= + = + + + + + + +



3 1 5 4 6 6 7 7
( ) ( )
com
XZ
u w
a c a c Y a X c Z a XY c YZ
Z X

= + = + + + + + + +


(2.1.9a)
(2.1.9b)
The shear strain
com
YZ
is equal to zero only when the coefficients of each terms in
(2.1.9a) vanish as:

; 0 ; 0 ; 0 ; 0
; 0 ; 0
7 7 5 5
4 6 2 3
= = = =
= + = +
c b c b
c b c b

(2.1.10a)
(2.1.10b)
Similarly, the shear strain
com
XZ
is equal to zero only when the coefficients of each
terms in (2.1.9b) vanish as:
12

; 0 ; 0 ; 0 ; 0
; 0 ; 0
7 7 6 6
4 5 1 3
= = = =
= + = +
c a c a
c a c a

(2.1.11a)
(2.1.11b)
Consequently, non-physical effects with compatible shear strains cause the so-
called shear locking for the standard elements by introducing artificial flexural stiffness.
This phenomenon is essential with the vanishing of the thickness of standard elements in
the modeling of bending dominated problems.

Table 2.1.1: Deformation modes of bilinear element (2D.Q1)
c
1
c
2
c
3
c
4
c
5
c
6
c
7
c
8

Mode









u 1 0 X 0 Z 0 XZ 0
w 0 1 0 X 0 Z 0 XZ

X

0 0 c
3
0 0 0 c
7
Y 0

Z

0 0 0 0 0 c
6
0 c
8
X

XZ
0 0 0 c
4
c
5
0 c
7
X c
8
Z















For the sake of more clarity, lets investigate deformation modes of a beam in
bending, Figure 2.1.2. Deformation of the beam is assumed to be independent of Y the
width direction. In that case, the displacement field (2.1.1) becomes:

( , ) 1 3 5 7
( , ) 2 4 6 8
X Z
X Z
u c c X c Z c XZ
w c c X c Z c XZ
= + + +
= + + +

(2.1.12)
It means the trilinear eight-node hexahedral element reduces to the bilinear four-
node quadrilateral element. Lets consider deformation modes of the displacement field
(2.1.12). All of those deformation modes are tabulated in Table 2.1.1.
In Table 2.1.1, c
1
and c
2
are rigid body modes. c
3
to c
6
are constant strain modes
and c
7
and c
8
are linear strain modes. When an in-plane bending deformation happens, it
means bending in X-direction around Z-axis, the mode No. 7 is active only the
coefficient c
7
is non-zero, thus leading to a parasitic linear shear strain
XZ
in X-direction
(see Table 2.1.1). In other words, it is impossible to find a linear combination of modes
that leads to a linear variation of
X
in Z-direction without being accompanied by shear
b) FEM representative
0
XZ
in general
a) Continuum mechanics
0
XZ
=
Figure 2.1.2: Pure bending of a rectangle
X
Z
X
Z
0
0
F
F
F
F
2a
M
M
T=0
T=0
2
b

13
strain
XZ
. This phenomenon is the manifestation of (transverse) shear locking. The same
is, in turn, true for the bending in Z-direction (see Table 2.1.1 - mode 8).
Lets investigate analytical solution of a beam of rectangular cross section which
is bent by two equal and opposite couples M (Figure 2.1.2). Stress components of the
beam are:

0
Z
X XZ
MZ
I

=
= =

(2.1.13)
where I is the area moment of inertia of the beam cross section.
From (2.1.13) and stress-strain relations we attain displacement components by
integrating strain-displacement relations, see [DUR58]. Finally, the analytical
formulations for displacements are:

( , ) 1 2
2 2
( , ) 1 3
( )
2
CM
X Z
CM
X Z
M
u XZ C Z C
EI
M
w X Z C X C
EI

= + +
= + +

(2.1.14)
where
1
C ,
2
C and
3
C are constants of integration. The superscript CM means that the
displacement field is calculated by the continuum mechanics.
If we impose the boundary conditions as: symmetric plane of the beam is the plane
which goes through X=0; vertical displacement at the 4 corners is equal to zero. Then, the
analytical solution of the problem in Figure 2.1.2 is:

( , ) 1
2 2 2 2 2 2 2 2
( , ) 1 2
1
( ) ( ) ( ) ( )
2 2 2
CM
X Z
CM
X Z
M
u XZ XZ
EI
M M
w a X b Z a X b Z
EI EI

= =
= =

(2.1.15)
where
1
M
EI
= and
2 2
M
EI

= are constants. The constant

2
is a function depends on
material properties, for Poissons ratio equal to zero that constant is 0
2
= .
The analytical strains calculated from (2.1.15) are:

1
2
0 if 0
2 Z if 0
0
CM
X
CM
Z
CM
XZ
Z
u w
Z X

=
=
=

= + =


(2.1.16)
Obviously, the solution (2.1.15) satisfies the pure bending condition zero shear
strain constraint.
For finite element solution, when pure bending occurs, only the mode 7 (Table
2.1.1) is active, it means:

( , ) 7
( , )
0
X Z
X Z
u c XZ
w
=
=

(2.1.17)
Hence, compare with the analytical solution, the form of error in the numerical
solution is:
14

2 2 2 2
( , ) 1 2
( ) ( ) ( )
X Z
Error w a X b Z = +
(2.1.18)












Therefore, the errors in strains are followed by the errors in displacement as:

2
7
0 if 0
( )
2 Z if 0
( )
Z
XZ
Error
Error c X

=
=

=

(2.1.19)
These errors cause parasitic stresses as shown in Figure 2.1.3. In fact, the
Error(
XZ
) causes shear locking, the Error(
Z
) causes Poisson thickness locking (will be
discussed in the next section).
In conclusion, shear locking happens because normal strains of linear elements are
coupled by shear strains. Elements do not have pure bending modes to behave correctly
for pure bending load cases. The consequence is there are parasitic shear strains appear
simultaneously with normal strains, which are physical in pure bending cases. These
shear strains are important compared to the normal strains.
Consider the pure bending problem under the point of view of the continuum
mechanics theory. In pure bending, the applied moment is constant and then shear stress
must vanish since the shear stress is the derivative of the bending moment with respect to
the axial coordinate. When solving the pure bending problem by the finite element
method, since the shear stiffness is often significantly greater than the bending stiffness,
the spurious shear absorbs a large part of the energy induced by the external forces and
the predicted deflections and strains are much too small. In other words, the additional
shear stress in the element (which does not occur in the actual beam) causes the element
to reach equilibrium with smaller displacements, i.e., it makes the element appear to be
stiffer than it actually is and gives bending displacements smaller than they should be.
Long, slender structures in bending have greater curvature than do short, deep ones, and
when modeled with low-order elements, will be affected more by shear
locking. Increasing the number of elements will allow a more accurate modeling of the
curvature, and reduce the effects of shear locking. Shear locking is prevented through the
use of high order elements.
0
x
y
0
x
y
b)
Z
( 0) a)
XZ

Figure 2.1.3: Errors in stresses in pure bending
z
z
15
2.1.1.3 Poisson thickness locking
Also in bending problems, e.g. bending in 0X direction around 0Z axis (Figure
2.1.2), a linear distribution of strain
CM
Z
over the thickness in the 0Z direction, see
(2.1.16b)
2
2 Z
CM
Z
= , is theoretically expected. However, the standard element (2.1.12)
gives only a constant strain (with respect to Z) as:

6 8
com
Z
w
c c X
z

= = +

(2.1.20)
This constant (with respect to Z) approximation of the strain
com
Z
in the thickness
direction is contrary to the linear variation along 0Z of
CM
Z
. In the real structure, due to
Poisson effect the normal strain (
CM
X
MZ
EI
= ), which linearly vary in the 0Z direction,
would cause a transverse normal strain also linearly varies along 0Z (
CM CM
Z X
= ).
However, it is not the case with
com
Z
calculated by the low-order interpolation function,
see (2.1.20). Consequently, the transverse normal stress in 0Z direction, which is
calculated as
[(1 ) ]
(1 )(1 2 )
com com com
Z Z X
E


= +
+

(2.1.21)
is not equal to the analytical solution along the thickness when bending occurs. It leads to
an undesired locking, which is known as Poisson thickness locking phenomenon.
In general, Poisson thickness locking is due to the resulting incorrect-linear
distribution of the normal stress in thickness direction. That locking effect does not
diminish with mesh refinement in all directions except the thickness direction (layers).

2.1.2 Solution for a locking free element
As discussed above, the low-order standard solid element suffers from locking
effects. It has been pointed out that the stiffening in the case of incompressibility is
caused by redundant terms in the normal strains. Using the SRI method, will be discussed
at the end of this chapter, is one of the best way for eliminating volumetric locking.
Restriction of the SRI method is that the applied material laws must allow a decoupling
stress and also strain into volumetric parts and deviatoric parts. Volumetric locking could
also be vanished by inserting suitable enhancing components in order that the
incompressible condition (2.1.3) be satisfied. The use of the EAS method with nine
volumetric modes (see Section 2.4), as proposed by Andelfinger and Ramm [AND93],
assures that all three normal strains
ii
consist of the same polynomial fields, then no
spurious constraint is produced. Disadvantage of this way is highly computational cost
for each element because the enhancing strains making the element system of equations
larger. In the thesis, only three EAS internal parameters are used to remove volumetric
locking for solid-shell element, see Chapter 3.
Shear locking for solid elements can be removed by the ANS method. Detailed
discussions are presented in Section 3.3.
To overcome Poisson thickness locking additional terms with linear distribution in
thickness direction must be introduced for the transverse normal strains. This also assures
16
the stress field
z
correctly varies linearly along the thickness in bending situations.
Using the EAS method, the terms (
3
,
1

3
,
2

3
) of EAS7 element (see Section 3.4 below)
can be used to circumvent Poisson thickness locking. In fact, the constant normal strain
in thickness direction is enhanced with a linear extension over the thickness and linear in
in-plane direction according to the EAS method.
In general, it is quite possible to use standard solid elements for the analysis of
shell-type structures if one can overcome the following problems as pointed out by
Wilson et al. [WIL73]:
1. Most solid elements have not had ability to represent accurately bending
moments.
2. Due to the full integration, i.e. 222 Gauss integration (see Figure 2.1.1),
the element will behave badly for isochoric material behavior, i.e. for high
values of Poisson's ratio or plastic behavior (due to volumetric locking, see
SRI method at the end of this chapter for a reference solution).
3. Errors in the transverse shear cause the element to be very stiff (transverse
shear locking).
4. For simulation of thin shells, because the thickness of the element is
relatively small compared to the in-plane dimensions there are relatively
large stiffness coefficients in the thickness direction of numerical problems
that are introduced. This effect makes the simulation problem ill-
conditioned.
The design of the solid-shell elements aims to overcome these disadvantages of
the standard solid element. Indeed, the first two problems can be solved by the use of the
EAS method. The third and fourth problems can be minimized by applying the ANS
method originally developed for thin shell elements. Practically, the solid-shell elements
adopt the EAS technique for the in-plane shear strains and in-plane normal strains, and
adopt the ANS technique for the transverse shear strains and transverse normal strain. As
a result, the solid-shell elements may overcome all of the above difficulties. Thanks to
intrinsic performance of the ANS method, the solid-shell element can be applied not only
for thin but also for moderately thick shell structures.

2.2 INTRODUCTION TO CONVENTIONAL SHELLS
In this section the conventional shell elements are briefly introduced. Some basic
concepts mentioned here are useful for building the solid-shell elements. We will briefly
discuss about disadvantages of the conventional shell elements compared with the solid-
shell elements, such as modified material models, variation of thickness strains,
contradictions of assumptions for shell elements, etc.
Basically, the conventional shell finite elements are developed from one of the two
following approaches:
1. Develop the formulation for shell elements by using classical strain,
displacement and momentum (or equilibrium) equations for shells to develop a weak
form of the momentum (or equilibrium) equations.
17
2. Develop the element directly from a continuum element by imposing the
structural assumptions on the weak form or on the discrete equations; this is called the
continuum based approach. For example, the kinematic assumptions will be imposed on
the discrete equations, i.e. the continuum finite element will be modified so that it
behaves like a shell [AHM70], [DVO84].
The first approach, also called classical shell theories, is difficult, particularly for
nonlinear shells, since the governing equations for nonlinear shells are very complex and
awkward to deal with. They are usually formulated in terms of curvilinear components of
tensors, and features such as variations in thickness, junctions and stiffeners are generally
difficult to incorporate. There is still a disagreement as to what are the best nonlinear
classical shell equations [STO95].
The continuum-based approach, on the other hand, is straight forward, yields
excellent results, is applicable to arbitrarily large deformations and is widely used in
commercial softwares and researches. The popular continuum-based method used in
structural analysis is called the degenerated continuum approach, see e.g. Ahmad et al.
[AHM70].

2.2.1 Classical shell theories
Earlier, a shell was considered as a curved form of a plate and its structural action
is a combination of stretching and bending [ZIE00b]. It is possible to perform a finite
element analysis of a shell by using what is called a facet representation - i.e. the shell
surface is replaced by a flat triangular and/or quadrilateral plate elements in which a
membrane stiffness (membrane element) is superposed on a bending stiffness (plate
bending element). Such a model is understandably inaccurate in the sense that with very
coarse meshes, they do not capture the bending-stretching coupling of thin shell behavior.
Hence, the motivation for designing elements is twofold: mid-surface curvature has to be
taken into account and the element has to capture the membrane-bending coupling
correctly. There are two types of kinematic assumptions, those that admit transverse
shear strains and those that don't. The theory which admit transverse shear strains are
called Mindlin-Reissner theories, whereas the theory which does not admit transverse
shear strains is called Kirchhoff-Love theory. The essential kinematic assumptions in
these shell theories are:
1. The normal to the mid-surface remains straight and normal (Kirchhoff-Love
theory).
2. The normal to the mid-surface remains straight (Mindlin-Reissner theory).
Shell theories, see [ZIE00b], proved that the Kirchhoff-Love assumptions are the
most accurate in predicting the behavior of thin shells. For thick shells, the Mindlin-
Reissner assumptions are more accurate because transverse shear effects become
important. Transverse shear effects are particularly important in composites. Mindlin-
Reissner theory can also be used for thin shells. In that case the normal will remain
approximately normal and the transverse shear strains will almost vanish.
The assumptions for Kirchhoff-Love shell theory are:
1) The shell is thin compared to the radius of curvature, i.e. t/R << 1.
2) The linear and angular deformations of the shell are small.
18
3) The transverse normal stress is negligible: 0 =
ZZ
.
4) The normals to the reference plane before deformation remain normal, straight
and inextensible after deformation.
The first assumption is the most basic since it implies the other three. It means the
assumption 1) cannot be violated without violating assumptions 2-4. Assumption 2) in
effect means that the state of the deformed shell can be related directly to the state of the
non-deformed shell. Assumption 3) is reasonable for thin shells (except for plasticity) and
has no further implications besides simplifying the derivation of the governing equations.
Assumption 4) has two implications; first, the inextensibility assumption implies zero
normal strain ( 0 =
ZZ
); second, the normal remains normal, this assumption eliminates
the possibility of transverse angular distortions and consequently, lead to neglect the
transverse shearing ( 0
XZ YZ
= = , see the coordinate system in Figure 2.1.1).
Mindlin-Reissner theory is applied for thin to thick shells. The kinematic
assumptions are:
1. The normal to the mid-surface remains straight throughout deformation.
2. The length of the normal remains unchanged throughout deformation.
3. Transverse normal stresses are negligible 0 =
ZZ
.
When Poissons ratio is not equal to zero the latter two assumptions are
contradictory because the normal must stretch when 0 =
ZZ
. Reissners and Mindlins
theories differ from each other in the way they solve this problem. While Reissner
assumes a cubic variation in thickness direction of the transverse normal stresses,
Mindlin manipulates the material law in order to comply with his assumptions. Although
conceptually different, both theories practically lead to the same results for transverse
displacements, shear forces and bending moments in actual structural analysis. The
method of Mindlin is simpler so it is popularly applied for numerical plate and shell
models. In order to construct the finite element formulas, the Mindlin method requires the
variational principle with only a displacement field but the Reissner method needs two
fields (displacement and stress) variational principle.
Although these elements have the advantage of being able to account for the
transverse shear that occurs for thick shell, low-order forms of these elements are subject
to shear locking.
Employing the first and second assumptions makes the transverse shear strains are
constant over thickness direction. Meanwhile, employing the third assumption requires
modifications of the 3D-material law. This work is not simple in such an approach,
especially for complicated material laws which are described only for 3D-continuum.
In order to use 3D-material laws directly, the normal stress in the thickness
direction must be taken into account. This leads to an interpolation of the extensible
director vector (defined in Figure 2.2.1), see [BET96], [BIS97].

2.2.2 Degenerated shell elements
There were various shell elements whose formulations are derived from the
degeneration concept introduced by Ahmad et al. [AHM70]. The core of this concept is
19
the discretization of a 3D mathematical model with 3D elements and their subsequent
reduction into 2D elements. As classified above, the degenerated shell elements are built
from a so-called continuum based approach (CM). In comparison the CM to shell theory,
it is not necessary to develop the complete formulation, i.e. developing a weak form,
discretizing the problem by using finite element interpolations, etc. The degeneration of
this 3D shell element is done by eliminating the nodes with the same (
1
,
2
) coordinates
into a single node located at the mid-surface of the element, as shown in Figure 2.2.1.
The procedure when creating a shell element using the degenerated solid approach
is to eliminate nodes by enforcing different constraints on the behavior of the element.
First, nodes on the mid-surface are removed (nodes 17 to 20, Figure 2.2.1),
corresponding to assuming constant transverse strains. Then, opposite nodes (1&9, 3&11,
5&13, 7&15; 2&10, 4&12, 6&14, 8&16) are linked by assuming equal displacements (u,
v and w) and assigning two rotational DOFs (
x
and
y
) to each pair of nodes. Finally,
the motion of each straight line is described by five DOFs in one node, lying on the
reference surface, Figure 2.2.1b.

Assumptions: For the shear deformable shells, the following assumptions are made:
1. The fibers (line connects bottom node with top node) remain straight. The unit
vector along each fiber is called a director vector;
2. The element is in a plane stress state, so 0 =
ZZ
;
3. The elongation of the fibers is governed by conservation of mass and/or the
constitutive equation.
The first assumption will be called the modified Mindlin-Reissner assumption. It
differs from what we call the classical Mindlin-Reissner assumption, which requires the
normal to remain straight; the fibers are not initially normal to the midline. For the CM
shell element to satisfy the classical Mindlin-Reissner assumptions, it is necessary for the
fibers to be aligned as closely as possible with the normal to the midline. This can be
accomplished by placing the slave nodes (nodes of the original solid element) so that the
fibers are as close to normal to the midline as possible in the initial configuration.
Otherwise the behavior of the degenerated shell element may deviate substantially from
classical Mindlin-Reissner theory and may not agree with the physical behavior.
Obviously, it is impossible to align the fibers with the normal exactly along the entire
length of the element when the motion of the continuum element is C
0
. Contrarily, if the
fibers are inclined too much with respect to the normal while the transverse normal strain
is taken into account, there is an effect call curvature thickness locking occurs, see
Chapter 3 for more details.
Instead of the third assumption, many authors assume that the fibers are
inextensible. Inextensibility contradicts the plane stress assumption: the fibers are usually
close to the z direction and so if 0 =
ZZ
, the strain in the z direction generally cannot
vanish. The assumption of constant fiber length is inconsistent with the conservation of
mass: if the shell element is horizontally stretched, it must become thinner to conserve
mass. Therefore, if the thickness strain is calculated through the constitutive equation via
the plane stress requirement, conservation of mass is enforced. The important feature of
the third assumption is that the extension of the fibers is not governed by the equations of
motion or equilibrium. From the third assumption, it follows automatically that the
20
equations of motion or equilibrium associated with the thickness modes are eliminated
from the system.

































The third assumption can be replaced by an inextensibility assumption if the
change in thickness is small. In that case, effect of the thickness strain on the position of
the slave nodes is neglected, so that the nodal internal forces do not reflect changes in the
thickness. The theory is then applicable only to problems with moderate strains (on the
order of 0.01, for instance).
The degenerated shell elements are in general cheap in computational cost due to
the reduced number of DOF and using coarse mesh. The major shortcoming of the
elements is the problems of locking for thin shells. However, the degenerated shell
elements are attractive since they propose a simple method without discretization of the
governing shell equation as in the case of the direct formulations - the classical shell
theories. We have presented above a brief discussion about shell elements. In the next
1
2
3
7
6
5
8
4
17
18
20
19
9
10
11
15
14
13
16
12

3

X
Y
Z
Solid element Slave node
1
2
3
7
6
5
8
4
9
10
11
15
14
13
16
12

3

X
Y
Z
1;9
2;10
3;11
7;15
6;14
5;13
8;16
4;12

3

X
Y
Z
Connect
opposite nodes
Delete mid-
surface nodes

3

X
Y
Z
5 DOFs/node
1
2
3
7
6
8
4
5
Degenerated shell element
Director
Master node
Figure 2.2.1: Degeneration of a 3D solid element into a
degenerated shell element
21
chapter we will discover performances of the one called solid-shell in comparison with
the degenerated and classical shell elements in simulating thin-walled structures.

2.3 INCOMPATIBLE DISPLACEMENT ELEMENT
Incompatible method, also called assumed displacement method, is derived from
the potential energy variational principle. In the thesis the incompatible method is
presented as a reference to motivate for the EAS method. Hence, the incompatible
method is briefly introduced and only considered in linear elastic theory.
The standard solid elements pose the following difficulties: locking phenomena
for bending and incompressible problems. By adding incompatible displacements to 3-D
isoparametric nodes, the mentioned difficulties are canceled.

2.3.1 Finite element formulation
Consider a continuum body occupying a volume V in a space of boundary surface
A. Assume that the body force and the tension force are conservative and the object is in
static state. Under the theory of linear elasticity, the principle of minimum potential
energy can be stated as:

( )
* *
( ) ( )
1
( )
2
com T com
V A V
dV dA dV

=

4
u u u
C u t u b
(2.3.1)
When the solid continuum is discretized into a finite number of elements, the
above variational is rewritten in the form as:

( )
4
1
1
( ) ( ) * *
2
e e
Nele
e T e e e e e
e
V A
dV dA

=

(

=
( `
(

)

u
Du C Du u b u t
(2.3.2)
where Nele : number of elements.
In the finite element formulation the element displacements
e
u are interpolated in
terms of nodal displacements that may be at both boundary nodes (Serendipity elements)
and internal nodes (Lagrange elements). Elements can also be formulated by adding to
the original element displacements
e
u , which are in terms of nodal displacements
e
U ,
higher-order displacements , which are not expressed in terms of nodal displacements
of the boundary nodes. For example, the displacements u and v for the standard four-
node quadrilateral element are based on bilinear interpolation functions. They are
incomplete in quadratic terms, see (2.1.12). Improvement of the performance of a four-
node element can be made by adding terms such that the displacements are complete in
quadratic terms. Wilson et al. [WIL73] suggested the addition of incompatible
displacements that vanish at all corner nodes. In these cases the element strain
e
Du =
can be expressed as
22
[ ]

=
e
e
u
e

U
B B Du

(2.3.3)
where

B B ,
u
are compatible strain matrix and incompatible strain matrix, respectively.

e
U is nodal displacement vector and
e
is incompatible mode vector of element:

] .. . [
] .. . [
2 1
2 2 2 1 1 1
inMod
e
Nnode Nnode Nnode
e
W V U W V U W V U
=
=

U

(2.3.4)
with Nnode is the number of nodes per element; inMod is the number of incompatible
modes per element.
Introducing (2.3.3) into (2.3.2) we have:

( )
1
1 1
( ) ( )
2 2
( ) ( )
Nele
e T e e e e e e T e e
uu u
e
e T e e T e
ext

= + +

u
U k U k U k
U f f

(2.3.5)
where:
- The standard stiffness matrix of the element is:
dV
u
T
V
u
e
uu
e
B C B k
4
0
) (

=
(2.3.6)
- The incompatible-standard stiffness matrix of the element is defined as:
dV
u
T
V
e
u
e
B C B k
4
0
) (

=


(2.3.7)
- The incompatible stiffness matrix of the element is defined as:
dV
T
V
e
e

B C B k
4
0
) (

=
(2.3.8)
- The standard nodal force vector is:
( ) ( )
e e
T T
e e e
ext u u
V A
dV dA

= +

f N b* N t *
(2.3.9)
- The incompatible nodal force vector is defined as:
( ) ( )
e e
T T
e e e
V A
dV dA

= +

f N b* N t *
(2.3.10)
with

N are the incompatible shape functions.

The incompatible vector
e
consists of internal variables hence it can be
condensed out of the variational formulation by setting 0 / = in order to get
e
in
function of
e
U as:
23

1
[ ] ( )
e e e e
u

= k k U f (2.3.11)
Then, the total potential energy
( ) u
is rewritten:

( )
1
1
( ) ( )
2
Nele
e T e e e T e
e

=
(
=
(

u
U k U U f
(2.3.12)
where the equivalent element stiffness matrix is

1
( ) ( )
e e e T e e
uu u u

= k k k k k (2.3.13)
and the equivalent nodal force vector is

1
( ) ( )
e e e T e e
ext u

= f f k k f (2.3.14)
Assembling
e
k into the global stiffness matrix K and
e
f into global nodal vector
F , finally, the total potential energy
( ) u
is:

( )
F U U K U
u
= ) ( ) (
2
1
T T

(2.3.15)
where:
- U is the global nodal displacement vector;
-
1
Nele
e
e=
=

K k ;
-
1
Nele
e
e=
=

F f .
Lets take the first variation of
( ) u
with respect to global displacement vector U
and impose it equal to zero,
( )
0 / = U
u
, the equation for displacement solution is
F U K = (2.3.16)
After the solution process, solving system (2.3.16), all the nodal displacements are
known. Other variables, strains and stresses, are obtained as in the standard manner.
In the next section, the EAS method, based on the three field Hu-Washizu
variational principle, is presented. The EAS method is considered as a generalized
approach of the incompatible method as pointed by Simo and Rifai [SIM90].

2.4 EAS ELEMENT

Due to their efficiency and simple geometry, low-order solid elements are often
preferred in structural mechanics. As mentioned above, the low-order standard
displacement elements exhibit, in many cases, severe stiffening effects known as locking.
Shear locking occurs when simulating thin-walled structures by the low-order standard
displacement elements, where pure bending modes are spoiled by parasitic shear strains.
Membrane locking is encountered in high aspect ratio elements when bending modes
24
cannot be separated from membrane strains and, thus, not allowing the verification of
pure inextensional modes. For incompressible or nearly incompressible conditions,
volumetric locking may also occur; in this case, deviatoric modes always come along
undesirably with volumetric strains. The class of EAS elements presented below allows
the systematic development of low-order elements with enhanced accuracy for coarse
meshes.
The EAS elements have been applied to simulate geometrically and materially
nonlinear problems due to the fact that they perform well in severe situations as the
nearly incompressible limit and pure bending situations. Compared with almost all finite
elements, the EAS elements show very good coarse mesh accuracy. In general, a low-
order free-locking element can be developed based on EAS technique.

2.4.1 Variational formulation
Initially proposed by Simo and Rifai [SIM90] for small strains, the EAS method,
which involves the three field variational principle of Hu-Washizu, was lately extended
to the finite strain theory by Simo and Armero [SIM92] and open to the
thermomechanically coupled behavior [ADA05]. When incorporated with the ANS
technique, that assumes directly on strain components, a formulation in terms of the
Green-Lagrange strains is however more favourable than the one based on the
displacement gradient. In the light of this consideration, the following form of the Hu-
Washizu principle is taken as the variational principle for a hyperelastic material:

( )
0 0
0
( ) int ( )
* * *
: [ ]
( )
u
com
ext , ,
V V
A A V
W dV dV
dA dA dV

= = + +
+
+


u S E u E S
S E E
u u t u t u b

(2.4.1)
where:
-
int
and
ext
: internal and external energies, respectively;
- The script * denotes prescribed values;
- u: admissible displacement field;
- S : admissible second Piola-Kirchhoff stress field;
- t b, : body force and surface traction vectors, respectively;
-
) ( E S
W : stored energy function;
-
com
) (u
E : admissible Green-Lagrange strain field. In the thesis, the letter com in
upper position designates for compatible quantities.
In (2.4.1), the body under consideration occupy a volume
0
V and has the boundary
u
A A A =

where

A denotes the prescribed traction (

*
t ) parts while
u
A denotes the
prescribed displacement (
*
u ) ones. As the thesis concentrates on quasi-static problems,
hence, the inertial force in (2.4.1) is not considered. Furthermore, the body force and
surface traction are assumed to be conservative.
In (2.4.1), three independent fields are the strain field E , the stress field S and the
displacement field u. According to the EAS method the independent strain field is
proposed as:
25

( )
mod com enh
= +
u
E E E E (2.4.2)
The strain tensor
enh
E in (2.4.2) is named the enhanced assumed strain field. The
strain tensor
mod
E is named the modified strain field. Introducing the modified strain
field
mod
E into (2.4.1) we have the Hu-Washizu principle as the variational basis for the
EAS method:

( )
0 0
0
( )
( )
* * *
: [ ]
( )
mod
u
com mod
int ext
, ,
V V
A A V
W dV dV
dA dA dV

= = + +
+
+


mod u
S E u E S
S E E
u u t u t u b

(2.4.3)
Once the modified strain tensor
mod
E is obtained, the gradient deformation tensor
mod
F can be consistently derived through the use of a polar decomposition, see Section
3.5.1. In (2.4.3), three fields u,
mod
E and S are independent, while the two fields t and
S relate together through the Cauchys stress theorem.
In order to pass the patch test, see details in Section 2.4.4, the approximation of
enh
E and S are chosen to satisfy the following orthogonality condition as proposed in
[SIM90]:

=
0
V
E S 0 : dV
enh

(2.4.4)
Applying (2.4.4) in combination with (2.4.2) the three field variational (2.4.3)
reduces to a two field variational principle as below:
( )
dV dA dA dV W
V A A V
u
enh com enh

+ =
+
0 0
*
*
) (
*
) (
) ( b u t u t u u
S
E E S E u,

(2.4.5)
The first variation
( )
enh

u,E
immediately follows:

( )
enh
int ext
= +
u,E

(2.4.6)
where:
-
int
: the variation of the internal term
: ) (
0
int
dV
V
mod enh com

+ = S E E
(2.4.7)
and the second Piola-Kirchhoff stress
mod
S is given by

mod
S
mod
W
E
S

=
(2.4.8)
-
ext
: the variation of the external term

0
* *
ext
A V
dA dV

=

u t u b
(2.4.9)
The orthogonality (consistency) condition (2.4.4) can be interpreted such that the
variationally consistent stress field of an EAS element is complementary to the enhanced
strain interpolation.
26
Following the finite element method, the approximation of the current geometry
vector x and displacement vector u at the element level are read as:

( ) ( )
h e
) , , (
3 2 1


x N x = and
( ) ( )
h e
U N u

=
(2.4.10)
where:
- superior index e refers to quantities at the element level;
- for eight-node hexahedral element [ ]
8 3 2 1 ) (
,..., , , N N N N N

= is matrix
of shape functions
(2.4.11)
in which:

(
(
(

=
I
I
I
I
N
N
N
0 0
0 0
0 0
N (2.4.12)
with the standard trilinear shape functions as

( )
( )( )( )
3 3 2 2 1 1
8
1
I I I I
1 1 1 N + + + =

; I=1-8 is nodal number. (2.4.13)

-
h
) , , (
3 2 1

x and
h
U are the vectors of nodal coordinates and nodal displacements,
respectively. They are defined as follows

] .. . [
] .. . [
8 8 8 2 2 2 1 1 1
8 8 8 2 2 2 1 1 1
W V U W V U W V U
z y x z y x z y x
h
h
=
=
U
x

(2.4.14)
- The superscript h refers to nodal values.
The displacement variation and increment are required by the linearization of the
variational. The displacement variation
( )
e

u are interpolated as:

( ) ( )
h e
) , , (
3 2 1

U N u

=
(2.4.15)
The displacement increment
( )
e

u are interpolated in the same manner:

( ) ( )
h e
) , , (
3 2 1

U N u

=
(2.4.16)
In the following sections, the superscript e in
e
u will be omitted.
The variation E and increment E of Green-Lagrange strain at the element level
are interpolated as:

( )
h com
h
) , , (
3 2 1

U B E
U
= and
( )
h com
h
) , , (
3 2 1

U B E
U
=
(2.4.17)
where the compatible strain-displacement matrix B is a function of displacements u, see
[ZIE00b], such as:
] ... [
8 2 1
com com com com
B B B B = (2.4.18)
with
27
, , , , , , , 0 0
, , , , , , 0 , 0
, , , , , , 0 0 ,
, , , , , , , , , , , , , , 0
, , , , , , , ,
, 0 ,
0 , ,
X i X X i X X i X X
Y i Y Y i Y Y i Y Y
Z i Z Z i Z Z i Z Z com
i
X i Y Y i X X i Y Y i X X i Y Y i X Y X
Z i X X i Z Z i X X i Z
Z X
Z Y
u N v N w N u
u N v N w N v
u N v N w N w
u N u N v N v N w N w N v u
u N u N v N v N w
w u
w v
= +
+ + +
+ +
(
(
(
(
(
(
(
(

B
, , , ,
, , , , , , , , , , , ,
Z i X X i Z
Y i Z Z i Y Y i Z Z i Y Y i Z Z i Y
N w N
u N u N v N v N w N w N
+
+ + +
(
(
(
(
(
(
(
(


(2.4.19)

In order to simplify the notation we are going to introduce Voigt notation. Thus,
components of Green-Lagrange strain vector
com
E and PK2 vector S are transformed
from the corresponding matrices with indices as listed in the table below.
Table 2.4.1: Index transformation
Matrix index 11 22 33 12 (21) 13 (31) 23 (32)
Vector index 1 2 3 4 5 6

The enhanced strains
enh
E , which will be presented in details below as functions
of enhanced strain interpolation matrix , see (2.4.45), and internal strain parameters ,
variation
enh
E and increment
enh
E of the enhanced strains are:
E =
enh
; E =
enh
; E =
enh
(2.4.20)
This variation and increment are required by the linearization of the weak form,
see (2.4.23).

2.4.2 Linearization of discrete weak form
Employing the Voigt notation (Table 2.4.1), we rewrite the internal term (2.4.7) at
the element level:

+ =
e
V
mod T enh
e
V
mod T com e
int
dV dV
enh
0
1 6
6 1
0
1 6
6 1
)
) ( ) ( S E S E
E (u,

(2.4.21)
The first variation of the external term (2.4.9) is


=
e
A
e
V
e
ext
dA dV

* *
t u b u .
(2.4.22)
We use
e
int
and
e
ext
to formulate the linearization of the weak form
e
)
enh
E (u,
by employing the truncated Taylor series about the k
th
iteration:

( )
( )
( )
( )
( ) ( )
) ) )
1
) )
| ,
,
| ,
enh enh enh
k+1 k+1 k k k k
enh enh
k k k k
e
e e e e enh
int ext
enh
k
e enh
D

= + = +

= +
(u ,E (u ,E (u ,E
(u ,E (u ,E
u E
u E
u E

(2.4.23)
where ( ) D is the Gateaux derivative operator, see Simo and Hughes [SIM98].
28
To alleviate the notation, the right subscript k designating the iterative index is
omitted. In order to calculate (u, E
enh
) we let the right hand side of (2.4.23) equal to
zero, the result is:

( ) ( ) ( )
( )
( )
( )
( )
( )
( )
)
- | ,
, ,
, ,
e e e enh
int ext
e e e
int ext int
h e h e
h e h e
D



+ =
+
= =

enh
(u ,E
u E
U U
U U

(2.4.24)
Introducing (2.4.15), (2.4.17) and (2.4.20) into (2.4.21), we get the internal virtual
work as:

( )
0 0
( ) ( )

h
e h T mod T mod
int
e e
V V
(h)T e (e)T e
int enh
dV dV

= +
= +

e
U ,
B U S S
U f f

(2.4.25)
with:

0
e T mod
int
e
V
dV =

f B S ;
0
e T mod
enh
e
V
dV =

f S
(2.4.26)
Introducing (2.4.15) into (2.4.22) of the external virtual work we get:

e
ext
(h)T e
ext
f U = (2.4.27)
where:
dA dV
e
A
T
e
V
T e
ext

+ =

* *
t N b N f
0

(2.4.28)
Observe that the element external force vector
e
ext
f (2.4.28) has the usual
expression of the standard displacement element.
Substituting (2.4.25) and (2.4.27) into (2.4.24) we have:

( ) ( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
,



e e
int int
e enh h e
int h e
e e
int enh
(h)T h (e)T h
h h
e e
int enh
(h)T e (e)T e
e e
D

= +


= + +


+ +

u E U
U
f f
U U U
U U
f f
U


(h)T e (e)T e h (h)T e (e)T e
uu u u
( ( = + + +

e
U k k U U k k

(2.4.29)
The constitutive tensor in the physical space is expressed through the following
stress-strain relationship:
[ ]
6 6
(

= =
mod
kl
(ij)mod
ijkl
C
E
S
C
4

(2.4.30)
In (2.4.29) the stiffness matrices are established and listed hereafter.
29
The standard stiffness matrix of the element includes the material part
e
mat
k and
geometrical part
e
geo
k that is

0 0
4 mod
( ) ,
h
e e
e
T
e e e int
uu mat geo h
V V
dV dV

= = + = +

U
k k k B C B B S
U
f

(2.4.31)
where, for a geometrical nonlinear theory the strain-displacement matrix B , see (2.4.19),
is a function of the displacements u. Furthermore, the strain-displacement matrix contains
the derivatives of the shape functions with respect to the global co-ordinates X in the
reference configuration. Components of the geometrical part
e
geo
k is defined as, see
(10.69) of [ZIE00b]:

( 24 24)
11 12 18
21 22 28 e
geo
81 82 88

(
(
(
=
(
(

G G G
G G G
k
G G G

(2.4.32)
where
IJ
G ( I , J = 1-8) is defined for a node combination I and J as
IJ IJ 3
G = G I ; with
3
I
is the unit matrix of dimension (33) and:

0
, ,
e
IJ I K KL J L
V
G N S N dV =

; , 1- 3 K L = .
(2.4.33)
Enhanced-compatible stiffness matrix of the element is
dV
T
V
T
e
int T e
u
e
u
e
B C

k k
e
f
) (
4
0

=
|
|

\
|

= =


(2.4.34)
where is defined in the next section, expression (2.4.45).
Enhanced stiffness matrix of the element is

0
4

e
e e mod
T
e enh enh
e mod e
V
dV

= = =


E
k C
E
f f

(2.4.35)
Combination of (2.4.23), (2.4.25), (2.4.27), (2.4.29) with (2.4.31) and (2.4.34) and
(2.4.35) we get the discrete linearized system of equations to solve for the increment U
h

and
e
, for more details see Klinkel and Wagner [KLI97]:

)
`

=
)
`

enh
ext
e
h
e e
u
e
u
e
uu
f
f f

U
k k
k k
int

(2.4.36)
The algorithm for solving the system (2.4.36) is listed in Chapter3 - Figure 3.4.1.
Since E
enh
is not required to enhance inter-element continuity, we could eliminate

e
at the element level before assembling the element matrices to the global matrices.
From (2.4.36), we withdraw the formula for
e
:
[ ] ( )
h e
u
e
enh
e e
U k k f + =


1
(2.4.37)
30
Introducing (2.4.37) into (2.4.36) we finally get the condensed element stiffness
matrix
) (e
T
k and element residual force vector r
e
:
[ ] [ ]
e
u
e
T
e
u
e
uu
e
T
k k k k k
1
= (2.4.38)
[ ] [ ]
e
enh
e
T
e
u
e
int
e
ext
e
f f f k k r
1
+ =

(2.4.39)
Assembling element matrices, the global system has the form:
R U K =
h
T
(2.4.40)
After condensing, the global system (2.4.40) has the similar form as the global
system of the standard displacement FEM.

2.4.3 EAS parameters
In this section we investigate the modified strain
( )
mod com enh
= +
u
E E E under the
framework of the EAS approach. The enhancing strain field for an element in the
Cartesian coordinate system
enh
E is usually assumed, see [KLI97], by

enh T enh

E T
J
J
E

=
0
0

(2.4.41)
where:
- J is the Jacobian matrix,

(
(
(

=
3 3 3
2 2 2
1 1 1
, , ,
, , ,
, , ,



Z Y X
Z Y X
Z Y X
J

(2.4.42)
-
0
J is the Jacobian matrix at the center of the element ) ( 0 = ;
- T is the transformation matrix that maps quantities in the physical space to the
natural space:

(
(
(
(
(
(
(
(

+ + +
+ + +
+ + +
=
) ( ) ( ) (
) ( ) ( ) (
) ( ) ( ) (
2 2 2
2 2 2
2 2 2

32 23 33 22 32 13 33 12 22 13 23 12 33 32 23 22 13 12
31 23 33 21 31 13 33 11 21 13 23 11 33 31 23 21 13 11
31 22 32 21 31 12 32 11 21 12 22 11 32 31 22 21 12 11
33 23 33 13 23 13
2
33
2
23
2
13
32 22 32 12 22 12
2
32
2
22
2
12
31 21 31 11 21 11
2
31
2
21
2
11
J J J J J J J J J J J J J J J J J J
J J J J J J J J J J J J J J J J J J
J J J J J J J J J J J J J J J J J J
J J J J J J J J J
J J J J J J J J J
J J J J J J J J J
T (2.4.43)
with ( , )
ij
J J i j .
-
0
T is natural-physical transformation matrix at the element center ) ( 0 = .
For enhancing strain interpolation in the natural coordinate system, the following
formulation is used:
M E
) (
=
enh

(2.4.44)
31
where is the vector of internal parameters. The dimension of is various, it depends
on the type of the free-locking EAS elements in volumetric, membrane and shear
responses. It is equal to the number of the additionally enhanced modes defined in Table
2.4.2 below. The natural - physical mapping is realized at elements center to obtain
unique values for the parameters
i
. Otherwise, values of the
i
parameters will vary
according to integration points. These internal strains fill in the available compatible
strain field to alleviate parasitic terms.
The additional factor J /
0
J in (2.4.41) is introduced to be able to enforce the
orthogonality condition (2.4.4). The justification of its use, as suggested by Taylor et al.
[TAY76], is based on the same considerations which led to the approximation of the
local-global transformation at the element center ) ( 0 = , i.e. using
0
T rather than the
transformation T . In fact, the factor J /
0
J relates the transformation of an infinitesimal
volume element to the element center as well, thus facilitating the enforcement of the
orthogonality condition for the constant stress and strain states, i.e. the patch test, see the
next section for a more detailed analyses.
After combination of (2.4.41) and (2.4.44), the enhancing strain field
enh
E in the
physical coordinate system is written as:
M T
J
J
E

) ( 0
0
= =
T enh
with
) ( 0
0

M T
J
J

T
=
(2.4.45)
The enhancing matrix M, according to Andelfinger and Ramm [AND93], is
defined:
54 50 45 40 35 30 25
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0
0 0
0 0
0 0 0
0 0 0
0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0
0 0 0
0 0 0
0 0
0 0
0 0
(
(
(
(
(
(
(
(

M

(2.4.46)
where
3 2 1
and ; .
In (2.4.46), non-zero terms in the three upper-rows are applied for enhancing the
additional modes of the normal strains; non-zero terms in the three lower-rows are
applied for enhancing the additional modes of the shear strains. For the sake of
comprehension, lets consider the compatible strain
11( )

in the natural space for the
eight-node solid element:

1 2 3 1 2 2 3 1 3 1 2 3
( ) 0 1 2 3 4 5 6 7
( ) 2 3 2 3
11( ) 1 4 6 7 1
com
u a a a a a a a a
u
a a a a

= + + + + + + +

= = + + +

(2.4.47)
When additionally enhanced by the non-zero terms in the first row of matrix M in
(2.4.46) the expression
11( )

will becomes a complete tri-linear polynomial as:

2 3 2 3 1 1 2 1 3 1 2 3
1 4 6 7 1 2 3 4
( ) ( )
mod com enh
11( ) 11( ) 11( )
E E
a a a a d d d d

= +
= + + + + + + +


(2.4.48)
32
where ( 1 4)
i
d i = are components of vector . The modified strains as (2.4.48) will help
the EAS element to satisfy the incompressible condition and pure bending condition
because there is no inconsistent term (see Section 2.1.1) in the enhanced strains.
Table 2.4.2: EAS elements
EAS
element
Additional modes
Detailed modes
Capability
EAS3v 3v + 0s
(only modes 2527 in (2.4.46) are
adopted to enhanced normal strains)
Free volumetric locking
EAS6s 0v + 6s
(modes 2833 in (2.4.46) are
adopted to enhanced shear strains)
Free shear locking
EAS9 3v + 6s
(2527) + (2833)
Free volumetric & shear locking
EAS12 6v + 6s
(2527;40,43,44) + (2833)
Improved incompressibility behavior
with respect to EAS9
EAS15 3v + 12s
(2527) + (2839)
Improved bending behavior with
respect to EAS9
EAS21 9v + 12s
(2527;4045) + (2839)
Totally free volumetric locking &
free shear locking
EAS24 9v + 15s
(2527;4045) + (2839;4648)
Totally free shear & volumetric
locking
EAS30 12v + 18s
(2527;4045;4951) +
(2839;4648;5254)
Totally free shear & volumetric
locking, applied for distorted mesh

The enhancing matrix M in (2.4.46), when introduced in the expression (2.4.45)
to calculate
enh
E , will make
enh
E satisfy the orthogonality condition (2.4.4). In other
words, this matrix M is designed in such a way that the EAS elements are locking free
while pass the patch test, see the next section for a detailed expression.
Consequently, as chosen in (2.4.46) the matrix M expands the compatible strain
field up to the complete tri-linear field (EAS30). It means the enhanced element has 30
additional modes and 54 modes in total (24 compatible modes and 30 enhanced modes).
However, the number of additional modes should be suitable to each problem so as to
limit the calculation time. The list in Table 2.4.2 gives some suggestions for reducing
additional mode elements.
Above, a formulation of the EAS elements in the Green-Lagrange strains has been
presented. This formulation was implemented in a MATLAB code. The numerical results
at the end of this chapter show performances of the EAS elements and also assure quality
of the implementation.
2.4.4 Patch test
The arguments in this section are valid for both the incompatible method (Section
2.3) and the EAS method. For
0
C elements, such as the eight-node hexahedral, the
33
method for deriving the incompatible shape functions

N (Section 2.3) for the

incompatible elements (or enhancing matrix M (2.4.46) for the EAS elements) is so
detail. Not only filling higher order terms in the standard displacement field but the
incompatible shape functions also have to satisfy the patch test requirement. That is the
necessary condition for an element to converge to the correct solution.
Designate the space of the enhanced strain field as
enh

and the space of

admissible strain field defined in the standard fashion as
com

. The enhanced strain

interpolation and the compatible strain interpolation are independent in a sense as:
} 0 { =
com enh



(2.4.49)
Consider the enhanced stiffness matrix (2.4.35) and use in (2.4.45), we have:

0 0
2
0 4 1 4
( ) 0 0 ( ) 2
( ) ( )
e e
T
e T
V V
dV dV

= =


J
k C M T C T M
J

(2.4.50)
In order to consider positive definite of
e

k , the constant values (

2
1
0 0
,

J T ) and
positive values (
2
J ) can be eliminated from the expression. In addition, rows of M, see
(2.4.46), have been assumed to be linearly independent. Since the constitutive matrix C
4

is positive definite, the assumption of linearly independent rows of M assure the
enhanced stiffness matrix dV
T
V
e
e
C k
4
0

(in fact, the integral

0
4
( ) ( )
e
V
dV

M C M ) also
positive definite.
Lets consider rigid body motions or constant strain conditions. In nonlinear
problems, we consider the increment quantities. Hence, the constant strain condition is
const = E . Denote
h
0
U as the set of nodal displacements, which corresponds to one of
the rigid body cases or a constant strain state. Then we also have const = =
h
0
E B U .
Denote
e
0
as the values of internal variables in the case of the motion
h
0
U . In order to
pass the patch test, the EAS element requires
e
to be zero whenever
h
U corresponds
to rigid body motions or constant strain conditions. Since the matrix
e

k is always
positive definite, the condition for
e
= 0 from the second set of equations in (2.4.36)
reduces to:

0 f U k = +
e
enh
h
0
e
u 0 ,
] [

0 S U B C = +

dV dV
e
V
T h
T
V
e
0
mod
0 0
4

0

(2.4.51)
Pay attention that the term (
h
0
4
U B C ) is in fact some constant stress state
mod
0
S
that is correspond to rigid body motions or constant strain conditions. Consequently, the
requirement for the patch test to be satisfied reduces to:

0
e
V
dV =

0
(2.4.52)
Calculation of the left hand side expression will be performed in the natural space
(
1 2 3
, , ) where the presence of the Jacobian (2.4.42) is evident as:

1 1 1
1 2 3
1 1 1
0
e
V
dV d d d

=

J
(2.4.53)
34
This will generally lead to non-zero value of dV
e
V

0
and hence the element will
not pass the patch test except when the element is a parallelepiped. In this latter case, the
Jacobian will consist of constants and dV
e
V

0
will be equal to zero and the patch test will
be passed. As a remedy, it was proposed (Taylor et al., [TAY76]) to replace J by the
constant values computed at the origin ( = 0) of natural coordinates as:

1 1 1 1 1 1
0 1 2 3 1 2 3
0 ( )
1 1 1 1 1 1
0
1 1 1 1 1 1
1 2 3 1 2 3
0 0 ( ) 0 0 ( )
1 1 1 1 1 1
0
T
e
V
T T
dV d d d d d d
d d d d d d

= =
= = =

J
J T M J
J
J T M J T M

(2.4.54)
In fact, the orthogonality condition (2.4.4) did imply expression (2.4.52) because:


= =
0 0
V V
S S E 0 ) ( : dV dV
enh

(2.4.55)
We see that expression (2.4.55) is the strong form of the patch test condition. The
condition in (2.4.55) is valid for an arbitrary stress field while the patch test condition is
only valid for a constant stress field
mod
0
S .
The use of
0
J to approximate J is equivalent to the introduction of a geometric
approximation by replacing the original hexahedral (for 2D: quadrilateral) element into a
parallelepiped of the same volume (for 2D: parallelogram of the same area).

2.4.5 Equivalence between EAS and Hellinger-Reissner elements
A so-called hybrid stress method can be derived from the Hellinger-Reissner (HR)
principle which consists of the stress field and the displacement field. Before comparing
with the EAS elements the formulation of HR element is briefly introduced. From the
Hu-Washizu principle in (2.4.1) one obtains the conventional stress-displacement HR
functional by eliminating S C E E
1 4
) (

= =
S
. The result is:

( )
0
0
4 1
int ( )
*
* *
1
[ ( ) ]
2
( )
u
com
ext ,
V
A A V
dV
dA dA dV

= +

u u S
+ = S C S + SE
u u t u t u b


(2.4.56)
The HR elements with compatible displacement and assumed stress fields that are
built as follow:

( ) ( )
h e
U N u =

(2.4.57)

( )
h h e
H
P P T S = =
0
(2.4.58)
where:
-
e
H
S is assumed stress at element level;
35
-
( )
P is the matrix that enhances the stress field;
-
( )
P T P =
0
with
0
T is defined in (2.4.43);
-
h
is vector of internal stress variables.
The displacement trial functions
( )
N should be compatible (
0
C continuous)
across inter-element boundaries because there are first derivatives of the displacement
field in the HR-functional. But stress trial functions P are not subject to derivation so
could be chosen to be incompatible (
1
C continuous). This help for eliminating the
internal stress variables
h
at the element level more easily.
For the sake of simplicity, the following abbreviations are defined:
dV
e
V
T
) ( ) (
0
1 4

=

P C P H
(2.4.59)
dV
e
V
T

=
0
) ( B P G
(2.4.60)
The element stiffness matrix of the HR elements is derived from HR principle has
the form as:
G H G k
e
=
1
) ( ) (
T

(2.4.61)
The (2.4.61) is achieved after eliminating stress parameters
h
from the system of
equations. Then, the element stiffness matrix (2.4.61) can be used for the standard
displacement formulation.
After the solution process, the stresses can be obtained at the element level as:
d G H P S =
1
) (
e
H
(2.4.62)
Simo and Rifai [SIM90] generalized the incompatible displacement method to the
EAS method. Both of these methods are the dual ones of the hybrid stress method.
Hence, the EAS and the incompatible displacement methods with respect to the hybrid
stress method are correspondent in some special cases. In this section some relationships
between these methods are introduced. These relationships provide a helpful theoretical
basis for development and exploitation of the incompatible and EAS methods and also
the hybrid method.
It was proven by inspection [AND93] that the stiffness matrix of the EAS
elements is equivalent to the stiffness matrix of HR elements if the polynomials in
( )
M
(2.4.46) and
( )
P (2.4.58) are complementary. It means a polynomial term used for a
strain component in
( )
M is not considered for the corresponding stress component in
( )
P .
Consequently, the hybrid stress and enhanced assumed strain fields are orthogonal to
each other as:

1 1 1
1 2 3
( ) ( )
1 1
( ) 0
T
d d d



=

-1
P M
or ( ) 0
e T enh
H
dV =

e
0
V
S E
(2.4.63)
36
In Table 2.4.3, a rectangle low-order element with 22 integration will provide an
equivalent stiffness matrix whether assumed by a strain field (EAS method) or by a stress
field (hybrid stress method).
Table 2.4.3: Equivalent bilinear EAS - HR elements
EAS element HR element
EAS7 :
( )
(
(
(

=
2 1 2 1
2 1 2
2 1 1
0 0 0 0
0 0 0 0 0
0 0 0 0 0

M

PS :

( )
(
(
(

=
0 0 1 0 0
0 0 1 0
0 0 0 1
1
2

P
EAS4:

( )
(
(
(

=
2 1
2
1
0 0
0 0 0
0 0 0

M
HR8:

( )
(
(
(

=
2 1
2 1
2 1
1
2
0 0
0 0
0 0

0 0 1 0 0
0 0 1 0
0 0 0 1

P
EAS0:

( )
[ ] 0 =

M
HR12:
( )
(
(
(

=
2 1 2 1
2 1
2 1
2 1
2 1
1
2
0
0 0 0
0 0 0 0

0 0
0 0
0 0

0 0 1 0 0
0 0 1 0
0 0 0 1

P



2.5 ANS ELEMENT

Application of the EAS method that is described in the preceding sections to shear
deformable elements does not work satisfactorily in all situations. Particularly in the case
of severely distorted meshes or too thin structures these elements do not perform well.
The method that is most widely used in these situations is the ANS method. In this
context, we focus on the problem of transverse shear locking in solid elements, although
the same have been done in a similar way for shell elements. There are also a number of
publications on application of the ANS concept to overcome volumetric locking, shear
locking and membrane locking. However, the ANS method is only effective in transverse
shear locking removal [BIS97].
The ANS abbreviation means Assumed Natural Strain. Here, strain components
are assumed in the natural (isoparametric) space. The principal idea of the ANS method
is to choose a certain interpolation for the transverse shear strains instead of deriving the
strains directly from the interpolation of the displacements. Hence, the method also
named mixed interpolation, it means both the interpolation for displacement field and
interpolation for strain field are required by the method. The bilinear ANS plate element
is the most widely used element in both scientific and commercial finite element
packages (e.g. ADINA, ANSYS). It is also known as MITC4 element (MITC = Mixed
Interpolation of Tensorial Components) or Dvorkin- Bathe element [BAT96].
37
The natural strains at an interior point of the MITC4 element are obtained by
linear interpolations of the strains on boundary lines. Finally, the physical strains, which
are required at numerical integration points for evaluating the element stiffness and
internal force arrays are obtained by tensorial transformation of the natural strain
components instead of the standard isoparametric derivative transformations. The
procedure of the natural strains and transforming tensorially to physical coordinates has
been found to play a key role in improving element performance when the mesh is
distorted or curved.
The ANS method can also be combined with the EAS method to improve the in-
plane bending behavior. For an explicit definition of an ANS element one has to specify
two things, namely:
- The number of nodes and the corresponding shape functions for the displacements
(which are the standard shape functions of a displacement element) and
- The number and location of the sampling points and the corresponding shape
functions.
The ANS elements do not contain any spurious zero energy mode and show good
convergence behavior [PAR86]. All the ANS elements can be used in linear analysis, and
in large displacement and large strain analyses, e.g. in the simulations of structural
problems and collapse of shells.
The ANS was conceived as one of several competing methods with which to solve
shear locking problems. Its most noteworthy feature is that, unlike many forms of
reduced integration elements, it produces no rank deficiency. Furthermore, it is easily
extendible to geometrically nonlinear problems.

2.5.1 Kinematics in natural coordinate system
The ANS method requires the interpolation of all assumed strains in natural
coordinate system. Therefore, it is necessary to define a convected description, which
naturally preserves the objectivity (in the convected description, the material base vectors
reflect the geometrical and kinematic aspects, hence, the corresponding components are
indifferent with respect to their material base vectors). To this end, lets denote the
position vectors of the reference configuration
0
and the current configuration
t
in
the local coordinate system by
( )
X and
( )
x , respectively. The convected basis vector
i
G and its components
ij
G in the initial basis system are defined by:
3 1 ; ; / = = = i G
ij j i
i
i
G G X G
(2.5.1)
while the contravariant vector
j
G and its components
ji
G are defined following:
3 1 ; ;
-1
= = = = i,j G G
i ji i
ji j j
i
j
i
G G G G G
(2.5.2)
Similarly, the convected basis vector
i
g and its components
ij
g in the current
basis system are defined:
3 1 ; ; / / = = + = = i g
ij j i
i
i
i
i
g g u G x g
(2.5.3)
38
and its contravariant vector
j
g and its components
ij
g are defined through the following
expressions:
3 1 ; ;
-1
= = = = i,j g g
j ij j
ij i j
i
j
i
g g g g g
(2.5.4)
The deformation gradient in the form of the convected vectors is calculated as:

( )
( )
( )
i
i
G g
X
x
F

=
(2.5.5)
while the Green-Lagrange strain tensor takes the following form

( ) ( ) ( )
j i
ij
j i
ij ij
T
E G g G G G G I F F E

= = =
2
1
) (
2
1
) (
2
1
(2.5.6)
or alternatively:

( )
3 1
2
1
=
|
|

\
|

= i,j
j i
j i
j
i j
i
G G
u u
G
u u
G E

(2.5.7)
In the context of large deformation, the ANS method modifies shear components
of Green-Lagrange strain tensor E . Hence, the variational equation should be written in
material configuration (or total Lagrange formulation) in terms of the Green-Lagrange
strain tensor E and its energy conjugated quantity S , the second Piola-Kirchhoff stress
tensor.


2.5.2 Classical ANS formulation
In this section we present the classical ANS technique for the removal of
transverse shear locking in an eight-node hexahedral element. For the sake of simplicity,
lets consider the case with a rectangular prismatic geometric configuration. It means the
physical space (X,Y,Z) is chosen to be identical to the natural space ) , , (
3 2 1
, see
(Figure 2.5.1).
Instead of the standard computation, which leads to shear locking, the transverse
shear strains
13
E and
23
E are assumed, according to Dvorkin and Bathe [DVO84], to be
interpolated through the use of certain sampling points as follows

) ( 23
1
) ( 23
1
23
) ( 13
2
) ( 13
2
13
) 1 (
2
1
) 1 (
2
1 ~ ~
) 1 (
2
1
) 1 (
2
1 ~ ~
B D YZ
C A XZ
E E E E
E E E E


+ + =
+ + =

(2.5.8)
where
) ( 23 ) ( 13 ) ( 13
, ,
B C A
E E E and
) ( 23 D
E are the natural shear strains at points A, C, B and
D situated on the mid-surface of the solid element (Figure 2.5.1), respectively. Values of
these sampling strains can be directly derived through the use of the covariant
components in the contravariant base vectors as by E
ij
in (2.5.6):
39

11 11 11
22 22 22
33 33 33
12 12 12
13 13 13
23 23 23
0.5( )
0.5( )
0.5( )

2 ( )
2 ( )
2 ( )
e e
e e
e e
e e
e e
e e
E g G
E g G
E g G
E g G
E g G
E g G

= =
` `

) )
E

(2.5.9)
Once the transverse shear strains are assumed, all assumed strain-displacement
matrices can be immediately formulated as:
]
~
,
~
,
~
,
~
,
~
,
~
,
~
,
~
[
~
8 7 6 5 4 3 2 1
B B B B B B B B B =

(
(
(
(
(
(
(
(
(
(
(

+ + + +
+ + + +
+ =
) ( 2 3 , 3 2 ,
1
) ( 2 3 , 3 2 ,
1
) ( 1 3 , 3 1 ,
2
) ( 1 3 , 3 1 ,
2
1 2 , 2 1 ,
3 3 ,
2 2 ,
1 1 ,
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1




~
B
T
I
T
I D
T
I
T
I
C
T
I
T
I A
T
I
T
I
T
I
T
I
T
I
T
I
T
I
I
N N N N
N N N N
N N
N
N
N
g g g g
g g g g
g g
g
g
g
B


; I=1-8
(2.5.10)
Then the stiffness matrix is formulated as in the standard manner. Expressions
(2.5.8-10) are only valid when the physical space is identical to the natural space, i.e.,
(X,Y,Z)
1 2 3
( , , ) .
For the general case, where the mid-surface quadrilateral is not a rectangle and the
X-Y frame is not aligned to the
1 2
frame, the natural shear strain components must
be firstly interpolated in the covariant space, as defined in (2.5.11). This allows taking
into account the element distortion.

1

Figure 2.5.2: Mid-surface of element in
isoparametric coordinates
Figure 2.5.1: ANS method illustration,
special case: X-Y aligns
1
-
2

0
(-1,-1) (1,-1)
(1,1)
(-1,1)
A
B
C
D
E F
H G G
C
B
F A
E
D
H
E
23
E
23(D)
E
13(C)
E
13(A)
O
~
E
23(B)
E
13
~
1
2
3
4
5
6
7
8
Y,
2

X,
1

Z,
3

40
From (2.5.8) and (2.5.9), the assumed shear strains in the general case can be
computed by:

+ +
+ +
=

+ +
+ +
=

=
) ( 23 23
1
) ( 23 23
1
) ( 13 13
2
) ( 13 13
2
) ( 23
1
) ( 23
1
) ( 13
2
) ( 13
2
23
13
) )( 1 ( ) )( 1 (
) )( 1 ( ) )( 1 (

) 1 ( ) 1 (
) 1 ( ) 1 (

~
2
~
2
~
B
e e
D
e e
C
e e
A
e e
B D
C A
s
G g G g
G g G g
E E
E E
E
E




E

(2.5.11)
where
) ( 23 ) ( 13 ) ( 13
, ,
D C A
E E E and
) ( 23 B
E are natural shear strains, evaluated by
displacement interpolations, at points A, C, D and B respectively, see Figure 2.5.1.
Clearly,
13
~
E is constant with respect to
1
and discontinuous at 1
1
= (between
elements), while
23
~
E is constant with
2
and discontinuous at 1
2
= .
The strain tensor can be equivalently expressed in both the natural space and the
physical (Cartesian) space. The transformation of the strain components between the
natural space and the physical space is done by using the transforming matrix
) ( 0
T as
defined in (2.4.43). In order to alleviate the shear locking, the natural shear strains
13
E
and
23
E are replaced by the assumed natural shear strains
13
~
E and
23
~
E before doing the
transformation from the natural space to the physical space. Finally, the physical assumed
strain vector is:


) )( 1 ( ) )( 1 (
) )( 1 ( ) )( 1 (
) (
) ( 5 . 0
) ( 5 . 0
) ( 5 . 0
~
2
~
2
2
~
2
~
2
2
~
) ( 23 23
1
) ( 23 23
1
) ( 13 13
2
) ( 13 13
2
12 12
33 33
22 22
11 11
23
13
12
33
22
11

+ +
+ +

=

B
e e
D
e e
C
e e
A
e e
e e
e e
e e
e e
T T
YZ
XZ
XY
ZZ
YY
XX
ANS
G g G g
G g G g
G g
G g
G g
G g
E
E
E
E
E
E
E
E
E
E
E
E


T T E

(2.5.12)
These assumed strains can be implemented in the standard solid element in a
straightforward manner. These assumptions allow the element to represent pure bending
modes without any spurious shear effect.
The modified shear strains lead to the new operator matrix

B (2.5.10) in the
natural space. However, the formulation can be slightly modified so that no explicit strain
evaluation at the sampling points is necessary in the numerical calculation. The physical
assumed strain-displacement matrix at node I of the eight-node solid element is:
41

(
(
(
(
(
(
(
(
(
(
(

+ + + +
+ + + +
+ =

) ( 2 3 , 3 2 ,
1
) ( 2 3 , 3 2 ,
1
) ( 1 3 , 3 1 ,
2
) ( 1 3 , 3 1 ,
2
1 2 , 2 1 ,
3 3 ,
2 2 ,
1 1 ,
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1




~
B
T
I
T
I D
T
I
T
I
C
T
I
T
I A
T
I
T
I
T
I
T
I
T
I
T
I
T
I
T ANS
I
N N N N
N N N N
N N
N
N
N
g g g g
g g g g
g g
g
g
g
T B



(2.5.13)

2.5.3 Some variational basis for the ANS method
The ANS element is locking-free, rank sufficient and distortion insensitive even
with coarse meshes as it has been pointed out by Park and Stanley [PAR86]. The ANS
method has originally been derived from engineering intuition without a convincing
variational background. Firstly, a restricted form of the method was proposed in 1969 for
four-node plane stress element by assuming a constant shear strain that is independent to
the direct strains [MIL90a]. In 1981 Hughes and Tezduyar [HUG81] used the method to
avoid shear locking for plates; later in 1984 it was applied successfully by Dvorkin and
Bathe to four-node shell element for geometric and material nonlinear analysis [DVO84].
In fact, the mathematical justifications, based on the Hu-Washizu and mixed functionals,
have been provided a couple of years later in separate publications, e.g. Militello and
Felippa [MIL90a,b].
In ANS method, there are two strain fields: the derived-displacement strain field
and the assumed natural strain field. If we consider these fields as independent fields, the
ANS formulation can be interpreted by a Reissner type functional: the functional that
uses the strains and displacements as independent fields. Departure from the three fields
general Hu-Washizu functional (2.4.1) the displacements u, stresses S and strains E are
independently varied, lets re-write the functional (2.4.1):

( )
dV dA dA
dV dV W
V A A
com
V V
ext , ,
u

+
+
+ + = =
0
0 0
*
*
*
) ( ) ( int
) (
] [
b u t u t u u
E E S
u E S S E u


(2.5.14)
From
( ) S E u , ,
one obtains the conventional stress-displacement Heillinger-Reissner
functional by eliminating E by:
S C E E
1 4
) (

= =
S
(2.5.15)
Another Reissner type, strain-displacement functional is obtained by eliminating
S through:
CE S S
4
= =
E
(2.5.16)
which yields:
42
( )
dV dA dA dV
V A A V
com T T
,
u

+ + =
0 0
*
*
*
) (
) ( ]
2
1
[ b u t u t u u CE E CE E
u
4 4
E u
(2.5.17)
Setting
com
) (u
E E = and
*
u u = on
u
A reduces
( ) E u,
to the potential energy functional

( )
dV dA dV
V A V
com T com

== + =
0 0
*
*
) ( ) ( int
) (
2
1
b u t u CE E
u
4
u ext u

(2.5.18)

Partial Strain Assumption
It is common practice to assume only a part of the strains to be independent fields.
For instance, with MITC4 element independent assumptions are only made for the
transverse shear strains [BAT96], whereas the bending strains are entirely derived from
displacements:
[ ]
T
b a
E E E = (2.5.19)
where
a
E stands for the assumed strain field and
com
b b
E E = stands for the derived-
displacement (bending) strain field.
The
( ) E u,
functional (2.5.17) requires obvious modification in the volume term:

( )
dV dA dA
dV
V A A
V
,
u

+
+
(

=
0
0
*
*
*
) (

5 . 0
5 . 0
] [
b u t u t u u
E
E E
C C
C C
E E
b
a
com
a
bb
4
ba
4
ab
4
aa
4
T
b
T
a E u
a

(2.5.20)
The resulting principles take a particularly simple form if the constitutive coupling
terms
ab
C and
ba
C vanish, in that case:

( ) ( ) ( ) ext b , a ,
+ + =
u E u E u
a a

(2.5.21)
where
( )
a
E u, a
is a mixed strain-displacement energy involving
a
E ;
( ) u b
is a potential
energy involving the
com
b
E ;
ext
is the external energy.
Up to now a compatible displacement field and a discontinuous strain field are
involved. Hence use
( )
a
E u,
is a suitable functional for the ANS method.
The element displacement field is interpolated as:

h
c
e
U N u
c
= (2.5.22)
where
c
N is the compatible shape functions as defined in (2.4.11),
h
c
U is the nodal
displacement vector as defined in (2.4.14).
The strain fields derived from the displacements are:
- bending strains:
=
com c h
b b c
E B U (2.5.23)
- shear strains:
=
com c h
s s c
E B U (2.5.24)
43
where
c
b
B and
c
s
B are parts of strain-displacement matrix that relate to bending strains
and shear strains, respectively (the letter c stands for compatible values).
The independent strains in
( )
a
E u,
are:
- bending strains as the derived-displacement bending strains (2.5.23);
- shear strains (the superscript a stands for ANS values):

a
a s
= E B a (2.5.25)
with
a
s
B is natural assumed strain-displacement matrix and a is strain coefficient vector.
Introduce (2.5.23) - (2.5.25) into (2.5.21) and carrying out the interpolations at the
element level we will have:

( )
( ) ( ) ( )
1 1
2 2
,
= +
h
c
T T
T
h cc h T aa ca h c
c c c
U a
U K U a K a Uh K a - U f (2.5.26)
where

( ) ( )
( ) ( ) ( ) dA dV dV
dV dV
e e e
e e
A V V
s
V
s
V
b


+ = =
= =

* t N * b N f B C B K
B C B K B C B K
T
c
T
c
c a
s
4
T
c
s
ca
a
s
4
T
a
s
aa c
b
4
T
c
b
cc
;
;

(2.5.27)
On performing the variations we obtain the matrix equation:

)
`

=
)
`

0
f
a
U
K - K
K K
h
c
c
aa T ca
ca cc
b
) (

(2.5.28)
From the second equation of (2.5.28) we obtained the shear strain coefficients:

h
c
h
c
U Q U K K a
c
T ca aa
= =

) ( ) (
1

(2.5.29)
Introduce (2.5.29) into (2.5.28) gives the statically condensed system:

c
c
aa T
c
cc
b
f U Q K Q K
h
c
= + ) (
(2.5.30)
In (2.5.30)
cc
b
K is the bending stiffness matrix, which is also obtainable from the
potential energy principle. While
c
aa T
c
Q K Q stands for the new shear stiffness matrix. The
system of equations (2.5.30) now contains only nodal displacement vector
h
c
U as in the
standard displacement method.
A variational justification of the ANS formulation as presented above has been
done by Militello and Felippa [MIL90a,b]. This study is based on two hybrid extensions
of Reissner-type functional that uses strains and displacements as independent fields.
However, the work of Militello and Felippa is not applicable to all types of material
models. The material is firstly required to be decoupled as in (2.5.20). Furthermore, the
proposal of Militello and Felippa, currently, is only valid to transverse shear locking
removal. Meanwhile the ANS method could apply for another locking effect, such as
curvature thickness locking (Chapter 3). By these arguments, we see that variational base
of the ANS method is still an open problem. However, as it has been pointed out
[BAT96] that a variational basis of an element might not exist, but whether the element is
44
useful and effective can of course be determined only by a deeper analysis of the
formulation.
Advantages of the ANS method are simplicity while remaining effective. The
ANS elements could be applied for both structural (plate, shell) and continuum (solid)
elements. Numerical results in literature show that ANS elements are locking-free, rank
sufficient and distortion insensitive even with coarse meshes. Furthermore, the method is
easy to be implemented in any code. Because of its attractions, the ANS method has been
being developed by many authors as Dvorkin and Bathe [DVO84], Park and Stanley
[PAR86], Betsch and Stein [BET95], Bathe et al. [BAT00], etc.


2.6 NUMERICAL RESULTS
This section investigates performances of the EAS and ANS elements. In the
following tests, the low-order standard solid element is designated as Q1. While the
standard solid element Q1, which employs the classical ANS technique [DVO84] for
alleviating transverse shear locking and curvature locking, is designated as ANS. The
EAS elements are designated as EASx, where x is the number of internal parameters.
The additional letter 2D. stands for elements in 2D, without this additional letter means
elements are in 3D. The ANS, EAS elements used in the following tests are implemented
in a MATLAB code, according to the theories presented in this chapter.

2.6.1 Membrane patch test
Lets consider a patch test as suggested by McNeal and Harder [MAC85] and
originally aimed to check the membrane behavior of plate and shell elements. In order to
adapt to 3D elements, the number of nodes has been doubled (Figure 2.6.1) as Vu-Quoc
and Tan [QUO03a]. An imposed displacement field at the boundary nodes is chosen to
cause a constant stress field in the plate. ,

Table 2.6.1: Interior nodal coordinates
5 6 7 8 13 14 15 16
X 0.04 0.18 0.16 0.08 0.04 0.18 0.16 0.08
Y 0.02 0.03 0.08 0.08 0.02 0.03 0.08 0.08
Z -h/2 -h/2 -h/2 -h/2 h/2 h/2 h/2 h/2

Practically, consider a rectangle plate of dimensions LWh = 0.240.120.001.
The material parameters are taken as E = 10
6
and = 0.25. In the original problem
designed for plane stress problems, McNeal and Harder [MAC85] employed the
following boundary conditions:

3
10 ) 2 / (

+ = Y X u ;
3
10 ) 2 / ( v

+ = Y X (2.6.1)
which lead to the corresponding reference solutions of constant strains and stresses:
45

3
10

= = =
XY Y X
;

X
=
Y
= 1333 ;
XY
= 400.
(2.6.2)

















Motivated by this result, the above boundary conditions (2.6.1) are also adopted
for the modified doubled-surface membrane patch test. Additionally, the condition w = 0
has also be imposed to a bottom node, e.g. node 1, to prevent rigid body motions.
Because the plate is too thin, L/h=240, it is possible to assure a constant stress state in the
plate when apply these conditions.
Numerical results show that the ANS element passes the modified patch test; i.e.
the computed stresses are constant all over the plate and consistent with (2.6.2). The
computed displacements of interior nodes also fully agree with (2.6.1). We also note that
the EAS elements also pass this patch test as previously reported in [KLI06], [QUO03a].

2.6.2 Out-of-plane bending patch test
Lets re-consider the above plate, Figure 2.6.1, but in bending situation. Again, a
patch test for 2D elements (plates and shells) is extended for 3D elements. In order to
create a constant stress state, the original boundary conditions, see [MAC85], for the
displacement ) (
0
w and rotations ) , (
y x
at the reference surface of the plate are:

3 2 2
0
3 3
10 ( ) / 2;
10 ( / 2) ; 10 ( / 2).
X Y
w X XY Y
w w
Y X X Y
Y X

= + +

= = + = = +


(2.6.3)
The boundary conditions (2.6.3) are not directly applicable to the solid elements
because they contain the rotations. If the cross sections of the plate are assumed plane (it
is plausible because the plate is thin) the boundary conditions can be modified, as in
[QUO03a], in such a way that:
; 2 / ) ( 10
2 2 3
0
Y XY X w w + + = =

(2.6.4)
Figure 2.6.1: Membrane patch test
1
9
2
10
3
11
4
12
8
16
7
15
6
14
5
13
X
Z
Y
L
h

W
46

3
10 ) 2 / (
2 2

+ = = Y X
h h
u
Y
;
3
10 ) 2 / (
2 2
v

+ = = X Y
h h
X
;
i.e. different displacements are imposed to the exterior nodes on the top (upper sign) and
bottom (lower sign) surfaces of the plate, respectively. The theoretical stresses, see
[MAC85], at the top and bottom surfaces of the plate are:
667 . 0 = =
Y X
; 200 . 0 =
XY
(2.6.5)

Table 2.6.2: Displacements of the interior nodes
Node u v w
5
8
10 500 . 2


8
10 000 . 2


6
10 400 . 1


6
8
10 750 . 9


8
10 000 . 6


5
10 935 . 1


7
7
10 000 . 1


8
10 000 . 8


5
10 240 . 2


8
8
10 000 . 6


8
10 000 . 6


6
10 600 . 9



Table 2.6.3: Normalized displacements at interior nodes
Node Displacement ANS EAS3v6s EAS12v18s
5 u
v
w
1.0000
1.0000
0.9999
2.6009
-2.5450
3.0543
4.6332
1.1749
3.8885
6 u
v
w
1.0000
1.0000
0.9999
0.6086
-0.1536
1.1603
0.1436
1.4310
1.3021
7 u
v
w
1.0000
1.0000
1.0000
0.6719
1.9783
1.1643
0.1815
0.7567
1.2923
8 u
v
w
1.0000
1.0000
1.0000
1.5000
1.8627
1.4216
0.2291
0.7849
1.7288

As also already reported by Vu-Quoc and Tan [QUO03a], all EAS elements are
unable to converge to the analytic solution listed in Table 2.6.2. The displacements of the
EAS12v18s in Table 2.6.3 are identical to those reported by Vu-Quoc and Tan
[QUO03a]. This helps to evaluate the quality of our EAS implementation in the
MATLAB code.
In contrast, the classical ANS enables the exact solution. This presents the superior
performance of the ANS techniques over the EAS ones in the removal of shear locking
and in working with distorted mesh, see Table 2.6.3.

2.6.3 Eigenvalues analyses of a rectangle
This is one of the basic tests. The eigenvalues of stiffness matrices of regular
meshes are calculated. We examine a single element of rectangle shape with length is
equal to unit, material properties are Youngs modulus E = 1000 and Poissons ratio =
0.
In order to check whether the element is free from shear locking, eigenvalues of
pure bending mode of the stiffness matrix are analysed. As references, lets take free-
47
shear locking elements as 2D.EAS4 (see Table 2.4.3), based on the EAS method,
proposed by Simo and Rifai [SIM90] and DSG element proposed by Bischoff et al.
[BIS03] to compare with the 2D.ANS element.
















The Poissons ratio is set to zero, = 0, in order to preventing the element from
volumetric locking and Poisson thickness locking. The results show that with various
aspect ratios (length is fixed, thickness is changed) the 2D.ANS element displayed 3 rigid
body modes and no spurious zero energy modes. Figure 2.6.2 shows that behavior of the
present 2D.ANS solid is completely coincident with results given by 2D.EAS4 and DSG.
It means all of them are free from shear locking at high aspect ratio. Meanwhile the
standard element 2D.Q1 shows stiffer (locking) behavior when aspect ratio increasing.

2.6.4 Circular cantilever beam at large displacements
The following example shows the applicability for thin 3D-beams. Lets consider
a circular cantilever of dimension R = 100, = 45 and cross section 11, see Figure
2.6.3. Material parameters are elastic modulus E =
7
10 and Poisson ratio = 0. The
cantilever is clamped at one end and loaded by a force P at the other extremity.

















Z
Y
X
R
P
Figure 2.6.3: Circular cantilever
Figure 2.6.2: Eigenvalue analyses of element stiffness matrices,
in-plane bending mode
1.0E-01
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
1 10 100 1000
Aspect Ratio Lx/Ly
E
i
g
e
n
v
a
l
u
e

2D.ANS/2D.EAS4/2D.DSG
2D.Q1
Lx
Ly
48
The problem has been modeled by Slavkovic et al. [SLA94], who used 18 internal
strain parameters to enhance the solid element. The problem has also been modeled by
Klinkel et al. [KLI97], who also used EAS solid element. Both the enhanced elements of
authors in [SLA94] and [KLI97] are EAS elements but enhanced by different modes.
However, these elements are shear and volumetric locking free. The large deformation
response will be calculated for different vertical tip loads. In this test, behavior of the
classical ANS method is very attractive. Figure 2.6.4 shows that displacements of the
cantilever discretized by only 10 ANS elements are quite comparable to the results of
Slavkovic and Klinkel with 16 elements. The ANS element is softer in displacement u
(curve 5) but stiffer in displacement v (curve 6) while give a well approximated
displacement w (curve 7) to the references, [SLA94] and [KLI97]. This is the result of
shear locking removal in thickness direction (z).

























2.6.5 Scordelis-Lo roof with rigid end diaphragms
Consider a shell of radius R = 25, thickness t = 0.25, length L = 50 and open angle
= 40 under a gravity load p = 90 (per unit area) distributed on the shell surface (Figure
2.6.5). Both ends of the shell are constrained with only a free movement in the axial
direction. The material parameters are: E = 4.32
8
10 and = 0.0. The vertical deflection
of the mid-side free edge v = 0.3024 is taken as the reference solution (McNeal and
Harder [MAC85]).
Due to the symmetry of the structure, only a single quarter of the shell is modeled.
Different types of discretization are considered together with various elements (Figure
2.6.6). The ANS element delivers a good solution with a rather coarse mesh (44). In
contrast, the EAS9 element, which has 9 incompatible modes, requires a finer mesh
(1616) to reach the correct solution.

Figure 2.6.4: Curved cantilever beam: displacements
0
100
200
300
400
500
600
700
10 20
30 40 50 60
u (EAS)
v
w
w (Q1)
Klinkel; Slavkovic
(16 elements)
3D.ANS-u
3D.ANS-v
3D.ANS-w
1
2
3
4
5
6
7
1 2 3
4 5 6 7
Centre displacement w (mm)
A
p
p
l
i
e
d

f
o
r
c
e

P

(
N
)

(10 elements)
49


































With an increase of enhanced modes, the EAS15 element appears to give a better
result more theoretically expected than the EAS9 element. However, instead of starting
from a low value of displacement at coarse meshes and then progressively increasing this
value with the refinement of mesh as seen for the EAS9 element, the EAS15 element
gives first a higher value of deflection at coarse meshes and then lower values due to a
mesh refinement. This can be explained by the effects of the high-order incompatible
modes (modes 34-39, see (2.4.46)), they make the EAS15 soft by fairly alleviating the
shear locking. As regards the Q1 standard element, a rather slow convergence is found.
Obviously, the shear locking contributes to this behavior.
Figure 2.6.6: Scordelis-Lo roof: convergence of finite element solution
Element per side
N
o
r
m
a
l
i
z
e
d

d
e
f
l
e
c
t
i
o
n

1
2
3
4
1
3
2
4
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
2 4 6 8 10 12 14 16
ANS
EAS9
EAS15
Q1
Rigid diaphragm
Free edge
Z
Y
X
L
R
Figure 2.6.5: Scordelis-Lo roof
p
50
2.6.6 Regular block with nearly incompressible material
In order to investigate the performance of the ANS and EAS elements in
volumetric locking conditions, a regular block of dimensions 10010050 clamped at
bottom and loaded by a uniform pressure of q = 250/unit area, acting on a top area of
2020 at the center is considered [AND93], see Figure 2.6.7. The material has an elastic
modulus
5
10 1 . 2 = E and Poisson ratio = 0.4999, i.e. nearly incompressible material.
Due to symmetry, only a quadrant of the block is modeled by a mesh of 555 elements.
















In Table 2.6.4 the vertical displacement simulated by different elements is listed.
Take the result of the free-volumetric locking SRI element as the reference. Obviously,
the volumetric locking response is observed with the standard element. Since volumetric
locking cannot be removed by the ANS techniques, the ANS element is nearly as stiff as
the standard element Q1 in this problem.
Table 2.6.4: Vertical displacement at the blocks center
Element Q1 ANS EAS3v6s EAS9v EAS21 EAS30 SRI
w 0.00160 0.00161 0.01136 0.01844 0.01907 0.01907 0.01966
w/w
SRI
0.08140 0.08190 0.57780 0.93790 0.97000 0.97000 1.00000

Compare to the EAS30, we see that EAS3v6s with 3 volumetric modes is rather
stiff. The EAS21 element with 9 volumetric modes gives as good result as the EAS30.
Note that the shear modes can assist the volumetric modes in volumetric locking
removal. It explains why the locking response can be more removed with the additional
introduction of shear enhanced modes. For example, the EAS21 with 12 shear modes
besides 9 volumetric modes gives a quite better result than the EAS9v with unique 9
volumetric modes. The SRI technique shows a better performance in removing the
incompressible locking in comparison with the EAS technique.



Figure 2.6.7: Regular block
100
1
0
0
5
0
q
O x
y
z
51
CONCLUSION
Through the numerical tests we see that the transverse shear locking treatment is
ideally suited by the ANS method. In the case the Poissons ratio is different from zero,
the ANS method gives less accuracy but always remains too much better than the
standard displacement-based method (see Section 2.6.5, for example). The EAS method
is also useful to shear locking removal but computational cost of the EAS method is more
expensive than the ANS because the EAS method requires calculation of internal
variables. The volumetric locking is effectively removed by the EAS method. The SRI is
completely suitable for volumetric locking removal but this method cannot pass the patch
test (see Appendix in the next page) so it is not taken into consider in the thesis. However
the SRI elements are still used by many authors because of its simplicity.
There was a combination between methods of ANS with EAS proposed by
Andelfinger and Ramm [AND93] for four-node degenerated-shell elements. In order to
improve the element performance, Andelfinger and Ramm use the EAS method for the
membrane and bending components, while the transverse shear component is formulated
according to the ANS method. To alleviate the shear locking, the shear strains are
referred to natural coordinates. Next, the combination of Andelfinger and Ramm was
continually extended to geometrically nonlinear Reissner-Mindlin shell by Bischoff and
Ramm [BIS97]. Authors in [BIS97] did apply the ANS method for avoiding curvature
thickness locking by use of interpolation functions for transverse normal strains at nodal
points instead of an evaluation at the integration quadrature points. Other example of
EAS-ANS combination results in a simple shell model, built directly from basis of ANS
shell model of Dvorkin and Bathe, where the EAS techniques are integrated for
membrane locking, was introduced by Slavkovic et al. [SLA94], etc.
Other successful combination of the EAS concepts and ANS method is the
generating of the solid-shell elements. The solid-shell concept was built on modifying
assumptions of the standard shell theory (Reissner-Mindlin shell model). In fact, the
solid-shell elements form a class of finite element models which are intermediate
between thin shell and conventional solid elements. Detail for the solid-shell elements are
presented in the chapter following.









52
APPENDIX of Chapter 2
Free-volumetric locking element - SRI element

If nearly incompressible material behavior is use, the elements suffer from
volumetric locking. Using the SRI method is one of the best ways for eliminating
volumetric locking [PON95]. By the way, use the SRI method with reduced integration
for volumetric part and full integration for deviatoric part assures that all three normal
strains satisfy the volumetric constraint 0 =
ii
.
Consider the SRI method, starting from the potential energy functional (2.5.18):

ext
V
u
dV W + =

0
) (
) (
com
E S

(A.1)
The first variation of the functional is:
0 :
0
= +

ext
V
dV S E
(A.2)
We split the strain E and stress S of (A.2) additively into volumetric (dilatational)
parts and deviatoric parts. The volumetric parts are:

1
3
ii
E i 1- 3 = =
v
E I (A.3)

1
3
ii
S i 1- 3 = =
v
S I (A.4)
And the deviatoric parts are:
=
d v
E E E
(A.5)
=
d v
S S S
(A.6)
Rewriting the expression (A.2) with tensors decoupled into volumetric and
deviatoric parts, we have:
0 0 0 0
: : : : 0
ext
V V V V
dV dV dV dV + + + + =

d d v v d v v d
E S E S E S E S
(A.7)
Assume that:

0 0
: : 0
V V
dV dV = =

d v v d
E S E S
(A.8)
So the weak form (A.7) becomes:

0 0
: : 0
ext
V V
dV dV + + =

d d v v
E S E S
(A.9)
The second integrand of the expression (A.9) is calculated by the reduced
integration, but this is only applied for in-plane components, 11 integration instead of
22 while the integration in thickness direction remains unchanged. Restriction of the
SRI method is that the applied material laws must allow a decoupling stress field and
strain field into volumetric parts and deviatoric parts. It means the tangent moduli tensor
4
C could be also split into a volumetric part and a deviatoric part as:
53

4
C =
4
C
d
+
4
C
v
(A.10)
As mentioned above, the SRI is the best method to remove volumetric locking. Its
severe problem is cannot pass the patch test. In order to understand the problem, lets
consider a cubic patch test.
The patch test for convergence is a fascinating area in the development of
nonconforming finite element methods. It has been intuitionally proposed by Irons since
the mid-1960s. By the early 1970s the test had became a powerful and practical tool for
evaluating and checking nonconforming elements. We consider the following example: a
unit cube modeled by seven elements - distorted mesh [MAC85]. Material parameters
are: elastic modulus E = 10
6
and Poisson ratio = 0.25.




















Table A1: Location of inner nodes
Coordinates Nodes
X Y Z
1 0.249 0.342 0.192
2 0.826 0.288 0.288
3 0.850 0.649 0.263
4 0.273 0.750 0.230
5 0.320 0.186 0.643
6 0.677 0.305 0.683
7 0.788 0.693 0.644
8 0.165 0.745 0.702

The outer nodes subject to following conditions:
u = 10
-3
(2X+Y+Z)/2
v = 10
-3
(X+2Y+Z)/2
w = 10
-3
(X+Y+2Z)/2
(A.11)
X
Z
Y
1
4
8
5
6
2
3
7
Figure A1. Cubic patch test
54
Figure A2. Stress in patch elements (METAFOR [MET08])

z
is spurious with SRI
z
is smooth with EAS
1.93e+3 1.95e+3 1.98e+3 2.01e+3 2.04e+3 2.00e+3 2.00e+3 2.00e+3 2.00e+3 2.00e+3
These conditions assure a uniform strain in the cubic. Reference solution of the
problem is analytically derived as:

X
=
Y
=
Z
=
XY
=
YZ
=
ZX
= 10
-3

X
=
Y
=
Z
= 2000

XY
=
YZ
=
ZX
= 400
(A.12)
The 3D element patch test is used to verify whether volume elements can exactly
reproduce a constant strain state for any configuration. If this is the case, then the element
will converge to the analytically exact solution (assuming that the materials are elastic
and deformations are small), as the mesh is refined. Depending on the element type and
problem, however, convergence may be too slow for practical purposes. The values of
stress at the Gauss points obtained with the EAS9 are presented in Table A2.

Table A2: Results of EAS9

X
=
X vol.
+
X dev.
Element

X vol.

X dev.

XY

XZ

1 1998.5 0.95E-5 399.5 399.5
2 1998.5 0.25E-5 399.5 399.5
3 1998.5 0.35E-4 399.5 399.5
4 1998.5 -0.11E-4 399.5 399.5
5 1998.5 -0.35E-4 399.5 399.5
6 1998.5 -0.68E-4 399.5 399.5
7 1998.5 0.23E-5 399.5 399.5
Ref. 2000 400 400

A quite similar trend is observed with different schemes of enhanced modes
(EAS15, EAS21, etc). In contrast with the SRI elements where the hydrostatic stress
vol.
is spurious and hence these elements fail to pass the patch test, the EAS elements really
pass the patch test, see Figure A2.

















As stated by McNeal and Harder [MAC85], if an element produces correct results
for the patch test, the results of any problem solved with that element will converge
55
toward the correct solution as the elements are subdivided. Many authors supposed that
an element that does not pass the patch test should not be trusted. On the other hand,
passing the patch test does not guarantee satisfaction since the rate of convergence may
be too slow for practical use. However, in the thesis the SRI element is not furthermore
considered because employment of the SRI technique may be the cause makes the solid-
shell elements cannot pass the bending patch test, see [CAR07] and [REE07].









































56
Chapter 3. SOLID-SHELL ELEMENTS FOR
FINITE DEFORMATION


INTRODUCTION

For the analysis of nonlinear mechanical behavior of structures, low-order
elements are widely applied because of their efficiency and simple geometry. However,
the standard pure displacement elements usually exhibit severe stiffening effects known
as locking. Concretely, in thin-walled structures pure bending modes are spoiled by the
parasitic shear strains, shear locking occurs. For nearly incompressible materials and
incompressible conditions, volumetric locking occurs, when deviatoric modes come
along undesirably with volumetric strains. In this work, we attempt to develop a low-
order solid-shell element, which is free from all kinds of locking effects. Indeed,
transverse shear locking and curvature thickness locking can be circumvented by using
the ANS method, while membrane locking and volumetric locking can be removed by
the EAS method.
The solid-shell elements are well applicable for geometrically nonlinear problems
([HAU98], [QUO03a,b]) or for both geometrically and materially (elastoplastic)
nonlinear problems ([HAU00], [TAN05], [JET08]). In comparison with other shell
elements, the solid-shell elements enable an easy connection with other continuum
elements due to their solid topology, (Figure 1.0.1).
The solid-shell elements here refer to the finite element models which are
applicable to shell analyses and possess no rotational DOFs. They are different from the
degenerated shell elements in the sense that the latter elements are equipped with both
translational and rotational DOFs. The solid-shell elements have the characteristics of
solid elements where strains can be extended up to complete trilinear fields by internal
parameters [AND93] or can be naturally assumed. There are several advantages of the
solid-shell elements compared to the degenerated shell elements. First, the solid-shell
elements are simpler in their kinematic and geometric descriptions. Second, no special
effort is required for matching the translational and rotational DOFs when a structure
consists of both solid and thin-walled regions. The laborious task of defining algebraic
constraints or introducing solid-to-shell transition elements can be exempted. Third, the
complication on handling finite rotational increments can be avoided. Nevertheless,
formulating the robust solid-shell elements is indeed more demanding than formulating
the degenerated shell elements. However, the latter elements are only plagued by shear
and membrane locking effects while the former elements are also bothered by Poisson
thickness and trapezoidal (curvature) locking effects, see Section 3.1.
Starting from the principle of shear locking removal by the available ANS method
[DVO84] an ANS technique with an alternative scheme of sampling points, which can be
employed for the solid-shell elements, is investigated. In fact, several ways can be
employed for the interpolation of natural strains such as linear interpolations [DVO84],
quadratic interpolations [HAU01], [PAR86]. In Section 3.3 an alternative bilinear
interpolation is introduced.
57
Table 3.0.1: Dominant features of degenerated shell and solid-shell
Degenerated shell Solid-shell
The kinematic DOF are the components
of the displacement vector and of the
extensible director vector of the
reference surface.
Locking effects (Poisson thickness
locking), which occur if a 3D material
law is used along with constant normal
thickness strains, can be avoided.

The kinematic degrees of freedom are
the components of the displacement
vector.

The stresses are evaluated from a 3D
material law. This feature is especially
useful for complicated nonlinear
constitutive equations.


3.1 LOCKING PHENOMENA WITH SOLID-SHELL
The well-known locking phenomenon of displacement based finite elements for
thin-walled beams, plates and shells is caused by unbalances of the trial functions. This
unbalance, described in innumerable papers, can be cured by either reduction or
enhancement of the DOF to strengthen the interpolations of variable fields.
With solid elements, there are essential locking effects as:
Shear locking;
Poisson thickness locking;
Curvature thickness locking (Section 3.1.1);
Volumetric locking.
Meanwhile, for shell elements, there are also some shell-typical locking
phenomena to face with:
Transverse shear locking;
Membrane locking.
The solid-shell elements are solid elements where the shell like-behaviors are
integrated so the elements can be coped with thin-walled problems. With the assumption
of straight normal to the element mid-surface, the solid-shell formulation can take into
account transverse shear effects.
If the solid-shell elements are applied for simulation of thick structures, the
locking effects they meet are similar to locking effects happen with solid elements. On
the contrary, being applied for thin and curve structure the solid-shell elements behave
similarly to shell elements, so the anti-locking techniques for shell elements would be
useful.
In general, methods to remedy locking effects may be classified as follows:
Incompatible displacement models by Wilson et al. [WIL73] and Taylor et al.
[TAY76], designed by the extension of the trial functions through additional
incompatible modes.
ANS elements by Hughes & Tezduyar [HUG81], Dvorkin and Bathe [DVO84].
58
Assumed stress elements by Pian [PIA86], based on the Hellinger-Reissner
functional to enrich the stress space.
Reduced or selected reduced integration techniques, e.g. [DOL00], to clear the
parasitic stresses by a modified numerical integration.
EAS elements by Simo and Rifai [SIM90], Simo and Armero [SIM92] based on
the Hu-Washizu functional and the extension of the strain tensor or the material
deformation gradient by additional terms.
The solid-shell element would suffer transverse shear locking and membrane
locking as the degenerated shell if we do not apply any one of the remedial methods as:
ANS, EAS, RI or assumed stress method. On the other hand, the solid-shell elements
would show volumetric locking as solid element if plasticity in either small or large
displacement occurs. To overcome those, the solid-shell formulation has to adopt the
above methods. In order to improve the element performance, it means remove locking
effects, the EAS method could be used for the membrane and bending strains, while the
transverse shear strain components are formulated according to the ANS method.
In Chapter 2 we did discuss about severe locking effects (volumetric locking,
Poisson thickness locking and shear locking) that happen to the low-order solid and, of
course, also happen to the solid-shell elements. Hereafter, under mechanical point of
view (Section 2.1.1), other locking phenomena happen to the solid-shell elements are
continuous discussed.

3.1.1 Curvature (trapezoidal) locking
A further locking effect observed for solid elements is the phenomenon of a so
called curvature locking or sometimes trapezoidal locking. The phenomenon is only
found in structures where the out-of-plane element edges are not perpendicular to the
mid-layer, which is the case for originally curved as well as for heavily deformed
structures. However, curvature locking only occurs when the elements include thickness
strains. It means continuum elements and some extensible-director degenerated shell
elements suffer this locking. Other degenerated shell elements, whose thickness strains
are equal to zero, do not suffer this locking. In brief, curvature locking happens when the
two following factors are present in the same time:
1) When element models include normal strains in thickness direction and
2) The out-of-plane element edges are not perpendicular to the mid-layer,
it will activate incompatible normal strains in thickness direction.
Trapezoidal locking is the least envisioned deficiency in the solid-shell element
development. This ignorance is probably due to the fact that curvature locking does not
occur in degenerated shell elements and flat plate geometry. Curvature locking was first
put forward by McNeal [MAC87]. Lets consider a simple case of four node element,
trapezoidal shape as Figure 3.1.1.
In the Cartesian coordinates, the element has the height H = 2; the average width
W = H, where is the element aspect ratio. The trapezoidal element relates to the
isoparametric space by:
X =
1
(1 -
3
) ; Z =
3
; = tan() (3.1.1)
59















Consider the analytical displacements for in-plane bending, see (2.1.15). For the
sake of simplicity we assume M/EI = 1; = 0 and we do not consider the constant term in
vertical displacement. Finally, the solutions are:
u
CM
= XZ
w
CM
= - X/2
(3.1.2)
or in the form of isoparametric coordinates
u
CM
= [
1

3
-
1
(
3
)
2
]
w
CM
= - [(
1
)
2
- 2(
1
)
2

3
+ (
1

3
)
2
]/2
(3.1.3)
The displacement within the element is calculated, see [MAC94], as:
u
a
= (
1

3
-
1
)
w
a
= - (1 - 2
3
+ )/2
(3.1.4)
The compatible strains in the element are:

3
3
2
3
1
3
( )

(1 )
( )
2 1
(1 )
a
X
a
Z
a
XZ

=
(
= +
(

(3.1.5)
While the analytical strains are:

3
0
2 0
CM
X
CM
Z
CM
XZ

=
=
=
(3.1.6)
In the case of rectangular, = 0, the strain components
X
and
Z
are correct. In
case 0, the error in
X
is small for small. The error in shear term
XZ
(causes
transverse shear locking) can be eliminated by the ANS method when shear strain is
interpolated by strains at sampling points in order to assure shear strains are equal to zero
in pure bending, see Chapter 2. Here we can see that apart from the inconsistent terms
(see Section 2.1.1.2) that cause transverse shear locking, the distorted mesh also causes a
similar effect. This locking effect becomes severe, due to parasitic strain
a
Z
, when the
curvature of the structure is high with respect to the thickness, i.e. when tan() = >>
2(1+)
2(1-)

3

2
2


X
Z
Figure 3.1.1: Four node element
H

=

2

a) In Cartesian space b) In isoparametric space
60
1. Curvature locking is a consequence of geometric irregular (distortion) on analysis
accuracy.
Curvature locking shows up if 2D or 3D solid elements are used to model curved,
thin-walled structures (the name trapezoidal locking reflects the fact that in these cases
the individual elements have a trapezoidal shape). When shell elements are built on
extensible director kinematics and are incorporated unmodified 3D constitutive models
these elements will also show severe locking behavior in the case of curved-thin shell
structures. We make conclusions that when multiple trapezoidal elements are used to
model bending problems, the transverse bending stress/strain mode, which should
physically vanish, is most detrimental to the element accuracy and leads to a deficiency.
In other words, the element accuracy drops substantially if trapezoidal meshes are used.
The oblique edges activate parasitic normal strains in thickness direction and may lead to
locking. The effect happens only for curved structures, and is severe for thin solid and
solid-shell elements. One method to resolve this problem is using a naturally assumed
strain interpolation of the normal strain in thickness direction as proposed by Bischoff
and Ramm [BIS97], Betsch and Stein [BET95]. The detail formulation is presented in
Section 3.4 below.

3.1.2 Membrane locking
In order to understand membrane locking, it is necessary to distinguish
extensional bending and inextensional bending. The term "inextensional bending"
refers to a class of plate and shell problems in which the potential energy is dominated by
flexural strains as opposed to extensional strains. It means the in-plane strains (
X
,
Y
and

XY
) become vanishingly small when compared to the bending strains (
Z
). In contrast,
the term extensional bending is referred if the mid-surface experiences significant
stretching or contraction; also called combined bending-stretching or coupled membrane-
bending.
Membrane locking, also known as inextensional locking, does only occur in
curved beam and shell elements when the curvature is large. It is sometimes confused
with shear locking and volumetric locking because these affect the membrane part of
shell elements. However, they are completely different phenomena.








For the sake of simplicity, in order to understand membrane locking easily, lets
consider a curved beam element of length 2l and radius of curvature R based on classical
thin beam theory, see Figure 3.1.2. The displacement degrees of freedom required are the
circumferential displacement u and the radial displacement w. The coordinate s follows
Figure 3.1.2: Curved beam element
= s/l R
w
u
s l
61
the middle line of the curved beam. The membrane strain and the bending strain are
described by the strain-displacement relations [PRA01]:
= u,
s
+ w/R
= u,
s
/R - w,
ss

(3.1.7)
Base on derivatives of (3.1.7), obviously, a C
0
description for u and a C
2

description for w are required. Kinematically admissible displacement interpolations for u
and w are:
u = a
0
+ a
1

w = b
0
+ b
1
+ b
2

2
+ b
3

3

(3.1.8)
where a
0
to b
3
coefficients are the generalized DOFs which can be related to the nodal
degrees of freedom U, W and W,
s
at the two nodes.
The strain field interpolations can be derived as
= (a
1
/l + b
0
/R) + (b
1
/R) + (b
2
/R)
2
(b
3
/R)
3

= (a
1
/Rl - 2b
2
/l
2
) (6b
3
/l
2
)
(3.1.9)
When the above curved element is applied for simulating an inextensional bending
problem the physical response requires that the membrane strain () tends to vanish. It
means:
a
1
/l + b
0
/R = 0
b
1
/R = 0
b
2
/R = 0
b
3
/R = 0
(3.1.10a)
(3.1.10b)
(3.1.10c)
(3.1.10d)
We can observe that constraint (3.1.10a) has terms participating from both the u
and w fields. It can therefore represent the condition ( 0) in a physical way. However,
the three remaining constraint (3.1.10b) to (3.1.10d) have no participation from the u
field. Lets examine what these three constraints imply for the physical problem. From
the three constraints we have the conditions (b
1
0), (b
2
0), (b
3
0). Each of these
the conditions in turn implies following conditions (w,
s
0), (w,
ss
0) and (w,
sss
0).
These are the spurious constraints due to b
1
, b
2
and b
3
must be different from zero, see
(3.1.8). Consequently, the exist of b
1
, b
2
and b
3
causes membrane locking. Apart from the
EAS method, there are two ways are popularly applied for membrane locking removal.
Use high-order approximations for in-plane displacements or use RI technique for in-
plane strains [STO82]. The first method requires a dramatical reduction of time step for
explicit time integration while the second method may cause spurious modes.
Membrane locking does not occur in flat elements, e.g. the four-node quadrilateral
shell only manifests membrane locking in wrapped configurations. With the solid-shell
elements, due to the coupling of the transverse normal strains with in-plane strains, when
the mesh is distorted, membrane locking may occur. Particularly, it is severe with large
aspect ratios of the elements. For the solid-shell element, this locking is overcomed
effectively by using the EAS method, see [HAU01], with an enhancing matrix M as:
62

(
(
(
(
(
(
(
(

=
0 0 0 0
0 0 0 0
0 0
0 0 0 0
0 0 0
0 0 0
2 1
2
1

M

(3.1.11)
where only in-plane strains, which are equivalent to rows 1, 2 and 4 of the enhancing
matrix M in (2.4.46), are enhanced.
In conclusion, membrane locking occurs if exist two conditions. First, the
transverse strains and the in-plane strains interlock. Second, the interpolations are unable
to model the inextensional behavior in inextensional bending problems. Consequently, a
stiffening effect occurs when pure bending deformations are accompanied by parasitic
membrane stresses.
In the following sections we will formulate a solid-shell element which
incorporates an alternative ANS technique that is different from the classical ANS
technique presented in Section 2.5. This element is free from all locking effects that have
been mentioned.


3.2 KINEMATICS OF SOLID-SHELL

For the use of the ANS and the EAS methods to design the free-locking solid-shell
elements the weak form is written in a local coordinate system, where the two axes X and
Y are aligned with the element mid-surface and the third axis Z is aligned with the
thickness direction. To ease for presentation, some kinematic definitions in Section 2.5.1
will be repeated in this section.
For the development of the low-order solid-shell element, we naturally adopt the
assumption of Naghdi for shells: the normal to the element mid-surface remains straight
but not necessarily normal during the deformation. Also adopted by Timoshenko beam
theory and the Mindlin-Reissner plate theory, this assumption is fulfilled by a linear
approximation of the in-plane displacements over the shell thickness [HAU98]. In this
context, the interpolation formulations for geometrical vectors can be read as

( )
1 2 1 2
3 3
( , ) ( , )
1
[(1 ) (1 ) ]
2
h h
u l
= +

X X X (3.2.1)

( )
1 2 1 2
3 3
( , ) ( , )
1
[(1 ) (1 ) ]
2
h h
u l
= +

x x x (3.2.2)
where: =
T
} , , {
3 2 1
=[-1,1][-1,1][-1,1], tri-unit cube in R
3
.
In the above formulas,
( )
X and
( )
x are the geometrical descriptions of the solid-
shell element in the initial and current coordinate systems, respectively. The vectors
1 2
( , )
h
u
X and
1 2
( , )
h
l
X are the position vectors of nodes in upper and lower surfaces,
respectively, of the element in the initial coordinate system. Similarly, the position
63

t


Figure 3.2.1: Initial configuration
0
, current configuration
t

and isoparametric configuration

) , , (
3 2 1

: unit vectors of isoparametric system
) , , (
3 2 1
e e e : unit vectors of global (physical) system

3

E
1

E
2

E
3

e
1

e
2

e
3

G
1

G
2

G
3

g
1

g
2

g
3

vectors of nodes in the current coordinate system are denoted by
h
u ) , (
2 1

x and
h
l ) , (
2 1

x .
Here the subscript l is the index for terms in lower surface and the subscript u is the
index for terms in upper surface.























The displacement vector u is calculated as

( ) ( ) ( )
X x u =
(3.2.3)
and the deformation gradient tensor F is defined by

( )
( )
( )

X
x
F

=
(3.2.4)
The Green-Lagrange strain tensor is defined as below

( )
( ) ( )
( ) ( )
( ) ( )
1 1
( ) [( ) ]
2 2
T T

= =




x x
E F F I I
X X

(3.2.5)
Originally proposed by Dvorkin and Bathe for small strains [DVO84], the ANS
method requires a modification of the shear components of the Green-Lagrange strain
tensor E when dealing with large deformation. Consequently, it is necessary to write the
variational equation in the material configuration (or the total Lagrange formulation) in
terms of the Green-Lagrange strain tensor E and its conjugated second Piola-Kirchhoff
stress tensor S .
Moreover, the ANS method requires the interpolation of all assumed terms in
natural coordinate system. Therefore, it is necessary to define a convected description.
Lets consider the position vectors
( )
X of the reference configuration
0
and the
64
Figure 3.2.2: Configuration of low-order solid-shell element
1
5
2
6
3
7
4
8
G
1
2
3
e
3
Upper surface
Lower surface
x
e
2
e
1
G
G
X
Y
Z
position vectors
( )
x of the current configuration
t
in the local coordinate system. As
illustrated in Figure 3.2.1, the convected basis vector
i
G and its components in the initial
basis system are defined by:
3 1 ; ; / = = = i G
j i ij
i
i
G G X G
(3.2.6)
while the contravariant vector
j
G and its components
ji
G are defined following:
3 1 ; ;
-1
= = = = i,j G G
i ji i
ji j j
i
j
i
G G G G G
(3.2.7)

















Similarly, the convected basis vector
i
g in the current basis system and its
components
ij
g are defined as:

j i ij
i
i
i
i
g
i
g g
u G x g
=
= + = = 3 1 ; / /

(3.2.8)
and its contravariant vector
j
g and its components
ij
g are defined through the following
expressions
3 1 ; ;
-1
= = = = i,j g g
j ij j
ij i j
i
j
i
g g g g g
(3.2.9)
Anticipating the transition to a shell element, one of the solid-element dimension
is identified as the shell thickness direction. We call the element face (1,2,3,4) as the
lower face, the element face (5,6,7,8) as the upper face, see Figure 3.2.2. All the edges 1-
5, 2-6, 3-7 and 4-8, the ones that connect the lower node with the upper nodes, are
thickness edges.
The deformation gradient in the form of the convected vectors becomes

( )
( )
( )
i
i
G g
X
x
F

=
(3.2.10)
while the Green-Lagrange strain tensor takes the following form
65

( ) ( ) ( )
j i
ij
j i
ij ij
T
E G g G G G G I F F E

= = =
2
1
) (
2
1
) (
2
1
(3.2.11)
or alternatively:

( )
3 1
2
1
=
|
|

\
|

= i,j
j i
j i j i j i
G G
u u
G
u u
G E

(3.2.12)


3.3 AN ALTERNATIVE ANS TECHNIQUE FOR TRANSVERSE
SHEAR LOCKING REMOVAL

The classical ANS technique presented in Chapter 2 has been widely adopted and
thus applied for shell [DVO84] and solid-shell elements ([HAU98], [HAU00], [HAU01],
[QUO03a]), i.e. 3D modeling. Its 2D counterpart can be obtained through a degeneration
of the 3D version, and is applicable to a 2D solid-shell element (Figure 3.3.1). The
2D.ANS element has only a single sampling point (A C) for the assumed shear strain
XZ
E
~
(2D case). Such a configuration can be also obtained through the application of a
selective-reduced integration to the shear part of the element. Consequently, the assumed
shear strain
XZ
E
~
becomes, in the 2D version, constant over the element, while it is not the
case for the 3D version.
We alternatively investigate another ANS technique, where the 3D and 2D
versions always feature a linearly assumed shear strain (along the thickness direction). To
this end, lets first start from the standard solid element to develop a new 3D ANS-solid
element (designated by ANSn). Its 2D counterpart (designated by 2D.ANSn) can be
immediately followed through a simple degeneration, see [NGU08].














3.3.1 Cubic hexahedral ANS element (ANSn)
Consider a single solid element of tri-unit geometry (size 222), see Figure 3.3.2.
For the sake of simplicity, the isoparametric space ( )
3 2 1
, , and the physical space
(X,Y,Z) are taken identical. Hence, it is possible to investigate the problem directly in the
physical space. Recall the Green-Lagrange strain components in (2.1.2):
Figure 3.3.1: Degeneration from 3D to 2D of ANS-solid element

3

A
B
D
C
E F
H
G X
Z
Y
X
Z

3

A C
E H F G
Degeneration
3D 2D
66

+ +
+ +
+ +
+ +
+ +
+ +
+

+
+
+
=

=
) v v (
) v v (
) v v (
) v v ( 5 . 0
) v v ( 5 . 0
) v v ( 5 . 0
v
v
v
2
2
2
, , , , , ,
, , , , , ,
, , , , , ,
, , , , , ,
, , , , , ,
, , , , , ,
, ,
, ,
, ,
,
,
,
) (
Z Y Z Y Z Y
Z X Z X Z X
Y X Y X Y X
Z Z Z Z Z Z
Y Y Y Y Y Y
X X X X X X
Y Z
X Z
X Y
Z
Y
X
YZ
ZX
XY
ZZ
YY
XX
com
u
w w u u
w w u u
w w u u
w w u u
w w u u
w w u u
w
w u
u
w
u
E
E
E
E
E
E
E
(3.3.1)
The displacement derivatives of displacement in 0X direction are defined as
follows

8 8
1
1 1
8 8
2
1 1
8 8
3
1 1
,
,
,
I I
I I
I I
X
I I
Y I I
I I
Z
I I
I I
I I
N N
U U
X
u
N N
u U U
Y
u
N N
U U
Z

= =
= =
= =






= =
` ` `


)




) )




(3.3.2)
If the element is subjected to a pure bending in 0X direction, Figure 3.3.3, and
assuming that the Poisson ratio is equal to zero, the following relations are hold:

8 1 I ; 0


8 7 6 5
4 3 2 1

= = =
= = = =
= = = =
I I
W V
U U U U U
U U U U U

(3.3.3)
and then

0 , , ,
0 , v , v , v
= = =
= = =
Z Y X
Z Y X
w w w

(3.3.4)


















Lets investigate the compatible shear strain
XZ
E , which can be directly derived
from (3.3.1) - (3.3.4):

Z X Z X Z XZ
u u u u u E , ) , 1 ( , , , 2 + = + = (3.3.5)
Figure 3.3.2: Solid element, special case (X,Y,Z) (
1
,
2
,
3
)
Y,
2

X,
1

Z,
3

1 2
Nodal coordinates:
1(-1,-1,-1) 5(-1,-1,1)
2(1,-1,-1) 6(1,-1,1)
3(1,1,-1) 7(1,1,1)
4(-1,1,-1) 8(-1,1,1)
3 4
5 6
7 8
67
where the second term represents the nonlinear geometrical quantities. Notice that:
( )
1
8
1
5 2 3
8
1

2
1
,

= =
+ =

=
I
I
I
I
I
I
Z
U U U
N
U
Z
N
u
(3.3.6)
] ) ( [
2
1
,
3
6 2 6 2
U U U U u
X
+ = (3.3.7)





















Thus, the standard element clearly presents a strain
XZ
E , which is a bilinear
function in both
1
and
3
, while this strain component must be physically zero in the
pure bending case. In other words,
XZ
E contributes to the so-called parasitic transverse
shear strain in pure bending problems. In order to remove this shear locking, we observe
that this parasitic shear strain is only equal to zero at 0
1
= , see (3.3.5) and (3.3.6) or line
u/Z in Figure 3.3.4. This motivates the use of sampling points, which are employed
later for the interpolation of assumed strains, on the vertical plane including the point
0
1
= and the axis
2
O . Indeed, the shear strain
XZ
E is equal to zero at the points
l u l
C A A , , and
u
C (Figure 3.3.4) under the pure bending condition. As a result, the physical
strain
XZ
E
~
is assumed to take the following form:

C
XZ
A
XZ XZ
E E E
~
) 1 (
2
1 ~
) 1 (
2
1 ~
2 2
+ + = (3.3.8)
where
A
XZ
E
~
and
C
XZ
E
~
are derived through the strains at sampling points on the face
) , , , (
l u u l
C C A A on Figure 3.3.4:

u l
u l
C
XZ
C
XZ
C
XZ
A
XZ
A
XZ
A
XZ
E E E
E E E
) 1 (
2
1
) 1 (
2
1 ~
) 1 (
2
1
) 1 (
2
1 ~
3 3
3 3


+ + =
+ + =

(3.3.9)



Figure 3.3.3: Pure bending in 0X direction
1
5
6
F F
F
U
1

U
5

U
6

Z,
3

X,
1

2
F
U
2

8
7
F F
U
8
U
7

4
F
U
4

3 F
U
3

Y,
2

68

























Similarly, the assumed shear strain
YZ
E
~
is interpolated as

B
YZ
D
YZ YZ
E E E
~
) 1 (
2
1 ~
) 1 (
2
1 ~
1 1
+ + = (3.3.10)
where
D
YZ
E
~
and
B
YZ
E
~
are assumed by using sampling points on the face ) , , , (
l u u l
B B D D ,
Figure 3.3.4, as

u l
u l
B
YZ
B
YZ
B
YZ
D
YZ
D
YZ
D
YZ
E E E
E E E
) 1 (
2
1
) 1 (
2
1 ~
) 1 (
2
1
) 1 (
2
1 ~
3 3
3 3


+ + =
+ + =

(3.3.11)
Interpolation of the transverse shear strains as (3.3.8) and (3.3.10) make sure of
preserving the bilinear variation of the assumed strains as the derived-displacement shear
strains.

The 2D version (Figure 3.3.5) of the ANSn element can be deduced in a
straightforward manner. Indeed, under pure bending in 0X direction, the bending mode is
activated so that w/X = w/Z = 0. Hence, the compatible transverse shear strain
becomes:

1 3
12 2 4 2 3
1 1
2 2 ( ) [1 ( )(1 )]
2 2
XZ
u u u
E E U U U U
Z Z X


= = + = + + +

(3.3.12)
Since this shear strain is equal to zero at 0
1
= , we can employ sampling points on
the vertical line through the point 0
1
= . For example, with the use of two sampling
points A
1 3
( 0, 1) = = and B
1 3
( 0, 1) = = , and the corresponding assumed shape
functions are:
Figure 3.3.4: Sampling points for ANSn element
u/Z
(in pure bending)
Strain sampling
points
1
2
5 6
A
l

Z,
3

A
u

4
8
7
X,
1

C
l

Y,
2

C
u

D
l

D
u

B
l

B
u

D B
A
C
O
l

O
u

O
3
E
H
F
G
Sampling points:
A
l
(0,-1,-1); A
u
(0,-1, 1)
B
l
(1, 0,-1) ; B
u
(1, 0, 1)
C
l
(0, 1,-1) ; C
u
(0, 1, 1)
D
l
(-1, 0,-1) ; D
u
(-1, 0, 1)
69

3 3
1 1
(1 ); (1 )
2 2
A B
N N

= = + (3.3.13)
The assumed shear strain
xz
E

can be linearly interpolated through two points A

and B:

A B
XZ A XZ B XZ
E N E N E

= +

(3.3.14)






















In the case of 3D elements, being applied for the pure bending in 0X direction, the
above ANS technique for the shear locking removal can be also employed for the pure
bending in 0Z direction. We however note that the bending in 0Z direction is not
considered in the conventional shell elements.
In the general case, where the element geometry is not regular, the shear term
needs to be evaluated in the natural space. The assumed strains are then transformed to
the physical space for the computation of the stiffness matrices. All of these problems are
addressed below.

3.3.2 Distorted hexahedral ANSn element
As a requirement, the order of the assumed strain field should be equal to the one
of the derivation of the displacement field so that the assumed strains may consistently
capture the strain field resulting from the derivation of the displacement in non-pure
bending modes. Notice that only the transverse shear strains ) and (
23 13
E E are considered
for shear locking removal by the ANS technique.
Consider the Green-Lagrange strain values, which can be derived from (3.2.11)
and (3.2.12) as:

j i j i j i ij ij ij
G g E

= =
u u
G
u u
G
(3.3.15)
1
2
4
3
X,
1

A
Figure 3.3.5: Sampling points for 2D.ANSn element
Z,
3

B
u/Z
(pure bending case)
Integration sampling
points
Strain sampling
points
Assumed shear
strain
XZ
E

(arbitrary case)
70
In order to determine the order of the function
ij
E with respect to the natural
variables
i
, we investigate the constituents of the strains in (3.3.15). The vector
i
G is a
bilinear function of the natural coordinates as follows:
{ }
T
k k k
k
i
K
T
i i i i
Z Y X
N Z Y X

8
1

=
)
`

=

G (3.3.16)
where
T
} , , {
k k k
Z Y X are reference nodal coordinate vector. The derivatives of the
displacement vector with respect to the natural coordinate
i
is
{ }
T
k k k
k
i
k
T
i i i i
W V U
N w v u

8
1

=
)
`

u
(3.3.17)
Based on (3.3.15) - (3.3.17) we can examine the order of the natural constituents
in the shear strain function
13
E as

3 1 3 1 3 1 13

=
u u
G
u u
G E
(3.3.18)


















Vectors
1
G and
1
u are derivatives of the shape functions with respect to
1
,
hence,
1
G and
1
u are functions of ) , , (
3 2 3 2
; similarly, vectors
3
G and
3
u
are functions of ) , , (
2 1 2 1
. Consequently,
13
E is a function of the natural coordinates
of the type:
] ) ( , ) ( , ) ( , , , , ) ( , , , , [
3 2 2 1 2 2 1 3 2 2 3 2 1 3 2 3 1 2 2 2 1 3 2 1
13
f E = (3.3.19)
Clearly, (3.3.19) shows that
13
E is a linear function of
1
. In order to remove shear
locking the naturally transverse assumed shear strain
13
E is interpolated on the face that
goes through 0
1
= and contains the axis
2
O , see plane (P) in Figure 3.3.6. On that
natural face the shear strain is quadratic with respect to
2
but linear with respect to
3

Figure 3.3.6: Distorted element - sampling points for strains

2

Z,
3

1
5
2
6
3
7
4
8
A
C
B
D
E
F
G
H
X
Y
L
U
A
A
B
B
C
C
D
D
L
L
L
U
U
U
P
O
L

O
U

71
as ] ) ( , , ) ( , , [
3 2 2 3 2 2 2 3 2
13
f E = . Hence, the naturally assumed shear strain
13
~
E can be
assumed to be a function whose order of natural variables is equal to or lower than the
order of natural variables in the derived-displacement strain (3.3.19). It can be:
Constant in thickness direction
3
O and linear in horizontal direction
2
O . This is
the classical ANS method [DVO84]. Hauptmann and Schweizerhof [HAU98],
Klinkel et al. [KLI06], Vu-Quoc and Tan [QUO03a] use this interpolation in their
solid-shell elements. This interpolation requires only two sampling points A and
C, see (2.5.8) and Figure 3.3.6.
Linear in both thickness direction
3
O and horizontal direction
2
O . This
interpolation requires two sampling points along both
3
O and
2
O
) , and , (
u l u l
C C A A , Figure 3.3.6. Similarly to (3.3.8), the natural assumed strain
13
~
E
is interpolated as

C A
E E E
13
2
13
2
13
~
) 1 (
2
1 ~
) 1 (
2
1 ~
+ + = (3.3.20)
where
A
E
13
~
and
C
E
13
~
are evaluated by using sampling points on the face that goes
through ) , , , (
u l u l
C C A A :

u l
u l
C C C
A A A
E E E
E E E
13
3
13
3
13
13
3
13
3
13
) 1 (
2
1
) 1 (
2
1 ~
) 1 (
2
1
) 1 (
2
1 ~


+ + =
+ + =

(3.3.21)
Linear in thickness direction
3
O and quadratic in
2
O direction. This
interpolation requires two sampling points along
3
O ) , and , (
u l u l
C C A A and three
sampling points along
2
O ) , , and , , (
u u u l l l
C O A C O A , Figure 3.3.6. This assumed
interpolation is the closest to the consistent displacement-derived strain
13
E . Our
numerical tests show that this quadratic interpolation for assumed strains in
2
O
axis significantly increases computational time with respect to the corresponding
linear interpolation while the improvement is not remarkable.

In the thesis, the second option is chosen to be investigated. Similar to
13
E
(3.3.19), the shear strain
23
E is a function of the natural coordinates as shown below
] ) ( , , ) ( , ) ( , , , , ) ( , , , [
3 2 2 1 3 2 1 3 2 1 2 2 1 3 2 3 1 2 1 2 1 3 2 1
23
f E = (3.3.22)
In order to overcome that shear locking occurs with the pure bending in 0Y
direction, we consider
23
E at 0
2
= . At that coordinate the shear strain
23
E is quadratic
with respect to
1
as ] ) ( , , ) ( , , [
3 2 1 3 1 2 1 3 1
23
f E = but
23
E is linear with respect to
3
.
Similar to (3.3.20), the natural assumed strain
23
~
E , if linearly interpolated in both
thickness direction
3
O and horizontal direction
1
O , is expressed as

B D
E E E
23
1
23
1
23
~
) 1 (
2
1 ~
) 1 (
2
1 ~
+ + = (3.3.23)
72
where
D
E
23
~
and
B
E
23
~
are linear functions of
3
with sampling points on faces parallel axis
2
O and go through a pair of points ) , (
u l
D D and ) , (
u l
B B , respectively:

u l
u l
B B B
D D D
E E E
E E E
23
3
23
3
23
23
3
23
3
23
) 1 (
2
1
) 1 (
2
1 ~
) 1 (
2
1
) 1 (
2
1 ~


+ + =
+ + =


(3.3.24)
In order to alleviate shear locking, the natural shear strains
13
E and
23
E are
replaced by the assumed natural shear strains
13
~
E and
23
~
E before doing the
transformation from the natural space to the physical space. Finally, the physical assumed
strain vector by the alternative ANS technique is:
{
}
{
}

] ) )( 1 ( ) )( 1 )[( 1 (
] ) )( 1 ( ) )( 1 )[( 1 (
2
1
] ) )( 1 ( ) )( 1 )[( 1 (
] ) )( 1 ( ) )( 1 )[( 1 (
2
1
) (
) ( 5 . 0
) ( 5 . 0
) ( 5 . 0
~
2
~
2
2
~
) ( 23 23
3
) ( 23 23
3 1
) ( 23 23
3
) ( 23 23
3 1
) ( 13 13
3
) ( 13 13
3 2
) ( 13 13
3
) ( 13 13
3 2
12 12
33 33
22 22
11 11
23
13
12
33
22
11

+ + +
+ + +
+ + +
+ + +

=

U L
U L
U L
U L
B
e e
B
e e
D
e e
D
e e
C
e e
C
e e
A
e e
A
e e
e e
e e
e e
e e
T T ANS
G g G g
G g G g
G g G g
G g G g
G g
G g
G g
G g
E
E
E
E
E
E




T T E
(3.3.25)
These assumed strains can be implemented in the standard solid element in a
straightforward manner. The physical assumed strain-displacement matrix at node I (I =
1-8) of the solid-shell element is similarly assumed:
{
}
{
}
(
(
(
(
(
(
(
(
(
(
(
(
(
(

+ + + + +
+ + + + +
+ + + + + +
+ + + + +
+
=

] ) )( 1 ( ) )( 1 )[( 1 (
] ) )( 1 ( ) )( 1 )[( 1 (
2
1
] ) )( 1 ( ) )( 1 )[( 1 (
] ) )( 1 ( ) )( 1 )[( 1 (
2
1




~
) ( 2 3 , 3 2 ,
3
) ( 2 3 , 3 2 ,
3 1
) ( 2 3 , 3 2 ,
3
) ( 2 3 , 3 2 ,
3 1
) ( 1 3 , 3 1 ,
3
) ( 1 3 , 3 1 ,
3 2
) ( 1 3 , 3 1 ,
3
) ( 1 3 , 3 1 ,
3 2
1 2 , 2 1 ,
3 3 ,
2 2 ,
1 1 ,
U L
U L
U L
U L
B
T
I
T
I B
T
I
T
I
D
T
I
T
I D
T
I
T
I
C
T
I
T
I C
T
I
T
I
A
T
I
T
I A
T
I
T
I
T
I
T
I
T
I
T
I
T
I
T ANS
I
N N N N
N N N N
N N N N
N N N N
N N
N
N
N
g g g g
g g g g
g g g g
g g g g
g g
g
g
g
T B





(3.3.26)

We can see that the ANS technique with alternative sampling points (the bilinear
assumed strains (3.3.25) and the classical ANS technique (2.5.12) have some common
and uncommon features:
73
For both techniques, the strain sampling points are taken at coordinate 0
1
= for
13
E and 0
2
= for
23
E .
For the alternative ANS technique the assumed strains are linear in the thickness
direction. Meanwhile for the classical ANS technique the assumed strains are
constant in the thickness direction. When strain is not too large, the two
techniques are similar because the strain at the mid-surface (the classical ANS)
will be the average value of the strains at the lower and the upper surfaces (the
alternative ANS presented here).
As being employed in the degenerated shell elements [AHM70], the classical
ANS technique uses sampling points on the mid-surface, which is usually
considered to be the reference plane. Contrarily, the alternative ANS technique
employs sampling points located on the physical, upper and lower surfaces, but
not on the reference plane of the solid element.

For distorted 2D.ANSn element, the shear strain is similarly assumed as (3.3.14),
Figure 3.3.5,

B A
E E E
12
2
12
2
12
) 1 (
2
1
) 1 (
2
1 ~
+ + = (3.3.27)
These assumed strains are then transformed to the physical space by the transform
matrix T in (2.4.43).
The 3D eight-node solid element which directly employs assumed strains as
(3.3.25) is hereinafter referred as ANSn. The element can be applied for distorted mesh
model. It is a 3D solid element which is free from shear locking for thin and moderate
thick-wall structures, see Section 3.6 below for numerical illustration.


3.4 COMBINED ANS-EAS SOLID-SHELL ELEMENT
The solid-shell element presented in this chapter possesses the performances of the
ANS elements in alleviating transversal shear locking and curvature locking. Further
more, to be free from volumetric locking, Poisson thickness locking (see [HAU98],
[HAU00]) and membrane locking the EAS method needs to be adopted. The solid-shell
elements which only employ the ANS techniques [FEL03] should not be applied for
incompressible deformation problems because they exhibit poor performance for
volumetric locking removal, see [BIS97].
In the above paragraphs, curvature locking was mentioned. In order to circumvent
curvature locking several authors, e.g. Bischoff and Ramm [BIS97] and Betsch and Stein
[BET95], suggested employing an assumed strain approximation for the strain
component
33
E . Corner nodes on the mid-surface play the role of the sampling points
(points E, F, G and H, Figure 3.3.6). The
0
C -continuous strain field is thus given by:

=
I
I I
E N E
) ( 33
) , (
33
2 1
~

; I = E, F, G, H (3.4.1)
74
where
) , (
2 1
I
N are the shape functions in mid-surface,
3
= 0, as:
( )( ) 1 1
4
1
2 2 1 1
2 1


I I
I
N + + =
) , (
(3.4.2)
Hauptmann et al. [HAU00] did show that, this kind of locking is minor compared
to other types, a bilinear interpolation of the transverse normal strain as (3.4.1) is enough
for subduing this locking. One simple way to overcome curvature locking is using fine
mesh on thickness direction.
Introducing the assumed natural strain
33
~
E (3.4.1) into the assumed strain vector
ANS
E
~
(2.5.12) or (3.3.25) we receive the strain field of an element that is free from both
transversal shear locking and curvature locking. The classical ANS element that is free
from transverse shear locking and curvature locking is given by:


) )( 1 ( ) )( 1 (
) )( 1 ( ) )( 1 (
) (
) ( 5 . 0
) ( 5 . 0
) ( 5 . 0
~
2
~
2
2
~
~
) ( 23 23
1
) ( 23 23
1
) ( 13 13
2
) ( 13 13
2
12 12
) ( 33 33
4
) , (
22 22
11 11
23
13
12
33
22
11
2 1

+ +
+ +

=


B
e e
D
e e
C
e e
A
e e
e e
e e
I
I
e e
e e
T T ANS
G g G g
G g G g
G g
G g N
G g
G g
E
E
E
E
E
E
I




T T E

(3.4.3)
where the transverse normal strain (3.4.1), which is assumed in order to alleviate
curvature thickness locking [BIS97], were rewritten as

) ( 33 33
4
) , (
33
) ( 5 . 0
~
2 1
I
e e
I
I
G g N E

=



(3.4.4)
Once the transverse shear strains are assumed, all strain-displacement matrices can
be similarly formulated. The classical ANS strain-displacement matrices are given by:

(
(
(
(
(
(
(
(
(
(
(
(

+ + + +
+ + + +
+
=

) ( 2 3 , 3 2 ,
1
) ( 2 3 , 3 2 ,
1
) ( 1 3 , 3 1 ,
2
) ( 1 3 , 3 1 ,
2
1 2 , 2 1 ,
3 3 ,
4
J
) , (
2 2 ,
1 1 ,
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1
) )( 1 (
2
1

) (


~
2 1
B
T
I
T
I D
T
I
T
I
C
T
I
T
I A
T
I
T
I
T
I
T
I
J
T
I
J
T
I
T
I
T ANS
I
N N N N
N N N N
N N
N N
N
N
g g g g
g g g g
g g
g
g
g
T B




(3.4.5)
with I = 1-8 (node number) and J = E, F, G, H.
The alternative assumed natural strains and the associate strain-displacement
matrices with curvature locking removal are formulated similarly to the classical ANS
ones (3.4.3) and (3.4.5).

75
An EAS element that is free from volumetric locking and membrane locking
[HAU01] has an enhancing matrix M such as

0 3 29 27 26 25
0 0 0 0 0
0 0 0 0 0
0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
2 1
3
2
1
(
(
(
(
(
(
(
(

M

(3.4.6)
An EAS element that is free from volumetric locking, membrane locking and
Poisson thickness locking [QUO03a] has an enhancing matrix M such as

45 4 4 0 3 29 27 26 25
0 0 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0
0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
2 1
3 2 3 1 3
2
1
(
(
(
(
(
(
(
(

M

(3.4.7)
The EAS element, which is interpolated by a matrix M such as (3.4.7), is
enhanced by seven internal parameters hence it is named EAS7. Recall that the
enhancing matrix (3.4.7) is extracted from matrix (2.4.46). Therefore, the EAS7 always
satisfies the patch test. The trilinear terms in modes 49, 50 and 51 of (2.4.46) are not
taken into account because they only improve element performance a little while making
the computational cost increase significantly.
Eventually, a completely original element that combines both the advantages of
the EAS and ANS formulations can be designed, simply by combining the ANS strain
field (3.3.25) and (3.4.1) with the 7 EAS modes (3.4.7) thus resulting in the so-called
SS7n. It means the solid-shell element with 7 EAS modes and the alternative ANS
technique. In this element, the resulting strain field is

enh ANS mod
) ( ) (
~
u
E E E + =
(3.4.8)
Similarly, a solid-shell element with 7 EAS modes (3.4.7) and the classical ANS
technique (3.4.3) results in the so-called SS7. This element has already been presented in
[QUO03a]. In the thesis, the EAS internal parameters are used to remove volumetric
locking for solid-shell element. Instead of 9 modes (25-27 and 40-45) as proposed in
[AND93], only three modes 25 - 27 of (2.4.46) are adopted for the solid-shell element to
limit computational cost.
Detail of the solid-shell implementation is given in Figure 3.4.1. In general, the
implementation is valid for the EAS element. If the compatible strains are assumed as in
(3.4.3) or (3.3.25) and then introduced in step 3 of the algorithm (Figure 3.4.1) we then
have the solid-shell element formulation.


76
ALGORITHM FOR SOLID-SHELL IMPLEMENTATION

1. Step 1: Initial values
k=0;
(k)
= U 0 ;
( ) k
= 0 ;
(k)
U from the last time step; tolerance Tol
2. Step 2: Update at element level for iteration (k +1):
Nodal displacement:
h
k
h
k
h
k ) ( ) ( ) 1 (
U U U + =
+

EAS parameters:

1
( 1) ( ) ( ) ( ) ( )
[ ] ( )
e e e e h e
k k k u k enh(k) k

+
= + k k U f
3. Step 3: at each Gauss point for each element,
a. calculate:
Compatible strains:
( 1)
com
k+
E as in (3.3.1)
Enhanced strains (2.4.45):

e
k
e
k
enh
k ) 1 ( ) 1 ( ) 1 ( + + +
= E where
J
J
M T
0
0
T e
=
ANS strains, e.g. (3.4.3) for the classical ANS method,:
if (solid-shell element)

enh
k
ANS
k
mod
k ) 1 ( ) 1 ( ) 1 (
~
+ + +
+ = E E E ;
( 1) ( 1)
ANS
k k + +
=

B B ;
mod
k
S mod
k
W
) 1 (
) 1 (
+
+

=
E
S
else (EAS element)

enh
k
com
k
mod
k ) 1 ( ) 1 ( ) 1 ( + + +
+ = E E E ;
( 1)
( 1)
mod S
k mod
k
W
+
+

S
E

end if
b. calculate tangent matrices and internal forces, Section 2.4.2

0 0
4
( 1) ( 1) ( 1) ( 1) ( 1)
( ) ( , )
e e
T
e e e mod
uu k mat k geo k k k
V V
dV dV
+ + + + +
= + = +

h
T
U
k k k B CB B S

0
4
( 1) ( 1)
( )
e
T
e
u k k
V
dV
+ +
=

k C B ;
0
4
( 1) ( 1)
( )
e
T
e
k k
V
dV
+ +
=

k C

0
int( 1) ( 1)
( )
e
e T mod
k k
V
dV
+ +
=

f B S ;
0
( 1) ( 1)
( )
e
e T mod
enh k k
V
dV
+ +
=

f S
dA dV
e e
0
A
T
k
V
T
k
e
1) ext(k

+ + +
+ =

* t N * b N f
c c ) 1 ( ) 1 (
) ( ) (
Update at element level [ ] [ ]
e
k u
e
k
T
e
k u
e
k uu
e
k T ) 1 (
1
) 1 ( ) 1 ( ) 1 ( ) 1 ( +

+ + + +
=

k k k k k
[ ] [ ]
e
k enh
e
k
T
e
k u
e
k nt i
e
k ext
e
k ) 1 (
1
) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( +

+ + + + +
+ = f f f k k r


Save EAS arrays: [ ] [ ]
e
k
e
k enh
e
k
T
e
k u ) 1 ( ) 1 (
1
) 1 ( ) 1 (
, , ,
+ +

+ +
k k f


4. Step 4: Assemble global matrices
1) (k 1) (k
,
+ +
R K .
5. Step 5: Solve the incremental displacement and check for global convergence

1
( 1) ( 1) ( 1) k k k

+ + +
= U K R
if Tol
k
<
+ ) 1 (
R or
( 1) ( 1) k k
Tol
+ +
< U R
goto next time step
else
k=k+1; return Step 2
end if
Figure 3.4.1: Solid-shell algorithm
















































77
3.5 ENHANCED QUANTITIES
3.5.1 Consistent deformation gradient
The solid-shell element presented in this chapter has been built on the modified
strain E
mod
(3.4.8). The associated modified deformation gradient F
mod
is required if the
element is implemented in a source code base on the updated Lagrange formulation or
when a material algorithm for large elastoplastic strains is needed.
The deformation gradient can be split into right-stretch tensor U and rotation
tensor R as:
RU F = (3.5.1)
Introduce U and R into the formulation for calculating Green-Lagrange strain
tensor (3.2.5) we have:
) (
2
1
] ) ( [
2
1
) (
2
1
2
I U I U R R U I F F E = = =
T T T
(3.5.2)
Clearly, Green-Lagrange strain tensor depends only on the right-stretch tensor U.
Thus, from the modified strain
mod
E we can derive the associated modified right-stretch
tensor U
mod
. According to Hauptmann et al. [HAU00], the modified deformation gradient
F
mod
is calculated as

mod mod
RU F = (3.5.3)
We see that calculation of F
mod
requires twice of polar decomposition. The first
time is calculation of rotation tensor R in (3.5.1). The second time is calculation of
modified right-stretch tensor U
mod
from
mod
E . These calculations will increase
computational cost of the algorithm when the deformation gradient F
mod
is required.

3.5.2 Local static condensation
The internal parameters of the EAS element are condensed out at the element
level, see (2.4.37), before assembling. As shown in the solid-shell algorithm (Figure
3.4.1), the internal parameters at iteration (k+1) is calculated by:

1
( 1) ( ) ( ) ( )
[ ] ( )
e e e e h e
k k k u k (k) enh(k)

+
= + k k U f
(3.5.4)
This procedure requires, for each element, that apart from the internal parameters
at iteration (k) -
e
k ) (
must be stored for the calculation of
e
k ) 1 ( +
, other quantities as
e
k ) (
k ,
e
k u ) (
k and
e
enh(k)
f also need to be kept. When solving a problem with a huge number of
DOFs, a large memory space must be reserved for storing these items. Simo et al.
[SIM93] did propose a local static condensation algorithm. According to this local
algorithm, the internal parameters
e
k ) 1 ( +
are not calculated at Step 2. Instead, they are
calculated at the end of Step 3b (Figure 3.4.1) from
e
k ) 1 ( +
k ,
e
k u ) 1 ( +
k and
e
1) enh(k +
f .
Consequently, it is not necessary to save EAS arrays that usually demand a significant
memory allocation. Details of the local static condensation algorithm are presented in
Figure 3.5.1.
78
LOCAL STATIC CONDENSATION ALGORITHM

Let {
h
n
U ,
h
n
U } be the correct solutions at time t
n
and {
h
k n ) ( 1 +
U ,
h
k n ) ( 1 +
U } be
solutions at a given iteration (k) within the interval [t
n
, t
n+1
]. Fix this iteration and
compute
e
k n ) ( 1 +
for each element by means of the following sub-iteration (at the
element level):

2. Step 1: Initial values
k = 1 ;
e
n
e
n
=
+ ) 1 ( 1
; tolerance tol ; Calculate
com
k ) 1 ( +
E
3. Step 2: Update at element level for iteration (k +1):
Compute:
enh
k ) 1 ( +
E ;
enh
k
com
k
mod
k ) 1 ( ) 1 ( ) 1 ( + + +
+ = E E E
Use constitutive equations to compute:
mod
k
S mod
k
W
) 1 (
) 1 (
+
+

=
E
S
4. Step 3:
Compute increment:
e
1) enh(k
e
k +

+ +
= f k
e
1) (k
1
) 1 (
] [

Test for convergence:
if ( tol
e
k enh
e
k
<
+ + ) 1 ( ) 1 (
f ) or (k 3)
stop
else
update =
+
e
k
e
k ) ( ) 1 (

k=k+1; return Step 2
end if
Figure 3.5.1: Static condensation algorithm [SIM93]





















In the algorithm of Simo et al. [SIM93], Figure 3.5.1, the internal parameters are
calculated by an approximation formulation:
) ( ] [
1
) ( ) ( ) 1 (
e
enh(k)
e
k
e
k
e
k
f k

+
=


(3.5.5)
Hence, a limited number of iterations ( 3) have to be realized to get a correct
solution.


3.6 NUMERICAL RESULTS AND DISCUSSION
In the thesis, the solid-shell, which adopts 7 EAS parameters (3.4.7) and the
classical ANS techniques (3.4.3), is designated by SS7. While the solid-shell element
with the alternative ANS technique that is presented in the above section, expression
(3.3.25), is designated by SS7n. The additional letter n stands for the elements that
employ the new alternative ANS technique.
79
3.6.1 Patch tests
The patch test has been originally proposed in the mid-sixties as a simple means to
proof convergence of an element. Beside a numerical verification there is also the
possibility of a theoretical analysis. The patch test checks, whether a constant distribution
of any state variable within an arbitrary element patch (i.e. a distorted mesh) can be
represented exactly. It is especially useful for finite element formulations which violate
the compatibility condition (and thus cannot be proven to be consistent), such as the
ANS, EAS, etc. Nature of the patch test is to verify an elements ability to represent a
constant strain/stress field, and thus ensure completeness and an ability to converge in the
limit as the element size decreases.
The membrane patch test (for membrane constant stress state) and the bending
patch test (for bending constant stress state) presented in Chapter 2 Section 2.6 are
satisfied by the classical ANS, SS7 elements and the proposed ANSn, SS7n elements.

3.6.2 Eigenvalue analysis of an incompressible cube
In order to estimate the behavior of the ANSn and SS7n elements at the nearly
incompressible limit an eigenvalue analysis of a unit cube is performed as in [AND93].
The material parameters are elastic modulus E = 1.0 and Poisson ratio = 0.4999. The
cube is considered in regular configuration, and furthermore, in distorted configuration to
check the sensitivity to distorted mesh of elements, see Figure 3.6.1.

Table 3.6.1 shows the eigenvalues of 18 deformable modes of the regular cube
(Figure 3.6.1a), the six zero eigenvalues for the six rigid body modes are not shown.
















The six eigenvalues of the constant strain-states, which are identical for all
elements, are printed in italic letters. Their corresponding eigenmodes can be identified
as three shear modes, two tension modes and one dilatation (or incompressible) mode
(Figure 3.6.2). The eigenvalues of the EAS3v, EAS6v and EAS9v, in Table 3.6.1 are
totally identical to those values of METAFOR [BUI02]. This assures the quality of our
EAS implementation in the MATLAB code.

Figure 3.6.1: Regular and distorted cubes
a) Regular shape b) Distorted shape
1
2
3
4
5
6
7
8
80
Table 3.6.1: Eigenvalues of regular cube
Mode Q1 EAS3v EAS6v EAS9v SS7 SS7n ANS ANSn
1 0.056 0.056 0.056 0.056 0.056 0.056 0.056 0.056
2 0.056 0.056 0.056 0.056 0.056 0.056 0.056 0.056
3 0.167 0.167 0.093 0.093 0.093 0.093 0.111 0.111
4 0.167 0.167 0.111 0.093 0.093 0.093 0.139 0.167
5 0.167 0.167 0.167 0.093 0.111 0.111 0.139 0.167
6 0.222 0,222 0.167 0.167 0.111 0.135 0.222 0.222
7 0.333 0.333 0.167 0.167 0.111 0.135 0.333 0.333
8 0.333 0.333 0.222 0.167 0.222 0.222 0.333 0.333
9 0.333 0.333 0.333 0.222 0.333 0.333 0.333 0.333
10 0.333 0.333 0.333 0.333 0.333 0.333 0.333 0.333
11 0.333 0.333 0.333 0.333 0.333 0.333 0.333 0.333
12 92.654 0.389 0.333 0.333 0.333 0.333 92.617 92.654
13 92.654 0.389 0.333 0.333 0.333 0.333 92.636 92.636
14 92.654 0.389 0.389 0.333 0.333 0.333 92.636 92.636
15 555.648 92,654 0.388 0.389 0.333 0.365 555.593 555.593
16 555.648 92,654 0.389 0.389 0.333 0.365 555.620 555.648
17 555.648 92,654 92.654 0.389 92.617 92.654 555.620 555.648
18 2500.00 2500.00 2500.00 2500.00 2500.00 2500.00 2500.00 2500.00



















For a volumetric-locking free behavior, the elements should contain only one
infinite eigenvalue, it is the eigenvalue of the dilatation mode (mode number 18,
eigenvalue = 2500 is considered as infinite). In the case of the standard elements Q1, six
deviatoric modes (modes 12-17) are always mixed up by volumetric strains which when
0.5 leads to six unrealistic infinite eigenvalues. The same consequences are found
with ANS and ANSn, it means Q1 and both ANS and ANSn are not volumetric locking
free. With the introduction of 3 enhanced modes, the most important part of volumetric
locking is removed in the EAS3v element, which has now a better volumetric locking
Figure 3.6.2: Two deformation modes
a) warping mode b) dilatation mode
1
2
5
6
4
3
8
7
1
2
5
4
3
8
7
81
response in comparison to the Q1. However, higher order parts in the interpolation
functions still exist and this might cause volumetric locking in certain cases. With the
additional enhanced modes, the EAS6v becomes nearly free volumetric locking.
Introduce more enhanced modes, for example EAS9v we obtains totally free volumetric
locking element. As we might observe, after removing volumetric locking, the infinite
eigenvalue becomes finite. The solid-shell elements, SS7 and SS7n, are almost free
volumetric locking, except a warping mode (Figure 3.6.2) has a moderate eigenvalue of
92.6 (mode 17).

Table 3.6.2: Distorted cube - location of nodes
1 2 3 4 5 6 7 8
x -2.5 +2.5 +2.5 -2.5 -0.5 +0.5 +0.5 -0.5
y -2.5 -2.5 +2.5 +2.5 -2.5 -2.5 +2.5 +2.5
z -1.5 -1.5 -0.5 -0.5 +1.5 +1.5 +0.5 +0.5


Table 3.6.3: Eigenvalues of distorted cube
Mode Q1 EAS3v EAS6v EAS9v SS7 SS7n ANS ANSn
1 0.056 0.056 0.056 0.056 0.057 0.059 0.059 0.069
2 0.378 0.378 0.391 0.298 0.067 0.070 0.133 0.135
3 0.401 0.391 0.312 0.276 0.171 0.164 0.202 0.217
4 0.493 0.464 0.422 0.412 0.163 0.181 0.372 0.344
5 0.622 0.614 0.464 0.454 0.284 0.414 0.428 0.589
6 0.679 0.674 0.578 0.553 0.442 0.440 0.591 0.633
7 1.109 0.935 0.614 0.578 0.484 0.667 0.606 0.814
8 1.185 1.109 0.935 0.811 0.660 0.675 1.055 1.031
9 1.783 1.457 0.989 0.925 0.919 0.958 1.436 1.730
10 1.722 1.646 1.457 0.988 1.392 1.448 1.704 1.826
11 12.916 1.724 1.477 1.473 1.805 1.677 177.914 9.241
12 202.125 3.129 1.662 1.614 1.789 1.754 652.063 197.282
13 464.475 3.706 1.724 1.686 3.040 3.035 901.394 771.135
14 1171.097 13.803 3.129 2.020 3.580 3.739 953.793 899.555
15 1764.644 204.498 3.689 3.061 6.863 6.906 2712.966 2721.249
16 3443.134 707.527 17.078 3.280 228.098 26.162 3995.422 4400.630
17 4268.216 1197.604 707.527 17.031 735.750 832.995 6639.616 6566.450
18 10673.478 10278.645 10278.645 10167.644 21513.640 21281.558 25092.928 25024.503

For distorted case (Figure 3.6.1b), the general trend is quite similar as in the
regular case. Table 3.6.3 shows the eigenvalues of 18 deformable modes of the distorted
cube. The ANS and ANSn elements and the Q1 standard element are locked because of
the presence of many infinite eigenvalues (modes 12-17). With the introduction of
enhanced modes in volume, volumetric locking is removed. However, in this distorted
configuration, the EAS elements cannot remove completely the locking effect. For
example, it exists two locking eigenmodes (modes 16, 17) in the EAS6v. Fortunately, the
eigenvalue is rather moderate and we expect a very mild locking. The solid-shell
elements, SS7 an SS7n, also show two locking eigenmodes (modes 16, 17) as the EAS6v.
Anyway, the ANSn and SS7n seem to be better than ANS and SS7, respectively, in this
distorted configuration, see Table 3.6.3 (by comparing eigenvalues of mode 11 of the
82
ANS and ANSn; and by comparing eigenvalues of mode 16 of the SS7 and SS7n). In the
distorted configuration, the SS7n and ANSn elements have only one more moderate
eigenvalue compare to their eigenvalues in the regular configuration, respectively. These
arguments improve that the ANSn and SS7n elements are less sensitive to distorted mesh
than the ANS and SS7 elements, respectively.
The distorted case of the cube was also analysed by Jetteur [JET08]. In [JET08]
the incompatible element, the SRI, EAS24 and EAS9 elements were invoked. Analysed
results in [JET08] showed that the incompatible element behaves similarly as the EAS3v
element. Volumetric locking was almost removed by the EAS9v element. Hence, in this
case the EAS9v and EAS24 give similar results. Only the SRI element is completely
volumetric locking free in this case.

3.6.3 Eigenvalues of a square plate
Eigenvalues of a square plate with zero Poissons ratio is investigated to check
performance of elements in bending [HAU00]. Dimensions and material parameters of
the plate are given in Figure 3.6.3.










Eigenvalues of elements are shown in Table 3.6.4. Result of EAS3DEAS solid-
shell element of Hauptmann et al. [HAU00] is also gathered to compare with the solid-
shell elements presented in this thesis. The EAS3DEAS element incorporated the
classical ANS technique for removal of transverse shear locking, and incorporated the
EAS method for membrane and Poisson thickness locking removals.











Figure 3.6.3: Square plate
a = 5
a
h

=
0
.
1

E = 10.92
= 0.0
X
Y
Z
1 2
5 6
4 3
8
7
1 2
5 6
4
3
8
7
Figure 3.6.4: Two bending modes
a) Bending in 0X b) Bending in 0Y
83
All of the elements have six rigid body modes. The standard element, Q1, suffers
transverse shear locking as we see its eigenvalues for bending modes in 0X and 0Y axes
(Figure 3.6.4) are infinite. These eigenvalues of Q1 are equal to 227.7 while the
correspondent values of non-locking elements are smaller than 1.0 (Table 3.6.4, modes
11 and 12). Furthermore, the Q1 element and also the EAS3v6s element may suffer a
mild locking caused by two warping modes (Figure 3.6.5) because these modes were
affected by transverse shear strains. The warping-mode eigenvalues of the Q1 and the
EAS3v6s are equal to 75.9 while the equivalent values of other elements in Table 3.6.4
are smaller than 1.0 (modes 9 and 10).

Table 3.6.4: Eigenvalues of the square plate
Eigen
mode
Q1 EAS3v6s ANS-SC SS7 ANSn SS7n
1 1365.0 1365.000 1365.000 1365.000 1365.000 1365.000
2 682.8 227.864 227.864 227.864 227.863 227.864
3 682.8 227.500 227.500 227.500 227.500 227.500
4 227.9 682.773 682.773 682.773 682.773 682.773
5 227.5 682.773 682.773 682.773 682.773 682.773
6 151.7 151.727 151.667 151.666 151.727 151.727
7 455.1 455.000 455.000 455.000 455.091 455.091
8 455.1 455.000 455.000 455.000 455.091 455.091
9 75.9 75.924 0.273 0.546 0.273 0.546
10 75.9 75.924 0.273 0.546 0.273 0.546
11 227.7 0.546 0.182 0.546 0.182 0.546
12 227.7 0.546 0.182 0.091 0.182 0.091
13 0.3 0.091 0.091 0.091 0.091 0.546
14 0.3 0.091 0.182 0.091 0.182 0.182
15 0.5 0.546 0.182 0.546 0.182 0.182
16 0.5 0.546 0.182 0.546 0.182 0.182
17 0.5 0.546 0.182 0.546 0.182 0.182
18 1.5E-4 1.454e-4 1.45e-4 1.454e-4 1.45e-4 1.45e-4

Table 3.6.5: Eigenvalues of the square plate - Summary
Eigenvalues (of deformable modes) Element Rigid body
modes < 0.1 < 1.0 Max.
Q1 1 6 7 8 12 1365
EAS3v6s 1 6 7 8 14 1365
EAS3DEAS
(Hauptmann)
1 6 7 9 10 16 1365
ANS 1 6 7 9 10 16 1365
ANSn 1 6 7 9 10 16 1365
SS7 1 6 7 9 10 16 1365
SS7n 1 6 7 9 10 16 1365




84









3.6.4 Square plate at large displacements and strains
Consider a square plate of dimension aah with a = 100 and h = 1. Young
modulus is E = 210
5
and Poissons ratio is = 0.3. The plate is considered in two cases:
fully clamped and simply supported at all edges. A concentrated force P is applied at the
center of the plate.















Due to symmetry, only one quarter of the plate is analysed, see Figure 3.6.6. By
adopting the assumption on small displacements and strains, analytical results for the
displacement at the center of the plate can be obtained thanks to Timoshenkos plate
theory. These analytical results are:
for the clamped plate: D Pa w / 0056 . 0
2
max
= ,
for the simply supported plate: D Pa w / 0116 . 0
2
max
= ,
(3.6.1a)
(3.6.1b)
where ) 1 ( 12 /
2 3
= Eh D is the flexural rigidity of the plate cross-section.
The analytical solution with the assumption of small displacements and strains
offers a reference to examine the correctness of the numerical results. The problem is
modeled by a coarse mesh of 551 elements. In both clamped and supported cases, the
ANS and ANSn give the same results, similarly, the SS7 and SS7n give the same results.
Hence, in this example we only expose results of the ANSn and SS7n.
Consider first the case where the plate is clamped, Figure 3.6.7. At a rather
moderate loading (P < 100), displacements and strains in the plate remain small. Hence
numerical results from large strain version should match the analytical solution. Results
are also compared to results from reduced integration (RI) element of Li and Cescotto,
[LI97], with automatic hourglass control. Notice that if the RI results match quite well
Figure 3.6.5: Two warping modes of EAS3v6s & Q1
1
2
5 6
4
3
8
7
1 2
5 6
4
3
8
7
Figure 3.6.6: Square plate geometry
a
a
h

P
85
the EAS15 and SS7n, this is at the expense of 3 layers over the plate thickness for the RI
while only one layer is required by other elements. The elements derived from the RI
technique undergo Poisson thickness locking. Therefore, for RI elements, more than one
layer should be used over thickness in order to obtain a good result of stress distribution
along thickness in bending dominated problems.































The EAS15, RI and SS7n elements are very satisfactory by approaching the
analytical curve up to a loading level about P = 100. With increasing loading,
geometrically nonlinear effects become important and this due to membrane effects
makes the plate stiffer. It explains why all numerical results are lower than the theoretical
ones. Also, large strains might happen and this will have an influence on the behavior of
the structure. The accuracy of the alternative ANS technique is confirmed when a very
good agreement is found in comparison of the SS7n element with the EAS15 and RI
elements being developed in the framework of large displacement and strains.
Consider now the case when the plate is simply supported, Figure 3.6.8. The
results of the EAS15, SS7n and the RI are rather close to each other. However, while the
EAS15 and SS7n results approach the analytical solution at a moderate loading (P < 100)
0

0.2
0.4
0.6
0.8
1.0
1.2
0 100 200 400 500
ANSn
SS7n
1.6
1.4
C
e
n
t
r
e

d
i
s
p
l
a
c
e
m
e
n
t

w

(
m
m
)

Timoshenko
EAS15
RI
Applied force P (N)
Figure 3.6.7: Displacement versus applied force for fully clamped plate
1
2
3
4
5
1
2, 3, 4
5
300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
0 50 100 150 200 250 300 350
ANSn
EAS15
Timoshenko
RI
SS7n
1
2
3
4
5 1
2, 3
4
5
Figure 3.6.8: Displacement versus applied force for simply supported plate
C
e
n
t
r
e

d
i
s
p
l
a
c
e
m
e
n
t

w

(
m
m
)
Applied force P (N)
86
as expected, it is not the case of the RI element. The deviation of the RI results from the
analytical solution at small strains. When the applied force is over 150N, the EAS15 and
SS7n continue giving the same result but they are a little stiffer than the RI element.
The behavior of the EAS15 in both clamped and simply supported cases are totally
identical to those values of METAFOR [BUI02]. This assures the quality of our EAS
implementation in the MATLAB code.



CONCLUSION
The ANSn and SS7n elements, which base on the alternative ANS technique,
satisfy the patch tests. By analysing eigenvalues we see that they are free from volumetric
locking and transverse shear locking.
The solid-shell elements are full-integrated scheme hence the stiffness matrix is
stable; it means there is not hourglass modes. It should be noted that the derivation of
the solid-shell stiffness matrix is carried out on the flattened (unwarped) solid geometry,
then globalized to the actual geometry. If the element is too warped or tapered, certain
tests are only approximately satisfied. Poor results can be expected if the element is
excessively warped. Unlike others plane-stress shell formulations for metal forming
simulation, the solid-shell elements provide a natural and efficient way for shell contact
problem since double-side surfaces of shell are available and the transverse normal stress
is included.
Featuring an appropriate combination of the ANS and the EAS methods dedicated
to alleviate locking effects, the solid-shell elements are free from locking due to parasitic
shear strains, distorted geometries and incompressible materials. Through an
investigation of the ANS method in the removal of shear locking, an alternative scheme
of sampling points, which enables a linear distribution instead of a constant value of
shear strains, is developed in this chapter. As it was revealed by numerical results, the
solid-shell elements with this alternative scheme offer a comparable performance in
comparison with that employed the classical ANS scheme.
In the next chapter performances of the ANS and ANSn, the SS7 and SS7n solid-
shell elements for elastic applications are continuously investigated with various linear
and nonlinear tests.












87
Chapter 4. ELASTIC APPLICATIONS


INTRODUCTION
In Chapter 3 a solid-shell element, which integrates the alternative ANS technique,
has been developed. In this chapter, various numerical tests are presented to demonstrate
the capabilities of the proposed solid-shell element. For the sake of clarity, all denotations
for elements in Chapter 2 and Chapter 3 are recalled below.
The standard solid element with full integration is designated by Q1, while the
solid element, which employs the classical ANS technique [DVO84] for alleviating
transverse shear locking and curvature locking, is designated by ANS. The solid-shell
element described in Chapter 3, which adopts 7 EAS parameters and the classical ANS
techniques (3.4.3), is designated by SS7, while the solid-shell element with the
alternative ANS technique presented in Chapter 3 (3.3.25) is designated by SS7n. The
additional letter n stands for the elements that employ the alternative ANS technique.
The EAS elements are designated by EASx, where x is the number of internal
parameters. For linear and nonlinear elasticity tests in this chapter, all of these elements
are implemented in a MATLAB code.
Table 4.0.1: Summary of employed elements
Name Type Description
Q1 Standard solid element Only compatible strains
ANS Classical ANS element Linear transverse shear strains (3.4.3)
ANSn New ANS element Bi-linear transverse shear strains (3.3.25)
EASx Enhanced assumed strain element x enhanced modes (Table 2.4.2)
SS7 Solid-shell element 7 enhanced modes (3.4.7)
Linear transverse shear strains (3.4.3)
SS7n New solid-shell element 7 enhanced modes (3.4.7)
Bi-linear transverse shear strains (3.3.25)


4.1 LINEAR APPLICATIONS
In this section we investigate performances of the ANSn and SS7n elements.
Various tests, which include shear, membrane and volumetric locking, are taken into
considered. All of tests in this section are linear problems. Nonlinear problems are
considered later, in section 4.2.
4.1.1 Cantilever beam under pure bending
Consider a cantilever of dimension 1011 clamped at left end and loaded by a
constant moment (induced by forces P = 0.5) at right end, Figure 4.1.1. This test
presented by Chandra and Prathap [CHA89]. The elastic modulus is E = 10
6
and
Poissons ratio is = 0.0. From the theory of the strength of materials, the analytical
solutions for vertical displacement w and maximum normal stress are:
88

6
ith 6 ; 10 6
2
2
max
4
2
bh
W w
W
M
EI
ML
w = = = = =

(4.1.1)
A mesh of one element is used. Since the structure undergoes pure bending, an
enhancement in volumetric locking is quite useless. Therefore, we will employ here 3
shear-enhanced modes. With this rather moderate enhancement, the EAS3s is already
able to attain the expected results both in displacement and normal stress. The stress
results in Table 4.1.1 are averages of absolute values at upper and lower surfaces of the
cantilever. Except the Q1 standard element is too stiff due to shear locking, all the other
elements, ANS, ANSn, SS7 an SS7n, are shear locking free.












Table 4.1.1. Normalized results
Q1 EAS3s ANS ANSn SS7 SS7n
w 0.0195 0.9998 0.9998 1.0000 0.9998 0.9998

max
0.0195 0.9998 0.9998 0.9998 0.9998 0.9998


4.1.2 Twisted beam with warping effects
In order to test the warping effect on elements McNeal and Harder [MAC85]
proposed the twisted cantilever in bending in-plane direction (P
v
= 1) and out-of-plane
direction (P
h
= 1), Figure 4.1.2. The cantilever length is L = 12, the width is w = 1.1 and
the thickness is t = 0.32. The cantilever is twisted 90 from root to tip. Youngs modulus
is E = 2910
6
and Poissons ratio is = 0.22.














P
P
Figure 4.1.1: Cantilever under pure bending
P
P
L=10
h
=
1

b=1
Y
X
Z
P
h

P
v

Figure 4.1.2: Twisted beam
w
t

89




















































The cantilever is fixed at left end and loaded by a unit force at right end. The
reference solutions are:
In-plane direction (P
v
= 1): w
ref
= 5.42410
-3
;
Out-of-plane direction (P
h
= 1): v
ref
= 1.75410
-3
.
The cantilever is modeled with a mesh of n11 elements, where n is number of
elements along the cantilever length.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1 2 3 4 5 6 7 8 9 10
ANSn
SS7n
EAS15
Q1
EAS21
1
2
3
4
1
2
3
4
5
5
Figure 4.1.4: Twisted beam, case of load along 0Y
N
o
r
m
a
l
i
z
e
d

t
i
p

d
e
f
l
e
c
t
i
o
n

Number of elements (n11)
Figure 4.1.3: Twisted beam, case of load along 0Z
N
o
r
m
a
l
i
z
e
d

t
i
p

d
e
f
l
e
c
t
i
o
n

Number of elements (n11)
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
1 2 3 4 5 6 7 8 9 10
ANS; ANSn
SS7; SS7n
EAS15
Q1
ASQBI
NEXHEX
1
2
3
5
6
7
1
2
3
4
5
6
EAS21 3
90
Numerical results for the case of bending in-plane direction are listed in Figure
4.1.3, where results of ASQBI element of Belytschko and Bindeman [BEL93] and
NEWHEX element of Fredriksson and Ottosen [FRE07] are also taken into considered.
Both ASQBI and NEWHEX elements are assumed strain elements where element
underintegration is adopted. The SS7 and SS7n elements converge with very coarse mesh
(n = 2) while the convergence with other elements requires finer mesh: n = 3 for ASQBI
and EAS21, n = 10 for NEWHEX and n = 9 for the EAS15. The ANS and ANSn
elements also converge as quickly as SS7. The Q1 element continues to show poor
performance in this bending case as usual.
Numerical results for the case of bending in out-of-plane direction are listed in
Figure 4.1.4. In this case, all of the elements (ANSn, SS7n and EAS) only converge with
rather fine mesh (n 5). However, with coarse meshes, the ANSn and SS7n elements
give better results than the EAS15. In contrary, the EAS21 converges as fast as the SS7n.
The Q1 element suffer shear locking hence show poor performance in this bending case.
Solid-shell in Samcef is a volumetric, quadrature shell element which can perform
the thickness deformation [JET08]. That element adopts the ANS method for shear
locking removal. Furthermore, 12 EAS modes are also adopted to remove other locking
effects. That solid-shell element passes the bending and membrane patch tests [JET08].
For this twisted beam problem, the solid-shell in Samcef converges with a mesh of
1221 elements (for both bending in-plane direction and bending out-of-plane
direction).

4.1.3 Clamped and simply supported plates under uniformed pressure
A square plate of dimension aah with fixed thickness h = 1 and various values
of width a = (10, 100, 1000) is modeled. All edges of the plate are clamped or simply
supported. A uniform pressure p = 1 loads on the upper face of the plate. This test is
presented by Chandra and Prathap [CHA89].











The symmetry of the structure allows to simulate a quarter of the plate, see Figure
4.1.5. The material properties are Poissons ratio = 0.3 and the elastic modulus E is
artificially dependent on the length a , see (4.1.3). The analytical result on the centre
deflection can be obtained [TIM59]:
Figure 4.1.5: Square plate under uniformed pressure
a
a
h

p
91

D
pa
w
4
01 . 0 = with
) 1 ( 12
2
3

=
Eh
D (4.1.2)
where D is the flexural stiffness of plate and dimensionless deflection depends on
boundary condition, see Table 4.1.2.
In order to respect the condition of small displacements, the centre deflection has
to be limited to a small value. Here, a typical value w = h/1000 is respected.
Correspondingly, the Youngs modulus takes the following value from (4.1.2):

2 2 4
3 3
(1 ) (1 )
12 12 0.01
D pa
E
h h w

= = (4.1.3)
Table 4.1.2. Dimensionless deflection (from [BLE00] and [TIM59])
h/a 0.1 0.01 0.001 0.0001
Plate thick thin very thin very thin
Clamped

0.15 0.126 0.126 0.126
Simply supported

0.424 0.406 0.406 0.406

For the comparison reason, the numerical dimensionless coefficient
num
, which is
defined below, derived from (4.1.2), will be employed:

4
100
pa
D w
num
num
=
(4.1.4)
Consider first the case where all the edges of the plate are clamped, Table 4.1.3.
Numerical analysis shows that the EAS15, SS7 and SS7n elements give very satisfactory
results for all cases ranging from thick to thin plates. While the Q1 element delivers a
poor prediction, especially when shear locking and Poisson thickness locking (due to
Poisson ratio is different from zero) becomes important with a decrease of the plate
thickness. The EAS9 element only gives good results for thick plate. When the aspect
ratio (a/h) is over 100, the EAS9 is too much worse than the ANS and ANSn elements.
However, the ANS and ANSn elements exhibit a stable tendency of convergence but
cannot reach the desired value because Poissons ratio is not equal to zero.
Table 4.1.3. Normalization of dimensionless deflection -
Clamped plate ( = 0.3)
h/a 0.1 0.01 0.001 0.0001
Mesh 22 44 22 44 22 44 22 44
Q1 0.38540 0.63140 0.00667 0.02428 0.00007 0.00025 0.007e-4 0.002e-3
EAS9 0.84413 0.95780 0.08126 0.59397 0.00087 0.01452 0.088e-4 0.148e-3
EAS15 0.93347 0.96627 0.96897 0.99579 0.96754 0.99413 0.95682 0.99960
ANS 0.78220 0.81200 0.78881 0.81254 0.78738 0.81095 0.79269 0.83643
ANSn 0.78220 0.81200 0.78881 0.81254 0.78738 0.81095 0.78809 0.81333
SS7 0.92860 0.96507 0.96278 0.99421 0.96103 0.99262 0.95087 1.02436
SS7n 0.92853 0.96507 0.96278 0.99421 0.96103 0.99262 0.96992 0.99897
92
Lets pass now to the case where the plate is simply supported at all edges, see
Table 4.1.4. Locking response occurs in the case of Q1 - standard element. The EAS15
element show again its performance with a quick convergence toward the expected
results but the EAS9 element continues to give poor result at high aspect ratio. The SS7n
and SS7 solid-shell elements show good performance as the EAS15 at coarse mesh for
thick and thin plate. As expected, the ANSn and ANS element behave better than the
EAS9, as in the clamped case when the aspect ratio (a/h) is over 100.

Table 4.1.4. Normalization of dimensionless deflection -
Simply supported plate ( = 0.3)
h/a 0.1 0.01 0.001 0.0001
Mesh 22 44 22 44 22 44 22 44
Q1 0.45014 0.69738 0.00916 0.03543 0.00009 0.00037 0.009e-4 0.004e-3
EAS9 0.95395 1.03089 0.14190 0.73649 0.00165 0.02737 0.165e-4 0.281e-3
EAS15 1.00977 1.03632 0.98173 0.99728 0.98097 0.99624 1.00235 0.99947
ANS 0.82675 0.84711 0.80011 0.81377 0.79944 0.81294 0.84025 0.78776
ANSn 0.82666 0.84653 0.80011 0.81377 0.79945 0.81294 0.83871 0.83123
SS7 1.00662 1.03589 0.97831 0.99648 0.97753 0.99544 0.99989 0.95143
SS7n 1.00649 1.03486 0.97831 0.99648 0.97754 0.99543 1.06946 1.04330


Table 4.1.5. Normalization of dimensionless deflection -
Clamped plate ( = 0.0)
h/a 0.1 0.01 0.001 0.0001
Mesh 22 44 22 44 22 44 22 44
Q1 0.30742 0.60525 0.00469 0.01718 0.00005 0.00017 0.047e-5 0.002e-3
EAS9 0.78011 0.91660 0.05836 0.50669 0.00062 0.01024 0.620e-5 0.103e-3
EAS15 0.90605 0.92896 0.98001 0.99804 0.97902 0.99690 0.97884 1.00560
ANS 0.88575 0.92410 0.95492 0.99190 0.95391 0.99079 0.95011 0.97469
ANSn 0.88571 0.92408 0.95492 0.99192 0.95391 0.99079 0.93525 1.00998
SS7 0.88575 0.92410 0.95492 0.99192 0.95391 0.99079 0.95666 0.96606
SS7n 0.88571 0.92408 0.95492 0.99192 0.95391 0.99079 0.94112 0.97986


Table 4.1.6. Normalization of dimensionless deflection -
Simply supported plate ( = 0.0)
h/a 0.1 0.01 0.001 0.0001
Mesh 22 44 22 44 22 44 22 44
Q1 0.39922 0.75090 0.00644 0.02526 0.00006 0.00026 0.006e-4 0.003e-3
EAS9 0.94032 1.04521 0.10388 0.66287 0.00116 0.01932 0.116e-4 0.197e-3
EAS15 1.02024 1.05355 0.98788 0.99888 0.98710 0.99770 1.02352 1.03863
ANS 1.00826 1.05032 0.97435 0.99565 0.97357 0.99446 1.08355 0.99346
ANSn 1.00769 1.04795 0.97435 0.99565 0.97355 0.99447 0.96104 1.01437
SS7 1.00951 1.05249 0.97435 0.99565 0.97358 0.99446 1.01366 0.96576
SS7n 1.00865 1.04932 0.97435 0.99566 0.97356 0.99447 1.00054 0.99384
93
In order to check performance of elements when there is no Poisson thickness
locking, lets impose = 0.0, see Tables 4.1.5 and 4.1.6. For both clamped and simply
supported cases, the EAS9 element continues to give poor results when a/h > 100. Hence
we can conclude that due to shear locking the EAS9 element shows bad performance.
The ANS and ANSn elements are only shear locking free. When there is no Poisson
thickness locking they behave as well as the SS7 and SS7n. Due to free from Poisson
thickness locking, performance of the other elements (SS7, SS7n and EAS15) are almost
similar to the case where Poisson thickness locking exists.

4.1.4 Square clamped plates with concentrated loads
A square plate of dimension aah = 1001001 is clamped at all edges and
loaded by a concentrated load P = 16.367 at the center. Material properties are E = 10
4

and = 0.3. Due to symmetry one quadrant of the plate is modeled with one layer of 22
and 44 elements. The structure has been modeled with EAS elements by Andelfinger
and Ramm [AND93].
By adopting the assumption on small displacements, analytical results on the
center displacement can obtained thanks to the Kirchhoffs plate theory:

D
Pa
w
2
0056 . 0 = with
) 1 ( 12
2
3

=
Eh
D (4.1.5)
where D is the flexural stiffness of the plate.

Table 4.1.7. Center deflection
Mesh 22 44
Q1 0.00683 0.02516
EAS9 0.07342 0.55351
EAS15 0.87357 0.97050
ANS 0.71124 0.79194
ANSn 0.71124 0.79194
SS7 0.86773 0.96848
SS7n 0.86773 0.96848
Theory 1.0


Applying (4.1.5), the theoretical displacement is imposed w = 1 at the center of the
plate. Without the presence of geometrical (nonlinear) stiffness the EAS15, SS7n and
SS7 elements give very good approximation, see Table 4.1.7.
This thin plate bending problem is modeled by coarse meshes, hence, the Q1
standard element is too stiff. Better results are given by the EAS9 element but with both
very coarse mesh 22 and finer mesh 44 the EAS9 is always worse than the ANSn and
ANS elements. The solid-shell elements who adopt the classical ANS technique (ANS
and SS7) or the alternative ANS technique (ANSn and SS7n) behave identically in this
test.

94
4.1.5 Simply supported square plate with various thickness and distorted mesh
This test illustrates the distorted mesh insensitivity of the solid-shell elements and
their performance in shear locking and Poisson thickness locking removal at high aspect
ratio (a/h 2000). Consider a plate of dimensions aah (Figure 4.1.6) with a = 20 and
different values of the thickness h (Table 4.1.8). The plate is simply supported along the
four edges and loaded by a unit concentrated force at the center. Youngs modulus is E =
10
7
and the Poissons ratio is = 0.25. The problem is symmetric, hence, only one
quarter of the plate is considered in two cases: regular mesh and distorted mesh of 44
elements (Figure 4.1.6).

















Table 4.1.8. Regular mesh - Normalization of displacement
h Q1 ASQBI NEW-
HEX
EAS9 ANSn,
ANS
SS7n,
SS7
1.5 0.667 1.089 1.089 1.070 0.965 1.080
1.0 0.454 1.039 1.039 1.014 0.921 1.034
0.5 0.173 1.009 1.009 0.935 0.894 1.005
0.1 0.008 0.999 0.999 0.377 0.885 0.995
0.05 0.002 0.999 0.999 0.133 0.884 0.995
0.01 8.4e-5 0.999 0.999 0.006 0.884 0.995

Table 4.1.9. Irregular mesh - Normalization of displacement
h Q1 ASQBI NEW-
HEX
EAS9 ANSn,
ANS
SS7n,
SS7
1.5 0.660 1.090 1.090 1.064 0.964 1.079
1.0 0.449 1.040 1.040 0.994 0.920 1.033
0.5 0.172 0.990 0.990 0.873 0.893 1.003
0.1 0.008 0.680 0.700 0.236 0.883 0.994
0.05 0.002 - - 0.073 0.883 0.993
0.01 8.5e-5 - - 0.003 0.883 0.993

Figure 4.1.6: A quarter of the plate - distorted mesh
(5,5,h/2)
x
y
O
6
0
0
6
z
P/4
95
The analytical solution of displacement at the center of this problem is given by
(3.6.1b). Numerical solutions are listed in Table 4.1.8 for the regular mesh and in Table
4.1.9 for the distorted mesh.
Results from the SS7n and ANSn elements are compared with results of the EAS9,
ASQBI of Belytschko and Bindeman [BEL93] and with NEWHEX of Fredriksson and
Ottosen [FRE07]. When the aspect ratio is less than 40 (i.e., h > 0.5), all the elements,
except Q1, give good results for both regular and distorted meshes. When the aspect ratio
is larger than 200 (i.e., h < 0.1) ASQBI and NEWHEX only give good results for regular
mesh while results of the EAS9 deteriorate rapidly even for regular mesh. On the
contrary, the behavior of ANSn and SS7n is stable with both distortion and high aspect
ratio (a/h = 2000). These elements can thus be considered as robust.

4.1.6 Pinched cylinder with rigid end diaphragms
Consider a cylinder of inner radius r = 300, thickness t = 3 and length L = 600, see
Figure 4.1.7. Youngs modulus is E = 310
6
and the Poissons ratio is = 0.3. The
concentrated forces F = 1 apply at the mid-length of the cylinder.

















Reference deflection w = 1.824810
-5
is coincident with the loaded points. Due to
symmetry only one-eighth of the cylinder is modeled. The structure is dominated by
inextensional bending response hence membrane locking may occur. Furthermore, the
thin structure with highly curve geometry also cause curvature thickness locking and
shear locking.
In this test a solid-shell element, ANS3DEAS, of Hauptmann and Schweizerhof
[HAU98] is used for comparison. The ANS3DEAS adopts the ANS method for
transverse shear locking removal and adopts the EAS method to enhance the membrane
strains. The ANS3DEAS element is superior in membrane dominated problems as stated
in [HAU98]. Mindlin shell and the solid-shell element of Samcef [JET08], see Section
4.1.2, are also taken into comparison.
Figure 4.1.8 shows that the vertical displacements of the alternative ANS method,
ANSn element (only with N < 100) and SS7n element, are better than ANS3DEAS and
Figure 4.1.7: Pinched cylinder with two rigid end diaphragms
L
F
F
r
L
/2
96
the EAS elements. The SS7n, EAS15 and ANS3DEAS elements converge with very fine
mesh (3232). The Samcef solid-shell element behaves as well as the SS7n element. The
Mindlin (Samcef) element converges at very coarse mesh; but does not provide the
correct result when the mesh is finer.





















4.1.7 Morley spherical shell
A benchmark test for shell elements of McNeal and Harder [MAC85] is
considered. The structure consists of a thin hemispherical shell, Figure 4.1.9. The middle
radius of the shell is R = 10, the thickness t = 0.04. Material properties are elastic
modulus E = 6.82510
7
and Poissons ratio = 0.3. Concentrated loads F of opposite
signs position at every 90 in the equatorial plane.
Table 4.1.10. Morley spherical shell - Normalized displacements at test point
Mesh Q1 EAS9 EAS15 ANSn,
ANS
SS7n,
SS7
22 1.064e-4 3.192e-4 4.255e-4 0.985 1.053
44 0.001 0.010 0.040 1.022 1.036
88 0.003 0.163 0.746 0.997 1.003
1616 0.010 0.750 0.989 0.991 0.998
3232 0.038 0.96 0.984 0.979 0.984

The theoretical displacement of test point is u = 0.0940. Because of the symmetry,
a one-fourth of the structure needs to be modeled. One element over thickness will be
fixed for all computation while different kinds of mesh in other directions will be tried.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1 10 100 1000
EAS9
EAS15
ANSn
SS7n
Figure 4.1.8: Convergence investigation for the pinched cylinder
Elements per area, N
N
o
r
m
a
l
i
z
e
d

d
i
s
p
l
a
c
e
m
e
n
t

ANS3DEAS
[HAU98]
1
2
3
4
5
1
2
3
4
5
Solid-shell
(Samcef)
6 Mindlin
(Samcef)
7
97
Computed results with the EAS elements as well as the solid-shell elements are
compared.







































As the spherical shell is very thin, 3D modeling may lead to an ill-conditioned
problem, since the distance between two corresponding nodes in the thickness direction is
too small in comparison to other directions in the case of a too coarse mesh. Hence, a
refinement of mesh will be helpful to handle this problem. Figure 4.1.10 shows results of
the solid-shell elements versus the EAS and standard elements with variety of meshes.
The EAS9 converges to the exact solution for a very fine mesh (3232) while the EAS15
converges for a coarser mesh (1616). However, the solid-shell elements give good
Figure 4.1.9: Morley spherical shell
Z
Y X
F=1
F= -1
(on quadrant)
0
Test point
S
y
m
S
y
m
Free
1
8

0
Free
R
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 5 10 15 20 25 30 35
Q1
EAS9
EAS15
ANSn
SS7n
Figure 4.1.10: Convergence of finite element solution
Element per side
N
o
r
m
a
l
i
z
e
d

d
e
f
l
e
c
t
i
o
n

1
2
3
4
5
1
2
3
4
5
98
results even at extremely coarse mesh (22). The classical ANS elements, ANS and SS7
elements, and the alternative ones, ANSn and SS7n elements, are in this case, totally
equivalent.

4.1.8 Thick-walled cylinder
Expansion of a thick cylinder with various Poissons ratio ( = 0-0.4999) as
described in [MAC85] is considered to investigate the performance of the solid-shell
elements in volumetric locking conditions. Elastic modulus is E = 1000. The inner radius
of the cylinder is R
i
= 3.0, the outer radius is R
o
= 9.0 and the thickness is t = 1.0, see
Figure 4.1.11. Plane strain conditions are assumed in the thickness direction. The inner
surface of the cylinder is loaded by a pressure q = 1/unit area.
A part of the cylinder, as described in Figure 4.1.11, is modeled by a 51 mesh.
The analytical solution of the problem is given by
] ) 2 1 ( / [
) (
) 1 (
2
2 2
2
r r R
R R E
pR
u
o
i o
i

+
=
(4.1.6)
Numerical solutions, at the inner radius, of elements are tabulated in Table 4.1.11.
The results show that the ANS and ANSn elements deliver identical results. The same
remark is also shown by the SS7 and SS7n elements. Obviously, volumetric locking
response can be observed for the Q1 element. Since volumetric locking cannot be
removed by the ANS techniques, the ANS elements response is nearly as stiff as the Q1
standard element for this problem.


















This volumetric locking can be removed with the use of the EAS internal
parameters, or more precisely with the introduction of enhanced volumetric modes.
Indeed, the ANS technique combined with the EAS technique to result in the solid-shell
elements, SS7 and SS7n. The solid-shell elements are free from not only shear locking
but also volumetric locking, see Table 4.1.11. Moreover, numerical results show the
Figure 4.1.11: Thick-walled cylinder
o
i
R


=
3
R


=
9
0
.
5
0
.
7
1
.
5
5
1
0

1
1
S
y
m
S
y
m
99
EAS9 elements, with 3 volumetric locking modes, and the EAS21, with 9 volumetric
locking modes, give almost exact results as the SS7n.
Table 4.1.11. Normalized radial displacement at R
i


Q1 ANSn,
ANS
EAS9 EAS21 SS7n,
SS7
0.0 0.993 0.996 0.998 0.998 0.998
0.25 0.989 0.992 0.996 0.996 0.996
0.30 0.987 0.990 0.995 0.995 0.995
0.49 0.845 0.848 0.991 0.991 0.991
0.499 0.358 0.360 0.990 0.990 0.990
0.4999 0.053 0.053 0.990 0.990 0.990


4.2 NONLINEAR APPLICATIONS
In this section we investigate performances of the ANSn and SS7n elements in
nonlinear problems. All locking effects (shear locking, membrane locking and volumetric
locking) are presented in problems.
4.2.1 Cantilever in large displacement
Consider a cantilever under transverse line load, see Figure 4.2.1. Geometry of the
cantilever are Lbh = 1010.1. The elastic modulus is E = 2.110
5
and Poissons ratio
is = 0.3. The cantilever is clamped at one end and suffer a line load q
0
= 1 at the other
end. Reference solution is numerical results of ANS3DEAS element (see Section 4.1.6)
of Hauptmann et al. [HAU01].













A discretization with a mesh of 1011 elements is used. The load - displacement
diagram in Figure 4.2.2 is obtained by using ten equal load steps. It is shown that when
the applied load is small or moderate,
0
3 . 0 q q < , the EAS15 element performs as well as
the SS7n and ANS3DEAS (Hauptmann et al. [HAU98]) elements while the ANSn
element is a little stiffer because of Poisson effect. When the applied load continues
increasing, the EAS15, due to transverse shear locking, becomes as stiff as the ANSn
element due to Poisson thickness locking. Meanwhile the other elements, SS7n and
ANS3DEAS, give identically better results than ANSn and EAS15 do.
L
b
h

q
0

Figure 4.2.1: Cantilever beam
test point
100

















4.2.2 Morley spherical shell large deformation case
The same data are given as the numerical test in Section 4.1.7 except the thickness
is thinner, t = 0.01 and the applied load is larger F=5.0. The mesh is composed of
16161 elements. The total load is applied in 15 equal steps. The problem was
considered by Vu-Quoc and Tan [QUO03a] and Klinkel et al. [KLI06] for investigating
behavior of their solid-shell elements. This test is considered as one of the most severe
bench-mark problems for nonlinear analysis of shell [QUO03a].




















The inward and outward displacements at the point A and B are plotted versus the
pinching load, see Figure 4.2.3. Both membrane and bending strains contribute to the
displacements at the load points. The structure is a doubly-curved shell with high aspect
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8
Q1
EAS15
ANSn, ANS
SS7n, SS7
ANS3DEAS
[HAU01]
Figure 4.2.2: Displacement of test point
w
A

q
0

q
1
2
3
4
5
1
2,5
3
4
Figure 4.2.3: Load-deflection curves for displacements at points A(u) and B(v)
EAS15-v
ANSn-u
ANSn-v
SS7n-u
SS7n-v
EAS15-u
u-[QUO03a]
v-[QUO03a]
Displacement u/R and v/R
P
i
n
c
h
i
n
g

f
o
r
c
e

F
/
2
1
2
3
4
5
6
7
8
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
1 2
3
4,8
5
6
7
101
B
A
Figure 4.2.4: Deformed hemisphere at F = 5
(without any magnification)
ratio (R/t = 1000), hence, curvature thickness locking, transverse shear locking and
membrane locking may simultaneously occur. From Table 4.2.1 the data show that the
presented solid-shell element (SS7n) is quite equivalent to the solid-shell element of Vu-
Quoc and Tan. Compare with the EAS9 and EAS15, the ANSn element, which is free
from shear locking and curvature thickness locking, deliver a very good result. Hence, we
can conclude that transverse shear locking and curvature thickness locking are the reason
for bad performance of the EAS9 and EAS15 elements in this problem.
















Table 4.2.1: Displacements due to pinched force F = 5
Element u
A
u
A
/u
ref
v
B
v
B
/v
ref

EAS9 2.30850 0.42100 1.69930 0.52117
EAS15 4.85402 0.88523 2.92634 0.89750
ANSn 5.29089 0.96491 3.19503 0.97991
SS7n 5.47885 0.99920 3.26135 1.00024
Reference
[QUO03a]
5.48331 1.00000 3.26055 1.00000


4.2.3 Slit annular plate under line force
A circular annular plate has a slit cut (line AB, Figure 4.2.5) along the radial
direction. The plate is clamped at one end of the slit and suffers a line force p = 0.8 at the
free end. The inner radius is R
i
= 6, the outer radius is R
o
= 10, the plate thickness is h =
0.03. Young modulus is E = 21.010
6
and Poissons ratio is = 0.0. The total load is
applied in 10 equal steps. A mesh of 6301 elements is used to model the plate.
Reference solution is numerical result of HS hybrid-stress solid-shell element of
Sze et al. [SZE02]. The HS element adopts the ANS method for transverse shear locking
and trapezoidal locking removals; and stress components are assumed independently
from the ones obtained from the displacement field. Numerical results of the HS element
are: vertical displacement at point A: w
A
= 13.618; vertical displacement at point B: w
B
=
17.257.
102






































The annular plate is a thin-walled structure (R/h > 300), hence, transverse shear
locking may appear. From numerical results in Figure 4.2.7, we see that the SS7n
element performs as well as the HS element of Sze et al. The ANSn element is slightly
stiffer than the SS7n and HS elements. Normally, when Poissons ratio is equal to zero, if
only transverse shear locking exists, the ANSn and SS7n should give similar results. In
this test, maybe membrane locking occurs, hence, the ANSn is little stiffer than the SS7n.
Due to transverse shear locking and high aspect ratio, the EAS9 element behaves too stiff
but still too much better than the Q1 standard element.
A
B
p=0.8
Fixed
Figure 4.2.6: The deformed configuration at maximum load
(without any magnification)
A B
p=0.8
Figure 4.2.5: Slit annular plate - initial configuration
R
i

R
o

103























CONCLUSION
In this chapter, various bench-mark tests (linear and nonlinear problems with
various locking effects) have been invoked to demonstrate performance for the SS7n
solid-shell element. The SS7n shows good performances for incompressible behavior and
for bending behavior of thin and thick-walled structures. However, all of the numerical
tests in this chapter are limited to linear material model. In comparison with the classical
shell elements, the solid-shell elements allow a straightforward integration of 3D material
models since they do not resort to the plane stress assumption. This advantage especially
becomes important for implementation of nonlinear material models. This argument is
assured in the next chapter, where performances of the SS7n element with nonlinear
material models are considered.












Q1-w
A

Q1-w
B

EAS9-w
A

EAS9-w
B

ANSn-w
A

ANSn-w
B

SS7n-w
A

SS7n-w
B

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 2 4 6 8 10 12 14 16 18
HS-w
A
[SZE02]
HS-w
B
[SZE02]
1
2
3
4
5
6
7
8
9
10
Vertical deflections at points A and B
N
o
r
m
a
l
i
z
e
d

o
f

l
o
a
d

Figure 4.2.7: Load-deflection curves for displacements at points A and B
1 2
3 4
5
6
7 8
9
10
104
Chapter 5. PLASTIC APPLICATIONS


INTRODUCTION
In the last chapter, applications of the SS7n element in linear elasticity and
nonlinear elasticity were presented. In order to exploit its performances in plasticity field,
the SS7n element has been also implemented in FEAP (Finite Element Analysis Program,
[TAY01]). In this chapter we investigate plastic behavior of the ANSn and SS7n
elements. First of all, available plastic theory in FEAP is briefly presented. Then,
numerical tests are investigated to look for differences in plastic behavior between ANS
and ANSn techniques. To carry out that work, stresses in largely plasticity-deformed
structures are analysed. Later, apart from the just mentioned tests, a special care is taken
for a springback simulation. Springback or elastic recovery relates to the change in shape
between the fully loaded and unloaded configurations that the material encounters during
a stamping operation. This results in the stamping component being out of tolerance and
can create major problems in the assembly or installation. Springback prediction of sheet
metal after forming is an important issue in controlling the manufacturing processes. To
this end, a benchmark test for high strength steel will be investigated with the ANSn and
SS7n elements.


5.1 FINITE STRAIN THEORY
5.1.1 Multiplicative split
Consider a body which contains a line vector dX before deformation (initial
configuration). After deformed to the current configuration, the line vector dX is
transformed to dx by a transformation mapping F (Figure 5.1.1). The line vector dX has
undergone both elastic and plastic deformation to be transformed to dx. The intermediate
configuration is defined as in which the line vector dx has been unloaded to a stress free
state, characterized by line vector dp. In other words, the line vector dX in initial
configuration has undergone purely plastic deformation to become the line vector dp in
the intermediate configuration. That pure plastic transformation can be expressed by:
X F p d d
p
= (5.1.1)
And the pure elastic transformation from the intermediate configuration to the
current configuration is realized by:
p F x d d
e
= (5.1.2)
Then we can write:
X F F p F x d d d
p e e
= = (5.1.3)
The finite strain plasticity formulation relies on the local multiplicative
decomposition of the deformation gradient F that is derived from (5.1.3) as:
105

p e
F F F =
(5.1.4)
where
e
F is the deformation caused by the elastic stretching and rotation and
p
F is the
plastic deformation. This is the classical multiplicative decomposition of Lee, see
[SIM88a].



















Due to the total Lagrange formulation of the variational equations for the solid-
shell elements, the plasticity model hereafter will be formulated using the right Cauchy
strain tensor. As pointed out by Simo [SIM88a], the return-mapping algorithm of
infinitesimal plasticity can be carried over to the presented formulation without any
modification. With the hyperelastic models, the elastic predictor in the return-mapping
algorithm is exactly calculated by using the strain energy function. The Green-Lagrange
strain tensor is defined relatively to the reference configuration as:
) (
2
1
) (
2
1
I C I F F E = =
T
with F F C
T
=
) (
2
1
) (
2
1
I C I F F E = =
p p pT p
with
p pT p
F F C =
(5.1.5)
where C is the right Cauchy strain tensor.
Consider a general form of strain energy function:
) , (
p
C C W W = (5.1.6)
Assuming hyperelastic response, the second Piola-Kirchhoff stress tensor S is
defined as:

C
C C
S
p

=
) , (
2
W
(5.1.7)

dX
dx
dp
X
x
F
F
F
e
p
Z
Y X
0
Figure 5.1.1: Schematic diagram of multiplicative decomposition
Initial (reference)
configuration
Current
configuration
Intermediate, stress-free,
configuration
106
5.1.2 Yield condition
Consider a yield surface defined in strain space with a general functional form
given by
0 ) , , ( Q C C
p
(5.1.8)
where Q is a suitable set of internal plastic variable vector.
The evolution of the internal plastic variable vector Q can be determined by a rate
equation in the form:
) , , ( Q C C H Q
p

= (5.1.9)
where ) , , ( Q C C H
p
is a prescribed function of the generalized plastic hardening moduli;
the initial condition is 0 Q = at the reference configuration. The term is plastic
consistency parameter.

5.1.3 Flow rule
As in the infinitesimal theory, in this section, beginning with the principle of
maximum plastic dissipation, the evolution of plastic flow is expressed directly in terms
of kinematic variables related to the multiplicative decomposition.
Without loss of generality, the elastoplastic behavior is assumed to be
characterized by variables {C,C
p
,Q}. Furthermore, assume that an elastoplastic potential
function can be decoupled into internal-independent contribution ) , (
p C
W C C and
internal-dependent contribution ) (Q
Q
W as:
) ( ) , ( Q C C
Q p C
W W W + = (5.1.10)
The plastic dissipation at the state defined by {C,C
p
,Q} is:
) : : ( ) , ; , , ( Q
Q
C
C
Q C Q C C

=
W W
D
p
p
p p p

(5.1.11)
In local form, the maximum plastic dissipation formulated in strain space may be
stated as follows. Give a state {C,C
p
,Q} among all admissible right Cauchy strain tensors
satisfying the yield criterion, the actual strain tensor C is the one for which plastic
dissipation attains its maximum. Lets consider the maximum plastic dissipation in point
of view of optimization theory, the problem may be stated as:
Maximize { ] : : [ ) , ; , , ( Q
Q
C
C
Q C Q C C

p
p
p
p p p
W W
D

= }
subject to } 0 ) , , ( {
6
= Q C C C
p C
R K
(5.1.12)
where
C
K is the space of admissible right Cauchy strain tensors at fixed plastic variables
{C
p
,Q}; tensor C

is symmetric, hence, consists of 6 independent components.
As shown in (5.1.10) and (5.1.11), with a fixed set of {C
p
,Q} Q
Q

W
constant,
the maximum of
p
D only depends on the term ] :
) , (
[
p
p
p C
W
C
C
C C

. Hence, the
107
maximum of
p
D is equivalent to the minimum of ] :
) , (
[
p
p
p C
W
C
C
C C

. Thus, the
maximum plastic dissipation problem can be changed to:
Minimize { ] :
) , (
[
p
p
p C
W
C
C
C C

}
subject to } 0 ) , , ( {
6
= Q C C C
p C
R K
(5.1.13)



































We can solve the problem (5.1.13) by the method of Lagrange multipliers to
inequality constraints. The Lagrange functional for the problem (5.1.13) is defined:
) , , ( :
) , (
Q C C C
C
C C
p p
p
p C
p
W
L +

(5.1.14)
STRAIN-BASED ELASTOPLASTIC CONSTITUTIVE MODEL

1. Step 1: Multiplicative decomposition

p pT p
T
p e
F F C
F F C
F F F
=
=
=

2. Step 2: Hyperelastic stress-strain relations

C
C C
S
p

=
) , (
2
C
W

3. Step 3: constitutive tensors
Elastic:

C C
C
4

=
C
W
2
4
Plastic

p
C
W
C C
M

=
2
4
4. Step 4: Flow rule

C
Q C C
S
C C C M S

=
=
) , , (
2
: ) , (
2
1
:
p
p
p p p

5. Step 5: Hardening law
) , , ( Q C C H Q
p

=
6. Step 6: Loading/unloading conditions
0 ; 0 ) , , ( Q C C
p

0 ) , , ( = Q C C
p

Figure 5.1.2: Finite strain theory
108
where is positive and belongs to the set of square integrable L
2
() functions, that
defined by the positive cone
p
K :

} 0 ) ( {
2
= L K
p

(5.1.15)
Resulting from (5.1.14) and (5.1.15), the Kuhn-Tucker conditions are:

1 ( , , )
0 ( , ) : 2
2
0
( , , ) 0
( , , ) 0
p p
p p p
p
p
L

= = =

=
)

C C Q
S M C C C
C C
C C Q
C C Q

(5.1.16)
where the constitutive tensor
p
p C
p
W
C C
C C
C C M

=
) , (
4 ) , (
2
.
The flow rule (5.1.16a) and loading/unloading conditions (5.1.16b-d) are
associative with the multiplicative decomposition (5.1.4). The presented theory is
summarized in Figure 5.1.2.

5.1.4 Elastoplastic tangent moduli
The requirement for the load point to remain on the yield surface during plastic
deformation is called the consistency condition. It enables to determine the plastic
multiplier . This condition states that plastic loading ( 0 ) requires
0 / ) , , ( = t
p
Q C C as:
0 : : 2 : 2 ) , , (
) , , (
=

= =

Q
Q
C
C
C
C
Q C C
Q C C

p
p
p
p
t
(5.1.17)
or
] : : 2 [ : 2 Q
Q
C
C
- C
C

p
p

(5.1.18)
Calculate
p
C

from (5.1.16a) then introduce the result into (5.1.18) we can derive
the expression for the plastic consistency parameter:

H
Q C
M
C
C
C
: : : 8
: 2
1

(5.1.19)
Time differentiating of the elastic constitutive equation (5.1.7) we have:

2 2
4 : 4 : : 4
W W

= + =



p
p
S C C A C -
C C C C C
(5.1.20)
where
2
4
W

A=
C C
and
2
:
W

p
p
C -
C C C
from (5.1.16a).
By inserting (5.1.19) into (5.1.20) we have the expression:
C A S

:
ep
= (5.1.21)
109
with the elastoplastic tangent moduli:
H
Q C
M
C
C C
- A A
p
: : : 8
8
1

ep



5.2 J2 MATERIAL MODEL
5.2.1 Multiplicative split and elastic response
The material model described above will be applied for J2 materials. The J2
models are well suited to the materials whose elastic volumetric response is uncoupled
with elastoplastic deviatoric response, this behavior is observed in metal in plasticity, for
instance.
Denote J = det( F ), the volume-preserving part of the deformation gradient part is
defined:
F F
3 / 1
= J (5.2.1)
with det( F ) = 1, it means F satisfies the incompressible condition.
The right Cauchy strain tensors which are associated with F and F are also
defined:

2/ 3 T
J

= = C F F C det( C ) = 1
and similarly,
2/ 3 p pT p p
J

= = C F F C
(5.2.2)
Account for uncoupled volumetric/deviatoric response the energy function
(5.1.10) is in the form:
) ( ) , ( ) ( ) , , ( Q C C Q C C
Q p dev vol p
W W J W W + + = (5.2.3)
The uncoupled energy function in (5.2.3) results in uncoupled volumetric
deviatoric stress-strain relationships. In metal plasticity the plastic deformation is
isochoric, i.e. fully incompressible. While elastic deformation is compressible and small
in many applications. As an example, an energy function from [SIM92b] is consisted as:

) exp( ) ( ) ( 2 /
] ) ( ) ( ) [(
2 / )] [ln( 2 / ) (ln
0
1
0
2
2
3
2
2
2
1
2
3 2 1
2
p y y p y y p iso
Q
e e e dev
e e e e vol
H W
G W
K J K W



+ + =
+ + =
= =

(5.2.4)
where
- G and K are constant. They are shear modulus and bulk modulus, respectively;
- > 0 is saturation exponent;
- ) ln(
e
i
e
i
= ; i = 1,2,3;
e
i
are principal elastic stretches (from eigenvalues of C);
- ) det(
e e
J F = ;
-
0 y
and
y
are the first yield stress and the saturation yield stress;
-
p
is the equivalent plastic strain;
- H
iso
is the linear hardening modulus.

110
Following (5.1.7), the second Piola-Kirchhoff, for J2 material, is decomposed into
the hydrostatic and deviatoric parts as:

2/ 3
( ) ( , )
2 2
( , )
DEV[2 ]
vol dev
dev
vol dev
W J W
W
Jp J

= +

= + = +

p
p
-1
C C
S
C C
C C
C S S
C

(5.2.5)
where J J W p
vol
= ) ( is the hydrostatic pressure. In (5.2.5) we did use the relation (see
[HOL00], page 41):

1 -
C
C
J
J
=

(5.2.6)
and the derivative (see [HOL00], page 229):
]
3
1
[
3 / 2 1 -
C C I
C
C
=

J (5.2.7)
and the denotation

1 -
)] ( : [
3
1
) ( ] [ DEV C C = (5.2.8)
In (5.2.5), the hydrostatic stress is presented by the term
-1
C Jp , and ] [ DEV gives
the physically correct deviator stresses in the reference configuration.

5.2.2 Flow rule and yield function
The yield function, for J2 material model can be assumed to only depend on the
deviatoric part of right Cauchy strain tensor C as:
0 ) , , ( Q C C
p
(5.2.9)
According to the arguments in the last section with C , the Kuhn-Tucker
conditions (5.1.16) leads to a flow rule as:

2 3
2 2 2 DEV[ ]
0
( , , ) 0
( , , ) 0
p - /
p
p
J

= = =

=
)

C
S
C C C C
C C Q
C C Q

(5.2.10)
where, similar to (5.1.16a),
p
S

is also set:
] : 2 [ DEV
2
3 2 p
p
dev
/ - p
W
J C
C C
S

(5.2.11)
The von Mises yield condition in form of spatial terms in stress space is:
0
3
2
] [ dev ) , ( + =
y
q Q

iso
H q =
(5.2.12)
where:
-
y
is the yield stress and
111
- the Kirchhoff stress tensor is the push-forward of the second Piola-Kirchhoff
stress S by:

T
FSF = (5.2.13)
- the hardening variables are:
] , [ Q q =
(5.2.14)
with is back stress which defines the location of the center of the yield surface. This
back stress is used to account for Bauschingers effect (kinematic hardening). The
isotropic hardening is characterized by the internal variable q with isotropic hardening
modulus H
iso
> 0. H
iso
is a function of isotropic hardening variable for nonlinear
hardening laws, and for linear isotropic hardening laws H
iso
is a constant.

The algorithm for the presented formulation is straightforward as follows.
Integrate the material (reference) description of the flow rule (5.2.10) by an integration
scheme, such as backward Euler difference scheme. Substitute the result into the
hyperelastic stress-strain relations (5.2.5). The detailed algorithm is listed in Figure 5.2.1.
The implementation of the hyperelastic formulation of J2-flow theory reduces to the
classical radial return with the elastic predictor computed by energy function evaluation.
We see that maximum of plastic dissipation (5.1.12) leads to a return mapping
algorithm, see Figure 5.2.1. That algorithm will look for the solutions on a path that
makes the plastic dissipation stationary. Concretely, within a typical time step a trial
elastic is calculated first. Then, the actual stress is defined in the closest-point projection
of the trial state onto the elastic domain. For J2 flow theory, the closest-point projection
boils down to the classical radial return method.























112
















































RETURN-MAPPING ALGORITHM FOR J2-FLOW THEORY

1. Step 1: Geometry update

1 1 1
1
3 1
1 1
1
; GRAD
+ + +
+ + +
+
=
= + =
+ =
n
T
n n
n
/ -
n n n
n n
J
F F C
F F U F F
U x x


2. Step 2: Elastic predictor (k=0)

n n
p
n n
dev
n
vol
n
p
n
p
n
W
J
J W
J
Q Q
C
C C
C
S
C C
=

=
=
+
+ + +
+
+
) 0 (
1
) 0 (
1 1 3 / 2 1 3 / 2 ) 0 (
1
) 0 (
1
) , (
2
) (


3. Step 3: Check for yielding
) , , (
) 0 (
1
) 0 (
1
) (
1
) (
1 + + + +
=
n
p
n
k
n
k
n
Q C C
IF
) (
1
k
n+
< TOL THEN
Set
) (
1 1
) ( ) (
k
n n + +
= and EXIT (to the next time step)
ELSE continue Step 4 and Step 5

4. Step 4: Compute plastic consistency parameter (see (5.1.19))
: : 8 2
) (
1
1 3 / 2 ) (
1
) (
1
k
n
p
k
n
k
n
J
+

+ + (

= H
Q C
M
C


5. Step 5: Update state variables

) (
1
) 1 (
1
) (
1
) 1 (
1
) (
1
1 ) 1 (
1
3 / 2 ) (
1
) 1 (
1
) (
1
) 1 (
1
3 / 2 ) (
1
) 1 (
1

: 2
2
k
n
k
n
k
n
k
n
k
n
k
n
k p
n
k p
n
k
n
k
n
k
n
k
n
J
J
+
+
+ +
+
+
+
+
+

+
+
+
+
+
+

+
+
+
+ =

=
H Q Q
C
M C C
C
S S

Set k = k+1 and RETURN Step 3
Figure 5.2.1: Material stress update algorithm
113
5.3 PLASTICITY APPLICATIONS
In the sections following we investigate performances of the SS7n element in field
of plasticity.
5.3.1 Cantilever at large elastoplastic deformation
Consider a cantilever beam of dimension Lbh = 1510.7 clamped at left end
and loaded by a distributed force (q
0
= 7) at right end, Figure 5.3.1. This test was
presented by Huh and Kim [HUH00]. The elastic modulus is E = 210
5
N/mm
2
and
Poissons ratio is = 0.3. Material law is elastoplastic with isotropic hardening as
p
1000 340 + = .















As proposed in the work of Huh and Kim [HUH00], a very fine mesh of 40101
elements (along L, h and b, respectively) is first employed. By adopting the assumption
on a plain strain state, these authors can employ a 2D modeling. Here, with 3D solid
elements to simulate 2D modeling, the width of beam (Y direction) will be presented by
only one finite element and the plain strain state will be obtained by setting to zero
displacement in the width direction. For the comparison between different approaches,
the displacement at the tip point A will be considered. The SRI, EAS9, EAS21
(METAFOR software, [MET08]), ANS, ANSn, SS7 and SS7n elements will be involved
in the computation.
The SRI element is one of the best elements in removing volumetric locking.
However, shear locking in this problem is important. Consequently, in Figure 5.3.2, we
see that the SRI element with 10 layers (i.e. 40101 elements) is stiffer than the RI
element [HUH00] or the EAS9 with 2, 4 or 10 layers. Results given by the EAS9 with 10
layers are converged to the results given by the EAS21 with 4 or 10 layers, see Figure
5.3.3. If only 1 layer is adopted, both the EAS 9 and EAS21 elements show stiff
behaviour, see Figures 5.3.2 and 5.3.3. The EAS21 always gives very good results in
both shear and volumetric locking removals, [AND93]. Hence, with a very fine mesh of
40101 elements, result of the EAS21 (METAFOR) element is considered as a
reference. We see that this reference is a little softer than the result of the RI element in
[HUH00].
X
Y
Z
L=15
b
h

q
0
= 7
Figure 5.3.1: Cantilever at large elasto-plastic deformation
A
114
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
0 2 4 6 8 10 12
EAS9,401
EAS9,402
EAS9,404
EAS9,4010
1
2
3
4
1
2
3 4
RI(4010, [HUH00]) 5
SRI(4010, METAFOR) 6
6
Figure 5.3.2: EAS9 (METAFOR) results with various elements along thickness
1
2
3 4
5
6
Next, the deformation is calculated by the ANS method with various layers of
element along the thickness direction. Consider the load-displacement curve, the ANS
and ANSn elements give identical results hence only the results of the ANSn are
presented here. We see that with 2 layers, the ANSn element is softer than the SRI when
P = qb 4.0N. The reason is due to shear locking. At load level P > 4.0N, the cantilever
is in larger deformation, effect of volumetric locking become more important, hence,
behavior of two-layered ANSn element is stiffer than the SRI element, Figure 5.3.4. With
4 layers, the ANSn element is always softer than the SRI element. When volumetric
locking becomes more serious (because of large plasticity deformation), the ANSn
element with 4 layers approaches the behavior of the SRI element. Increasing number of
layers to 10, ANSn behavior becomes too soft. Compare to the reference result (EAS21),
behaviour of the ANSn is too stiff. We see that for this problem, the ANSn element,
which is only free from shear locking, should not be used.



















Now we investigate the problem with the SS7n element (because the SS7 and
SS7n elements give identical results). From Figure 5.3.5 we see that result of the SS7n
element, with 2 layers along the thickness, is only close to the result of the EAS21 when
shear locking is important (i.e., P 4.0N). When P > 4.0N, SS7n with 2 layers is a little
stiffer than the EAS21 and RI elements. Increasing layers along the thickness to 4 or 10,
the SS7n even becomes stiffer. If a very coarse mesh is adopted, i.e. with only 1 layer
along the thickness, the SS7n is too stiff. Hence, in this problem, 2 layers of SS7n are the
most suitable choice.
115









































1
2
3
4
5
RI(4010, [HUH00]) 6
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
0 2 4 6 8 10 12
EAS21,401
EAS21,402
EAS21,4010
EAS9,4010
EAS21,404
Figure 5.3.3: EAS21 (METAFOR) results with various elements along thickness
1
2
3
4
5
6
ANSn(4010) 1
ANSn(404)
0
1
2
3
4
5
6
7
0 2 4 6 8 10 12
ANSn(402)
2
3
EAS21(4010, METAFOR)
5
Q1(4010)
4
RI(4010, [HUH00])
6
1
2
7
5 4
6
Figure 5.3.4: ANS results with various elements along thickness
Displacement v (mm)
F
o
r
c
e


P
(
N
)

3
SRI(4010, METAFOR)
7
116
























It was presented in Chapter 3 that the ANS and ANSn elements are different in
assumed strained components E
xz
and E
yz
. These elements will give identical results if
transverse shear strains are not important. As in this problem, the ANS and ANSn
elements give identical results because the cantilever is not too thin (aspect ratio 20),
hence, transverse shear stresses given by the ANS and ANSn are similar, see Figures
5.3.6-7. Consequently,
xz
stress given by the SS7 and SS7n elements are similar, see
Figures 5.3.8-9.


Table 5.3.1. Convergence of
xz
(MPa)
4011 4021 4041 40101 40151
min max min max min max min max min max
EAS9 -400 369 -259 266 -250 245 -247 262 -246 250
EAS21 -400 369 -262 262 -251 248 -251 258 -246 250
SS7 - - -224 181 -235 226 -242 240 -246 243
SS7n - - -225 185 -235 225 -241 240 -245 243


Distribution of
xz
stress in a mesh of 40101 given by the EAS9 element
(METAFOR, Figure 5.3.10) and the solid-shell elements (SS7 and SS7n, Figures 5.3.8
and 5.3.9) are similar. However, the ultimate values given by the EAS9 element is larger
in comparison with the solid-shell elements (-247MPa vs. -240MPa and 262MPa vs.
240MPa). The convergence value of
xz
is found in Table 5.3.1 with both solid-shell and
EAS21 elements (bold letters) with mesh of 40151 elements. Data in Table 5.3.1
shows that the EAS21 and EAS9 elements go to the convergence from higher values. In
contrary, the SS7 and SS7n elements go to the convergence from lower values.
SS7n, SS7 (402)
SS7n, SS7 (404)
0
1
2
3
4
5
6
7
0 2 4 6 8 10 12
SS7n, SS7 (4010)
Figure 5.3.5: SS7n results with various elements along thickness
Displacement v (mm)
F
o
r
c
e


P
(
N
)

1
2
3
EAS21(4010, METAFOR) 5
RI(4010, [HUH00]) 6
1
2
3
5
6
SS7n, SS7 (401)
4
7
4
117
Figure 5.3.7: ANSn element -
xz
(mesh 40101)
Figure 5.3.6: ANS element -
xz
(mesh 40101)




















































118
Figure 5.3.9: SS7n element -
xz
stress (mesh 40101)
(the result is shown in the structural coordinate system)
Figure 5.3.8: SS7 element -
xz
stress (mesh 40101)
(the result is shown in the structural coordinate system)




















































119
Figure 5.3.10: EAS9 element -
xz
stress (METAFOR, mesh 40101)



















































120
5.3.2 Pinched cylinder at large elastoplastic deformations
Consider a cylinder of inner radius r = 300, thickness t = 3 and length L = 600,
Figure 5.3.11. Youngs modulus is E = 310
3
and Poissons ratio is = 0.30. A couple of
opposite, concentrated forces F applied at the mid-length of the cylinder. Both ends of the
cylinder are pinched, only a free movement in the axial direction Y is possible. Material
law is elastoplastic with isotropic hardening as 24.3 300
p
= + .














We investigate deflection coincident with the point load against loads. The
deflection response is strongly dominated by inextensional circumferential bending. Due
to symmetry, only one eighth of the cylinder is modeled. A result with 3D elements and
mesh of 40401 elements proposed by Wriggers et al. [WRI96b] will be employed as
the reference, see Figure 5.3.13. Wriggers did use 3D enhanced element presented by
Simo and Armero in [SIM92a]. That element is equivalent to the EAS12 (6v + 6s)
element, see Table 2.4.2. It means the reference element is volumetric and shear locking
free. Lets investigate convergence of the EAS12 element. We also see that the EAS12
element of METAFOR, Figure 5.3.12, converges with a mesh of 40401 elements. The
EAS12 (40401 elements) of METAFOR gives an identical result as the reference
[WRI96b], Figure 5.3.13.
















F
F
L
r
Figure 5.3.11: Pinched cylinder
X
Y
Z
Figure 5.3.12: EAS12 element (METAFOR)
0
500
1000
1500
2000
2500
3000
3500
0 50 100 150 200 250
1
2
3
4 EAS12 (40401)
EAS12 (32321)
EAS12 (24241)
EAS12 (16161)
1
2
3
4
121

















This cylinder is a thin shell structure (R/t = 100 >> 20) and subject to large
deformation. Consequently, transverse shear strains are important. In this problem results
of the ANS and ANSn elements are not totally coincident, see Figure 5.3.13.
Now we consider the resultant stresses from the ANS and ANSn techniques with a
coarse mesh: 16161 elements. Lets concentrate in ultimate transverse shear stresses at
the left end of the cylinder and at the area where the load is applied. All results are
considered in the structural coordinate system. Data in Figures 5.3.14 - 15 show the
xz

stresses calculated by the SS7 and SS7n elements. Reference result of the
xz
stress is in
Figure 5.3.20b. Pay attention that the concentrated load is only applied at the center node.
As a consequence, the element containing the center node is singular due to the applied
load is unphysical. Hence, in following analyses we will consider stresses at point B
instead of stresses at point A, see Figure 5.3.14.
When using the same coarse mesh (16161), the SS7n gives the results, which
are closer to the results of the reference - EAS12 (40401) - than the SS7 does, see
Table 5.3.2. In Table 5.3.2, transverse shear stresses of the EAS12 (16161) are
extracted from Figures 5.3.16 - 17. In that table, transverse shear stresses of the ANS and
ANSn elements (16161) are also introduced. We see that transverse shear stresses
given by the ANSn are closer to the reference than the result of ANS. Transverse shear
stresses predicted by the EAS12 (16161) are worse than the values predicted by the
SS7n (except the minimal value of
yz
), Table 5.3.2.

Table 5.3.2. Normalized transverse shear stresses
(reference values are results of EAS12 - 40401 element)
ANS
(16161)
ANSn
(16161)
SS7
(16161)
SS7n
(16161)
EAS12
(16161)
Reference
values
Max 0.854 0.854 0.695 0.894 0.642 15.1

xz

Min 0.291 0.648 0.559 0.862 0.836 -31.8
Max 0.569 0.643 0.816 0.863 0.580 25.5

yz

Min 0.557 0.708 0.587 0.925 1.042 -49.6

Data from Figures 5.3.18 -19 show
yz
stress calculated by the SS7 and SS7n
elements. Reference result of
yz
stress is in Figure 5.3.21b. When a coarse mesh is
6 EAS12 (METAFOR, 16161)
6
0
500
1000
1500
2000
2500
3000
3500
0 50 100 150 200 250
Displacement w
A
p
p
l
i
e
d

f
o
r
c
e

F
3D (Wriggers, 40401)
SRI (METAFOR, 40401)
Figure 5.3.13: Pinched cylinder: Force-Displacement
ANS (16161)
ANSn (16161) 1
2
3
4
1
2
3
4
EAS12 (METAFOR, 40401) 5
5
122
Figure 5.3.15: SS7n element -
xz
(mesh 16161)
adopted, compare with the reference stresses in areas I and II (Figures 5.3.18, 19 and
21b) we can see that the SS7n element give less worse results than the SS7 element. In
detail, stress in area I given by the reference is 10.4 while values given by the SS7 and
SS7n are 7.16 and 9.63, respectively. Stress in area II given by the reference is -5.0 while
values given by the SS7 and SS7n are -1.9 and -2.7 respectively. The maximal and
minimal values of
yz
are listed in Table 5.3.2.










































Figure 5.3.14: SS7 element -
xz
(mesh 16161)
A
B
123
Figure 5.3.17: EAS12 element -
yz
(EAS12,METAFOR, 16161)
(due to singularity at load point, min. value has been imposed to -51.7)
Figure 5.3.16: EAS12 element -
xz
(EAS12,METAFOR, 16161)
(due to singularity at load point, max. value has been imposed to 26.0)
















































124
Figure 5.3.19: SS7n element -
yz
(mesh 16161)
II
II
I
Figure 5.3.18: SS7 element -
yz
(mesh 16161)
I
II
II




















































125
Figure 5.3.20a: Reference -
xz
(EAS12,METAFOR, 40401)
Figure 5.3.20b: Reference -
xz
(EAS12,METAFOR, 40401)
(due to singularity at load point, max. value for drawing has been imposed to 43.3)




















































126
Figure 5.3.21a: Reference -
yz
(EAS12,METAFOR, 40401)




















































Figure 5.3.21b: Reference -
yz
(EAS12,METAFOR, 40401)
(due to singularity at load point, min. value for drawing has been imposed to -49.6)
II
II
I
127
Figure 5.3.22: SS7 element - von Mises stress (mesh 16161)
Figure 5.3.23: SS7n element - von Mises stress (mesh 16161)
As analysing above, the differences in transverse shear stresses calculated by the
classical ANS technique and the alternative ANS technique lead to difference in von-
Mises stress and Load-Displacement curve calculated by the SS7 and SS7n elements, see
Figures 5.3.22-24.














































128
Next, we compare results of the SS7 and SS7n elements with results of solid
element (EAS12) and shell element presented in [WRI96b]. The shell element in
[WRI96b] is based on a quasi-Kirchhoff-theory, which means that the assumption of the
classical Kirchhoff-Love kinematics is respected via a penalty constraint. A reduced
integration for the penalty term was applied in order to obtain locking-free behavior in
bending dominated problems.
In Figure 5.3.24 the load-displacements curves for different elements are plotted.
In general, the results of the SS7 and SS7n (16161) elements are a little stiffer than the
results of 3D calculation (EAS12, 40401 elements). However, with the same coarse
mesh (16161) the solid-shell elements, SS7 and SS7n, are comparable to the
conventional shell element (RI shell elements, [WRI96b]) and better than results of the
EAS12, Figure 5.3.24.
The SS7n element shows a load decrease (at w 200) as RI shell element does (at
w 210). This consequence arises as a result of the relatively coarse mesh. Hence, it
makes the plastic zone cannot developed continuously. This consequence can be canceled
by using a finer mesh as appear for the EAS12 results (1616 vs. 4040). In Figure
5.3.24 the curve of the SS7 element is smoother than the SS7n.






























0
500
1000
1500
2000
2500
3000
3500
0 50 100 150 200 250
Displacement w
A
p
p
l
i
e
d

f
o
r
c
e

F

3D (Wriggers, 40401)
RI Shell (Wriggers 16161)
Figure 5.3.24: Pinched cylinder solid-shell elements
SS7 (16161)
SS7n (16161) 1
2
3
4
1
2
3
4
Snap-through
EAS12 (METAFOR, 40401) 5
5
6 EAS12 (METAFOR, 16161)
1
2
3
4
5
6
6
129
For this problem, behavior of the EAS12 and EAS21 (with both coarse mesh and
fine mesh) are similar, see Figures 5.3.24 - 25. Hence, it is reasonable to use 12 internal
parameters for the EAS as Wriggers et al. [WRI96b].



















By this numerical test we see that the SS7n element predicts transverse shear
stresses better than the SS7 element, Table 5.3.2. For analyzing a thin-walled structure
with a coarse mesh, both solid-shell elements, SS7 and SS7n, can work as well as the
conventional shell elements, e.g. the shell element in [WRI96b]. The other solid-shell
elements, such as the solid-shell element of Samcef (20201 elements) [JET08]
provides a result which approximates the result of 3D element of Wriggers (40401
elements). Hence, using the solid-shell elements to simulate thin-walled structures will
provide a real 3D model with a cheap computational cost.


















0
500
1000
1500
2000
2500
3000
3500
0 50 100 150 200 250
Displacement w
A
p
p
l
i
e
d

f
o
r
c
e

F

Figure 5.3.24: Pinched cylinder solid-shell elements
SS7 (16161)
SS7n (16161) 1
2
3
4
1
2
3
4
EAS12 (METAFOR, 40401)
5
5
EAS21 (METAFOR, 40401)
EAS21 (METAFOR, 16161)
130
5.3.3 Springback of unconstrained cylindrical bending
The following benchmark initially proposed at the NUMISHEET02 conference
[NUM02] has been chosen as a reference case for comparing finite element formulations
with various time integration schemes. This benchmark is recommended to investigate
springback analysis and complex contact treatment. It consists of an initially flat blank
bent into a cylindrical shape and then unloaded. Because there is no blankholder, the
problem is called unconstrained bending. The initial geometry and the loading process
are described in Figure 5.3.26 (left side). The loading process is stopped when punch and
die are concentric (right side of Figure 5.3.26).
















After the loading phase, the unloading phase takes place and some springback
occurs. The amount of springback is quantified in the following way, see Figure 5.3.27.
- Angle between line AB and line CD (Figure 5.3.27a) before and after springback
at the final stroke of 28.5 mm.
Other specifications of the problem are:
- The tools are assumed rigid.
- Blank dimension are: length: 120.0mm; thickness: 1mm; width: 30mm.
- Plane strain is assumed during all simulations.
- Friction coefficient: = 0.1482.
- The punch speed is kept constant between: (1-50) mm/sec.
- Total punch stroke: 28.5 mm.
- Blank material: isotropic steel with mechanical properties:

0.25177
645.24( 0.0102) = + MPa;
E = 217.500 GPa ; = 0.3.

Figure 5.3.26: Initial position (left) and final position (right) for stamping
R
R
R
Punch
Blank
Die
2
3
1
R =23.5 mm
1
R =25 mm
2
R = 4 mm
3
2
9
131

















First of all a parameter study are undertaken with all elements (EAS, SRI, ANS
and solid-shell) in order to determine the sensitivity of the numerical solution with
respect to some numerical parameters. Indeed we considered the influence on the results
of the discretization, the penalty parameters for contact treatment and the time integration
scheme with their associated numerical parameters.
All computations with the EAS and SRI elements were carried out by using
METAFOR software [MET08]. All computations with the ANS and solid-shell elements
were carried out by using FEAP program [TAY01]. This is a plane strain problem, hence,
instead of using the SS7n element its 2D version - 2D.SS4n element will be adopted. The
2D.SS4n is resulted from the 2D.ANSn in (3.3.27) and the EAS4 element, Table 2.4.3.


Simulation with 2D.EAS7 elements
Table 5.3.3 is a survey of some representative results. First of all the blank is
divided into three zones (see Figure 5.3.28) whose lengths are respectively given by 21,
18 and 21 mm, respectively. Inside any of the three zones any element has the same
length. Then each of the zones has been discretized by imposing n
2
elements through the
thickness and respectively n
1
equal elements in the first zone and n
3
and n
4
in the two
remaining zones.

The different numerical parameters are then systematically explored. First, mesh
size was changed to find out a good mesh quality (Table 5.3.3/Cases 1-4). After having
set the mesh, the penalty parameters are varied (Table 5.3.3/Cases 5-6). Then, integration
scheme is changed from quasi-static to implicit Chung-Hulbert (Table 5.3.3/Cases 6-8)
with parameters are
M
= -0.97 and
F
= 0.01. This choice for
M
and
F
satisfies the
unconditional stability and second order accuracy of Chung-Hulbert scheme [PON99].
Number of elements through the thickness direction was also considered (Table
5.3.3/Cases 6-9). Each input parameter was changed to get a better and better result. The
Table 5.3.3/Case 10 with very fine mesh is introduced to get a reference solution.
The areas which are sensible to mesh size are deformed areas (meshed by n
1
, n
2
, n
3

elements). We only survey the changes of n
1
, n
2
and n
3
because the area with n
4
elements
is neither deformed nor contacted by the punch or die (see Figure 5.3.28).
a) between line AB and line CD
2
0
2
0
B
A
D
C
Unit: mm
E
Punch
F
0
b) between 2 contacted points
Figure 5.3.27: Definition of angles
(point E and point F are the farthest contact points from the centerline)
132
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank




















Then, when a fine enough discretization has been attained for the mesh, the
penalty parameters is varied. For reasonable variations, it can be seen in Figure 5.3.30
that there is almost no difference in the force curves. The conclusion is the same when
the implicit Chung-Hulbert scheme has been used instead of the quasi-static algorithm.
As far as one avoids using only one element through the thickness (Case 9 in Figure
5.3.29) one can also see that the results are quite stable. All the punch force -
displacement curves, except the case of using 1 element through the thickness, nearly
coincide with the reference curve - Case 10 (Table 5.3.3). So we can conclude that with
2 elements through the thickness the mesh is refined enough.



Table 5.3.3. Calculation with 2D.EAS7 element
Case
Integration
scheme
Penalty
Mesh Size
n
1
; n
2
; n
3
; n
4

Note
1 QS 1.010
3
30; 2; 40; 20
2 QS 1.010
3
60; 2; 40; 20
3 QS 1.010
3
30; 2; 80; 20
4 QS 1.010
3
30; 2; 100; 20
5 QS 5.010
2
30; 2; 100; 20
6 C-H 5.010
2
30; 2; 100; 20 Convergence of 2D.EAS7
7 C-H 5.010
2
30; 3; 100; 20
8 C-H 3.010
2
30; 3; 100; 20
9 QS 1.010
3
30; 1; 100; 20
10 QS 1.010
3
150; 7; 150; 30 Reference
* Bold numbers or letters indicate what has changed from one case to another
C-H: Chung - Hulbert scheme
QS: Quasi - static
Figure 5.3.28: Discretization of the model: definition of the 3 zones
R
R
R
Punch
Blank
Die
2
3
1
n2 elements
n4 elements n3 elements n1 elements
133
50
100
150
200
250
P
u
n
c
h

f
o
r
c
e

(
N
)

0
5 10 15 25 30
Figure 5.3.29: Punch force vs. punch displacement EAS element
(quasi-static; Table 5.3.3/cases 1 to 5; penalty =1.010
3
)
20
Punch displacement (mm)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
case 2: EAS-60x2x40x20
case 1: EAS-30x2x40x20
case 3: EAS-30x2x80x20
case 4: EAS-30x2x100x20
case 5: : EAS-30x2x100x20-QS;
penalty=5.010
2
300




















For mesh discretization parameter n
3
is the most sensitive. Indeed, a rough
discretization in this zone leads to oscillations in the punch force versus punch
displacement curve as can be seen in Figure 5.3.29 (Case 2). This can be easily explained
since it is the lower side of elements located in this zone that do have a sliding contact
with the shoulder of the die (radius R
3
- see Figure 5.3.28). As far as the contact with
punch is considered the radius R
1
is much larger than R
3
so it is much less sensitive to
discretization. Actually, if n
3
is too small the number of nodes in contact with the die can
be reduced to one. As a consequence, this leads to oscillations in the curve. As can be
seen in Figure 5.3.29, n
3
= 100 lead to almost no oscillation.





















P
u
n
c
h

f
o
r
c
e

(
N
)

Figure 5.3.30: Punch force vs. punch displacement EAS element
(Table 5.3.3/cases 6 to 10)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Case 6: EAS-30x2x100x20
C-H; penalty=5.010
2
(convergence)
Case 7: EAS-30x3x100x20
C-H; penalty=5.010
2

Case 8: EAS-30x3x100x20
C-H; penalty=3.010
2

Case 9: EAS-30x1x100x20
QS; penalty=1.010
3

Case 10: EAS-150x7x150x30
QS; penalty=1.010
3
(Reference)
0
50
100
150
200
250
300
5 10 15 20 25 30
Punch displacement (mm)
134
In Figure 5.3.30, we see that the results got by using different penalty parameters,
which are 2 times of difference (5.010
2
and 110
3
), are generally similar. For this
problem, solutions got by quasi-static scheme and dynamic scheme are almost similar.
Solutions with 2 or 3 elements on the thickness are similar, except the case of using 1
element on the thickness - Case 9 makes the behavior stiffer.



Simulation with 2D.SRI elements
In this section, a similar study was carried out but this time using 2D.SRI
elements. Different results are tabulated in Table 5.3.4.
First of all, we started the 2D.SRI computations with a reasonable mesh for the
2D.EAS7 simulation (reasonable means that the force curve is quite close to the reference
one in this case), e.g. n
1
= 30; n
2
= 2; n
3
= 100; n
4
= 20 (see Table 5.3.3/Case 6). As can
be seen in Figure 5.3.32, the resulting curve exhibits a lack of smoothness. Changing the
numerical parameters while keeping n
1
= 30 does not affect too much the results (see
again Figure 5.3.32), while increasing n
1
will lead to the reference solution - see Figure
5.3.33 - for n
1
equal to or larger than 100.
As a first conclusion, we can state that even if 2D.SRI element are much cheaper
than 2D.EAS7 and converge to the reference solution as the mesh is refined, they should
be used with some care as they do not exhibit a coarse mesh accuracy as the 2D.EAS7
does. The 2D.SRI element converges with a mesh of 100210020 element (Table
5.3.4/Case 6). Meshes of cases 7 and 8 are very fine, so they are the best solutions of
2D.SRI elements.


























Figure 5.3.31: Reference solution of SRI element
(Table 5.3.4 /case 7 equivalent plastic strain, simulation with METAFOR)
0.0211 0.0158 0.0105 0.0052 -6.70e-005
135
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank



Table 5.3.4. Calculation with 2D.SRI element
Case
Integration
scheme
Penalty
Mesh
n
1
; n
2
; n
3
; n
4

Note
1 QS 5.010
3
30; 2; 100; 20
2 QS 5.010
2
30; 2; 100; 20
3 C-H 1.010
3
30; 2; 100; 20
4 C-H 1.010
3
30; 3; 100; 20
5 QS 1.010
3
30; 1; 100; 20
6 C-H 1.010
3
100; 2; 100; 20
Convergence of
2D.SRI
7 QS 1.010
3
150; 5; 150; 20
8 C-H 1.010
3
185; 7; 185; 30























Simulation with 2D.ANS elements
Before investigate springback with the solid-shell elements, lets consider
springback behavior of 2D.ANSn elements. For this springback test, behavior of the ANS
and ANSn is similar so only results of the 2D.ANSn are presented in Table 5.3.5. We
investigate the ANS computations with a mesh with which the EAS element converged,
i.e. n
1
= 30; n
2
= 2; n
3
= 100; n
4
= 20 (see Table 5.3.3/Case 6). Due to volumetric locking
happens with the 2D.ANSn element, we should start with a higher number of element
along the thickness to reduce locking, e.g., n
2
= 3. However, as seeing in Figure 5.3.35,
the resulting curve still exhibits locking.
Case 1: SRI-30x2x100x20
QS; (Penalty 5.010
3
)
0
50
100
150
200
250
300
5 10 15 20 25 30
Case 2: SRI-30x2x100x20
QS; (Penalty 5.010
2
)
Figure 5.3.32: Punch force vs. punch displacement SRI element
(Penalty=1.010
3
; Table5.3.4/cases 1 to 4)
P
u
n
c
h

f
o
r
c
e

(
N
)

Case 3: SRI-30x2x100x20
C-H; (Penalty 1.010
3
)
Case 4: SRI-30x3x100x20
C-H; (Penalty 1.010
3
)
Punch displacement (mm)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Reference
(EAS-150x7x150x30)
Convergence of EAS-30x2x100x20
C-H; penalty=5.010
2

136
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank





















The punch force - displacement curve with the 2D.ANSn element is stiffer than
the reference curve, see Figure 5.3.35. Increasing number of element along the thickness
n
2
= 4 (Table 5.3.5/Case 2), the resulting curve is closer to the reference one. To get more
stringent result, let's trying increasing element number in contact area with punch, n
1
=
100 (Table 5.3.5/Case 3). Data in Figure 5.3.34 shows that increasing n
1
has only effect
of reducing oscillation of contact force. Meanwhile, increasing number of element along
the thickness, n
2
= 5 (Table 5.3.5/Case 4), we get better result. The converged result of
the ANSn element is obtained with penalty parameter equal to 1.0e+2, Table 5.3.5/Case
4. If we use a higher penalty value (5.0e+2) the computational cost increasing while
result is not improved, Figure 5.3.34-36. The consequence is the same when integration
scheme is changed from quasi-static to Newmark (because the Chung-Hulbert scheme is
not available in FEAP).


Table 5.3.5. Calculation with 2D.ANSn element
Case
Integration
scheme
Penalty
Mesh
n
1
; n
2
; n
3
; n
4

Note
1 QS 1.010
2
30; 3; 100; 20
2 QS 1.010
2
30; 4; 100; 20
3 QS 1.010
2
100; 4; 100; 20
4 QS 1.010
2
30; 5; 100; 20
Chosen result of
2D.ANSn
5 QS 1.010
2
100; 5; 100; 20
6 QS 5.010
2
30; 5; 100; 20
7 Newmark 1.010
2
30; 5; 100; 20
0
50
100
150
200
250
300
5 10 15 20 25 30
Case 5: SRI-30x1x100x20
QS
Case 6: SRI-100x2x100x20
C-H (convergence)
P
u
n
c
h

f
o
r
c
e

(
N
)

Figure 5.3.33: Punch force vs. punch displacement SRI element
(Penalty=1.010
3
; Table5.3.4/cases 5 to 8)
Case 7: SRI-150x5x150x20
QS
Case 8: SRI-185x7x185x30
C-H
Punch displacement (mm)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Reference
(EAS-150x7x150x30)
137
Figure 5.3.34: Chosen solution of 2D.ANSn element
(Table 5.3.5 /case 6 von Mises stress, simulation with FEAP)
















































n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Reference
(Table5.3.3/Case 10)
ANS-30x3x100x20
(Table5.3.5/Case 1)
ANS-30x4x100x20
(Table5.3.5/Case 2)
ANS-100x4x100x20
(Table5.3.5/Case 3)
ANS-30x5x100x20
(Table5.3.5/Case 4)
5 10 15 20 25 30
Figure 5.3.35: Punch force vs. punch displacement ANSn element
(Table5.3.5/cases 1 to 4)
Punch displacement (mm)
50
100
P
u
n
c
h

f
o
r
c
e

(
N
)

150
200
250
300
0
0
138
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank

















We see that ANS computation with 5 layers of element along the thickness gives
the same maximal punch force as the one of the EAS element (30210020), see
Figures 5.3.35-36. However, behaviors of the ANS and EAS elements are different. The
ANS element is stiffer when punch displacement is smaller than 20mm. When punch
displacement is larger than 20mm, ANS behavior is softer. This consequence is
reasonable because the ANS element is only shear-locking free.


Simulation with solid-shell element
Finally, springback prediction for this unconstrained bending problem is
investigated with the solid-shell element: 2D.SS4n. With this test, behavior of the solid-
shell elements using the classical technique and the alternative ANS technique is similar.
Hence, only results of the later are presented in Table 5.3.6. We begin the computation
with a coarse mesh for EAS simulation, i.e. n
1
= 30; n
2
= 2; n
3
= 40; n
4
= 20 (Table
5.3.3/Case 1). Then, the mesh is made finer to get better result. Data from Figure 5.3.38
shows that the solid-shell element converged with rather coarse mesh, n
1
n
2
n
3
n
4
=
6027020.


Table 5.3.6. Calculation with solid-shell element
Case
Integration
scheme
Penalty
Mesh
n
1
; n
2
; n
3
; n
4

Note
1 QS 1.010
2
30; 2; 40; 20
2 QS 1.010
2
50; 2; 60; 20
3 QS 1.010
2
60; 2; 70; 20
Convergence of
solid-shell
4 QS 5.010
2
60; 3; 70; 20
5 Newmark 1.010
2
60; 2; 70; 20
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Reference
(Table5.3.3/Case 10)
ANS-100x5x100x20
(Table5.3.5/Case 5)
ANS-30x5x100x20
(Table5.3.5/Case 6)
ANS-30x5x100x20
(Table5.3.5/Case 7)
5 10 15 20 25 30
Figure 5.3.36: Punch force vs. punch displacement ANSn element
(Table5.3.5/cases 5 to 7)
Punch displacement (mm)
50
100
P
u
n
c
h

f
o
r
c
e

(
N
)

150
200
250
300
0
0
139
Figure 5.3.37: Reference solution of 2D.SS4n element
(Table 5.3.6 /case 3 von Mises stress, simulation with FEAP)





















After having set the mesh, the penalty parameters are varied (Table 5.3.6/Cases 4).
Then, integration scheme is changed from quasi-static to Newmark (Table 5.3.6/Cases 5).
It can be seen in Figures 5.3.38 - 39 that there is almost no difference in the force curves.
The conclusion is the same when the Newmark scheme has been used instead of the
quasi-static algorithm. We see that when the punch force is larger than 200N the solid-
shell element is a little softer than the reference result.





















Figure 5.3.38: Punch force vs. punch displacement 2D.SS4n element
(Table5.3.6/cases 1 to 3)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
Reference
(Table5.3.3/Case 10)
SS4n-30x2x40x20
(Case 1)
SS4n-50x2x60x20
(Case 2)
SS4n-60x2x70x20
(Case 3. convergence)
5 10 15 20 25 30
Punch displacement (mm)
50
100
P
u
n
c
h

f
o
r
c
e

(
N
)

150
200
250
300
0
0
140


















Validation of simulation results
In order to validate results of the solid-shell element, the force curve obtained has
been compared to both experimental (BE - Figure 5.3.40) and numerical (BS - Figure
5.3.41) results published in the NUMISHEET02 proceeding [NUM02]. As can be seen
from those figures, numerical results from the 2D.SS4n match quite well the
experimental reference results (curve BE-01). It should be noted that numerical results
exhibit a quite large dispersion which can be attributed to the variety of finite element
codes, as well as the variety of elements (shell, continuum quads, triangles) and time
integration algorithm (implicit, quasi-static).




















Figure 5.3.39: Punch force vs. punch displacement 2D.SS4n element
(Table5.3.6/cases 4 and 5)
n elements
1
n elements
3
n elements
4
n elements
2
Mesh on Blank
EAS-30x2x100x20
(Table5.3.3/Case 6)
SS4n-100x5x100x20
(Case 4)
SS4n-30x5x100x20
(Case 5)
5 10 15 20 25 30
Punch displacement (mm)
50
100
P
u
n
c
h

f
o
r
c
e

(
N
)

150
200
250
300
0
0
0
50
100
150
200
250
300
0 5 10 15 20 25 30
350
400
450
500
550
600
BE-01
BS-13
BS-14
BS-15
BS-16
BS-17
Figure 5.3.40: Punch force vs. punch displacement [NUM02]
(BS = Benchmark simulation result)
P
u
n
c
h

f
o
r
c
e
,

k
N

Punch displacement (mm)
2D.SS4n-6027020
(Table 5.3.6/Case 3)
2D.EAS7-30210020
(Table5.3.1/Case 6)
141















When punch force is smaller than 200N results of the 2D.EAS7 and 2D.SS4n are
coincident but they are a little different from the experiment results, BE-01. When punch
force is larger than 200N, only the 2D.EAS7 gives the identical results with results of the
BE-01. Meanwhile, the 2D.SS4n is a little softer than the 2D.EAS7.

As a more local result of the benchmark, it was also asked to evaluate the angle -
see Figure 5.3.27 and Figure 5.3.42 for definition and illustration for different punch
stroke, i.e. 7, 14, 21 and 28.5 mm. Results delivered by the 2D.SS4n are very closed to
the experimental values (BE-01), see Figure 5.3.43. The convergence results of the other
element are also presented in Figure 5.3.43. See Table 5.3.7 - 10 for angle at different
punch strokes in all cases of simulation with the 2D.EAS7, 2D.SRI, 2D.ANSn, 2D.SS4n
elements.


Table 5.3.7. Springback angles - calculation with 2D.EAS7 element
Angle () between line
AB and line CD
Angle () between 2 farthest
points (from centerline) at stroke Case
before SB after SB 7mm 14mm 21mm 28.5mm
1 21.864 33.440 16.536 60.298 111.108 158.726
2 21.184 33.436 18.688 62.120 111.228 158.494
3 21.376 33.338 16.964 60.054 112.298 158.406
4 21.378 33.082 16.786 60.182 112.514 158.804
5 21.882 33.850 20.154 63.634 113.386 158.970
6 21.974 33.788 20.102 63.652 113.468 158.904
7 21.630 33.404 20.158 63.550 113.480 158.872
8 21.056 33.968 20.326 63.666 114.338 159.800
9 21.438 35.510 20.322 63.458 113.466 158.702
10 21.344 33.358 18.880 62.260 112.432 158.278





P
u
n
c
h

f
o
r
c
e
,

N

EAS-30210020
(Table5.3.1/Case 6)
2D.SS4n-6027020
(Table 5.3.6/Case 3)
Punch displacement (mm)
Figure 5.3.41: Punch force vs. punch displacement [NUM02]
(BE = Benchmark Experiment result)
BE-01
BE-02
BE-03
BE-04
0
50
100
150
200
250
300
350
0 5 10 15 20 25 30
142
B
E
-
0
1

B
E
-
0
2

B
E
-
0
3

B
E
-
0
4

B
S
-
0
1

B
S
-
0
2

B
S
-
0
3

B
S
-
0
4

B
S
-
0
5

B
S
-
0
6

B
S
-
0
7

B
S
-
0
8

B
S
-
0
9

B
S
-
1
0

B
S
-
1
1

B
S
-
1
2

B
S
-
1
3

B
S
-
1
4

B
S
-
1
5

B
S
-
1
6

2
D
.
E
A
S
7

2
D
.
S
R
I


BE01
BE01
BE01
BE01
7mm
14mm
21mm
28.5mm
Figure 5.3.43: Angle between 2 contact points which are the
farthest from the centerline [NUM02]
Participants
A
n
g
l
e

(
0
)

0
20
40
60
80
100
120
140
160
180
200
2
D
.
A
N
S
n

2
D
.
S
S
4
n

















































of open angle
Figure 5.3.42: Definition of the angle between 2 contact points
which are the farthest from the centerline
contact area
143

Table 5.3.8.: Springback angles - calculation with 2D.SRI element
Angle () between line
AB and line CD
Angle () between 2 farthest
points (from centerline) at stroke Case
before SB after SB 7mm 14mm 21mm 28.5mm
1 21.246 34.086 16.988 60.106 111.550 157.048
2 21.704 34.654 20.258 63.340 113.228 159.652
3 21.652 34.348 20.112 63.606 112.612 158.816
4 21.514 33.400 16.786 60.028 112.446 158.598
5 21.144 35.726 20.238 63.560 113.476 158.806
6 21.466 34.884 19.246 63.052 112.640 159.002
7 21.288 33.430 18.806 62.132 112.388 158.790
8 21.344 33.812 18.550 62.206 112.464 158.462


Table 5.3.9. Springback angles - calculation with 2D.ANSn element
Angle () between line
AB and line CD
Angle () between 2 farthest
points (from centerline) at stroke Case
before SB after SB 7mm 14mm 21mm 28.5mm
1 22.000 49.800 16.000 62.000 116.000 158.000
2 23.000 40.600 18.000 64.000 116.000 159.000
3 23.000 40.600 19.000 64.000 116.000 158.000
4 23.000 37.000 19.000 63.000 116.000 159.000
5 23.000 37.000 19.200 61.500 114.000 159.000
6 21.500 36.000 18.000 63.000 114.000 159.000
7 21.600 36.500 18.000 63.000 114.000 159.000


Table 5.3.10. Springback angles - calculation with 2D.SS4n element
Angle () between line
AB and line CD
Angle () between 2 farthest
points (from centerline) at stroke Case
before SB after SB 7mm 14mm 21mm 28.5mm
1 23.000 34.500 25.000 66.000 117.000 160.000
2 23.000 33.500 24.000 67.500 117.000 161.000
3 23.000 33.500 23.000 66.000 117.000 157.500
4 22.500 32.500 24.000 66.000 114.500 159.000
5 22.500 32.500 22.000 66.000 115.500 159.000




Springback simulation
In order to evaluate the springback, the tools are progressively removed and the
resulting opening angle, as defined in Figure 5.3.27a is measured as shown in Figures
5.3.44 and 5.3.45.


144
Figure 5.3.44: Open angle between the lines AB and CD before spring back
(Table 5.3.3/Case 6 - equivalent plastic strain - METAFOR)
0.0248 0.0183 0.0119 0.00540 -0.00107
( open angle)
10,987
















































Figure 5.3.45: Open angle between the lines AB and CD after spring back
(Table 5.3.6/Case 3 - equivalent stress - FEAP)
( open angle)
16.750
145
A
n
g
l
e

(

)

Participants
B
E
-
0
1

B
E
-
0
2

B
E
-
0
3

B
E
-
0
4

B
S
-
0
2

B
S
-
0
1

B
S
-
0
3

B
S
-
0
4

B
S
-
0
5

B
S
-
0
6

B
S
-
0
7

B
S
-
0
8

B
S
-
0
9

B
S
-
1
0

B
S
-
1
1
A

B
S
-
1
1
B

B
S
-
1
2

B
S
-
1
3

B
S
-
1
4

B
S
-
1
5

2
D
.
E
A
S
7

2
D
.
S
R
I

B
S
-
1
6
A

B
S
-
1
6
B

Figure 5.3.46: Angle before and after springback at the final punch
stroke of 28.5 mm
0
5
10
15
20
25
30
35
40
45
Average
Average
50
Before SB
After SB
2
D
.
A
N
S
n

2
D
.
S
S
4
n



























Data in Tables 5.3.7 - 5.3.10 are used to compare results of the 2D.EAS7, 2D.SRI,
2D.ANSn and 2D.SS4n with the other sources in [NUM02]. Figure 5.3.46 presents
curves for angles before and after springback. Concretely:
- Figure 5.3.46 shows that almost open angles before and after springback
concentrate around the mean values, 22 with respect to open angles before
springback and 34 with respect to open angles after springback.
- Simulated results of the 2D.EAS7 and 2D.SS4n almost conform to major
softwares, concretely, 21.974 (2D.EAS7) and 23.000 (2D.SS4n) with respect
to open angles before springback and 33.788 (2D.EAS7) and 33.500
(2D.SS4n) with respect to open angles after springback. For data of all cases see
Tables 5.3.7 -5.3.10.












146
CONCLUSION

In this chapter some plasticity problems were invoked to investigate performances
of the solid-shell elements. Only free from shear locking, the ANS element is not suitable
to the plasticity tasks. The SS7n element, in general, gives similar results as the SS7,
except the cases where large aspect ratio and large deformation occur simultaneously
(Section 5.3.2). Being easy for contact handling, the solid-shell elements are well
applicable to metal forming problem, such as springback prediction. As we have seen in
Section 5.3.3, the simulation accuracy can be improved effectively if a good performance
elements is used with a suitable methodology for choosing numerical parameters. In the
last section, the 2D.EAS7, 2D.SRI, 2D.ANSn and 2D.SS4n elements are used to predict
springback due to elastic recovery of the material under elastoplastic deformation.
Comparing between models, the best one costs minimum computational calculation while
gives a correct solution. Correlation between the numerical solution of the solid-shell
element and the EAS element, a good conformability was found with the experimental
solution (BE-01). However, with a limited number of enhanced modes, the solid-shell
element proposes a lower computational cost than the EAS elements. Hence, the metal
forming simulation with the solid-shell element promises a confident and economical tool
for springback prediction in industrial manufacturing.

In conclusion, through all numerical tests, it is possible to state that the solid-shell
elements are useful for simulating shell structures with a wide range of thickness (from
thin (1000 > R/t > 20) to moderately thick (R/t < 20)). Advantages of the elements are:
- Connecting of the solid-shell with other continuum elements is simple;
- Easy handling of variable thicknesses;
- Accuracy over a wide range of thickness;
- No more need for mid-surface extraction;
- Easier contact handling.



















147
Chapter 6. GENERAL CONCLUSIONS AND
FUTURE WORKS


CONCLUSIONS
In the thesis, locking phenomena which happen with low-order solid elements
have been analysed. The free shear locking ANS element and its performance have been
introduced. As showed in Chapter 2, the EAS element can circumvent all locking effects
but with high computational cost. Hence, an element which can circumvent all locking
effects but with lower computational cost than the EAS element is preferred. In chapter 3,
the solid-shell element (designated as SS7) has been presented as a combination of the
classical ANS element [DVO84] and the EAS element [SIM90]. In comparison with the
classical shell elements, the solid-shell elements allow a straightforward integration of
3D material models since they do not resort to the plane stress assumption. Moreover,
their solid topology offers an effective doubly sided contact handling possibility. In
comparison with the EAS solid elements, the solid-shell elements require only a limited
number of EAS parameters, i.e. 7 modes, and the computational effort is thus reduced.
Those features, which are truly a combination of the strong points of both the ANS and
EAS methods, render the solid-shell elements very attractive.

The thesis concentrates in developing the solid-shell element. As a result, an
alternative ANS technique has been presented in Chapter 3. This alternative ANS
technique can incorporate with the EAS method to result in a new solid-shell element,
named SS7n. This alternative ANS technique assumes that the transverse shear strains are
linear in the thickness direction and in an in-plane direction. Meanwhile, following the
classical ANS technique, the assumed transverse shear strains are only linear in one in-
plane direction. The classical and the alternative ANS techniques were systematically
compared together. Both of them can assist the solid-shell elements, SS7 and SS7n, pass
the membrane and bending patch tests. Performances of the SS7 and SS7n elements have
been investigated for both bulk problems and shell-like structures, ranging from thin to
moderately thick structures. Theory in Chapter 3 and the plasticity test with pinched
cylinder in Chapter 5 proved that the new solid-shell element, SS7n, leads to a better
approximation of strains and stresses in the thickness direction than the SS7 element
(with the classical ANS technique). Also, the SS7n element is less sensitive to distorted
mesh than the SS7 element, see numerical test 3.6.2.

Numerical results in the thesis show that the SS7n solid-shell element is well
adapted to most engineering problems. In particular, for pure bending problems and
under nearly incompressible condition the SS7n element exhibits comparable or superior
performances with respect to their original ANS and EAS counterparts.





148




FUTURE WORKS
The solid-shell SS7 and SS7n elements have been implemented in a MATLAB
code. This MATLAB code is helpful to investigate linear and nonlinear elastic tests. The
solid-shell SS7 and SS7n elements have been also implemented in FEAP [TAY01] to
exploit the available plasticity material models and contact algorithm in FEAP. These
facilities are necessary for plasticity and springback tests in Chapter 5. However,
plasticity behavior and springback prediction are enormous subjects. In the thesis, only
some basic aspects of these subjects have been investigated. The following works should
be pursued:
- Investigating capabilities of the classical ANS technique and the alternative ANS
technique in approximating transverse stresses and strains. The alternative ANS
technique allows a better approximation of the stress field in the thickness
direction, hence would lead to a more accurate evaluation of stress-based criteria
such as delamination criteria in composite materials.
- Exploitation of the SS7 and SS7n in metal forming and springback prediction;
These engineering problems are usually too complex (deformation rate, contact
condition, material hardening, distorted mesh, etc.) and include a large number of
DOFs. Hence, an effective (less computational cost and more accuracy) element
is truly demanded.
- Apart from the springback prediction, there are also several interesting topics in
metal forming field, such as tearing, wrinkling, limit forming curve, etc.
Performances of the solid-shell elements with these topics may be perspective.
- In Chapter 4 we did see that the SS7n element is well applicable to structural
problems. For some specific applications, such as collapse of shells, structural
stability, etc. peformances of the SS7n would be useful.

The solid-shell theory in this thesis will be useful for researchers who want to
apply 3D solid elements to simulate thin-walled structures. In order to facilitate research
works in various topics, the solid-shell elements should be implemented in a software
which had various material models, integration schemes and contact algorithms.













149
REFERENCE



[A]

[ADA05] L. Adam, JP. Ponthot, Thermomechanical modeling of metals at finite
strains: first and mixed order finite elements, International Journal of
Solids and Structures 2005; 42/21-22: 5615-5655.
[AHM70] S. Ahmad, B.M. Irons, and O.C. Zienkiewicz, Analysis of thick and thin
shell structures by curved elements, Int. J. Numer. Meth. Engng. 1970; 2:
419-451.
[AND93] U. Andelfinger, E. Ramm, EAS-elements for two-dimensional, three-
dimensional, plate and shell structures and their equivalence to HR-
elements, Int. J. Numer. Meth. Engng. 1993; 36: 1311-1337.

[B]

[BAR91] F. Barlat, D.J. Lege, J.C. Brem, A six component yield function for
anisotropic materials, Int. J. of Plasticity, 7: 693-712, 1991.
[BAT96] K.J. Bathe, Finite Element Procedures, Prentice Hall, 1996.
[BAT00] K.J. Bathe, A. Iosilevich and D. Chapelle, An evaluation of the MITC shell
elements, Computers & Structures, 75 (1): 1-30, 2000.
[BAT03] K.J. Bathe, P.S. Lee and J.F. Hiller, Towards Improving the MITC9 Shell
Element, Computers & Structures, 81: 477-489, 2003.
[BEL93] T. Belytschko, LP. Bindeman, Assumed strain stabilization of the eight
node hexahedral element, Computer Methods in Applied Mechanics and
Engineering 1993; 105: 225-260.
[BET95] P. Betsch, E. Stein, An assumed strain approach avoiding artificial
thickness straining for a non-linear 4-node shell element,
Communications in Numerical Methods in Engineering 1995; 11: 899-909.
[BET96] P. Betsch, E. Gruttmann, E. Stein, A 4-node finite shell element for the
implementation of general hyperelastic 3D-elasticity at finite strains,
Comput. Methods Appl. Mech. Engrg. 130 (1996) 57-79.
[BIS97] M. Bischoff, E. Ramm, Shear deformable shell elements for large strains
and rotations, Int. J. Numer. Meth. Engng. 40, 44274449 (1997).
[BIS99] M. Bischoff, E. Ramm, D. Braess, A class of equivalent enhanced
assumed strain and hybrid stress finite elements, Computational
Mechanics, Vol. 22, pp. 443-449, 1999.
[BIS03] M. Bischoff, F. Koschnick, K.U. Bletzinger, Stabilized DSG elements A
new paradigm in finite element technology, 4
th
European LS-DYNA users
conference, Germany 2003.
[BIS05a] M. Bischoff, K.U. Bletzinger, Interaction of locking and element stability
at large strains, VIII International Conference on Computational Plasticity
- COMPLAS VIII, E. Onate and D.R.J. Owen (Eds), Barcelona, 2005.
[BIS05b] M. Bischoff, K.U. Bletzinger, Advanced Finite Element Methods, Master
Program on Computational Mechanics - Technical University of Munich,
Summer Term 2005.
150
[BLE00] K.U. Bletzinger, Theory of Plates: Part II: Plates in bending, Lecture
Notes - Winter Semester, Technische Universitt Mnchen, 2000
[BRA98] D. Braess, Enhanced assumed strain elements and locking in membrane
problems, Comput. Methods Appl. Mech. Engrg. 165 (1998) 155-174.
[BUC93] M.L. Bucalem and K.J. Bathe, Higher-Order MITC General Shell
Elements, Int. J. Numer. Meth. Engng. 1993; 36: 3729-3754.
[BUI02] Q.V. BUI, Validation of enhanced assumed strains (EAS) 8-node brick
elements in METAFOR - Universit de Lige, LTAS-Milieux Continus
& Thermomcanique, 2002.
[BUI04] Q.V. Bui, L. Papeleux, J.P. Ponthot, Numerical simulation of Springback
using enhanced assumed strain elements, Journal of Materials Processing
Technology 2004; 153-15:314-318.

[C]

[CAO02] Y.P. Cao, N. Hu, J. Lu, H. Fukunaga and Z.H. Yao, A 3D brick element
based on HuWashizu variational principle for mesh distortion, Int. J.
Numer. Meth. Engng. 2002; 53: 25292548.
[CAR07] RPR. Cardoso, JW. Yoon, M. Mahardika, S. Choudhry, RJ. Alves de
Sousa, RA. Fontes Valente. Enhanced assumed strain (EAS) and assumed
natural strain (ANS) methods for one-point quadrature solid-shell
elements. International Journal for Numerical Methods in Engineering
2007; 69:627-663.
[CHA89] S. Chandra and G. Prathap, A field-consistent formulation for the eight-
noded solid finite element, Computers and Structures 33 (1989) 345-355.
[CHO01] C.K. Choi, K.Y. Chung, T.Y. Lee, A direct modification method for
strains due to non-conforming modes, Structural Engineering and
Mechanics 2001; Vol. 11: 325340.
[CHU92] K. Chung, Shah K. Finite element simulation of sheet metal forming for
planar anisotropic metals. International Journal of Plasticity 1992; 8: 453
- 476.

[D]

[DJO05] J.K. Djoko, B.P. Lamichhane, B.D. Reddy and B. I. Wohlmuth,
Conditions for equivalence between the Hu-Washizu and related
formulations, and computational behavior in the incompressible limit,
Institut fur Angewandte Analysis und Numerische Simulation (IANS),
2005.
[DOL00] S. Doll, K. Schweizerhof, R. Hauptmann, C. Freischlger, On volumetric
locking of low-order solid and solid-shell elements for finite elasto-
viscoplastic deformation and selective reduced integration, Engineering
Computations, vol. 17, pp 874-902, 2000.
[DUR58] A.J. Durelli, E.A. Phillips and C.H. Tsao, Introduction to the theoretical
and experimental analysis of stress and strain - New York : McGraw-Hill,
1958.
[DVO84] E.N. Dvorkin, K.J. Bathe, A continuum mechanics based four-node shell
element for general non-linear analysis, Engrg. Comput. 1 (1984) 7788.
151
[DVO95] E.N. Dvorkin, D. Pantuso and E.A. Repetto, A formulation of the MITC4 shell
element for finite strain elasto-plastic analysis, Comput. Meth. Appl.
Mechs. Engrg., 125, 17-40, 1995.

[F]

[FEL02] C.A. Felippa and K.C. Park, Fitting Strains and Displacements by
Minimizing Dislocation Energy, 6th International Conference on
Computational Structures Technology, Prague - Czech Republic, September
2002.
[FEL03] C.A. Felippa, Advanced Finite Element Methods (ASEN 5367)
(http://titan.colorado.edu/courses.d/AFEM.d/), Colorado, 2003.
[FEL05a] C.A. Felippa, Supernatural Quad4: A template formulation Report No.
CU-CAS-05-06, Dept. of Aerospace Engineering Sciences and Center for
Aerospace Structures, Univ. of Colorado, 2005.
[FEL05b] C.A. Felippa, A template tutorial, Ch. 3 of Computational Mechanics
Theory and Practice, ed. by K. M. Mathisen, T. Kvamsdal and K. M.
Okstad, CIMNE, Barcelona, 2005.
[FLU73] W. Flugge, Stresses in shells, Springer-Verlag - NewYork, 1973.
[FRE07] M. Fredriksson and N. S. Ottosen, Accurate eight-node hexahedral
element, Int. J. Numer. Meth. Engng.; 72: 631-657, 2007.
[FRE96] C. Freischlger, K. Schweizerhof, On a systematic development of
trilinear three-dimensional solid elements based on Simo's enhanced strain
formulation, Int. J. Solids Struct.; Vol. 33, No. 20-22: 2993-3017, 1996.

[H]

[HAC89] W.L. Hacker, H.L. Schreyer, Eigenvalue analysis of compatible and
incompatible rectangular four-node quadrilateral elements, Int. J. Numer.
Meth. Engng.; 28: 687703, 1989.
[HAN98] P. Hansbo, A new approach to quadrature for finite elements
incorporating hourglass control as a special case, Comput. Methods
Appl. Mech. Engrg. 158: 301-309, 1998.
[HAR02] M. Harnau, K. Schweizerhof, About linear and quadratic Solid-Shell
elements at large deformation, Computers and Structures 80: 805
817, 2002.
[HAU98] 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. Meth. Engng 42,
49-69, 1998.
[HAU00] R. Hauptmann, K. Schweizerhof, S. Doll, Extension of the solid-shell
concept for application to large elastic and large elastoplastic
deformations, Int. J. Numer. Meth. Engng; 49: 1121-1141, 2000.
[HAU01] R. Hauptmann, S. Doll, M. Harnau, K. Schweizerhof, Solid-shell
elements with linear and quadratic shape functions at large deformations
with nearly incompressible materials, Computers and Structures; 79
: 1671-1685, 2001.
152
[HIL90] R. Hill, Constitutive modelling of orthotropic plasticity in sheet metals, J.
Mech. Phys. Solids, 38 (3) 405-417, 1990.
[HIL00] R. Hill, Plastic anisotropy and the geometry of yield surface in stress
space, J. Mech. Phys. Solids, 48: 1093-1106, 2000.
[HOG76] M. Hogge, Transfert de chaleur et contraintes thermiques dans les
structures par la mthode des lments finis. PhD thesis (in French),
Universit de Lige, Lige, Belgium, 1976.
[HOL96] GA. Holzapfela, R. Eberleinb, P. Wriggers, HW. Weizscker. Large strain
analysis of soft biological membranes: Formulation and finite element analysis.
Comput. Methods Appl. Mech. Engrg, 132 (1-2): 45-61, 1996.
[HOL00] G.A. Holzapfel, Nonlinear solid mechanics, Jonh Wiley & Sons, Ltd.,
Chichester 2000.
[HUG81] T.J.R. Hughes, T.E. Tezduyar, Finite elements based upon Mindlin plate
theory with particular reference to the four-node bilinear isoparametric
element, J. Appl. Mech., 48: 587-596, 1981.
[HUH00] H. Huh, S.H. Kim, Stress evaluation with numerical integration schemes
in finite element analysis of elasto-plastic bending, Commun. Numer.
Meth. Engng, 16 (11): 755 767, 2000.

[J]

[JET08] P. Jetteur, Elment Solid shell dans Samcef, Rapport 252, Samtech
(2008).
[JIA97] Z.P. Jiao, T.H.H. Pian, S. Yong, A new formulation of isoparametric finite
elements and the relationship between hybrid stress element and
incompatible element, Int. J. Numer. Meth. Engng 40, 15-27, 1997.

[K]

[KAS00a] E.P. Kasper, R.L. Taylor, A mixed-enhanced strain method
Part I: Geometrically linear problems, Computers and Structures 75
237-250, 2000.
[KAS00b] E.P. Kasper, R.L. Taylor , A mixed-enhanced strain method
Part II: Geometrically nonlinear problems, Computers and Structures
75 251-260, 2000.
[KLI97] S. Klinkel and W. Wagner, A geometrical non-linear brick element based
on the EAS-method, Int. J. Numer. Meth. Engng 40, 4529-4545, 1997.
[KLI02] S. Klinkel and S. Govindjee, Using finite strain 3D-material models in
beam and shell elements, Engineering Computations, Vol. 19 No. 8,
pp. 902-921, 2002.
[KLI06] S. Klinkel, F. Gruttmann and W. Wagner, A robust non-linear solid shell
element based on a mixed variational formulation, Comput. Methods
Appl. Mech. Engrg. 195 179201, 2006.
[KUH00] D. Kuhl, E. Ramm, Time Integration in the Context of Energy Control and
Locking Free Finite Elements, Comput. Methods in Engrg. Vol. 7, Issue 3
299-332, 2000.



153
[L]

[LAU00] N. Lautersztajn-S and A. Samuelsson, Further discussion on four-node
isoparametric elements in plane bending, Int. J. Numer. Meth. Engrg., 47,
129140, 2000.
[LEM90] J. Lemaitre and J.L. Chaboche, Mechanics of solid materials, Trans. by
B. Shrivastava, Cambridge University Press, Cambridge, UK, 1990.
[LI97] K.P. Li and S. Cescotto, An 8-node brick element with mixed
formulation for large deformation analyses, Comp. Methods Appl. Mech.
Engrg., 141, 157-204, 1997

[M]

[MAC85] R.H. McNeal and R.L. Harder, A proposed standard set of problems to test
finite element accuracy, Finite Elements in Analysis and Design, Volume
1, Issue 1, April 1985, Pages 3-20.
[MAC87] R.H. McNeal, A theorem regarding the locking of tappered four-noded
membrane elements, Int. J. Numer. Meth. Engng. 1987; 24: 1793-1799.
[MAC92] R.H. McNeal, On the limits of finite element perfectability, Int. J. Numer.
Meth. Engng.; 35: 1589-1601, 1992.
[MAC94] R.H. McNeal, Finite elements : their design and performance (Mechanical
engineering series/ 89), Marcel Dekker, New York, 1994.
[MAL69] L.E. Malvern, Introduction to the mechanics of a continuous
medium, Prentice-Hall, Englewood Cliffs, New Jersay, 1969.
[MET08] METAFOR. A finite element software for large strain analysis,
http://www.ltas-mnl.ulg.ac.be/dokuwiki/doku.php, Lige, 2008.
[MIE04] C. Miehe, N. Apel, Anisotropic elastic-plastic analysis of shells at large
strains. A comparison of multiplicative and additive approaches to
enhanced finite element design and constitutive modelling, Int. J. Numer.
Meth. Engng.; 61: 2067-2113, 2004.
[MIL90a] C. Militello and C.A. Felippa, A variational justification of the assumed
natural strain formulation of finite elementsI. Variational
principles, Computers and Structures 34: 431-438, 1990.
[MIL90b] C. Militello and C.A. Felippa, A variational justification of the assumed
natural strain formulation of finite elements II. The C
0
four-node plate
element, Computers and Structures 34: 439-444, 1990.

[N]

[NGU08] N.H. Nguyen, V.N. Pham, M. Hogge and J.P. Ponthot. An assumed
natural strain technique for 2D solid-shell elements. ACOMEN 2008,
Editors: M. Hogge, R. Van Keer, L. Noels, L. Stainier, J.-P. Ponthot, J.-
F. Remacle, E. Dick. University of Lige, Belgium, 26-28 May 2008.
[NUM02] NUMISHEET 2002, in: D.Y. Yang, S.I. Oh, H. Huh, Y.H. Kim (Eds.),.
Proceedings of the 5th International Conference and Workshop on
Numerical Simulation of 3D Sheet Forming Processes - Verification of
Simulation with Experiments. Jeju Island, Korea, 2002.
154
[O]

[OWE86] D.R.J. Owen and E. Hinton, Finite elements in Plasticity: theory and
practice, Pineridge Press Ltd., Swansea, UK. 1986.

[P]

[PAP98] P. Papadopoulos and J. Lu, A general framework for the numerical
solution of problems in finite elasto-plasticity, Comput. Methods Appl.
Mech. Engrg. 1998, 159: 118.
[PAP01] P. Papadopoulos and J. Lu, On the formulation and numerical solution
of problems in anisotropic finite plasticity, Comput. Methods Appl.
Mech. Engrg. 2001, 190 (37-38): 48894910.
[PAR86] K.C. Park and G.M. Stanley, A curved C
0
shell element based on assumed
natural coordinate strains, J. Appl. Mech. 1986, 53: 278290.
[PIA86] T.H.H. Pian, Relations between incompatible displacement model and
hybrid stress model, Int. J. Numer. Meth. Engng. 1986; 22: 173-181.
[PIA06] T.H.H. Pian, C.C. Wu, Hybrid and incompatible finite element
methods, Boca Raton: Chapman & Hall/CRC, 2006.
[PIL99] R. Piltner and R.L. Taylor, A systematic construction of B-bar functions
for linear and non-linear mixed-enhanced finite elements for plane
elasticity problems, Int. J. Numer. Meth. Engng. 1999; 44: 615-639.
[PIL00] R. Piltner, An implementation of mixed enhanced finite elements with
strains assumed in Cartesian and natural element coordinates using sparse
B -matrices, Engng. Comput 2000; Vol. 17 8: 933-949.
[PON94] J-P. Ponthot and M. Hogge, On relative merits of implicit / explicit
algorithms for transient problems in metal forming simulation. In
International Conference on Numerical Methods for Metal Forming in
Industry, Vol 2, pp 128-148, Baden-Baden, Germany, Sept. 1994.
[PON95] J-P. Ponthot, Tratement unifi de la Mcanique des Milieux Continus
solides en grandes transformations par la mthode des lments finis.
PhD thesis (in French), Universit de Lige, Lige, Belgium, 1995.
[PON99] J-P Ponthot and D. Graillet, Efficient implicit schemes for the treatment of
the contact between deformable bodies: Application to shock-absorber
devices, International Journal of Crashworthiness; 4: 273-286, 1999.
[PRA85] G. Prathap, The poor bending response of the four-node plane stress
quadrilateral, Int. J. Numer. Meth. Engng.; 21: 825-835, 1985.
[PRA01] G. Prathap, Finite element analysis as computation, On-line book of
http://www.cmmacs.ernet.in/, Bangalore, 2001.
[PUS00] M.A. Puso, A highly efficient enhanced assumed strain physically
stabilized hexahedral element, Int. J. Numer. Meth. Engng.; 49:1029-1064,
2000.

[Q]

[QUO03a] L. Vu-Quoc and X.G. Tan, Optimal solid shells for non-linear analyses of
multilayer composites. I. Statics, Comput. Methods Appl. Mech. Engrg.
192 (2003) 9751016.
155
[QUO03b] L. Vu-Quoc and X.G. Tan, Optimal solid shells for non-linear analyses of
multilayer composites. II. Dynamics, Comput. Methods Appl. Mech.
Engrg. 192 (2003) 10171059.

[R]

[RAO01] K.M. Rao and U. Shrinivasa, A set of pathological tests to validate new
finite elements, Sadhana Vol. 26, Part 6, December 2001, pp. 549590.
Printed in India.
[REE05] S. Reese, On a physically stabilized one point finite element formulation
for three-dimensional finite elasto-plasticity, Comput. Methods Appl.
Mech. Engrg. 194 (2005) 46854715.
[REE07] S. Reese, A large deformation solid-shell concept based on reduced
integration with hourglass stabilization. International Journal for
Numerical Methods in Engineering 2007; 69:1671-1716.

[S]

[SCH02] J. Schrder, F. Gruttmann and J. Lblein, A simple orthotropic finite
elastoplasticity model based on generalized stressstrain
measures, Computational Mechanics 30 (2002), 4864.
[SCI02] F.M. Sciarra, Relations between enhanced strain methods and the HR
method, Comput. Methods Appl. Mech. Engrg. 191 (2002), 26612677.
[SIM86] J.C. Simo, T.J.R. Hughes, On the variational foundations of assumed
strain methods, ASME J. Appl. Mech. 1986; 53: 5154.
[SIM88a] J.C. Simo, A frame work for finite strain elastoplasticity based on
maximum plastic dissipation and the multiplicative decomposition: Part I.
Continuum formulation, Comput. Meth. Appl. Mechs. Engrg., 66, 199-
219, 1988.
[SIM88b] J.C. Simo, A frame work for finite strain elastoplasticity based on
maximum plastic dissipation and the multiplicative decomposition: Part II.
Computational aspect, Comput. Meth. Appl. Mechs. Engrg., 68, 1-31,
1988.
[SIM90] J.C. Simo, M.S. Rifai, A class of mixed assumed strain methods and the
method of incompatible mode, Int. J. Numer. Meth. Engng. 1990;
29: 15951638.
[SIM92a] J.C. Simo, F. Armero, Geometrically nonlinear enhanced-strain mixed
methods and the method of incompatible modes, Int. J. Numer. Meth.
Engng. 1992; 33: 1413-1449.
[SIM92b] J.C. Simo, Algorithms for static and dynamic multiplicative plasticity that
preserve the classical return mapping schemes of the infinitesimal
theory Comput. Methods Appl. Mech. Engrg. 1992; 99(1): 61 - 112.
[SIM93] J.C. Simo, F. Armero, R. L. Taylor, Improved versions of assumed
enhanced strain tri-linear elements for 3D finite deformation
problem, Comput. Methods Appl. Mech. Engrg. 110 (1993) 359386.
[SIM98] J.C. Simo, T.J.R. Hughes, Computational Inelasticity - Interdisciplinar
Applied Mathematics Vol. 7, Springer Verlag, New York, 1998.
156
[SLA94] R. Slavkovic, M. Zivkovic, M. Kojic, Enhanced 8-node three-dimensional
solid and 4-node shell elements with incompatible generalized
displacements, Commun. Numer. Meth. Engng. 10 (1994) 699709.
[STE94] P. Steinmann, C. Miehe, E. Stein, On the localization analysis of
orthotropic Hill-type elastoplastic solids, J. Mech. Phys. Solids, 42
(1994) 1969-1994.
[STO82] H. Stolarski, T. Belytshcko, Membrane locking and reduced integration
for curved elements, Journal of Applied Mechanics. 49 (1982) 172176.
[STO83] H. Stolarski, T. Belytshcko, Shear and membrane locking in curved C
0

elements, Comput. Methods Appl. Mech. Engrg. 41 (1983) 279-296.
[STO95] H. Stolarski, T. Belytschko, S.H. Lee, A review of shell finite elements and
corotational theories, Computational Mechanics Advances. 2 (1995) 125-
212.
[SZE00] K. Y. Sze, L. Q. Yao, A hybrid stress ANS solid-shell element and its
generalization for smart structure modelling. Part I*solid-shell element
formulation, Int. J. Numer. Meth. Engng. 2000; 48: 545564.

[T]

[TAN05] X.G. Tan and L. Vu-Quoc, Efficient and accurate multilayer solid-shell
element: Non-linear materials at finite strain, Int. J. Numer. Meth. Engng.
63: 21242170 (2005).
[TAY76] R.L. Taylor, P.J. Beresford, E.L. Wilson, A non-conforming element for
stress analysis, Int. J. Numer. Meth. Engng. 10, 1211-1219 (1976).
[TAY00] R.L. Taylor, A mixed-enhanced formulation fortetrahedral finite
elements, Int. J. Numer. Meth. Engng. 47, 205-227 (2000).
[TAY01] R.L. Taylor, FEAP - A Finite Element Analysis Program: Theory Manual,
University of California, Berkeley, 2001,
http://www.ce.berkeley.edu/~rlt/feap/.
[TIM59] S.P. Timoshenko and S. Woinowsky-Krieger, Theory of plates and shells
- 2th Edition, Engineering Societies Monographs, New York: McGraw-Hill,
1959. US.

[V]

[VLA49] V. Vlasov, Some new problems on shells and thin structures,
Washington : National advisory committee for aeronautics, 1949.

[W]

[WIL02] K.J. Willam, Constitutive models for engineering materials,
Encyclopedia of Physical Science & Technology 3rd Edition, Volume 3
pp. 603-633, Academic Press, 2002.
[WIL73] E.L. Wilson, R.L. Taylor, W.P. Doherty, J. Ghaboussi, Incompatible
displacement modes, in: S.J. Fenves, N. Perrone, A.R. Robinson, W.C.
Sehnobrich (Eds.), Numerical and Computer Models in Structural
Mechanics, Academic Press, New York, 1973.
157
[WRI96a] P. Wriggers, S. Reese, A note on enhanced strain methods for large
deformations, Comput. Methods Appl. Mech. Engrg. 135 (1996) 201-209.
[WRI96b] P. Wriggers, R. Eberlein and S. Reese, A comparison of three-dimensional
continuum and shell elements for finite plasticity, Int. J. of Solids Structures
Vol. 33, No 20-22, p 3309-3326, 1996.

[Y]

[YAN00] H.T.Y Yang, S. Saigal, A. Masud, R.K. Kapania, A survey of recent shell
finite elements, Int. J. Numer. Meth. Engng. 47, 101-127 (2000).
[YOO99] J.W. Yoon, D.Y. Yang and K. Chung, Elasto-plastic finite element method
based on incremental deformation theory and continuum based shell
elements for planar anisotropic sheet materials, Computer Methods in
Applied Mechanics & Engineering 174, 23-56 (1999).

[Z]

[ZIE00a] O.C. Zienkiewicz and R.L. Taylor, The Finite Element Method , Volume
1: The Basis - 5th Edition, Butterworth-Heinemann, 2000. UK.
[ZIE00b] O.C. Zienkiewicz and R.L. Taylor, The Finite Element Method , Volume
2: Solid mechanics - 5th Edition, Butterworth-Heinemann, 2000. UK.
































158

CONTENTS


NOTATIONS AND SYMBOLS..................................................................................................i
FIGURE LIST ........................................................................................................................... iii
TABLE LIST..............................................................................................................................vi
Chapter 1. INTRODUCTION.................................................................................................1
INTRODUCTION.......................................................................................................................1
1.1 OBJECTIVES OF THE THESIS ............................................................................2
1.2 APPLICATION DOMAIN .....................................................................................4
SUMMARY OF THE THESIS...................................................................................................5
Chapter 2. BACKGROUND ON THE DEVELOPMENT OF SOLID-SHELL
ELEMENTS...........................................................................................................7
INTRODUCTION.......................................................................................................................7
2.1 THREE DIMENSIONAL STANDARD ELEMENT.............................................7
2.1.1 Difficulties with low-order standard elements ........................................................8
2.1.2 Solution for a locking free element .......................................................................15
2.2 INTRODUCTION TO CONVENTIONAL SHELLS ..........................................16
2.2.1 Classical shell theories ..........................................................................................17
2.2.2 Degenerated shell elements ...................................................................................18
2.3 INCOMPATIBLE DISPLACEMENT ELEMENT ..............................................21
2.3.1 Finite element formulation ....................................................................................21
2.4 EAS ELEMENT....................................................................................................23
2.4.1 Variational formulation .........................................................................................24
2.4.2 Linearization of discrete weak form......................................................................27
2.4.3 EAS parameters .....................................................................................................30
2.4.4 Patch test................................................................................................................32
2.4.5 Equivalence between EAS and Hellinger-Reissner elements ...............................34
2.5 ANS ELEMENT ...................................................................................................36
2.5.1 Kinematics in natural coordinate system...............................................................37
2.5.2 Classical ANS formulation....................................................................................38
2.5.3 Some variational basis for the ANS method .........................................................41
2.6 NUMERICAL RESULTS.....................................................................................44
2.6.1 Membrane patch test..............................................................................................44
2.6.2 Out-of-plane bending patch test ............................................................................45
2.6.3 Eigenvalues analyses of a rectangle ......................................................................46
2.6.4 Circular cantilever beam at large displacements ...................................................47
2.6.5 Scordelis-Lo roof with rigid end diaphragms........................................................48
2.6.6 Regular block with nearly incompressible material ..............................................50
CONCLUSION .........................................................................................................................51
APPENDIX of Chapter 2 ..........................................................................................................52
Chapter 3. SOLID-SHELL ELEMENTS FOR FINITE DEFORMATION.....................56
INTRODUCTION.....................................................................................................................56
3.1 LOCKING PHENOMENA WITH SOLID-SHELL.............................................57
3.1.1 Curvature (trapezoidal) locking.............................................................................58
3.1.2 Membrane locking.................................................................................................60
3.2 KINEMATICS OF SOLID-SHELL......................................................................62
3.3 AN ALTERNATIVE ANS TECHNIQUE FOR TRANSVERSE SHEAR
LOCKING REMOVAL.........................................................................................65
159
3.3.1 Cubic hexahedral ANS element (ANSn)...............................................................65
3.3.2 Distorted hexahedral ANSn element .....................................................................69
3.4 COMBINED ANS-EAS SOLID-SHELL ELEMENT..........................................73
3.5 ENHANCED QUANTITIES ................................................................................77
3.5.1 Consistent deformation gradient............................................................................77
3.5.2 Local static condensation.......................................................................................77
3.6 NUMERICAL RESULTS AND DISCUSSION...................................................78
3.6.1 Patch tests ..............................................................................................................79
3.6.2 Eigenvalue analysis of an incompressible cube ....................................................79
3.6.3 Eigenvalues of a square plate ................................................................................82
3.6.4 Square plate at large displacements and strains.....................................................84
CONCLUSION .........................................................................................................................86
Chapter 4. ELASTIC APPLICATIONS ..............................................................................87
INTRODUCTION.....................................................................................................................87
4.1 LINEAR APPLICATIONS...................................................................................87
4.1.1 Cantilever beam under pure bending.....................................................................87
4.1.2 Twisted beam with warping effects.......................................................................88
4.1.3 Clamped and simply supported plates under uniformed pressure.........................90
4.1.4 Square clamped plates with concentrated loads ....................................................93
4.1.5 Simply supported square plate with various thickness and distorted mesh...........94
4.1.6 Pinched cylinder with rigid end diaphragms .........................................................95
4.1.7 Morley spherical shell ...........................................................................................96
4.1.8 Thick-walled cylinder............................................................................................98
4.2 NONLINEAR APPLICATIONS ..........................................................................99
4.2.1 Cantilever in large displacement ...........................................................................99
4.2.2 Morley spherical shell large deformation case.................................................100
4.2.3 Slit annular plate under line force........................................................................101
CONCLUSION .......................................................................................................................103
Chapter 5. PLASTIC APPLICATIONS.............................................................................104
INTRODUCTION...................................................................................................................104
5.1 FINITE STRAIN THEORY................................................................................104
5.1.1 Multiplicative split...............................................................................................104
5.1.2 Yield condition ....................................................................................................106
5.1.3 Flow rule..............................................................................................................106
5.1.4 Elastoplastic tangent moduli................................................................................108
5.2 J2 MATERIAL MODEL.....................................................................................109
5.2.1 Multiplicative split and elastic response..............................................................109
5.2.2 Flow rule and yield function................................................................................110
5.3 PLASTICITY APPLICATIONS.........................................................................113
5.3.1 Cantilever at large elastoplastic deformation ......................................................113
5.3.2 Pinched cylinder at large elastoplastic deformations ..........................................120
5.3.3 Springback of unconstrained cylindrical bending ...............................................130
CONCLUSION .......................................................................................................................146
Chapter 6. GENERAL CONCLUSIONS AND FUTURE WORKS................................147
CONCLUSIONS.....................................................................................................................147
FUTURE WORKS..................................................................................................................148
REFERENCE ..........................................................................................................................149