Comput Mech

DOI 10.1007/s00466-011-0653-y
ORIGINAL PAPER
Extended rotation-free shell triangles with transverse shear
deformation effects
Francisco Zrate Eugenio Oate
Received: 28 January 2011 / Accepted: 9 October 2011
Springer-Verlag 2011
Abstract The paper extends recent work of the authors to
include transverse shear effects on rotation-free triangular
element for plates (Oate and Zrate in Int J Numer Meth-
ods Eng 83(2):196227, 2010). Two newshell triangular ele-
ments are presented, the EBST+and the EBST+1. Transverse
shear deformation effects are important for thick shells, as
well when the shell is laminated or formed by composite
material. The ingredients for the element formulation are:
a Hu-Washizu type mixed functional and linear interpola-
tion for the displacement eld. In both elements presented
a nite volume approach is used for computing the bend-
ing moments and the curvatures over a patch of elements.
The nodal translational degrees of freedom of the original
enhanced basic shell triangle (EBST) are extended with the
two shear deformation angles via two different approaches.
The rst one uses a linear interpolation of the rotation angles
inside the element (EBST+) and the second one assumes a
constant eld for the rotation angles (EBST+1). For the thin
shell case the shear angles vanish and the newelements repro-
duce the good behaviour of the original thin EBST element.
As a consequence the elements can reproduce the solutions
for thick to thin shells situations without exhibiting shear
locking. The numerical solution for the thick shell case can
be found iteratively starting from the deection values for
the Kirchhoff theory using the original thin EBST element.
F. Zrate E. Oate (B)
Centre Internacional de Mtodes Numrics en
Enginyeria (CIMNE), Barcelona, Spain
e-mail: [email protected]
F. Zrate E. Oate
Universitat Politcnica de Catalunya (UPC), C. Gran Capitn s/n,
Campus Nord, Barcelona 08034, Spain
e-mail: [email protected]
Examples of the good performance of the new rotation-free
shell triangles are given.
Keywords Rotation-free shell triangle EBST element
Thick and thin shells Finite elements Transverse shear
deformation
1 Introduction
The need for efcient shell elements is essential for solving
large scale and industrial problems such as those occurring
in the analysis of shell structures in civil, mechanical, aero-
space and naval engineering among others. Development of
efcient and robust shell elements has been a eld of con-
stant research in the last four decades. The idea of using
translational variables at the nodes is not new at all; for
a comprehensive review of shell elements see [130]. In
essence all methods attempt to express the curvature eld
over an element in terms of the displacements of a collection
of nodes belonging to a patch of adjacent elements. Oate
and Cervera [13, 14] proposed a general procedure of this
kind combining nite element and nite volume concepts
for deriving thin plate elements of triangular and quadrilat-
eral shapes with translational degree of freedom only. This
approach has recently formalised and extended to shell anal-
ysis in [16, 17]. The basic components of the new elements
are a mixed Hu-Washizu formulation, a standard discretiza-
tion of the shell surface into three node triangles, a linear
FE approximation of the deexion eld within each trian-
gle and a nite volume (FV) type approach for computing
the curvature and bending moment elds within appropriate
non-overlapping control domain. For the triangular elements
considered here, the control domain is formed by the element
itself and the three adjacent triangular elements dening a
1 3
Comput Mech
patch of elements. This allows us dening a quadratic inter-
polation of the geometry from the position of the six nodes.
In this article, the geometry of the four element patch is
expressed as a quadratic polynomial in terms of the coor-
dinates of the six nodes in the patch. This denes a lin-
ear eld for the displacement gradients and the membrane
strains, similarly as for the EBST element [26]. A nite vol-
ume approach is used for computing the curvatures over the
central triangle in the patch in terms of the displacement gra-
dients at the mid-side nodes. The transverse shear effects are
introduced by enhancing the original EBST element with the
two shear angles via two different procedures: In the EBST+
element a linear interpolation of the rotation angles inside
the element is used, while in the EBST+1 element the shear
angles assumed to be constant inside the elements.
The content of this paper includes a brief description of
the formulation of the EBST+ and EBST+1 elements, and
several examples of analysis of thick and thin shells showing
the good behaviour of the elements.
2 Basic theory
2.1 Reissner-Mindlin folded plate theory
Let us consider the plate of Fig. 1 dened in the local coor-
dinate system x

, y

, z

. This plate domain is considered to

pertain to a folded shell, or to a curved shell modelled by an
assembly of at elements. We will assume Reissner-Mindlin
conditions to hold. The displacement eld can therefore be
written in terms of the mid-plane displacement variables, i.e.
u

(x

, y

, z

) = u

0
(x

, y

) z

x
(x

, y

)
v

(x

, y

, z

) = v

0
(x

, y

) z

y
(x

, y

)
w

(x

, y

, z

) = w

0
(x

, y

) (1)
with

x
=
w

+
x
,
y
=
w

+
y
(2)
or

=
_

x

y

_
T
= w

(3)
with

_

x

_
T
and

=
_

x

y

_
T
(4)
where u

, v

, w

are the local cartesian displacements of a

shell point, u

0
, v

0
, w

0
are the local cartesian displacements
of a point in the middle plane of the shell,
x
,
y
are the rota-
tions and
x
,
y
are angles coinciding with the transverse
shear deformations (Fig. 1).
Fig. 1 Sign convention for the deection and the rotations in a shell
The generalized membrane, bending and shear strain vec-
tors are dened as [31]

m
=
_

x

y

x

_
T
=
_
u

0
x

,
v

0
y

,
_
u

0
y

+
v

0
x

__
T
= L
m
u

0
(5a)

b
=
_

x

y

x

_
T
=
_

x

,

y

,
_

+

y

__
T
= L
b

(5b)

s
=
_

z

y

_
T
=
_
w

x
,
w

_
T
= L
s
_
u

_
(5c)
where
_

x

y

x

_
,
_

x

y

x

_
and
_

z

y

_
are
the local membrane, bending and transverse shear strains
respectively and
L
m
=
_
_
_

0 0
0

y

0
_

_,
L
b
=
_
_
_

0
0

y

_, L
s
=
_
0 0

x

1 0
0 0

y

0 1
_
(6a)
u

0
=
_
u

0
v

0
w

0
_
T
(6b)
Substituting Eqs. 2 into 5c gives

z
=
x
,
y

z
=
y
(6c)
i.e. the angles
x
,
y
coincide (with opposite sign) with the
transverse shear deformations. In the following
x
and
y

will be called transverse shear angles (or shear angles in

short).
1 3
Comput Mech
The set of governingequations canbe expressedinintegral
formstarting fromthe following Hu-Washizu type functional
[18, 32]

=
1
2
_
_
_ _
A

T
m
D

m
dA +
_ _
A

T
b
D

b
dA
+
_ _
A

T
s
D

s
dA
_
_
+
_ _
A
_
L
m
u

m
_
T

m
d A
+
_ _
A
_
L
b

b
_
T

b
d A
+
_ _
A
_
L
s

s
_
T

s
d A

_ _
A
_
q
T
u

0
_
d A
n
c

i =1
u
T
i
p
i
(7)
where the distributed loading is q =
_
q
x
q
y
q
z

_
T
; p
i
=
_
p
x
p
y
p
z

_
T
is a concentrated point force acting at point i ,
n
c
is the number of points with external concentrated forces
and Ais the area of the shell. Concentrated bending moments
have been excluded for simplicity.
In Eq. 7

m
=
_
N
x
N
y
N
x

_
T
is the membrane force
vector,

b
=
_
M
x
M
y
M
x

_
T
is the bending moment
vector,

s
=
_
Q
x
Q
y

_
T
is the shear force vector. The
membrane bending and shear constitutive matrices for the
isotropic case are:
D

m
=
Et
_
1
2
_
_
_
1 0
1 0
0 0
1
2
_
_
,
D

b
=
Et
3
12
_
1
2
_
_
_
1 0
1 0
0 0
1
2
_
_
,
D

s
= Gt
_
1 0
0 1
_
(8)
where t is the shell thickness and E, and G are the Young
modulus, the Poissons ratio and the transverse shear modu-
lus, respectively and is the shear correction factor; = 5/6
has been taken for isotropic shells.
2.2 Using the deection and the shear angles
as main variables
We will express Eqs. 57 in terms of the local deexions u

0
and the shear angles

. The resulting shell theory contains

the standard expressions of Kirchhoff thin shell theory plus
additional transverse shear deformation terms.
Substituting Eq. 2 into 5b and 5c gives

b
= L
b

=
_

2
w

0
x
2
+

x

2
w

0
y
2
+

y

, 2

2
w

0
x

+
_

+

y

__
T
= L
w
u

0
+ L
b

(9a)

s
=
_

x

y

_
T
=

(9b)
with L
b
dened in Eq. 6a and
L
w
=
_
_
_
0 0

2
x
2
0 0

2
x
2
0 0

2
x

_ (10)
Substituting Eqs. 3 and 9 into the functional of Eq. 7 gives

=
1
2
_
_
_ _
A

T
m
D

m
dA +
_ _
A

T
b
D

b
dA

_ _
A

s
dA
_
_
+
_ _
A
_
L
m
u

m
_
T

m
d A
+
_ _
A
_
L
w
u

0
+ L
b

b
_
T

b
d A
+
_ _
A
_
L
s

s
_
T

s
d A

_ _
A
_
q u

0
+
T
m
_
d A
n
c

i =1
u

i
p
i
(11)
Note that by making the local shear angles equal to zero we
recover the variational form of the standard Kirchhoff thin
folded shell theory [3135].
The constitutive equations can be obtained by taking the
variation of

with respect to

m
,

b
and

s
. This gives
Membrane constitutive equation
_ _
A

T
m
_
D
m

m
_
dA = 0 (12)
Bending constitutive equation
_ _
A

T
b
_
D
b

b
_
dA = 0 (13)
Shear constitutive equation
_ _
A

T
s
_
D
s

s
_
dA = 0 (14)
1 3
Comput Mech
The variation of

with respect to

b
gives the relationship
between bending strains, deection and shear angles, i.e.
_ _
A

T
b
_
L
w
u

0
+ L
b

b
_
dA = 0 (15)
Finally the equilibrium equations can be found by taking the
variation of

with respect to u

0
and

, i.e.
_ _
A
_
L
m
u

0
_
T

m
dA +
_ _
A
_
L
w
u

0
_
T

b
dA

_ _
A
_
u

0
q
_
dA
n
c

i =1
u

i
p
i
= 0 (16a)
_ _
A
_
_
L
b

_
T

b

T

s
_
dA = 0 (16b)
Equations 1216 are the basis of the nite element discreti-
zation dened in the next sections.
3 Finite element/nite volume discretization
The main difference between the new EBST+ element and
the EBST+1 one is the discretizations chosen for the shear
angles. Both elements have three nodes with the local dis-
placements being the translational degrees of freedomat each
node. For the EBST+ element there are two shear angles at
each node, while for the EBST+1 element there are only two
shear angles at the center of the element. These angles can
be considered as rotational degrees of freedom.
For the modelling of bending effects a patch of elements
is dened by the central triangle and the three adjacent ele-
ments (Fig. 2). This patch allows us to dene the curvature
eld within the central triangle by evaluating the rotations

at the mid-side nodes of the central element in terms of the

six nodal displacements and the shear angles of the central
element for the EBST+ element, and in terms of the six nodal
displacements and the shear angle at the center of each of the
four element patch for the EBST+1 element.
Figure 3 shows the local Cartesian coordinate system for
the patch. For the central element the unit vector t
1
(associ-
ated to the local coordinate
1
) is directed along side 3; t
3
(associated to the normal coordinate
3
) is the unit normal to
the plane and, nally, t
2
= t
3
t
1
.
For the EBST+ and the EBST+1 elements the nodal vari-
ables are interpolated within each 3-noded node triangle as
(from now onwards we will remove the zero sub indexes, as
all the displacements are referred to the shell mid plane):
u

=
_
_
u

_
_
=
3

i =1
L
i
u

i
(17a)
EBST+ element

=
_

_
=
3

i =1
L
i

i
(17b)
Fig. 2 Element patch and
kinematic variables for the
EBST+ (a) and EBST+1 (b)
(0,0)
(1,-1)
(-1,1)
(0,1)
(1,1)
(1,0)
(a) (b)
Fig. 3 Patch of four elements (a) in spatial coordinates and (b) in natural coordinates
1 3
Comput Mech
EBST+1 element

=
_

_
=
(e)

(17c)
where L
i
are the standard linear shape functions of the 3-no-
ded triangle given by [1, 34]
L
i
=
1
2A
e
_
a
e
i
+ b
e
i
x

+ c
e
i
y

_
, a
e
i
= x

e
j
y

e
k
x

e
k
y

e
j
, b
e
i
= y

j k
, c
e
i
= x

kj
(18a)
where
y

j k
= y

e
j
y

e
k
, x

e
kj
= x

e
k
x

e
j
i, j, k = 1, 2, 3 (18b)
In Eqs. 17a, 17b and 17c
u

i
=
_
u

i
v

i
w

i
_
T
,

i
=
_

i

y

i
_
T
,
(e)

=
_

(e)
x

(e)
y

_
T
(18c)
The geometry of the four-element patch in the reference
conguration is described by a quadratic interpolation as
=
_
x, y, z
_
T
=
6

i =1
N
i
(
1
,
2
)
i
(19a)
The non-standard quadratic shape functions for the six
nodes in the patch are [26]:
N
1
=
3
+
1

2
, N
4
=

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

3
, N
5
=

1
2
(
1
1) (19b)
N
3
=
2
+
3

1
, N
6
=

2
2
(
2
1)
with
3
= 1
1

2
The interpolation of Eq. 19a allows us computing the dis-
placement gradients at selected points in the patch. This com-
putation is performed at three points over the boundaries of
the central element of the patch. These points are located at
the mid-point of each side, denoted G
1
, G
2
and G
3
in Fig. 3b.
This choice has the following advantages [26]:
Gradients at these points depend only on the nodes per-
taining to the two elements adjacent to the side. This can
be veried by sampling the derivatives of the shape func-
tions at each mid-side point.
When gradients are computed at the common mid-side
point from two adjacent elements, the same values are
obtained, since the coordinates of the same four points
are used. That is, the gradient at the mid-side points is
independent of the element where they are computed. A
side-oriented implementation of the nite element will
lead to a unique evaluation of the gradients per side.
The Cartesian derivatives of the shape functions are com-
puted by the standard expression:
_

x
N
i

y
N
i
_
= J
T
_

1
N
i

2
N
i
_
(20)
where the Jacobian matrix is
J =
_

1
t
1

2
t
1

1
t
2

2
t
2
_
(21)
With the previous definitions the deformation gradient on the
mid surface is:
_

x

y

_
=
_

1

2

_
J
1
(22)
3.1 Membrane eld
The membrane strains within the central triangle are obtained
using a linear assumed strain eld

m
(
1
,
2
) =
3

i =1
L
i

i
m
(23)
where

i
m
are the membrane strains computed at the three
mid side points G
j
( j = 1, 2, 3). The gradient of the mid-
surface at the mid side point is computed from the quadratic
interpolation by

x

Gj
=
3

i =1
[(
x
N
i
)
i
]
j
+
_
(
x
N
i +3
)
i +3
_
j
(24)

y

Gj
=
3

i =1
__

y
N
i
_

i
_
j
+
__

y
N
i +3
_

i +3
_
j
From the standard expression of the Green-Lagrange strains
and using Eqs. 23 and 24 gives the membrane strain vector
as [26]

m
=
3

i =1
L
i

i
m
=
3

i =1
L
i
_
_

x

Gi

x

Gi
1

y

Gi

y

Gi
1
2
x

Gi

y

Gi
_
_
(25)
Membrane constitutive equation

e
_ _
A

e
T
m
_
D
m

e
m

e
m
_
dA = 0 (26)
Recalling that the virtual membrane deformations are arbi-
trary we obtain the following constitutive equation for the
assumed membrane forces for each element.

e
m
= D
e
m

e
m
, D
e
m
=
1
A
e
_ _
A
D
m
d A (27)
where D
e
m
is the average membrane constitutive matrix for
the element.
1 3
Comput Mech
3.2 Curvatures and bending moments eld
As it is typical in rotation-free plate and shell triangles [13
18, 22, 25, 26, 35] we describe the bending moments and the
curvatures by constant elds within the triangle, i.e.

b
=

e
b
,

b
=

e
b
(28)

b
=

e
b
,

b
=

e
b
where ()
e
denotes constant values within the triangle.
Equations 28 are introduced into the governing integral
Eqs. 13,15,16 as shown next.
Bending constitutive equation

e
_ _
A

e
T
b
_
D
b

e
b

e
b
_
dA = 0 (29)
Recalling that the virtual curvatures are arbitrary we obtain
the following constitutive equations for the assumed (con-
stant) bending moments for each element.

e
b
= D
e
b

e
b
, D
e
b
=
1
A
e
_ _
A
D
b
d A (30)
where D
e
b
is the average bending constitutive matrix for the
element.
Relationship between the curvatures, the deection and
the shear angles.
Substituting Eq. 28 into Eq. 15 gives

e
_ _
A

eT
b
_
L
w
u

+ L
b

e
b
_
dA = 0 (31)
As the virtual bending moments are arbitrary and

e
b
is con-
stant within each element we obtain

e
b
=
1
A
e
_ _
A
e
_
L
w
u

+ L
b

_
d A (32)
We describe next the computation of the curvatures via Eq. 32
for the EBST+ and EBST+1 elements.
3.2.1 EBST+ element
For the EBST+ element the term L
w
u

in Eq. 32 can be inte-

grated by parts to give
__
A
e
_
L
w
u

_
d A
=
_

e
_
_
n
x
0
0 n
y

n
y
n
x

_
_
w

d
e
=
_

e
Tw

d
e
(33a)
with
T =
_
_
n
x
0
0 n
y

n
y
n
x

_
_
(33b)
Using Eqs. 24 and 33a and noting that w

i
= t
3
u
i
where
u
i
=
_
u
i
v
i
w
i
_
T
is the global displacement vector of a
point, we can write Eq. 32 in terms of the global displace-
ment vector u and the shear angles as:

e
b
=
1
A
e
_

e
T
_

x

Gi
t
3

y

Gi
t
3
_
ud
+
1
A
e
_ _
A
e
L
b

d A (34)
and n
x
, n
y
are the components to the outward unit normal n
to the boundary of the element
e
. Equation 34 expresses the
bending strains in terms of the global displacement and the
derivatives of the shear angles. These derivatives are constant
within each element for a linear interpolation of

.
3.2.2 EBST+1 element
Having in mind the different interpolations used for the shear
angles in the EBST+1 element and following the same solu-
tion path as for the EBST+ element; the integral in Eq. 32
can be written as
__
A
e
_
L
w
u

+ L
b

_
d A
=
_

e
_
_
n
x
0
0 n
y

n
y
n
x

_
_
_
w

_
d
e
(35a)
From Eqs. 24 and 35a we can write Eq. 32 as:

e
b
=
1
A
e
_

e
T
__

x

Gi
t
3

y

Gi
t
3
_
u +
_

Gi

Gi
__
d (35b)
In Eq. 35b
x

Gi
and
y

Gi
are the shear angles at every
mid-side node Gi (Fig. 3). These angles are computed as the
average of the values in the two adjacent elements to each
mid-side node.
3.3 Shear forces-shear angles relationship
From Eq. 9b we have

s
=

Therefore
D
s

s
=

s
= D
s

(36)
1 3
Comput Mech
3.4 Equilibrium equations
3.4.1 EBST+ element
Substituting Eqs. 27, 30, 34 and 36 into 16 and noting that the
assumed resultant stress and strains elds are constant within
each triangle gives the two following equilibrium equations:

e
_
_
_ _
A
e
_
L
m
u

_
T
d A
_
_

e
m
+

e
_
_
_

e
T
_

x

Gi
t
3

y

Gi
t
3
_
d
_
_

e
b

_ _
A
e
u

qdA
n
c

i =1
u

i
p
i
= 0 (37a)

e
_
_
_ _
A
e
_
L
b

_
T
dA
_
_

e
b

_
_
_ _
A
e

T
dA
_
_

e
s
= 0 (37b)
3.4.2 EBST+1 element
Substituting Eqs. 27, 30, 35b and 36 into 16 noting that the
assumed resultant stress and strains elds are constant within
each triangle gives:

e
_
_
_ _
A
e
_
L
m
u

_
T
d A
_
_

e
m
+

e
_
_
_

e
T
_

x

Gi
t
3

y

Gi
t
3
_
d
_
_

e
b

_ _
A
e
u

qdA
n
c

i =1
u

i
p
i
= 0 (38a)

e
_
_
_

e
T
_

Gi

Gi
_
d
_
_

e
b

_
_
_ _
A
e

T
dA
_
_

e
s
= 0 (38b)
The sums in Eqs. 37 and 38 extend over all the elements in
the mesh.
4 Element implementation
In order to simplify the implementation of the EBST+ and
EBST+1 elements we will assemble the displacements in the
global coordinate system and the shear rotations in the local
coordinate system. The formulation of both elements is very
similar. However, for clarity purposes their implementation
is presented separately for each element.
4.1 EBST+ element
The steps for the implementation of the EBST+ element are
the following.
4.1.1 Bending strain matrices for the EBST+ element
From Eq. 32 we can write

e
b
=
e
w
+
e

(39a)
where

e
w
=
1
A
e
_

e
T
_

x

Gi
t
3

y

Gi
t
3
_
ud,

=
1
A
e
_ _
A
e
L
b

d A (39b)
The numerical integration of the boundary integral of
e
w
in
Eq. 39b results in a sum over the integration points on the
element boundary

e
w
=
1
A
e
_

e
T
_

x

Gi
t
3

y

Gi
t
3
_
ud
=
1
A
e
3

i =1
l
i
_
_
n
x
0
0 n
y

n
y
n
x

_
_
_

x

Gi
t
3

y

Gi
t
3
_
u (40)
Equation 40 can be simplied as

e
w
= 2
3

i =1
_
_

x
L
i
0
0
y
L
i

y
L
i

x
L
i
_
_
_

x

Gi
t
3

y

Gi
t
3
_
u (41)
Expanding the terms of Eq. 41 and rearranging the resultant
expression in a matrix formwe obtain the following equation
describing the generalized bending strain matrix B
w

e
w
= 2CGUu = B
w
u
e
(42a)
with B
w
= 2CGU (42b)
and
C =
_
C
1
C
2
C
3
_
, C
i
=
_
_

x
L
i
0
0
y
L
i

y
L
i

x
L
i
_
_
(43a)
1 3
Comput Mech
G=
_
_
_
_
_
_
_
_

x
N
1
|
G1

x
N
2
|
G1

x
N
3
|
G1

x
N
4
|
G1
0 0

y
N
1

G1

y
N
2

G1

y
N
3

G1

y
N
4

G1
0 0

x
N
1
|
G2

x
N
2
|
G2

x
N
3
|
G2
0
x
N
5
|
G2
0

y
N
1

G2

y
N
2

G2

y
N
3

G2
0
y
N
5

G2
0

x
N
1
|
G3

x
N
2
|
G3

x
N
3
|
G3
0 0
x
N
6
|
G3

y
N
1

G3

y
N
2

G3

y
N
3

G3
0 0
y
N
6

G3
_

_
(44)
U =
_
_
_
_
_
_
_
_
t
1
3
t
2
3
t
3
3
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 t
1
3
t
2
3
t
3
3
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 t
1
3
t
2
3
t
3
3
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 t
1
3
t
2
3
t
3
3
0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 t
1
3
t
2
3
t
3
3
0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 t
1
3
t
2
3
t
3
3
_

_
(45)
Substituting into (43a) the expression of
x
L
i
(Eq. 18a)
gives:
C
i
=
1
2A
e
_
_
b
e
i
0
0 c
e
i
c
e
i
b
e
i
_
_
(43b)
In Eq. 42 u
e
=
_
u
T
1
u
T
2
u
T
3
u
T
4
u
T
5
u
T
6
_
T
lists the nodal
global displacements of the four element patch, with u
i
=
_
u
i
v
i
w
i
_
T
.
For the second term in the r.h.s. of Eq. 39a we obtain:

=
1
A
e
__
A
e
L
b

d A
= B

e
=
_
B
1
B
2
B
3
_

e
(46)
with

e
=
_
_
_

3
_

_,

i
=
_

i
_
and B
i
=
1
A
e
__
A
e
L
b
L
i
d A (47)
Recalling that L
i
are the linear shape functions yields:
B
i
=
_
_

x
L
i
0
0
y
L
i

y
L
i

x
L
i
_
_
=
1
2A
e
_
_
b
e
i
0
0 c
e
i
c
e
i
b
e
i
_
_
= C
i
(48)
Finally substituting Eqs. 42 and 47 into 39a and this into 30
gives:

e
b
= D
e
b
_
B
w
u
e
+ B

e
_
(49)
4.1.2 Shear strain matrices for the EBST+ element
The shear strainmatrixis obtainedbysubstitutingthe approx-
imation for

from Eqs. 17 into Eq. 36 to give:

s
= D
s

= D
s
_
3

i =1
L
i

e
i
_
= D
s
N

e
(50)
with
N =
_
L
1
0 L
2
0 L
3
0
0 L
1
0 L
2
0 L
3
_
=
_
N
1
N
2
N
3
_
(51)
4.2 EBST+1 element
Many of the expressions dened in the previous section can
be used for the EBST+1 element. The main differences are
described next.
4.2.1 Bending strain matrices for the EBST+1 element
The expression for the curvature
e
w
and the corresponding
curvature matrix B
w
coincide with the expressions for the
EBST+ element (Eqs. 40 and 42).
Following Eq. 35b the curvature
e

due the rotation angles

is obtained as:

=
1
A
e
_

e
T
_

Gi

Gi
_
d (52)
The numerical integration of the boundary integral of
e

in
Eq. 52 results in a sum over the integration points on the
element boundary, i.e.

=
1
A
e
_

e
T
_

Gi

Gi
_
d
=
1
A
e
3

i =1
l
i
_
_
n
x
0
0 n
y

n
y
n
x

_
_
_

Gi

Gi
_
(53)
Equation 53 can be simplied as

= 2
3

i =1
_
_

x
L
i
0
0
y
L
i

y
L
i

x
L
i
_
_
_

Gi

Gi
_
(54)
Expanding the terms of Eq. 54 and rearranging the resultant
expression in matrix form we obtain the following equation
describing the generalized bending strain matrix B

= 2CU

= B

(55)
1 3
Comput Mech
where B

= 2CU

, the expressions for matrix C is given in

Eq. 43a and the expression for U

is
U

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

_
(56)
Matrix U

is obtained by computing the shear angles

x

Gi
and
y

Gi
in Eq. 35b at a mid-side point as the mean value
of the shear rotation angles of the two adjacent elements
sharing the side. Another strategy is to transform the local
shear angles to the global coordinate system and evaluate
their mean value in this system. Then, transform back the
angles to the local coordinate system, neglecting the nor-
mal rotation to the element surface. We have found that a
faster way with no significant loose of accuracy is evalu-
ating directly the mean value considering that the two adja-
cent local systems do not undergo big changes. This assump-
tion holds if the mesh is ne enough and the shell surface is
smooth.
InEq. 55vector

=
_

M
x

M
y

1
x

1
y

2
x

2
y

3
x

3
y

_
T
lists all the shear angles of the four-element patch according
to Fig. 2.
Finally the bending moments in Eq. 30 are computed as:

e
b
= D
e
b
_
B
w
u
e
+ B

_
(57)
4.2.2 Shear strain matrices for the EBST+1 element
The derivation of the shear strain matrix is straightforward
due to the constant values of the shear rotations inside the
element (Eq. 17c). Therefore

s
= D
s

= D
s

(e)

(58)
4.2.3 Membrane strain matrices for the EBST+
and EBST+1 elements
In both elements the membrane eld is described in the same
way. Recalling that the membrane deformations were dened
in Eq. 25, the membrane strain matrix at each mid side point
G
i
can be obtained by
_

m
_
i
=
_
_

xy
_
_
i
=
_
_

x

T
Gi
0
T
3x1
0
T
3x1

y

T
Gi

y

T
Gi

x

T
Gi
_
_
G u
e
= (B
m
)
i
u
e
(59)
Fig. 4 Local cartesian systemfor the treatment of boundary conditions
Fig. 5 Cylindrical roof under uniform load. Geometry, boundary con-
ditions and meshes
where matrix G is given in Eq. 44. From Eq. 59 we deduce
(B
m
)
i
=
_
_

x

T
Gi
0
T
3x1
0
T
3x1

y

T
Gi

y

T
Gi

x

T
Gi
_
_
G (60)
4.2.4 Stiffness matrix and nodal force vector
for the EBST+ and EBST+1 elements
The derivation of the stiffness matrices is similar for both
elements.
Substituting Eqs. 49, 50 and 59 into the equilibrium
Eqs. 37a and 37b for the EBST+ element or Eqs. 57, 58 and
59 into the equilibrium Eqs. 38a and 38b for the EBST+1
element yields.
1 3
Comput Mech
Table 1 Cylindrical roof under uniform weight (normalized displacement at point B)
Scordelis Lo shell normalized
vertical
EBST EBST+ EBST+1 TLQL QLQL QLLL
Displacement at point B W
A
/W
F
A
W
A
/W
F
A
W
A
/W
F
A
W
A
/W
F
A
W
A
/W
F
A
W
A
/W
F
A
N Nodes Elements N Nodes Elements
5 5 36 50 0.979 0.981 0.981 4 4 25 32 0.728
10 10 121 200 0.980 0.981 0.981 4 4 25 16 1.687 1.108
20 20 441 800 0.984 0.985 0.985 8 8 81 168 0.661
30 30 961 1800 0.987 0.988 0.988 8 8 81 84 1.174 0.951
40 40 1681 3200 0.988 0.990 0.990 16 16 289 512 0.876
16 16 289 256 1.075 0.972
32 32 1089 2048 0.951
32 32 1089 1024 1.058 0.989
W
F
A
[36] = 0.3024

e
_
_
_ _
A
e
_
u
e
_
T
B
T
m
D
e
m
B
m
u
e
d A
_
_
+

e
_
u
e
_
T
B
T
w
D
e
b
_
B
w
u
e
+ B

e
_
A
e

_ _
A
e
__
3

i =1

N
i
u

_
q
_
dA
n
c

i =1
u

i
p
i
= 0
(61a)

e
_

e
_
T
_
B
T

D
e
b
_
B
w
u
e
+ B

e
_
A
e
+
_
_
_ _
A
e

N
T
D
s

Nd A
_
_

e
_
_
= 0 (61b)
After simplication of the virtual nodal deections and the
virtual nodal shear angles we nally obtain the matrix system
of equilibrium equations as
K
u
u + K
u

= f
u
(62)
K
T
u
u + K

= 0
where vectors u and

contain the nodal displacements (in
global axes) and the nodal local shear angles for the whole
mesh. The rest of the matrices and vectors are assembled
from the element contributions given by
_
K
e
u
_
i j
=
_
K
e
m
_
i j
+
_
K
e
b
_
i j
=
_ _
A
e
B
T
mi
D
e
m
B
mj
d A + B
T
wi
D
e
b
B
T
wj
A
e
_
K
e
u
_
i j
= B
T
wi
D
e
b
B
T
j
A
e
(63a)
_
K
e

_
i j
= B
T
i
D
e
b
B
T
j
A
e
+
_ _
A
e

N
i
D
e
s

N
j
d A for the EBST+ (63b)
_
K
e

_
i j
= B
T
i
D
e
b
B
T
j
A
e
+
i j
D
e
s
A
e
for the EBST + 1
(63c)
where
i j
is the Kronecker delta
The element stiffness matrices K
e
b
and K
e
u
can be explic-
itly computed from the strain and constitutive matrices for
the element. Matrix K
e
m
for the EBST+ element is evaluated
with just one integration point and this has shown to give a
well conditioned membrane stiffness matrix, free of spuri-
ous zero-energy modes. The exact computation of
_
K
e

_
i j
for the EBST+ element requires a 3 Gauss point quadrature.
Excellent results have been obtained for both elements in
all examples analyzed using a reduced one-point integration
rule for
_
K
e

_
i j
which indicates no advantage in using a full
quadrature for these matrices. The simple one-point quadra-
ture allows computing all the element stiffness matrices in
Eq. 63b explicitly.
The equivalent nodal force vectors for both elements for
a distributed load q is
_
f
e
_
i
=
_ _
A
L
i
qdA (64)
A point load p
i
acting at a node i contributes the following
terms to the i th component of the global force vector
_
f
e
_
i
= p
i
(65)
The reactions (the nodal forces and the bending moments)
can be computed at the prescribed nodes from the nodal dis-
placement and rotations eld in the usual manner.
1 3
Comput Mech
-20.000
-10.000
0
10.000
20.000
30.000
40.000
50.000
60.000
70.000
80.000
0 5 10 15 20 25 30 35 40
TLQL QLQL
QLLL EXACT
EBST EBST+
EBST+1
-50
0
50
100
150
200
250
300
0 5 10 15 20 25 30 35 40
TLQL QLQL
QLLL EXACT
EBST+ EBST+1
-2200
-1800
-1400
-1000
-600
-200
200
0 5 10 15 20 25 30 35 40
TLQL QLQL
QLLL EXACT
EBST EBST+
EBST+1
0,6
0,8
1,0
1,2
1,4
1,6
1,8
10 100 1000 10000
EBST TLQL
QLQL QLLL
EBST+ EBST+1
Normalized vertical
displacement at point A
Mx 40 x 40 mesh
A
B
Qx 40 x 40 mesh

A
B
Ny 40 x 40 mesh

A B

A
Mx

Qx
Ny
Fig. 6 Cylindrical roof bending moments, shear and normal force dis-
tribution
5 Solution scheme
5.1 EBST+ element
An iterative computation of the displacements and shear
angles is considered in the implementation of the EBST+
element. This leads to memory savings and it has also the
advantage that for the rst iteration the value of the nodal
displacements corresponds to the thin shell solution of Kir-
chhoff theory, which is accurate enough for many practical
cases. The iteration solution scheme is as follows.
Fig. 7 Open semi-sphere dome under point loads. Geometry, bound-
ary conditions and meshes used
Step 1 Compute the nodal deections
1
u
K
u
1
u = f
u

1
u (Kirchoff thin shell solution) (66)
Step 2 Compute
i

, i 1
K

= f

K
T
u
i
u
i

(67)
Step 3 Compute the element shear angles
i

e
i

e
=
1
3
_
i

e
1
+
i

e
2
+
i

e
3
_
(68)
where
i

e
are the mean nodal shear angles for element e
and the i th iteration.
Step 4 Compute the smoothed nodal shear angles
i

k
i

k
=
1
n
k
n
k

j =1
i

j
, k = 1, N (69)
where n
k
is the number of elements sharing the node with
global numbering k and N is the total number of nodes in
the mesh.
Step 5 Compute
i
u, i > 1
K
u
i
u = f
u
K
u
i 1

k

i
u (70)
Return to step 2
Convergence of this iterative scheme for the deection
eld is quite fast (24 iterations), even for thick shells as
shown in the examples presented.
5.1.1 EBST+1 element
For the EBST+1 element the iterative scheme is simpler
because there is no need to perform an intermediate smooth-
1 3
Comput Mech
Table 2 Open semi-sphere dome under point loads (normalized displacement at point B)
Hemispherical dome normalized
horizontal
EBST EBST+ EBST+1 TLQL QLQL QLLL
Displacement at point B U
B
/U
8
B
U
B
/U
8
B
U
B
/U
8
B
U
B
/U
8
B
U
B
/U
8
B
U
B
/U
8
B
N M Nodes Elements N M Nodes Elements
4 5 36 40 0.815 0.237 0.237 4 4 25 32 0.667
8 10 99 160 0.805 0.807 0.807 4 4 25 16 1.086 0.054
16 20 357 640 0.951 0.952 0.952 8 8 81 168 0.484
32 40 1353 2560 0.979 0.979 0.979 8 8 81 84 1.022 0.688
40 50 2091 4000 0.984 0.984 0.984 16 16 289 512 0.968
16 16 289 256 0.989 0.989
32 32 1089 2048 0.999
32 32 1089 1024 0.999 0.999
U
s
B
[37] = 0.093
ing of the shear rotations. The iterative scheme is therefore
as follows:
Step 1 Compute the nodal deections
1
u by Eq. 66
Step 2 Compute
1

, i 1 by Eq. 67
Step 3 Compute
i
u, i > 1 by Eq. 70
Return to step 2
6 Boundary conditions
One particular feature of the rotation-free shell formulation
is the treatment of the boundary conditions when one of the
triangles belonging to the patch is missing [1618, 25, 26].
This is taken into account by ignoring the contribution of this
element when computing matrix B
w
for the EBST+ element
(Eq. 42) and matrices B
w
and B

for the EBST+1 element

(Eqs. 42 and 55) [18, 25, 26, 35].
6.1 Clamped edge (u = 0, = 0)
Prescribing the condition u = 0 is straightforward as the
displacements are the natural degrees of freedom of the ele-
ment. Thetreatment of thezerorotationconditionis explained
below.
6.1.1 EBST+
The condition of zero rotations at a clamped edge introduces
some modications in the terms of Eqs. 42. The condition of
= 0 along the clamped side |
s
implies that
|
s
= w

|
s
+ |
s
= 0 w

|
s
= |
s
(71)
where s is the side direction (Fig. 4).
Due the linear variation of the shear angles inside the cen-
tral element, the value along the clamped side is
|
s
=
1
2
_

i
+

j
_
(72)
Introducing Eqs. 71 and 72 into the definition of the curva-
tures Eq. 41 gives

e
w
= 2
2

i =1
_
_

x
L
i
0
0
y
L
i

y
L
i

x
L
i
_
_
_

x

Gi
t
3

y

Gi
t
3
_

1
A
e
l
s
2
T
3
_

i
+

j
_
(73)
Which leads to

e
w
= B
w
u
e
+ B

e
(74)
The bending stress eld of Eq. 50 is modied as

e
b
= D
e
b
_
B
w
u
e
+
_
B

+ B

e
_
= D
e
b
_
B
w
u
e
+

B

e
_
(75)
where

= B

+ B

(76)
The new matrix

B

substitutes matrix B

in Eq. 55
The above modications are only needed for elements
with a clamped edge.
6.1.2 EBST+1 element
Prescribing the condition of zero rotations is simpler than
for the EBST+ element. The condition of = 0 along the
clamped side |
s
implies that
|
s
= w

|
s
+ |
s
= 0 w

|
s
= |
s
= 0 (77)
1 3
Comput Mech
-2,5
-2,0
-1,5
-1,0
-0,5
0,0
0,5
1,0
1,5
0 9 18 27 36 45 54 63 72
TLQL QLQL
QLLL EXACT
EBST+ EBST+1
0,0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
0 9 18 27 36 45 54 63 72
TLQL QLQL
QLLL EXACT
EBST EBST+
EBST+1
-2,0
-1,5
-1,0
-0,5
0,0
0,5
1,0
1,5
2,0
0 10 20 30 40 50 60 70 80 90
TLQL QLQL
QLLL EXACT
EBST EBST+
EBST+1
0,0
0,2
0,4
0,6
0,8
1,0
1,2
10 100 1000 10000
EBST TLQL
QLQL QLLL
EBST+ EBST+1
Normalized horizontal
displacement at point B
M 40 x 50 mesh
M 40 x 50 mesh
A

A
B
C
A
Q 40 x 50 mesh
C


B
M

exact
M
Q

Fig. 8 Hemispherical dome Bending moments, shear and axial force
distribution
In this case it is possible to impose the trivial condition (77)
as both w

and are approximated by a constant eld.

Condition (77) is simply imposed by making (w

|
s
)
Gi
=
(|
s
)
Gi
= 0 in Eqs. 44 and 54.
F=1
y
z
18
S
y
m
m
e
t
r
y
R = 10.0
t = 0.25
E = 6.825 10
7
= 0.3
F
ree
F=1
A
B
S
y
m
m
e
t
r
y
x
Same mesh
configuration as for
the open semi-sphere
in figure 7
Fig. 9 Open thick dome under point loads. Geometry and boundary
conditions
6.2 Simply supported edge (u = 0,
s
= 0)
The condition u = 0 at the nodes laying on a simply sup-
ported (SS) edge is prescribed when solving the global sys-
tem of equations as for the clamped case. Prescribing u = 0
at an edge node also automatically implies that
w

s
is zero
along the SS edge direction s. This is equivalent to assuming
a soft simply support condition (i.e. w

= M
ns
= 0 at the
SS edge) [1, 35, 33].
The hard support condition requires prescribing the tan-
gential rotation
s
to a zero value. This means

s
=
w

s
+
s
= 0
s
=
w

s
(78)
Prescribing w at the SS nodes gives automatically
_
w

s
_
i
=
0 and hence the shear angle (

s
)
i
must be also prescribed to
a zero value. This condition can be imposed by transform-
ing the Cartesian shear strains to the boundary axes s, n and
making (

s
)
i
= 0 at each SS boundary node.
Note that prescribing just (w

)
i
= 0 at the SS nodes and
letting (

s
)
i
free reproduces the soft support condition for
Reissner-Mindlin theory [1, 33].
6.3 Symmetry condition (
n
= 0)
The condition of zero normal rotation (
n
= 0) is imposed
by neglecting the contributions from the normal rotation
w

n
at the symmetry edge when computing Eq. 41. The condition

n
=
w

n
+
n
= 0 for the thick case is treated as explained
for the clamped edge.
7 Examples
We present several examples showing the good performance
of the new EBST+ and EBST+1 rotation-free elements.
1 3
Comput Mech
Fig. 10 Open thick dome
under point loads. Normalized
horizontal displacement of point
B
THICK HEMISPHERICAL DOME
NORMALIZED HORIZONTAL EBST+ EBST+1 QLQL
DISPLACEMENT AT POINT B
N x M NODES ELEMENTS N x M NODES ELEMENTS
4 x 5 30 40 0,794 0,795 4 x 4 25 16 0,667
8 x 10 99 160 0,910 0,911 8 x 8 81 84 0,484
16 x 20 357 640 0,953 0,955 16 x 16 289 256 0,968
32 x 40 1353 2560 0,977 0,981 32 x 32 1089 1024 0,999
40 x 50 2091 4000 0,982 0,992
3D FEM Solution
0,40
0,50
0,60
0,70
0,80
0,90
1,00
0 500 1000 1500 2000 2500
EBST+
EBST+1
QLQL
Normalized horizontal
displacement at point B

B
nodes
u
B
/ u
B
F
u
B
= 4,0028E-4
F
u
B
/ u
B
F
u
B
/ u
B
F
Fig. 11 Pinched cylinder. Geometry, boundary conditions and meshes
The analysis has been performed in all cases using the
iterative solution scheme described in Sect. 5. Convergence
of the deection eld measured in L2 norm with an error
norm of 10
3
was obtained in 23 iterations in all examples
studied.
7.1 ScordelisLo roof
This is a classical example widely used to test the perfor-
mance of shell elements. In this example an effective mem-
brane interpolation is of primary importance. Hence this is a
good test to asses the new elements. The geometry is a cylin-
drical roof supported by a rigid diaphragm at both ends and
it is loaded by a uniform dead weight. Only a quarter of the
structure is meshed due to symmetry conditions [36].
Figure 5 shows the geometry and the meshes used. In
Table 1 we present the normalized vertical displacement of
point B. In Fig. 6 the normalized displacement is plotted and
compared with results obtained with the EBST element as
well as with the TLQL, QLQL and QLLL shell elements
based on Reissner-Mindlin theory [33]. The distribution of
M
x
, Q
x
and N
y
along side AB for the 40 40 mesh is
also shown in Fig. 6. Note the excellent behaviour of the two
new rotation-free shell elements.
7.2 Open semi-sphere dome under point loads
Finite elements with initially curved geometry are prone to
suffer the so called membrane locking [1, 33]. The new
EBST+ and EBST+1 elements have a quadratic interpola-
tion of the geometry, then it may suffer from this problem
[26]. In Fig. 7 the geometry, the meshes used and the mate-
rial properties and boundary conditions are shown. Only one
quarter of the mesh is used due the symmetry. In Table 2 the
normalized horizontal displacement of point B is presented.
In Fig. 8 the normalized displacement and the distribution of
M

along the line AB and M

, and Q

along the line AC

are plotted and compared with the same elements, as in the
previous example.
From the results shown it can be seen that both elements
present a slight membrane looking for coarse meshes. How-
ever membrane locking vanishes for ne meshes.
1 3
Comput Mech
7.3 Thick open semi-sphere dome under point loads
This example tests the capability of the new formulation to
model thick shell problems. For this purpose we analyze a
thick open semi-sphere dome under point loads. The geom-
etry and load conditions are exactly the same as for Example
7.2 except for the thickness of the shell that has a value of
0.25 (Fig. 9).
The results obtainedare comparedwitha 3DFEMsolution
obtained with a very ne mesh of 1512024 8-noded hexahe-
dral elements. Figure 10 shows that the normalized displace-
ments have an excellent convergence ratio for the EBST+
and EBST+1 elements.
The Reissner-Mindlin solution obtained with the QLQL
element is also presented, for comparative purposes.
7.4 Cylindrical shell under central point loads
Figure 11 shows the geometry of the cylindrical shell, the
material properties and the loading. Figure 12 shows the con-
vergence of the central deection and the distribution for M

and Q

along side BA and the distribution of N

y
along
side DC for the 50 50 mesh. In Table 3 the normalized
central displacement is plotted.
Once again the performance of the new EBST+ and
EBST+1 elements is excellent, with a slightly advantage in
favour of the later.
8 Concluding remarks
We have presentedtwoprocedures for extendingthe enhanced
rotation-free shell triangle (EBST), so as to account with
shear deformation effects. Two newtriangular shell elements
(the EBST+ and the EBST+1) have been presented using a
linear and constant approximations of the shear angle within
the 3-noded triangle.
The two shell triangles have very similar features, their
implementation share many common steps and both have a
excellent behaviour for thin and thick shell analysis.
A detailed study of the numerical results of the problems
solved in this paper indicates a slightly advantage in favour
of the EBST+1 element.
The iterative solution scheme presented allows one to
introduce the effect of shear deformation in a progressive
(iterative) manner starting fromthe initial thin shell solution.
The formulation is therefore useful for shell analysis with
a variety of thickness and materials. A particular applica-
tion of the new triangular shell elements is the analysis of
composite laminated shells for which shear deformation
effects are relevant.
-0,006
-0,005
-0,004
-0,003
-0,002
-0,001
0,000
0,001
0,002
0 20 40 60 80
TLQL QLQL
QLLL EXACT
BST EBST+
EBST+1
-0,009
-0,008
-0,007
-0,006
-0,005
-0,004
-0,003
-0,002
-0,001
0,000
0,001
0 10 20 30 40 50 60 70 80 90
TLQL QLQL
QLLL EXACT
EBST+ EBST+1
-0,04
-0,02
0,00
0,02
0,04
0,06
0 10 20 30 40 50 60 70 80 90
TLQL QLQL
QLLL EXACT
EBST EBST+
EBST+1
0,00
0,20
0,40
0,60
0,80
1,00
1,20
1,40
0 500 1000 1500 2000 2500 3000
EBST TLQL
QLQL QLLL
EBST+ EBST+1
Normalized central
deflection at point B
M 50 x 50 mesh
B
D
Q 50 x 50 mesh
B
D

A
C
N y 50 x 50 mesh

B
M
Q
N y
Fig. 12 Pinched cylinder deformation bending moments (M

) shear
force (Q

) and axial force (N

y
)
The formulation is also suitable for implementing an adap-
tive solution scheme where the shear angles are introduced in
structures (or zones of a structure) where the effect of shear
deformation is of importance.
1 3
Comput Mech
Table 3 Pinched cylinder (normalized vertical displacement at point B)
Point load cilinder normalized ver-
tical
EBST EBST+ EBST+1 TLQL QLQL QLLL
Displacement at point B U
B
/U
F
B
U
B
/U
F
B
U
B
/U
F
B
U
B
/U
F
B
U
B
/U
F
B
U
B
/U
F
B
N Nodes Elements N Nodes Elements
5 5 36 50 0.743 0.745 0.745 4 4 25 32 0.088
10 10 121 200 1.229 0.231 0.232 4 4 25 16 0.099 0.044
20 20 441 800 1.173 1.174 1.176 8 8 81 168 0.493
30 30 961 1800 1.106 1.105 1.110 8 8 81 84 0.614 0.373
40 40 1681 3200 1.071 1.072 1.076 16 16 289 512 0.877
50 50 2601 5000 1.052 1.053 1.058 16 16 289 256 0.943 0.756
32 32 1089 2048 0.955
32 32 1089 1024 1.014 0.921
U
F
B
[36] = 0.18248
Acknowledgment This research was partially supported by the pro-
ject SAFECON of the European Research Council.
References
1. Zienkiewicz OC, Taylor RL (2005) The nite element method for
solid and structural mechanics, 6th edn. Elsevier, Amsterdam
2. Stolarsky H, Belytschko T, Lee SH (April 1995) A review of
shell nite elements and corotational theories. In: Computational
mechanics advances, vol 2. North-Holland, Amsterdam
3. Idelsohn S, Oate E, Devorkin EN (eds) Proceedings of IACM IV
world congress on computational mechanics. CIMNE, Barcelona
(1998)
4. Uguraz AC (1981) Stresses in plates and shells. McGrawHill, New
York
5. Nay RA, Utku S (1972) An alternative to the nite element method.
In: Brebbia et al (eds), Variational methods in engineering, vol
1. International Conference Proceedings (Book), University of
Southampton, USA
6. Barnes MR (1977) Form nding and analysis of tension space
structure by dynamic relaxation. Ph.D. Thesis, Dept. of Civil Engi-
neering, The City University, London
7. Phaal R, Calladine CR (1992) A simple class of nite elements for
plate and shell problems. I: elements for beams and thin plates. Int
J Numer Methods Eng 35:955977
8. Phaal R, Calladine CR (1992) A simple class of nite elements
for plate and shell problems. II: An element for thin shells with
only translational degrees of freedom. Int J Numer Methods Eng
35:979996
9. Rio G, Tathi B, Laurent H (1994) A new efcient nite element
model of shell with only three degrees of freedom per node, Appli-
cations to industrial deep drawing test. In: MJM Barata Marques
(ed) Recent developments in sheet metal forming technology. 18th
IDDRG Biennial Congress, Lisbon
10. Abel JF, Cooke JR(eds) (2008) Proceedings of the 6th International
Conference on Computation of Shells and Spatial Structures IASS-
IACM Ithaca, NY, USA
11. Hrabok MM, Hrudey TM (1984) A review and catalogue of plate
bending nite elements. Comput Struct 19:479495
12. Hampshire JK, Topping BHV, Chan HC (1992) Three node trian-
gular elements with one degree of freedom per node. Eng Comput
9:4962
13. Oate E, Cervera M (1993) Derivation of thin plate bending ele-
ments with one degree of freedom per node. Eng Comput 10:543
561
14. Oate E, Cervera M, Zienkiewicz OC (1994) A nite volume for-
mat for structural mechanics. Int J Numer Methods Eng 37:181
201
15. Idelsohn S, Oate E (1994) Finite volumes and nite elements: two
good friends. Int J Numer Methods Eng 37:33233341
16. Oate E, Zrate F (June 26, 1997) New thin plate and shell tri-
angles with translational degree of freedom only. Presented at IU-
TAM/IACM Symposium, Viena Austria
17. Zrate F (1996) New nite element for plate and shell analysis. (in
Spanish), Ph.D. Thesis, Univ. Politcnica de Catalunya, Barcelona
18. Oate E, Zrate F (2000) Rotation-free plate and shell triangles.
Int J Numer Methods Eng 47:557603
19. Sabourini F, Bounet M (1995) Analysis of plates and shells with
a simplied three node triangular element. Thin Walled Struct
21:209223
20. Cirak F, Ortiz M (2000) Subdivision surfaces: a new paradigm
for thin-shell nite element analysis. Int J Numer Methods Eng
47:20392072
21. Cirak F, Ortiz M (2001) Fully C1-conforming subdivision ele-
ments for nite deformations thin-shell analysis. Int J Numer Meth-
ods Eng 51:813833
22. Flores FG, Oate E (2001) A basic thin shell triangle with only
translational DOFs for large strain plasticity. Int J Numer Methods
Eng 51:5783
23. Guo YQ, Gatis W, Naceur H, Batoz JL (2002) An efcient dkt
rotation free shell element for spring simulations in sheet metal
forming. Comput Struct 80:22992312
24. Engel G, Garikipati K, Hughes TJR, Larson MG, Mazzei L, Tay-
lor RL (2002) Continuous/discontinuousnite element approxima-
tion of fourth-order elliptic problems in structural and continuum
mechanics with applications to thin beams and plates, and strain
gradient elasticity. Comput Methods Appl Mech Eng 191:3669
3750
25. Oate E, Flores FG (2005) Advances in the formulation of the
rotation-free shell triangle. Comput Methods Appl Mech Eng
194:24062443
26. Flores FG, Oate E (2005) Improvements in the membrane behav-
iour of the three node rotation-free BST shell triangle using
an assumed strain approach. Comput Methods Appl Mech Eng
194:907932
1 3
Comput Mech
27. Brunet M, Sabourin F (2006) Analysis of a rotation-free 4-node
shell element. Int J Numer Methods Eng 66:14831510
28. Sabourini F, Brunet M (2006) Detailed formulation of the rotation-
free triangular element \S3 for general purpose shell analysis. Eng
Comput 23:469502
29. Grdsback M, Tibert G (2007) A comparison of rotation-free tri-
angular shell elements for unstructured meshes. Comput Methods
Appl Mech Eng 196:50015015
30. Oate E, Flores FG, NeamtuL (2007) Enhancedrotation-free basic
shell triangle. Applications to sheet metal forming. In: Oate E,
Owen R(eds) Computational plasticity. Springer, The Netherlands
pp 239265
31. Timoshenko SP (1979) Theory of plates and shells. McGraw Hill,
New York
32. Zienkiewicz OC, Taylor RL (2005) The nite element method for
solid and structural mechanics, 6th edn. Elsevier
33. Oate E (2012) Structural analysis with the nite element method.
Linear statics, vol 2. Beams, plates and shells. CIMNE, Barcelona
34. Oate E (2011) Structural analysis with the nite element method.
Linear statics, vol 1. Basis and solids. CIMNE, Barcelona
35. Oate E, Zrate F (2010) Extended rotational-free plate and beam
elements with shear deformation effects. Int J Numer Methods Eng
83(2):196227
36. Flgge N (1962) Stresses in shells. Springer, Berlin
37. Simo JC, Fox DD, Rifai MS (1989) On a stress resultant geomet-
rically exact shell model. Part II: the linear theory; computational
aspects. Comput Methods Appl Mech Eng 73:5392
1 3