the displacements prescribed for `in-plane' forces do not aect the bending de-
at shell element as combination of a plane stress element and a plate bending
and bending strain are not coupled in the energy expression within the elements.
13 in 27 .
results can be obtained with the at shell element Chapter 3 in 26 and Chapter
elements. As the size of the at elements decreases, convergence of the element
at least in part, and these can be simply reproduced. Shell structures with
displacement each node is modeled with three displacement and three rotation
at shell nite element that incorporates membrane and bending components of
This thesis describes the formulation and testing of a four node quadrilateral
Introduction
CHAPTER
, parallel to the plane of the plate at each node. And then, the membrane
and
with plate bending element 26, 27 , we know that for plane stress actions, the
In classical formulations of at shell element that combine plane stress element
assumptions of at shell nite elements. Numerical experiments with the shell
with drilling degree of freedom, and a bending component based upon Kirchho
Section 1.2.
component including the vertex rotation perpendicular to the plane of the plate
Sections 1.2 and 1.3 describe the classical formulation of at shell elements
for at shell nite elements that combine plane stress element with plate bending
ui , (Ui)
v i , (Vi)
yi , (My i)
xi ,(Mx i)
wi , (Wi)
f f p g = K p
f q p g
1.2.1
stiness matrix K p
, nodal forces ff pg, and element displacement fqpg, where
x , y
displacements perpendicular to the plane of the shell element, and the variables
in-plane displacements along the x and y axes respectively, the variable w for
Figure 1.1: A at shell element subject to plane membrane and bending action.
z i ,(M z i )
each typical node i. These modeling assumptions are shown in Figure 1.1. A
9
=
and
8
Ui
p
ffi g =
Vi
:
9
=
, for i = 1 2 3 4.
9
=
and
8
Wi
ffib g = Mxi
: Myi
9
=
, for i = 1 2 3 4.
1.2.2
fqi g
to the surface of the at shell, given by z , is not included in the denition
element. Notice that in the classical formulation, the rotation of the normal
8
wi
where fqibg = xi
: yi
ff b g = K b
fq b g
given uniquely by the nodal displacements in the z direction, w, and the two
Here we use the superscript `p' to denote in-plane deformation of the shell
8
ui
p
with fqi g =
vi
:
1.2.3
shows what will happen in the mathematical model. The column will displace
freedom perpendicular to the plane of the oor. As such, the column torsional
columns rotate about their axis by the same amount. Use of the abovementioned
the oor slab rotates due to external loadings, compatibility requires that the
1.2.6
area where oor slabs in a building are supported by columns. In the real build-
stiness cannot be connected to the shell element stiness 10, 11 . Figure 1.3
0 0 0
0 0
0
0
Krsb
0 0 0
0
0 0 0
0 0
0 0
0 0
Krsp
membrane elements are excessively sti. Figure 1.2 shows a second application
lar shear-loaded cantilever beam. The numerical experiments reveal that these
Felippa 6 reports that Turner et al. 25 and Taig 21 have used this
2
66
66
66
66
66
6
Krs
= 666
66
66
66
66
64
3
77
77
77
77
77
77
77
77
77
77
77
75
ff g = K fq g
1.2.5
where
ffi g
Elevation
zi
direction equations of
0
Kz
zi
does not a
ect
directions di
er, and a transformation is accomplished, then the global sti
ness
matrix is singular. Detection of this singularity is dicult. There are two simple
7
and
normal to the element plane are added. Progress in this direction was rst made
modeling information. If, on the other hand, the local and global coordinate
!,
The diculties described in Section 1.1 vanish when nodal rotational parameters
6.
to the plane of the element. These are the so-called drilling degree of freedom
the stresses, and indeed, is uncoupled from all others equilibrium equations, any
sti
ness matrix is achieved after a local-to-global coordinate transformation, and
equation Kz
coplanar only.
coordinate directions of these elements happens to coincide with the global ones,
of straight boundaries of cylindrical shaped shells 26, 27, 10. When the local
elements occur when elements meeting at a node are coplanar or nearly coplanar.
the at shell elements do not include rotational parameter with this class of
Figure 1.3: Finite element model of simple table using shell element having only
ve degree of freedom per node.
Plan
Incompatible
Rotation
Displaced Structure
the formulation of shell nite elements having six degrees of freedom per node.
in the translational degrees of freedom, but may not rotate by the same amount
un 1
u t2
u n2
-21 (1 - 2)
u t1
- -21 ( 1 - 2 )
ut
t
= 1 , sl u 1 + sl u 2
n12
1.3.2
1.3.1
n
= u 2 , u 1 =l. Dierentiating
= 1 , s u 1 + s u 2 + 4 s 1 , s u
n12
where u
un
l
un12 =--(
1- 2)
n12
n12
= , u 1 + u 2 + 4 1 , 2s u
l
l
l
l
4
2
s
= + 1 , u
1.3.4
1.3.3
Also, we get
therefore
@s
@un
= , l !1 , !2
8
n12
10
j0 , = , 12 !1 , !2
n12
n12
, !2 = , 8l u
u
!1
@un
n12
j0 = + 4l u
4
j
=
, u
@s
l
@s
@un
1.3.8
1.3.7
1.3.6
1.3.5
where !1 and !2 are the vertex rotations at nodes 1 and 2, respectively. Since
@u
,!2 + !1 = @u
j
,
j0
@s
@s
@s
@un
= 12 !1 , !2
1.3.9
and v, in terms of
1.3.10
1.3.12
1.3.11
11
direction to the element side and the x-axis. We can use a similar technique to
9
=
3 8 9
77 u =
75
: v
8 9 2
3 8
u = 66 C ,S 77 u
75
: = 64
v
S C : u
9 2
= 6 C
6
= 64
,S
and
8
u
:u
the systems x-y and n-t. A schematic of the required coordinate transformation
ut
quantities, u and v , and one vertex rotation quantity, ! are obtained through
the nodal parameters along the edge of the element i.e. two nodal translation
= 1 , sl u 1 + sl u 2 , 2s 1 , sl !1 , !2
un
where !1 and !2 are the so-called vertex rotation parameters at nodes 1 and 2,
@un
jl ,
@s
t
un
u
= vu v !
u = uui vi !i
1.3.13
With the coordinate transformation in plane, we can now write the quadratic
u=
i=1
8
X
12
Ni ui
vi
n-t.
ut
i=1
X
v = Ni vi
1.3.15
i = 6 8
Ni = 12 1 , 21 + i
i = 1 2 3 4
m n = 8 5 5 6 6 7 7 8
Ni = 41 1 + i 1 + i , 12 Nm , 12 Nn
i = 5 7
Ni = 12 1 , 21 + i
1.3.16
13
into the quadratic interpolation elds of the entire element with mid-side dis-
ing the expressions for u and v interpolations in the boundaries, i.e. equations
displacement interpolation elds u and v within the entire element may be de-
functions are
Ni are shape functions of the 8-node Serendipity element 26. The shape
In equations 1.3.14 and 1.3.15, and are parametric coordinates 27, and
and
10, 16.
14
For this reason, it is concluded that a better way is to use the true rotations,
is not equal to the true rotations at nodes, even though it can be related to it.
element based on the ! connector will have a defect. The new nodal connector
16
15
gree of freedom. The rst approach was rst reported by Sabir in 1985 20 . He
tially suggested by Hughes and Brezzi 11, 12 . The latter approach employs
This chapter discusses two approaches for producing the so-called drilling de-
of the vertex bisecting the angle between adjacent edges of the nite element.
In fact, numerical experiments indicate that only rectangular elements are well
unfortunately suer from geometrical restrictions and zero energy modes 10
.
behaved, and provide accurate results 10
. The second approach, as presented
2.0.1
, Undeformed Plate
v/ x
- u / y
= 1 @v , @u :
2 @x @y
Unlike the denition of the vertex rotation ! given in equation 1.3.4, the
to a class of nite elements that performs better than those element mentioned
In this chapter we derive a at shell nite element model that contains nodal
CHAPTER
2.1.1
2.1.2
a2
17
left with nine constants to represent strain deformation in the element. These
used three of these in the representation of the rigid body movements, we are
then the shape functions should contain twelve independent constants. Having
node quadrilateral nite elements. If each node has three degrees of freedom
Now let's consider the shape functions needed model displacements in a four
a2
Notice that equations 2.1.2 are described in terms of three components a1 and
8
u = a1 , a3y
: v = a2 + a3 x
If "x = "y = "xy = 0, then the equations above can be integrated to obtain
8
"x = @u=@x
"y = @v=@y
: "xy = @u=@y + @v=@x
Based on the strain analysis, Sabir derived 20 a rectangular membrane element
2.1.3
"y
@x2
+@
"xy
@x@y
=@
2.1.4
8
"x = a4 + a5 y + a11 y 2 + 2a12 xy 3
"y = a6 + a7 x + ,a11 x2 , 2a12 x3 y
: "xy = a8 + a9 x + a10y + a5x + a7 y:
2.1.5
a7
and
a9
18
strain behavior. The higher order bracketed terms are added in such a way that
constant strain states. These state ensure convergence as the nite element grid
as
attempt to mitigate the problem, Sabir suggested interpolations for the strain
@ 2 "x
@y 2
We observe that, if the terms of equation 2.1.3 are dierentiated, they satisfy
"y = a7 + a8 x + a9 y
: "xy = a10 + a11 x + a12 y
8
"x = a4 + a5 x + a6 y
and "xy . As a rst-cut, we could assume that the three strain components
satisfy:
"y
nine constants are to be distributed among the three components of strain "x,
2.1.7
2.1.8
19
vxi yi and i = xi yi with i = 1 2 3 4, into equation 2.1.6 and 2.1.8
2
6 1 0 ,y x xy 0 0 y=2 0 y2=2 xy2 x2 y3
x = 664
0 1 x 0 0 y xy x=2 x2 =2 0 ,x2 y ,x3 y2
3
77
75
where displacements
fug = xfAg
8
u = a1 , a3y + a4x + a8y=2 + asxy + a10 y2=2 + a11 xy2 + a12 x2 y3
2.1.6
: v = a2 + a3 y + a6 y + a8 x=2 + a7 xy + a9 x2 =2 , a11 x2 y , a12 x3 y2
for u and v are obtained by integrating equations 2.1.5, and then adding the
1
= x
3
12
fAg
12
12
12
1
0 0 1
0 1 xi
yi=2
0
yi =2
xi yi
2
,xi yi ,xi yi
xi yi
yi xiyi xi =2 xi =2
0 ,xi =2 0 yi=2
xi yi
1 0 ,yi xi
where
20
N = xx,
2.1.11
2.1.10
Unlike the naive approximation for displacements, the resulting matrix x is not
2
66
66
xi = 66
66
4
2.1.9
vector, and
fq g
gives,
3
77
77
77
77
5
21
freedom. The variational formulation is due to Hughes and Brezzi 11, 12. It
formulation that includes an independent rotation eld for the drilling degree of
Flat shell nite elements may be formulated through the use of a variational
problem are modeled with non-rectangular nite element meshes 20, 10.
accuracy of these nite elements is not satisfactory signicantly when the same
`close' to being rectangular. Unfortunately, Frey also reports 10 that numerical
10 report good numerical performance with quadrilateral elements that are
on the assumption of the strain states is non-conforming, Sabir 20 and Frey
Even though this membrane component with drilling degree of freedom based
element formulations, if a inverse of parameter matrix x exists when the form of
is the matrix of shape functions. Observe that all formulations of nite elements
2.2.3
2.2.4
= skew r u
symm = C symm u
r
2.2.2
skew = 0
2.2.1
div + f = 0
2.2.7
2.2.6
2.2.5
22
Z
u = 1 symmr u C symmr ud
2 Z
Z
+ 12 jskewr u , j2d , u fd
described as
2.2.8
The variational formulation suggested by Hughes and Brezzi 11, 14, can be
= symm + skew
symm = 12 + T
skew = 21 T
Consider the elastic boundary value problem for a body described by region .
symm
u C symm
ud
u fd
+
skew
r u , T
skew
r u , d
0 = D
u
u
2.2.9
i=1
4
X
Ni
i
23
Ni
= 14
1 + i
1 + i
=
i = 1 2 3 4
2.2.10
where n34 is a outward normal direction to the element side 3 , 4. The inde-
independent rotation eld over the entire element 14. Consider a 4-node
the Allman-type interpolation for displacement eld and the standard bilinear
The shell nite element with drilling degree of freedom is derived by combining
formulation is
34
3
where,
ij
24
2.2.14
lij =
x2ij + yij2 21
Cij = cos ij = yl ij
ij
Sij = sin ij = xl ij
2.2.16
2.2.15
2.2.13
,
yij = yj yi
9
=
2.2.11
2.2.12
,
9
8
8
= X
Cij
lij
+
N
k
j
i
8
k=5
: Sij
xij = xj xi
4
X
8
ui
=
Ni
i=1
vi
:
8 9
u
=
u =
:v
n34
k = 6 8
Nk = 12 1 + k 1 , 2
i=1
4
X
2.2.18
2.2.17
8
9
@Ni =@x 0
=
Bi = 0
@Ni =@y
: @N =@y @N =@x
i
i
25
2.2.19
9
=
is the nodal translation parameters, and i is the nodal
8
ui
where ui =
: vi
8
9
@u=@x
=
symmr u
=
@v=@y
: @u=@y + @v=@x
k = 5 7
i j = 1 2 2 3 3 4 4 1
Nk = 12 1 , 2 1 + k
k = 5 6 7 8
1 l S @Nl , l S @Nm , N
+ 16
ij ij
i
@x ik ik @x
1 l C @Nl , l C @Nm
gi = , 16
ij ij
@y ik ik @y
biui + gii
1
@N
1
@N
i
i
bi = , 2 @y 2 @x
i=1
4
X
26
where,
skewr u , =
2.2.24
2.2.23
2.2.22
2.2.21
2.2.20
9
8
lij Cij @Nl =@x , lik Cik @Nm =@x
Gi = 1
lij Sij @Nl =@y , lik Sik @Nm =@y
8
: lij Cij @Nl =@y , lik Cik @Nm =@y
+ lij Sij @Nl =@x , lik Sik @Nm =@x
G
B
b
b
27
the formulation is insensitive to the value of used at least for several orders
modulus value
12. Numerical studies performed by Taylor have shown that
` '. For isotropic elasticity, it is suggested that may be taken as the shear
g g
, with
m =
K +
P
28
where
and
are given in equations 2.2.25 and 2.2.26.
m
q =
f
rium and the stiness matrix in equation 2.2.25 is the regular element stiness
symmetric stresses will not be identically zero in general, and thus will play a
2.2.26
where
is general external forces,
is nodal parameter vector, dened as
2.2.24, and
matrix without the modication term. In the discrete case, however, skew-
b
b d
T
is the penalty,
= f
1
2
3
4 g, and
i = f
i
zero. It follows that the rst term in the equations 2.2.9 expresses equilib-
tions 2.2.9 is set to zero | this asserts that the skew-symmetric stresses are
problem dependent
11. For instance, suppose that the second term in equa-
i and
i = f
i
ig
B
2.2.25
where,
B
B d
From equations 2.2.9 and 2.2.21, the second term in equation 2.2.9
= f
1
2
3
4 g
i and
i are as dened in equations 2.2.19 and 2.2.20.
where
is the constitutive matrix and
=
T
matrix,
The rst term in the variational equations 2.2.9 produces the element stiness
Ube = 12
Ae
T Db dAe
Xn Ube
3.0.1
29
Here, Ube is the element strain energy due to bending. Ae is the element area.
with
U=
on the discretization of the strain energy. The model neglects the transverse
using the discrete Kirchho technique. The DKQ element formulation is based
The plate bending component of the shell element corresponds to the 12 DOF
CHAPTER
1 ,
0 0
1
2
1
1
3
7
7
7
7
7
7
7
7
5
3.0.2
,2@ 2 w=@x@y
,@ 2 w=@y2
,@ w=@x
9
=
3.0.3
3.1.2
y = , @w
@x = ,x
30
3.1.1
x = @w
@y = y
scribes the rotations of the normal to the undeformed middle surface in the x -
an interpolation for the independent nodal rotation elds x and y that de-
=
:
where E , and h are the Young's modulus, Poisson's ratio and thickness, re-
Db =
2
6
6
3
Eh 666
121 , 2 666
4
for
node3
(x3 , y 3)
j,
@w=@s
w
x
y
i j
23
@n
j+
@n
@w
j
j = 12
!
ni
+
nj
j,
@w=@n
@w=@n
a1
a2 s
a3 s
31
a4 s
n23
3.1.4
3.1.3
at mid-side of
w
node2 x
(x2 ,y 2 )
y
= 1 2
2 3
3 4
4 1
j = 12
@w=@s
@n
@w
@w
when
node1
(x1 , y 1) x
y
node4 w
(x 4, y 4) x
y
with a cubic displacement interpolation along the element sides, shown in Figure
sides are eliminated by averaging the corresponding corner nodes values. Rota-
At the middle point of element side, the rotation components along the element
@s
@w
=
a2
+2
a3 s
+3
a4 s
Generally, we assign
lij
lij
@s
@w
lij
1
4 = 2
a
= 0
+
+
a2 sj
@s
@w
@s
@w
32
xi
yj
yi
a4 sj
a4 sj
+
+3
2
a3 sj
a3 sj
j,
j+
@s
@w
xj
3
a4 si
a4 si
+3
a3 si
wj
lij
wi
lij
j +3 ,
wj
j ,2 ,
@s
@w
wi
= , 2 + , 2 1 2
a3 si
a2 si
j = 2+2
a1
1 ,2
3 =
a2
sj
si
@s
@w
wj
@s
a1
j = 2+2
@w
wi
So,
3.1.5
@w j + @w j , 2 wj , wi :
@s i @s j
lij
3.1.6
9
=
3.1.8
3.1.7
pression
y =
x =
i
N i y
N i x
33
i=1
8
X
i=1
8
X
3.1.10
3.1.9
8
,@x =@x
=
,@y =@y
: ,@x =@y , @y =@x
= , 41 si + sj + 32 wj l, wi
ij
@w j = , 1 @w j + @w j + 3 wj , wi
@s k
4 @s i @s j 2
lij
At the midpoint of the sides, sk = lij =2, the @w=@s expression can be obtained
ij
+ 3ls2
@w = @w j + 2s ,2 @w j , @w j + 3 wj , wi
@s
@s i lij
@s i @s j
lij
i = 6 8
i = 5 7
m n = 8 5 5 6 6 7 7 8
Shape Function
34
@w j =
@s k s
i = 1 2 3 4
x = @w
@y ji = y
y = , @w
@x ji = ,x
3.2
of the quadrilateral element with straight sides again, see Figure 3.1.
3.2.3
3.2.2
3.2.1
xi and yi are transitory nodal variables aected at the corner and middle-nodes
i = 1 2 3 4
Ni = 41 1 + i 1 + i , 12 Nm , 12 Nn
Ni = 12 1 , 21 + i
Ni = 12 1 , 21 + i
The shape function Ni , where and are parametric coordinates 27,
3.2.4
= x nij
lij
ij
Ck
yij
35
= xj , x i
xij
Sk
= yj , yi
other words, the DKQ technique is appropriate for thin plates only.
3 Convergence towards the thin plates theory will be obtained for any element
2 The Kirchho assumptions are satised along the entire boundary of the
3.1.7.
1 w varies independently along the element sides. At the four corner nodes, the
We note that
= nk
k = 5 6 7 8
@w
jk
@n
38
77
x
75
: y
9 2
38
n
= 66 C ,S 77
=
6
7
4 S C 5
: s
9 2
= 66 C
=6
4 ,S
and
yi
9
=
9
=
3.2.6
3.2.5
ni
nj
xi
yi
= @w
j
@n
!
1
@w
@w
= 2 @n j + @n j
= 1 +
2
= 1 C + S + C
2
xj
+S
yj
3.2.7
write
sk
36
ij
= @w j
@s
!
1
@w
@w
= ,
j
+ j + 3w ,w
4 @s
@s
2 l
Similarly, from equations 3.2.3 and 3.1.7, and transformation 3.2.5, we can
nk
yk
xi
8
x
: y
8
n
: s
and
i j = 1 2 2 3 3 4 4 1
3.2.8
xk
yk
ij
ij
37
ij
a k = Ck S k = ,
3
3 xij yij
4
4 lij2
1
1
1
1
bk = Ck2 , Sk2 = yij2 , x2ij =lij2
2
4
2
4
Sk
xij 2
ck =
=,l
l
= S k n + Ck s
= 3 Ck Sk x + 1 Sk2 , 1 Ck2 y + 3 Ck Sk
4
2
4
4
1
1
3
wj , wi
2
2
+ 2 Sk , 4 Ck y + 2 Ck l
yj
= Ck n , Sk s
= 1 Ck2 , 1 Sk2 x + 3 Ck Sk y + 1 Ck2 , 1 Sk2
2
4
4
2
4
3
3
wj , w i
+ Ck Sk y , Sk
4
2
lij
and
can write
xj
3.2.11
3.2.10
3.2.9
From above equations 3.2.7 and 3.2.8, and transformation 3.2.6, as well I
k = 5 6 7 8
= , 14 s + s + 32 wj l, wi
ij
1
= , 4 ,Sk x + Ck y , Sk x + Ck y + 32 wj l, wi
ij
ij
and
3.2.13
3.2.12
i=1
XN
i xi
38
3.2.14
= 21 Sk2 , 14 Ck2x , 34 Ck Sk y
+ 21 Sk2 , 14 Ck2x , 34 Ck Sk y + 32 Ck wj l, wi
ij
3
= dk x , ak y + dk x , ak y , 2 ek wj , wi
1
4
= 43 Ck Sk x , 12 Ck2 , 14 Sk2y
+ 43 Ck Sk x , 12 Ck2 , 14 Sk2y , 32 Sk wj l, wi
ij
3
= ak x , bk y + ak x , bk y , 2 ck wj , wi
yk
xk
ij
1
1
4
2
Ck
yij 2
ek = , = ,
l
l
1
2
y
Ni yi
i=1
i=1
i=1
Ni xi
Ni xi
Ni yi
k =5
k =5
X
X
+
= hH ifqg
X
X
X
Nk xk
Nk dk xi
Nk yk
yi
yi
xj
xj
yj
yj
3.2.16
3.2.15
, b + a , b , 32 c w , w
, a + d , a , 23 e w , w
Nk ak xi
k =5
k =5
X
X
+
Ni yi
and
39
3.2.18
3.2.17
where hH i and hH i are the shape functions. In component form,
and
i=1
i=1
X,
X
,
= hH ifqg
with
= 23 c6 N6 , c5N5
= a6 N6 + a5 N5
H4x
H5x
= a8 N8 + a7N7
= ,N4 , b8 N8 , b7 N7
x
H12
= 23 c8 N8 , c7 N7
x
H10
x
H11
= ,N3 , b7 N7 , b6 N6
= a7 N7 + a6 N6
= c7 N7 , c6N6
3
2
H9x
H8x
H7
x
= ,N2 , b6 N6 , b5 N5
= ,N1 , b5 N5 , b8 N8
H3x
H6x
= a5 N5 + a8 N8
= 32 c5 N5 , c8N8
H2x
H1x
40
H12
H11
H10
H9
H8
H7
H6
H5
H4
H3
H2
H1
= ,a8 N8 , a7 N7 = ,H11
= N4 + d8N8 + d7N7
= 32 e8N8 , e7N7
= ,a7 N7 , a6 N6 = ,H8
= N3 + d7N7 + d6N6
= e7 N7 , e6 N6
3
2
= ,a6 N6 , a5 N5 = ,H5
= N2 + d6N6 + d5N5
= 32 e6 N6 , e5 N5
= ,a5 N5 , a8 N8 = ,H2
= N1 + d5N5 + d8N8
= 32 e5 N5 , e8 N8
3.2.19
41
the displacements and rotations of the shell mid-surface are independent vari-
approach suggested by Kanok-Nukulchai 18. Unlike the at shell element pre-
pared to a four node at shell element in ANSYS-5.0, which has six degrees
In case studies a-d, performance of the at shell nite element is com-
analysis.
for analytical solutions to exist, and because they produce displacements in the
Applications a and c have been selected because they are simple enough
Numerical Examples
CHAPTER
12
16
8
6
22 4 4
12
12
10
20
6
10
12
42
Irregular mesh
82
Irregular mesh
41
Regular mesh
41
P = 40
l = 48
h = 12
E = 30000
= 0.25
A shear-loaded cantilever beam, as shown in Figure 4.1, has been used as a test
shell element. Application c is also computed using a quadrilateral at shell
for the transverse shear energy. The result is a so-called bilinear degenerated
P l3 4 + 5P l
3EI + 2Eh :
4.1.1
distorted meshes.
43
described herein gives reasonably accurate and rapidly convergent results with
racy than ANSYS-5.0. The numerical results also suggest that the shell element
irregularly shaped meshes, the present shell element provides much greater accu-
The numerical results from this test problem indicate that, with the same
bilinear element is a rectangular constant strain element without any nodal ro-
the Allman element is a rectangular element with the vertex rotation and the
The Sabir element is a rectangular element with the drilling degree of freedom,
mary of numerical results, with the asterisk * denoting the irregular mesh.
the literature for the performance of other elements. Table 4.1 contains a sum-
are compared against the theoretical solution, and numerical results reported in
tion 4.1.1 gives w = 0:3553. The numerical results for this shell
nite element
Substituting the material and section properties selected in Figure 4.1 into equa-
w=
From elasticity 13, 23, the analytical solution for the tip displacement is
tion, and also for a irregular meshes of four and sixteen quadrilateral elements.
44
wc = 0:00406
q0a4
D
forming loading, shown in Figure 4.2. The exact transverse displacement at the
Consider a square plate simply supported on all four edges subjected to a uni-
and b a concentrated point load at the center of the plate. For each load case,
4.2. Two load cases are considered a a uniform loading over the entire plate,
Consider the square plate simply supported on four edges, as shown in Figure
Table 4.1: Comparison in some results of the tip displacement, w, for the short
cantilever beam.
Meshes
4 1 8 2 16 4 4 1 8 2
Present element
0:3445 0:3504 0:3543 0:3066 0:3455
Error to theoretical solution 3:039 1:379 0:282 13:707 2:758
ANSYS5.0
0:2424 0:3162 0:3449 0:2126 0:2996
Sabir 20
0:3281 0:3454 0:3527 |
|
Allman 2
0:3026 0:3394 0:3512 |
|
Bilinear element
0:2424 0:3162 0:3447 |
|
a = 10"
S.S.
S.S.
n=2
D=
12 1 , 2
Eh3
4.2.2
4
1:06405
0:000489
1:0492
46
45
4.2.3
P a2
D
quarter of the plate is taken for numerical computation. Regular meshes on the
wc = 0:0115999
center, wc, from the plate theory 7, 24, can be expressed as
4.2.2
where P is the concentrated loading at the center, a is the length of edge of the
wc = 1:064045
105
10,1 in
:
D = 1:1446886
2
1:06027
0:355
1:0044
Table 4.2: The transverse displacements at the center of the square plate simply
supported on 4 edges under uniform load over the entire plate with dierent
meshes and the comparations with the exact theoretical solution.
Meshes N
Displacements wc 10,1
ANSYS-5.0.
are closer to the theoretical solution than predicted by the shell element from
in Table 4.2. Once again, the present element generates displacements that
Because the plate geometry is symmetric about x-axis and y-axis, only one
and
and 4.2.2
gives
and
where q0 is the uniforming loading, a is the length of edge of the square plate
a = 10"
= 0. 3
h = 0. 5"
q 0 = 300 psi
7
E = 10 psi
S.S.
S.S.
4
3:12850
2:908
3:0777
8
3:06664
0:873
3:0518
47
loads at the anges of the free end in opposite directions along y, as shown in
the top face, as shown in Figure 4.4. The third is under two level concentrated
4.3. Second, displacements are computed for a uniform load on the center line of
ments due to a concentrated load at the center of the free end, as shown in Figure
an I-shape cross section. Three load cases are considered. The rst is displace-
In the third example, displacements are computed for a cantilever beam having
Table 4.3: The transverse displacements at the center of the square plate simply
supported on 4 edges under concentrated point load at the center with dierent
meshes and the comparations with the exact theoretical solution.
2
3:32666
9:426
3:1574
z
y
N=2
h
x
48
3
PL
w = PL
+
3EI A G
h = 5"
P = 400 blf
E = 107 psi
= 0.3
t = 0.25"
L = 40"
b = 10"
The solution of the transverse displacement at the free end, w, from the beam
Meshes N
Displacements wc 10,1
Error to theoretical solution
ANSYS5.0 wc 10,1
loaded with point load P at the center of the free end shown in Figure 4.3.
Displacements are computed for a cantilever beam with I-shape cross section
plate. With regular meshes N = 2, 4 and 8 See Figure 4.2, the transverse
4.2.4
computed using commercial nite element tools, ANSYS-5.0 and SAP-90, and
Figure 4.5. The numerical solutions of these three cases for all meshes are also
Once again, numerical displacements are computed for only a quarter of the
wc = 3:0401019 10,1in
4.3.2
49
4.4. The solution of the transverse displacement at the free end, w, from the
distributed line load q0 along the center line of the top face, as shown in Figure
Also, I look at a cantilever beam with I-shape cross section under a uniformly
close agreement.
50
from beam bending theory and computed by elements in this thesis are in very
From the two cases of the I-beam above, we can see that the displacements
Table 4.5.
using the elements described in this thesis convergent rates are tabulated in
w = 2:22585 10,2
L4 + q0 L2
w = q80EI
2A G
of convergence for numerical results. These results are tabulated in the same
where similarly the second term represents shear eect. q0 is the unique load, and
Table 4.4: The transverse displacements at the free end of the I-shape section
cantilever beam under concentrated point load at the center of the free end with
dierent meshes and their convergent rates.
Meshes N
2
4
8
16
Displacement w
10,2 ,2:65646 ,2:80859 ,2:85107 ,2:85482
ANSYS5.0 w
10,2 ,1:9158 ,2:5457 ,2:7789 ,2:8424
SAP90 w
10,2 ,2:2862 ,2:6772 ,2:8150 ,2:8486
N = wN w, wN ,1
N
using the elements described in this thesis. Results are tabulated in Table 4.4.
web. Suppose that the shear modulus is as E=G = 2:5. Substitute values of I ,
of the beam. I = 33:8802 is modulus of the area and Aw = 1:1875 is area of the
where the second term represents shear eect. P is the load, and L is the length
z
q0
y
N=2
b
1
h
x
at
52
From the examples of the I-beam and the folded plate, it can be observed
= 1, 2 and 4 are used and the results, the transverse displaceat points 1 and 2, and their convergent rates are tabulated in Table
meshes with
As the third example, I consider a folded plate, as shown in Figure 4.6. The
Table 4.6: The transverse displacements at point 1 of the I-shape section cantilever beam under two lever concentrated loads at the anges of the free end in
opposite directions along y with dierent meshes and their convergent rates.
51
Meshes N
2
4
8
16
Displacement 10,1 2 45142 2 50528 2 51482 2 52135
ANSYS5.0 10,1 1 1280 1 8943 2 3378 2 5135
SAP90 10,1
2 0218 2 3010 2 4562 2 5195
ments
Table 4.5: The transverse displacements at the free end of the I-shape section
cantilever beam under uniform load along center line of the top face with dierent meshes and their convergent rates.
Meshes N
2
4
8
16
Displacement
10,2 ,2 19184 ,2 22248 ,2 23926 ,2 24607
ANSYS5.0
10,2 ,1 6003 ,2 0237 ,2 1862 ,2 2339
SAP90
10,2 ,2 0861 ,2 1831 ,2 2228 ,2 2413
by using the elements described in this thesis and their convergent rates are
point 1
see Figure 4.5 are computed by using the elements described in this
Displacements are computed for a cantilever beam having I-shape cross section,
Figure 4.4: Cantilever I-beam under a uniformly distributed line load along the
center line of the top face.
h = 5"
q0 = 20 bl/in
E = 10
= 0.3
t = 0.25"
L = 40"
b = 10"
7 psi
N=2
b
1
53
Table 4.7: The horizontal displacements at point 1 of the I-shape section cantilever beam under two lever concentrated loads at the anges of the free end in
opposite directions along y with dierent meshes and their convergent rates.
Meshes N
2
4
8
16
Displacement 10,1 1 38753 1 46572 1 48824 1 49583
ANSYS5.0 10,1 0 6299 1 1112 1 3743 1 4601
SAP90 10,1
1 0434 1 3116 1 4407 1 4833
Figure 4.5: Cantilever I-beam under two level concentrated loads at the anges
of the free end in opposite directions along y.
P = 1600 bl
h = 5"
E = 107 psi
= 0.3
t = 0.25"
L = 40"
b = 10"
z
1
N=2
q0
1
,1 38009
,1 3654
2
,1 41003
,1 4068
4
,1 42237
,1 4209
2
,1 35207
,1 3514
4
,1 36062
,1 3604
54
Table 4.9: The transverse displacements at point 2 of the folded plate simply
supported on two opposite sides under uniform load along the center line with
dierent meshes and their convergent rates.
Meshes N
Displacement 10,1
ANSYS5.0 10,1
Table 4.8: The transverse displacements at point 1 of the folded plate simply
supported on two opposite sides under uniform load along the center line with
dierent meshes and their convergent rates.
Meshes N
Displacement 10,1
ANSYS5.0 10,1
b = 50"
L = 50"
q0 = 4000 lb/in
t = 0.1"
h = 25"
E = 10 7psi
= 0.3
55
integrating the shape functions. One advantage of the second approach is mass
lumped mass matrix. A second method is to form the consistent mass matrix by
computed in at least two ways. The easiest approach is to simply form the
nite element. The mass matrix is needed for dynamic analyses it can be
Future work will include the formulation of a mass matrix for the at shell
other shell nite elements. The at shell elements shows excellent performance
Numerical experiments have been conducted to assess the accuracy and relia-
the membrane element in more than one way. In this project we have introduced
with the drilling degrees of freedom, and a discrete Kirchho plate nite element.
This thesis has presented the formulation of a four-node thin at shell nite
CHAPTER