You are on page 1of 12

PBE(1)

Plate Bending Elements



PBE.1 Introduction to Plate Bending
Consider a plate loaded in the transverse direction only,

In linear elastic analysis, the following assumptions are made regarding displacements:
z
x
y
w w(x, y) (independent of z, neglecting )
u (x, y) z
(linear in z with zero values at mid-thickness)
v (x, y) z
= |
|
|
=

|
`
|
=

)

(PBE.1)

Note that
x
is rotation about the y-axis, while
y
is rotation about the x-axis.
If the normal strain
z
is neglected, the strain state within an infinitesimal volume is given by:
PBE(2)
y
x x
x xy
y
y xz x
yz y
u u v
z , z
x x y x y x
v u w w
z ,
y y z x x
v w w
,
z y y
| | |
= = = + = +
|
|

\ |
|
|
= = = + = +
|

|
|

|
= + = +
|

(PBE.2)
Because the strains (
x
,
y
&
xy
) are linear functions of (z), generalized strains independent
of (z) can be used to represent any strain state:
x
x
y
y
y
x
xy
xz x
yz y
x
y
y x
w
x
w
y



+ = =
` `


+
)

)
(PBE.3)
The generalized stresses corresponding to are,
x
y
xy
x
y
m
m
m D
q
q







= =
`






)
(PBE.4)
where,
m
x
, m
y
are bending moments per unit length
m
xy
is twisting moment per unit length
q
x
, q
y
are shear forces per unit length.
PBE(3)

For isotropic elastic material, D can be expressed as:
( )
1,1 1,2
2,1 2,2
3,3
4,4
5,5
3
1,2 2,1 3,3
1,1 2,2
2
4,4 5,5
D D 0 0 0
D D 0 0 0
0 0 D 0 0 D
0 0 0 D 0
0 0 0 0 D
where,
D D 2D
Et
D D
1
12 1
Et 1
D D
1.2 2(1 )
|
(
|
(
|
(
|
(
=
|
(
|
(
|
(

|
|
|
|
|
= = = = =
|


|
|
|
= =
|
+

(PBE.5)
The factor (1.2) in the denominator of
4,4
D and
5,5
D is to account for the fact that
xz
and
yz

are not constant over the thickness, but are rather parabolic functions of (z) according to
mechanics of materials theory. The factor (1.2) is obtained by equating the strain energy
based on constant
xz
and
yz
to that obtained from mechanics of materials theory.
Since the generalized strains and stresses in (PBE.3) and (PBE.4) are independent of (z), the
solution of plate bending problems can be performed in the (x,y) plane.
PBE.2 Kirchhoff Elements (Thin Plates)
For thin plates, transverse shear strains
xz
and
yz
are assumed negligible. Therefore,
PBE(4)
xz x x
yz y y
w w
0
x x
w w
0
y y
|
= + = =
|

|
|

= + = = |
|


(PBE.6)
Substituting (PBE.6) in (PBE.3), the three remaining generalized strains can be expressed in
terms of the derivatives of (w) only:
2
x
2
2
y
2
2
xy
w
x
w
y
2 w
x y







= =
` `


)


)
(PBE.7)
Hence, only moment generalized stresses need to be considered:
x
y
xy
m
m D
m




= =
`



)
(PBE.8)
where D is the top-left 33 sub-matrix of the 55 matrix given in (PBE.5).
PBE.2.1 Continuity requirements
Since the strains in (PBE.7) are second derivatives of (w), C
1
continuity in (w) is required.
With finite element representations, this can be guaranteed if:
1. C
1
continuity in (w) is satisfied within individual elements
2. C
1
continuity in (w) is satisfied at nodal points. This is achieved if (w,
x
,
y
) are
used as nodal freedoms
3. C
1
continuity in (w) is satisfied along the edges between adjacent finite elements.
PBE(5)
PBE.2.2 4-noded rectangular element
This element has 12 local degrees of freedom
*
:
1 1
x1 x1
y1 y1
4 4
x4 x4
y4 y4
w q
m
m
d , f
w q
m
m






= =
` `






) )
(PBE.9)
Therefore, 12 terms from Pascal's triangle must be used for the interpolation of (w).

The shape functions for (w) are determined through a procedure similar to that of the beam
element (i.e. N = x C
-1
):
1 4
2 5
3 6
w Nd
where,
TV SV
N (2TV 2SU 4ST 4UV) , N (2SV 2TU 4ST 4UV)
32 32
TV SV
N ( 4aST) , N ( 4aST)
32 32
TV SV
N ( 4bUV) , N (4bUV)
32 32
continued
= |
|
|
|
= = + +
|
|
|
| = =
|
|
|
= =
|
|

(PBE.10)

*
Note that q
1
, m
x1
, m
y1
, etc..., are not values per unit length.
PBE(6)
7 10
8 11
9 12
continued
SU TU
N (2SU 2TV 4ST 4UV) , N (2TU 2SV 4ST 4UV)
32 32
SU TU
N (4aST) , N (4aST)
32 32
SU TU
N (4bUV) , N ( 4bUV)
32 32
x a x a y b y b
with, S , T , U , V
a a b b
|
|
|
|
= = + +
|
|
|
= =
|
|
|
= = |
|
|
+ +
= = = =
|

(PBE.10)
These shape functions do not satisfy the C
1
continuity requirements for edges opposite to the
corresponding nodes. For example, applying
x1
causes non-zero
w
y
| |
|

\
on edge 34.

This can be explained by considering the plate subjected to
x1
and
y3
only. If
w
x
| |
|

\
along
edge 14 is uniquely defined by
x1
, then
2
w
x y
| |
|

\
along edge 14 is uniquely defined by
x1
.
Also, if
w
y
| |
|

\
along edge 34 is uniquely defined by
y3
, then
2
w
y x
| |
|

\
along edge 34 is
uniquely defined by
y3
. Since
2
w
x y
| |
|

\
is identical to
2
w
y x
| |
|

\
for polynomial shape
functions of (w), this implies that at node (4)
2
w
x y
| |
|

\
is uniquely defined by
x1
and
uniquely defined by
y3
, which is impossible. Therefore, the assumptions that
w
x
| |
|

\
along
edge 14 is uniquely defined by
x1
, and that
w
y
| |
|

\
along edge 34 is uniquely defined by

y3
, are false. Hence, the requirement of C
1
continuity in (w) is not satisfied by this element.
To ensure C
1
continuity in (w), an additional
2
w
x y
| |
|

\
freedom is required at each node,
which increases the number of local freedoms to 16. However, enforcing the same
2
w
x y
| |
|

\

at nodes between adjacent elements having different thicknesses corresponds to different
PBE(7)
twisting moments m
xy
at these locations, which is not true for the real structure, and hence
leads to stiff solutions.
Despite the violation of the C
1
continuity requirements with the original 12 degree-of-
freedom element, it satisfies the Patch test discussed later, and hence performs well. It should
be noted, however, that the stiffer response compared to the real solution is no longer
guaranteed with this element.
PBE.2.3 The Patch test
The Patch test is a procedure for checking the performance of finite elements. The general
idea is that if elements can model constant strain conditions, then their use with finer meshes
ensures steady convergence to the real solution.
The test consists of assembling a patch of elements of different geometries such that at least
one node is completely surrounded by elements, and applying at the boundary nodes either
prescribed displacements or consistent loads that correspond to a state of constant strain.

For thin plate elements, the constant strain states correspond to constant
2 2 2
2 2
w w w
, ,
x y x y
| |
|

\
.
Solution is first obtained for the displacements at nodal freedoms which are not prescribed or
restrained, and the stresses and strains are computed within each individual element. The
Patch test is passed if, at every point in every element, computed strains agree exactly with
the assumed constant strain state.
Elements that pass the patch test are guaranteed to satisfy continuity requirements under
constant strain conditions. Therefore, even though such elements may not satisfy continuity
requirements under general strain conditions, their use within a fine mesh, where the strain
states within elements are almost constant, would lead to a reasonable approximation.
However, the fundamental characteristic of stiffer response with coarser meshes is not
necessarily retained by such elements.
PBE(8)
PBE.3 Reissner-Mindlin Elements (Thin to Moderately Thick Plates)
These elements account for transverse shear strains
xz
and
yz
, and hence can be used to
model thin up to moderately thick plates. The generalized strains and stresses are as given by
(PBE.3) and (PBE.4):
x
x x
y
y y
y
x
xy xy
xz x x
yz y y
m
x
m
y
m , D
y x
w
q
x
w
q
y



+ = = = =
` ` `


+
) )

)
(PBE.11)
PBE.3.1 Shape functions
Since the strains in (PBE.11) are first derivatives of (w,
x
&
y
), only C
0

continuity is
required for these fields. The presence of transverse shear strains makes (
x
&
y
)
independent of
w w
&
x y
| |
|

\
. Therefore, there is no requirement for satisfying C
1
continuity
in (w) as is the case with thin plate elements, and hence the associated complexities are
avoided.

Since only C
0
continuity is required, the same shape functions can be used for all fields, and
isoparametric elements can be utilised to model different geometric shapes:
1
1
1 m
1 m
m
m
x
y
0 0 x
0 0 y
x
y



(

=
` `
(

)



)

(PBE.12)
PBE(9)

The displacement and rotation fields u within the element are interpolated from the values at
the nodal freedoms using the same shape functions:
1
x1
y1 1 m
x 1 m
m y 1 m
xm
ym
i
w
w 0 0 0 0
u 0 0 0 0 Nd
w 0 0 0 0
where, shape function for node (i)
|
|

|

|
(
|

(
= = =
` ` |
(
|
(

)
|

|
|

)
|
|
|
=

(PBE.13)
Serendipity shape functions can be used for (
i
) similar to plane stress/strain elements. Note
that (w) is interpolated independently of the rotations (
x
&
y
).
PBE.3.2 Generalized strains
The generalized strains can be obtained in terms of d by combining (PBE.11) and (PBE.13),
and transforming the derivatives of (
i
) with respect to ( & ) to derivatives with respect to
(x & y):
x
1
y
x1
y1
y
x
m
xm x
ym
y
x
w
y
B Bd
y x
w
w
x
w
y



= = =
+ ` `


)
(PBE.14)
where,
PBE(10)
i i
1,3i 1 3,3i 4,3i 2
i i
2,3i 3,3i 1 5,3i 2
4,3i 1 5,3i i
1 y y
B B B
det J
1 x x
B B B
det J
B B
with, J jacobian matrix given in (HOP.7)

| | |
= = =
| |

\
|
|
| | |
= = = +
|
|

\
|
|
= =
|
|
|
=

(PBE.15)
PBE.3.3 Stiffness matrix
Similar to isoparametric plane stress/strain elements, the stiffness matrix is integrated
numerically with respect to ( & ):
( )
1
1
T T
i
i
1
1
k B DB det J d d w B DB (det J)

| |
=
|
\


(
(
(
(
(


(PBE.16)
Here as well, spurious mechanisms appear if less than a minimum number of Gauss points
are used. Furthermore, the local system is chosen as the global system since the effect of
element orientation is explicitly accounted for in B and (det J).
PBE.3.4 Shear locking
Consider a beam subjected to a linearly varying moment, as shown. Because of the bending
moment, end cross-sections undergo a rotation (
B
). All cross-sections undergo additional
rotation (
S
) due to constant transverse shear force. For thin beams (t << a), shear
deformation is much smaller than bending deformation (
S
<<
B
).

If we model the above beam with a 3-noded Reissner-Mindlin element in which (w) and ()
are interpolated separately, the solution predicts zero (w) because (w) is only a quadratic
function of (x) and the loading is asymmetric.
PBE(11)

The transverse shear deformation is:
dw w
since 0
dx x
| |
= = =
|

\

which is significantly larger than the true solution. This is referred to as shear locking, and
leads to much stiffer prediction of end rotations.
The same discussion applies to Reissner-Mindlin plate bending elements. The effect of shear
locking can be reduced if the contribution of shear to the stiffness matrix is integrated using a
reduced number of Gauss points.
PBE(12)
Tutorial PBE-1

1. Choose from Pascal's triangle the terms required for the
interpolation of (w) for the 8-noded thin plate element
shown.


2. Discuss why a term representing the thickness (t) is not required in (PBE.16).
3. Show that the following displacement vectors d represent spurious mechanisms for the 4-
noded isoparametric square plate element integrated with one Gauss point at the centre:
1 0 0 0
0 1 0 1
0 0 1 1
1 0 0 0
0 1 0 1
0 0 1 1
d , , ,
1 0 0 0
0 1 0 1
0 0 1 1
1 0 0 0
0 1 0 1
0 0 1 1





=
` ` ` `






) ) )

)

Sketch these mechanisms.
4. Model the plate shown below with 10 rectangular 4-noded isoparametric elements using
ANSYS. Repeat the analysis using the 8-noded elements. Compare the results of
displacement at the point of loading obtained from the two formulations with hand
calculation estimates, and explain possible discrepancies.

You might also like