5.3 Stress Analysis Using Mapped Four Node Elements: 5.3.1 The Strain Matrix For The Four Node Element
5.3 Stress Analysis Using Mapped Four Node Elements: 5.3.1 The Strain Matrix For The Four Node Element
5.3 Stress Analysis Using Mapped Four Node Elements: 5.3.1 The Strain Matrix For The Four Node Element
In this section we develop the finite element analysis that begins with a square Master
Element and, by mapping shape functions, transforms this into a quadrilateral of desired
shape. We present the analysis using a 4-node element and with shape functions
derived from linear interpolation. It is possible in this way to bring out the essential
features of the analysis whilst keeping the number of variables and algebra within
manageable bounds. However the application of this type of element is seen to best
advantage only when it is applied using elements with more nodes and with higher
order interpolation functions which give varying stresses over the element.
We first develop the matrix that relates the strain at a point in a 4-node element (as
shown in the figure below) to the nodal displacements.
= B u (5.20)
(u1 , v1 ) at node 1 and similarly for the other nodes, so the nodal displacement vector is
u = u1 v1 u 2 v2 u3 v3 u 4 v4
T
(5.21)
58
At a point (x, y) in the element the elastic strains are given by the same elastic strain
vector that we derived for the triangular element. We reproduce it here.
u
x x
( x, y) = y = v (5.22)
y
xy u v
y + x
We are going to find the explicit expressions for the strain components.
v = N1v1 + N 2 v2 + N 3 v3 + N 4 v4 (5.23)
We recall that the N terms in these equations are functions of and , and the
displacements at the 4 nodes are constants for the element. So u and v are functions of
and .
x = N1 x1 + N 2 x2 + N 3 x3 + N 4 x4
y = N1 y1 + N 2 y 2 + N 3 y3 + N 4 y 4 (5.24)
Again the terms in N are functions of and , and the co-ordinates of the 4 nodes are
constants for the element. So x and y are functions of and .
Within the real element, u and v vary with x and y, but x and y are functions of co-
ordinates and in the Master Element because x and y have been derived by
mapping from it. So in mathematical terms, u and v are functions of x and y which are in
turn functions of and , or
59
u = f u [ x( , ), y ( , )] and v = f v [ x( , ), y ( , )]
Therefore we have
u u x u y
= +
x y
and
u u x u y
= + (5.25)
x y
u x y u
x
u = x
y u (5.26)
y
The 2 2 matrix in [ ] has an important role in the mapping process and is called a
Jacobian by mathematicians. It is usually presented with the partial derivatives
replaced by equivalent symbols
J 11 J 12
J J 22
21
So the alternative form of the above matrix equations (5.26) is
u u
J 11 J 12 x
u = (5.27)
J 21 J 22 u
y
The derivatives u and u which we need to get the elastic strains from the nodal
x y
u u u
−1
x J 11 J 12 1 J 22 − J 12
u = u =
J 22 J 11 u
(5.28)
J 21 det J − J 21
y
60
where
det J = J 11 J 22 − J 12 J 21 (5.29)
Similarly we have
v v
x 1 J 22 − J 12
v = − J (5.30)
det J 21 J 11 v
y
The above two matrix equations contain all the partial derivatives that we need to
evaluate the three elastic strains. It is tidy to collect them all together in a single matrix.
u u
x
u J 22 − J12 0 0 u
y 1 − J 21 J11 0 0
v = (5.31)
det J 0 0 J 22 − J12 v
x 0 0 − J 21 J11
v v
y
All the terms on the right of the matrix equations (5.31) are known and therefore are the
left-hand side derivatives. These are the ones we seek for strain matrix. So we are
now in a position to evaluate. Returning therefore to strain matrix
u
x x
v
y =
y
xy u v
y + x
The three strain components are respectively
u 1 u u
= ( J 22 − J 12 )
x det J
v 1 v v
= (− J 21 + J 11 ) (5.32)
y det J
u v 1 u u v v
+ = (− J 21 + J 11 + J 22 − J 12 )
y x det J
61
We collect the above three equations together in the strain matrix
u
u
x x J 22 − J 12 0 0 u
v 0
J 11
1 (5.33)
y = = 0 − J 21
v
y det J − J 21 J 11 J 22 − J 12
xy u v
y + x v
The terms in this matrix equation contain only
1. the derivatives of the shape functions, which are known,
2. the nodal co-ordinates of the real element, which are known,
3. the nodal displacements, which are the principal unknowns that we seek.
We need finally to go through some tedious manipulation to make the above matrix
equation more explicit and get it in the form that we shall use for computation. First we
look at the vector of derivatives
T
u u v v
u = N 1u1 + N 2 u 2 + N 3u 3 + N 4 u 4
so
u N 1 N 2 N 3 N 4
= u1 + u2 + u3 + u4
u N 1 N 2 N 3 N 4
also = u1 + u2 + u3 + u4
and v = N1v1 + N 2 v2 + N 3 v3 + N 4 v4
so
v N 1 N 2 N 3 N 4
= v1 + v2 + v3 + v4
also
62
v N 1 N 2 N 3 N 4
= v1 + v2 + v3 + v4
= G u (5.34)
We can derive explicit expressions for the derivatives of the shape functions in G .
1 N 1 1 N 1 1
N1 = (1 − )(1 − ) so = − (1 − ) , = − (1 − )
4 4 4
N 2
= (1 − ) , N 2 = − 1 (1 + )
1 1
N2 = (1 + )(1 − ) so
4 4 4
1 N 3 1 N 3
N3 = (1 + )(1 + ) so = (1 + ) , 1
= (1 + )
4 4 4
N 4
= − (1 + ) , N 4 = 1 (1 − )
1 1
N4 = (1 − )(1 + ) so
4 4 4
(5.35)
If we substitute the expressions for the shape function derivatives in terms of and
− (1 − ) 0 (1 − ) 0 (1 + ) 0 − (1 + ) 0
− (1 − ) 0 − (1 + ) 0 (1 + ) 0 (1 − ) 0
G = 1
4 0 − (1 − ) 0 (1 − ) 0 (1 + ) 0 − (1 + )
0 − (1 − ) 0 − (1 + ) 0 (1 + ) 0 (1 − )
(5.36)
63
Finally we express the terms from the Jacobian matrix in terms of and .
x
J 11 = but x = N1 x1 + N 2 x2 + N 3 x3 + N 4 x4
so
x N 1 N 2 N 3 N 4
= x1 + x2 + x3 + x4
Therefore
x 1
J 11 = = [−(1 − ) x1 + (1 − ) x 2 + (1 + ) x3 − (1 + ) x 4 ] (5.37)
4
Similarly we find
y 1
J 12 = = [−(1 − ) y1 + (1 − ) y 2 + (1 + ) y 3 − (1 + ) y 4 ] (5.38)
4
x 1
J 21 = = [−(1 − ) x1 − (1 + ) x 2 + (1 + ) x3 + (1 − ) x 4 ] (5.39)
4
y 1
J 22 = = [−(1 − ) y1 − (1 + ) y 2 + (1 + ) y 3 + (1 − ) y 4 ] (5.40)
4
We can now return to the elastic strain matrix equation and express it as
u
x x J 22 − J 12 0 0
v 0 J 11 G u
1
y = = 0 − J 21 (5.41)
y det J
xy u v − J 21 J 11 J 22 − J 12
y + x
If we let
J 22 − J 12 0 0
1
A = 0 0 − J 21 J 11 (5.42)
det J
− J 21 J 11 J 22 − J 12
then we have
= AG u
64
or = Bu where B = AG (5.43)
this relates the elastic strains to the nodal displacements and is the relation we set out
to establish at the beginning of this section.
With the relation between the nodal displacements and the elastic strains, established
for the four-node element the way forward to the stiffness matrix follows the same basic
analysis that we developed for the triangular element. The important difference is the
following:
In the triangular element analysis we developed an expression for the elastic strain
energy generated by the virtual displacements. This involved an integral through the
volume of the element. Since the integrand was constant throughout the element, the
integral was easily worked out as the integrand times the element volume. We now
wish to analyse a situation with the four-node element where the stress, therefore the
integrand varies across the element so the integration is more complex.
We recall that
F
Wext = *e
T e
Wint = ( x, y )* ( x, y)dV = B DBu dV = B dV
T
*e
T T e *e T
T
V V V
B dV = F
T
(5.44)
V
and = DBu
so
B DBudV = F
T
65
But the nodal displacements in the vector u are constant so we take it outside the
integral sign then
B DB dV u = F
T
V
This is of the form
K u = F
where K is the stiffness matrix and so
K = BT DBdV (5.45)
V
To establish the stiffness matrix we need to perform the integral of the above equation
through the volume of the element and the function to be integrated is B DB . We
T
If in the stiffness matrix integral we let the small volume dV in the real element have an
area dx.dy in a sheet of uniform thickness t then
dV = tdx dy so the stiffness matrix is
K = t B DBdx dy
T
(5.46)
The evaluation of this integral is complicated by the fact that the matrices in the
integrand are expressed in terms of and not x and y. the way forward is to perform
the integral in the Master Element, where the co-ordinates are and , and we use the
mapping process to relate the element dx.dy in the real element to the corresponding
element d d in the Master Element, as the following diagram shows.
66
It turns out that the area of the real calculus volume element is
det J = J 11 J 22 − J 21 J 12 (5.29)
Physically det J is the ratio of the area of the real calculus element to the area of the
Master calculus element. Since the thickness t of the sheet is the same for both,
The advantage of this transform from x and y to and is that the integration is
performed over the same element, the simple square Master Element with sides of 2
units, irrespective of the size and shape of the real element.
The above stiffness matrix cannot be integrated directly so numerical integration has to
be used. We now explain how to achieve this.
67