ISOPARAMETRIC TRANSFORMATION

and
NUMERICAL INTEGRATION

LECTURE 12
1
ISOPARAMETRIC ELEMENTS

r
x=  xi Li ()
i=1

For a linear transformation r = 2

x = x1 N1 () + x2 N2 ()

= x1 (1- ) + x2 (1 + )
2 2

2
3
For example an element whose x co-ordinates
are given by x1 = 3 & x2 = 7

Then x1 = x1 (1 -  ) + x2 (1 + )
2 2

3= 3 (1 - ) + 7 (1 + )
2 2
or 6 = 3 – 3 + 7 + 7
or 4 = -4
or = -1
ie the point xi = 3 transforms to  = -1 in natural
co-ordinate space 4
similarly x2 = x1 (1 - ) + x2 (1 +)
2 2
7 = 3 (1 - ) + 7 (1 + )
2 2
14 = 3 - 3 + 7 + 7
4 = 4 or =1
The point x2 = 7 in Cartesian space gets
transformed to 2 = +1 in Natural co-ordinate
space. Similarly every point in X space
transforms to a corresponding point in 
space

5
So the transformation
r

X = Ni xi () transforms the geometry

i=1
from Cartesian space to Gaussian space

Similarly we have the approximation of the

field variable in terms of shape functions
expressed as s
u =  ui Ni ()
i=1
6
Here ‘r’ - the number of nodes used for
geometric transformation

‘s’ - the number of nodes used for

approximation of field variable.

In general the polynomial used for geometric

transformation need not be of the same
order as that used for the field variable
approximation.
7
 In other words two sets of nodes exists for
the same region or element.

 One set of nodes for co-ordinate

transformation from Cartesian space to
natural co-ordinate space

 One set of nodes for approximating the

variation of the field variable over the
element.

8
Depending upon the relationship
between these two polynomials
elements are classified into three
categories as

sub parametric elements r<s

iso-paramatric elements r=s
super-parametric elements r>s

9
10
11
12
13
Field variable approximation

Geometric Transformation

14
15
16
Jacobian of Transformation
Among the 3 cases given above Isoparametric
are more commonly used due to their
advantages which include the following:

i) Quadrilateral elements in (x,y) coordinates

with curved boundaries get transformed to a
rectangle of (2 x 2) units in (, ) co-ordinates

ii) Numerical integration is more easily

performed as limits of integration vary from –1
to +1 for all elements.
17
We have seen that determination of the
stiffness matrix requires the computation of
derivative of shape functions with respect to
‘x’. However as the shape functions
(Interpolation functions) are expressed in
terms of  &  co-ordinates (natural co-
ordinates) we use the chain rule.
dN1 = dN1 d = dN1 1
dx d dx d dx / d

= dN1 1 = J-1 dN1

d J d 18
Here J = dx/d is the ‘Jacobian’ of
transformation from Cartesian space to
natural co-ordinate space. It can be
considered as the scale factor between the
two co-ordinate systems.

19
 Jacobian of transformation for 2 Noded
Linear Element

For a 2 Noded element the shape functions

are given by

N1 () = (1 - )
2

N2 () = (1+)
2

20
Now x = N1x1 + N2 x2
= (1 - ) x1 + (1 +) x2
2 2
dx = J = -1 x1 + 1 x2
d 2 2
= (x2 – x1) = L
2 2

Here (x2 – x1) represents the length of the

element. So the Jacobian of transformation
for a 2 noded element is given by L/2
21
3- Noded Quadratic element:-

N1 = -/2 (1- )
N2 = (1- ) (1+ )
N3 = /2 (1+ )
u = N 1 u 1 + N2 u 2 + N3 u 3 &
x = N1 x1 +N2 x2 + N3 x3

= -/2(1- )x1 + (1- )(1+ )x2 +/2(1+ )x3

22
J = dx = -1 +2 -2 1 + 2 x
Jacobian of transformation for 2-D
elements:-
In the case of two dimensional elements the
shape functions Ni are functions of both x &
y. When we obtain the same using Natural
co-ordinates the shape functions will be
functions of  & . In order to derive the
stiffness matrices we need to evaluate the
derivatives with respect to x and y. We
therefore apply the chain rule to get
23
Ni = Ni x + Ni y
 x  y  ----- (1)

Ni = Ni x + Ni y

 x  y 

or in Matrix notation
Ni x y Ni
   x
=
Ni x y Ni
   y
24
Ni = [J] Ni ------ (2)
 x
Ni Ni
 y

Here ‘J’ is the Jocobian of transformation

from Cartesian to Gaussian space. This
gives the relationship between the
derivatives of Ni with respect to the global
and local co-ordinates.
From (2) we obtain
25
Ni = [J]-1 Ni ------ (3)
x 
Ni Ni
x 

Hence the Jacobian Martrix [J] must be

non-singular

[J] = x y
  ------------------(4)
x y
  26
We know that x =  Ni (,) xi ----- (5)
i =1
m
y =  Ni (,) yi
i=1
m m
 x =  xi Ni y =  yi Ni ---- (6)
 i = 1   i = 1 
m m
x =  xi Ni y =  yi Ni
 i = 1   i = 1 
27
Substituting equation (6) in (4) we get

[J]=  xi Ni  yi Ni

 
 xi Ni  yi Ni
 

= N1 N2 N3 ….. Nm x 1 y1

    x2 y 2
N1 N2 N3 ….. Nm x m ym
   
28
In general the Jacobian of transformation
in 3D is given by

 x y z 
    

x y z 
J  
   
 x y z 
 
    
29
  x y z 
    

x y z 
J  

   
 x y z 
 
    

30
Problem:
Evaluate the Cartesian co-ordinate of the
point P which has local co-ordinates  = 0.6
and  = 0.8 as shown in the Figure.

31
32
Given: Natural co-ordinates of point P
 = 0.6
 = 0.8
Cartesian co-ordinates of point 1,2,3 and 4

x1 3 ; y1 2
x 2 9 ; y 2 4
x3 6 ; y 3 8
x 4 4 ; y 4 5

33
To Find: The Cartesian co-ordinates of the point
P (x,y)
Solution:
Shape functions for quadrilateral element are,
1
N 1  (1   ) (1   )
4
1
N 2  (1   ) (1   )
4
1
N 3  (1   ) (1   )
4
1
N 4  (1   ) (1   )
4 34
Substituting the values

1
 N 1 (0.6,0.8)  (1  0.6) (1  0.8) 0.02
4
1
 N 2 (0.6,0.8)  (1  0.6) (1  0.8) 0.08
4
1
 N 3 (0.6,0.8)  (1  0.6) (1  0.8) 0.72
4
1
 N 4 (0.6,0.8)  (1  0.6) (1  0.8) 0.18
4
35
Co  ordinate, x N 1 x1  N 2 x 2  N 3 x3  N 4 x 4
0.02(3)  0.08(9)  0.72(6)  0.18(4)
x  5.82

Co  ordinate, y N 1 y1  N 2 y 2  N 3 y 3  N 4 y 4
0.02 (2)  0.08(4)  0.72(8)  0.18(5)
y 7.02
Co  ordinates are ( x, y ) (5.82,7.02) 

36
Problem 1
Evaluate [ J ] at   2 for the linear
quadrilateral element shown in Fig.

37
38
Given:
Natural co-ordinates at point, P
1 1
  0.5 ;   0.5
2 2
Cartesian co-ordinates of point 1,2,3 & 4
x1 4 ; y1 4
x 2 7 ; y 2 5
x3  8 ; y 3 10
x 4 3 ; y 4 8
39
To Find:1.Jacobian matrix [J].

Solution: Jacobian matrix for quadrilateral

element is given by,

 x y 
    J 11 J 12 
J   x 
 J   
 y   J 21 J 22 
   
40
 1
J 11  [ (1   ) x1  (1   ) x 2  (1   ) x3  (1   ) x 4 ]
4
1
J 12  [ (1   ) y1  (1   ) y 2  (1   ) y 3  (1   ) y 4 ]
4
1
J 21  [ (1   ) x1  (1   ) x 2  (1   ) x3  (1   ) x 4 ]
4
1
J 22  [ (1   ) y1  (1   ) y 2  (1   ) y 3  (1   ) y 4 ]
4

41
 1
J 11 (0.5,0.5)  [ (1  0.5) 4  (1  0.5) 7  (1  0.5) 8  (1  0.5) 3]
4
2.25
1
J 12 (0.5,0.5)  [ (1  0.5) 4  (1  0.5) 5  (1  0.5)10  (1  0.5) 8]
4
0.875
1
J 21 (0.5,0.5)  [ (1  0.5) 4  (1  0.5) 7  (1  0.5) 8  (1  0.5) 3]
4
0.25
1
J 22 (0.5,0.5)  [ (1  0.5) 4  (1  0.5) 5  (1  0.5)10  (1  0.5) 8]
4
2.375

42
  J 11 J 12 
  J   
 J 21 J 22 

 2.25 0.875
 
 0.25 2.375

43
 Stiffness Matrix for a 2 Noded Axial
Element
[K] =  BT D BAdx
0

[B] = du = dN = 1 dN
dx dx J d

= 2 dN1 dN2
L d d
= 2 d (1 -) d (1+)
L d 2 d 2
= 2 -1 1 -1 1
L 2 2 = L L 44
 +1

[K] = A ∫ –1/L E <-1/L 1/L> J d

-1 1/L
+1

= EA ∫ –1/L < -1/L 1/L > L/2 d

-1 1/L

+1

= EAL ∫ 1/L2 -1/L2 d

2 -1 -1/L2 1/L2
+1

= EA 1 –1 ∫ d = EA 2 1 -1 45
Problem:
For the four noded rectangular element
shown if Fig. determine the following:
i) Jacobian matrix
ii) Strain-Displacement matrix
iii)Element stresses

Take E = 2 x 105 N/mm2; v = 0.25;

u =[0, 0, 0.003, 0.004, 0.006, 0.004, 0, 0]T
= 0 ;  = 0
Assume plane stress condition.
46
47
Cartesian co-ordinates of point 1,2,3 & 4

x1 0 ; y1 0
x 2 2 ; y 2 0
x3  2 ; y 3 1
x 4 0; y 4 1

Young’s modulus, E = 2 x 105 N/mm2

Poisson’s ratio v = 0.25

48
  0 
 0 
 
0.003
 
0.004
Displacement , u  
0.006
0.004
 
 0 
 0 
 
Natural Co  ordinates,  0,  0

To Find: 1. Jacobian matrix, J.

2. Strain Displacement, [B]
3. Element stress,  49
Solution:
 J 11 J 12 
J   
 J 21 J 22 
1
J 11  [ (1   ) x1  (1   ) x 2  (1   ) x3  (1   ) x 4 ]
4
1
J 12  [ (1   ) y1  (1   ) y 2  (1   ) y 3  (1   ) y 4 ]
4
1
J 21  [ (1   ) x1  (1   ) x 2  (1   ) x3  (1   ) x 4 ]
4
1
J 22  [ (1   ) y1  (1   ) y 2  (1   ) y 3  (1   ) y 4 ]
4
50
 1 1
J 11 (0,0)  [0  2  2  0] ; J 12 (0,0)  [0  0 1  1]
4 4
1 0

1 1
J 21 (0,0)  [0  2  2  0] ; J 22 (0,0)  [ 0  0  1 ]
4 4
0 0.5

51
  J 11 J 12 
  J   
 J 21 J 22 

1 0 
Jacobian matrix, [ J ]  
 0 0.5
 J 1 0.5  0  0.5
52
Strain- Displacement matrix for quadrilateral
element is,

 J 22  J 12 0 0 
1 1
 [ B ]   0 0  J 21 J 11  
J 4
  J 21 J 11 J 22 
 J 12 

  (1   ) 0 (1   ) 0 (1   ) 0  (1   ) 0 
  (1   ) 0  (1   ) 0 (1   ) 0 (1   ) 0 
 
 0  (1   ) 0 (1   ) 0 (1   ) 0  (1   )
 
 0  (1   ) 0  (1   ) 0 (1   ) 0 (1   ) 

53
  1 0 1 0 1 0 1 0
 0.5 0 0 0  1 0  1 0
1  1 1 0 1 0 
 [B ]   0 0 0 1  
0.5 4 0  1 0 1 0 1 0  1
 0 1 0.5 0  
0 1 0 1 0 1 0 1

  0. 5 0 0.5 0 0.5 0  0.5 0 

1  
 0  1 0  1 0 1 0 1
0.5 4  
  1  0.5  1 0.5 1 0.5 1  0.5

 1 0 1 0 1 0 1 0
0.5  
 0  2 0  2 0 2 0 2
0.5 4  
  2  1  2 1 2 1 2  1

54
  1 0 1 0 1 0 1 0
[B ] 0.25  0  2 0  2 0 2 0 2 
  2  1  2 1 2 1 2  1

Element stress,  = [D] [B] {u}

Stress  strain relationship matrix ,
 
 1 v 0 
E
[ D]  2
v 1 0 
1 v  1 v
0 0 
 2  55
  
1 0 . 25 0
2 10 5  
 2
 0.25 1 0 
1  (0.25)  1  0.25 
 0 0 
 2 

 1 0.25 0 
3  
213.3310  0.25 1 0 
 0 0 0.375
4 1 0 
213.3310 3 0.25  1 4 0 
 0 0 1.5
56
 4 1 0 
3  
[ D] 53.3310  1 4 0 
 0 0 1.5

Substituting the values in Element stress

equation

 D B d 

57
 4 1 0
   53.33 10 3 1 4 0 

 0 0 1.5
 1 0 1 0 1 0 1 0
0.25  0  2 0  2 0 2 0 2 
  2 1  2 1 2 1 2  1
 0 
 0 
 
0.003
 
0.004
  
0.006
0.004
 
 0 
 0 
  58
  4  2 4  2 4 2  4 2 
3 
53.3310 0.25   1  8 1  8 1 8  1 8  
  3  1.5  3 1.5 3 1.5 3  1.5
 0 
 0 
 
0.003
 
0.004
  
0.006
0.004
 
 0 
 0 
  59
 0.036
3  
  13.33310 0.009
0.021
 
480
 
   120  N / m 2

280
 

60
 NUMERICAL INTEGRATION
In the isoparametric formulation of higher
order elements we see that the strain-
displacement matrix [B] is given by
[B] = du = dN [] = 1 d[N]
dx dx J d

= 1 d (- + 2 1 – 2  + 2)
J d 2 2

61
Here J = (-1 + 2 -2 1 + 2)
2 2
Therefore Matrix [B] is a function of , with
polynomials in  in its denominator because of
the 1/J factor. Hence the equation (A) cannot
be integrated to give on the solution. Hence
we resort to numerical integration.

62
So evaluation of integrals of the form
b
 F(x) dx becomes difficult or impossible in
a
cases where the integrand F has functions of
x in both numerator denominator.

The basic idea behind whatever numerical

integration technique we may employ is that of
obtaining a function P(x) which is both a
suitable approximation of F(x) and simple
enough to integrate.
63
Referring to Fig the variation of F(x) is
b
shown. Evaluation of the Integral  F(x) dx
a
will yield the area under the F(x) curve
between points x1 (= a) & x2 (= b).

64
 “Trapezoidal rule”,
b
h

a
F ( x)dx  ( y 0  y8  ( y1  y 2  ........y 7 ))
2

65
66
Trapezoidal Rule Simpsons Rule

67
b
h

a
F ( x)dx  ( y 0  y8  4( y1  y 3  ........y 7 )  2 ( y 2  y 4  ........y 6 )
3

68
Gauss Quadrature:- Amongst the several
schemes available for evaluating the area
under the curve F(x) between two points the
gauss quadrature method has proved to be
most useful for isoparametric elements. As in
isoparametric formulation, the limits of the
integral are always from –1 to +1, the problem
in gauss integration is to evaluate the integral
+1
I=  F() d.
-1

69
The simplest and probably the crudest way to
evaluate the integral is to sample or evaluate
F() at the mid point of the interval and to
multiply this by the length of the element
which is ‘2’ [because 1 = -1 & 2 = 1 &
(2 – 1) = 2]
 F(x) dx = I = 2 fi

This result will be exact only if the actual

function happens to be a straight line.

70
One point formula

71
We can extend the same to take two sampling points
or three etc.Generalization of this relation gives
+1
I=  F() d = w1f1 + w2f2 + ….. wnfn
-1
n
=  wif (i)
i=1
Here wi is called the ‘weight’ associated with the i th
point and n is the number of sampling points. The
Table (1) gives the sampling points and the
associated weights (wi) for Gauss quadrature.

72
No.of Location Weight Wi
points

1 1 = 0.00000 2.00000
2 1,2=0.57735 1.000000
3 1,3=0.77459 0.55555
73
 =
Thus to approximate the integral I, the
function f() is evaluated at each of several
locations i, and each f(i) is multiplied by
the approximating weights w. The
summation of these products gives the
value of the integral. The sampling points
are generally located symmetrically with
respect to the center of the interval.
Symmetrically paired points have the same
weight wi.

74
As an example consider the evaluation of the
Integral I using 2 sampling points i.e. n = 2.

I  (1.0) ( f at  = - 0.577350269189626) +
(1.0) (f at  = + 0.577350269189626)

75
76
In general if we know that the integral to be
evaluated is of order p then the number of
sampling points required n is given by the
relation

2n-1 = p

The calculated number of sampling points can

be rounded off to the nearest integer

77
78
Problem 1


2
Evaluate the integral I  ( 2  x  x ) dx and
1
compare with exact solution.
1

Given: Integral, I (2  x  x 2 ) dx

1

2
 f ( x) 2  x  x

To Find: The integral I by using Gauss

quadrature.
79
Solution:
We know that , the given integrand is a
polynomial of order 2.

So, 2n-1 = 2
 2n =3
 n = 1.5  2

For two point Gaussian quadrature,

1
x1  0.577350269 w1 1
3
1
x 2   0.577350269 w2 1
3 80
 2
f ( x ) 2  x  x
2
f ( x1 ) 2  x1  x 1
2
2  (0.577350269)  (0.577350269)

f ( x1 )  2.9106836
w1 f ( x1 ) 12.9106836
 w1 f ( x1 )  2.9106836
81
 2
f ( x 2 ) 2  x 2  x 2
2
2  (0.577350269)  ( 0.577350269)
f ( x 2 ) 1.755983
w2 f ( x 2 ) 11.755983
w2 f ( x 2 ) 1.755983

w1 f ( x1 )  w2 f ( x 2 ) 2.9106836 1.755983
4.666666
1
 (2  x  x ) dx 4.666666
2

82
1
Exact Solution:

(2  x  x
2
) dx 2 x 
1
1  
1 2
 x
2
1
1   
1 3
3
x
1
1
1

1 1
2[1  ( 1)]  [1  (1)]  [1  ( 1)]
2 3
4.666666

83
Using Gauss Quadrature evaluate the
following integral using 1 2 and 3 point
Integration.

i)
ii)

iii)

84
 ,=0.57735

n
F(,) =  f (i ,i ) wiwj
i=1

85