NAOE 3103

FINITE ELEMENT METHOD

Lecture 16
Two-Dimensional Isoparametric Elements
and Numerical Integration

Md. Habibur Rahman

Lecturer
Department of Naval Architecture and Marine Engineering (NAME)
Bangladesh University of Engineering and Technology (BUET)
Two-Dimensional Isoparametric Elements and Numerical Integration
Element Stiffness Matrix
The element strain energy in the body is given by-
 1 T  1 T 
U e     dV  te    dA ... ... ... (i) where
  is the thickness of
2V 2 e the element.
Now the strain-displacement relations are-
 u 
 
 x 
  v 
   ... ... ... (ii )
  y 
 u v  
   
 y x  
We have,  u   u 
 x  1  J 22  J12    
 u       ... ... ... (iii )
  det J   J 21 J11   u 
 y    
 v   v 
 x  1  J 22  J12    
 v     J   ... ... ... (iv)
  det J  21 J11   v 
 y     2
Two-Dimensional Isoparametric Elements and Numerical Integration
From equation (ii), (iii), and (iv), we get-
 u 
  
  where
 u 
      J 22  J12 0 0 
  A   ... ... ... (v)  1 
 v  A  0 0  J 21 J11  ... ... ... (vi)
det J  
     J 21 J11 J 22  J12 
 
 v 
  

We have, u  N1q1  N 2 q3  N 3q5  N 4 q7

... ... ... (vii )
v  N1q2  N 2 q4  N 3q6  N 4 q8
u
 1    q1   1    q3   1    q5   1    q7 
1 1 1 1

 4 4 4 4

u 1 1 1 1 
   1    q1   1    q3   1    q5   1    q7 
 4 4 4 4 
 ... ... ... (viii )
v 1 1 1 1
   1    q2   1    q4   1    q6   1    q8 
 4 4 4 4 

v 1 1 1 1
   1    q2   1    q4   1    q6   1    q8 
 4 4 4 4  3
Two-Dimensional Isoparametric Elements and Numerical Integration
We can write  u 
  
 
 u 
   
   Gq ... ... ... (ix)
 v 
  
 
 v 
  
where
  1    0  1   0  1   0   1  0 
 1   1    0  1   0 1   0  1   0 

G 
4 0   1   0  1   0  1  0   1   
 
 0   1   0  1   0  1   0 1  
From equation (v) and (ix), we get-
   
  Bq  B  AG 
 

4
Two-Dimensional Isoparametric Elements and Numerical Integration
Now the stress is given by
 
  DBq
is
  a (3×3) material property matrix. The strain energy becomes-
 1 T 
U e   te    dA
e e
2
1 T  
1 1  

 

  q te   B DB det J d  d  q
T

e 2  1 1 
1 T  e
 q K q
e 2

where
  is the element stiffness matrix.
1 1

 
e T  
K  te   B DB det J d d
1 1

The element stiffness matrix is of dimension (8×8).

5
Two-Dimensional Isoparametric Elements and Numerical Integration

Force Vectors (Body Force)

Body
  force, that is distributed force, per unit volume, contributes to the global load
vector, . Consider the body force term appearing in the potential energy equation,

1 1  

 
T  T  T 
 u f dV  te  u f dA  te   u f det J d d
V e 1 1

Using
  and considering the body force , we get -
T  T e
 u f dV   q f
V e

where the (8×8) elemental body force vector is given by -

 1 1 T   fx 
 
e 
f  te    N det J d  d   
 1 1   fy 

6
Two-Dimensional Isoparametric Elements and Numerical Integration

Force Vectors (Traction Force)

Assuming
  a constant traction force in units of force per unit of surface area-

e t l T
T  e 23 0,0, Tx , Ty , Tx , Ty ,0,0  [  = length of the edge 2 -3]
2

7
Two-Dimensional Isoparametric Elements and Numerical Integration

Numerical Integration
Considering the problem of numerically evaluating a non-dimension integral of the
form:
1
I  f    d ... ... ... (i)
1

Now let consider the n-point approximation.

(Gaussian approach)
1
I  f    d
1

 w1 f  1   w2 f   2   ... ... ...  wn f   n  ... ... ... (ii)

where
  , , … … … , are the weights and , , … … … , are the sampling points or gauss
points.

8
Two-Dimensional Isoparametric Elements and Numerical Integration
Numerical Integration (One-Point Formula)
Considering the formula with n = 1 as;
1
I  f    d  w f    ... ... ... (i)
1
1 1

Since
  there are two parameters and , we consider requiring the formula in equation
(iii) to be exact when is a polynomial of order 1.

Thus if f     a0  a1 ; then we require-

1

Error a
1
0  a1  d  w1 f  1   0 ... ... ... (ii )

Error  2a0  w1  a0  a11   0

Error  a0  2  w1   w1a11  0 ... ... ... (iii )

We
  see that error is zero if or .
For any general function, then we have-
1
I  f    d  2 f  0 
1
which is seem to be familiar mid-point rule
(Figure)
9
Two-Dimensional Isoparametric Elements and Numerical Integration
Numerical Integration (Two-Point Formula)
Considering the formula with n = 2 as;
1
I  f    d  w f     w f    ... ... ... (i)
1
1 1 2 2

Since
  there are four parameters , and , , we consider requiring the formula in equation
(i) to be exact when is a polynomial of order 3.

Thus if f     a0  a1  a2 2  a3 3 ; then we require-

1
Error a
1
0  a1  a2 2  a3 3  d   w1 f  1   w2 f   2    0

1
 a1 2 a2 3 a3 4 
  a0 
2

3

4    w1 f  1   w2 f   2    0
  1
 a a a   a a a 
  a0  1  2  3     a0  1  2  3     w1 f  1   w2 f   2    0
 2 3 4  2 3 4 
 2a 
  2a0  2    w1  a0  a11  a212  a313   w2  a0  a1 2  a2 2 2  a3 23    0
 3 
2 
 a0  2  w1  w2   a1  w11  w2 2   a2   w112  w2 2 2   a3  w113  w2 23   0
3 
10
Two-Dimensional Isoparametric Elements and Numerical Integration

2 
a0  2  w1  w2   a1  w11  w22   a2   w112  w22 2   a3  w113  w22 3   0 ... ... ... (ii )
3 

We can see that the error is zero if

w1  w2  2
w11  w2 2  0
2
w112  w2 2 2 
3
w113  w2 23  0

These linear equations have the unique solution-

1
w1  w2  1 &  1   2   0.5773502619 ... ... ...
3

Note: The n-point Gaussian quadrature will provide an exact answer if f is a

polynomial of order (2n-1).

11