Finite Element Method (FEM)

Module Code:

Lecture on Multi Physics of FEM, 15 September 2019

By: Dr. Mesay Alemu Tolcha

Faculty of Mechanical Engineering, JiT, Ethiopia.

1/14
Weighted Residual Methods/ Discretization Principle

Galerkin’s in Finite Element Method/Discretization methods

• The classical Galerkin’s Method can be applied to the finite

element method.
• An approximate solution is assumed in the form,
n
X
∗
y (x) = yi Ni (x) (1)
i =1

where yi is the value of the solution function at x = xi and Ni (x) is

a corresponding shape function.

• Consider the linear element given in Figure 1.

2/14
Weighted Residual Methods/ Discretization Principle...

x
i j
Figure 1: Linear Element

For the linear element, distribution of y can be given as,

y (x) = ax + b (2)
Equation (2) can be written with the help of shape functions as,
y (x) = N1 (x)yi + N2 yj (3)
xj −x x−xi
where N1 (x) = xj −xi and N2 (x) = xj −xi

3/14
Weighted Residual Methods/ Discretization Principle...

Example 1
Use Galerkin’s method to formulate a linear finite element for
solving the differential equation
 
d dy
x − 4x = 0
dx dx
where 1 ≤ x ≤ 2 with y (1) = y (2) = 0

Solution, Analytical solution of the problem is

3
y (x) = x 2 − ln(x) − 1 (4)
ln(2)
The finite element solution proceeds in the following way. Assuming
a linear element with y (x) = Ni (x)yi + Nj (x)yj , the residual
equation for the element is
4/14
Weighted Residual Methods/ Discretization Principle...

Zxj    
d dy
Nk x − 4x dx = 0, k = i , j (5)
dx dx
xi

Expansion of this equation gives,

Zxj   Zxj
dy
Nk d x − Nk 4xdx = 0, k = i , j
dx
xi xi

Integration by part of the first term gives us,

  Z xj Zxj
dy x dy
Nk x |xji − x dNk − Nk 4xdx = 0, k = i , j
dx xi dx
xi

Substitution of the approximate solution,

5/14
Weighted Residual Methods/ Discretization Principle...

Solution...
  Z xj   Zxj
dy x dNi dNj
Nk x |xji − x yi + yj dNk − Nk 4xdx = 0, k = i , j
dx xi dx dx
xi

Or
Z xj     Zxj
dNi dNj dy xj
x yi + yj dNk = Nk x |xi − Nk 4xdx = 0, k = i , j
xi dx dx dx
xi

dNi dNj 1
Here, dx = −1
xj −xi and dx = xj −xi

6/14
Weighted Residual Methods/ Discretization Principle...

Solution...
The above equation can be split in to two equations: one for k = i
and other for k = j
Z xj     Zxj
dNi dNj dy xj
x yi + yj dNi = Ni x |xi − Ni 4xdx = 0
xi dx dx dx
xi

Z xj     Zxj
dNi dNj dy x
x yi + yj dNj = Nj x |xji − Nj 4xdx = 0
xi dx dx dx
xi
Further simplification gives us,
Zxj xj
1 4
Z
dy
x(xi − yi )dx = −xi |xi − (xj − x)xdx
(xj − xi )2 dx xj − xi xi
xi

7/14
Weighted Residual Methods/ Discretization Principle...

Solution...
Zxj xj
1 4
Z
dy
x(xi − yi )dx = xj |xj − (xj − x)xdx
(xj − xi )2 dx xj − xi xi
xi

Integration of all terms gives us,

 3
4xi − 6xi2 xj − 2xi

xi − xj dy
(yi − yj ) = −xi |xi − (6)
2(xj − xi ) dx 3(xj − xi )
" #
xi − xj dy 4xj3 − 6xj2 xi − 2xi
(yi − yj ) = xj |xj − (7)
2(xj − xi ) dx 3(xj − xi )

8/14
Weighted Residual Methods/ Discretization Principle...

Solution...
The last two equations can be written in a more compact form as,
)  3 2 3

 X dy | − 4xi −6xi xi +1 +2xi +1 
" #(
xi + xi +1 1 −1 yi i dx xi 3(xi +1 −xi )
= 3 2 3
2 (xi +1 − xi ) −1 1 yi +1 Xi +1 dy |x − 4xi +1 −6xi +1 xi +2xi 
dx i +1 3(xi +1 −xi )

Consider now dividing the domain in to four elements as shown

below.

x
1 2 3 4 5
Figure 2: Example 1

The last equation can be written for each element as

Element 1:
9/14
Weighted Residual Methods/ Discretization Principle...

" # ( )  dy 
4x13 −6x12 x2 +2x23 
1 −1 y1 X | −
1 dx x1 3(x2 −x1 )
K1 = 4x23 −6x22 x1 +2x13 
−1 1 y2 X2 dy |x −
dx 2 3(x2 −x1 )

Element 2:
" # ( )  dy 
4x23 −6x12 x3 +2x33 
1 −1 y2 X | −
2 dx x2 3(x3 −x2 )
K2 = 4x33 −6x32 x2 +2x23 
−1 1 y3 X3 dy |x −
dx 3 3(x3 −x2 )

Element 3:
" # ( )  dy 
4x33 −6x32 x4 +2x43 
1 −1 y3 X | −
3 dx x3 3(x4 −x3 )
K3 = 4x43 −6x42 x3 +2x33 
−1 1 y4  X4 dy
dx |x4 − 3(x4 −x3 )

10/14
Weighted Residual Methods/ Discretization Principle...

Element 4:
" # ( )  dy 
4x43 −6x42 x5 +2x53 
1 −1 y4 X | −
4 dx x4 3(x5 −x4 )
K4 = 4x53 −6x52 x4 +2x43 
−1 1 y5  X5 dy
dx |x5 − 3(x5 −x4 )

The next step is to assemble the element equations to get a global

one.

Solution ... on next page

11/14
Weighted Residual Methods/ Discretization Principle...

  
k1 −k1 0 0 0 y1 
 
−k1 k1 + k2 0 0
 
k2 y 2


  
 y3  =
 0 −k2 k2 + k3 −k3 0 

 0 0 −k3 k3 + k4 −k4   y4 
  

 

 
0 0 0 −k4 k4 y5 
4x13 −6x12 x2 +2x23
 
dy


 X 1 |
dx 1x − 3(x2 −x1 )



 4x 3 −6x 2 x +2x 3 4x 3 −6x 2 x +2x 3 
2 1 2 3
−

 2 1 2 3

 3(x2 −x1 ) 3(x3 −x2 )

 


4x33 −6x32 x2 +2x23 4x33 −6x32 x4 +2x43
3(x 3 −x 2 ) − 3(x4 −x3 )
4x43 −6x42 x3 +2x33 4x43 −6x42 x5 +2x53 

 
−

 


 3(x4 −x3 ) 3(x5 −x4 ) 

3 2 3
 X dy | − 4x5 −6x5 x4 +2x4 

 

5 dx x5 3(x5 −x4 )

12/14
Weighted Residual Methods/ Discretization Principle...

Solution ...
Applying boundary conditions,
Since y1 and y5 are known, the above equation reduces to,
    4x23 −6x22 x1 +2x13 4x23 −6x22 x3 +2x33

−

k1 + k2 k2 0  y2 
  3(x −x )
 3 22 1 3
  3(x3 −x2 )



4x3 −6x3 x2 +2x2 3 2
4x3 −6x3 x4 +2x4 3
 −k2 k2 + k3 −k3  y3 = −
 
3(x3 −x2 ) 3(x4 −x3 )
 4x43 −6x42 x3 +2x33 4x43 −6x42 x5 +2x53 
   
0 −k3 k3 + k4 y4
  
−

3(x4 −x3 ) 3(x5 −x4 )

13/14
Weighted Residual Methods/ Discretization Principle...

Figure 3: Solution - Example 1

14/14