Scribd
Upload
34 views14 pages

Fem Formulation

This document describes finite element formulation and weighted residual methods. It discusses direct formulation, approximation of solutions, and residual errors. Four weighted residual methods are presented: point collection, subdomain collection, least squares, and Galerkin. It also introduces the variational formulation and conditions for functionals and functionals to be linear, quadratic, and positive definite.

Uploaded by

michael Ketsela
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
34 views14 pages

Fem Formulation

This document describes finite element formulation and weighted residual methods. It discusses direct formulation, approximation of solutions, and residual errors. Four weighted residual methods are presented: point collection, subdomain collection, least squares, and Galerkin. It also introduces the variational formulation and conditions for functionals and functionals to be linear, quadratic, and positive definite.

Uploaded by

michael Ketsela
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

FEM FORMULATION

 
Direct Formulation
Weighted Residual Methods

Example with a single governing equation with only one independent variable
f [T ( x)]  0 in 

T is the function sought , function of x only


 is the domain of the region governed by f
Boundary conditions
g1 [T ( x)]  0 in 1
g 2 [T ( x)]  0 in 2
1 and 2 are parts of the boundary of 
Approximation of the solution with a T ' function:
n
T '  T ' ( x; a1 , a 2 ,, a n )   ai N i ( x)
i 1
Cont--

which has one or more unknown( but constant) parameters a1 , a 2 , a n satisfies exactly
the boundary conditions. No surprise if the approximation does not satisfy the equation
exactly! We will get a residual error:
f [T ' ( x; a1 , a 2 , , a n )]  R( x; a1 , a 2 ,  , a n )
The method of weighted residuals requires that the parameters a1 , a 2 , a n be determined
satisfying:


 Wi ( x)R( x; a1 , a 2 ,, an )d  0
where the functions Wi (x) are the n arbitrary weighting functions
Cont.----

The conditions of the weighting functions is generally left to a personal judgment


The most popular weighted-residual methods are:
1) Point collection
2) Subdomain collection
3) Least squares
4) Galerkin
Point Collection
The weighted functions Wi (x) are  ( x  xi ) and defined such that
b
  ( x  x )dx  1
a
i for x  xi

b
  ( x  x )dx  0
a
i for x  xi
Substitution of this choice of wi (x) gives:

  ( x  x ) R( x; a , a

i 1 2 ,  a n ) d  0 for i  1,2,, n

which is evaluated at n collection points x1 , x 2 ,, x n results n algebraic equations in n


unknowns

R( x1 ; a1 , a 2 ,, a n )  0
R( x 2 ; a1 , a 2 ,, a n )  0

R( x n ; a1 , a 2 , , a n )  0
Subdomain Collection

The weighted functions Wi (x) are:


1 for x in 1
W1 ( x )  
0 for x not in 1

1 for x in  2
W2 ( x )  
0 for x not in  2
Substitution of this choice of wi (x) gives the following n integral equations

 R( x; a , a
1
1 2 ,  a n ) d 1  0

 R( x; a , a
2
1 2 , a n ) d 2  0


 R( x; a , a
n
1 2 ,  a n ) d n  0
Least Squares

The method of least squares requires that the integral I of the square of the residual R be
minimized. That is:
I   [ R( x; a1 , a 2 , , a n )] 2 d be a minimum, or equivalently

I  
     d
2 2
[ R ( x; a1 , a 2 ,  , a n )] d [ R ( x; a1 , a 2 ,  , a n )]
ai ai  
ai
Carrying out the differentiations and simplifying, we have:
R R R
 R ai d   R a2 d    R an d  0
which means that the weighting functions are:
R
 R ai d  0 i  1,2,, n
Galerkin

The weighting functions wi (x) are N i (x)


Therefore for the Galerkin method of the weighted residuals we have:
 N i ( x) R( x; a1 , a2 ,, an )d  0 i  1,2,, n

Remember that our approximation of the T (x) function is:
n
T  T ' ( x; a1 , a 2 , , a n )   ai N i ( x)
i 1
Variational Formulation

A functional is linear
u u
 (u , )   ( a1u  a 2 ) d
x 
x
A functional is quadratic if all terms are of second order, for example

u 2
 

 (a1
x
)  a 2 u 2 ) d
(

For a purely quadratic functional


u 
u  u 
1  
   u   D   d
2  x   x 

  

 D is a symmetric matrix independent of u

u 
 (u )  u 
 
   u   D   d

 x   x 

  

 is positive definite if D is positive definite matrix that is all the value of D are positive
Variational Formulation

A functional is linear
u u
 (u , )   ( a1u  a 2 ) d
x 
x
A functional is quadratic if all terms are of second order, for example

u 2
 

 (a1
x
)  a 2 u 2 ) d
(

For a purely quadratic functional


u 
u  u 
1  
   x
u   D    d
2    x 

   
 D is a symmetric matrix independent of u

u 
 (u )  u 
 
   u   D   d

 x   x 

  

 is positive definite if D is positive definite matrix that is all the value of D are positive

You might also like