Scribd
Upload
39 views

Finite Element Methods

The document discusses the finite element method (FEM) for solving boundary value problems (BVPs). FEM involves: (1) Discretizing the domain into small elements, (2) Approximating the solution within each element using shape functions like linear polynomials, (3) Applying Galerkin's method locally within each element to obtain a system of equations relating the nodal values. This leads to a global system of equations whose solution provides the approximate solution to the BVP over the entire domain. FEM overcomes limitations of Ritz and weighted residual methods for complex domains by using piecewise approximations rather than a single global approximation.

Uploaded by

Chiranjit Sau
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
39 views

Finite Element Methods

The document discusses the finite element method (FEM) for solving boundary value problems (BVPs). FEM involves: (1) Discretizing the domain into small elements, (2) Approximating the solution within each element using shape functions like linear polynomials, (3) Applying Galerkin's method locally within each element to obtain a system of equations relating the nodal values. This leads to a global system of equations whose solution provides the approximate solution to the BVP over the entire domain. FEM overcomes limitations of Ritz and weighted residual methods for complex domains by using piecewise approximations rather than a single global approximation.

Uploaded by

Chiranjit Sau
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

CE 601: Numerical Methods

Lecture 36

Finite Element Methods

Course Coordinator:
Dr. Suresh A. Kartha,
Associate Professor,
Department of Civil Engineering,
IIT Guwahati.
In the last class, we discussed on the approximate methods
to solve a boundary-value problem (say)
d2y
2
Qy F ; Appropriate BCs
dx
They were
Rayleigh-Ritz method based on calculus of variations.
Collocation method based on residuals.
Galerkin weighted residual method.
In Galerkin weighted residual method, the methodology involves:
d2y
Identify the differential equation to be solved (e.g. here Qy F)
dx 2
Approximate the actual solution y ( x) with apprximate soln.
I
y ( x), where y ( x) Ci yi
i 1

yi are trial functions usually linearly independent polynomials


that satisfy the given boundary conditions.
Define the residual functions W j ( x); j 1, 2,3,
galerkin suggested that results will be better if we choose the trial
functions yi ( x) as the weighing functions.
Perform the integration using each weighing function
x2

i.e. W j ( x) R( x) dx 0; j 1, 2,3,
x1

Solve the system of weighted residual integrals for the coefficients


Ci ; (i 1, 2,3, )
Usually collocation method is not that much used.
If the variational functional is known a priori and if we want to solve
in the domain (usually solid mechanics problem), Rayleigh-Ritz method
preferred.
If the differential equation is already specified in the domain,
the Galerkin weighted residual method is preferred (fluid mechanics).
The Finite-Element Method for B.V. problems
The Rayleigh-Ritz method and Galerkin weighted residual method
approximate the solution y ( x) for the entire domain.
As linearly independent trial functions are applied for the whole
domain D( x), the accuracy falters for larger domains.
Or else you might have to use higher degree polynomials as trial
functions.
It is a natural approach to increase the degree of the polynomial
trial functions to increase the accuracy.
However on increase of polynomial degree, complexities may
arrive as the approximation may not work within two desired points.
Recall in the class POLYNOMIAL APPROXIMATIONS, we have
seen that it is better to use piece-wise lower degree polynomials in
smaller domians rather that going for a higher degree polynomial
for the entire domian (splines).
We will use the same philosophy for the approximate
solutions of boundary-value problem while using Rayleigh-Ritz (RR)
or Galerkin Weighted Residual method (GWRM). How?
That is the entire solution domain D( x) is discretised into
small small pieces - called elements - and the RR or GWRM are
applied in each of these elements.
This is the Finite-Element Method.
Note:
If the variational functional of a problem is already known
in advance (solid mechanics) it is better to use Rayleigh-Ritz
FEM.
If the governing differential equation is known in advance
(fluid mechanics) it is better to use Galerkin Weighted Residual
FEM.
Galerkin FEM
We now describe Galerkin FEM to solve the simple linear BV-ODE
d2y
2
Qy F (with appropriate BCs)
dx
The concept involves:
The overall outline involves:

The actual GWRM for the entire domain.


M
y Ci yi ( x) C1 y1 ( x) C2 y2 ( x) CM yM ( x)
i 1
b
W j ( x) R( x) dx 0 I (C j )
a

R ( x) is evaluated by using y( x).


In FEM:

The domain is discretised into discrete nodes (1), (2), (3),...,( I )


Again the domain consists of I -1 elements
(elements can be linear, quadratic etc.)
x(i ) xi 1 xi (element length)
Also total integration, I I (1) I (2) I ( I 1)
Summation of integrals for all the elements 1,2,..., I 1.
xi 1
(i )
i.e. I W ( x) R( x)dx, integration over an element.
xi

On applying this equation for all the weights W j ( x) within the element i ,
we will get a set of equations relating the nodal values within each element.
Domain discretisation
D( x) into I nodes, I 1 elements
x(i ) xi 1 xi
The exact solution y( x) y ( x)
y ( x) Sum of series of local interpolating polynomials y (i ) ( x); i 1, 2,3,.., I 1
These are valid within each element.
I 1
y ( x) y ( i ) ( x)
i 1

Local interpolating polynomial y (i ) ( x) yi N i(i ) ( x) yi 1 N i( i )1 ( x)


where yi Nodal values of y at node i & yi 1 Nodal values of y at node i 1
N i( i ) ( x) & N i( i )1 ( x) Linear interpolating polynomials within element i.
N i( i ) ( x) 1.0 at x xi
N i(i ) ( x) 0.0 at x xi 1

N i(i )1 ( x) 0.0 at x xi
N i(i )1 ( x) 1.0 at x xi 1
x xi 1 x xi
y (i ) ( x) yi y i 1
x(i ) x(i )
Ni are called shape functions.
d2y
To solve the BV-ODE 2
Qy F
dx
d2y
Define R( x) 2
Qy F
dx
b b
d2y
I ( y ( x)) W j ( x) 2
Qy F dx 0 W j ( x)( y " Qy F )dx
a
dx a
b b b
b
W j y " dx y 'W j ' dx y 'W j y 'W j ' dx yb 'W j (b) ya 'W j ( a)
a
a a a

If the Neumann b.c.'s are given: yb ' 0, ya ' 0


b
You have I ( y ( x)) ( y 'W j ') (QyW j FW j ) dx 0
a

Again, I ( y ( x)) I (1) I (2) I (I 1)


yb 'WI (b) ya 'WI (a) 0
xi
(i )
1
dW j
I y' QyW j FW j dx
xi
dx
xi 1(i ) x xi x
Aslo, y ( x) (i )
yi (i )
yi 1
x x
As in Galerkin method, the weighing functions are same as shape
functions N i( i ) ( x) and N i( i )1 ( x).
N i(i ) ( x) 0.0 for x xi 1 &x xi
N i(i )1 ( x) 0.0 for x xi 1 &x xi
xi 1
(i ) d
I y ' ( N i( i ) ( x)) QyN i( i ) FN i( i ) dx 0 ( A)
xi
dx
xi 1
(i ) d
Also, I y ' ( N i( i )1 ( x)) QyN i( i )1 FN i( i )1 dx 0 ( B)
xi
dx
(A) and (B) are element equations.

You might also like