MANE 4240 & CIVL 4240

Introduction to Finite Elements

Prof. Suvranu De

Numerical Integration in 1D
Reading assignment:

Lecture notes, Logan 10.4

Summary:

• Newton-Cotes Integration Schemes

•Gaussian quadrature
Axially loaded elastic bar
y A(x) = cross section at x
b(x) = body force distribution
F (force per unit length)
x E(x) = Young’s modulus
x
x=0 x=L
x1 x2
For each element 1 2
Element stiffness matrix
x2 x2
kij   Bi EB j Adx d1x
d2x
k   B EB Adx
T
x1 x1

dN i ( x)
where Bi 
dx
Only for a linear finite element


x2

x1
T
B EB Adx 
1
 x 2  x1 
2
 1 1 x2
 1 1  x1 AEdx 
 

x1
x2
AEdx  x x 
2
1
1
2
 1 1
 1 1 
 

Element nodal load vector

x2
f b i   N i b dx
x2
f b   N b dx
T
x1 x1

Question: How do we compute these integrals using a computer?

Any integral from x1 to x2 can be transformed to the following
integral on (-1, 1)
1
I   f (r ) dr
1

Use the following change of variables

1 r 1 r
x x1  x2
2 2
Goal: Obtain a good approximate value of this integral
1. Newton-Cotes Schemes (trapezoidal rule, Simpson’s rule, etc)
2. Gauss Integration Schemes

NOTE: Integration schemes in 1D are referred to as “quadrature

rules”
Trapezoidal rule: Approximate the function fr by a straight line
g(r)that passes through the end points and integrate the straight
line
gr f

f fr

r
-1 1
1 r 1 r
g (r )  f (1)  f (1)
2 2
1 1
I   f (r ) dr   g (r ) dr  f (1)  f (1)
1 1
•Requires the function f(x) to be evaluated at 2 points
(-1, 1)
• Constants and linear functions are exactly integrated
• Not good for quadratic and higher order polynomials

How can I make this better?

 Simpson’s rule: Approximate the function f(r) by a parabola g(r)
that passes through the end points and through f(0) and integrate
the parabola f
fr gr

f
f(
r
-1 1
r (r  1) r (1  r )
g (r )  f (1)  (1  r )(1  r ) f (0)  f (1)
2 2
1 1 1 4 1
I   f (r ) dr   g (r ) dr  f (1)  f (0)  f (1)
1 1 3 3 3
•Requires the function f(x) to be evaluated at 3 points
(-1,0, 1)
• Constants, linear functions and parabolas are exactly
integrated
• Not good for cubic and higher order polynomials

How to generalize this formula?

Notice that both the integration formulas had the general form
M
I   f (r ) dr   Wi f (ri )
1

1
i 1

Weight Integration point

Trapezoidal rule: Accurate for polynomial of
W1  1 r1  1
M=2 degree at most 1 (=M-1)
W2  1 r2  1

Simpson’s rule: W1  1 / 3 r1  1 Accurate for polynomial of

M=3 W2  4 / 3 r2  0 degree at most 2 (=M-1)
W2  1 / 3 r2  1
Generalization of these two integration rules: Newton-Cotes
• Divide the interval (-1,1) into M-1 equal intervals using M points
• Pass a polynomial of degree M-1 through these M points (the
value of this polynomial will be equal to the value of the function
at these M-1 points)
• Integrate this polynomial to obtain an approximate value of the
integral
f
fr
f
gr
r
-1 1
With ‘M’ points we may integrate a polynomial of degree ‘M-1’
exactly.

Is this the best we can do ?

With ‘M’ integration points and ‘M’ weights, I should be able to

integrate a polynomial of degree 2M-1 exactly!!
Gauss integration rule

See table 10-1 (p 405) of Logan

 Gauss quadrature
M
I   f (r ) dr   Wi f (ri )
1

1
i 1

Weight Integration point

How can we choose the integration points and weights to exactly

integrate a polynomial of degree 2M-1?

Remember that now we do not know, a priori, the location of the

integration points.
 Example: M=1 (Midpoint qudrature)
1
I   f (r ) dr  W1 f (r1 )
1

How can we choose W1 and r1 so that we may integrate a

(2M-1=1) linear polynomial exactly?
f (r )  a0  a1r
1
1
f (r ) dr  2a0

But we want
1
1
f (r ) dr  W1 f (r1 )  a0W1  a0W1r1
Hence, we obtain the identity

2a0  a0W1  a1W1r1

For this to hold for arbitrary a0 and a1 we need to satisfy 2
conditions
Condition 1 : W1  2
Condition2 : W1r1  0

i.e., W1  2; r1  0
 1
For M=1 I   f (r ) dr  2 f (0)
1

fr
f
gr

r
-1 1
Midpoint quadrature rule:
• Only one evaluation of fris required at the midpoint of the
interval.
• Scheme is accurate for constants and linear polynomials
(compare with Trapezoidal rule)
 Example: M=2
1
I   f (r ) dr  W1 f (r1 )  W2 f (r2 )
1

How can we choose W1 ,W2 r1 and r2 so that we may integrate a

polynomial of degree (2M-1=4-1=3) exactly?
f (r )  a0  a1r  a2 r 2  a3r 3
1 2
1 f (r ) dr  2a0  3 a2
But we want
1
 1
f (r ) dr  W1 f (r1 )  W2 f (r2 )

 2 2
 
 a0 W1  W2   a1 W1r1  W2 r2   a2 W1r1  W2 r2  a3 W1r1  W2 r2
3 3

Hence, we obtain the 4 conditions to determine the 4 unknowns
(W1 ,W2 r1 and r2 )
Condition 1 : W1  W2  2
Condition 2 : W1r1  W2 r2  0
2 2
2
Condition 3 : W1r1  W2 r2 
3
3 3
Condition 4 : W1r1  W2 r2  0

Check that the following is the solution

W1  W2  1
1 1
r1   ; r2 
3 3
 1 1 1
For M=2 I   f (r ) dr  f ( )  f ( )
1
3 3
1 1
f ( ) f( )
3 3
fr

* * r
-1 1

• Only two evaluations of fris required.

• Scheme is accurate for polynomials of degree at most 3
(compare with Simpson’s rule)
 Exercise: Derive the 6 conditions required to find the
integration points and weights for a 3-point Gauss quadrature
rule

Newton-Cotes Gauss quadrature

1. ‘M’ integration points are 1. ‘M’ integration points are
necessary to exactly integrate necessary to exactly integrate a
a polynomial of degree ‘M-1’ polynomial of degree ‘2M-1’
2. More expensive 2. Less expensive
3. Exponential convergence,
error proportional to  1  2 M
 
 2 M 
Example
1
I   f (r )dr where f (r )  r 3  r 2
1
Exact integration
2
I Integrate and check!
3
Newton-Cotes
To exactly integrate this I need a 4-point Newton-Cotes
formula. Why?
Gauss
To exactly integrate this I need a 2-point Gauss formula.
Why?
Gauss quadrature:
1
I  f (r )dr
1

1 1
 f ( ) f ( )
3 3
2
 Exact answer!
3
Comparison of Gauss quadrature and Newton-Cotes for the
integral 1
I   cos(2x) dx
1

Newton-Cotes

Gauss quadrature
In FEM we ALWAYS use Gauss quadrature
Linear Element 2
1

r  1 r 1

Stiffness matrix
r 1
1  1  1 1
 1
 1  1  1
k   B EBAdr  
T
  AEdr   AEdr   
r  1
4   1 1  1  1  4   1 1 
Nodal load vector
1
f b i   N i b dr
1
f b   N b dr
T
1 1

Usually a 2-point Gauss integration is used. Note that if A, E and b

are complex functions of x, they will not be accurately integrated
 Quadratic Element

Nodal shape functions 1 2 3

r (r  1) r  1 r 0 r 1
N1 ( r ) 
2
N 2 (r )  (1  r 2 ) You should be able to derive these!
r (r  1)
N 3 (r ) 
2
Stiffness matrix
r 1 r 1

 B EBAdr  AE  B Bdr Assuming E and A are constants

T T
k
r  1 r  1

dN  dN1 dN 2 dN 3   (2r  1) (2r  1) 

B     2r
dr  dr dr dr   2 2 
 r 1 r 1

 B EBAdr  AE  B Bdr
T T
k
r  1 r  1

1
 (r  1/2) 2
 2r(r  1/2) (r 2
 1/4) 
 
 AE   2r(r  1/2) 4r 2  2r(r  1/2)dr
-1  2 
 (r 2
 1/4)  2r(r  1/2) (r  1/2) 

Need to exactly integrate quadratic terms.

Hence we need a 2-point Gauss quadrature scheme..Why?