CSC475: Numerical Computing

Tanveer Ahmed Siddiqui

Department of Computer Science

COMSATS University, Islamabad
 Lecture # 26

Numerical Integration

Department of Computer Science 2

 Recap

Department of Computer Science 4

 Today Covered
,

After completing this lecture, you should be able

to know
 Motivation/Why Numerical Integration
 Relative Error for Numerical Integration
 Rectangle and midpoint methods
 Newton-Cotes Closed Quadrature Formula
 Trapezoidal Rule
 Error Estimate in Trapezoidal Rule

Department of Computer Science

Department of Computer Science 6
 What is Integral?
,

 Integration, in mathematics, technique of

finding a function g(x) the derivative of
which, Dg(x), is equal to a given function f(x).
 This is indicated by the integral sign “∫,” as in
∫f(x), usually called the indefinite integral of the
function.
 The symbol dx represents
an infinitesimal displacement along x; thus
∫f(x)dx is the summation of the product of f(x)
and dx.

Department of Computer Science

 What is Integral?
,

 Finding an Integral is the reverse of

finding a Derivative

Department of Computer Science

 What is Integral?
,

The definite integral, written

with a and b called the limits of integration,
is equal to g(b) − g(a), where Dg(x)
= f(x).

Department of Computer Science

 What is Integral?
,

 A definite integral assigns numbers to

functions in a way that can describe
displacement, area, volume, and
other concepts that arise by
combining infinitesimal data.
 But it is easiest to start with finding
the area under the curve of a
function like this: b
Area   f(x)dx
a
Integral = area under the curve

Department of Computer Science

 Integration or anti derivative

Indefinite Integrals Definite Integrals

Indefinite Integrals of a Definite Integrals are

function are functions numbers.
that differ from each
other by a constant.
1 2 1
x 1
x2 0 xdx  2 
2
 x dx  2  c 0

Department of Computer Science 11

 Warmup

Department of Computer Science 14

 The need for numerical integration

Department of Computer Science 15

 Challenge yourself

Department of Computer Science 16

 Motivation

 Let a train start from rest with a velocity as given

in the following table for time t
Time(t) 0 10 30 40 55 60
Velocity(V) 0 20 35 43 48 52

 Can we find the distance travelled in 60

seconds?
 Yes(How)
 Use Numerical Integration

Department of Computer Science 17

 What is numerical integration?

Department of Computer Science 18

 Numerical Integration

 Numerical integration is the approximate

computation of an integral using numerical
techniques.
 The numerical computation of an integral is
sometimes called quadrature.
 Ueberhuber (1997, p. 71) uses the word "quadrature"
to mean numerical computation of a
univariate integral, and "cubature" to mean numerical
computation of a multiple integral.

Department of Computer Science 19

 Approaches to Numerical Integration

Department of Computer Science 20

 Approaches to numerical Integration

 In numerical integration, the range of integration

(b – a) is divided into a finite number of
intervals.
 Various methods have been developed for
carrying out numerical integration.
 In each of these methods, a formula is derived
for calculating an approximate value of the
integral from discrete values of the integrand.

Department of Computer Science 21

 Approaches to numerical Integration

 The methods can be divided into groups:

 Open methods and closed methods.
 Newton-Cotes integration formulas
 Gaussian Quadrature formulas

Department of Computer Science 22

 Closed methods

 In closed integration methods, the endpoints of

the interval (and the integrand) are used in the
formula that estimates the value of the integral.

 The trapezoidal and Simpson's methods are

closed methods

Department of Computer Science 23

 Open methods

 In open integration methods, the interval of

integration extends beyond the range of the
endpoints that are actually used for calculating
the value of the integral.

 The midpoint method and Gauss quadrature are

open methods.

Department of Computer Science 24

 Open and Closed methods

 Difference between closed and open methods

Department of Computer Science 25

 Newton-Cotes Methods

 In Newton-Cote Methods, the function is

approximated by a polynomial of order n.
 Computing the integral of a polynomial is easy.

a
b
f ( x)dx  
a
b
a
0  a1 x  ...  a n x 
n
dx
b (b 2  a 2 ) (b n 1  a n 1 )
a
f ( x)dx a0 (b  a)  a1
2
 ...  an
n 1

Department of Computer Science 26

 Newton-Cotes Methods

 When the original integrand is an analytical

function, the Newton-Cotes formula replaces it
with a simpler function.
 When the original integrand is in the form of
data points, the Newton-Cotes formula
interpolates the integrand between the given
points.
 Most commonly, as with the trapezoidal method
and Simpson's methods, the Newton-Cotes
integration formulas are polynomials of different
degrees.
Department of Computer Science 27
 Other Methods

Department of Computer Science 28

 Other Methods

 Then, the integral is evaluated by

direct analytical methods from calculus.
 This procedure requires numerical methods for
finding F(x) (Chapters 6 and 8), but may not
require a numerical method to evaluate the
integral if F(x) is an integrable function.

Department of Computer Science 29

 Relative Error

Department of Computer Science 30

 Relative Error

Department of Computer Science 31

 How we Compute definite Integration

 A simple intuitive approach is to plot the function

on a grid (Fig. PT6.5) and count the number of
boxes that approximate

Department of Computer Science 32

 How we Compute definite Integration

 A simple intuitive approach is to plot the function on a

grid (Fig. PT6.5) and count the number of boxes that
approximate the area.
 This number multiplied by the area of each box provides
a rough estimate of the total area under the curve.
 This estimate can be refined, at the expense of
additional effort, by using a finer grid.

Department of Computer Science 33

 Why Numerical methods for Integration

 Another commonsense approach is to divide the

area into vertical segments, or strips, with a
height equal to the function value at the
midpoint of each strip (Fig. PT6.6). The area of
the rectangles can then be calculated and
summed to estimate the total area.

Department of Computer Science 34

 Why Numerical methods for Integration

 In this approach, it is assumed that the value at

the midpoint provides a valid approximation of
the average height of the function for each strip.
 As with the grid method, refined estimates are
possible by using more (and thinner) strips to
approximate the integral.

Department of Computer Science 35

 Numerical Integration formulas

Department of Computer Science 36

 Numerical Integration formulas

Department of Computer Science 37

 Numerical Integration formulas

Department of Computer Science 38

 Numerical Integration formulas

Department of Computer Science 39

 Numerical Integration formulas

 three straight-line segments are used to approximate the

integral. Higher-order polynomials can be utilized for the
same purpose.

Department of Computer Science 40

 TRAPEZOIDAL RULE

 TRAPEZOIDAL RULE
 When n=1

Department of Computer Science 41

 TRAPEZOIDAL RULE

 When n=2

Department of Computer Science 42

 TRAPEZOIDAL RULE

 When n=4

Department of Computer Science 43

 TRAPEZOIDAL RULE

 generalization

Department of Computer Science 44

 Numerical Integration formulas

 TRAPEZOIDAL RULE

Department of Computer Science 45

 Example

 Solution

Department of Computer Science 46

 Your Turn

 TRAPEZOIDAL RULE

Department of Computer Science 47

 Your Turn

 TRAPEZOIDAL RULE

Department of Computer Science 48

 SIMPSON’S 1/3 RULE
 SIMPSON’S RULE
 It is most commonly formula for the
approximation integration.
 Similar to trapezoidal rule (trapezoidal method relies on
approximating the integrand by a straight line.)

 Divide required area into n number of strips with

equal width (even n)
 Gives better approximation then trapezoidal rule
 SIMPSON’S 1/3 RULE (uses quadratic polynomial to
approximate integrand)

 SIMPSON’S 3/8 RULE (uses cubic polynomial to

approximate integrand)
Department of Computer Science 49
 SIMPSON’S 1/3 RULE

 A better approximation can possibly be obtained

by approximating the integrand with a nonlinear
function that can be easily integrated
 The coefficients of a quadratic polynomial can be
determined from three points

Department of Computer Science 50

 Numerical Integration formulas

 SIMPSON’S 1/3 RULE

 if two conditions are satisfied:

 • The subintervals must be equally spaced.
 • The number of subintervals within [a, b] must be an
even number.

Department of Computer Science 51

 Example

 SOLUTION

Department of Computer Science 52

 Your Turn

 SIMPSON’S 1/3 RULE

Department of Computer Science 53

 Your Turn

 SIMPSON’S 1/3 RULE

Department of Computer Science 54

 Numerical Integration formulas

 SIMPSON’S 3/8 RULE

3

Department of Computer Science 55

 Example

 SOLUTION

Department of Computer Science 56

 Your Turn

 SIMPSON’S 3/8 RULE

3

Department of Computer Science 57

 Your Turn


 SIMPSON’S 3/8 RULE
3

Department of Computer Science 58

 Numerical Integration formulas

 BOOLE’S RULE

Department of Computer Science 59

 Example

 SOLUTION

Department of Computer Science 60

 Your Turn

 BOOLE’S RULE

Department of Computer Science 61

 Numerical Integration formulas

 WEDDLE’S RULE

Department of Computer Science 62

 Example

 SOLUTION

Department of Computer Science 63

 Your Turn

 WEDDLE’S RULE

Department of Computer Science 64

 Recap: Integration methods so far
,

 Approximate f(x) with a polynomial that

could be easily integrated.
 One point: rectangular and midpoint methods
 Two points: Trapezoidal method
 More points: Simpson’s method

 Use equally spaced points

Department of Computer Science

 Error Bound
,

 It is often helpful to be able to determine an

upper bound for the error in an approximation
of an integral.
 Error bound for Midpoint and Trapezoidal Rule

 We can use these bounds to determine the

value of n necessary to guarantee that the error
in an estimate is less than a specified value.
Department of Computer Science
 Example
,

Department of Computer Science

 Example
,

Department of Computer Science

 Example
,

Department of Computer Science

 Up-Shot
,

Department of Computer Science

Order of accuracy of Trapezoidal and Simpsons Rule
,

 The trapezoidal rule of integration of a

definite integral is of O(h2), while that of
Simpson’s 1/3 and 3/8 rules are of fourth-
order accurate.

Department of Computer Science

 Warm-up
,

 Find the value of for the following

tabular values by taking h = 0.1 ,0.2, 0.4,
0.8 using Trapezoidal Rule

Department of Computer Science

 How to improve accuracy using Trapezoidal Rule
,

 How can we get a better estimate of

integral?
 We can get a better estimate by combining
two Trapezoid estimates with intervals h and
h/2
 How we combine these two estimates?
 By successive application of Richardson’s
extrapolation
 To obtain a more accurate estimate of the value of an
integral from two less accurate estimates
 For example, two estimates with an error O(h2) can be used
for calculating an estimate with an error O(h4)
Department of Computer Science
 Romberg's Integration

Successive application of
Richardson's Extrapolation

Department of Computer Science 74

 Romberg Integration
,

Department of Computer Science

 Romberg Integration Algorithm
,

Department of Computer Science

 When do we stop?

STOP if
I (n, n)  I (n, n  1)  
or
After a given number of steps,
for example, STOP at I(4,4)

Department of Computer Science 77

 Romberg Integration
,

The computations are continued until the successive values are close to each
other.
The general extrapolation formula used in this scheme is

Department of Computer Science

 Up-Shot: Romberg Method

Estimates using Trapezoid method intervals of size h, h/2, h/4, h/8 ...


b
are combined to improve the approximation of f(x) dx
a

First column is obtained R(0,0)

using Trapezoid Method R(1,0) R(1,1)

R(2,0) R(2,1) R(2,2)

R(3,0) R(3,1) R(3,2) R(3,3)

The other elements
are obtained using
the Romberg Method
Department of Computer Science 79
 Example

Department of Computer Science 80

 Example

Department of Computer Science 81

 Example

Department of Computer Science 82

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature

Department of Computer Science 84

 Gauss Quadrature: Motivation

Trapezoid M ethod :
 n 1 

 
b
 
f ( xi )   f ( x0 )  f ( xn ) 
1
f ( x )dx  h 
2
a  i 1 
It can be expressed as :


n


b
f ( x )dx  ci f ( xi )
a
i 0
 h i  1,2,..., n  1
where ci  
 0.5 h i  0 and n
Department of Computer Science 85
 General Integration Formula


n


b
f ( x )dx  ci f ( xi )
a
i 0
ci : Weights xi : Nodes
Problem :
How do we select ci and xi so that the formula
gives a good approximation of the integral?

Department of Computer Science 86

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature
,

Department of Computer Science

 Gauss Quadrature
,

Department of Computer Science

 Double Integration

Department of Computer Science 93

 Double Integration
,

 An integral of the

d b

  g(x, y) dx dy = VOLUME
c a

Department of Computer Science

Department of Computer Science 95
 Double Integration
,

 How we evaluate double integral using

Trapezoidal Rule or Simpson rule?
 We employ either trapezoidal rule or
Simpson’s rule, repeatedly with respect to
one variable at a time over a rectangular
region bounded by the lines x = a, x =b, y
= c, y = d d b

y=d
  g(x, y) dx dy
c a

y=c

x=a x=b
Department of Computer Science
 Double Integration
,

 An integral of the
d b

  g(x, y) dx dy = VOLUME
c a

 How we evaluate double integral using

Trapezoidal Rule or Simpson rule?
 We employ either trapezoidal rule or
Simpson’s rule, repeatedly With respect to
one variable at a time over a rectangular
region bounded by the lines x = a, x =b, y
= c, y = d
Department of Computer Science
 Double Integration

Department of Computer Science 98

 Example

Department of Computer Science 99

 Example

Department of Computer Science 100

 Example

Department of Computer Science 101