BSC207 / BSC208

Numerical Methods in Engineering

Numerical Integration

1
Learning outcomes
• Implement single application of Newton-Cotes formula
(Trapezoidal rule, Simpson’s 1/3 rule, Simpson’s 3/8 rule)
• Implement composite Newton-Cotes formulas.
• Use the trapezoidal rule to integrate unequally spaced data

2
Integration
• The inverse process to differentiation in calculus is
integration.
• Mathematically, integration is represented by
"
𝐼 = # 𝑓 𝑥 𝑑𝑥
!
• which stands for the integral of the function f(x) with respect
to the independent variable x, evaluated between the limits
x = a to x = b.
• The function 𝑓(𝑥) is referred to as the integrand.

3
Integration

4
Main uses
• Two Main Uses
• Summing things up
• Averaging quantities
• a) Discrete Data Integration
n

åy i
y= i =1
n

• b) Continuous Data Integration

b
ò f ( x ) dx
y = a b-a

5
Integrate analytically
• One way to do the integration is to analytically obtain a
closed form solution
t
z (t ) = ò v(t )dt
0
t
gm æ gcd ö
z (t ) = ò tanh çç × t ÷÷dt
0
cd è m ø
m é æ gcd öù
z (t ) = ln êcosh çç t ÷÷ú
cd êë è m øúû

• What if we encounter a function we cannot integrate

analytically?

6
Newton-Cotes Formulas
• The Newton-Cotes formulas are the most common
numerical integration schemes.
• Generally, they are based on replacing a complicated
function or tabulated data with a polynomial that is easy to
integrate:
" "
𝐼 = # 𝑓 𝑥 𝑑𝑥 ≅ # 𝑓𝑛 𝑥 𝑑𝑥
! !
where fn(x) is an nth order interpolating polynomial.

7
Newton-Cotes Examples
• The integrating function
can be polynomials for
any order—for example,
(a) straight lines or (b)
parabolas.
• The integral can be
approximated in one step
or in a series of steps to
improve accuracy.

8
Numerical integration
• If integration is the same as summing, can we do it
numerically?
• Take this simple function: f(x) = 2

f(x)
2

What is the Integral

between x=a and x=b?

a b x

9
Numerical integration
• What if the function is not flat?
• We can approximate using Trapezoid

f1 f2
Area = x(f1+f2)/2

10
 Trapezoidal rule
• The trapezoidal rule is the first
of the Newton-Cotes closed
integration formulas; it uses a
straight-line approximation for
the function:
"
𝐼 = # 𝑓𝑛 𝑥 𝑑𝑥
!

"
𝑓 𝑏 −𝑓 𝑎
𝐼=# 𝑓 𝑎 + (𝑥 − 𝑎) 𝑑𝑥
! 𝑏−𝑎

𝑓 𝑎 + 𝑓(𝑏)
𝐼 = (𝑏 − 𝑎)
2
11
Trapezoidal rule error
• The large curvature of f(x) between a=0 and b=0.8 results
in a massive error!

12
Example
Integrate numerically
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
from a = 0 to b = 0.8.

13
Example

14
Trapezoidal rule error
• Can we estimate the truncation error if there is no analytical
solution?

• The truncation error estimate of the Trapezoidal Rule is:

1
Et = - f ¢¢(x )(b - a )
3

12
• where a < x < b,
• f ” is the 2nd derivative of the function being integrated.

15
Trapezoidal rule error
• In actual situations, we would have no foreknowledge of the
true value.
• Therefore, an approximate error estimate is required.

1 $$ %
𝐸# = − 𝑓 𝜉 𝑏 − 𝑎
12

𝑓 $$ 𝑥 = −400 + 4050𝑥 − 10800𝑥 & + 8000𝑥 %

We need average value of 𝑓′′(𝜉)

16
Trapezoidal rule error
1 $$ %
𝐸# = − 𝑓 𝜉 𝑏 − 𝑎
12

𝑓 $$ 𝑥 = −400 + 4050𝑥 − 10800𝑥 & + 8000𝑥 %

17
Estimating Truncation Error
• Functions that have quickly changing gradients will have
large Trapezoidal Rule errors

f ( x) = 0.2 + 25 x - 200 x 2 + 675x 3 - 900 x 4 + 400 x 5

f " ( x) = -400 + 4050 x - 10800x 2 + 8000 x 3
Integral is more
f ( x) = 0.2 + 25 x - 200 x + 675x - 900 x
2 3 4
erroneous with high
curvature models
f " ( x) = -400 + 4050 x - 10800x 2

f ( x) = 0.2 + 25 x Linear f(x) has

zero truncation
f " ( x) = 0 error!

18
Numerical integration
• Can we estimate the integral more accurately by using
multiple segments?

• The error formula indicates that the error is dependent upon

the curvature of the actual function as well as the distance
between the points.

• Error can thus be reduced by breaking the curve into parts.

19
Composite Trapezoidal Rule
• With n input data points there are n-1 segments
• The segment width h = (b – a) / (n – 1)
• The total integral can be calculated by integrating each
segment and then adding them together
Assume Data Points
I = ò f ( x ) dx
xn
are evenly spaced at
x1
distance h apart
= ò f ( x ) dx + ò f ( x ) dx +  + ò f ( x ) dx
x2 x3 xn

x1 x2 xn-1

f ( x2 ) + f ( x1 ) f ( x3 ) + f ( x2 ) f ( xn -1 ) + f ( xn )
= ( x2 - x1 ) + ( x3 - x2 ) +  + ( xn - xn -1 )
2 2 2
f ( x2 ) + f ( x1 ) f ( x3 ) + f ( x2 ) f ( xn -1 ) + f ( xn )
=h +h ++ h
2 2 2
hé n -1
ù
= ê f ( x1 ) + 2å f ( xi ) + f ( xn )ú
2ë i =2 û
Composite
Trapezoidal Rule 20
Example
• Use a two-segment Trapezoidal Rule to integrate
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
from a=0 to b=0.8

21
Example
• Use a four-segment Trapezoidal Rule to integrate
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
from a=0 to b=0.8

22
Composite Trapezoidal Rule Example
• Use a eight-segment Trapezoidal Rule to integrate
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
from a=0 to b=0.8

23
Comparison Table
Step
Method Points Segments
Size
Result Et et
Analytical - - - 1.640533 - -
Trapezoidal Rule

Composite
Trapezoidal Rule

Integrate multiple segments reduces the truncation error significantly

Smaller step size leads to better accuracy (smaller truncation error)

24
Dealing with Unequal Segments
• In many engineering situations, the independent data is not
evenly spaced

• This is because in many cases the experimenter has no

control over the independent variables

• Having unequal segments is also useful as it allows finer

integration in parts of the curve that “turns” sharply

25
Composite Trapezoidal Rule with Unequal
Segments
• The previous Composite Trapezoidal equation relied on evenly
spaced data points at distance h apart
hé n -1
ù
I = ê f (x1 ) + 2å f (xi ) + f (xn )ú
2ë i =1 û
• The Composite Trapezoidal Rule still works on unequal
segments, just not as “neat”
Unequal Segment
I = ò f ( x ) dx
xn
Composite
x1
Trapezoidal Rule
= ò f ( x ) dx + ò f ( x ) dx +  + ò f ( x ) dx
x2 x3 xn

x1 x2 xn-1

f ( x2 ) + f ( x1 ) f ( x3 ) + f ( x2 ) f ( xn -1 ) + f ( xn )
= ( x2 - x1 ) + ( x3 - x2 ) +  + ( xn - xn -1 )
2 2 2
f ( x2 ) + f ( x1 ) f ( x3 ) + f ( x2 ) f ( xn -1 ) + f ( xn )
=h +h ++ h Can’t do this
2 2 2
hé n -1
ù anymore!
= ê f ( x1 ) + 2å f ( xi ) + f ( xn )ú
2ë i =2 û

26
Simpson’s Rules
• One drawback of the Trapezoidal Rule is that the truncation
error is related to the second derivative of the function.

• Using 2nd and 3rd degree polynomials help reduce

truncation error

• The formulas that result from taking the integrals under

these polynomials are called Simpson’s rules.

27
Trapezoidal rule
• Trapezoidal Rule (Composite, n=4)

Degree 1: linear lines joining data points together

28
Simpson’s rules

• Simpson’s Rules fit quadratic and cubic curves to multiple

data points
Simpson’s 1/3 Rule Simpson’s 3/8 Rule
(Degree 2) (Degree 3)

Similar to 2-segment Similar to 3-segment

composite trapezoidal composite trapezoidal
rule, but uses parabola rule, but uses x3 curve
curve instead of instead of straight line
straight line

29
Simpson’s 1/3 Rule
• Simpson’s 1/3 rule uses a degree 2 polynomial (x2) that
passes through three points
2 segments
• The integration is performed on this degree 2 polynomial,
which produces the following equation

I = ò f n (x ) dx
x3

x1

x1 x2 x3

30
Simpson’s 1/3 Rule
• The integration is performed on this degree 2 polynomial,
which produces the following equation

I = ò f n ( x ) dx
x3
h=
(b - a )
x1
n -1
h
I = [ f (x1 ) + 4 f ( x2 ) + f ( x3 )]
3 Simpson’s 1/3 Rule

• Another way to write Simpson’s 1/3 Rule is

f (x1 ) + 4 f (x2 ) + f (x3 )
I = (b - a)
6

Width
* Mean Height

31
Simpson’s 1/3 Rule Example
• Consider the polynomial
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
• Integrate it from a=0 to b=0.8 using Simpson’s 1/3 Rule

32
Comparing Simpson’s 1/3 Rule with Trapezoidal Rule

Composite
Trapezoidal
εt = 34.85%

Simpson’s 1/3 Rule

εt = 16.64%

33
Comparison Table
• In general, the Simpson’s 1/3 rule is more accurate than the
Trapezoidal rule with the same n and h

Step
Method Points Segments
Size
Result Et et
Analytical - - - 1.640533 - -
Trapezoidal Rule

Composite
Trapezoidal Rule

Simpson’s 1/3
Rule

34
Simpson’s 1/3 Rule Error
• An estimate for the truncation error of a single application of
Simpson’s 1/3 rule is
1
f (x )(b - a)
(4 ) 5
Et = -
2880
where a < x < b

• The error is dependent upon the fourth-derivative of the

function being integrated and the integration limits
• Cubic (x3) function f(x) can be integrated perfectly with
Simpson’s 1/3 Rule
d d d d

x3 dx
x2 dx
x dx
constant dx
0

35
Comparing Truncation Errors
• The truncation error estimate of the single application of
Trapezoidal Rule is:
1
Et = - f ¢¢(x )(b - a )
3

12
• An estimate for the truncation error of a single application of
Simpson’s 1/3 Rule is:
1
f (4 ) (x )(b - a)
5
Et = -
2880

Much smaller errors

than trapezoidal

36
Composite Simpson’s 1/3 Rule
• Like the Trapezoidal rule, we can use multiple segments to
perform a Simpson’s 1/3 integration
• This improves integration accuracy

• However, there must be an odd number of points (n = 3, 5,

7 … etc)
• In other words, even number of segments

• Because of the heavy weighting of the internal points, the

formula is a little more complicated than the trapezoidal rule

37
Composite Simpson’s 1/3 Rule

h
I= [ f (x1 ) + 4 f (x2 ) + f (x3 )]
3

38
 Composite Simpson’s 1/3 Rule
• The integral is broken up two segments at a time
I = ò f (x ) dx = ò f (x ) dx + ò f (x ) dx +  + ò f (x ) dx
xn x3 x5 xn

x1 x1 x3 x n -2

• Each integral is replaced with Simpson’s 1/3 Rule

h h h
I = [ f x1 + 4 f x2 + f x3 ] + [ f x3 + 4 f x4 + f x5 ] +  + [ f (xn-2 ) + 4 f (xn-1 ) + f (xn )]
( ) ( ) ( ) ( ) ( ) ( )
3 3 3
• There is a common factor of h/3 with the odd and even f(x)
terms overlapping (1,4,2,…4,2,4,1 pattern)
é n -1 n-2
ù Composite
f (x1 ) + 4 å f (xi ) + 2 å f (x j ) + f (xn )ú
hê
I= Simpson 1/3
3ê i = 2 , 4 , 6 ,... j =3, 5, 7 ,...
ú
Rule
êë i , even j , odd úû

39
Composite Simpson’s 1/3 Rule
é n -1 n-2
ù
f (x1 ) + 4 å f (xi ) + 2 å f (x j ) + f (xn )ú
hê
I=
3ê i = 2 , 4 , 6 ,... j =3, 5, 7 ,...
ú
êë i , even j , odd úû

x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11
40
Composite Simpson’s 1/3 Rule Example
• Lets revisit this polynomial
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
• We compared a single application of Simpson’s 1/3 Rule to
a two-segment Composite Trapezoidal Rule

• The single Simpson’s 1/3 Rule integration used two

segments (n=3) but reduced the error of the Composite
Trapezodial rule from 34.85% to 16.64%

• Can we improve this result further by applying the

Simpson’s 1/3 Rule twice (n=5, that is, four segments)

41
Composite Simpson’s 1/3 Rule Example
• Lets revisit the polynomial from a few slides ago
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
• Integrating from a=0 to b=0.8 using TWO applications of
the Simpson’s 1/3 Rule (5 points, 4 segments)
n=5
f(x4)
h = 0 .2 f(x3)
x1 = 0.0
f(x2)
x 2 = 0 .2
f(x5)
x3 = 0.4 f(x1)
x 4 = 0 .6
Simpsons 1/3 Simpsons 1/3
x5 = 0.8

42
Composite Simpson’s 1/3 Rule Example

43
Composite Simpson’s 1/3 Rule Example

Single
Simpson’s 1/3 Rule

Composite
Simpson’s 1/3 Rule
(Two Passes)

44
Comparison Table

Step
Method Points Segments
Size
Result Et et
Analytical - - - 1.640533 - -
Trapezoidal Rule

Composite
Trapezoidal Rule

Simpson’s 1/3
Rule
Composite
Simpson’s 1/3
Rule

45
Simpson’s 3/8 Rule
• Simpson’s 3/8 rule corresponds to using degree 3 polynomials
(x3) that pass through four points
I = ò f n ( x ) dx
x4
Simpson’s 3/8
x1
Rule
3h
I = [ f ( x1 ) + 3 f ( x2 ) + 3 f ( x3 ) + f ( x4 )]
8
h=
(b - a)
where n -1

• Generally, Simpson’s 3/8 rule

is used with the 1/3 rule when
there is an even number of
points (n=4,6,8 … etc)

46
Simpson’s 3/8 Rule Error
• The Simpson’s 3/8 Rule truncation error can be estimate
using
1
Et = - f ( 4 ) (x )(b - a ) 5
6480

• As the denominator is larger than the error equation for

Simpson’s 1/3 Rule, the 3/8 rule is a little bit more accurate

• However, as the 1/3 rule only requires 3 points, it is more

commonly used
• The 3/8 Rule is usually applied once in combination with
the 1/3 Rule when dealing with an even number of
points (n = 4,6,8 etc)

47
Simpson’s 3/8 Rule Example
• Reusing the same polynomial as before
f ( x) = 0.2 + 25 x - 200x + 675x - 900x + 400x
2 3 4 5

• Again, integrating from a=0 to b=0.8 but using Simpson’s

3/8 Rule instead of Simpson’s 1/3 rule
• For a single application of the 3/8 Rule, we need to have 4
evenly spaced points (n = 4, h = 0.8/3)
• Evaluating f(x) at these points give
f (0) = 0.2 f (0.2667) = 1.432724
f (0.5333) = 3.487177 f (0.8) = 0.232

48
Simpson’s 3/8 Rule Example

49
Comparison Table
• As expected, the Simpson’s 3/8 rule result is more accurate
than Simpson’s 1/3 but not better than the four-segment
(n=5) Composite Simpson’s 1/3 result
Step
Method Points Segments
Size
Result Et et
Analytical - - - 1.640533 - -
Trapezoidal Rule

Composite
Trapezoidal Rule

Simpson’s 1/3 Rule

Composite
Simpson’s 1/3 Rule
Simpson’s 3/8

50
Combining Simpson’s 1/3 and 3/8 Rule

• What if we want to integrate f(x)

over 5 segments, n=6?
• Simpson’s 1/3: n is not odd!
• Composite Simpson’s 1/3: 1
missing segment!
• Simpson’s 3/8: 2 extra
segments!
• Composite Simpson’s 3/8: 2
missing segments!
• We can combine the Simpson’s
1/3 AND 3/8 rules to integrate
f(x) from a=0 to b=0.8

51
Combining Simpson’s 1/3 and 3/8 Rule
• Since n = 6, h = (0.8 - 0) / 5 = 0.16, which gives the
following function values
f (0) = 0.2 f (0.16) = 1.296919
f (0.32) = 1.743393 f (0.48) = 3.186015
f (0.64) = 3.181929 f (0.8) = 0.232

52
Combining Simpson’s 1/3 and 3/8 Rule

53
Comparison Table
Step
Method Points Segments
Size
Result Et et
Analytical - - - 1.640533 - -
Trapezoidal Rule

Composite
Trapezoidal Rule

Simpson’s 1/3 Rule

Composite
Simpson’s 1/3 Rule
Simpson’s 3/8
A Simpson’s 1/3
+
A Simpson’s 3/8

54
Romberg Integration
• Technique designed to attain efficient numerical integrals of
functions.
• Based on successive application of the trapezoidal rule.
• An error correction technique similar to Richardson’s
Extrapolation for numerical differentiation.
• The formula for Romberg Integration is
4 1
𝐼 ≅ 𝐼 ℎ& − 𝐼(ℎ')
3 3
where ℎ = (𝑏 − 𝑎)/(𝑛 − 1)

55
Example
Evaluate the integral of
f ( x) = 0.2 + 25 x - 200x 2 + 675x3 - 900x 4 + 400x5
from a = 0 to b = 0.8.

56
Multiple Integrals
• Multiple integrals can be determined numerically
by first integrating in one dimension, then a
second, and so on for all dimensions of the
problem.

57
Example
• Suppose that the temperature of a rectangular heated plate
is described by the following function:
𝑇 𝑥, 𝑦 = 2𝑥𝑦 + 2𝑥 − 𝑥 & − 2𝑦 & + 72
• If the plate is 8 m long (x-dimension) and 6 m wide (y-
dimension), compute the average temperature of the plate.

58
Example

59