NUMERICAL INTEGRATION

b
▪ It uses numerical procedures to evaluate  a
f ( x ) dx approximately if F could not be
expressed in terms of elementary functions such as polynomial, rational or
trigonometric functions.

A. The Trapezoidal Rule

For the trapezoidal element, its area Ak is

y
f ( xk − 1 ) + f ( xk )
Ak = xk
2
f(xk )
f(xk – 1)
xk x
xk – 1 xk

In general, for a regular partition of an interval [a, b] on which a function f is continuous, the
so-called Trapezoidal Rule is given by

b f ( x0 ) + f ( x 1 ) f ( x 1 ) + f ( x2 ) f ( xn− 1 ) + f ( xn )

a
f ( x ) dx  x
2
+ x
2
+ + x
2

b−a
Since x = , then
n

b b-a
 a
f ( x ) dx   f ( x 0 ) + 2f ( x 1 ) + 2f ( x 2 ) + … + 2f ( xn-1 ) + f ( xn ) 
2n 

where:
x0 = a, xn = b and xk = a + kx, k = 0, 1, …, n

Examples
6
1. Use the trapezoidal rule with n = 4 to find an approximation to  2
x 2 − 2 dx .

Solution:
Here, f ( x ) = x2 − 2

If we use n = 4: x = (6 – 2)/4 = 1

x0 = 2 f ( x0 ) = 22 − 2 = 2

x1 = 3 f ( x1 ) = 32 − 2 = 7

x2 = 4 f ( x 2 ) = 4 2 − 2 = 12

x3 = 5 ( )
f x 3 = 5 2 − 2 = 23

x4 = 6 f ( x4 ) = 6 2 − 2 = 34

Consequently,

( ) ( )
6 b-a 

2
x 2 − 2 dx 
2n 
f ( x0 ) + 2 f ( x 1 ) + 2 f ( x 2 ) + 2 f x 3 + f x 4 

 6 6−2
 2
x 2 − 2 dx   2 + 2 7 + 2 12 + 2 23 +
2(4) 
34 

 14.81

2 dx
2. Approximate 
1 x
by the Trapezoidal rule for n = 1, n = 2 and n = 6.

Solution:
1
f (x) =
x

n = 1: x = (2 – 1)/1 = 1

x0 = 1 f(x0) = 1/1 = 1
x1 = 2 f(x1) = 1/2

b−a 2−1 1
= =
2n 2(1) 2

2 dx 1 1
 1
  f ( x0 ) + f ( x 1 )  = ( 1 + 0.5 ) = 0.75
x 2 2

n = 2: x = (2 – 1)/2 = 0.5

x0 = 1 f(x0) = 1
x1 = 1 + ½ = 3/2 f(x1) = 2/3
x2 = 3/2 + ½ = 2 f(x2) = 1/2

b−a 2−1 1
= =
2n 2(2) 4

2 dx 1
 1 x

4
 f ( x 0 ) + 2 f ( x 1 ) + f ( x 2 ) 

1
=  1 + 2 ( 2 / 3 ) + 0.5  = 0.7083
4

n = 6: x = (2 – 1)/6 = 1/6

x0 = 1 f(x0) = 1
x1 = 1 = 1/6 = 7/6 f(x1) = 6/7
x2 = 7/6 + 1/6 = 4/3 f(x2) = 3/4
x3 = 4/3 + 1/6 = 3/2 f(x3) = 2/3
x4 = 3/2 + 1/6 = 5/3 f(x4) = 3/5
x5 = 5/3 + 1/6 =11/6 f(x5) = 6/11
x6 = 2 f(x6) = 1/2

b−a 2−1 1
= =
2n 2(6) 12
 ( ) ( ) ( )
2 dx 1 
 1

x 12 
f ( x0 ) + 2 f ( x 1 ) + 2 f ( x 2 ) + 2 f x 3 + 2 f x 4 + 2 f x5 + f ( x6 ) 

1
  1 + 2 ( 2 / 6 ) + 2 ( 3 / 4 ) + 2 ( 2 / 3 ) + 2 ( 3 / 5 ) + 2 ( 6 / 11 ) + 0.5 
12
 0.6949

Error for Trapezoidal Rule

b
Suppose I =  a
f ( x ) dx and Tn is an approximation to I using n trapezoids. We define

the error in the method to be En = I − Tn .

Theorem: Upper Bound for the Error

If there exists a number M > 0 such that f " ( x )  M for all x in [a, b], then

M (b − a)
3
En 
12n2

Note:
1. En is inversely proportional to n2. Thus, if the number of trapezoids is doubled, the error
E2n is less than ¼ the error bound for En.
2. The formula can be utilized to determine the number of trapezoids that will yield a
specified accuracy.

Examples
2 dx
1. Determine a value of n that give approximation to  1 x
accurate to two decimal places.

Solution:
1 2
f (x) = f '' ( x ) =
x x3

Since f'' decreases on [1, 2], it follows that f''(x)  f''(1) = 2. Thus,
M = 2, (b – a) = 2 – 1 = 1

Therefore,
M (b − a)
3
En 
12n2
2(1)
3
0.005 
12n2
12 ( 0.005 ) 1

2 n2
1
0.03 
n2
100
n 6
3

1
2. Approximate 1/2 cos x dx by the Trapezoidal rule so that the error is less than 0.001.

Solution:
 f ( x ) = cos x
1  sin x 
f '' ( x ) =  − cos x 
4x  x 


For x in the interval [1/2, 1]:

sin x
0 1
x
0  cos x  1

Consequently,
1
f '' ( x ) 
4x
1
f '' ( x ) 
2

Thus,
M = ½, (b – a) = ½
M (b − a)
3
En 
12n2
3
11
 
22
0.001 
12n2
125
n2   5.21
24
n  2.282

Use n = 3 (odd)
n = 3: x = (1 – ½)/3 = 1/6

x0 = ½ f(x0) = cos 1/2

x1 = ½ + 1/6 = 2/3 f(x1) = cos 2/3
x2 = 4/6 + 1/6 = 5/6 f(x2) = cos 5/6
x3 = 4/6 + 1/6 = 1 f(x3) = cos 1

1
1 1−
 cos x dx  2  cos 1 / 2 + 2 cos 2 / 3 + 2 cos 5 / 6 + cos 1 
1/2 2(3)  
 0.3244

B. Simpson’s Rule
▪ Numerical integration using area under a parabolic arc connected by three points

To begin, let us find the area under an arc of a parabola that passes through three points P 0(x0,
y0), P1(x1, y1) and P2(x2, y2), where x0 < x1 < x2 and x1 – x0 = x2 – x1 = h. As shown in the figure, this
can be done by finding the area under the graph of y = Ax2 + Bx + C on the interval [-h, h] so
that P0, P1 and P2 have coordinates (-h, y0), (0, y1) and (h, y2), respectively.
 y
P1(0, y1)
P0(-h, y0) P2(h, y2)

x
-h h

The interval [-h, h] is chosen for simplicity; the area in question does not depend on the location
of the y-axis. Clearly,

−h ( Ax )
h
2
Area = + Bx + C dx
h
 x3 x2 
=A +B + Cx 
 3 2 
  −h

=
h
3
(
2 Ah 2 + 6C )
But, since the graph is to pass through (-h, y0), (0, y1) and (h, y2), we must have
y0 = Ah2 – Bh + C (a)
y1 = C (b)
y2 = Ah2 + Bh – C (c)

Solving equations (a), (b) and (c) for A, B and C, we find 2Ah2 = y0 + y2 – 2y1. Thus,

h
Area = ( y0 + 4 y 1 + y 2 )
3

Now, let us suppose y = f(x) is continuous on [a, b] and the interval is partitioned into n
subintervals of equal width x = (b – a)/n, where n is an even integer. On each interval [xk – 2, xk]
of width 2x, we approximate the graph of f by an arc of a parabola through points Pk – 2, Pk – 1,
and Pk on the graph that corresponds to the end points and midpoint of the subinterval. If Ak
denotes the area under the parabola on [xk – 2, xk], then

x
Ak =  f ( xk − 2 ) + 4 f ( xk − 1 ) + f ( xk ) 
3 

Thus, Simpson’s rule consists of summing all the Ak:

x x 
( ) ( )
b

a
f ( x ) dx 
3
 f ( x 0 ) + 4 f ( x 1 ) + f ( x 2 )  +
3 
f ( x2 ) + 4 f x3 + f x4  +

x
+  f ( x n − 2 ) + 4 f ( x n − 1 ) + f ( x n ) 
3 

b−a
Using = x
n

( )
b b-a
a f ( x ) dx  3n 
f ( x 0 ) + 4f ( x 1 ) + f ( x 2 ) + 4f x 3 + … + 2f ( x n-2 ) + 4f ( x n-1 ) + f ( x n ) 

Example:
2 dx
Approximate 
1 x
by Simpson’s rule for n = 4.

Solution:
n = 4: x = (2 – 1)/4 = 1/4 = 0.25

x0 = 1 f(x0) = 1/1 = 1
x1 = 1.25 f(x1) = 1/1.25 = 0.8
x2 = 1.5 f(x2) = 1/1.5 = 0.67
x3 = 1.75 f(x3) = 1/1.75 = 0.57
x4 = 2 f(x4) = 1/2 = 0.5

b−a 2−1 1
= =
3n 3(4) 12

( ) ( )
b b-a 
a f ( x ) dx  3n 
f ( x0 ) + 4 f ( x 1 ) + 2 f ( x 2 ) + 4 f x 3 + f x 4 

1
=  1 + 4 ( 0.8 ) + 2 ( 0.67 ) + 4 ( 0.57 ) + 0.5 
12 
1
=  1 + 3.2 + 1.34 + 2.28 + 0.5 
12
= 0.6933

Error for Simpson’s Rule

b
If I = 
a
f ( x ) dx , then E n = I − S n

Theorem:
If there exists a number M > 0 such that f 4 ( x )  M for all x in [a, b], then

M (b − a)
5
En 
180n4

Example
2 dx
Determine a value of n so that the Simpson’s rule will give an approximation to 
1 x
that
is accurate to two decimal places.

Solution:
1 24
f (x)= f 4 (x) =
x x5

Since f 4 ( x )  f 4 ( 1 ) = 24 , hence M = 24
 M (b − a)
5
En 
180 n4
24 ( 1 )
5
 0.05
180 n4
80
n4   26.67
3
n  2.27

Since n is even, use n = 4.

Exercises 8
I. Use Trapezoidal rule to obtain an approximation to the given integral for the indicated value
of n.

2 dx  /4

1 sin x
1. 1 3
x +1
n=5 2.

0
tan x dx n=3 3.
 0 x
dx n=5

II. Trapezoidal Rule problems

2 dx
1. Determine the number of trapezoids needed so that an approximation to −1 x + 3 is
accurate to two decimal places.
2. Use Trapezoidal Rule so that an approximation to the area under the graph of
1
f (x)= on [0, 2] is accurate to two decimal places.
1 + x2
3. The domain of f(x) = 10x is the set of real numbers and f(x) > 0 for all x. Use the
Trapezoidal Rule to approximate the area under the graph of f on [-2, 2] with n = 4.
4. The so-called moment of inertia I of a three-bladed ship’s propeller is given by
3  3  4.5 2
I=
2g
+
g 1 r Adr

Where  is the density of the metal, g is the acceleration of gravity, and A is the area of
a cross section of the propeller at a distance r ft from the center of the hub. If  = 570
lb/ft3 for bronze, use the data below and the Trapezoidal Rule to find an approximation
to I.
r(ft) 1 1.5 2 2.5 3 3.5 4 4.5
A(ft2) 0.3 0.5 0.62 0.7 0.6 0.5 0.27 0

III. Use the Simpson’s Rule to obtain an approximation to the given integral for the indicated
value of n.

1. 1 2.  sin x 3.  /2 dx
−1
x 2 + 1 dx n=2
0 x +
dx n=6  /4 2 + sin x n=2

IV. Simpson’s Rule problems

dx 3
1. Using Simpson’s rule, determine n so that the error in approximating
x
is less than  1
10-5. Compare with the n needed in the Trapezoidal rule to give the same accuracy.
3 dx
2. Find an upper bound for the error in approximating
0 2x + 1
by Simpson’s rule with n 
= 6.