30
Numerical Computation Lecture 16: Numerical Integration United International College

Numerical Computation Lecture 16: Numerical Integration United International College

Embed Size (px)

Citation preview

Page 1: Numerical Computation Lecture 16: Numerical Integration United International College

Numerical Computation

Lecture 16: Numerical Integration

United International College

Page 2: Numerical Computation Lecture 16: Numerical Integration United International College

Last Time

• During the last class period we covered:– Numerical Approximations to Derivatives– Readings: • Pav, Chapter 7

Page 3: Numerical Computation Lecture 16: Numerical Integration United International College

Today

• We will cover:– Numerical Integration (Quadrature)– Readings: • Pav, Chapter 8, sections 8.1-8.3• Moler,

Page 4: Numerical Computation Lecture 16: Numerical Integration United International College

The Definite Integral

• The definite integral is the total value or

summation of f(x) over a range of x.

• The integration symbol is actually a stylized capital S intended to signify the connection between integration and summation.

b

a

dxxf )(

Page 5: Numerical Computation Lecture 16: Numerical Integration United International College

The Definite Integral

f(x)

x

I f x dxa

b

ba

Page 6: Numerical Computation Lecture 16: Numerical Integration United International College

Upper, Lower Sums

• We approximate the definite integral by splitting the interval from a to b into n subintervals a = x0 < x1 < x2 < . . . < xn-1 < xn = b

• This is called a partition P of [a,b]. • On each subinterval [xi-1 ,xi ] define the smallest value

of f(x) to be mi and the largest value of f(x) to be Mi . mi = min{ f(x) | xi-1 < x < xi }

Mi = max{ f(x) | xi-1 < x < xi }

Page 7: Numerical Computation Lecture 16: Numerical Integration United International College

Upper, Lower Sums

• Then, we can define Lower and Upper approximating sums for the integral: )(),(

)(),(

1

1

0

1

1

0

ii

n

ii

ii

n

ii

xxMPfU

xxmPfL

Page 8: Numerical Computation Lecture 16: Numerical Integration United International College

Upper, Lower Sums

• Definition: A function f(x) is Riemann-integrable over [a,b] if

over all possible partitions P. Here the words “sup” and “inf” can be thought of as the same as maximum and minimum.

• Theorem (from calculus): If f(x) is continuous on [a,b], then it is Riemann-integrable on [a,b]

),(inf),(sup PfUPfLPP

Page 9: Numerical Computation Lecture 16: Numerical Integration United International College

Upper, Lower Sums

• Note: While this is a good definition, in practice it is often hard to find the max and min of f(x) on each subinterval.

• Idea: Use a simpler approximation that is between the Upper and Lower approximations.

Page 10: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Trapezoidal Rule

• Over a single interval [a,b] we can approximate f(x) by a line and find the area under this line. This area looks like a trapezoid.

Page 11: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Trapezoidal Rule

• The area under this trapezoid is just

)(2

)()(ab

bfaf

Page 12: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Trapezoidal Rule

• To get a better approximation for f(x) over [a,b] we create a partition with equal sized subintervals:

a = x0 < x1 < x2 < . . . < xn-1 < xn = b with the width of every subinterval = h. That is,

xi+1 - xi = h for all i. Adding up all the trapezoidal areas we get

1

01

1

0

))()((2

1)()(

1 n

iii

b

a

n

i

x

x

hxfxfdxxfdxxfi

i

Page 13: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Trapezoidal Rule

• Factoring out the h we get

• This is the Composite Trapezoidal Rule

1

01))()((

2)(

n

iii

b

a

xfxfh

dxxf

Page 14: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Trapezoidal Rule

• Example: Approximate using the composite trapezoidal rule with n=4. • Solution: h = (2-0)/4 = ½. Partition is 0 < 0.5 < 1.0 < 1.5 < 2.0 Trapezoidal Rule gives

2

02 1

1dx

x

))]0.2()5.1(())5.1()0.1((

))0.1()5.0(())5.0()0([(2

2/1

1

12

02

ffff

ffffdxx

1038.1

)]2.013

4()

13

45.0()5.08.0()8.01[(

4

1

1

12

02

dx

x

Page 15: Numerical Computation Lecture 16: Numerical Integration United International College

Error in Trapezoidal Rule

• Theorem: (Section 8.2.1 in Pav)

• Proof: (Read section 8.2.1 in Pav)• Note: The trapezoidal rule will be exact (no error) if

the function f(x) is linear. (Why?)

bawhere

fhab

E

''12

)( 2

Page 16: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Midpoint Rule

• The area defined by a rectangle of height f((a+b)/2) is

))(2

( abba

f

a b(a+b)/2

Page 17: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Midpoint Rule

• Create a partition: a = x0 < x1 < x2 < . . . < xn-1 < xn = b with the width of every subinterval = h. Adding up all

the midpoint areas we get

• This is the Composite Midpoint Rule

1

0

11

0

))2

(()()(1 n

i

iib

a

n

i

x

x

xxfhdxxfdxxf

i

i

Page 18: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Simpson’s Rule

• Instead of using linear approximations to f(x), we can use a 2nd order Lagrange polynomial approximation L(x). To do this, we will need three points, a,b, and x1 halfway between a and b.

a=x0 c=x1

f(x)

b=x2

L(x)

Page 19: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Simpson’s Rule

• The area under L(x) can be an approximation to the integral

2

0

21202

101

2101

200

2010

21

)()(

x

x

b

a

b

a

dxxfxxxx

xxxxxf

xxxx

xxxxxf

xxxx

xxxx

dxxLdxxf

Page 20: Numerical Computation Lecture 16: Numerical Integration United International College

Simple Simpson’s Rule

• After integrating and simplifying we get

210 46

)()( xfxfxf

abdxxf

b

a

Page 21: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Simpson’s Rule

• Create a partition with an even number of subintervals: a = x0 < x1 < x2 < . . . < xn-1 < xn = b

with the width of every subinterval = h. On every pair of subintervals, [x0 , x1], [x1 , x2 ], etc, we use Simpson’s Rule:

x0 x2 x

f(x)

x4h h xn-2h xn

…...

hx3x1 xn-1

Page 22: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Simpson’s Rule

• We get:

)()(4)(2

)4f(x)(2)f(x4

)2f(x)f(x4)2f(x)f(x4)f(x3

)f(x)4f(x)f(x6

2h

)f(x)f(x4)f(x6

2h)f(x)f(x4)f(x

6

2h

f(x)dxf(x)dxf(x)dxf(x)dx

12

12i21-2i

43210

n1n2n

432210

x

x

x

x

x

x

b

a

n

2n

4

2

2

0

nnn

i

xfxfxf

xf

h

Page 23: Numerical Computation Lecture 16: Numerical Integration United International College

Composite Simpson’s Rule

• Simplifying, we have

• This is the Composite Simpson’s Rule

• Note: The error in Simpson’s rule is O(h4)

]24[3

f(x)dx1

..5,3,1

2

..6,4,20

b

a

n

i

n

jnji xfxfxfxf

h

Page 24: Numerical Computation Lecture 16: Numerical Integration United International College

Matlab – Trapezoidal Rule

function sum = trapezoid( f, a, b, n )% trapezoid computes the trapezoidal rule% approximation to f(x) over [a,b] using n % equally spaced subintervalsh = (b-a)/n;x = a + (h .* (0:n)); % vector of partition x valuessum=0.0;for i = 1:n sum = sum + (h/2.0)*(f(x(i)) + f(x(i+1)));endend

Page 25: Numerical Computation Lecture 16: Numerical Integration United International College

Matlab – Trapezoidal Rule

>> trapezoid(inline('x^2'),1,2,64)ans = 2.3334

Page 26: Numerical Computation Lecture 16: Numerical Integration United International College

Romberg Integration

• Idea: Use the Richardson Extrapolation technique to improve integral approximations.

• Consider the Trapezoidal Rule using 2n subintervals of [a,b]. Then, h = (b-a)/ 2n and the approximation will be

• We know that this approximation has error O(h2). Thus,

12

01)()(

22

1)(

n

iiinxfxf

abn

...)()( 66

44

22 hahahadxxfn

b

a

Page 27: Numerical Computation Lecture 16: Numerical Integration United International College

Romberg Integration

• This is exactly the kind of expression we had for Richardson Extrapolation in Chapter 7. Let

...)()( 66

44

22 hahahadxxfn

b

a

14

)1,1()1,(4),(

m

m mnRmnRmnR

)()0,( nnR

Page 28: Numerical Computation Lecture 16: Numerical Integration United International College

Romberg Intergation

• Compute the table of values:

• This is called Romberg Integration

Page 29: Numerical Computation Lecture 16: Numerical Integration United International College

Romberg Intergation

• Example: Approximate The exact value =

dxxe x4

0

2

926477.5216

02.008.027.075.202.139

95.521601.521720.521768.521995.535525.0

84.522414.522975.525676.57645.0

68.549998.567079.72881

41.82402.121422

7.238474

43210

)()()()()( 108642

error

h

h

h

h

h

mmmmm

hOhOhOhOhO

Trapezoid

Page 30: Numerical Computation Lecture 16: Numerical Integration United International College

Romberg Intergation

• Note: In the column for m=1 we have

• If we calculate this out, we get exactly the Simpson’s Rule approximations for the integral.

3

)1()(4

3

)0,1()0,(4)1,(

nnnRnRnR

84.522414.522975.525676.57645.0

68.549998.567079.72881

41.82402.121422

7.238474

43210

)()()()()(

'108642

h

h

h

h

mmmmm

hOhOhOhOhO

sSimpsonTrapezoid