17
St. John's University of Tanzania MAT210 NUMERICAL ANALYSIS 2013/14 Semester II INTEGRATION Richardson's Extrapolation & Romberg Integration Kaw, Chapter 7.04 http://nm.mathforcollege.com/topics/romberg_method.html

MAT210/Integration/Romberg 2013-14

Embed Size (px)

DESCRIPTION

Lecture slides introducing Romberg Integration based on Chapter 7.04 of Prof. Anton Kaw's Numerical Methods textbook. Some parts of this presentation are based on resources at http://nm.MathForCollege.com, primarily http://mathforcollege.com/nm/topics/romberg_method.html

Citation preview

Page 1: MAT210/Integration/Romberg 2013-14

St. John's University of Tanzania

MAT210 NUMERICAL ANALYSIS2013/14 Semester II

INTEGRATIONRichardson's Extrapolation & Romberg Integration

Kaw, Chapter 7.04http://nm.mathforcollege.com/topics/romberg_method.html

Page 2: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 2 of 17

● Approximating Error was not exact● Error for multi-segment Trapezoidal was

known to be

● What if we tried to say that the next approximation for the integral was

● In the example the “predicted” error was 51 and the “actual” error was 48, but at least we jump from error of ~50 to one of ~3

Recall

112h2 f ''

In+1=In+predicted error

Page 3: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 3 of 17

Richardson's Extrapolation● That is the idea behind a technique known

as Richardson's Extrapolation● Jump to a new approximation using an

approximation of the error● It works out in this integration case because

the error is nearly a function of n alone

Et=112h2 f ' '= 1

n2[(b−a)2 f ' '

12 ]⇒Et≈

Cn2

Page 4: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 4 of 17

Improved approximationI≈In+

Cn2

I≈I 2n+C

(2n)2⇒ 4 I≈4 I 2n+

Cn2

⇒ 3 I ≈ 4 I 2n−In

True Value, TV = I ≈ I 2n+I 2n−In

3

Richardson's Extrapolation for True Value

Page 5: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 5 of 17

Apply it to the Example

Exact is 11061m, so error is only 0.00904%!

Page 6: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 6 of 17

More on the Error

Now recall Simpson's 1/3 results:

Richardson's Extrapolation beats them both

Page 7: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 7 of 17

Romberg Integration● Romberg Integration takes Richardson's

Extrapolation and builds a recursive algorithm around it

● It uses the added fact that

to create the recursion formula

Et=A1h2+A2h

4+A3h6+…

Page 8: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 8 of 17

Finding Recursion

The next term in the error expansion

Combining like before yields

Page 9: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 9 of 17

Recursion Formula and Error● The process can repeated to produce:

● j = Level of accuracy● k = Order of the extrapolation

● k = 1 – Trapezoidal rule, O(h²) error● k = 2 – 1st Level of Romberg, O(h4) error● k = 3 – 2nd level of Romberg, O(h6) error

Page 10: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 10 of 17

Applied to the Example● Results from the 1,2,4 & 8

segment Trapezoidal Rule● This is the hard work● The rest is easy & recursive

Page 11: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 11 of 17

Pictorial View

Page 12: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 12 of 17

Spreadheet View

Page 13: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 13 of 17

Not just known functions!

Page 14: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 14 of 17

Not just known functions!

Page 15: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 15 of 17

Not just known functions!

Page 16: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 16 of 17

Not just known functions!

Page 17: MAT210/Integration/Romberg 2013-14

MAT210 2013/14 Sem II 17 of 17

Good enough to be “standard”