38
Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Embed Size (px)

Citation preview

Page 1: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Today’s class

• Numerical Integration• Newton-Cotes

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

1

Page 2: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Discuss Mid-term Exam 1

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

2

0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

Average = 74.5, Standard dev = 17.6

Max = 98 Min = 35

Histogram of scores

Page 3: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Discuss Mid-term Exam 1

1 (30) Use Taylor series to approximate at x=0.5 until t 1%, using a base point x=0. (Use 4 significant digits with rounding, and the true value is 0.4055)

 

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

3

Page 4: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Discuss Mid-term Exam 1

2 (30) Use the false-position method with initial guesses of 50 and 70 to determine x to a level of a 0.5%. (Use 4 significant digits with rounding)

 

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

4

351653.0135

xex

Page 5: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Discuss Mid-term Exam 1

3 (20) Use the Newton-Raphson method with an initial guess of 6 to determine a root of the following function to a level of a 0.01% (use 4 significant digits with rounding)

 

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

5

05.045.51 32 xxx

Page 6: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Discuss Mid-term Exam 1

4 (20) Use the Gauss-Seidel method to solve the following linear equation system to a tolerance of s = 5% with a starting point (0, 0, 0).(Use 4 significant digits with rounding)

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

6

2028

3473

3862

321

321

321

xxx

xxx

xxx

Page 7: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Numerical Integration

• Numerical technique to solve definite integrals

• Find the area under the curve f(x) from a to b

• Another way to put it is that we need to solve the following differential equation

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

7

Page 8: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Numerical Integration• Where do we use numerical techniques?

• When you can’t integrate directly• When you have a sampling of points

representing f(x) as with the experiments results

• Graphical techniques• Grid approximation• Strip approximation

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

8

Page 9: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Graphical Techniques

• Grid approximation

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

9

Page 10: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Graphical Techniques

• Strip approximation

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

10

Page 11: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Numerical Integration

• Numerical techniques• Treat integral as a summation

• Basic idea is to convert a continuous function into discrete function

• The smaller the interval, the more accurate the solution but also at more computational expense

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

11

Page 12: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Numerical Integration

• Newton-Cotes formulas• Approximate function with a series of

polynomials that are easy to integrate

• Zero-order approximation is equivalent to strip approximation

• First order approximation is equivalent to trapezoidal approximation

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

12

Page 13: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Newton-Cotes Formulas

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

13

Page 14: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Newton-Cotes Formulas

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

14

Page 15: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

15

Page 16: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

16

Page 17: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

17

Page 18: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

• Evaluate integral• Analytical solution

• Trapezoidal solution

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

18

Page 19: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

19

Page 20: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

• Trapezoidal solution• To increase accuracy, shorten up the interval

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

20

Page 21: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

21

Page 22: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

• If you split the interval into n sub-intervals, each sub-interval has a width of h = (b – a)/n

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

22

Page 23: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

• More accurate as you increase n• Double n and reduce the error by a factor of

four

• However, if n is too large, you will start encountering round-off error and the integral solution can diverge

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

23

Page 24: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s Rules

• Second and third-order polynomial forms of the Newton-Cotes formulas

• Simpson’s 1/3 Rule• Use three points on the curve to form a second

order polynomial

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

24

Page 25: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s Rules

• Approximate the function by a parabola

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

25

Page 26: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s Rules

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

26

Page 27: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s Rules

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

27

Page 28: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s 1/3 Rule

• Error

• Multiple Application

• n must be even

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

28

Page 29: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s 3/8-Rule

• Approximate with a cubic polynomial

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

29

Page 30: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s 3/8-Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

30

Page 31: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s 3/8-Rule

• Error

• Multiple Application

• n must be a multiple of 3

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

31

Page 32: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

32

Simpson’s 3/8-Rule

• Both the cubic and quadratic approximations are of the same order

• The cubic approximation is slightly more accurate then the quadratic approximation

• Usually not worth the extra work required• Only use the 3/8 rule if you need odd

number of segments• Can combine with 1/3 rule on some segments

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

32

Page 33: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Example

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

33

Page 34: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Integration with unequal segments• We have until now assumed that each

segment is equal• If we are integrating based on

experimental or tabular data, this assumption may not be true

• Newton-Cotes formulas can easily be adapted to accommodate unequal segments

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

34

Page 35: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoid Rule with Unequal Segments

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

35

Page 36: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Trapezoidal Rule

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

36

Page 37: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Simpson’s Rule with Unequal Segments

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

37

Page 38: Today’s class Numerical Integration Newton-Cotes Numerical Methods Lecture 12 Prof. Jinbo Bi CSE, UConn 1

Next class

• Numerical Integration

• HW5 due Tuesday Oct 21

Numerical MethodsLecture 12

Prof. Jinbo BiCSE, UConn

38