ME 361-Interpolation

ME – 361 NUMERICAL METHODS FOR ENGINEERS
INTERPOLATION
1.00 In engineering and science, one often
has a number of data points, obtained
0.80
by sampling or experimentation, which
0.60 represent the values of a function, f(x),
for a limited number of values of
0.40
the independent variable, x. It is often
required to interpolate, i.e., estimate
0.20
the value of that function for an
f(x) = ?
0.00 intermediate value of the independent
0.0 2.0 4.0 6.0
x variable.
Remember thermodynamic tables and linear interpolations.
Prof. Dr. Faruk Arınç Fall 2020

INTERPOLATION
f(x) = 2x - 10 => g(x) = ? A closely related problem is

25 the approximation of a complicated
20 function by a simple function. Suppose
15 the formula for some given function is
10 known, but too complicated to evaluate
5 efficiently. A few data points from the
0 original function can be interpolated to
-5 produce a simpler function which is still
-10 fairly close to the original. The resulting
-15 gain in simplicity may outweigh the loss
-6 -4 -2 0 2 4 6
from interpolation error.
INTERPOLATING POLYNOMIALS
Given a set of data points (tabulated data) [xi , f(xi)] , i = 0 to N

(or f(x) in (a , b) converted to a set of N+1 data points)
No random statistical errors.
Question: Find f(x), x not being one of the data points.
Answer: Replace the data with a polynomial of some chosen, small degree, n,
f (x)  Pn (x) = c1 + c 2 x + c 3 x 2 + ... + c n x n = g(x)
Find all the constants, ci’s, and evaluate Pn(x).
Questions: How do you find the constants, ci’s?
And how do you find or estimate the error?

Example:
xi 0 1 2 3 4 5 6
f(xi) 1 0.368 0.135 0.050 0.018 0.007 0.002
1.00
0.80
At x = 2.5
0.60
f(x) = f (2.5) = ?
Error = ?
0.40
0.20
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0

Solution:
 Choose a polynomial of some (small) degree, Pn(x);
 Decide on the range (the data points to be used);
 Estimate the error.
Questions: How do you find the constants, ci’s?
and, how do you find (or estimate) the error?
 Forward
Answer: Use Newton-Gregory difference tables  Backward
 Central
Isaac Newton James Gregory

1642 - 1726 1638 - 1675

Forward (and backward) differences:


Newton-Gregory Formulae
Forward
Define  f i = f i+1 - f i First forward difference at i
Differences
2 f i =  f i+1 -  f i Second forward difference at i
n f i = n-1 f i+1 - n-1 f i
Backward
Differences  f i  f i - f i -1 First backward difference at i
2 f i   f i -  f i -1 Second backward difference at i
 n f i   n1 f i -  n1 f i - 1

x - x0 x - x0
Define s= = where h is constant
x h
Newton-Gregory Forward Polynomial:
s (s - 1) 2 s (s - 1) ... (s - n  1) n
Pn (x)  f 0  s  f 0   f 0  ...   f0
2! n!
s (s - 1) (s - 2) ... (s - n) n 1 (n 1)
E (x)  h f ( )
Error:
(n  1)!
(x - x 0 ) (x - x1 ) ... (x - x n ) (n 1)
E (x)  f ( ) , x 0    x n
(n  1)!
To estimate the error, replace the (n+1)th derivative with (n+1)th divided difference.

Difference Table for f(x) = e-x

x f (x)
0 1
1 0.367879441
2 0.135335283
3 0.049787068
4 0.018315639
5 0.006737947
6 0.002478752

x f (x) Δf
0 1
- 0.632120559
1 0.367879441
- 0.232544158
2 0.135335283
- 0.085548215
3 0.049787068
- 0.031471429
4 0.018315639
- 0.011577692
5 0.006737947
- 0.004259195
6 0.002478752

x f (x) Δf Δ2 f
0 1
- 0.632120559
1 0.367879441 0.399576401
- 0.232544158
2 0.135335283 0.146995943
- 0.085548215
3 0.049787068 0.054076785
- 0.031471429
4 0.018315639 0.019893738
- 0.011577692
5 0.006737947 0.007318497
- 0.004259195
6 0.002478752

x f (x) Δf Δ2 f Δ3 f
0 1
- 0.632120559
1 0.367879441 0.399576401
- 0.232544158 - 0.252580458
2 0.135335283 0.146995943
- 0.085548215 - 0.092919158
3 0.049787068 0.054076785
- 0.031471429 - 0.034183048
4 0.018315639 0.019893738
- 0.011577692 - 0.012575241
5 0.006737947 0.007318497
- 0.004259195
6 0.002478752

x f (x) Δf Δ2 f Δ3 f Δ4 f
0 1
- 0.632120559
1 0.367879441 0.399576401
- 0.232544158 - 0.252580458
2 0.135335283 0.146995943 0.1596613
- 0.085548215 - 0.092919158
3 0.049787068 0.054076785 0.05873611
- 0.031471429 - 0.034183048
4 0.018315639 0.019893738 0.021607807
- 0.011577692 - 0.012575241
5 0.006737947 0.007318497
- 0.004259195
6 0.002478752
Example: At x = 2.5 f(x) = f(2.5) = ?
Choose a polynomial of some degree, Pn(x): Say, n = 3
Which four points? Approximately symmetrical around the point of interest
Find the constants of the polynomial: Use the difference table and the formula
Evaluate the polynomial at x = 2.5. This is the interpolation.
How do you estimate the error? Use the difference table and the formula


0 1
- 0.632120559
1 0.367879441 0.399576401
- 0.232544158 - 0.252580458
2 0.135335283 0.146995943 0.1596613
x = 2.5 ? - 0.085548215 - 0.092919158
3 0.049787068 0.054076785 0.05873611
- 0.031471429 - 0.034183048
4 0.018315639 0.019893738 0.021607807
- 0.011577692 - 0.012575241
5 0.006737947 0.007318497
- 0.004259195
6 0.002478752

0 1
- 0.632120559
x0 =1 0.367879441 0.399576401
- 0.232544158 - 0.252580458
x1 = 2 0.135335283 0.146995943 0.1596613
x = 2.5 ? - 0.085548215 - 0.092919158
x2 = 3 0.049787068 0.054076785 0.05873611
- 0.031471429 - 0.034183048
x3 = 4 0.018315639 0.019893738 0.021607807
- 0.011577692 - 0.012575241
5 0.006737947 0.007318497
- 0.004259195
6 0.002478752
x - x0 x-1
Define s  = =x-1
h 1
s (s - 1) 2 s (s - 1) (s - 2) 3
P3 (s)  f0  s f0   f0   f0
2! 3!
P3 (x)  0.368   x - 1  - 0.233  

 x - 1 (x - 2)
0.147 
2
 x - 1 (x - 2) (x - 3)
  - 0.093 
6
(x - 1) (x - 2)  x - 3  (x - 4)
E (x)  f (4) ( )
Error: 4!
(x - 1) (x - 2)  x - 3  (x - 4)  4 f0
E (x) 
4! h4

Difference Table for f (x) = x2
x f (x) f 2 f 3 f
0 0
1
1 1 2
3 0
2 4 2
5 0
3 9 2
7 0
4 16 2
9
5 25

Difference Table for f (x) = x2
x f (x) f 2 f 3 f
0 0
1
1 1 1.9
2.9 0.3
2 3.9 2.2
5.1 - 0.3
3 9 1.9
7 0.1
4 16 2
9
5 25

Polynomial Interpolation with Unequally-spaced Data
Lagrange Form of a Polynomial:
(x - x1 ) (x - x 2 ) ... (x - x n-1 ) (x - x n )
Pn (x) = f (x 0 )
(x 0 - x1 ) (x 0 - x 2 ) ... (x 0 - x n-1 ) (x 0 - x n )
(x - x 0 ) (x - x 2 ) ... (x - x n-1 ) (x - x n )
+ f (x1 )
(x1 - x 0 ) (x1 - x 2 ) ... (x1 - x n-1 ) (x1 - x n )
+ ...
(x - x 0 ) (x - x1 ) ... (x - x n-2 ) (x - x n )
+ f (x n-1 )
(x n-1 - x 0 ) (x n-1 - x1 ) ... (x n-1 - x n-2 ) (x n-1 - x n )
(x - x 0 ) (x - x1 ) ... (x - x n-2 ) (x - x n-1 )
+ f (x n )
(x n - x 0 ) (x n - x1 ) ... (x n - x n-2 ) (x n - x n-1 )

Divided differences:
f i 1 - f i
D fi  First-order forward divided difference
x i 1 - x i
D f i1 - D f i
D2 fi  Second-order forward divided difference
x i2 - x i
D n -1 f i 1 - D n -1 f i
D fi 
n
n'th-order forward divided difference
x in - x i

Divided Difference Table for f(x) = e-x
x f (x)
0 1
1 0.367879441
3 0.049787068
6 0.002478752

Divided Difference Table for f (x) = e-x
x f (x) Df
0 1
-0.632120559
1 0.367879441
-0.159046186
3 0.049787068
-0.015769439
6 0.002478752

x f (x) Df D2 f
0 1
-0.632120559
1 0.367879441 0.157691457
-0.159046186
3 0.049787068 0.02865535
-0.015769439
6 0.002478752

x f (x) Df D2 f D3 f
0 1
-0.632120559
1 0.367879441 0.157691457
-0.159046186 -0.021506018
3 0.049787068 0.02865535
-0.015769439
6 0.002478752

Newton-Gregory Interpolating Polynomial with Forward Divided Differences:
Pn (x)  f(x i )  Dfi (x - x i )  D 2fi (x - x i ) (x - x i1) 

...  Dnfi (x - x i ) (x - x i1) ... (x - x in-1)
n
f (n 1) ( )
Error: E (x)  f (x) - Pn (x)   (x - x i )
i 0 (n  1) !
n
E (x)   (x
i 0
- x i ) D n 1
fi

Example: At x = 2.5 f(x) = f(2.5) = ?
Choose a polynomial of some degree, Pn(x): Say, n = 3
There are only four points in the given data.

Which four points?
So, we use them all.
Use the divided difference table

Find the constants of the polynomial:
and the formula for divided differences
Evaluate the polynomial at x = 2.5. This is the interpolation.
How do you estimate the error? Use the divided difference table
and the related formula

P3 (x)  f(x 0 )  (x - x 0 ) Dfi  (x - x 0 ) (x - x1 ) D2 fi 

(x - x 0 ) (x - x1 ) (x - x 2 ) D3 fi
P3 (x)  1  x  - 0.6321206   x (x - 1)  0.15769146 

 x (x - 1) (x - 3)  - 0.0215060 
(x - 1) (x - 2)  x - 3  (x - 4)
E (x)  f (4) ( )
Error: 4!
E (x)  (x - 1) (x - 2)  x - 3  (x - 4) D 4f0
Note that the fourth divided difference, D4f, does nor exists. So, what do you do?

Example: An object is suspended in a wind tunnel and the force is measured for
various levels of wind velocity. The results are tabulated below.
V, m/s 10 20 30 40 50 60
F, N 25 70 380 550 610 1220
1400
(a) Find the force at V = 32 1200
m/s using a second- 1000
800
degree interpolating
600
polynomial, P2(V).
400
(b) Estimate the error. 200
0
0 20 40 60 80

V F ΔF Δ2F Δ3F Δ4F Δ5F

10 25
45
20 70 265
310 -405
Difference table: 30 380 -140 435
170 30 195
40 550 -110 630
60 660
50 610 550
610
60 1220
Interpolating polynomial: P2 (V)  70  31 (V - 20) - 0.7 (V - 20) (V - 30)

P2 (32)  425.2 N
(V - V0 ) (V - V1 ) (V - V2 )  3F
Error estimate: E(x)  3
= - 0.96
3! h

V F DF D2F D3F D4F D5F

10 25
4.5
20 70 1.325
31 -0.0675
Divided 30 380 -0.7 0.00181
17 0.005 0.00001625
difference table: 40 550 -0.55 0.00263
6 0.11
50 610 2.75
61
60 1220
Interpolating polynomial: P2 (V)  70  31 (V - 20) - 0.7 (V - 20) (V - 30)

P2 (32)  425.2 N
Error estimate: E(x)  (V - V0 ) (V - V1 ) (V - V2 ) (D 3F) = - 0.96

Rules of Thumb for Polynomial Interpolation
 The error of polynomial interpolation is reduced by making the range as

symmetric as possible around the point of interpolation.
 The truncation error will decrease but round-off will increase as the degree of
the polynomial increases.
 A local irregularity in the tabulated data can cause amplified distortions at

points far from the irregularities. Such amplification increases with the degree
of the interpolating polynomial.
 If there is a choice, smaller step size, h, leads to higher accuracy.


ME 361-Interpolation

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ME 361-Interpolation

Uploaded by

Copyright:

Available Formats

ME – 361 NUMERICAL METHODS FOR ENGINEERS

Remember thermodynamic tables and linear interpolations.

Prof. Dr. Faruk Arınç Fall 2020

f(x) = 2x - 10 => g(x) = ? A closely related problem is

Given a set of data points (tabulated data) [xi , f(xi)] , i = 0 to N

Question: Find f(x), x not being one of the data points.

f (x)  Pn (x) = c1 + c 2 x + c 3 x 2 + ... + c n x n = g(x)

Find all the constants, ci’s, and evaluate Pn(x).

Questions: How do you find the constants, ci’s?

And how do you find or estimate the error?

Prof. Dr. Faruk Arınç Fall 2020

 Choose a polynomial of some (small) degree, Pn(x);

 Decide on the range (the data points to be used);

 Estimate the error.

Questions: How do you find the constants, ci’s?

and, how do you find (or estimate) the error?

Isaac Newton James Gregory

Prof. Dr. Faruk Arınç Fall 2020

Forward (and backward) differences:

Prof. Dr. Faruk Arınç Fall 2020

Prof. Dr. Faruk Arınç Fall 2020

2 f i =  f i+1 -  f i Second forward difference at i

n f i = n-1 f i+1 - n-1 f i

2 f i   f i -  f i -1 Second backward difference at i

Prof. Dr. Faruk Arınç Fall 2020

Newton-Gregory Forward Polynomial:

Prof. Dr. Faruk Arınç Fall 2020

Difference Table for f(x) = e-x

Difference Table for f(x) = e-x

Difference Table for f(x) = e-x

Difference Table for f(x) = e-x

Difference Table for f(x) = e-x

Example: At x = 2.5 f(x) = f(2.5) = ?

Choose a polynomial of some degree, Pn(x): Say, n = 3

Which four points? Approximately symmetrical around the point of interest

Evaluate the polynomial at x = 2.5. This is the interpolation.

Prof. Dr. Faruk Arınç Fall 2020

Difference Table for f(x) = e-x

Difference Table for f(x) = e-x

Newton-Gregory Forward Polynomial:

P3 (x)  0.368   x - 1  - 0.233  

Prof. Dr. Faruk Arınç Fall 2020

Difference Table for f (x) = x2

Prof. Dr. Faruk Arınç Fall 2020

Difference Table for f (x) = x2

Prof. Dr. Faruk Arınç Fall 2020

Polynomial Interpolation with Unequally-spaced Data

Lagrange Form of a Polynomial:

Prof. Dr. Faruk Arınç Fall 2020

Prof. Dr. Faruk Arınç Fall 2020

Divided Difference Table for f(x) = e-x

Prof. Dr. Faruk Arınç Fall 2020

Divided Difference Table for f (x) = e-x

Prof. Dr. Faruk Arınç Fall 2020

Divided Difference Table for f (x) = e-x

Prof. Dr. Faruk Arınç Fall 2020

Divided Difference Table for f (x) = e-x

Prof. Dr. Faruk Arınç Fall 2020

Newton-Gregory Interpolating Polynomial with Forward Divided Differences:

Pn (x)  f(x i )  Dfi (x - x i )  D 2fi (x - x i ) (x - x i1) 

Prof. Dr. Faruk Arınç Fall 2020

Example: At x = 2.5 f(x) = f(2.5) = ?

Choose a polynomial of some degree, Pn(x): Say, n = 3