9/8/2021

Introduction to Scientific
Computing

9/8/2021 Son Dao, PhD 1

Introduction

Son Dao, PhD 2

Son Dao, PhD 1

 9/8/2021

My advice
• If you don’t let a teacher know at what level
you are by asking a question, or revealing
your ignorance you will not learn or grow.
• You can’t pretend for long, for you will
eventually be found out. Admission of
ignorance is often the first step in our
education.
– Steven Covey—Seven Habits of Highly Effective People

Son Dao, PhD 3

Why use Numerical Methods?

• To solve problems that cannot be solved exactly

x u2
1 

2 e

2
du

Son Dao, PhD 2

 9/8/2021

Why use Numerical Methods?

• To solve problems that are intractable!

Steps in Solving an
Engineering Problem

Son Dao, PhD

Son Dao, PhD 6

Son Dao, PhD 3

 9/8/2021

How do we solve an engineering

problem?

Problem Description

Mathematical Model

Solution of Mathematical Model

Using the Solution

Son Dao, PhD 7

Example of Solving an
Engineering Problem

Son Dao, PhD 8

Son Dao, PhD 4

 9/8/2021

Bascule Bridge THG

Son Dao, PhD 9

Bascule Bridge THG

Hub

Trunnion

Girder

Son Dao, PhD 10

10

Son Dao, PhD 5

 9/8/2021

Trunnion-Hub-Girder
Assembly Procedure

Step1. Trunnion immersed in dry-ice/alcohol

Step2. Trunnion warm-up in hub
Step3. Trunnion-Hub immersed in
dry-ice/alcohol
Step4. Trunnion-Hub warm-up into girder

Son Dao, PhD 11

11

Problem

After Cooling, the Trunnion Got Stuck

in Hub
Son Dao, PhD 12

12

Son Dao, PhD 6

 9/8/2021

Why did it get stuck?

Magnitude of contraction needed in the trunnion

was 0.015” or more. Did it contract enough?

Son Dao, PhD 13

13

Video of Assembly Process

Unplugged Version VH1 Version

Son Dao, PhD 14

14

Son Dao, PhD 7

 9/8/2021

Consultant calculations

 D  D    T
D  12.363"
  6.47  106 in / in / o F
T  108  80  188 o F

D  (12.363)(6.47  106 )( 188)

 0.01504"
Son Dao, PhD 15

15

Is the formula used correct?

D  D    T T(oF) α (μin/in/oF)
-340 2.45
-300 3.07
-220 4.08
-160 4.72
D  D  T
-80 5.43
0 6.00
40 6.24
80 6.47

Son Dao, PhD 16

16

Son Dao, PhD 8

 9/8/2021

The Correct Model Would Account for

Varying Thermal Expansion Coefficient

Tc

D  D   (T )dT
Ta

Son Dao, PhD 17

17

Can You Roughly Estimate the

Contraction?
Tc

D  D   (T )dT Ta=80oF; Tc=-108oF; D=12.363”

Ta

Tc

D  D   (T )dT
Ta

Son Dao, PhD 18

18

Son Dao, PhD 9

 9/8/2021

Can You Find a Better Estimate for

the Contraction?
Tc

D  D   (T )dT
Ta

Ta = 80oF
Tc = -108oF
D = 12.363"

Son Dao, PhD 19

19

Estimating Contraction
Accurately
Change in diameter
(D) by cooling it in dry
ice/alcohol is given by
Tc

D  D   (T ) dT
Ta

Ta = 80oF
Tc = -108oF
D = 12.363"
  1.2278  10 5 T 2  6.1946  10 3 T  6.0150
D  0.0137"
Son Dao, PhD 20

20

Son Dao, PhD 10

 9/8/2021

So what is the solution to the

problem?
One solution is to immerse the trunnion in liquid nitrogen
which has a boiling point of -321oF as opposed to the
dry-ice/alcohol temperature of -108oF.

D  0.0244"

Son Dao, PhD 21

21

Revisiting steps to solve a problem

1) Problem Statement: Trunnion got stuck in
the hub.
2) Modeling: Developed a new model
Tc

D  D   (T )dT
Ta

3) Solution: 1) Used trapezoidal rule OR b)

Used regression and integration.
4) Implementation: Cool the trunnion in liquid
nitrogen. Son Dao, PhD 22

22

Son Dao, PhD 11

 9/8/2021

THE END

Son Dao, PhD 23

23

Introduction to Numerical Methods

Mathematical Procedures

Son Dao, PhD 24

24

Son Dao, PhD 12

 9/8/2021

Mathematical Procedures
• Nonlinear Equations
• Differentiation
• Simultaneous Linear Equations
• Curve Fitting
– Interpolation
– Regression
• Integration
• Ordinary Differential Equations
• Other Advanced Mathematical Procedures:
– Partial Differential Equations
– Optimization
– Fast Fourier Transforms
Son Dao, PhD 25

25

Nonlinear Equations
How much of the floating ball is under water?

Diameter=0.11m
Specific Gravity=0.6

x 3  0.165 x 2  3.993  10 4  0
Son Dao, PhD 26

26

Son Dao, PhD 13

 9/8/2021

Nonlinear Equations
How much of the floating ball is under the water?

f ( x )  x 3  0.165 x 2  3.993  104  0

Son Dao, PhD 27

27

Differentiation
What is the acceleration
at t=7 seconds?

 16  104  dv
v(t)  2200 ln
16  10 4
 5000t
  9.8t a
  dt
Son Dao, PhD 28

28

Son Dao, PhD 14

 9/8/2021

Differentiation
What is the acceleration at t=7 seconds?
Time (s) 5 8 12
Vel (m/s) 106 177 600

dv
a
dt

Son Dao, PhD 29

29

Simultaneous Linear Equations

Find the velocity profile, given

Time (s) 5 8 12
Vel (m/s) 106 177 600

v (t )  at 2  bt  c, 5  t  12
Three simultaneous linear equations
25a  5b  c  106
64a  8b  c  177
144a  12b  c  600

Son Dao, PhD 30

30

Son Dao, PhD 15

 9/8/2021

Interpolation
What is the velocity of the rocket at t=7 seconds?
Time (s) 5 8 12
Vel (m/s) 106 177 600

Son Dao, PhD 31

31

Regression
Thermal expansion coefficient data for cast steel

Son Dao, PhD 32

32

Son Dao, PhD 16

 9/8/2021

Regression (cont)

Son Dao, PhD 33

33

Integration
Finding the diametric contraction in a steel shaft when
dipped in liquid nitrogen.

T fluid

D  D   dT
Troom

Son Dao, PhD 34

34

Son Dao, PhD 17

 9/8/2021

Ordinary Differential Equations

How long does it take a trunnion to cool down?

d
mc  hA(   a ),  (0)   room
dt
Son Dao, PhD 35

35

Son Dao, PhD 18

 9/8/2021

Measuring Errors

Son Dao, PhD

Why measure errors?

1) To determine the accuracy of
numerical results.
2) To develop stopping criteria for
iterative algorithms.

2 Son Dao, PhD

Son Dao, PhD 1

 9/8/2021

True Error
 Defined as the difference between the true
value in a calculation and the approximate
value found using a numerical method etc.

True Error = True Value – Approximate Value

3 Son Dao, PhD

Example—True Error
The derivative, f (x) of a function f (x) can be
approximated by the equation,
f ( x  h)  f ( x)
f ' ( x) 
h

If f ( x)  7e and h  0.3
0.5 x

a) Find the approximate value of f ' ( 2)

b) True value of f ' ( 2)
c) True error for part (a)

4 Son Dao, PhD

Son Dao, PhD 2

 9/8/2021

Example (cont.)
Solution:
a) For x  2 and h  0.3
f ( 2  0.3)  f ( 2)
f ' (2) 
0 .3
f (2.3)  f (2)

0.3
7e 0.5( 2.3)  7e 0.5( 2)

0.3
22.107  19.028
  10.263
0 .3

5 Son Dao, PhD

Example (cont.)
Solution:
b) The exact value of f ' ( 2) can be found by using
our knowledge of differential calculus.
f ( x)  7e 0.5 x
f ' ( x )  7  0.5  e 0.5 x
 3.5e 0.5 x
So the true value of f ' ( 2) is
f ' ( 2)  3.5e 0.5( 2 )
 9.5140
True error is calculated as
Et  True Value – Approximate Value
 9.5140  10 .263   0.722

6 Son Dao, PhD

Son Dao, PhD 3

 9/8/2021

Relative True Error

 Defined as the ratio between the true
error, and the true value.
True Error
Relative True Error ( t ) =
True Value

7 Son Dao, PhD

Example—Relative True Error

Following from the previous example for true error,
find the relative true error for f ( x)  7e 0.5 x at f ' (2)
with h  0.3
From the previous example,
Et  0.722
Relative True Error is defined as
True Error
t 
True Value
 0.722
  0.075888
9.5140
as a percentage,
t  0.075888  100%  7.5888 %

8 Son Dao, PhD

Son Dao, PhD 4

 9/8/2021

Approximate Error
 What can be done if true values are not
known or are very difficult to obtain?
 Approximate error is defined as the
difference between the present
approximation and the previous
approximation.
Approximate Error ( E a ) = Present Approximation – Previous Approximation

9 Son Dao, PhD

Example—Approximate Error
For f ( x)  7e 0.5 x at x  2 find the following,
a) f (2) using h  0.3
b) f (2) using h  0.15
c) approximate error for the value of f (2) for part b)
Solution:
a) For x  2 and h  0.3
f ( x  h)  f ( x)
f ' ( x) 
h
f ( 2  0.3)  f ( 2)
f ' ( 2) 
0. 3

10 Son Dao, PhD

10

Son Dao, PhD 5

 9/8/2021

Example (cont.)
Solution: (cont.)
f (2.3)  f (2)

0.3
7e 0.5( 2.3)  7e 0.5( 2)

0.3
22.107  19.028
  10 .263
0.3
b) For x  2 and h  0.15
f (2  0.15)  f (2)
f ' ( 2) 
0.15
f (2.15)  f ( 2)

0.15

11 Son Dao, PhD

11

Example (cont.)
Solution: (cont.)
7e 0.5( 2.15)  7e 0.5( 2 )

0.15
20.50  19.028
  9.8800
0.15

c) So the approximate error, E a is

Ea  Present Approximation – Previous Approximation
 9.8800  10 .263
 0.38300

12 Son Dao, PhD

12

Son Dao, PhD 6

 9/8/2021

Relative Approximate Error

 Defined as the ratio between the
approximate error and the present
approximation.
Approximate Error
Relative Approximate Error ( a) =
Present Approximation

13 Son Dao, PhD

13

Example—Relative Approximate Error

For f ( x)  7e at x  2 , find the relative approximate
0.5 x

error using values from h  0.3 and h  0.15

Solution:
From Example 3, the approximate value of f (2)  10.263
using h  0.3 and f (2)  9.8800 using h  0.15
Ea  Present Approximation – Previous Approximation
 9.8800  10 .263
 0.38300

14 Son Dao, PhD

14

Son Dao, PhD 7

 9/8/2021

Example (cont.)
Solution: (cont.)
Approximate Error
a 
Present Approximation
 0.38300
  0.038765
9.8800
as a percentage,
 a   0 . 038765  100 %   3 . 8765 %

Absolute relative approximate errors may also need to

be calculated,
a | 0.038765 |  0.038765 or 3.8765%

15 Son Dao, PhD

15

How is Absolute Relative Error used as a

stopping criterion?
If |a |  s where s is a pre-specified tolerance, then
no further iterations are necessary and the process is
stopped.

If at least m significant digits are required to be

correct in the final answer, then
|a | 0.5  10 2 m %

16 Son Dao, PhD

16

Son Dao, PhD 8

 9/8/2021

Table of Values
For f ( x)  7e at x  2 with varying step size, h
0.5 x

h f (2) a m
0.3 10.263 N/A 0

0.15 9.8800 3.877% 1

0.10 9.7558 1.273% 1

0.01 9.5378 2.285% 1

0.001 9.5164 0.2249% 2

17 Son Dao, PhD

17

Son Dao, PhD 9

 9/8/2021

Sources of Error

Son Dao, PhD

Two sources of numerical error

1) Round off error
2) Truncation error

1
 9/8/2021

Round-off Error

Round off Error

• Caused by representing a number
approximately

1
 0.333333
3
2  1.4142...

2
 9/8/2021

Problems created by round off error

• 28 Americans were killed on February 25,

1991 by an Iraqi Scud missile in Dhahran,
Saudi Arabia.
• The patriot defense system failed to track
and intercept the Scud. Why?

Problem with Patriot missile

• Clock cycle of 1/10 seconds was
represented in 24-bit fixed point
register created an error of 9.5 x
10-8 seconds.
• The battery was on for 100
consecutive hours, thus causing
an inaccuracy of

s 3600s
 9.5  108  100hr 
0.1s 1hr
 0.342s
6

3
 9/8/2021

Problem (cont.)
• The shift calculated in the ranging system
of the missile was 687 meters.
• The target was considered to be out of
range at a distance greater than 137
meters.

Effect of Carrying Significant

Digits in Calculations

4
 9/8/2021

Find the contraction in the

diameter
Tc

D  D   (T )dT
Ta

Ta=80oF; Tc=-108oF; D=12.363”

α = a0+ a1T + a2T2

9

Thermal Expansion
Coefficient vs Temperature
T(oF) α (μin/in/oF)
-340 2.45
-300 3.07
-220 4.08
-160 4.72
D  D  T
-80 5.43
0 6.00
40 6.24
80 6.47

10

10

5
 9/8/2021

Regressing Data in Excel

(general format)

Alpha ( in/in/ o F)
8.00

4.00

0.00
-400 -200 0 200

T ( o F)

α = -1E-05T2 + 0.0062T + 6.0234 11

11

Observed and Predicted Values

α = -1E-05T2 + 0.0062T + 6.0234
T(oF) α (μin/in/oF) α (μin/in/oF)
Given Predicted
-340 2.45 2.76
-300 3.07 3.26
-220 4.08 4.18
-160 4.72 4.78
-80 5.43 5.46
0 6.00 6.02
40 6.24 6.26
80 6.47 6.46 12

12

6
 9/8/2021

Regressing Data in Excel

(scientific format)

Alpha ( in/in/ F)
8.00

o
4.00

0.00
-400 -200 0 200
o
T ( F)

α = -1.2360E-05T2 + 6.2714E-03T + 6.0234

13

13

Observed and Predicted Values

α = -1.2360E-05T2 + 6.2714E-03T + 6.0234
T(oF) α (μin/in/oF) α (μin/in/oF)
Given Predicted
-340 2.45 2.46
-300 3.07 3.03
-220 4.08 4.05
-160 4.72 4.70
-80 5.43 5.44
0 6.00 6.02
40 6.24 6.25
80 6.47 6.45 14

14

7
 9/8/2021

Observed and Predicted Values

α = -1.2360E-05T2 + 6.2714E-03T + 6.0234
α = -1E-05T2 + 0.0062T + 6.0234
T(oF) α (μin/in/oF) α (μin/in/oF) α (μin/in/oF)
Given Predicted Predicted
-340 2.45 2.46 2.76
-300 3.07 3.03 3.26
-220 4.08 4.05 4.18
-160 4.72 4.70 4.78
-80 5.43 5.44 5.46
0 6.00 6.02 6.02
40 6.24 6.25 6.26
80 6.47 6.45 6.46
15

15

THE END

16

16

8
 9/8/2021

Truncation Error

17

17

Truncation error
• Error caused by truncating or
approximating a mathematical
procedure.

18

18

9
 9/8/2021

Example of Truncation Error

Taking only a few terms of a Maclaurin series to

x
approximate e
x 2 x3
e  1  x    ....................
x

2! 3!
If only 3 terms are used,
 x2 
Truncation Error  e  1  x  
 x

 2! 

19

19

Another Example of Truncation

Error

Using a finite x to approximate f (x )

f ( x  x)  f ( x)
f ( x) 
x

secant line
P

tangent line

Figure 1. Approximate derivative using finite Δx

20

20

10
 9/8/2021

Another Example of Truncation

Error

Using finite rectangles to approximate an

integral.
y

90

2
y=x
60

30

0 x
0 1.5 3 4.5 6 7.5 9 10.5 12

21

21

Example 1 —Maclaurin series

1.2
Calculate the value of e with an absolute
relative approximate error of less than 1%.
1.2 2 1.2 3
e1.2  1  1.2    ...................
2! 3!
n
e 1.2 Ea a %
1 1 __ ___
2 2.2 1.2 54.545
3 2.92 0.72 24.658
4 3.208 0.288 8.9776
5 3.2944 0.0864 2.6226
6 3.3151 0.020736 0.62550

6 terms are required. How many are required to get 22

at least 1 significant digit correct in your answer?
22

11
 9/8/2021

Example 2 —Differentiation

f ( x  x)  f ( x)
Find f (3) for f ( x)  x using f ( x) 
2

x
and x  0.2
f (3  0.2)  f (3)
f ' (3) 
0 .2
f (3.2)  f (3) 3.2 2  32 10.24  9 1.24
     6.2
0. 2 0.2 0 .2 0.2

The actual value is

f ' ( x )  2 x, f ' (3)  2  3  6

Truncation error is then, 6  6.2  0.2

Can you find the truncation error with x  0.1 23

23

Example 3 — Integration

Use two rectangles of equal width to

approximate the area under the curve for
f ( x)  x 2 over the interval [3,9]
y

90
9

 dx
2
x
2
y=x
60

30 3

0 x
0 3 6 9 12

24

24

12
 9/8/2021

Integration example (cont.)

Choosing a width of 3, we have

9

x dx  ( x 2 ) (6  3)  ( x 2 ) (9  6)
2
x 3 x 6
3
 (3 2 )3  (6 2 )3
 27  108  135
Actual value is given by
9 9
 x 3   93  33 
3
2
x dx       234
 3 3  3 
Truncation error is then
234  135  99
Can you find the truncation error with 4 rectangles?
25

25

13
 9/8/2021

Binary Representation

How a Decimal Number is

Represented

257.76  2 10 2  5 101  7 100  7 101  6 102

1
 9/8/2021

Base 2

 (1 23  0  2 2  1 21  1 2 0 ) 
(1011.0011) 2   1 2 3

4 
  (0  2  0  2  1 2  1 2 ) 10
 11.1875

Convert Base 10 Integer to

binary representation
Table 1 Converting a base-10 integer to binary representation.

Quotient Remainder
11/2 5 1  a0
5/2 2 1  a1
2/2 1 0  a2
1/2 0 1  a3
Hence
(11)10  (a3 a 2 a1 a0 ) 2
 (1011) 2

2
 9/8/2021

Start

Integer N to be
Input (N)10
converted to binary
format

i=0

Divide N by 2 to get
quotient Q & remainder R

i=i+1,N=Q ai = R

No
Is Q = 0?

Yes

n=i
(N)10 = (an. . .a0)2

STOP

Fractional Decimal Number

to Binary
Table 2. Converting a base-10 fraction to binary representation.

Number Number after Number before

decimal decimal
0.1875  2 0.375 0.375 0  a1
0.375  2 0.75 0.75 0  a 2
0.75  2 1.5 0.5 1  a 3
0.5  2 1.0 0.0 1  a 4

Hence
(0.1875)10  (a1a 2 a 3a 4 ) 2
 (0.0011) 2

3
 9/8/2021

Start

Fraction F to be
Input (F)10
converted to binary
format
i  1

Multiply F by 2 to get
number before decimal,
S and after decimal, T

i  i  1, F  T
ai = R

No
Is T =0?

Yes

n=i
(F)10 = (a-1. . .a-n)2

STOP

Decimal Number to Binary

11.187510   ?.? 2
Since
(11)10  (1011) 2
and
(0.1875)10  (0.0011) 2

we have
(11.1875)10  (1011.0011) 2

4
 9/8/2021

All Fractional Decimal Numbers

Cannot be Represented Exactly
Table 3. Converting a base-10 fraction to approximate binary representation.
Number Number
Number after before
decimal Decimal
0.3  2 0.6 0.6 0  a1
0.6  2 1.2 0.2 1  a 2
0.2  2 0.4 0.4 0  a 3
0.4  2 0.8 0.8 0  a 4
0.8  2 1.6 0.6 1  a 5

(0.3)10  (a1a 2 a3 a 4 a5 ) 2  (0.01001) 2  0.28125

Another Way to Look at

Conversion

Convert 11.187510 to base 2

1110  23  3
 23  21  1
 23  21  20
 1 23  0  2 2  1 21  1 20
 10112

10

10

5
 9/8/2021

0.187510  23  0.0625

 2 3  2  4
 0  2 1  0  2  2  1 2 3  1 2  4
 .00112

11.187510  1011.00112
11

11

6
 9/8/2021

Floating Point Representation

Floating Decimal Point : Scientific Form

256.78 is written as  2.5678  10 2

0.003678 is written as  3.678  10 3
 256.78 is written as  2.5678  10 2

1
 9/8/2021

Example

The form is
sign  mantissa 10exponent
or
  m  10 e
Example: For
 2.5678  10 2
  1
m  2.5678
e2
3

Floating Point Format for Binary

Numbers

y    m  2e
  sign of number 0 for  ve, 1 for - ve
m  mantissa 12  m  10 2 
1 is not stored as it is always given to be 1.
e  integer exponent

2
 9/8/2021

Example
9 bit-hypothetical word
the first bit is used for the sign of the number,
the second bit for the sign of the exponent,
the next four bits for the mantissa, and
the next three bits for the exponent

54.7510  110110.112  1.10110112  25

 1.10112  1012
We have the representation as

0 0 1 0 1 1 1 0 1
mantissa exponent
Sign of the Sign of the
number exponent

Machine Epsilon
Defined as the measure of accuracy and found
by difference between 1 and the next number
that can be represented

3
 9/8/2021

Example

Ten bit word

Sign of number
Sign of exponent
Next four bits for exponent
Next four bits for mantissa
0 0 0 0 0 0 0 0 0 0  110
Next
number 0 0 0 0 0 0 0 0 0 1  1.00012  1.0625 10

mach  1.0625  1  2 4

Relative Error and Machine

Epsilon
The absolute relative true error in representing
a number will be less then the machine epsilon
Example
0.0283210  1.11002  2 5
 1.11002  2 0110  2

10 bit word (sign, sign of exponent, 4 for exponent, 4 for mantissa)

0 1 0 1 1 0 1 1 0 0
Sign of the exponent mantissa
Sign of the
number
exponent

1.11002  2 0110  2
 0.0274375
0.02832  0.0274375
a 
0.02832
 0.034472  2  4  0.0625
8

4
 9/8/2021

IEEE 754 Standards for Single

Precision Representation

IEEE-754 Floating Point

Standard
• Standardizes representation of
floating point numbers on
different computers in single and
double precision.

• Standardizes representation of
floating point operations on
different computers.

10

5
 9/8/2021

One Great Reference

What every computer scientist (and even if

you are not) should know about floating point
arithmetic!

http://www.validlab.com/goldberg/paper.pdf

11

IEEE-754 Format Single

Precision

32 bits for single precision

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Sign Biased Mantissa (m)

(s) Exponent (e’)

.
Value  ( 1) s  1 m 2  2 e ' 127

12

12

6
 9/8/2021

Example#1
1 1 0 1 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Sign Biased Mantissa (m)

(s) Exponent (e’)

Value   1  1. m 2  2 e ' 127

s

  1  1.101000002  2(10100010) 2 127

1

  1  1.625  2162127
  1 1.625 235  5.5834 1010

13

13

Example#2
Represent -5.5834x1010 as a single
precision floating point number.
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?

Sign Biased Mantissa (m)

(s) Exponent (e’)

  1  1. ?   2  ?
 1
 5.5834  10 10

14

14

7
 9/8/2021

Exponent for 32 Bit IEEE-754

8 bits would represent
0  e  255
Bias is 127; so subtract 127 from
representation
 127  e  128

15

15

Exponent for Special Cases

Actual range of e
1  e  254
e  0 and e  255 are reserved for special numbers
Actual range of e
 126  e  127

16

8
 9/8/2021

Special Exponents and Numbers

e  0 all zeros
e  255 all ones
s e m Represents
0 all zeros all zeros 0
1 all zeros all zeros -0
0 all ones all zeros 
1 all ones all zeros 
0 or 1 all ones non-zero NaN

17

IEEE-754 Format

The largest number by magnitude

1.1........12  2127  3.40 1038

The smallest number by magnitude
1.00......02  2 126  2.18 1038
Machine epsilon

 mach  2 23  1.19  107

18

18

9
 9/8/2021

Propagation of Errors

Propagation of Errors
In numerical methods, the calculations are not
made with exact numbers. How do these
inaccuracies propagate through the calculations?

1
 9/8/2021

Example 1:
Find the bounds for the propagation in adding two numbers. For example
if one is calculating X +Y where
X = 1.5 ± 0.05
Y = 3.4 ± 0.04
Solution
Maximum possible value of X = 1.55 and Y = 3.44

Maximum possible value of X + Y = 1.55 + 3.44 = 4.99

Minimum possible value of X = 1.45 and Y = 3.36.

Minimum possible value of X + Y = 1.45 + 3.36 = 4.81

Hence
4.81 ≤ X + Y ≤4.99.

Propagation of Errors In Formulas

If f is a function of several variables X 1 , X 2 , X 3 ,......., X n1 , X n

then the maximum possible value of the error in f is

f f f f
f  X 1  X 2  .......  X n 1  X n
X 1 X 2 X n 1 X n

2
 9/8/2021

Example 2:
The strain in an axial member of a square cross-
section is given by
F

h 2E
Given
F  72  0.9 N
h  4  0.1 mm
E  70  1.5 GPa

Find the maximum possible error in the measured

strain.

Example 2:
Solution
72

(4  10 3 ) 2 (70  10 9 )
 64.286  10 6
 64.286 

  
  F  h  E
F h E

3
 9/8/2021

Example 2:
 1  2F  F
 2  3  2 2
F h E h hE E h E
Thus
1 2F F
E  F  3 h  2 2 E
h2E h E h E
1 2  72
  0.9   0.0001
(4 103 ) 2 (70 109 ) (4  103 ) 3 (70 109 )
72
  1.5 109
(4  103 ) 2 (70 109 ) 2
 5.3955 
Hence
 (64.286   5.3955  )
7

Example 3:
Subtraction of numbers that are nearly equal can create unwanted
inaccuracies. Using the formula for error propagation, show that this is true.

Solution
Let
z  x y
Then
z z
z  x  y
x y
 (1)x  (1)y
 x  y
So the relative change is
z x  y

z x y
8

4
 9/8/2021

Example 3:
For example if
x  2  0.001
y  2.003  0.001

z 0.001  0.001

z | 2  2.003 |
= 0.6667
= 66.67%

5
Taylor Series Revisited
 What is a Taylor series?
Some examples of Taylor series which you must have
seen
x2 x4 x6
cos(x)  1     
2! 4! 6!

x3 x5 x7
sin( x)  x     
3! 5! 7!
x2 x3
e  1 x 
x
 
2! 3!

2
 General Taylor Series
The general form of the Taylor series is given by
f x  2 f x  3
f x  h  f x   f x h  h  h 
2! 3!
provided that all derivatives of f(x) are continuous and
exist in the interval [x,x+h]

What does this mean in plain English?

As Archimedes would have said, “Give me the value of the function
at a single point, and the value of all (first, second, and so on) its
derivatives at that single point, and I can give you the value of the
function at any other point” (fine print excluded)

3
 Example—Taylor Series
Find the value of f 6 given that f 4  125, f 4  74,
f 4  30, f 4  6 and all other higher order derivatives
of f x  at x  4 are zero.
Solution:
h2 h3
f x  h  f x   f x h  f x   f x   
2! 3!
x4
h  64  2

4
 Example (cont.)
Solution: (cont.)
Since the higher order derivatives are zero,
22 23
f 4  2  f 4  f 42  f 4  f 4
2! 3!
 2 2   23 
f 6  125  742  30   6 
 2!   3! 
 125  148  60  8
 341
Note that to find f 6 exactly, we only need the value
of the function and all its derivatives at some other
point, in this case x  4

5
 Derivation for Maclaurin Series for ex
Derive the Maclaurin series
x2 x3
e  1 x 
x
 
2! 3!
The Maclaurin series is simply the Taylor series about
the point x=0
h2 h3 h4 h5
f x  h  f x   f x h  f x   f x   f x   f x   
2! 3! 4 5
2 3 4
h5
f 0  h  f 0  f 0h  f 0  f 0  f 0  f 0  
h h h
2! 3! 4 5

6
 Derivation (cont.)
Since f ( x)  e x , f ( x)  e x , f ( x)  e x , ... , f n ( x)  e x and
f n ( 0)  e 0  1

the Maclaurin series is then

(e0 ) 2 (e0 ) 3
f (h)  (e )  (e )h 
0 0
h  h ...
2! 3!
1 1
 1  h  h 2  h3 ...
2! 3!
So,
x 2 x3
f ( x)  1  x    ...
2! 3!

7
 Error in Taylor Series
The Taylor polynomial of order n of a function f(x)
with (n+1) continuous derivatives in the domain
[x,x+h] is given by
h2 hn
f x  h  f x   f x h  f ' ' x     f x   Rn x 
n 
2! n!
where the remainder is given by
 
Rn x 
 x  h
n 1
 n 1
f c
(n  1)!
where
x  c  xh
that is, c is some point in the domain [x,x+h]

8
 Example—error in Taylor series
The Taylor series for e at point x  0 is given by
x

x 2 x3 x 4 x5
e 1 x     
x

2! 3! 4! 5!
It can be seen that as the number of terms used
increases, the error bound decreases and hence a
better estimate of the function can be found.
How many terms would it require to get an
approximation of e1 within a magnitude of
true error of less than 10-6.

9
 Example—(cont.)
Solution:
Using n  1 terms of Taylor series gives error bound of
Rn x  
 x  h
n 1
f n 1 c  x  0, h  1, f ( x)  e x
n  1!

Rn 0 
0  1
n 1
f n 1 c 
n  1!

  1
n 1
ec
n  1!
Since
x  c  xh
0  c  0 1 1
 Rn 0 
e
0  c 1 (n  1)! (n  1)!

10
 Example—(cont.)
Solution: (cont.)
So if we want to find out how many terms it would
require to get an approximation of e within a
1

magnitude of true error of less than 106 ,

e
 106
(n  1)!
(n  1)! 106 e
(n  1)! 106  3
n9
So 9 terms or more are needed to get a true error
less than 106
11