3/15/21

KA20602 NUMERICAL ANALYSIS

Chapter 1
INTRODUCTION TO NUMERICAL ANALYSIS
AHMAD NURFAIDHI RIZALMAN (PhD)
Civil Engineering Programme
, No. 66 Level 2 Block A, Faculty of Engineering
( 088 – 320000 (ext. 3033) + [email protected]
Prepared by: Ahmad Nurfaidhi Rizalman

CONTENT

q Introduction to Numerical Method

q Significant Figures, Accuracy and Precision
q Round-off Error
q Truncation Error
q Error Analysis

Sample Questions
Solution to Exercise

Prepared by: Ahmad Nurfaidhi Rizalman

1
 3/15/21

INTRODUCTION

Prepared by: Ahmad Nurfaidhi Rizalman

INTRODUCTION
There are three (3) methods to solve engineering problems:
Experiment Simulation Mathematical Model

Outcome is expressive but Define and solve the Expresses the problems in
involves lots of equipment problems graphically to mathematical terms (i.e.
and facilities which are characterize the behavior of formula). Solutions are
expensive. systems. derived analytically or
numerically.

Prepared by: Ahmad Nurfaidhi Rizalman

2
 3/15/21

INTRODUCTION
There are two (2) types of mathematical models
ANALYTICAL NUMERICAL
• Produce exact solutions. • Produce approximate solutions.
• Limited to linear problems with simple • Capable to solve nonlinear problems
geometry and low dimensionality. with complex shapes and high
dimensionality.
• Thus, it involves large numbers of
tedious arithmetic calculations.
• This method has increased dramatically
due to the development of digital
computers and software, i.e. MATLAB,
Excel, SPSS, and etc.

Prepared by: Ahmad Nurfaidhi Rizalman

INTRODUCTION
There are eight (8) types of numerical methods, including
a) Nonlinear Equations (Roots of Equations).
b) Linear Equations
c) Optimization
d) Curve Fitting
e) Interpolation
f) Numerical Differentiation and Integration
g) Ordinary Differential Equations
h) Partial Differential Equations

Prepared by: Ahmad Nurfaidhi Rizalman

3
 3/15/21

INTRODUCTION
a) Nonlinear Equations (Roots of Equations)
These problems are concerned with the value of
a variable or a parameter that satisfies a single
nonlinear equations. (i.e. solve 𝑓 𝑥 = 0 for 𝑥).

They are valuable in engineering design contexts

where it is often impossible to explicitly solve
design equations for parameters.

Prepared by: Ahmad Nurfaidhi Rizalman

INTRODUCTION
b) Linear Algebraic Equations
They are also concerned with values that satisfy
equations. But, in this approach, a set of values is
sought that simultaneously satisfies a set of linear
algebraic equations.

Such problems include mathematical modelling

of large systems of interconnected elements such
as structures and fluid networks.

For instance, given the 𝑎’s and the 𝑐′s, solve

𝑎!!𝑥! + 𝑎!"𝑥" = 𝑐!
𝑎"!𝑥! + 𝑎""𝑥" = 𝑐"
for the 𝑥’s.

Prepared by: Ahmad Nurfaidhi Rizalman

4
 3/15/21

INTRODUCTION
c) Optimization
The problem involve determining value(s) of an
independent variable that correspond to a “best,”
or optimal, value of a function.

For instance, in the figure above, the optimization

involves identifying maxima and minima.

Prepared by: Ahmad Nurfaidhi Rizalman

INTRODUCTION
d) Curve Fitting
Curve fitting (or regression) is employed where
there is a significant degree of error associated
with the data.

Experimental results are often of this kind. It is

done by deriving a single curve that represent the
general trend of the data without matching any
individual points.

Prepared by: Ahmad Nurfaidhi Rizalman

10

5
 3/15/21

INTRODUCTION
e) Interpolation
Interpolation is used where the objective is to
determine intermediate values between
relatively error-free data points.

Such is usually the case for tabulated information.

It is done by fitting a curve directly through the
data points and use the curve to predict the
intermediate values.

Prepared by: Ahmad Nurfaidhi Rizalman

11

INTRODUCTION
f) Numerical Differentiation and Integration
Numerical integration is determination of the
area under a curve.

For instance, the determination of the centroids

of an object. In addition, it plays an important
role in the solution of differential equations.

For example, to find the area of the curve, the

following equations are employed.
$
𝐼 = ∫# 𝑓 𝑥 𝑑𝑥

Prepared by: Ahmad Nurfaidhi Rizalman

12

6
 3/15/21

INTRODUCTION
g) Ordinary Differential Equations
ODE are of great significance in engineering
practice. For instance, the acceleration of a falling
body (rate of change of velocity).

Two types of problems are addressed: initial-

value and boundary-value problems. It also
include the computation of eigenvalues.

Given
%& (&
%'
≈ ()
= 𝑓(𝑡, 𝑦)
Solve for 𝑦 as a function of 𝑡.
𝑦*+! = 𝑦* + 𝑓 𝑡* , 𝑦* ∆𝑡

Prepared by: Ahmad Nurfaidhi Rizalman

13

INTRODUCTION
h) Partial Differential Equations
They are used to characterize engineering
systems where the behavior of a physical quantity
is couched in terms of its rate of change with
respect to two or more independent variables.

The common approaches include finite difference

method and finite element method.

For instance, solve for 𝑢 as a function of 𝑥 and 𝑦.

,! - ,! -
,' ! + ,&!
= 𝑓(𝑥, 𝑦)

Prepared by: Ahmad Nurfaidhi Rizalman

14

7
 3/15/21

SIGNIFICANT FIGURE, ACCURACY AND PRECISION

Prepared by: Ahmad Nurfaidhi Rizalman

15

SIGNIFICANT FIGURE, ACCURACY AND PRECISION

• Significant figures (or digit) is defined as numbers that can be used with confidence.
They correspond to the number of certain digits plus one estimate digit.

• The concept of significant figures will have relevance to the definition of accuracy
and precision.

Prepared by: Ahmad Nurfaidhi Rizalman

16

8
 3/15/21

QUESTION 1.1
How many significant figures in the following numbers?
a) 0.01688
b) 0.001688
c) 0.0001688
d) 0.000000001688
e) 37900
f) 3.79 x 104
g) 3.790 x 104
h) 3.7900 x 104

Prepared by: Ahmad Nurfaidhi Rizalman

17

SOLUTIONS TO QUESTION 1.1

(a) – (d) all have four significant figures.
(e) 37900 may have three, four, or five significant figures, depending on whether the
zeros are known with confidence.

(f), (g) and (h) are three, four, or five significant figures, respectively.

Prepared by: Ahmad Nurfaidhi Rizalman

18

9
 3/15/21

SIGNIFICANT FIGURE, ACCURACY AND PRECISION

Increasing accuracy Accuracy refers to how close a
particular measurement to the
Inaccurate Accurate
exact value, while precision
and and refers to consistency of
repeated measurement.
Increasing precision

imprecise imprecise

Thus, the best measurement is

accurate and precise.

From the figure, the arrow on

Inaccurate Accurate each target can be taught as
and precise and precise the prediction of a numerical
technique, and the bull’s eye is
the exact value.

Prepared by: Ahmad Nurfaidhi Rizalman

19

SIGNIFICANT FIGURE, ACCURACY AND PRECISION

• In numerical methods (analysis), the results obtained should be sufficiently accurate to
meet the requirements of a particular engineering problems. They should also be precise
enough for adequate engineering design.

• But, since the numerical methods to obtain approximate solutions, it is expected the
results consist of certain errors. There are two (2) major form of unavoidable numerical
errors – (1) Round-off Errors, and (2) Truncation Errors.

Prepared by: Ahmad Nurfaidhi Rizalman

20

10
 3/15/21

ROUND-OFF ERRORS

Prepared by: Ahmad Nurfaidhi Rizalman

21

ROUND-OFF ERROR
• Round-off error arise when the numbers having limited significant figures are used
to represent exact numbers.

• Some numbers cannot be expressed by a fixed number of significant figures such as

!
.
= 0.3333333 … , 𝑒 = 2.71828182 … . and 𝜋 = 3.1415926.

• In the calculations, round-off errors occur when the digits to the right of the decimal
are discarded. Say for and 𝜋 computer can store 4 significant figures. Therefore,
a) Chopping-off à 𝜋 = 3.1415
b) Rounding-off à 𝜋 = 3.1416

Prepared by: Ahmad Nurfaidhi Rizalman

22

11
 3/15/21

QUESTION 1.2
Perform each of the computation below (i) exactly, (ii) using four-digit chopping
arithmetic, and (iii) four-digit rounding arithmetic.
. 0
a) / + 1
. 0
b) /
−1

Prepared by: Ahmad Nurfaidhi Rizalman

23

SOLUTIONS TO QUESTION 1.2

Solutions to 1.2a)
. 0 !!
(i) /
+ 1
= 1 2"
. 0
(ii) /
+ 1 = 1.261904 … ≈ 1.261
. 0
(iii) /
+ 1 = 1.261904 … ≈ 1.262

Solution for 1.2b)

. 0 !/
(i) /
− 1
= − 2"
. 0
(ii) − = −0.404761 … ≈ −0.4047
/ 1
. 0
(iii) /
− 1 = −0.404761 … ≈ −0.4048

Prepared by: Ahmad Nurfaidhi Rizalman

24

12
 3/15/21

QUESTION 1.3
Given an irrational number 𝜋 = 3.141592654 …, find the five-digit representations of 𝜋
by
a) Chopping-off technique,
b) Rounding-off technique.

Prepared by: Ahmad Nurfaidhi Rizalman

25

SOLUTIONS TO QUESTION 1.3

a) Chopping-off: 𝜋 ≈ 0.31415×10!
b) Rounding-off: 𝜋 ≈ 0.31416×10!

Prepared by: Ahmad Nurfaidhi Rizalman

26

13
 3/15/21

TRUNCATION ERROR

Prepared by: Ahmad Nurfaidhi Rizalman

27

TRUNCATION ERROR
• Truncation error arise when approximation are used to represent exact mathematical
procedure or expression. It originates from the technique of replacing a complicated
function with a truncated Taylor series. For example, the infinite Taylor series of
'! '" '# '$ '%
𝑒 ' = 1 + 𝑥 + "! + .! + 2! + 0! + ⋯ + 4!

might be replaced with just first five terms

'! '" '#
𝑒 ' ≈ 1 + 𝑥 + "! + .! + 2!

• Truncation error is introduced whenever a number computational uses a formula

involving discrete values. Unlike round-off error, which is controlled by the hardware and
computer language, truncation error is under the control of the programmer.

Prepared by: Ahmad Nurfaidhi Rizalman

28

14
 3/15/21

QUESTION 1.4
!/" !
Given that ∫5 𝑒 ' 𝑑𝑥 = 0.544987104184. Find the approximation by replacing the
! '# '& ''
integrand 𝑓 𝑥 = 𝑒 ' with the truncated Taylor series 𝑒 ' ≈ 1 + 𝑥 " + "!
+ .!
+ 2!
.
Give the answer in 3 decimal places using round-off technique.

Prepared by: Ahmad Nurfaidhi Rizalman

29

SOLUTIONS TO QUESTION 1.4

!/" ! !/" '# '& ''
∫5 𝑒 ' 𝑑𝑥 ≈ ∫5 1 + 𝑥" + "!
+ .!
+ 2!
𝑑𝑥 Term-by term integration, yields
':!/"
'" '$ '( ')
= 𝑥+ .
+ 0("!) + /(.!) + 9(2!)
':5
! ! ! ! !
= " + "".. + "$.0("!) + "(./(.!) + ").9(2!)
! ! ! ! !
= " + "2 + ."5 + 0./1 + !!509"
= 0.54498672
≈ 0.545.

Prepared by: Ahmad Nurfaidhi Rizalman

30

15
 3/15/21

ERROR ANALYSIS

Prepared by: Ahmad Nurfaidhi Rizalman

31

ERROR ANALYSIS
In the practice of numerical analysis, it is important to note that computed solutions are
not exact mathematical solutions. Perfect accuracy in most computational processes is
impossible. Therefore, it is imperative to make certain approximations, thus called as
errors.

The error in a computed quantity is defined by the relationship between exact solution (or
true value) and the approximate value. It is expressed as follows:

Error = 𝑡𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒

= 𝑥) − 𝑥#

Absolute Error = 𝑡𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑣𝑎𝑙𝑢𝑒

= 𝑥) − 𝑥#

Prepared by: Ahmad Nurfaidhi Rizalman

32

16
 3/15/21

ERROR ANALYSIS
The relative error is a measure of the error in relation to the size of the true being sought.
Mathematically, it is expressed as:
<==>=
Relative Error = )=-< ?#@-<
'" A'#
= '$

True percent
)=-< ?#@-< A#BB=>'*C#)< ?#@-<
relative error, 𝜀) = )=-< ?#@-<
× 100%

Approximate percent
B=<D<4) #BB=>'*C#)*>4 AB=<?*>-D #BB=>'*C#)*>4
relative error, 𝜀# = B=<D<4) #BB=>'*C#)*>4
×100%

Prepared by: Ahmad Nurfaidhi Rizalman

33

QUESTION 1.5
Find the absolute error and relative error for values given below:
a) Let 𝑥E = 3.141592 and 𝑥F = 3.14
b) Suppose that the true value is 1000000 and our estimation is 999996.
.
c) The answer obtained from analytical solution is ; meanwhile the answer
"05555
from from numerical method is 0.000009.

Prepared by: Ahmad Nurfaidhi Rizalman

34

17
 3/15/21

SOLUTIONS TO QUESTION 1.5

a) Absolute error = 𝑥) − 𝑥#
= 3.141592 − 3.14
= 0.001592
'" A'#
Relative error = '"
5.55!09"
= ..!2!09"
≈ 0.000507

b) Absolute error = 1000000 − 999996

=4
2
Relative error = !555555
= 0.000004
Prepared by: Ahmad Nurfaidhi Rizalman

35

SOLUTIONS TO QUESTION 1.5

.
c) Absolute error = "05555 − 0.000009
= 0.000003
5.55555.
Relative error = 5.5555!"
= 0.25

Prepared by: Ahmad Nurfaidhi Rizalman

36

18
 3/15/21

QUESTION 1.6
Suppose that you have the task of measuring the length of a table and a handphone,
from measurement, you got 110.0 𝑐𝑚 and 11.0 𝑐𝑚, respectively. Find
a) Absolute error,
b) Relative error (in percentage), for each case.

Prepared by: Ahmad Nurfaidhi Rizalman

37

SOLUTIONS TO QUESTION 1.6

a) The absolute error for measuring the table is:
𝑥) − 𝑥# = 110.0 − 109.5 = 0.5 𝑐𝑚.

The absolute error for measuring the handphone is:

𝑥) − 𝑥# = 11.0 − 10.5 = 0.5 𝑐𝑚.

b) The relative error for measuring the table is

' A' 5.0
= "' # = !59.0 ≈ 0.000456 = 0.0456%.
"

The relative error for measuring the handphone is

' A' 5.0
= "' # = !5.0 ≈ 0.0476 = 4.76%.
"

Prepared by: Ahmad Nurfaidhi Rizalman

38

19
 3/15/21

SAMPLE QUESTIONS

Prepared by: Ahmad Nurfaidhi Rizalman

39

QUESTION 1.7
How many significant figures in the following numbers?
a) 929
b) 0.0929
c) 0.00929
d) 0.00000000929
e) 512000
f) 5.12 x 105
g) 5.120 x 105
h) 5.12000 x 105

Solution:
a) - d): Three significant figures
e) may have three, four, five or six significant figures.
f), g) and h) are three, four, or six significant figures, respectively.
Prepared by: Ahmad Nurfaidhi Rizalman

40

20
 3/15/21

QUESTION 1.8
Perform each of the computation below (i) exactly, (ii) using three-digit chopping arithmetic,
and (iii) three-digit rounding arithmetic.
2 !
a) 0
+ .
2 !
b) ×
0 .

Solution:
"
a)(i) 1 !0 (ii) 1.13 (iii) 1.13
b)(i) 4⁄15 (ii) 0.266 (iii) 0.267
Prepared by: Ahmad Nurfaidhi Rizalman

41

QUESTION 1.9
Find the representation of the following numbers by chopping–off technique and rounding-off
technique.
a) 97450677
b) e = 2.718281828
c) 0.2553075606
d) mn = 1.67492716 x 10-27

Solution:
a)(i) 97450600 (ii) 97450700
b)(i) 2.71828 (ii) 2.71828
c)(i) 0.255307 (ii) 0.255308
d)(i) 1.67492 ×10A"/ (ii)1.67493 ×10A"/
Prepared by: Ahmad Nurfaidhi Rizalman

42

21
 3/15/21

QUESTION 1.10
!/2 ' !
Given that ∫5 𝑒 𝑑𝑥 = 0.2553074606 . Find the approximation by replacing the integrand
! ! '% '&
𝑓 𝑥 = 𝑒 ' with the truncated Taylor series 𝑒 ' ≈ 1 + 𝑥 " + "!
+ .!
. Give the answer in 3
decimal places using round-off technique.

Solution:
0.255
Prepared by: Ahmad Nurfaidhi Rizalman

43

QUESTION 1.11
Find the absolute error and relative error for values given below:
a) Let 𝑥E = 2.7182818 and 𝑥F = 2.7182.
b) Suppose that the true value is 97450 and the estimation is 97000.
c) The answer obtained from analytical solution is 0.000764, meanwhile the answer from
numerical method is 0.000760.

Solution:
a) 0.0000818, 0.00003009
b) 450, 0.004618
c) 0.000004, 0.0005236
Prepared by: Ahmad Nurfaidhi Rizalman

44

22
 3/15/21

QUESTION 1.12
If 𝑥 = 2 is approximated by 𝑥̅ = 1.94, find the true error and relative error.

Solution:
True error = 0.06, Relative error = 0.03
Prepared by: Ahmad Nurfaidhi Rizalman

45

QUESTION 1.13
.
If 𝑥 = 𝜋 is approximated by 𝑥̅ = /, find the actual error and relative percent error.

Solution:
True error = 2.71302, Relative percent error = 86.36%

Prepared by: Ahmad Nurfaidhi Rizalman

46

23
 3/15/21

QUESTION 1.14
Given a function 𝑓 𝑥 = sin(𝑥) . Compute the following Taylor’s polynomial at 𝑥 = 0 for
a) 𝑃! 𝑥 , 𝑃" 𝑥 , 𝑃.(𝑥), and 𝑃G(𝑥),
b) If the above Taylor’s polynomial is used to approximates 𝑠𝑖𝑛 27> , then find the true error
𝐸) of each of these approximations.

Prepared by: Ahmad Nurfaidhi Rizalman

47

SOLUTIONS TO QUESTION 1.14

a) Since 𝑥5 = 0, then
! ! !
𝑃4 𝑥 = 𝑓 0 + 𝑓 H 0 𝑥 + " 𝑓 HH 0 𝑥 " + .! 𝑓 . 0 𝑥 . + ⋯ + 4! 𝑓 4 (0)(𝑥)4
!
= ∑4I:5 I! 𝑓 I (0)(𝑥)I .

A systematic way to calculate these polynomials is by constructing a table for the derivatives.
! ! ! !
Therefore 𝑃! 𝑥 = 0, 𝑃" 𝑥 = 𝑥, 𝑃. 𝑥 = 𝑥 − 1 𝑥 . and 𝑃G 𝑥 = 𝑥 − 1 𝑥 . + !"5 𝑥 0 − 0525 𝑥 /.

b) The true value sin 275 = 0.4540. To use these polynomials to approximate sin 27> , we
J
need to take 𝑥 in radians as 𝑥 = 27> × "/' = 0.4712. Hence, the solution is
𝑠𝑖𝑛 27> − 𝑃! 27> = 0.4540
𝑠𝑖𝑛 27> − 𝑃" 27> = 0.0172
𝑠𝑖𝑛 27> − 𝑃. 27> = 0.0002
𝑠𝑖𝑛 27> − 𝑃G 27> = 3.4662 × 10A0

Prepared by: Ahmad Nurfaidhi Rizalman

48

24
 3/15/21

QUESTION 1.15
Find the second and fourth order Taylor’s polynomial, 𝑃"(𝑥) and 𝑃2(𝑥), centered at 𝑥 = 𝑥5, for
each of the following functions
a) 𝑓 𝑥 = 𝑒 ' , 𝑥5 = 1
(
b) 𝑓 𝑥 = 𝑥 ) , 𝑥5 = 5

Prepared by: Ahmad Nurfaidhi Rizalman

49

SOLUTIONS TO QUESTION 1.15

a)
!
𝑃" 𝑥 = 𝑒 ! 1 + 𝑥 − 1 + " (𝑥 − 1)"
1 1 1 1
𝑃0 𝑥 = 𝑒 ! 1 + 𝑥 − 1 + (𝑥 − 1)"+ (𝑥 − 1).+ (𝑥 − 1)2+ (𝑥 − 1)0
2 6 24 120

b)
1 2
𝑃" 𝑥 = 5!/. 1 + 5A! 𝑥 − 5 − 5A"(𝑥 − 5)"
3 18
𝑃" 𝑥
1 2 5 A. 5
= 5!/. Q1 + 5A! 𝑥 − 5 − 5A"(𝑥 − 5)"+ 5 (𝑥 − 5).− 5A2 𝑥 − 5 2
3 18 162 1458
11
+ 5A0(𝑥 − 5)0R
104976

Prepared by: Ahmad Nurfaidhi Rizalman

50

25
 3/15/21

QUESTION 1.16
Compute the following Taylor’s polynomial, 𝑃4 (𝑥) centered at 𝑥 = 0 of 𝑦 = cos 𝑥 " for
a) 𝑃!(𝑥)
b) 𝑃"(𝑥)
c) 𝑃2(𝑥)

Prepared by: Ahmad Nurfaidhi Rizalman

51

SOLUTIONS TO QUESTION 1.16

𝑛 𝑓 4 (𝑥) 𝑓 4 (0)
0 cos 𝑥 " 1
1 −2𝑥 sin 𝑥 " 0
2 −4𝑥 " cos 𝑥 " − 2 sin 𝑥 " 0
3 8𝑥 . sin 𝑥 " − 12𝑥 𝑐𝑜𝑠 𝑥 " 0
4 16𝑥 2 cos 𝑥 " + 48𝑥 " sin 𝑥 " − 12 cos 𝑥 " −12

!" 2
Therefore 𝑃2 𝑥 = 1 − "2 𝑥

Prepared by: Ahmad Nurfaidhi Rizalman

52

26
 3/15/21

QUESTION 1.17
A civil engineer has measured the height of a 15 floor building as 3940 m and the working
height of each beam as 35 m while the true values are 3945 m and 40 m, respectively.
Compare their true error and true percent relative error.

Prepared by: Ahmad Nurfaidhi Rizalman

53

SOLUTIONS TO QUESTION 1.17

True error in measuring the height of the building =5m
True percent relative error in measuring the height of the building = 0.13%

True error in measuring the height of the beam =5m

True percent relative error in measuring the height of the beam = 13%

Although the true errors are the same, the true percent relative errors differ by 100%. Hence,
there is something wrong in the measurement of the height of the beam by using true error.
Thus, it should be done more accurately.

Prepared by: Ahmad Nurfaidhi Rizalman

54

27