Sie sind auf Seite 1von 23

Chapter 1

NUMERI CAL METHODS


1.1 I ntroduction
Q1. What is numerical Analysis/methods?
Ans.: Branch of science that deals with
Study of methods, techniques or algorithms for obtaining
approximate solutions for complex mathematical problems.
Numerical analysis can be considered as
1. Science: to solve complicated mathematical problems
2. Art: to develop an algorithm and apply it to a particular
problem.
Q2. Why do we need Numerical Methods?
Ans.: Two ways to solve mathematical problems:
Closed form solutions.
Ex1.: Integration by parts, by partial fraction, by substitu-
tion of variables.
Ex2.: Solution dierential equations.
Ex3.: Solution of small set of linear equations, (2 to 5 equa-
tions).
1
2 CHAPTER 1. NUMERICAL METHODS
Numerical Solutions. Not all problems have closed form so-
lution.
Ex1.: nt = c
d|
Ex2.:
o
Z
0
r
4
c
a
(c
a
1)
2
dr
1.2 Power Series Representations
Some functions can be written as
)(r) = c
0
+ c
1
r + c
2
r
2
+ c
3
r
3
+ + c
a
r
a
(1.1)
Examples:
Ex1.:
c
a
= 1 + r +
r
2
2!
+
r
3
3!
+
r
4
4!
+ +
r
a
n!
+
=

X
l=0
r
l
/!
, |r| <
=

X
l=1
r
(l1)
(/ 1)!
, |r| <
Ex2.:
sin(r) = r
r
3
3!
+
r
5
5!

r
7
7!
+
=

X
l=0
(1)
l
r
2l+1
(2/ + 1)!
, |r| <
=

X
l=1
(1)
l1
r
2l1
(2/ 1)!
, |r| <
Q: On what bases can we write these series? Or, how can we nd the
coecients c
0,
c
1
, , c
a
in equation (1.1)?
Ans.: Using Taylor Series.
1.3. TAYLOR SERIES EXPANSION 3
1.3 Taylor Series Expansion
Taylor series expansion of )(r) about a real point c is given as
)(r) = )(c) + )
0
(c)(r c) +
)
00
(c)
2!
(r c)
2
+
)
000
(c)
3!
(r c)
3
+
)
(4)
(c)
4!
(r c)
4
+ (1.2)
=

X
l=0
)
(l)
(c)
/!
(r c)
l
(1.3)
where )
(l)
(c) =
d
l
)(r)
dr
l
evaluated at r = c.
When c = 0, Series (1.3) is called Maclaurin Series.
Example 1 : Let )(r) = sin(r), nd the Maclaurin series?
Solution:
)(r) = sin r )(0) = 0
)
0
(r) = cos r )
0
(0) = 1
)
00
(r) = sin r )
00
(0) = 0
)
000
(r) = cos r )
000
(0) = 1
)
(4)
(r) = sin r )
(4)
(0) = 0
Then
sin r = )(0) + )
0
(0)r +
)
00
(0)
2!
r
2
+
)
000
(0)
3!
r
3
+
)
(4)
(0)
4!
r
4
+
= 0 + r 0
r
3
3!
+ 0 +
= r
r
3
3!
+
=

X
l=0
(1)
l
r
2l+1
(2/ + 1)!
=

X
l=1
(1)
l1
r
2l1
(2/ 1)!
4 CHAPTER 1. NUMERICAL METHODS
Q1: Is this method practical?
Ans.: No.
Q2: What is the solution, then?
Ans.: Taylor series with Remainder.
1.4 Taylor Series with Remainder
The practical form of Taylor series is
)(r) =
a
X
l=0
)
(l)
(c)
/!
(r c)
l
+ 1
a+1
(1.4)
where
1
a+1
=
)
(a+1)
()
(n + 1)!
(r c)
a+1
is called the remainder or the least upper bound for the error. [r, c]
is chosen such that |)
(a+1)
()| is maximum. Note that if n is large, the
error is small and vise-versa.
Example 2 )(r) = c
a
can be written in Maclaurin Series form as
c
a
= 1 + r +
r
2
2!
+
r
3
3!
+
r
4
4!
+ +
r
a
n!
+
r
a+1
(n + 1)!
+
r
a+2
(n + 2)!
+
= 1 + r +
r
2
2!
+
r
3
3!
+
r
4
4!
+ +
r
a
n!
+ 1
a+1
=
a
X
l=0
r
l
/!
+ 1
a+1
where
1
a+1
=
)
(a+1)
()
(n + 1)!
r
a+1
but )
(a+1)
() = c

,then
1
a+1
=
c

(n + 1)!
r
a+1
[0, r].
1.4. TAYLOR SERIES WITH REMAINDER 5
So if r and n are given we can calculate an upper bound for the error.
Some times, the error and r are given and it is required to nd the
number of terms.
Example 3 Determine how many terms are needed to compute c cor-
rect to 15 decimal places (rounded)?
Solution: In this question, )(r) = c
a
where r = 1 and c = 0.
c
a
= 1 + r +
r
2
2!
+
r
3
3!
+
r
4
4!
+ +
r
a
n!
+ 1
a+1
=
a
X
l=0
r
l
/!
+ 1
a+1
1
a+1
=
)
(a+1)
()
(n + 1)!
r
a+1
, [0, 1].
Since
)(r) = c
a
)
0
(r) = c
a
)
00
(r) = c
a
.
.
.
)
(a+1)
(r) = c
a
so )
(a+1)
() = c

. Then
1
a+1
=
c

(n + 1)!
r
a+1
, [0, 1].
It is required to nd n such that the error is less or equal 0.510
15
.
Therefore we have the following
c

(n + 1)!
r
a+1
0.5 10
15
where r = 1 (given) and is chosen in the interval[0, 1] such that

)
(a+1)
()

is maximum. When = 1, c

is maximum.
6 CHAPTER 1. NUMERICAL METHODS
Then
c
1
(n + 1)!
(1)
a+1
0.5 10
15
c
(n + 1)!
0.5 10
15
Solving for n by trial and error by substituting dierent integer
values for n in
c
(n + 1)!
0.5 10
15
n 1Ho
10 0.680986 9099 10
7
0.5 10
15
20 0.532047 7 10
19
<< 0.5 10
15
16 0.764233 3164 10
14
0.5 10
15
17 0.424574 0647 10
15
< 0.5 10
15
18 0.223460 0341 10
16
<< 0.5 10
15
We stop here and take n = 17,but the number of terms= n+1 =
18 terms (Why?).
In conclusion, to evaluate c
1
correct to 15 decimal places we evaluate
the series c =
17
P
l=0
1
l!
1.5 Taylor series in terms of ( +)
In Taylor series expansion of )(r) about point c with remainder
)(r) =
a
X
l=0
)
(l)
(c)
/!
(r c)
l
+ 1
a+1
(1.5)
replace r by r + / and c by r we get
)(r + /) =
a
X
l=0
)
(l)
(r)
/!
(r + / r)
l
+ 1
a+1
=
a
X
l=0
)
(l)
(r)
/!
/
l
+ 1
a+1
1.6. ALTERNATING SERIES 7
where r [c, /], and / is any value such that (r + /) [c, /]. What
is the use of this form? Read Example 4 in your text book. Also this
form will be used extensively when we talk about the dierentiation of
functions in topic 6.
1.6 Alternating Series
If c
0
c
1
c
2
c
3
c
a
0 for all n, and lim
a
c
a
= 0, then
the alternating series
c
0
c
1
+ c
2
c
3
+
converges, that is
lim
a
a
X
l=0
(1)
l
c
l
= lim
a
o
a
= o
where
o , Total sum of the series, n
o
a
, The sum of the rst n + 1 terms.
Moreover, the error due truncation is bounded by (important)
|o o
a
| c
a+1
(1.6)
This means that the error is less or equal to the magnitude of the rst
omitted term in the alternating series.
Remark 1 Alternating Series is special cases of Taylor Series.
Remark 2 A series is called alternating if the following two conditions
are satised
1. The magnitude of the terms are decreasing.
2. The signs of the terms are changing, i.c., +, , +, ,and so on.
8 CHAPTER 1. NUMERICAL METHODS
Example 4 Let ))r) = ln r.
(a) Does )(r) has Maclurin Series expansion?
(b) Find the Taylor series expansion of )(r) about point 1.
(c) How many terms are needed to approximate ln(1.1) with error less
or equal to 10
4
?
Solution:
(a) For )(r) = lnr, the Maclaurin series of )(r) does not exist about
point 0, because )(0), )
0
(0), are not dened..
(b) The expansion of )(r) about point c = 1 is derived as follows:
)(r) = ln r )(1) = 0
)
0
(r) = r
1
)
0
(1) = 1
)
00
(r) = r
2
)
00
(1) = 1
)
000
(r) = 2r
3
)
000
(1) = 2.1
)
(4)
(r) = 3.2r
4
)
(4)
(1) = 3.2.1
)
(5)
(r) 4.3.2r
5
)
(5)
(1) = 4!
Then
ln r = 0 + 1(r 1)
1
2!
(r 1)
2
+
2!
3!
(r 1)
3

3!
4!
(r 1)
4
+
4!
5!
(r 1)
5

= (r 1)
(r 1)
2
2
+
(r 1)
3
3

(r 1)
4
4
+
(r 1)
5
5

=
a
X
l=1
(1)
l1
(r 1)
l
/
+ 1
a+1
(c) This is an alternating series if (r 1)
l
< /, i.c., (r 1) < 1. Since
r = 1.1, then r 1 = 1.1 1 = 0.1.]. Therefore, the error is less
or equal to the rst omitted term, i.e.
|1
a+1
| =
(r 1)
a+1
n + 1
1.6. ALTERNATING SERIES 9
and
(r 1)
a+1
n + 1
10
4
where r = 1.1
(0.1)
a+1
n + 1
10
4
10
(a+1)
n + 1
10
4
10
(a+1)
(n + 1)10
4
n + 1 10
(a+1)+4
n + 1 10
a+3
By trial and error n = 3. (The number of terms here = n = 3
only).
Example 5
(a) Obtain the rst three nonzero terms of the Taylor series expansion
of co-(r) about c = :,2. Now obtain the general expression that
includes all the terms.
(b) A function )(r) is approximated using the following innite series
)(r) = r
r
2
3
+
r
3
5
2

r
4
7
3
+
r
5
9
4

How many terms are needed to compute )(1) correct to 6-decimal
places (rounded)?
Solution:
(a)
c =

2
)(r) = cos(r) )(

2
) = 0
)
0
(r) = sin(r) )
0
(

2
) = 1
)
00
(r) = cos(r) )
00
(

2
) = 0
)
000
(r) = sin(r) )
000
(

2
) = 1
)
(4)
(r) = cos(r) )
(4)
(

2
) = 0
)
(5)
(r) = sin(r) )
(5)
(

2
) = 1
10 CHAPTER 1. NUMERICAL METHODS
)(r) = )(c) + )
0
(c)(r c) +
)
00
(c)
2!
(r c)
2
+
)
000
(c)
3!
(r c)
3
+ +
)
l
(c)
/!
(
= 0 (r
:
2
) + 0 +
1
3!
(r
:
2
)
3
+ 0
1
5!
(r
:
2
)
5
= (r
:
2
) +
1
3!
(r
:
2
)
3

1
5!
(r
:
2
)
5
=

X
l=1
(1)
l
1
(2/ 1)!
(r
:
2
)
2l1
=

X
l=0
(1)
l+1
1
(2/ + 1)!
(r
:
2
)
2l+1
(b):
)(r) = r
r
2
3
+
r
3
5
2

r
4
7
3
+
r
5
9
4
=

X
l=0
(1)
l
r
l+1
(2/ + 1)
l
, if / starts from .c:o
=

X
l=1
(1)
l1
r
l
(2/ 1)
l
, if / starts from onc
The error is given by 1
a+1
c
a+1
and|c
a+1
| =

r
a+1+1
(2(n + 1) + 1)
a+1

r
a+2
(2n + 3)
a+1

=
1
(2n + 3)
a+1
, for r = 1. To nd number of terms,
we rst nd n such that
1
(2n + 3)
a+1

1
2
10
6
, (6 dicimal places)
1.6. ALTERNATING SERIES 11
By trial and error we get:
n = 1
1
(2n + 3)
a+1
= .0 4 0.5 10
6
= 2
1
(2n + 3)
a+1
= 0.2 9155 10
2
0.5 10
6
= 3
1
(2n + 3)
a+1
= 0.1 5242 10
3
0.5 10
6
= 4
1
(2n + 3)
a+1
= 0.6 2092 10
5
0.5 10
6
= 5
1
(2n + 3)
a+1
= 0.2 0718 10
6
< 0.5 10
6
So n = 5 and the number of terms = 6, since the summation
starts from 0.
Example 6 The Taylor Series expansion of )(r) about point c
is given as )(r) = )(c) +)
0
(c)(rc) +
)
00
(c)
2!
(rc)
2
+ +
)
()
(c)
l!
(r
c)
l
+ =
P

l=0
)
()
(c)
l!
(r c)
l
(a) Determine the Taylor series expansion about point c =
:
4
for the function )(r) =sin 3r
(b) A function )(r) is approximated by )(r) = r
a
2
3
4
+
a
3
5
4

a
4
7
4
+ .Find an upper bound for the error if 4 terms are used
to approximate )(1)?
(c) Determine the minimum number of terms needed to
compute )(1) correct to 8 decimal places for the function in
(b).
Solution (a) c =
:
4
)(r) = sin 3r )(
:
4
) =
1

2
)
0
(r) 3 cos 3r )
0
(
:
4
) = 3
1

2
)
00
(r) = 3
2
sin 3r )
00
(
:
4
) = 3
2 1

2
)
000
(r) = 3
3
cos 3r )
000
(
:
4
) = 3
3 1

2
)
(4)
(r) = 3
4
sin 3r )
(4)
(
:
4
) = 3
4 1

2
12 CHAPTER 1. NUMERICAL METHODS
Then
f(x) = f(c) +f
0
(c)(x c) +
f
00
(c)
2!
(x c)
2
+ +
f
(k)
(c)
k!
(x c)
k
+
=
1

1 3(r
:
4
) 3
2
(r
:
4
)
2
2!
+ 3
3
(r
:
4
)
3
3!
+ 3
4
(r
:
4
)
4
4!

=
1

X
k=0
(1)
l
3
2l
(r
:
4
)
2l
(2/)!
+

X
k=0
(1)
l+1
3
2l+1
(r
:
4
)
2l+1
(2/ + 1)!

(b)
f(x) = x
x
2
3
4
+
x
3
5
4

x
4
7
4
+
=

X
l=1
(1)
l1
r
l
(2/ 1)
4
=

X
l=0
(1)
l
r
l+1
(2/ + 1)
4
4 terms, means n = 4 if / starts from one. Then
1
a+1
= |c
a+1
| =
r
a+1
(2(n + 1) 1)
4
1
4+1
= |c
4+1
| =
r
4+1
(2(4 + 1) 1)
4
=
1
4+1
(9)
4
= 0.15242 10
3
or
4 terms, means n = 3 if / starts from zero. Then
1
a+1
= |c
a+1
| =
r
a+1
(2(n + 1) + 1)
4
1
4+1
= |c
4+1
| =
r
3+1
(2(3 + 1) + 1)
4
=
1
3+1
(9)
4
= 0.15242 10
3
1.6. ALTERNATING SERIES 13
(c) c
1
2
10
8
. Then
1
a+1
= |c
a+1
| c
|c
a+1
| 10
8
r
a+1
(2(n + 1) 1)
4
10
8
, if / starts from 1.
1
(2n + 1)
4
10
8
(2n + 1)
4
10
8
2n + 1 100
n
100 1
2
= 49.5 = 50
So the number of terms is equal to n = 50 terms.
Or
|c
a+1
| 10
8
r
a+1
(2(n + 1) + 1)
4
10
8
, if / starts from 0.
1
(2n + 3)
4
10
8
(2n + 3)
4
10
8
2n + 3 100
n
100 3
2
= 48.5 = 49
So the number of terms is equal to n + 1 == 49 + 1 = 50 terms.
14 CHAPTER 1. NUMERICAL METHODS
1.7 More Examples on Taylor Series
Example 7 Find the Maclaurin series expansion of )(r) = c
a
2
.
Example 8 Find the smallest non-negative integer : such that the
Taylor series expansion of )(r) =

r 1 about point : exist.


Example 9 Determine the 1
c|
two nonzero terms of the series expan-
sion about .c:o of
(c) c
cosa
(/) sin(cos r)
Example 10 In the Taylor series expansion of )(r) = 3r
2
7 +cos r,
what is the coecient of r
2
? Answer: 2.5.
Example 11 Find the Taylor series expansion of sin r in powers of
(r

6
)?
Example 12 For small r, the approximation sin r ' r is used. For
what values of r is this a good approximation if the accuracy required
is to be less or equal 0.5 10
14
. Repeat this problem if sinr is approx-
imated by sin r ' r
a
3
3!
.
Example 13 Given )(r) = c
a
.
(a) Find the error term if Maclaurin series is assumed?
(b) If r = 1.2, nd the value of that gives maximum error?
(c) Find the number of terms if the error is to be less or equal to 0.01?
1.8 Exercise
Q1. If cos r =
P

l=0
(1)
l a
2
2l!
, how many terms are needed to compute
cos r for |r| <
1
2
accurate to 12 decimal places (rounded)?(n 6,
but number of terms 7, since / starts from zero ).
1.8. EXERCISE 15
Q2. A function ) is dened by )(r) =
P

l=1
(1)
l a

l
4
. Determine the
minimum number of terms needed to compute )(1) with error
less than 10
8
? (n 99).
Q3. What is the least number of terms required to compute : as 3.14
(rounded) using the series : = 4
4
3
+
4
5

4
7
+ ? (Note: First
nd the c
a
term, then 1
a+1
= ...). [n 399, number of terms
400 ].
Q4. Consider the function )(r) = c
2a
+ c
3a
.
(a) Expand ) (r) into Taylors Series around point 0 ?
(b) If )(1) is approximated by 5 terms of the series, nd an upper
bound of the error in this approximation?
(c) How many terms are needed if the error is to be less than
10
5
?(answer: 17, 42.65)
Q5. Given the function )(r) = cos r + sin r,
(a) Find the rst four nonzero terms of the Taylor series expan-
sion of )(r) around point zero?
(b) What is the maximum error if four terms are used in approx-
imation )(1)?
Q6. The Taylor Series expansion of )(r) about point c is given as
)(r) = )(c)+)
0
(c)(rc)+
)
00
(c)
2!
(rc)
2
+ +
)
()
(c)
l!
(rc)
l
+ =
P

l=0
)
()
(c)
l!
(r c)
l
(a) Determine the Taylor series expansion about zero for the func-
tion )(r) = c
oa
+
1
r + 1
?
(b): A function ) is approximated by )(r) = 1
a
2
3
4
+
a
4
5
4

a
6
7
4
+
. Determine the minimum number of terms needed to compute )(1)
correct to 8 decimal places.
16 CHAPTER 1. NUMERICAL METHODS
Chapter 2
SOURCES OF ERRORS
AND ERROR ANALY SI S
2.1 Number Representation
Numbers are represented in dierent bases as follows:
1. Decimal System:
The base is = 10, the digits used are:0, 1, 2, , 9.
2. Binary System:
The base is = 2, the digits used are:0, 1.
3. Octal System:
The base is = 8, the digits used are:0, 1, 2, , 7.
4. Hexadecimal System:
The base is = 16, the digits used are:0, 1, 2, , 9, , !, C, 1, 1.
In any system, a real number r is written as
r =
a
X
l=0
c
l

l
+

X
l=1
/
l

l
(2.1)
17
18CHAPTER 2. SOURCES OF ERRORS AND ERROR ANALYSIS
where
a
X
l=0
c
l

l
is the integer part and

X
l=1
/
l

l
is the fractional part.
See the summary in the appendix for conversion from one system
to another.
2.2 Normalized oating-Point
Any real number r can be written as
r = c
o
c
1
c
2
c
a
/
1
/
2
/
n
(2.2)
such as r = 29.3121. But in scientic notation r is written as
r = 0.d
1
d
2
d
3
10
a
, d
1
6= 0 (2.3)
In computer terminology, this representation of r is called normalized
oating-point.
In general, if r is a decimal number it is written as
r = : 10
a
,
1
10
: < 1 (2.4)
where
: is called mantissa
n is called exponent
For binary system
r = 2
n
,
1
2
< 1 (2.5)
i.e.,
= (0./
1
/
2
)
2
2
n
, /
1
= 1 and /
i
= 1 or 0 for i 1 (2.6)
In some computers |:| (127)
10
.
Q: Can any number be stored exactly in computers?No. (Why?)
Some numbers may have nite expansion in one system but have
innite expansion in another system.
2.3. MEASURING THE ERROR 19
Example 14 The number (
1
10
)
10
= (0.1)
10
= (0.00011 0011 0011 )
2
.
This number has a nite expansion in the Decimal system but has in-
nite expansion in the binary system.
In computers, a number is stored in a word. A word has a nite
length, i.e., a nite number of bits.
Numbers that can be represented (stored) in computers are divided
into three cases:
1. machine numbers: numbers that can be stored (t) exactly in
computer word, (No error)
2. not machine numbers: numbers that can not be stored (t)
exactly in computer word, (Rounding/Truncation error).
3. There are some numbers that can not be stored in computers
because they are small or big.
If the number is small, the computer will give under-ow error.
If the number is big, the computer will give over-ow error
Example 15 If the largest/smallest number that can be represented by
a computer is 2
129
, then
r = 2
571
gives oc: )/on error. In this case, the computer
will stop execution of the program.
r = 2
571
gives nndc: )/on error. In this case, the computer
will set r = 0 and will continue.
Some computers use double words or triple words to deal with
small or big numbers.
2.3 Measuring the error
The error is measured by
The absolute error: it is dened as
Absolute error = |True value-Approximate value| (2.7)
20CHAPTER 2. SOURCES OF ERRORS AND ERROR ANALYSIS
The relative error: it is dened as
Relative error =
|True value-Approximate value|
True value
(2.8)
The percentage error: it is dened as
Relative error 100 (2.9)
2.4 Errors in Arithmetic Operations
Let r be a real number and dene )/(r) to the oating point machine
number, i.e. )/(r) is the approximation of r that is stored in the
computer. Let o denote the relative error in r when it is stored in the
computer, i.e.,
o =
|r )/(r)|
r
(2.10)
Then we have three cases:
1. r is a machine number, i.e., r will be stored exactly in the com-
puter. That is o = 0 and r = )/(r).
2. r is not a machine number but within the range of computer, in
this case r will not be stored exactly in the computer but it will
be rounded or truncated.. )/(r) will be stored instead of r. The
relation between r and )/(r) is given by r = )/(r)(1 + o).
3. r is not a machine number and outside the range of computer, in
this case r will not be stored and an overow / underow error
will occur.
Let denote any operation in the set

+, , ,,

, then for
any two machine numbers r and
)/(r ) = (r )(1 + o) (2.11)
Here r = )/(r) and = )/() since r and are machine numbers. If
r and are not machine numbers but within the range of computer
2.4. ERRORS IN ARITHMETIC OPERATIONS 21
word, we have
)/(r) = r(1 + o)
)/() = (1 + o)
)/(r ) = ()/(r) )/())(1 + o)
= {r(1 + o) (1 + o)} (1 + o)
Example 16 Let r, and . be real numbers. What is the relative error
in computing r. if
(a) r, and . are machine numbers.
(b) r, and . are not machine numbers but within the range of com-
puter word.
Solution:
(a) Since all variables are machine numbers, then
)/(r) = r
)/() =
)/(.) = .
and
)/(r) = r(1 + o)
and
)/()/(r)..) = ()/(r)..)(1 + o)
= (r(1 + o)..)(1 + o)
= ..(1 + o)
2
= ..(1 + 2o + /iqc: tc::-)
' r.(1 + 2o), ignore higher powes of o
So the relative error in computing r. is 2o.
22CHAPTER 2. SOURCES OF ERRORS AND ERROR ANALYSIS
(b) Since all variables are not machine numbers but within the range
of computer word, then
)/(r) = r(1 + o)
)/() = (1 + o)
)/(.) = .(1 + o)
and
)/(r) = )/(r))/()(1 + o)
= r(1 + o)(1 + o)(1 + o)
= r(1 + o)
3
and
)/()/(r)..) = {)/(r).)/(.)} (1 + o)
=

r(1 + o)
3
.(1 + o)

(1 + o)
= r.(1 + o)
5
' r.(1 + 5o)
The relative error is approximately equal to 5o.
2.5 Loss of Signicance
In the real number r = 0.321498 10
5
, the digit 3 is called the most
signicant digit (MSD) and the digit 8 is called the least signicant
digit(LSD).
Exact numbers can be represented with any number of signicant
digits.
While measured numbers have limited number of signicant dig-
its.
The number of signicant digits in measured numbers depend on
the accuracy of the machine used in measurements.
2.6. COMPUTER-CASED LOSS OF SIGNIFICANCE 23
Denition 17 A number r
o
is said to approximate a number r
|
to n
signicant digits if n is the largest integer that satisfy
|r
|
r
o
|
r
|
<
1
2
10
a
(2.12)
Example 18 Let r
|
= 3.141592 and r
o
= 3.142, then
|r
|
r
o
|
r
|
=
|3.141592 3.142|
3.141592
= 0.129870 4606 10
3
<
1
2
10
3
Then r
o
approximate r
|
correctly to 3 signicant digits. While if r
o
=
3.145,then
|r
|
r
o
|
r
|
=
|3.141592 3.145|
3.141592
= 0.108480 0318 10
2
<
1
2
10
3
which means that r
o
approximate r
|
correctly to 2 signicant digits only.
2.6 Computer-Cased loss of Signicance
Just have an idea of this topic from your textbook.

Das könnte Ihnen auch gefallen