In practice, the applications of Mathematics ultimately results in numerical form. These results may be from the evolution of formulae, solutions of equations or inferences drawn from tabulated data.
Numerical analysis may be described as that branch of Mathematics which provides with convenient methods for obtaining numerical solutions to such problems.
Suppose for a certain experiment the heights of a set of university students are measured. The numbers representing the heights are only approximations, true to two or three decimal places. In general the data represent approximations. Sometimes, the process used to deal with the data is approximate. So, the error in a computed result may be due to errors in the data or errors in the methods or both.
Numerical analysis deals with methods of which errors in computation is reduced to a minimum. With the advent of computers the demand for numerical methods increased rapidly in the applications to engineering and scientific fields. Today numerical methods have become an indispensable tool.
The numbers and e etc written in this form are exact numbers. If is written as 1.732, then it is an approximate number. We can also write as 1.73205, which is a better approximation. However, we cannot write the exact value of by finite number of digits, because has infinite number of decimals. Thus we deal with two types of numbers exact and approximate.
The digits used to express a number meaningfully are called significant figures or significant digits. For instance the digits are all significant digits in any number. But 0 may or may not be a significant figure. If 0 is used to fix the decimal point or to fill places of unknown or discarded digits, then 0 is not a significant figure.
In the number of 0.000567, the significant figures are 5, 6, 7. The zeros are not significant, because they are used to fix the decimal point. But in the number 4.5037 all the five digits are significant. Here 0 is also a significant digit.
In 0.5000, only 5 is the significant figure.
The numbers 0.0028, 0.000035 contain only two significant figures 2, 8, and 3, 5 respectively.
We know is a never ending decimal. To use such a number in a practical computation we must cut down to a usable form such as 1.7 or 1.73 or 1.732 etc. This process of cutting off the extra digits is called rounding off of numbers.
Usually rounding off a number is done according to the following rules.
“To round off a number to n significant figures, discard all the digits to the right of the nth digit, and if the (n + 1)th digit is
A number rounded off according to this rule is said to be correct to n significant places.
For example, the following numbers are rounded off to five significant figures.
23.876345 becomes 23.876
0.876345 becomes 0.87634
4.782250 becomes 4.7822
1.823156 becomes 1.8232
76.69954 becomes 76.700
In numerical analysis, often, we have to perform a sequence of arithmetical operations on numbers. During the computation retain one more figure (or decimal place) than that given in the data and round off after the last operation has been performed. When this practice is followed, no attention is paid to rounding off rule.
In numerical computations we always look for the accuracy of the result obtained.
The size of the error in the computed value is usually expressed in two ways:
∴
and it is denoted by Er.
∴
Sometimes, we express as percentages which will enable us for comparison.
∴ and it is denoted by Ep.
⇒
Note:
If a number is correct to n significant figures, then the relative error is
In other words, a number is an approximation to x to n significant figures if
In numerical analysis errors are classified into two major categories.
Inherent error: It is the error that is present in the statement of the problem before its analysis and solution. The inherent errors arise due to the simplication assumptions that are used in the mathematical formulation of the problem or due to the errors in measurements of the parameters of the problem.
Truncation error: It is due to those errors caused by the method. For example:
If is approximated by the cubic and computed then the error in the result is due to truncating the series.
The remainder after the fourth term in the Maclaurin’s series of is the truncation error.
We shall state some theorems without proof.
Theorem 1.1
If a number is correct to n significant figures and if a1 is the first significant figure, then the relative error is less than i.e.
Theorem 1.2
If x is a number having n decimal digits and if is got by truncating to k digits, then the absolute error is
Absolute error due to rounding off to k digits
Theorem 1.3
If the relative error of any number is then the number is correct to n significant figures.
Example 1
If the number 23.876 is correct to 5 significant figures, then find the relative error.
Solution
Given the number is 23.876.
∴ the first significant figure is 2 and the number of significant digits is 5, n = 5, a1 = 2
∴ the relative error
Example 2
Consider the number 52.43, which is correct to four significant figures. Find Ea and Er. Also find the percentage error.
Solution
Given the number is 52.43. The first significant number is 5 and the number of significant figures is 4
∴
∴ by theorem 1 the relative error
⇒
The nth place is
∴ the absolute error = 0.005
Percentage error = 0.0002 × 100 = 0.02
Example 3
Find the absolute error and relative error in takinπ
Solution
Given
∴
Example 4
Round off the numbers 784652 and 78.4625 to four significant digits and compute Ea, Er, Ep.
Solution
Rounding off to 4 significant figures, we get x1 = 784600
Rounding off to 4 significant figures, we get x1 = 78.4600
Example 5
A person measured a length as 3450 cm, where as its actual length is 3445 cm and another length as 145 cm where as its actual length is 140 cm Compare their absolute and relative errors.
Solution
Though the absolute errors are same for both measurements, their relative errors differ more than 24 times.
Let be a function of several variables and be differentiable. Let Δx1, Δx2,…,Δxn be the errors in the variables so that the error in y is Δy.
We have to find the error relation.
We know that the total differential
∴ the error relation is
The relative error of y is
∴ The maximum value of
Note: Suppose then
The differential relation is
∴ the error relation is
The relative error in y is
Maximum relative error is
Example 6
Let Find the maximum error and relative error in y if x1 = x2 = x3 = 1 and
Solution
Given
∴
∴ the error relation is
∴ the maximum error relation Δy is
But
At the point
and given that
When
∴ maximum error is
= 0.06 + 0.3 + 0.12 = 0.21
Maximum relative error
The general error formula can be used to find error in fundamental operations of arithmetic.
Example 7
Find the sum of the numbers 681.32, 521.7, 94.853, 5.9271, 0.0034, each being correct to its last digit. Also find the absolute error.
Solution
Here 52.7 has one decimal place
The absolute error =
This number is the one with greatest absolute
So, we round off all the numbers to two decimals.
∴ the round off numbers are 681.32, 94.85, 5.93, 0.00
So, their sum S = 681.32 + 521.7 + 94.85 + 5.93 + 0.00 = 1303.8
Ea = sum of absolute errors in each number
Rounding off error is 0.01.
∴ total error in S is 0.07 + 0.01 = 0.08
∴ S = 1303.8 ± 0.08
Example 8
Evaluate the number correct to 4 significant digits and find its absolute and relative error.
Solution
Evaluate the numbers to 4 significant digits using calculator ,
∴
We know that the absolute error in a number correct to 3 decimal places is
∴ the absolute error in the sum of the 3 numbers is
This shows that the sum is correct to 3 significant figures only,
∴
Then
Power series expansion of a function is a very useful technique in theory and applications. The general method for expanding functions into power series is by means of Taylor’s formula.
This is called Maclaurin’s formula.
The remainder after n terms is denoted by Rn and
If as then the series converges.
If we approximate by the first n terms, then the maximum error committed in the truncation is Rn.
Conversely, if the accuracy required is given in advance, the we can find the number of terms n to be taken.
where or then the error committed in truncating is less than the first term neglected
∴
where
If then
If then
Error Rn is maximum when
If maximum. relative
Example 9
Compute truncating after the third term. Find the error.
Solution
where
Put
∴
and
which is correct to 7 decimal places, because in the first significant value is 4 occurring in the 8th place.
Note: Using calculator
Example 10
Find the number of terms n to be taken in the expansion of correct to 7 significant figures, when x = 1.
Solution
The Maclaurin’s series for
Maximum relative
Maximum relative error, when is
For 7 significant figure accuracy, we must have, by theorem 2
This means we have to take 11 terms of the series for 7 significant figure accuracy.
Exercises 1.1
Answers 1.1
3.133.158.36