You are on page 1of 18

Introduction to Numerical

Analysis
Analysis in Mathematics
• Solving a problem through equations.
• Equations reduced to answers.
• Via applying various procedures of algebra, differential equations or
like.

2
Numerical Analysis
• Solving problems via equations.
• Equations reduced to answers.
• Via applying only simple arithmetical procedures.
• Addition.
• Subtraction.
• Multiplication.
• Division.
• Comparison.
• Simple operations performed by computers also.

3
Formally Defining Numerical Analysis
• Numerical Analysis (NA) is concerned with derivation, implementation
and analysis of various numerical methods to find out the
approximate solutions of complex scientific and engineering
problems.

4
What is a Numerical Method?
• Process of reducing the original problem to the series of similar steps so
that the computation becomes automatic.
• This provides estimates closer to exact analytical solutions but obviously
with some error introduced.
• Not a human error.
• Nor a blunder.
• But a discrepancy between exact and approximate values.
• Computers program numerical methods to find approximate solutions to
complex problems at rapid rate with greater degree of accuracy.
• Calculating value of pi up till 100 decimal places is only possible via NM.
• Programmable calculators with limited memory but a ROM having instructions to
find out the most accurate numerical solutions.
5
Software Packages for Performing NA
• IMSL (International Mathematical & Scientific Library) with many
routines written in FORTAN and C++.
• LAPACK (Linear Algebra Package).

6
Accuracy & Uncertainty
• Measuring exact values of physical quantities is very difficult.
• If height of person is measured as 5’8”, how accurate it is:
• How straight the person stood?
• Just woke up or not?
• Hair style?
• Wearing shoes or not?
• ……
• Even if circumstances are specified, there is still the chance of uncertainty.
• Uncertainty in measuring scale.
• Reading floating point values.
• ……
• Finding out uncertainty is more difficult than calculating the actual value.
7
Error Analysis
• Study of errors.
• Evaluation/Estimation of errors.
• Suggest ways to minimize or eliminate errors.

8
Error
•  Difference between the actual value of a quantity and its approximate
computed value.
• If x’ is actual value and x is approximate computed value, then :

• Known as error of computation.

9
Sources of Error
• Gross errors
• Rounding errors.
• Truncation errors.
• Inherent errors.
• Propagated errors.

10
Gross Errors
• Caused by human mistake or computers.
• Can be negligible in some cases but can have devastating effects in
others.
• Misreading of figures. E.g. 34.75 read as 34.57 mistakenly.
• Use of inaccurate formula or algorithm in solving a particular problem.
• Using inaccurate data.
• Can be avoided if proper care be taken.

11
Rounding Errors
• Error introduced by rounding off numbers to limited number of
decimal places.
• Numerical methods when programmed to run on computer can
introduce rounding errors because computers may not represent
some numbers exactly.
• Impractical to specify the distance between two points as 15.2967 metres.
More practical to specify it as 15 metres. Rounding error of 0.2967 is
introduced.
• Rounded off value of pi can introduce errors in certain cases.

12
How smaller Rounding error is produced?
• Following rules of rounding off.
• Given a number, and we want to round it off to only ONE decimal
place:
• If first discarded digit is less than 5, then previous digit remains unchanged.
E.g. 56.44 rounded off to 56.4
• If discarded digit is greater than 5, then previous digit is increased by 1. E.g.
56.46 rounded off to 56.5
• If discarded digit is exactly 5, then previous digit remains unchanged if it is
even and increased by 1 if odd.
• 56.45 rounded off to 56.4
• 56.75 rounded off to 56.8

13
Accumulated Round Off Error
• Rounding error present in final result of numerical computation.
• Not always the some of local round off errors.

14
Truncating Errors
• Replacing infinite series or repetitive items in one series with the
finite one or one with fewer items.
• Most important source of error.
• E.g. Taylor series expansion.

15
Inherent Errors
• Present in problem statement itself before finding its solution.
• Causes:
• Simplified assumptions.
• Data obtained from measurements of certain physical quantities.
• ……..

16
Propagated Errors
• Error in succeeding steps of process due to an earlier error.
• Magnified error at the last stage of process makes this process
unstable.
• For process to be stable, errors must die out at earlier stages.

17
The
End
18

You might also like