CFD

You might also like

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 4

Starting our study on computational fluid dynamics we must first understand the term.

It simply implies the use of computers in order to represent and solve the partial
differential equations that govern the fluid flow, gathering the experimental data of
flow and post processing it into the computer(that strictly does not fall in this
domain).
In order to understand what is the meaning of solve we can simply state that to find
the value of the independent variable that satisfies an mathematical(algebraic or
trigonometric or whatever form the equation may be in)equation or to find such a
value/s that leave a residue if they do not completely satisfy the equation.
We can also have equations of the form as f( u(x),x)=0.
This has a solution which is not a number but another simpler equation.if we want to
solve CFD problems on computers then we must have a way to represent equations or
functions to the computer.
CLOSED FORM SOLUTIONS
ANALYTIC SOLUTIONS
Such solution means that the solution is NOT an infinite series.Functions of such
kinds are polynomial functions, transcendental functions and the linear combination
of these. However there are some functions which cannot be represented in form a
integral form(like integrating e raised to x^2) or differential form without a closed
form solution. So we need some or the other way of representing these kind of
functions on computers.
REPRESENTING MATHEMATICAL ENTITIES ON THE COMPUTER
Starting from the basics , as in calculus we will first learn how to represent real
numbers on the computer.
Numbers on a computer are stored as binary numbers using 0 or 1.
For example if we have two bits so we can have 4 distinct combinations from them
that can be used to represent 4 different symbols or numbers on the computer.
Similarly for 4 bits we get 2^4=16 distinct symbols and so on.
Thus we see however large the number of bits we select we can represent only a
subset of the set of Integers. So there lies the problem as to how we are going to
represent real numbers.
We can scale our problem so that the solution always lies in (0,1).
e.g. we can represent numbers ranging from -2^31 to 2^31 after dividind them by a
suitable power of 10 so that all of them lie in (0,1).
This method is called fixed point arithmetic.it is very fast but sometimes inaccuracy
results due to the loss of the MSB as illustrated
0.125-0.12=0.005
Taking numbers as ,1/3,1/4.1/9 and seeing what all of them have recurring and
non recurring representations.In binary representations we cannot represent recurring
decimal numbers.Recurring decimal representation in decimal gives a truncation error
and it leaves the LSB as doubtful.

FLOATING POINT REPRESENTATION


We work on

You might also like