MATLAB MATLAB: AN INTRODUCTION

FINITE ELEMENT METHODS ME G512

HISTORY

Created originally as a set of FORTRAN y Cleve Moler, a subroutines in the 70s by scientist in Mathworks Inc. Rewritten in C with extended capabilities in the 80s (including graphics). Distributed by Mathworks, Inc. (www mathworks com) (www.mathworks.com)

ABOUT MATLAB

Interactive Software used in various areas of engineering and scientific applications. applications Not a computer language though it does most of the work of a computer language. E Easy to t l learn. Does not require the in depth knowledge of computer programming like compiling and linking. (Can be looked upon as an advantage and also as a disadvantage). Plotting g Interactive mode may reduce computational speed.

Basic Features
COMMAND WINDOW

When we run MATLAB b Wh by d double bl clicking li ki on MATLAB on th the desktop, it creates one or more windows on the computer monitor. Out of these, COMMAND WINDOW is the place where we interact with MATLAB. The prompt >> is displayed in the command window and when the command window is active, a blinking cursor should appear on the right side of the prompt. Simple Computations can be performed by typing the command and pressing enter. Ex. S= 1+2, fun = sin (pi/4)

Workspace and Help

As we work in the Command Window, MATLAB remembers the commands we enter as well as the variables we create. They reside in the Workspace. Workspace The command whos/ who on the command window shows all he variables in the current session. The help p system y in MATLAB. On the command window type yp help command to get help on a particular command; say for example help LU. In case one doesnt remember the entire command, one can type look for incomplete command name. The command helpwin opens a new window which gives various help options. For example helpwin zeros gives example on zeros. The command demo can be used to demonstrate examples.

Simple Math

Like in a calculator, MATLAB can do simple math. Addition (+), Multiplication(*), Subtraction(-), Division (/ or \), Exponentiation (^) Expressions are evaluated from left to right with p operation p having g highest g p precedence exponentiation followed by multiplication, division having equal precedence, followed by subtraction and addition having equal precedence. Parentheses ( ) can be used to alter these rules of precedence. Variables: case sensitive, 31 characters, start variable name with i h a letter. l Special Variables: ans, pi, i and j etc.

Script m files
Why is this needed?

Ill Illustration: i Creating C i a script i file, fil saving i and d executing i (running) ( i ) it. i
COMMENTS, PUNCTUATION AND ABORTING EXECUTION

Number of commands increases. Re evaluate a number of commands for more than one value of one or more variables. All M-files should have .m extension. For example ex1.m

Semicolon at the end of a command suppresses the printing of computed results. All text after a p percent sign g (%) is taken as a comment statement. Multiple commands can be placed on one line if they are separated by commas or semicolons.

Commas tell MATLAB to display results; semicolons suppress printing. To continue expressions or commands onto additional lines, we use in succession succession. Comment lines and variable names cannot be split between two lines. MATLAB p processing g can be interrupted p at any y moment by y pressing Ctrl-C.

COMPLEX NUMBERS No special handling for complex numbers. (See ex2.m).

Mathematical Functions
sqrt, sin, cos, sinh, asin, acos, exp, log etc. (Note: Can handle Specialized mathematical functions ( Bessel, legendre, beta function etc))

NUMBER DISPLAY FORMATS By default, if a result is an integer, g , it is displayed p y as an integer. g if real number, four digits to right of decimal place is displayed. Can change g this default behaviour (ex3.m). ( )

Script m files-revisited

echo on and echo off- for displaying the commands i the in th command d window i d as th they are read d and d evaluated in the script M file.

Helpful p during g debugging. gg g

input command - When computations for a variety of cases is desired, the input command allows one to prompt for input as the script file is executed.

(See ex4.m)

Array and Array Operations

Simple arrays :

Start with a left bracket bracket, enter the desired values separated by spaces (or commas), and then close it using a right bracket. Colon notation

x=first:last x=first:increment:last linspace command x=linspace (first, last, n) Note: These arrays have one row and multiple columns columns.

See ex5.m.

Vectors and operations

Column vectors
Specify it element by element and separate values with semicolons. N te: Se Note: Separating ti elements ele e t by b spaces e or commas specifies e ifie elements ele e t in i different diffe e t columns, l whereas he e separating elements by semicolons specifies elements in different rows. b=a transposes; length(a) gives length of the vector a a *a a. a is for component wise application of the operator a on a; if a=[1 2 3]; a.*a=[1 4 9] a.^m gives m th power of components of a a.\b gives component wise division of elements of a by elements in b a *b b gives dot product of the vectors a and b, b cross(a,b) gives cross product of 2 three dimensional vectors. unique(a) gives the same values as in array a but without repetitions. sort(a) sorts a in ascending order, for matrices sorts columns in ascending order. order setdiff(a,b) returns vector of values in a but not in b. See ex6.m

Matrix Operations

Formation of matrices
Commas or spaces are used to separate elements in a specific row and semicolon are used to separate individual rows. Alternately, pressing return or enter key while entering a matrix also tells MATLAB to start a new row. row Ex: A=[1 2 3 4; 5 6 7 8; 1 1 2 3] denotes the matrix with first row as 1 2 3 4, second row as 5 6 7 8, third row as 1 1 2 3. To extract a submatrix B consititing of row 1 and 3, columns 2 and 4, we use the command B=A([1 3], [2 4])

Matrix Operations(Contd..)
To interchange rows 1 and 3 of A use the row indices together with colon operator: C=A([3 2 1], :). The colon operator : stands for all columns or all rows. A(:) creates a vector version of a matrix A. To delete a row(column) use the empty vector operator [ ]; for ex. A(:, 2)=[]. To insert the second column back, A=[A(:,1) [2 6 1] A(:,3) A(:,4)] Let d= [1 2 3]; D= diag(d) is a diagonal matrix with diagonals entries in d. (See ex7.m)

Scalar Array Mathematics Addition, subtraction, multiplication(.*), division by a scalar(./) applied to all elements in the array/ matrix. Array Array Mathematics & Matrix functions size(A) gives order of the matrix. A gives transpose of A. A(:, 3) gives the third column of A. A(1, :) gives first row of A. A(1,:) +A(3,:) adds first and third row of A. C=A+B adds two matrices A and B. C=A-B C AB C=A*B inv(A) gives inverse of A. A*inv(A) gives identity matrix. chol(A) h l(A) gives i the h Cholesky Ch l k decomposition d i i of f a positive i i definite d fi i matrix i A. A det(A) gives determinant of matrix A. rank(A) gives rank of matrix A.

eye(n) generates identity matrix of order n. TRACE(A) gives summation of diagonal elements of a square matrix A. zeros(n,m) produces an n x m zero matrix. rand(n,m) d( ) defines d fi an n x m matrix t i having h i random d entries. ti If v is a vector of dimension n, diag(v) produces a diagonal matrix whose elements are components of v. The eigenvalue problem of a matrix is defined by A= where is the eigenvalue i l of f the th matrix t i A and d is i the th corresponding di eigenvector. i t eig(A) i (A) gives the eigenvalues of matrix A and [V,D]=eig(A) produces V matrix whose columns are eigenvectors and the diagonal matrix D whose values are eigenvalues of the matrix A. Th eigenvalue-eigenvector The i l i t decomposition d iti of f a diagonalizable di li bl matrix t i A is i calculated using A=VDV^{-1}: V*D/V; note that right division is used as inverse is expensive. LU decomposition command decomposes a matrix into a combination of upper and lower triangular matrices respectively. respectively [L,U]=lu(A).

QR decomposition command decomposes a matrix into a combination of an orthonormal and upper triangular matrix. [Q,R]=qr(A). (See ex8.m) Though one can write routines to compute adj(A), cofactor (A) for a matrix A, they are hardly used in practical applications involving large scale computations.

Solution of linear equations : Ax=y Assuming the system has a unique solution solution, we can obtain x as: x=inv(A)*y (computes inverse explicitly) x=A\y (uses LU decomposition) (See ex9.m) Computing the CPU time (in seconds) : T=cputime; Operations. TT=cputime-T; TT cputime T;

Loops and Logical statements

for loop for k=array commands end Multiple for loops is also possible. for i=1:100 i 1:100 for j=1:50 for k=1:50 a(i ,j)=b(i,k)*c(k,j)+a(i,j); j j j end end end

Loop and logical statements

While while condition statements end

if, elseif, else, end if condition #1 statement #1 elseif condition #2 statement #2 else statement #3 end

Preallocating Arrays
for and while loops that incrementally increase the size of a data structure each time through the loop can adversely affect performance and memory use. Repeatedly resizing arrays often requires that MATLAB spend extra time looking for larger contiguous blocks of memory and then moving the array into those blocks. You can often improve on code execution time by preallocating the maximum amount of space that would be required for the array ahead of time. The following code creates a scalar variable x, and then gradually increases the size of x in a for loop instead of preallocating the required amount of memory at the start: x = 0; for k = 2:1000 x(k) = x(k-1) + 5; end Change the first line to preallocate a 1 1-by-1000 by 1000 block of memory for x initialized to zero. This time there is no need to repeatedly reallocate memory and move data as more values are assigned to x in the loop: x = zeros(1, 1000); for k = 2:1000 x(k) (k) = x(k-1) (k 1) + 5 5; end

Vectorization
One way to make your MATLAB programs run faster is to vectorize the algorithms you use in constructing the programs programs. Where other programming languages might use for loops or DO loops, MATLAB can use vector or matrix operations. A simple example involves creating a table of logarithms. x=0.01; for k = 1:1001 y(k) = log10(x); x = x + .01; end A vectorized version of the same code is: x = .01:.01:10; y = log10(x); For more complicated code, vectorization options are not always so obvious. (S ex10.m) (See 10 )

Function Subroutines

Special type of m file. Receives input data through input argument list. Returns results to the caller through an output argument list. The general form is as follows:

function[outarg1,outarg2,]=fname(inarg1,inarg2,.) %Comment lines executable code return (See ex11.m and dist2.m)

Plotting T Tools - 2D plots

Simple plotting
( See graph1.m)

Subplot can be used for plotting more than one set of axes in a single figure. Subplots are created with the help of the command subplot(m,n,p) This creates m x n subplots in the current figure, figure arranged in m rows and n columns, and selects subplot p to receive all current plotting commands. For ex, subplot(2,3,4) will create six subplots in the current figure and will make subplot 4 (the lower left one) as the current subplot. (See graph2.m) graph2 m) Plot can be used for plotting more than curve in the same figure. (See graph3.m) loglog to plot using log-log scales, semilogx to plot using logscale along xaxis. i

Plotting tools -3D

A simple 3D plot is shown in graph4.m Another examples p of 3D g graph p in g graph5.m p and g graph6.m p Animation- A sequence of graphs in motion is shown in firstmovie.m Modification in figures, adding arrows, text etc can be done directly in the .fig file using the editing tools built into the Figure Tool Bar. Te conventions Tex con entions can be used sed for legends; for instance \pi. \pi \alpha etc etc.

Polynomials
The command polyval evaluates a polynomial at one or several points. roots(p) p computes p the zeros of p( p(x). ) To check correctness of this result, evaluate p(x) at roots r to get err=polyeval(p,r) close to 0. poly(r) reconstructs polynomials from the roots. conv(p1,p2) gives the coefficients of the polynomial given by the product of two polynomials whose coefficients are contained in the vectors p1 and p2. polyder(p) l d ( ) gives i the th coefficient ffi i t of f p(x). ( ) polyint(p) gives coefficients of integral of p(x). (See ex12.m)

fzero for zeros of non-linear

equations

fzero(fun,x0) adopts Dekker-Brent method to compute the zero of the function fun, where fun can be defined as an inline i li function f ti or in i a .m file. fil Syntax: f=inline(fun f inline( fun ,variable); variable ); x0 x0=.. [alpha,res,flag,iter]=fzero(fun,x0) p is the root, , res=residual, , iter= number of iterations, , alpha flag is negative means fzero cant find the zero.

Symbolic Tools
Tool box symbolic provides the commands of diff int taylor which allows us to obtain the analytical expressions of the derivative, indefinite integral and the Taylor polynomial respectively of a given function function.
(See ex13.m)

Numerical Integration g (composite ( p midpoint quadrature formula)

Let I(f)=ab f(x) dx where f is a continuous function in [a,b]. Partition P titi [ [a,b] b] i into t subintervals bi t l Ik=[x [ k-1,xk], ] k=1,,M. k 1 M In each subinterval Ik, we approximate f by a constant polynomial p y interpolating p g f at the middle p point of Ik. M Impc(f)=H * f((xk+xk-1)/2), H=(b-a)/M k 1 k=1 (see ex14.m and midpointc.m for this, inline functions and feval)

Trapezoidal rule

Let I(f)=ab f(x) dx where f is a continuous function in [a,b]. P titi [ Partition [a,b] b] i into t subintervals bi t l Ik=[x [ k-1,xk], ] k=1,,M. k 1 M In each subinterval Ik, we approximate f by a linear polynomial p y interpolating p g f at the nodes xk k-1 1 , xk.
M

Itc(f)=H/2 *
k=1

f(xk)+f(xk-1), H=(b-a)/M

(see numint2.m for use of inbuilt trapz function)

Gauss-Lobatto quadrature formula

We use an inbuilt function quadl to compute an integral using Gauss-Lobatto quadrature formula. The function is defined as an inline object. Call quadl(fun,a,b). (fun-function, a ,b limits of integration) g ) The number of subintervals is not requested as it is automatically computed so as to ensure that quadrature error is below 10-3. quadl(fun,a,b,tol) specifies a different tolerance level. (see ex16.m)

Newtons method for non-linear systems

To compute the roots of nonlinear systems of the following form:

f1(x1,x2,.,xn)=0, f2(x1,x2,.,xn)=0, f3(x1,x2,.,xn)=0, ---------------------, fn(x ( 1,x2,.,xn)=0, ) 0 where fi are nonlinear functions, using Newtons method, use zero=fsolve(fun zero=fsolve( fun ,x0) x0) This allows the computation of one zero of the nonlinear system defined through fun starting from x0 as initial guess. (See ex17.m)

General Instructions for Numerical Analysis Programs

Declare the array variables to zero. Use clear all statement in the beginning of the code. Use long g double for real arrays. y Declare the matrix as sparse in case you are working with a sparse matrix to save memory and computation time. For ex. a sparse matrix of dimension 10* 100 is initialized by A = sparse(10,100). Huge identity matrices can be also declared as sparse matrix by using A A=speye(100). speye(100). A full matrix A can be transformed to a sparse matrix by using the command A=sparse(A). In case, a sparse matrix needs to be transformed to a full matrix, use the command dA A=full(A). f ll(A) Use vectorization as far as possible. Assign space in memory for arrays and matrices in the beginning.