Index out of bounds Error

Hey I have the following code which I am unable to compile as I'm getting out of bound error.

in the line

du=abs(ukp1(:,:)-u(:,:));

clear all
clc
L=5;
H=5;
delx=0.2;
dely=0.2;
ERRORMAX=0.01;
W=input(' Enter the relaxation parameter \omega < 2 ');
h=zeros(L/delx+1,H/dely+1);
G=h;
G(1,[1:5])=0;
G(1,[6:26])=100;
G([1:15],26)=100;
G([16:26],26)=0;
G(26,:)=0;
G(:,1)=0;
u(1,[1:5])=0;
u(1,[6:26])=100;
u([1:15],26)=100;
u([16:26],26)=0;
u(26,:)=0;
u(:,1)=0;
u([2:L/delx-1],[2:H/dely-1])=input('Enter the guess solution :' );
beta=delx/dely;
A=W;
B=-2*(1+beta^2);
C=A;
error=1;

while error>ERRORMAX
    for i=2:L/delx
        for j=H:dely
            D(i,j)=-(1-W)*(2*(1+beta^2))*u(i,j)-(W*beta^2*(u(i+1,j)+u(i-1,j)));
            h(i,j)=C/(B-A*h(i,j-1));
            G(i,j)=(D(i,j)-A*G(i,j-1))/(B-A*h(i,j-1));
        end
        for k=H/dely:-1:2
           ukp1(i,k-1)=-h(i,k-1)*u(i,k)+G(i,k-1);
        end
    end

du=abs(ukp1(:,:)-u(:,:));
Error=sum(du);
error=sum(Error);
disp(ukp1(:,[1 6 11 16 21 26]));

end

[Walter Roberson]

If you had said which line, I might have tried to answer. But I'm going to bed instead.

[pepsodent]

I just edited my question. I am trying to solve Line Gauss Sidel method for Elliptical PDE (laplace with boundary condition by Implicit method

Answer 1

[Walter Roberson]

You are not getting an "out of bound" error, you are getting a "matrix dimension must agree" error on the subtraction.

>> size(ukp1)

ans =

      25    24

>> size(u)

ans =

      26    26

You initialize u([16:26],26)=0; so u is going to be at least 26 by 26, but your computation for ukp1 has k=H/dely to start and that is 5/0.2 which is 25 . And then you subtract 1 from k when using it as the index into ukp1, so your maximum index based on that is going to be 24. Therefore you are not creating ukp1 as the same size as u, and you cannot subtract the two matrices.

[pepsodent]

Yeah I m getting this error now as I have edited the original code. If I run the Original Code I was getting the out of bound error. So I modified it . Is there any way to fix it ? First it involves Tri diagonal solver and iterating till the solution converges.

[Walter Roberson]

It can be fixed in the "Rewrite the program from scratch" sense. Except the second time write it with comments. And with the matrix bounds calculated and placed into variables instead of being used as expressions like L/delx all over the place. And you should add code that cross-checks the bounds and gives an error message if they are not consistent.