Matrix Operation Returns NaN

Hello, I am working on a project where I analyze temperature change with a few different methods. One of them calls for a set of equations to be done repetitively, using a matrix. My code that is relavant is below:

 while flag == 0
    time=1;
    for i=2:1:n
        if i==n
           %This can be ignored for now, the else statement below is the issue.
           %Tnew(i) = T(i) + (alpha).*(0.01).*(-1/deltax).*(((T(i)-T(i-1)./(deltax))-b+                                            (h/k.*deltax).*(T(i)-TAir)));
 else
            TF(i) = T(i)-2*T(i)+T(i-1);
            Tnew(i) = T(i) + ((alpha).*(0.01)).*((TF(i)/deltax)- b.*(T(i)-TAir));
        end
        error(i) = abs((Tnew(i)-T(i))/T(i))*100;
    end
    for i=1:1:n
        T(i) = Tnew(i);
    end
    time = time + 1;
    if max(error)<0.01
       flag=1;
    end
    if time>100
       disp('Error');
       break 
    end
 end

The code in the if else statement is giving me trouble. I pulled the TF part out, to analyze it. For some reason, it returns NaN for every member of the matrix. I may be missing something small, but after looking through it for hours, I am unsure where the error lies.

Thank you for any help

Answer 1

[Jan]

Use the debugger to finde the problem. Set a breakpoint at the line, which defines TF and inspect the parts in the command window. Find out, which term creates the NaNs. Perhaps deltax is zero?

We cannot do this for you, because we do not have your data and the code does not run by copy&paste.

Notes:

• Do not shadow the important function error by a variable.
• This code:

for i=1:1:n
  T(i) = Tnew(i);
end

should be beautified:

T(1:n) = Tnew(1:n);

or if applicable:

T = Tnew;

[Daniel Platt]

Thank you for the input, I did try using the debugger and could not figure it out. I guess a question I have is, is there anything inherently wrong using

TF(i) = T(i)-2*T(i)+T(i-1); to accomplish this?

This is definitely where the problem occurs. I know the deltax value is not zero, it is checked upstream in the code. I also understand that without all of the code, it is difficult to diagnose/cannot run. It is rather lengthy, and has some information I am not necessarily comfortable distributing.

[Star Strider]

So long as you’re in the loop, no. But it could be simplified to:

TF(i) = T(i-1) - T(i);