Why am I getting inf even though there is no divide by zero?

When I execute the folowing code if I take the value of 'k' as 2 or 3, the answer for any value for 'alpha' array becomes inf if the quotient is greater than 1. I am also not able to take the sum of the elements of array 's' as that gives this error: 'Array indices must be positive integers or logical values.'. Can someone tell me what is causing this?

A=[14.5463 14.224 14.2043]
B=[2940.46 2945.467 2972.64]
C=[273.22 224 209]
X=[0.25 0.35 0.4]
P=80
n=3
to=0
Pi_sat=[0 0 0]
k=1
alpha=[0 0 0]
s=[0 0 0]
y=[0 0 0]
for i=1:n
    Ti_sat(i)=B(i)/(A(i)-log(P))+C(i) 
    to=to+X(i)*Ti_sat(i)
end
while 1
    for i=1:n
        Pi_sat(i)=A(i)-B(i)/(to+C(i))
        alpha(i)=Pi_sat(i)/Pi_sat(k)
    end
    s=X.*alpha
    Pk_sat=P/sum(s)
    tnew=B(k)/(A(k)-log(Pk_sat))+C(k)
    e=abs((tnew-to)/to)
    if(e<0.001)
        break
    end
    to=tnew
end
fprintf('\nTemperature=%f',tnew)
for i=1:n
    y(i)=X(i)*Pi_sat(i)/P
    fprintf('\ny(%i)=%f',i,y(i)) 
end

Answer 1

[Matt J]

I am also not able to take the sum of the elements of array 's' as that gives this error: 'Array indices must be positive integers or logical values

You have created a variable names "sum" somewhere, so that now Matlab is treating it as a variable.

Other than that, the infs are indeed caused by a divide by zero condition, see below.

while 1
    for i=1:n
        Pi_sat(i)=A(i)-B(i)/(to+C(i));
        
        alpha(i)=Pi_sat(i)/Pi_sat(k);

        if isinf(alpha(i))
            Pi_sat([i,k])
            keyboard
        end
    end

end

ans = 1×2

10.8783 0

Unable to run the 'fevalJSON' function because it calls the 'keyboard' function, which is not supported for this product offering.

[Evan]

Thank you so much. I am new to matlab so I didn't realize that such a small aspect would affect my final answer by this much.

Answer 2

[Torsten]

while 1
        Pi_sat=A-B./(to+C)
        alpha=Pi_sat/Pi_sat(k)
    ...
        

instead of

while 1
    for i=1:n
        Pi_sat(i)=A(i)-B(i)/(to+C(i))
        alpha(i)=Pi_sat(i)/Pi_sat(k)
    end
    ...

Answer 3

[Steven Lord]

To determine why an Inf is generated even though you don't believe there should be any division by zero performed by your code, set an error breakpoint to stop when a NaN or Inf value is generated then run your code. If MATLAB stops at a line of code, check the values on that line to see if they contain the values you expect them to contain. If they don't, set breakpoints on earlier and earlier lines of code and rerun your code to allow you to look at the variables as your code runs and hopefully determine when and why the differences between what you expect and what's actually stored in the variables are introduced.

[Jai]

But, when the value is too large to get overflowed it will return "inf". how to handle this please help

Thanks.