Function value at starting guess must be finite and real.

Hi everybody. I created a code as under the this text.

hiz(1)=18.54;
ra=36:1:71

for g=2:1:numel(ra)+1
         WR=C*cos(egim(ra(g-1)))
         WS=C*sin(egim(ra(g-1)))
         WL=@(hiz1) C*((hiz1+hiz(ra(g-1)))/2).^2
         WB=@(hiz1) C*( hiz1-hiz(ra(g-1))).^2/(mes(ra(g))-mes(ra(g-1)))
         P=C*((yuk(ra (g)))- (yuk(ra(g-1))))
         T=C*mes((ra(g))-mes(ra(g-1)))
         TOTAL= @(hiz1) WR+WL(hiz1)+WS+WB(hiz1)
         TOTAL_LOSS= @(hiz1)  TOTAL(hiz1)*(mes(ra(g))-mes(ra(g-1)))
         hiz1_function=@(hiz1) C*hiz(ra(g-1))^2-TOTAL_LOSS(hiz1)+T-C*hiz1.^2-P
         x1 = [-100  100];
for k1 = 1:length(x1);
    hiz_rts(g-1,k1) = fzero(hiz1_function, x1(k1))
end

end

C is the constant value of the algorithm. When this code starts to run, MATLAB gives me Function value at starting guess must be finite and real. error. How can i achieve this problem? Yesterday, i used this code without for loop and it didn't give me an error message.

Thanks..

Answer 1

[Guillaume]

Well, the error message is clear, the function you're passing to fzero must return a finite and real value at the starting point. It's not the case for one of your starting x1 value, hence the error.

We can't tell you which value nor the cause since we don't know the values of mes, hiz, egim. Possibly mes(ra(g))-mes(ra(g-1)) is 0 in some case, and WB ends up being Inf. You should be able to test it yourself by looking at the actual values.

In any case, your g loop will fail at the last iteration since ra(g) is not valid when g = numel(ra)+1

[Volkan Yangin]

mes(ra(g))-mes(ra(g-1)) is the gap between two points and it is constant as 10. So, it never equals to 0. I have deleted the +1 at the g=numel(ra)+1, but the problem didn't achieved.

And i have tried to run the code by changing x1, but it gives me same error message.

[Volkan Yangin]

hiz_rts has 2 values on Workspace as -11,7286001575164 18,4808283485018. Algorithm finds the hiz_rts at g=2, but 3,4,5. etc can't run.

[Guillaume]

Well, look at the output of

hiz1_function(-100)
hiz1_function(100)

What is it? It's either going to be +Inf, -Inf, NaN or a complex value. Work back through your function terms to see which one causes that.

I thought it might be a division by zero, but it was just a guess. As I said, without the values of the unknown variables, there's no way for us to tell.

[Volkan Yangin]

Yes, they go to NaN with -100 and 100. Thank you Guillaume. I am working on parameters now.

[Volkan Yangin]

Guillaume, the problem on the WB, as you think. Algorithm don't know hiz(ra(g-1)) values. hiz(1) has been known only. I think, if i will be able to add the new value of hiz, algoritm runs correctly.