fminbnd not working for storing values in array.

I keep getting this error.

Error in ScratchWork (line 12)

[Vmin(i),DragMin(i)] = fminbnd(Drag,0,15000);

I don't know what wrong.

this is the code

s = 0.6;  
W = linspace(12000,20000,100);
%array of W values given 
Drag = @(V) (0.01*s*V^2) + (0.95/s).*((W./V)^2);

Vmin = zeros(1,length(W));
DragMin = zeros(1,length(W));

for i = 1:1:length(W)     %incrementing for loop by 1 to the length of array W
[Vmin(i),DragMin(i)] = fminbnd(Drag,0,15000);
%calling fminbnd function for each new W value to store in each V and D with respect to ii
end
hold on
grid on
plot(W, Vmin, 'r', W, DragMin, 'g');
legend('Minimum Velocity', 'Minimum Drag');
xlabel('Weight(W)');
title('Sensitivity Analysis'); 

Answer 1

[Walter Roberson]

Your W is a vector, so in

Drag = @(V) (0.01*s*V^2) + (0.95/s).*((W./V)^2);

then you are calculating a vector, but fmincon needs to a scalar.

for i = 1:1:length(W)     %incrementing for loop by 1 to the length of array W
  Drag = @(V) (0.01*s*V^2) + (0.95/s).*((W(i)./V)^2);
  [Vmin(i),DragMin(i)] = fminbnd(Drag,0,15000);
  %calling fminbnd function for each new W value to store in each V and D with respect to ii
end

[Alexa Shumaker]

I forgot about the W(i) in the equation. It works now. Thank you. :D

Answer 2

[Star Strider]

There are two errors.

The first is with respect to your needing to vectorise your ‘Drag’ function, and this will work:

Drag = @(V) (0.01*s*V.^2) + (0.95/s).*((W./V).^2);

and the second is that ‘Drag’ must return a scalar value, and this will work for that:

[Vmin(i),DragMin(i)] = fminbnd(@(V)norm(Drag(V)),0,15000);

Although I strongly suspect that what you intend is to define ‘Drag’ as:

Drag = @(V,W) (0.01*s*V.^2) + (0.95/s).*((W./V).^2);

and then optimise it in the loop as:

[Vmin(i),DragMin(i)] = fminbnd(@(V)Drag(V,W(i)),0,15000);

This does not require that you re-define ‘Drag’ in every iteration of your loop. Just call it with the new value of ‘W(i)’ instead. This is likely much more efficient.

[Walter Roberson]

You do not need to vectorize Drag. For fminbnd

fun is a function that accepts a real scalar x and returns a real scalar f

Redefining Drag each iteration is more efficient, as then fminbnd is only calling through one anonymous function instead of two (Drag and the wrapping anonymous function.)