FMINCON requires all values returned by functions to be of data type double

Hello, I'm coding a project for optimization in probability engineering but I got an error with the function fmincon. I'm a newbie using Matlab so don't really know what the error is all about.

Here is my main function:

% Initialize parameters
d0 = [0.2,0.2,3]; %starting point, b=h=0.1, l=2
lb = [0.1,0.1,2]; %lower bounds for design variables
ub = [0.4,0.8,20]; %upper bounds for design variables
option = optimset('Display','iter'); %set options to show the optimization history 
d = fmincon('Obj_fun',d0,[],[],[],[],lb,ub,'constr_fun',option); %call the optimizer

% Analysis the optimal point
t = d(1);
h = d(2);
l = d(3);
obj = t*h*l;
c = constr_fun(d); %calculate the constraint function
disp(['The optimal point = ',num2str(d)]);
disp(['The objective function = ',num2str(obj)]);
disp(['The constraint functions = ',num2str(c)]);

And here are my 2 others functions for using fmincon:

% Construction function
function [c,ceq]=constr_fun(d)
b = d(1);
h = d(2);
l = d(3);
Pf_mean = 300; Pf_std = 30; Pt_mean = 450; Pt_std = 50; Ps_mean = 100e4; Ps_std = 10e4;
syms u1 u2 u3
Pf_sample = Pf_mean+u1*Pf_std;
Pt_sample = Pt_mean+u2*Pt_std;
Ps_sample = Ps_mean+u3*Ps_std;
c = Ps_sample - ((6*Pf_sample*l/b*h^2).^2 + (Pt_sample/h*b^2*(3+1.8*b/h)).^2).^0.5; %constraint equation
ceq = []; %no equality constraint

% Objective function
function obj = Obj_fun(d)
b = d(1);
h = d(2);
l = d(3);
obj = b*h*l;

Answer 1

[Alan Weiss]

Generally, in order for us to be able to help you need to report the entire error message that you receive, everything in red. Additionally we would need to see your function call and function definitions.

However, there are some obvious programming errors here. You declare some symbolic variables. Don't do that. You don't need them. Most likely, symbolic variables are variables that you want to optimize. They should be the variables in your d vector.

Also, you need to write a separate function for your nonlinear constraints. Something like this:

function [c,ceq] = confun(d)
b = d(1);
h = d(2);
l = d(3);
% Give an expression in d for Ps_sample, etc. Then
c = Ps_sample - ((6*Pf_sample*l/b*h^2).^2 + (Pt_sample/h*b^2*(3+1.8*b/h)).^2).^0.5; %constraint equation
ceq = []; %no equality constraint
end

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation