How to use fmincon for dynamic optimisation?

Essentially, I'm trying to find the optimum temperature profile that maximises a value (Mean crystal size), subject to several constraints using the fmincon function.

The fmincon function is as follows:

[opt_design_var,fval] = fmincon(@(design_var) obj_function(design_var,timeSet,CSD0, other_inputs),...
  design_var0,*CONSTRAINTS*);

where the design_var is the temperature profile, which is an array of temperatures at constant discrete time intervals, and the obj_function returns the mean particle size (the value that needs to be maximised), given the temperature profile and other variables.

The initial temperature profile estimate design_var0 looks like this:

After each set of iterations, the calculated temperature profile, opt_design_var, is used as the new initial temperature profile guess for the next set of iterations. This is repeated until convergence.

The problem is that the minimum value it converges on, I know for a fact is not the minimum.

Any help is greatly appreciated, thanks in advance.

Answer 1

[Walter Roberson]

fmincon is a local minimizer. You need to use something from the Global Optimization Toolbox.

Possibly patternsearch() might be useful for you. I do not happen to know much about it, though.