fmincon nonlinear constraints not being optimized

Im working on modeling battery gassing. I have IR absorption data from research papers. I know the ratios of these absorption intensities are proportional to the ratios of the number of moles of component gases and thus the average parasitic currents for each component. When i try to optimize my chemical reaction equations to the absorption data my nonlinear constraints do not change. My non linear constraints are of the form:

ceq3=totalcurrent/totalcurrent313K-0.265873214;
ceq4=totalcurrent/totalcurrent333K-0.06471225;

but the final values are just the constants ie ceq3=-0.26 and ceq4=-0.065

How do i get the solver to do this better? I have tried messing with the contraint tolerances and function tolerance.

Answer 1

[Alan Weiss]

Your code contains a lot of lines of the form

averagecurrentH2(isnan(averagecurrentH2))=0;

When I run your entire code in debugging mode I see that many of these lines are active; in other words, many of the calculations are returning NaN and then returning 0. I suspect that you don't really want all these NaN values, and that something is going wrong with your calculation, and that is why fmincon gets stuck.

FYI, I set a 0 objective function

fun = @(x)0;

as recommended in Converged to an Infeasible Point.

I suggest that you debug your nonlinear constraint function. And also, when that is debugged, set your objective to the square of the difference, not the absolute value of the difference:

error = norm((averagecurrentCO2+averagecurrentCO+averagecurrentC2H4+averagecurrentC2H6+averagecurrentO2+averagecurrentPOF3+averagecurrentH2) - 1e-3)^2;
% not
error=abs((averagecurrentCO2+averagecurrentCO+averagecurrentC2H4+averagecurrentC2H6+averagecurrentO2+averagecurrentPOF3+averagecurrentH2)-1*10^-3);

Alan Weiss

MATLAB mathematical toolbox documentation

[Anosh Mevawalla]

i did that and is not having any effect my ceq is still converging to

ceq =

-0.0000

0.0000

-0.2659

-0.0647

for some reason it is not considering the constant term subtracted.