MATLAB: Objective function problem using “fmincon”

fminconOptimization Toolbox

Dear Community,
I work on the optimization of the lift-coefficient of an airfoil. My goal is to know how to do that with a genetic algorithm, but I prefer beginnig with something easier.
Here is my main code : 
[Xinit,Yinit,XU,YU,XL,YL] = airfoil_coord('NACA0012'); %load Xinit=[Xlower,Xupper] Yinit=[Ylower,Yupper] : (x,y)coordinates NACA0012
fun = @(YN)airfoilcoord(YN,CLtarget,Xinit,alpha,Re,M) 
CLtarget = 0.6; %I assume here that {CLinit=0.4610} so {CLtarget=0.6} would be an optimization
alpha =4; %Flight conditons


Re = 32000000; %Flight conditons
M = 0.1764; %Flight conditons
% lb = transpose(ones(1,199)*(-0.08)); %I assume to not enforce a lower and upper bounds
% ub = transpose(ones(1,199)*(0.08)); 
% Code optimization
[Yopti]=fmincon(fun,Yinit,[],[],[],[]);
Here are the functions used in the code: 
function error = airfoilcoord(YN,CLtarget,Xcoord, alpha, Re, M)
% 
CL = run_xfoil(Xcoord, YN, alpha, Re, M); %I want the YN vector to change in every loop
error = (CLtarget - CL)
function [CL] = run_xfoil(xcoord, ycoord, alpha, Re, M)
 
for i=1:length(ycoord)
    ycoordxfoil(i)=ycoord(200-i);
end
for i=1:length(xcoord)
    xcoordxfoil(i)=xcoord(200-i);
end
Profil = [transpose(xcoordxfoil) , transpose(ycoordxfoil)];
%Necessary to use the XFOIL software
% Function xfoil in the directory  
 [pol,foil] = xfoil(Profil, alpha, Re, M, 'oper/iter 5000');
 
 CL = pol.CL; %Return a double which is the CL for the airfoil represented as : (xcoord,yccord)
Here is my problem :
When I evaluate each member alone, it works well. But when I run the code, there is not improvement, the CL and the vector YN stay the same at each loop. I think there is a little mistake in my code but I can't find where..
Can someone help me ? Thank you for reading and the time you'll take to answer !
Teva

Best Answer

Whenever you have an optimization problem, a good first step is to verify that IF you pass in two slightly different set of parameters to your objective, you get a different result. We cannot test that, since we don't have the complete code. So, does the software produce a result that does not change for slightly different inputs?
Related Question