Different results from different MATLAB versions (R2017a and R2018a)

Hello,

I am using 'gamultiobj' for multiobjective optimization problem. I am awared of that two different MATLAB versions (R2107a and R2018a) give different results for the exactly same code. I checked the options of the function such as cross-over fraction, mutation function etc., and both of them have same options. What is the reason of this situation? Thank you for your attention.

[Walter Roberson]

Can you attach your code and any relevant data for testing?

[Hasan Eren Bekiloglu]

nvars=6;
lb=[0.45 318.8 1.5 1.2 3 0];
ub=[0.75 1920 9 1.5 20 15];

rng default % for reproducibility

options = optimoptions(@gamultiobj,'Display','iter');

[x,fval,exitflag,output,population,scores]=gamultiobj(@ORC,nvars,[],[],[],[],lb,ub,options);

You can see the code and an answer as an example above. I preferred to use default options and I checked robustness of my code; I've taken exactly same results a lot of times when I used same version of MATLAB. For example, the red solution in the figure is obtained with R2017a and R2017b, but the blue result is gathered with R2018a.

[Walter Roberson]

I find some papers that talk about development of Organic Rankine Cycle (ORC) simulation in MATLAB or Simulink, but I can't seem to find code at this time.

[Hasan Eren Bekiloglu]

However, my problem is not related to 'ORC' function. For example,

fitnessfcn=@(x)[x(1)^2/4+x(2)^2/4,(x(1)*(1-x(2))+10)];

nvars = 2;lb = [0 0]; ub = [1 8];

rng default % for reproducibility

options = optimoptions(@gamultiobj,'Display','iter','TolFun',10^-4,'ParetoFraction',0.35);

[x,fval,exitflag,output]= gamultiobj(fitnessfcn,nvars,[],[],[],[],lb,ub,options);

figure plot(fval(:,1),fval(:,2),'r*') xlabel('Obj_1') ylabel('Obj_2') title('Pareto Front') legend('Pareto front')

The code I published above also gives different results, and you can see the difference in the figure.

Very basic function even gives different results with rng default. I wonder the reason.