i need to make optimization using genetic algorithm to reduce the minimum between desired deformation and simulated deformation

i need to make optimization using genetic algorithm to reduce the minimum between desired deformation and simulated deformation

i used ansys software for making simulation and the parameters (Height , depth , width )

but i have problem that the time of optimization took too long time and the error is too big .. can i know where is the problem ?

Thanks

function error = objectiveFunction(x, desiredDeformation)

% x(1) = Height, x(2) = Depth, x(3) = Width

% Modify the journal file with the new input parameters

Height = x(1);

Depth = x(2);

Width = x(3);

% Read and modify the journal file

fid = fopen("E:\Master\part.wbjn",'r');

f = fread(fid,'*char')';

fclose(fid);

f = strrep(f,'Width' , num2str(Width));

f = strrep(f,'Height' , num2str(Height));

f = strrep(f,'Depth' , num2str(Depth));

fid = fopen('finaljournal.wbjn','w');

fprintf(fid,'%s',f);

fclose(fid);

% Run Ansys using the updated journal file

system('"C:\Program Files\ANSYS Inc\v231\Framework\bin\Win64\runwb2.exe" -X -R "finaljournal.wbjn"');

% Load the ANSYS output data (Total Deformation)

simulatedDisplacement = readmatrix("E:\Master\part.csv", 'Range', 'E8:P8');

% Calculate the error between simulated and desired deformation

error = sqrt(mean((simulatedDisplacement - desiredDeformation).^2)); % RMSE as the error metric

end

% Define desired deformation and force data (example values)

desiredDeformation = [1.25066872840705e-09, 2.50133745681411e-09, 6.25334385783699e-09, 1.25066877156740e-08, 1.87600315915789e-08, 2.50133754313480e-08, 3.12667193063999e-08, 3.75200631831578e-08, 4.37734070582097e-08, 5.00267508626959e-08, 5.62800948083135e-08, 6.25334386127997e-08];

desiredForce = [2, 4, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100];

% Define number of variables and bounds

nVars = 3; % Number of variables: Height, Depth, Width

lb = [10, 5, 10]; % Lower bounds for Height, Depth, Width

ub = [100, 10, 100]; % Upper bounds for Height, Depth, Width

% Set up the genetic algorithm options with a time limit of 10 minutes

options = optimoptions('ga', ...

'Display', 'iter', ...

'MaxTime', 600, ... % Set the maximum time to 600 seconds (10 minutes)

'PlotFcn', @gaplotbestf);

% Run the optimization, passing the correct objective function

[x_opt, fval] = ga(@(x) objectiveFunction(x, desiredDeformation), nVars, [], [], [], [], lb, ub, [], options);

% Display optimized parameters

disp('Optimized Parameters:');

fprintf('Height: %.2f\n', x_opt(1));

fprintf('Depth: %.2f\n', x_opt(2));

fprintf('Width: %.2f\n', x_opt(3));

% Load the optimized data from Ansys

optimizedDisplacement = readmatrix("E:\Master\part.csv", 'Range', 'E8:P8');

optimizedForce = readmatrix("E:\Master\part.csv", 'Range', 'Q8:AB8');

% Plot comparison

figure;

hold on;

plot(optimizedForce, optimizedDisplacement, '-o', 'DisplayName', 'Optimized Displacement');

plot(desiredForce, desiredDeformation, '-x', 'DisplayName', 'Desired Displacement');

xlabel('Force');

ylabel('Displacement');

title('Comparison between Optimized and Desired Force-Displacement Curves');

legend('show');

grid on;

Answer 1

[MULI]

Hi @noura,

To optimize parameters (Height, Depth, Width) using a genetic algorithm, you can follow these steps:

Each iteration involves running ANSYS via a journal file and retrieving the results, which can be computationally expensive. To address this:

• MATLAB supports parallel computation in “ga” using the UseParallel option. This requires the Parallel Computing Toolbox. You can refer to this link for more information on parallel processing: https://www.mathworks.com/help/gads/how-to-use-parallel-processing.html
• Use a coarser mesh or reduce the number of load steps during optimization to speed up simulations.

A high error between simulated and desired deformation can be reduced by following these steps:

• Ensure that the search space (parameter bounds) includes feasible designs that align with physical and simulation constraints.
• Verify that parameters (Height, Depth, Width) are correctly updated in the journal file before each simulation.
• Use an appropriate metric like Root Mean Square Error (RMSE) to measure the difference between the simulated and desired deformation. https://www.mathworks.com/help/deeplearning/ref/deep.metric.rmsemetric.html

By following these steps, you can reduce computational costs, minimize error, and achieve more accurate optimization results.