Process Optimization using a trained ANN (Neural Networks) Model by Genetic Algorithm (ga)

Hi

I am doing optimization of a process unit. I have already trained an ANN model. It has 16 inputs and 11 outputs. I have generated the neural network function matrix for simulation of the ANN model using built-in MATLAB feature (in ANN toolbox). The objective is to minimize the cost of the process with respect to 4 product streams (present among 11 outputs).

function z = ObjectiveFun(x)
p = [103 91 85 80]'; % price of product streams
c = 50; % cost of feed stream
x1=ProcessANNFunctionMatrix(x); % ANN function for simulation of the ANN network
x=x1(7:10);
z = -sum((x.*p(1:4))) + (x1(1).*c);

end

Now, what I am doing is to use the above objective function as input in the ga (available in MATLAB optimization toolbox), implement the lower and upper bounds of variables x and x1. However, it seems that the answer is not correct.

I want to achieve the following:

• Optimize the product flow rates (obtained as outputs of the trained ANN model) based on the objective function and constraints.

How should I implement the trained ANN model with an objective function and constraints in the ga for optimization of product flow rates? I have also attached the trained ANN function matrix for reference.

Looking forward to your responses.

Regards,

[Walter Roberson]

What leads you to suspect that it is not getting the correct answer?

[Zaid Rana]

When I run it, the ga only stucks at objective function value of -5320. It does not improve. The value remains the same from 1st to 100th iteration.