how to optimize neural network input using GA? (maximizing the input)

I am new to this field and any help can be useful. I'm currently doing process optimization for furnace operation - in MATLAB. I am trying to relate the process parameters with the efficiency of the heater such as e.g.:

Inputs (Process parameters)

heat gain heater draft pressure excess air Outputs

heater efficiency

I am first training a Neural Network model in order to predict the efficiency from the process parameters. I got good results and for now I am not interested in the performance.

Next I want to optimize (maximize) the input parameter heat gain using Genetic Algorithm. using heat balance,my fitness function (object function) is

heat gain = (Qf/m)*[1-(.5/(21-x1)+.8)*(x2-Ta)] Qf is fired duty excess air = x1 m is mass flow T is fluid temperature = x2 Ta ambient temperature

only excess air is an input in the neural network, other parameters were not used in the network I don't have constrains but I have LB & UB

function y = fitness(x)

    y = (Qf/m)*[1-(.5/(21-x1)+.8)*(x2-Ta)];

end

next i set the number of design variables and their upper and lower bounds:

nvars = 2; % Number of variables LB = [3 280]; % Lower bound UB = [4 310]; % Upper bound

[x,fval] = ga(@fitness_func,nvars,[],[],[],[],LB,UB,@(x));

I still don't know how to fit the network input into the GA (to maximize input)

I hope I've explained the problem in clear way. As I said at the beginning,I am new on this subject and any help is appreciated