I want optimum combination of three inputs which will provide me optimum values of eight outputs

I have prctically done a test on a Variable Compression Ratio Diesel Engine with three inputs (load, blend, compression ratio) and my eight outputs are BP,BSFC,Efficiency,CO emission,HC emission,CO2 emission,O2 emission,NOx emission. I have 125 combinations of the three inputs as well as eight outputs. Out of the eight outputs higher values of BP,Efficiency and O2 emission are desirable. At the same time the other five outputs need to be small. So I want the optimum values of the three inputs for what I can get the best result.

[Oleg Komarov]

And the question is? (How do I do it, is not a feasible one...too generic)

Have you tried out smt? Show the code (model) and a brief synthetic example of your inputs and desired output.

[j dr]

As I understand you have 1 3x125 matrix of "input results" and a related 8x125 matrix of "output result"

What is it you want to optimize and how ?

Most likely you're gonna have to ponderate your criteria...

Example, you could produce a value V=(BP)*(Efficiency)*(O2)/BSFC)/(CO)/(HC)/(CO2)/(NOx)

this value would be large for what you specified, but a result of 30% efficiency and 10 Co emissions would give a better result than an efficiency of 50% and 20 CO emissions (because (50/20) 2.5 < 3 (30/10)

...

Answer 1

[Greg Heath]

This is not a job for a NN. The net tries to yield the specified training output when the corresponding training input is applied.

All you need to do is search the training data for the desired type of outputs. Then find the corresponding inputs.

Greg