Dear Himanshukumar R. Patel

I am a PHD candidate in Tabriz university of Iran .

I am interested in your researches presented in the paper:

“Neural Network Based Control Framework for SISO Uncertain System: Passive Fault Tolerant Approach '' and

“Fault-Tolerant Controller Comparative Study and Analysis for Benchmark Two-Tank Interacting Level Control System”

I would like to ask you some questions and I would appreciate your kind responses.

I implement exactly your method in my code, In data generation layer for example, at first run I set the control valve value to ‘0.5’ for a healthy system and ‘0.505’ for faulty system (1% actuator fault) and then record residual signal which is a time domain signal whose amplitude is zero at the beginning and increases with the passage of time

Then in Neural Network training stage, I have some problems:

Residual signal has a time dimension and the other four inputs, which are statistical characteristics, are scalar data, so How can these 5 inputs be given to the neural network? The output ‘M’ dimension is also a question for me, Should the output also be 0.5 in this example?

So dear friend, what is your suggestion or solution?

How can I get 10000 data for neural network training?

How many times should I repeat the test?

What values should be assigned to the control valve?

I will be so happy if you can guide me