Data Prediction & Comparative Error Analysis of Compressible Fluid Flow using Neuro-Fuzzy Modelling

Nitrogen gas under pressure released from storage is expected to flow at a set pressure at the downstream end of the control valve. The dynamic changes in pressure and temperature are measured and the Proportional- Integral-Derivative (PID) error-based prediction model gives feedback signal to the control valve to attain accurate downstream pressure. It was observed that the error percentage in a PID controller is very much higher and hence the system never reaches the target pressure. As it is a highly dynamic system having few seconds as the processing time, less error percentage is desired to attain the set point in a short period of time. Statistical model of the real time data was done through regression and compared with the ANFIS modeling results. It could be found that the machine learning process with ANFIS modeling resulted in very less error when compared with the existing PID controller and the traditional statistical modeling.