Comparing methods for system identification on the quadruple tank

Abstract

System Identification in control engineering has been the field of interest for the research and development to find the optimal model of dynamic system. Both MIMO and SISO systems can be modeled using different system identification techniques with satisfactory result using input and output data and easy to implement the control strategies. System identification of a quadruple Tank which is a MIMO system with two user inputs and two outputs and adjustable zero. Using first principle, non-linear mathematical model is developed and it is linearized. Input and output data are taken from the real system using LabVIEW and Stability, observability and controllability are analyzed. Both minimum and non-minimum phase with the change in valve parameters are analyzed. Using input and output data from the real process system identification model is developed by DSR, PEM and N4SID methods. All these identification analysis are done using MATLAB software while the input and output data taken using model developed in LabVIEW. Model developed using all three methods are compared against each other both by simulation and experimentally. Model developed are validated individually using new set of data and the quality of methods are again compared against each other using indices such as MAE, RMSE. Finally, DSR method of identification is found out to be the best method among all three methods and it is suggested to proceed for the implementation of control strategies and further analysis.