System Identification of Batch Polymerization Reactor by Means of Estimator Algorithms and Performance Analysis of Autoregressive Exogenous Models

Abstract

Batch reactor is an irreplaceable process of many industries. An essential task in any industrial process is model identification. Any mathematical model, including transfer functions, state spaces, and differential equations, can be used to depict a process. In some cases, due to our limited knowledge of the model parameters, the process couldn’t be represented mathematically. In these situations, instead of white and gray box technique the black box model technique has been used, in which the process input and output data are subjected to various estimation methods for the model parameter identification.  Here, least mean square and recursive least square algorithms are chosen as estimators. This work concentrates on ARX, NARX, and NARMAX model identification of reactor temperature T_r of batch process. The reactor temperature data is collected by solving process differential equation. Those data are subjected to estimation algorithms to find the ARX, NARX, and NARMAX model parameters. Identified model outputs are validated with process reactor temperature T_r data. The best model and estimation algorithm are highlighted from the computational study for future work.