Controlling the Level of Nonlinear Process Using Soft Computing Technique and Model Identification is done by Radial Basis Function

Abstract

The prerequisite  for  efficient and accurate  control  of  industrial  applications  is driving the  field of system identification to  face  continuous  challenges in  providing better  models of physical phenomena. Approximate linear models are used in most of the industrial controllers, which lowers the control performance. Systems encountered in practice are mainly nonlinear systems.  Model identification of a nonlinear system is usually complicated when compared with linear systems. Hence, it is vital to develop a simple and practical technique for nonlinear process modeling and identification. This model is further used to control the nonlinear process. The nonlinear process used in this work is conical tank process. Conical tanks are widely used in the process industries because of its shape that contributes better drainage for solid mixtures, slurries and viscous liquids. The conical tank is a well-known system, which is having high non linearity, due to the variation in the area with respect to height, controlling the level of conical tank is a challenging task for many industries. The main objective is to model the conical tank using neural network with different controllers. Radial Basis Function Neural Network is used for modeling the Conical Tank. Using Radial Basis Function model the Level of the conical tank is controlled by Neural Network Model Predictive Control (NMPC), Direct Synthesis Proportional Integral (DSPI) and Skogestad (SK) method. These controllers are simulated using Matlab /Simulink. The controllers are designed and compared based on the performance analysis enactment both in performance analysis and time domain specification so it has been chosen as a best controller for the real time application.