Design and implementation of an industrial generalized predictive controller on multivariable processes via programmable logic controllers

Abstract

This paper addresses design and implementation of an Industrial Generalized Predictive Controller with the aid of artificial neural networks (ANNs) for multivariable processes via industrial Programmable Logic Controllers (PLCs). Nowadays, Although PLCs are the widely accepted computer-based industrial controllers due to their remarkable advantages, implementation of advanced control techniques is not easily possible on available PLCs. Therefore, searching ways of implementing these control techniques via PLCs even with low computational power can promote the product quality in process industries. The key novelty of this paper is, implementing an Industrial Generalized Predictive Control via PLC, in which the controller parameters are approximated by using artificial neural network (ANN). Furthermore, a fast programming technique based on simplified algorithms using IEC 61131-3 standard is investigated. In addition, laboratory experiments are carried out for a two-input two-output semi-industrial level and temperature process in the PLC Laboratory of Shiraz University. Implementation of traditional PID controllers on the same PLC is also investigated for comparative purpose. Experimental results reveal the superiority of the proposed approach over the traditional PID controllers in both set point tracking and control action. Besides, the results show that such advanced techniques can be implemented on PLCs with normal computational power without bearing the high cost of upgrading the PLCs.