Analisadores Virtuais e Controle Avançado para uma Planta Industrial de Polietileno Linear e Avaliação de seus Benefícios Econômicos

Abstract

This work presents the necessary structuring for the implementation of an advanced quality controller of the type Model Predictive Control (MPC) for an industrial plant producing linear polyethylene using “Sclairtech” technology. Through this control, the dispersions of the Melt Index (MI) and density quality variables will be minimized and the consumption of catalysts, co-catalysts, deactivators and adsorbent bed will be optimized. A conceptual project was developed where all the premises for the control strategy are explained. An economic analysis was then developed to validate the economic viability of implementing the control, in accordance with the best global practices. Furthermore, a new technique for defining the optimal architecture for an Artificial Neural Network (ANN) model was developed and presented. This technique was applied to develop two virtual analyzers based on ANN to predict the MI quality and density variables. Finally, using historical data, industrial tests and empirical operational knowledge about the process, models were developed to simulate the developed control system and controllers. In this context, a new model development methodology for MPC controllers is presented, which also explicitly uses the experience of process experts. The results show a very adequate performance from a technical point of view, and a very substantial and viable economic return, demonstrating the viability of the study and project developed.