Controle multivariável aplicado a uma coluna de alta pureza com recompressão de vapor.

Abstract

Industrial processes are multivariable systems (MIMO) consisting of multiple input variables and multiple output variables, where the interaction between these variables is an inherent characteristic of these processes. The model identification in this type of process is an important step in the implementation of the control system, especially when it comes to MPC controllers, which incorporate an explicit process model. The identification procedure performed in this research proposes a different approach to what has been seen in the literature. While the most common is to stimulate the manipulated variables, the proposed new approach aims to stimulate the process from disturbances in the setpoints of the controlled variables, closed loop, in order to obtain a better assessment of the effects of controlled and manipulated variables. The purpose of this study is to compare two control strategies: A classic control strategy and a multivariable control strategy based on model (MPC). The case study is a column separation of propylene / propane with high purity vapor recompression. When compared with the decentralized control the MPC control, it is observed that the MPC performs better, this statement, based on the values of IAE performance index (Integral Absolute Error) for the two proposals analyzed.