Smith predictor-based PI control of a wet granulation process

Abstract

The need for prediction and reference updating in feedback control of a wet granulation process is addressed. The granulation process is often modeled as a multi-input multi-output (MIMO) linear model with dead-time. Industrial implementation of granulation process poses strict constraints on the process inputs & outputs. The presence of dead-time and the physical necessity of the input-output constraints are the key challenges of the wet granulation control. These challenges motivated the use of model predictive control (MPC) for such processes. In this work, a Smith predictor-based proportional-integral (PI) controller is designed for the dead-time compensation. Accompanied with the reference updating method to handle the physical constraints. The regulation and reference tracking control problems are assessed via closed-loop simulations of the wet granulation model. The ability of the proposed control approach of dead-time compensation and coping with input/output constraints is rigorously proved. The current approach is compared to MPC of a similar granulation process and found superior in terms of output stability, performance and reference tracking.