Design of decentralised Smith predictor for multivariable non-square processes with multiple time delays

In this paper, multivariable Smith predictor is designed in a decentralised sense for multivariable non-square systems with multiple time delays where the number of process inputs is higher than the number of process outputs. As the relative gain array method is not applicable for MIMO non-square processes, pairing for input and output variables is carried out using the block relative gain technique. The controllers are designed individually based on the corresponding process models. For the delay-compensated system, decentralised multivariable proportional-integral controllers are designed by simplified internal model control method. The proposed method is applied to shell standard control problem (three inputs and two outputs) and mixing tank (three inputs and two outputs) and the performance is compared with that of decentralised multivariable control system without any delay compensator. Simulation studies were carried out for both servo and regulatory problems. Robustness studies were carried out for uncertainty in all the process model parameters. The decentralised delay-compensated system with decentralised controllers showed better performance than that of uncompensated system.