Internal model control of cumene process using analytical rules and evolutionary computation

Abstract

Cumene is used as precursor for production of many organic chemicals and as thinner in paints & lacquers. Its production process involves one of the large-scale manufacturing processes with complex kinetics. Different classical control strategies have been implemented and compared for the cumene reactor in this process. As a system with large degrees of freedom a novel approach for extracting the state space model from the COMSOL Multiphysics implementation of the system is adopted here. Internal Modern Control (IMC) based PI and PID controllers are derived for the system. To derive the controller setting the system is reduced to the FOPDT and SOPDT model structure using Skogestad half rules. The integral time is modified to obtain the excellent set point tracking and faster disturbance rejection. From the analysis it can be stated that PI controller suits more for this specific process. Particle Swarm Optimization (PSO) algorithm, an evolutionary computation technique is also used to tune the PI settings. The PI controllers with IMC, Zeigler Nichols and PSO tuning are compared and it can be concluded that PSO PI controller settles at 45s without any oscillations and settles down faster for the disturbance of magnitude 0.5 applied at t=800s, through it is computationally intensive compared to other controller strategies.