Critical Assessment of Control Strategies for Industrial Systems with Input-Output Constraints

Safety and economic considerations pose constraints on inputs and outputs of industrial processes. A conventional feedback control system is not equipped to handle these constraints, and hence, industrial practitioners resort to the addition of ad hoc elements such as selector blocks, in conjunction with more sophisticated control strategies such as split range control. However, this approach lacks an organized and prudent methodology and is usually carried out with the aid of experiential knowhow of plant operators and control engineers. Some recent developments in the literature address this issue, however simultaneously alluding to model predictive control (MPC) as a more natural approach for such systems. In this work, we evaluate the efficacy of the aforementioned approach of simple ad hoc additions to accommodate constraints in a process. To this end, with the aid of a continuously stirred tank reactor model, we illustrate the ease of deployment of selector blocks for handling constraints in simple systems and assess their limitations as the system becomes more complex. Next, we consider a heat exchanger employed in a South -East Asian processing facility where the performance of the currently used selector-based architecture is deemed unsatisfactory by the operators. The current control structure was evaluated for conformity with the approaches suggested in the recent literature and was found to be sound; hence, the unsatisfactory performance stems from the inherent limitations of the proportional integral control structure and/or improper tuning. The implementation of MPC on the exchanger system leads to performance amelioration and enhanced robustness.

Automated summary

Safety and economic constraints in industrial processes are not well handled by conventional control systems, leading to the addition of ad hoc elements and more sophisticated control strategies. Recent developments suggest model predictive control (MPC) as a more natural approach. This work evaluates the efficacy of ad hoc additions to handle constraints and applies it to a continuously stirred tank reactor model and a heat exchanger system, demonstrating the limitations of the approach and the improvement achieved with MPC.