An Enhanced Smith-Predictor-Based Control Law for a MIMO CSTR with Multiple Time Delays

This paper introduces an enhanced Smith-predictor-based control system (SPCS) for multi-input multi-output (MIMO) time delay processes. A conventional SPCS for an MIMO process is composed of a Smith predictor matrix providing predicted outputs and an array of classical feedback controllers. These controllers act on predicted error signals, calculated by subtracting the predicted output from the reference signal at the time of operation. Two major shortcomings were idented for the underperformance of conventional SPCSs for MIMO time-delay processes: (i) the employed classical feedback controllers are designed based on trade-off. Moreover, their use may lead to a windup phenomenon, which adversely influences the control performance of conventional SPCSs; (ii) a predicted output generated by a Smith predictor corresponds to a future time and is therefore not concurrent with the reference at the time of operation. Thus, in conventional SPCSs, the control error is generated using two asynchronous signals. This article proposes two enhancements to tackle the aforementioned dual shortcomings in SPCSs for MIMO time-delay systems. The proposed control system is shown to distinctly outperform a conventional SPCS with proportional-integral-derivative (PID) feedback controllers. The case study is a catalytic stirred tank reactor (CSTR) with three inputs (feed and water flow rates and auxiliary temperature), two outputs (output flow concentration and temperature) and three different time delays. The presented CSTR model, used for the design and simulations, is more comprehensive than any CSTR model reported in the literature.