MONITORING AND DIAGNOSING PROCESS-CONTROL PERFORMANCE - THE SINGLE-LOOP CASE

This paper presents a hierarchical method for monitoring and diagnosing the performance of single-loop control systems based primarily on typical operating plant data. It (1) identifies significant deviations from control objectives, (2) determines the best achievable performance with the current control structure, and (3) identifies steps to improve the current performance. Within the last point, the method can isolate whether poor performance is due to the feedforward loop or the feedback loop. If in the feedback loop, it is sometimes possible to determine whether the cause of poor performance is plant/model mismatch or poor tuning. The methods are based on simple but rigorous statistical analysis of plant time series data using autocorrelation and cross correlation functions. The theoretical basis of the method is developed, and it is applied to simulation studies which clarify the principles. Then, results of studies on two industrial processes are reported. The first is a heat exchanger feedback temperature controller, and the second is a feedforward-feedback tray temperature controller in a 50-tray distillation column. The initial diagnosis and subsequent control performance improvements are reported for both cases.