Control performance monitoring of chemical industrial process based on model residual

With the increase in the scale of modern industry, control systems for chemical processes have become increasingly complex and intelligent, with more stringent requirements for control performance. This paper proposes a control performance monitoring method based on model residual for chemical process. First, the control performance index is constructed by canonical variable analysis. Then, the process model/disturbance monitoring index is proposed with the estimated disturbance innovation and the benchmark model residual. Next, in order to monitor the process-model mismatch, the process model monitoring index is defined using the ratio of the estimated actual disturbance innovation to the actual model residual. The disturbance monitoring index is proposed using the JS divergence of the actual model residual and the benchmark model residual, so as to detect the disturbance dynamics. Finally, the effectiveness of the proposed method is demonstrated by Wood-Berry Distillation Column.