Evidential KNN-based Performance Monitoring Method for PID Control System

It is interesting and challenging to monitor the performance of closed-loop systems to refine the tuning of a controller implemented in practice. This paper intends to propose a new pattern recognition rule based method in the framework of D-S evidence theory to monitor the performance of proportional- integral-derivative (PID) controller. Based on some priori training patterns explored from three typical responses of the controlled systems, the evidential K-Nearest Neighbor (EKNN) classifier is utilized to recognize the underlying patterns of the closed-loop systems controlled by PID controller online. In contrast to some existing methods, our proposed method provides an insight of monitoring control performance from viewpoint of imprecision and uncertainty. More interestingly, the proposed method uses a simple but in nature one synchronized index instead of multiple performance indexes that have been commonly used. Some simulations validate our conclusions.