Fault detection in model predictive controller

Real-time monitoring and maintaining model predictive controller (MPC) is becoming an important issue with its wide implementation in the industries. In this paper, a measure is proposed to detect faults in MPCs by comparing the performance of the actual 'controller with the pedormance of the ideal controller. The ideal controller is derived from the dynamic matrix control @MC) in an ideal work situation and treated as a measure benchmark. A detection index based on the comparison is proposed to detect the state change of the target controller. This measure is illustrated through the implementation for a water tank process.