Actuator fault diagnosis of a class of nonlinear systems in closed-loops: a case study

This paper presents an actuator fault diagnosis approach for a class of nonlinear systems in closed-loops based on a continuous stirred tank reactor (CSTR). The purpose of the control is to track the reactant concentration set point. A new concept of "equivalent bias" is proposed to model the actuator faults. Both the states and the equivalent bias are online estimated by a strong tracking filter (Zhou et al., 1992). The estimated equivalent bias is then evaluated via a modified Bayes classification based algorithm to detect and diagnose the actuator faults, and the estimated states are used to form the PID control law. Four kinds of actuator faults are tested by Monte Carlo simulations, which demonstrate the effectiveness of the proposed approach.