Joint Fault Detection, Identification, and State Estimation Based on Conditional Joint Decision and Estimation

This paper presents an approach to fault detection, identification, and state estimation (FDISE) for a dynamic system with abrupt total or partial failures. FDISE includes both decision and estimation and they are highly coupled. Decision includes fault detection and identification (FDI), while estimation is for failure magnitude and system state. Correct FDI benefits estimation and accurate estimation can facilitate FDI. Moreover, detection is binary with only two possibilities, but identification involves multiple hypotheses each for a specific fault. Thus, FDISE is a challenging joint decision and estimation (JDE) problem. Our recently proposed JDE approach provides a general and flexible framework to solve such a problem. In this paper, a joint FDISE (JFDISE) is proposed based on the conditional JDE method. In JFDISE, an interacting multiple-model estimator is used to obtain the hypothesis probabilities and the state estimates. Furthermore, the difficulty caused by the incompatibility of the decision set and the hypothesis set is handled. Simulation results for JFDISE of a flight control system with sequential actuator failures show that the JFDISE outperforms the decision then estimation and the separate estimation and decision methods in terms of a joint performance measure.