State-based fault diagnosis of discrete-event systems with partially observable outputs

Abstract

This paper presents a state-based method for solving the problems of diagnosis and diagnosability of discrete-event systems (DES) with partially observable outputs, due to the lack or limitations of sensors. The diagnoser used for diagnosis consists of two parts: a state estimator and a failure decision-maker. The state estimator makes the state estimation of a system based on the observed output sequence and transfers the estimation to the failure decision-maker that determines whether a fault occurs or not. Moreover, the state or condition (failure status) of the system is not required to be known when launching the diagnoser; thus the system and the diagnoser do not have to be initialized simultaneously, i.e., the diagnoser may be initialized at any moment while the system is operational. Under the premise that the outputs of a system are partially observable, the notion of diagnosability is given and an efficient algorithm for verification of diagnosability is designed with the polynomial computational complexity with respect to the number of system states. Finally, the proposed algorithm is applied to a pump-valve-controller system.