Asynchronous Fault Diagnosis of Stochastic Discrete-Event Systems in Industrial Applications

In practice, it is often infeasible to initialize the diagnoser and the system under diagnosis synchronously. Moreover, many real-world systems tend to be uncertain and imprecise. In this study, we investigate asynchronous fault diagnosis of stochastic discrete-event systems (DESs). First, we introduce the notion of asynchronous A-diagnosability, which characterizes the ability to diagnose a fault with an arbitrarily large probability by observing a finite number of events when the diagnoser and the system are asynchronously initialized. Then, we propose a diagnosis structure called asynchronous stochastic diagnoser and derive a necessary and sufficient condition for asynchronous A-diagnosability. Furthermore, we discuss and prove the relationship between asynchronous A-diagnosability and synchronous A-diagnosability. Using the presented diagnosis framework, we have developed a diagnostic system specifically tailored for an aircraft. This case study exemplifies the effectiveness of the proposed method in computing the probability of a system being in a faulty state based on the observed event sequence under unknown initial conditions. Additionally, in cases where the diagnoser misses an observation, it can be restarted to record the system’s behaviors without requiring the restart of the system.