Hierarchical Active Fault Diagnosis for Stochastic Large Scale Systems with Coupled Faults

The paper deals with the active fault diagnosis of large scale stochastic systems with faults modeled as mutually dependent Markov chains. The system is described by multiple models representing fault-free and faulty behavior of the system. The aim of the active fault detector in addition to detecting the faults is to excite the system to improve the detection quality. The algorithm consists of two stages: the off-line design of the Bellman function providing the optimal excitation and the on-line estimation, which generates the decisions and selects the optimal excitation according to the Bellman function. In particular, the paper focuses on the online estimation and proposes an algorithm in the hierarchical architecture. The local nodes estimate the continuous state of the subsystems, select the optimal excitations and send local likelihoods to the central node. The central node generates the decisions and submits the respective model probabilities to the local nodes. The performance of the proposed algorithm is validated using a simple numerical example.