Decentralized fault detection in a mechatronic plant using identified models

Abstract

Recently, techniques for Discrete Event Systems (DES) identification with the objective of fault detection have been proposed in the literature. DES identification consists of computing a monolithic model capable of simulating the observed fault-free behavior generated by the system. Thus, a fault is detected when a discrepancy is observed between the evolution of the system and the estimation of the model. However, when working with systems whose information is distributed, it is necessary to implement a decentralized fault diagnosis architecture in which local diagnosers are computed based on their own observations. Therefore, this paper proposes a decentralized fault detection architecture using identified models. As a motivational example, a mechatronic plant system with decentralized control is proposed and used to verify the proposed method. The formalism used to represent the local models of each subsystem is the Deterministic Automaton with Outputs and Conditional Transitions (DAOCT). Finally, intermittent and permanent faults are simulated in the equipment of the system to verify the efficiency of the proposed method in detecting faults.