Real-time monitoring and diagnosis of a train door mechatronic system

For train manufacturers, a technical, economical, and societal key issue is the real-time monitoring and diagnosis of a train door in new generation of urban trains. Therefore, this paper proposes to deal with the modeling and diagnosis of a train door mechatronic system. For this purpose, the system is first detailed and shows a coupling between electrical and mechanical domains. As the modelling of such a mechatronic system needs a multidisciplinary and unified tool, the bond graph representation is chosen. Moreover, this model contains key components needed to reproduce the door movements and frictions. Thanks to this bond graph model, a global model-based FDI (Fault Detection and Isolation) is proposed for residual computations. Finally, the characterization of residuals for certain internal failures is investigated and the ability of the proposed diagnostic model to detect these failures and isolate some of them is demonstrated. The methodology employed is intended to be applied in a near future in real-time on the practical system.