Augmented analytical redundancy relations to improve the fault isolation

Abstract

Fault detection and isolation (FDI) is an essential task that allows avoiding the fault consequence on the performance of the system. The bond graph, as a modelling tool, provides through its structural and causal properties, an automatic generation of analytical redundancy relations (ARRs). These relations are used for diagnosis applications, they constitute the mathematical constraints that are used to verify the coherence between the process measurements and the reference of the system behaviour represented by the model. The classical ARR diagnosis approach allows to both detect and isolate the defective component in the system. In this work, the main goal is to increase the number of isolable faults by increasing the number of ARRs, using the output of the bond graph model along with the measured output of the real system. The innovative interest in this work is that the number of the isolated faults can be improved without the addition of more sensors. Following the general discussion of the proposed method, a robotic subsystem (traction of an omnidirectional mobile robot) is considered to validate the proposed procedure. Two faulty scenarios are then presented and discussed using both the classical and the proposed approach.The proposed method is able to isolate 3 faults that can not be isolated using classical ARR.