Fault Diagnosis in Hybrid Systems using Possible Conflicts

Abstract

Hybrid systems, whose behavior exhibit continuous and discrete event dynamics, are present in many industrial environments. The complexity of such behavior makes the online fault diagnosis task very challenging. In this work we propose an accurate and timely online fault diagnosis approach for hybrid systems. Our approach uses the Hybrid Bond Graph modeling methodology to avoid the precompilation of all possible modes in the system. Structural decomposition of Hybrid Bond Graph models, obtained by means of Possible Conflicts, together with its implementation using Block Diagrams, allows the diagnosis system be quickly reconfigured when a mode change occurs. Moreover, this diagnosis approach has the ability to track the behavior of the system during the transitory states caused by mode changes. Diagnosis results using a four-tank system demonstrate the capabilities of the proposal.