Fault Diagnosis of Uncertain Systems using Bond Graph Approach
Diagnostic des Systèmes Incertains par l'Approche Bond Graph
Résumé
Abstract
This work deals with robust fault diagnosis in presence of parameter uncertainties using bond graph model in LFT form (Linear Fractional Transformations). An approach for ARRs and adaptive thresholds generation is developed and implemented on appropriate software. The diagnosis performances are improved using a residual sensitivity analysis, which allows defining sensitivity indexes of parameter uncertainties and fault detectability indexes. The fault detectable value can be estimated due to the energetic aspect of the bond graph tools.
The developed method is applied in real time on two industrial applications with different complexities:
􀀹 a mechatronic system, in order to detect and to isolate backlash phenomenon by distinguishing fault from parameter uncertainties
􀀹 an energetic process of steam generation, which is a non stationary system with a complex parameter space.
This work deals with robust fault diagnosis in presence of parameter uncertainties using bond graph model in LFT form (Linear Fractional Transformations). An approach for ARRs and adaptive thresholds generation is developed and implemented on appropriate software. The diagnosis performances are improved using a residual sensitivity analysis, which allows defining sensitivity indexes of parameter uncertainties and fault detectability indexes. The fault detectable value can be estimated due to the energetic aspect of the bond graph tools.
The developed method is applied in real time on two industrial applications with different complexities:
􀀹 a mechatronic system, in order to detect and to isolate backlash phenomenon by distinguishing fault from parameter uncertainties
􀀹 an energetic process of steam generation, which is a non stationary system with a complex parameter space.