Hybrid Bond Graph for robust diagnosis - Application to hydraulic system

Abstract

This paper presents a method for robust Fault Detection and Isolation (FDI) of Hybrid systems. The scientific interest of proposed method is use only one tool (Bond Graph) for modelling and diagnosis taking into account parameter uncertainties and hybrid aspect of the monitored system. For this task is proposed Hybrid Bond Graph (HBG) approach in Linear Fractional Transformation form (LFT). Based on behavioral, causal and structural properties the generated Analytical Redundancy Relations (ARRs) are robust with respect to parameter uncertainties presented by an adaptative threshold. Furthermore those fault indicators called Generalized ARRs are valid at all mode and derived systematically from Diagnostic Hybrid Bond Graph (DHBG). An application to a hydraulic is used to illustrate this method.