Hybrid Bond Graph for robust diagnosis to measurement uncertainties

Abstract

This paper deals with robust Fault Detection and Isolation (FDI) to measurement uncertainties using Hybrid Bond Graph (HBG) approach to improve the robustness of detection in presence of measurement uncertainties. The scientific interest of this work is use of one tool (Bond Graph) not only for modelling of hybrid system and measurement uncertainties but also for generation of robust fault indicators and thresholds. For this task all the measurement uncertainties are modelled on the Hybrid Bond Graph (HBG) model in derivative preferred causality. Based on the structural and causal proprieties of the bond graph tool, the generated Analytical Redundancy Relations (ARRs) are robust with respect to measurement uncertainties presented by thresholds. Furthermore, those fault indicators called Generalized Analytical Redundancy Relations (GARRs) are valid at all modes and derived systematically from an HBG model with a specific causality assignment. This causality assignment is guided by the preferred causality of controlled junctions. An HBG with such causality assignment is named Diagnostic Hybrid Bond Graph (DHBG). A systematic causality assignment procedure, named Sequential Causality Assignment Procedure for Hybrid Systems Diagnosis (SCAPHD) is developed to facilitate the derivation of GARRs. The SCAPHD extends the classical SCAP by introducing the concept of preferred causality of controlled junctions. An application to a hydraulic is used to illustrate this method.