Bond graph UIO approach for fault components monitoring

Abstract

In the literature, Unknown Input Observers methods for Fault Detection and Isolation deal with input (actuators) and output (sensors) faults. The used model is given under state equations format. The dynamic bond graph model as multidisciplinary tool and unified language is directly derived from the physical system where the different parts of the system can be explicitly displayed. The actuators (sources of effort MSe or flow MSf), the sensors (effort De or flow Df detectors) and the physical behaviors modeled by inertia (I), storage (C) and dissipative phenomena (R) are graphically represented by the bond graph model. Furthermore in addition to the behavioral properties, the causal and structural properties are already used for control and diagnosis analysis. In this paper the innovative interest consists in the use of UIO approach for detection, isolation and estimation of component fault which may affect the system. The construction of UIO is based on causal and structural properties of bond graph model (causal path, observability, controllability, invariant zero). The experimental results are implemented on a real hydraulic system.