A comparative study of energetic model-based fault detection using HBG and COG formalisms

Abstract

Summary The problem of modeling and fault detection in an electromechanical system having a graphical representation is considered. For this system's dynamics, motivated by a desire to provide a precise fault detection procedure using a unified energy‐based framework, 2 energy‐based graphical formalisms are presented and compared: the Hamiltonian Bond Graph and the Causal Ordering Graph. Firstly, easily calculated from the Hamiltonian Bond Graph representation covering the causal energetic paths, the energetic residual generators are systematically deduced by comparing the energy quantity in both normal and faulty situations. Secondly, the causal and structural properties of the Causal Ordering Graph tool can be used to design an observer devoted to fault detection so as to derive directly energy‐based residual generators in terms of faults. To highlight the effectiveness and applicability of 2 proposed approaches, simulation results on DC motor are provided and discussed. The performances of these fault detection schemes are compared and discussed in detail with respect to existing methods.