Model-based process diagnosis: Bond Graph and Signed Directed Graph tools

This paper deals with a fault detection and isolation for linear systems using causal graphical approaches. The monitoring based on qualitative models (Signed Directed Graph) enhance fault localization by analyzing initial and final responses of system variables due to deviations, these dynamic behaviors are deduced using causal paths and propagations from the fault node to system variable node. However, the quantitative models (Bond Graph), as a unified multidisciplinary tool, is adopted to detect and isolate faults by generating analytical redundancy relations (ARRs) through covering causal paths from unknown variables to known variables. The interest and efficiency of the proposed methods are illustrated via a numerical example.