A Comparison of Bond Graph Model-Based Methods for Fault Diagnosis in the Presence of Uncertainties: Application to Mechatronic System

This paper deals with comparing three methods for robust fault diagnosis that generate their residuals using bond graph model. These methods are the causality inversion method, a sensor data combinations method, and a faults/residuals sensitivity relations method. In addition, both parameter and measurement uncertainties are considered to generate the adaptive residual thresholds. Through simulation on a mechatronic system, the presented methods are studied under sensor and parameter faults. The results of the case study are compared for gaining practical insights about the applicability and performance of these methods. The results show that the faults/residuals sensitivity relations method has a better diagnosis performance as compared to the other methods.