Robust Fault Diagnosis by Using Bond Graph Approach

This paper deals with the generation of fault indicators and residual thresholds in the presence of parameter uncertainties by using a bond graph representation in linear fractional transformation (LFT) form. The residuals' sensitivity analysis, which is based on the fault detectability indexes, is used for residuals evaluation. The developed algorithms are applied on an electromechanical test bench system for on-line fault detection and isolation. Simulation and experimental tests show the influence of dead zone magnitude and variations of system parameters on the residuals.