Bond Graph model based for robust diagnosis of Electromechanical actuators

This paper belongs to a series of articles as part of the Electro-Mechanical Actuator Health Management applied to avionic systems and carried out in the framework of an industrial project. The previous first and published part of the project concerned the structural diagnosability analysis with bond graph (BG): which faults can be detected and isolated with any need of numerical calculation. However those structural results must be verified using analytical redundancy based on BG model. The performance depends mainly on the instrumentation architecture and model accuracy for online monitoring. The fault indicators mainly named Analytical Redundancy Relations (ARR) are generated using the variable elimination theory thus model inversion. The interest of the present paper is to use the BG model not only for diagnosability analysis before industrial application (as proposed in the previous research) but also for online implementation of robust fault indicators with respect to parameter uncertainties generated from the dynamic graphical model. Finally the structural qualitative results will be confronted with quantitative results. The developed methodology is illustrated using an industrial controlled Electro Mechanical Actuator test Bench.