Robust fault detection based on bond graph UIO observer

Abstract

Motivated by recent concerns in building a totally graphical unknown input observer (UIO), a bond graph approach to design an UIO observer for linear systems with unknown inputs is re-evaluated in this paper. By using the Bond graph methodology and the geometric theory as the main tools, a simple and systematic design procedure is depicted. Compared with the classical UIO design, the built observer is a none other than the famous Luenberger observer but with higher freedom degree that ensures the decoupling of the disturbance and the estimation error. In the past, it was believed that the method is valid for control purpose. Here, a new result is obtained, namely that the method is valid also for robust fault detection purpose. On the other hand, the method is found with no previous simulation results. In this paper, a proof with simulations results on an electromechanical system will show the performance of the approach.