A new robust fault detection method based on combination of bond graph and unknown input observer

In this paper, a new active robust fault detection method based on combination of bond graph and Unknown Input Observer (UIO) is proposed. The proposed two-stage method will satisfy some desirable criteria of a fault detection system. In the first stage, an UIO based on derived state space representation form from the bond graph model is used to estimate the states and outputs of the system, which are insensitive to disturbances in the system. Then, the obtained Analytical Redundancy Relations (ARRs) are considered based on the output estimation error of the observer, which are sensitive to faults while are robust against the disturbances. This form of residuals is called Error-based Analytical Redundancy Relations (EARRs), which further becomes robust against parametric uncertainties by defining adaptive thresholds on the parameters’ values of bond graph model. Simulation results for an active suspension system are provided to show the great performance of the proposed method.