Fault Detection, Identification and Estimation in the EHA System using multiple model estimation

In this paper, a model-based Fault Detection, Identification and Estimation (FDIE) scheme has been developed for the condition monitoring of the electro-hydraulic actuator (EHA) system. The scheme combines the use of a model of the EHA system with the multiple-model (MM) estimation algorithm to evaluate if fault is present, its cause and its severity. Using a multivariate nonlinear stochastic model, the EHA system dynamic responses are simulated in the presence of different types of faults. The FDI and estimation in the EHA system is to determine the system state over time and extrapolate the trend of state evolution in the future time given a stream of observations. The state to be estimated is assumed to be hybrid, i.e., it consists of both discrete and continuous components. System operational modes (nominal or faulty) are described as the discrete components of the hybrid state, while the continuous components track the dynamic behavior of the system to be monitored. The MM estimation algorithm makes use of the extended Kalman filter (EKF) technique to generate estimates of states and key physical parameters, which are related to faults in the EHA system. The proposed fault detection and identification (FDI) is formulated as a hybrid interacting multiple-model estimation scheme. The interaction scheme between multiple models is introduced into the MM estimation algorithm to yield more robust detection and estimation. Estimates of the key physical parameters in the EHA system are assessed against baseline values and fused with the FDI results for higher level monitoring purposes. Two parameters of interests, namely torque motor equivalent resistance and the effective bulk modulus are investigated for the EHA system condition monitoring purpose. The simulation results highlight the considerable potential of the proposed technique for achieving improved condition monitoring of the EHA system.