Actuator leakage fault detection and isolation based on extended Kalman filtering scheme.

摘要

In this research, the dynamics of a typical class of hydraulic actuators is studied and actuator leakage Fault Detection and Isolation (FDI) schemes are proposed based on the Extended Kalman Filtering (EKF) algorithm.; A dynamic model of the hydraulic actuation test rig combined with an actuator friction model is first derived. The parameters of the friction model are obtained by careful experiments. Simulations conducted to validate the model shows that the simulation errors are bounded within 10% of actual measurements and, within the normal operating bandwidth, the dynamic model adequately represents the actuator used in the test rig.; Based on the validated dynamic model, an offline actuator leakage fault detection scheme employing EKF algorithm is developed. Simulations and experiments are carried out to verify the effectiveness of the proposed FDI scheme without the presence of external disturbance. With a residual generation scheme that compared the actual measurements and the EKF estimates, three types of leakage - the cylinder chamber leakage at either side of the actuator and the leakage between the two chambers - are tested. The leakage faults are detected and identified by tracking the variation of the residual errors and an offline actuator FDI scheme is developed. The leakage identification patterns are recognized.; To further study the performance of the EKF based FDI scheme in normal operating conditions with external disturbances, the system model is improved to estimate any arbitrary load. Actuator friction is regarded as part of the external disturbance in this phase and, together with the environmental load, becomes a time-varying parameter of the system. Simulations and experiments show that the external load, including the actuator friction, can be accurately estimated. Different types of testing signals are applied to investigate the effectiveness of the FDI scheme. Experiments show that the leakage fault identification patterns are consistent with the conclusion obtained in the unloaded mode. A potential online FDI scheme and an offline actuator friction estimator are further proposed.