A hybrid fault diagnosis method of electromechanical actuator combining model-based and data-driven

Aiming at the current popular fault methods' disadvantages on diagnosing faults existing in electromechanical actuator, this paper presents a novel hybrid fault diagnosis approach combinmg Model-Based and Data-Driven. Firstly, the extended Kalman filter is used to estimate states of monitored system using available input and output variables. The residual vector corresponding to the specific fault is generated by comparing the actual measured value and estimated value. Secondly, the original accelerometer signals collected from monitored system are decomposed by EEMD and a group of the intrinsic mode functions without mode mixing are obtained. The characteristic vector is constructed by computing the energy index of every IMF. Finally, the characteristic vector is combined and acts as the training samples to train BP neural network fault classifier, therefore, the fault diagnosis is achieved. The example analysis proves that the hybrid fault diagnosis method proposed in this paper is more accurate in diagnosing direct drive EMA faults.