Fault diagnosis of hydraulic retraction system based on multi-source signals feature fusion and health assessment for the actuator

In order to solve the problems that traditional diagnostic method is heavily dependent on the signal processing techniques and expert experience, and the diagnostic accuracy is difficult to have big improvement anymore with the accumulation of operational data, which cannot meet the needs of fault diagnosis in the big data age, a multi-source signals feature fusion method by deep learning model is proposed in this paper. The stacked denoising autoencoders (SDAE) is used to extract the abstract and deep features from original features, and then locality preserving projections (LPP) is used for dimensionality reduction to complete the feature fusion. Finally, the fused low-dimensional features act as inputs to the support vector machine (SVM) to realize the failure detection and fault location of typical fault modes of the landing gear hydraulic retraction system. The inhibitory effect of the feedback control on the incipient fault is discussed as well. Moreover, a severity assessment method is presented considering the gradual degradation of leakage fault of the actuator. The diagnostic results show that the proposed method has a better feature fusion ability and higher diagnostic accuracy. The health assessment model can evaluate the health state of the actuator. The significance of this paper is to provide a feasible idea for the fault diagnosis of the landing gear hydraulic retraction system and health assessment of the actuator.