Abstract: Aiming at the problem that difficulty of identifying the fault of aviation hydraulic pipeline, an intelligent fault diagnosis method of hydraulic pipeline based on nonlinear adaptive Kalman filter (NAKF) and depth belief network (DBN) was proposed. Firstly, on the basis of the traditional Kalman filter(KF), the least square method is used to modify the Sigma points constructed, the influence of Gaussian distribution on Sigma points was eliminated, and the nonlinear adaptive Kalman filter was proposed.Then, the random noise of vibration signals measured in aviation hydraulic pipeline was removed, the parameters of the deep belief network model were designed, and the hydraulic pipeline data set was input into the deep belief network model for training.Finally, based on the same sample data, support vector machine(SVM) and back propagation neural network(BPNN) were used for training and processing respectively. Classification accuracy were used to comprehensively evaluate the classification performance of the three fault diagnosis models. The results show that the accuracy of NAKF-DBN hydraulic pipeline intelligent fault model can reach 99.72%, the average accuracy of traditional support vector machine model and backpropagation neural network model is less than 95%, and the accuracy of DBN network without NAKF filtering is even lower, only 86.58%. The effectiveness of NAKF-DBN model for hydraulic pipeline fault identification is verified, which provides a new idea for intelligent diagnosis of aviation hydraulic pipeline.

Key words:  hydraulic transmission circuit; support vector machine(SVM); back propagation neural net work(BPNN); deep belief network (DBN); nonlinear adaptive Kalman filter (NAKF);intelligent fault model