Research On Fault Diagnosis Of Aviation Hydraulic Pipeline Based On DTCWT And DBN

In recent years,with the rapid development of my country’s aviation industry,the safety of aircraft has also been paid more and more attention.Data shows that about50% of the failures on the aircraft are related to the aviation hydraulic system.It can be seen that the safety of the aircraft hydraulic system is particularly important.As the aviation hydraulic pipeline system is an important part of the aircraft hydraulic system,this article takes the aviation hydraulic pipeline system as the research object,and uses two algorithms of dual tree complex wavelet and deep confidence network to diagnose and analyze aviation hydraulic pipelines.In turn,the purpose of accurate diagnosis of early typical failures of aviation hydraulic pipelines is realized.First,introduce the theories of wavelet transform,discrete wavelet transform and Mallat algorithm.Aiming at the characteristics of high degree of nonlinearity of aviation hydraulic pipeline vibration signal,unclear fault characteristics,severe external noise interference,etc.,through the comparison of soft threshold and hard threshold functions.And the simulation signal denoising effect analysis,choose the double-tree complex wavelet soft threshold denoising method.Then it explains the development process of deep learning and four typical deep learning models,and focuses on the model structure of deep belief networks,including restricted Boltzmann machines and Softmax classifiers,and unsupervised layer-by-layer pre-training algorithms.Secondly,vibration tests were performed on the 15 working conditions designed on the comprehensive experimental platform of aviation hydraulic pipelines,using piezoelectric acceleration sensors and data acquisition systems,and the single faults and mixed faults of straight pipes and elbows were tested.Frequency domain analysis to find out the characteristics of the vibration amplitude changes of straight pipes and elbow pipes under different working conditions.The vibration amplitude of the axial crack failure of the elbow pipe is obviously greater than the vibration amplitude of the axial crack failure of the straight pipe,regardless of whether it is a straight pipe or an elbow pipe.The vibration amplitude of pit failure is greater than that of crack failure,and mixed failures of the same tube will have a certain mutual influence.The analysis of the vibration amplitude data shows that for the early minor failures,the vibration amplitude changes are not obvious.It is difficult to accurately identify the early failures of aviation hydraulic pipelines using only the time-frequency domain analysis method.Finally,according to the two algorithm theories of dual-tree complex wavelet and deep belief network,a program and data modeling suitable for aviation hydraulic pipelines are written using Matlab platform.Two algorithms of dual-tree complex wavelet and discrete wavelet are used to denoise the measured experimental signals.At the same time,the signal-to-noise ratio and mean square error are used to quantitatively compare the noise reduction effects of the two algorithms.The results prove that the dual-tree complex wavelet The noise reduction effect of is better than that of discrete wavelet.The experimental data after the double-tree complex wavelet de-noising is divided into training samples and prediction samples,and the optimal experimental plan is selected by orthogonal experimental design,so as to determine the optimal parameters of the deep confidence network,and realize it through the classification model of the deep confidence network In order to classify and accurately identify early typical faults of hydraulic pipelines,the classification accuracy rate is as high as 99.2%.It is proved that the combination of dual-tree complex wavelet and deep belief network is feasible for the early diagnosis of typical aviation hydraulic pipeline faults,and provides an effective solution for the fault diagnosis of aviation hydraulic pipeline.