Research On Fault Diagnosis Method For Rolling Bearing Of Helicopter Swashplate Based On DCAE-CNN

The swashplate is an important part of the helicopter transmission system,the rolling bearing is the core component of the swashplate,if the failure occurs,it will bring safety hazards to the flight of the helicopter.Therefore,studying the fault diagnosis method of rolling bearings has important practical significance for ensuring the safe flight of helicopters.The swashplate rolling bearing has the characteristics of many balls,large size and low speed,the traditional bearing fault diagnosis method is complicated and the recognition rate is not high.Deep learning as an artificial intelligence technology has unique advantages in dealing with big data tasks,it has achieved good results in many areas.In this paper,a deep convolutional autoencoder(DCAE)and convolutional neural network(CNN)is used to study the rolling bearing fault diagnosis method.The main work contents and research results are as follows:(1)Introduced the relevant theoretical basis.Firstly,the vibration signal collected by the swashplate rolling bearing has nonlinear and non-stationary characteristics,the time-frequency analysis method of wavelet transform is introduced.Secondly,the basic structure and classification principle of CNN are introduced.Finally,the structure and denoise principle of DCAE are introduced.(2)For the helicopter swashplate rolling bearing has the characteristics of many balls,large size and low speed,and the working environment is complex and the noise interference is large,a deep learning image denoising method based on DCAE is proposed.Firstly,the wavelet time-frequency diagram in different noise environments is constructed.Then,the DCAE network model for image denoising is designed,the parameters such as the number of network layers,the size of the underlying convolution kernel,the learning rate and the loss function are determined through experimentally.The results show that DCAE has better denoising effect regardless of whether the time-frequency diagram contains large or small noise.(3)A fault diagnosis method for swashplate rolling bearing based on DCAE-CNN is proposed.Firstly,DCAE is trained by using time-frequency diagrams in different noise environments,reduce the interference of noise on the time-frequency diagram.Secondly,the optimal network structure and parameters of CNN are determined through experiments.Finally,the CNN is used to classify bearing faults after denoising time-frequency map.the real fault data of the research team swashplate rolling bearing and Case Western Reserve University rolling bearing are used to conductdiagnostic experiments,the results show that the proposed method has a better diagnostic effect under different noise environments,especially in the high noise environment,has a higher fault recognition rate compared with other deep learning methods.