Research On Fault Diagnosis Method Of Helicopter Swashplate Rolling Bearing Based On Time-frequency Representations And CNN

Helicopter swashplate is the key component to control the flight condition of the helicopter.Once the cracks and wear of the rolling bearings occur,there is a great safety risk to the flight of the helicopter.Research can accurately identify ball bearing diagnostic methods in time and ensure flight safety is of great significance.This article takes Luoyang LYC Bearing Co.,Ltd.as the object of study for the vibration signal collected by the simulated failure experiment of a helicopter swashplate rolling bearing.The shorttime Fourier transform is used to construct the corresponding time-frequency representations.The recognition of the time-frequency representations through the convolutional neural network(CNN)achieves the purpose of fault diagnosis.The main work content and research results are as follows:(1)Introduction to related basic theory.Firstly,the short-time Fourier transform method and normalization method are used to construct time-frequency representations of arbitrary size,and the gray value of the time-frequency representations is obtained.The constructed time-frequency representations can simultaneously represent the timefrequency domain information of the vibration signal.Second,The system elaborated the structure,composition and function of CNN.(2)A CNN network model for fault diagnosis of helicopter swashplate rolling bearings was designed.First,determine the most suitable CNN network structure through experiments with four convolution layers,four pooled layers,a convolution kernel size of 3×3,and a number of convolution kernels of 3/14/812,respectively.Pooling,using 50% Dropout,Softplus function as its activation function.And for the problem of fault diagnosis of helicopter automatic tilting roller bearing,the experimental verification of the CNN network designed by this method has high recognition effect,strong generalization ability and low overfitting effect.(3)A method for fault diagnosis of helicopter swashplate rolling bearing based on short-time Fourier transform and CNN is proposed.First,the diagnostic data is segmented,and the corresponding time-frequency representations is constructed for each segment of data,and the data set is divided into training data,verification data,and test data.Then,training data is used to train the CNN.When the trained model reaches the required recognition rate of the verification data or reaches the maximum number of iterations,the training is stopped and the model is saved.Finally,use the test data to test the model and get the experimental results.Through a large number of experiments,it is proved that the method has good generalization ability and can still achieve high accuracy in the case of small samples.At the same time,it can still obtain better recognition results for different data sets.