Study On Joint Time-frequency Diagnosis Of Civil Aviation Engine Bearing Fault

Vibration fault signal processing and improved fault diagnosis and recognition accuracy are two major technical difficulties which civil aviation engine rolling bearing fault diagnosis research is faced with.The decomposition of the vibration signal of the rolling bearing is a prerequisite for fault diagnosis.Therefore,the signal processing model and the fault diagnosis model are constructed to improve these two issues respectively.The specific work is as follows1?A signal adaptive decomposition and reconstruction model which is achieved by combining Beetle Antennae Search(BAS)algorithm and Variational Modal Decomposition(VMD)is proposed to improve the accuracy of vibration signal decomposition and reconstruction of rolling bearings.It's true that the manual set of the second penalty factor and the number of decomposition layers has an influence on the effect of signal processing.The best two-parameter combination of VMD decomposition is optimized by the way that the position vector of the left and right is iterated in the BAS algorithm,and then the Intrinsic Mode Function(IMF)is acquired by adaptive decomposition method.In addition,the envelope characteristic demodulation method is used to obtain the characteristic frequency information of the fault in the IMF component,which is compared with the a priori fault frequency to evaluate the effect of adaptive decomposition,and the proportion of the impact component is computed to verify the accuracy of signal reconstruction.Results show that the BAS-VMD adaptive decomposition and reconstruction model obtains the IMF component with rich fault frequency components based on the best combination of decomposition parameters.In addition to doubling the frequency bandwidth of the rolling element fault,the reconstruction accuracy of the health state,outer ring fault and inner ring fault signals are also increased by 2.16%,0.41%and 2.28%respectively,which results in establishing the foundation for fault diagnosis in its two-dimensional time-frequency image2?A fault diagnosis deep leaning model based on two-way paralleling two-dimensional convolutional neural network(TWP-2D-CNN)is put forward to improve the recognition rate of rolling bearing fault diagnosis.A batch normalization algorithm(BN)is introduced between the two-way convolutional layer and the pooling layer,dropout algorithm incorporated between three fully connected networks,which aims at enhancing the non-linear mapping ability of the model to the features of two-dimensional time-frequency images.Meanwhile,while retaining the basic convolutional layer,a dilated convolution layer structure is added to expand the receptive field for two-dimensional joint time-frequency images.Through the t-sne visualization of the output value of the middle layer,the generalization ability of different diagnostic models in the layer-by-layer mapping expression is assessed,and the robustness is tested under noise of different intensities.The results of diagnosis indicate that based on the training and testing of the TWP-2D-CNN model,the diagnostic recognition rates of joint time-frequency fault images is increased to 99.803%and 99.396%,respectively.And a recognition rate of 95.934%is achieved under three different noise levels,which is better than the corresponding value of the basic 2D-CNN model and the deep neural network(DNN).