Research On Hob Vibration Signal Noise Reduction And Feature Enhancement And Wear State Identification Method

In the long-term high-speed hobbing process,the hob will inevitably produce wear,and if the tool wear is not detected in time,it will lead to the deterioration of the surface quality and accuracy of the machined gears.Therefore,in order to ensure the quality of gear processing,it is necessary to carry out research on the identification of the wear state of hobbing tools.This paper takes the hob of high-speed dry-cutting hobbing machine as the research object,and carries out the research of hob vibration signal noise reduction and feature enhancement method and hob wear state identification method for the problem of hob wear state identification under complex working conditions.The main research contents of this paper include:(1)Research on hob vibration signal decomposition method based on variational mode decomposition.In the early stages of wear,the wear-related characteristics of the hob are weak and the signal contains a lot of ambient noise and other vibration sources,resulting in the relevant characteristics being masked.In order to fully extract the processing feature information masked by noise,the vibration signal needs to be decomposed first.Aiming at the problem that the empirical modal decomposition and its improvement algorithm have the phenomenon of mode mixing when decomposing the signal,the mean energy entropy is used as the fitness function,and the advantages of the gradient-based optimizer are combined to propose a hob vibration signal decomposition method based on the parameter adaptive variational mode decomposition,which realizes the adaptive setting of the hyperparameters of the variational mode decomposition(the number of modes K and the quadratic penalty factor α)and the optimal decomposition.(2)Research on hob vibration signal noise reduction and feature enhancement method.Aiming at the problem that the modulation and impact characteristics of hob vibration signals lead to the evaluation bias of traditional time-domain statistical feature indexes in the evaluation of signal components.An enhanced periodic modulation intensity index based on autocorrelation analysis is proposed for the attribute evaluation of submodular components,and a signal weighting reconstruction strategy is further proposed to realize the noise reduction and feature enhancement of hob vibration signal,and the effectiveness of the proposed method is verified by envelope spectrum analysis of the results obtained by other methods.(3)Research on hob wear state recognition method based on deep convolutional neural network.In view of the problem that most of the current methods require a priori knowledge to extract the corresponding wear features,a hob wear recognition method based on short-time Fourier transform and deep convolutional neural network is proposed on the basis of signal noise reduction and processing feature enhancement,which effectively reduces the dependence on feature extraction means and a priori knowledge and realizes the end-to-end hob wear recognition.The proposed method has a higher recognition rate than other comparative methods.In addition,the effects of different hyperparameters on model performance are analyzed and the proposed model’s focus on learning different classes of features is visually explained using the Grad-CAM visualization method.