Research On Intelligent Diagnosis Method Of Planetary Gearbox Based On Deep Learning

As a key component of rotating machinery,the condition monitoring and fault diagnosis analysis of planetary gearbox are of great significance,not only for safe production and economic growth,but also for avoiding industrial accidents and reducing casualties.Using modified experience wavelet transform,sparse coding shrinkage,convolutional neural network,transfer neural network and other technical methods,this paper takes the vibration signal of the gear components in the planetary gearbox as the research object,and conducts research from two aspects of signal feature extraction and intelligent diagnosis.The paper proposes a method based on deep learning to intelligently identify the fault of planetary gearboxes.The main research contents of the paper are as follows:(1)Introduced the background and research significance of the subject,expounded the current research status of planetary gearbox fault diagnosis at home and abroad,and investigated the research work done by domestic and foreign scholars on planetary gearbox fault diagnosis from the aspects of signal feature extraction and intelligent classification methods.Then it leads to the research object and research direction of this article.(2)Systemically carry out research on the failure mechanism of planetary gearbox gear parts.Starting from the common failure modes of gears and their causes,the characteristics of vibration signals generated by different gear components are analyzed,and the sun gear,planetary gear,and ring gear failure simulations are established.built a planetary gearbox failure test bed,collect vibration data under different speeds and different test beds,and constructed a complete dataset.(3)Empirical Wavelet Transform is inadequate in the early weak diagnosis fault of planetary gearboxes under strong noise background,mainly due to improper segmentation of the signal spectrum,which cannot effectively determine the number of modal components.Consequently,an early fault diagnosis method of planetary gearbox is proposed,which combines the Modified Empirical Wavelet Transform and Adaptive Sparse Coding Shrinkage algorithms.It is applied to the fault diagnosis of planetary gearbox,and the effectiveness of the proposed method is proved by simulation signals and actual collected signals.(4)Aiming at the identification of different fault states of planetary gearbox gears and the rational selection of convolutional network parameters,a gear fault diagnosis method based on MEWT and CNN is proposed.First,the MEWT is used to decompose the signal,and one-dimensional data samples are generated by overlapping sampling,and the network structure experiment and parameter experiment are performed to determine the convolutional network and the network parameters of each layer.The effectiveness of the proposed fault classification method is verified through multiple types of fault data sets,and different degrees of noise are added to the training data to verify the noise immunity of the proposed method.The data sets under different working conditions verify the transferability of the proposed method.(5)In the case of actual fault diagnosis,there is less labeled data,training data and test data are distributed differently,and the diagnosis effect of normal volume integral network model is not good.The Transferable fault diagnosis method based on MEWT and Transferable Convolutional Neural Network is proposed.The validity of the proposed method is verified through the data sets collected by two different test benches,and the fault diagnosis of the planetary gearbox across equipment is realized.