Research On Gearbox Fault Diagnosis Method Based On Deep Learning

In recent years,under the background of intelligent manufacturing and Big data,the fault diagnosis technology of mechanical equipment is developing in the direction of efficiency and intelligence.As an indispensable part of mechanical equipment,the fault detection and analysis research of gearbox is of great significance to the safe and reliable operation of equipment and the avoidance of casualties.At present,the research on gearbox fault diagnosis methods based on deep learning is a hot topic in the field of fault diagnosis.However,the existing fault diagnosis methods have the problems of low diagnostic accuracy and weak generalization ability.Therefore,this paper uses Convolutional neural network and Transfer learning to study the problem of gearbox fault diagnosis,and proposes new fault diagnosis models and methods to improve the accuracy of gearbox fault diagnosis and the performance of the model.The main content of the paper research is as follows:(1)Introduce the research background and significance of the topic,elaborate on the current research status of gearbox fault diagnosis at home and abroad,and summarize and analyze the research on gearbox fault diagnosis from both machine learning and deep learning,thus leading to the research content of this article.(2)Analyzed the forms of failure of components such as gears and bearings in the gearbox and the causes of gearbox failures.The structure and training process of each layer of the Convolutional neural network model in the deep learning algorithm,as well as the basic concepts,classifications and methods of Transfer learning,are introduced in detail,providing a theoretical basis for the following research.(3)Aiming at the problems of low accuracy and weak generalization ability of classical one-dimensional Convolutional neural network model in gearbox fault diagnosis,a gearbox fault diagnosis method based on improved one-dimensional Convolutional neural network model is proposed.The first convolution layer of the model uses a large convolution kernel,and the rest uses a small convolution kernel to obtain a larger Receptive field.After the convolution layer,a batch normalization layer is added to improve the Rate of convergence of the model.In the forward propagation,a drop out method is added to the full connection layer to suppress the problem of over fitting.Adam’s adaptive algorithm is used to optimize the parameters of the model.The experimental results show that compared to other comparative methods,the proposed method has significant improvements in fault diagnosis accuracy and model generalization ability.(4)The method proposed in Chapter 3 is mainly applicable to fault diagnosis under the same working condition sample data.In order to study the gearbox fault diagnosis problem under different working condition sample data,a fault diagnosis method based on Transfer learning is proposed.In this method,the improved one-dimensional Convolutional neural network model in Chapter 3 is used as the pre training model of Transfer learning.After the pre training model is trained with one working condition sample data,the model is transferred to different working condition sample data for testing.Experiments based on sample data of bearings and gears show that,compared with the method proposed in Chapter 3,this method also achieves high fault diagnosis accuracy,and the Rate of convergence of the model is faster and the stability is good.