Research On Fault Diagnosis Method Of Gearbox Based On Improved Convolution Neural Network

As one of the key components of modern machinery,gearbox has been widely used in various industrial fields.With the growth of operating time,the gear,bearing,and other components inside the gearbox are likely to undergo varying degrees of wear,which can cause serious safety accidents due to faults at any time.Therefore,accurate diagnosis of the operating status of gearbox is one of the important means to ensure industrial production safety.However,the working environment of gearbox is complex,the working conditions are variable,and there are noise disturbances.At the same time,the structure of gearbox used in modern industry is more precise and complex,and there are multiple components that are damaged to form composite faults,which makes it difficult to extract and identify weak fault features.To address the above issues,this paper proposes three models based on improved Convolutional neural networks to achieve end-to-end intelligent fault diagnosis for gearbox.The main research work is as follows:(1)Aiming at the problem that Convolutional neural network are difficult to deeply dig temporal features,a CNN-ABi GRU fault diagnosis method based on Convolutional neural network,Bidirectional gated recurrent unit network,and Attention mechanism is proposed.Convolutional neural networks can extract local features from vibration signals,while Bidirectional gate recurrent unit network can dig temporal features from signal data.Combining the two can extract high-level features with local and temporal features.The Attention mechanism then filters the extracted features to remove redundant features,and finally completes fault classification through Softmax.This paper selects the gearbox data set from Southeast University and the rolling bearing data set from Case Western Reserve University for experimental verification.The experimental results show that the method has high diagnostic accuracy for gearbox.(2)Aiming at the problem that it is difficult to collect sufficient data for deep learning model training in some operating conditions of gearbox in a real industrial environment,a fault diagnosis method for gearbox under cross operating conditions based on CNN-ABi GRU and Transfer learning is proposed.Transfer learning allows the model to train the ability to extract fault features on a sufficiently sampled condition dataset.After training,a small number of samples can be fine-tuned on a sparsely sampled condition dataset to achieve fault identification on the corresponding dataset.This paper selects the gearbox data set from Southeast University and the bearing data set from Case Western Reserve University for experimental verification.The experimental results show that the method can achieve high accuracy cross working condition fault diagnosis.(3)To solve the problem of difficult identification of fault signals caused by mixed noise in gearbox,a CARes Net-Bi GRU fault diagnosis method based on Residual neural network,Channel attention mechanism,and Bidirectional gate recurrent unit network is proposed.The Residual neural network,as an improved Convolutional neural network,solves the problem of network degradation caused by too many layers of the network,making it possible for fault diagnosis models to build deeper network structures to enhance the ability of feature extraction.In combination with the channel attention mechanism,it can adaptively adjust the characteristics of the importance of feature channels in the Residual neural network,effectively suppressing noise interference.This paper selects the gearbox dataset from Southeast University and the industrial gearbox dataset from the Prediction and Health Management System Data Challenge Competition 2009 to complete the experimental verification of the proposed method.The experimental results show that the method has strong noise resistance.