小波变换和深度残差收缩网络在齿轮箱故障诊断中的应用OA北大核心CSTPCD

Accurate fault diagnosis of gears is an effective means to ensure stable and reliable operation of rotating machinery.Aiming at the problem of gear fault classification in gearboxes under strong noise environment,a fault diagnosis model based on continuous wavelet transform and deep residual shrinkage network is proposed.Firstly,wavelet transform is used to analyze the vibration data of one-dimensional time series,and it is converted into a two-dimensional time-frequency map as the input of the deep residual shrinkage network(DRSN).Secondly,based on the multi-layer convolutional neural network,the cross-layer identity connection in the residual structure is added to solve the problem of gradient disappearance and explosion,and then the adaptive threshold sub-network is used to achieve soft threshold noise reduction.Finally,the time-frequency map of the fault sample is used as the input of the diagnosis model to achieve fault classification.The experimental results show that compared with other models,the fault diagnosis method is easier to identify fault features,and the classification accuracy rate reaches 99.15%.