【摘要】 针对现有滚动轴承故障诊断算法在噪声背景下诊断精度低、稳定性差的问题，提出了基于FSWT-DRSN的滚动轴承故障诊断方法。首先，将一维振动信号进行频率切片小波变换，生成时频图。其次，结合软阈值、注意力机制和深度残差网络构建深度残差收缩网络(DRSN)，实现对含噪样本自适应设置阈值，提升网络的抗噪性和稳定性。最后，将预处理过的时频图输入模型进行训练，实现滚动轴承故障分类。与现有的SVM、CNN和ResNet算法相比，FSWT-DRSN诊断精度更高且稳定性好，在噪声干扰下有出色的诊断性能。更多还原

【Abstract】 Aiming at the problems of low accuracy and poor stability of existing bearing fault diagnosis algorithms under the background of noise,a fault diagnosis method of rolling bearing based on FSWT-DRSN( frequency slice wavelet transform-deep residual shrinkage network) is proposed in this paper. Firstly,one-dimensional vibration signal is transformed by frequency slice wavelet to generate time-frequency graph. Secondly,a deep residual shrinkage network is constructed by combining soft threshold,attention mechanism and deep residual network to realize adaptive threshold setting for noisy samples and improve the anti-noise performance and stability of the network. Finally,the pretreated time-frequency graph is input into the model training to realize fault classification of rolling bearings. Compared with the existing SVM,CNN and ResNet algorithms,FSWTDRSN has the highest diagnostic accuracy and the best stability,which proves that FSWT-DRSN has excellent diagnostic performance under noise interference.更多还原