强噪声工况下滚动轴承的CDAE-ResBiLSTM故障诊断方法

打开文本图片集

关键词：滚动轴承；故障诊断；强噪声工况；卷积降噪自编码器；双向长短期记忆网络；残差收缩单元中图分类号：TH133.33 文献标志码：A 文章编号：1671-6841（2025）05-0069-09DOI：10. 13705/j. issn.1671-6841.2023244

Abstract：The strong noise environment during the operation of rolling bearings pose dificulties in extracting data features.Aiming at the problem of fault diagnosis of roling bearings with strong noise conditions，an improved residual shrinkage building unit （IRSBU） was constructed using residual network， semi-soft threshold function，APReLU activation function，and SENet attntion mechanism．A rolling bearing fault diagnosis method combining convolutional denoising autoencoder （CDAE） and improved residual shrinkage bi-directional long short term memory network （ResBiLSTM）was proposed. Firstly， Gaussian noise was added to the one-dimensional raw signal to simulate strong noise conditions，and the noisy dataset was input into CDAE for feature extraction. Then，the low dimensional denoising features of the hidden layer encoded by the encoder was input into ResBiLSTM for fault diagnosis. Finally，the proposed method was experimentally validated using the Case Western Reserve University bearing dataset （CWRU） and the Xi ' an Jiaotong University bearing dataset （XJTU-SY）. The experimental results showed that the CDAE-ResBiLSTM model had good feature extraction ability and noise resistance.

Key words：roling bearing；fault diagnosis； strong noise condition；convolutional denoising autoencoder; bi-directional long short term memory network; residual shrinkage building unit

0 引言

滚动轴承是旋转机械的重要部件之一，在工程任务、列车出行等多个领域有着显著的应用。（剩余11952字）