基于多特征的Res-BiLSTM滚动轴承故障诊断

Res-BiLSTM Rolling Bearing Fault Diagnosis Based on Multiple Features

针对故障诊断中存在的由于特征提取单一或特征提取缺失导致的诊断准确率较低这一问题,提出一种基于多特征的Res-BiLSTM滚动轴承故障诊断方法。首先通过时域分析法和残差网络对时域特征值和空间特征进行提取并进行特征融合,其次将融合的特征值输入双向长短期记忆网络进行时序特征提取,最后利用Softmax进行分类。实验结果表明,该故障诊断方法增强了故障诊断网络的特征提取能力,进而提高了诊断准确率。

To address the problem of low diagnostic accuracy due to single feature extraction and missing feature extraction in fault diagnosis,Res-BiLSTM rolling bearing fault diagnosis method based on multiple features is proposed.Firstly,the time domain feature values and spatial features are extracted and feature fusion are performed by time domain analysis and residual network,then the fused feature values are input to the bidirectional long and short-term memory network for temporal feature extraction,and finally Softmax is used for classification.The experimental results show that this fault diagnosis method enhances the feature extraction ability of the fault diagnosis network,thus improving the diagnosis accuracy.