基于深度迁移自编码器的变工况下滚动轴承故障诊断方法

Bearings Fault Diagnosis Method Based on Deep Transfer Auto-encoder under Variable Working Conditions

摘要:

在实际工业场景下的轴承故障诊断，存在轴承故障样本不足，训练样本与实际信号样本存在分布差异的问题。本文提出一种新的基于深度迁移自编码器的故障诊断方法FS-DTAE，应用于不同工况下的轴承故障诊断。该方法首先采用小波包变换进行信号处理与特征提取；其次，采用提出的基于朴素贝叶斯与域间差异的特征选取（FSBD）方法对统计特征进行评估，选取更有利于跨域故障诊断和迁移学习的特征；然后，利用源域特征数据训练深度自编码器，将训练得到的模型参数迁移至目标域，再利用目标域正常状态样本对深度迁移自编码器模型进行微调，微调后的模型用于目标域无标签特征数据的故障分类。最后，基于CWRU轴承故障数据开展不同工况下故障诊断实验，结果表明，所提出的FS-DTAE方法能够有效提高不同工况下的故障诊断准确率。

Abstract:

Bearing fault diagnosis in the actual industrial scene, there are some problems, such as the lack of bearing fault samples, and the distribution difference between the training samples and the actual signal samples. A new fault diagnosis method based on deep transfer auto-encoder is proposed in this paper, which is applied to the fault diagnosis of bearings under different working conditions. Firstly, wavelet packet transform is used for signal processing and feature extraction; Secondly, the proposed feature selection method based on Naive Bayes and difference between domains (FSBD) is used to evaluate the statistical features and select the features that are more conducive to cross-domain fault diagnosis and transfer learning; the source domain feature data is used to train the deep auto-encoder, and parameters of the trained model are migrated to the target domain. Then, the normal state samples of the target domain are used to fine-tune the deep transfer auto-encoder model, and the fine-tuned model is used for fault classification of the target domain unlabeled feature data. Finally, based on the CWRU bearing fault data, fault diagnosis experiments under different working conditions are performed. The results show that the proposed FS-DTAE method can effectively improve the fault diagnosis accuracy under different working conditions.