| 朱怡琳1，2， 库鹏博1，2， 张守京1，2.基于CBAM的元学习小样本变工况轴承故障诊断[J].轻工机械,2025,43(1):55-62 |
| 基于CBAM的元学习小样本变工况轴承故障诊断 |
| CBAM Based Deep Meta Learning for Few Shot Variable Operating Condition Bearing Fault Diagnosis |
| |
| DOI：10.3969/j.issn.1005 2895.2025.01.008 |
| 中文关键词: 轴承 故障诊断 小样本 变工况 元学习 卷积注意力模块 |
| 英文关键词:bearing fault diagnosis few shot varying working conditions meta learning CBAM（Convolutional Block Attention Module） |
| 基金项目:西安市现代智能纺织装备重点实验室课题(2019220614SYS021CG043)。 |
|
| 摘要点击次数: 7 |
| 全文下载次数: 8 |
| 中文摘要: |
| 针对在实际生产中轴承故障数据的样本量少、不同故障类别样本分布不均衡等影响轴承故障诊断的问题，课题组提出了一种基于注意力机制的元学习方法。首先通过连续小波变换将一维振动信号转化为二维图像，并将二维图像作为网络的输入，然后采用卷积注意力模块（Convolutional Block Attention Module，CBAM）预训练特征提取器和分类器对大规模已知轴承故障的轴承数据进行预训练，增强网络对时频图特征的表示能力和分类性能，从而提高轴承故障诊断的准确性和鲁棒性；考虑了多种工况在CWRU数据集上进行交叉验证。结果显示该方法在小样本故障诊断的1 shot和5 shot任务上优于其他故障诊断方法，表明所提方法具有较高的鲁棒性和泛化性。 |
| 英文摘要: |
| Aiming at the problems that affect bearing fault diagnosis in actual production, such as small sample size of bearing fault data and uneven distribution of samples of different fault categories. A meta learning method based on the attention mechanism was proposed by research group. First, the one dimensional vibration signal was converted into a two dimensional image by a continuous wavelet transform and used as input to the network. Then, the feature extractors and classifiers were pre trained by using the Convolutional Block Attention Module (CBAM), and the large bearing datas with known bearing faults were used for the pre training, which enhances the network′s ability to represent the features of time frequency diagrams and its classification performance to improve the accuracy and robustness of the diagnosis of bearing faults. A variety of different operating conditions were considered for cross validation on the CWRU dataset. The results show that the proposed method is superior to other fault diagnosis methods in 1 shot and 5 shot tasks of small sample fault diagnosis, indicating that the proposed method has high robustness and generalization. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
