基于CNN的变工况滚动轴承故障诊断研究

针对滚动轴承工作环境多变和样本不足导致故障诊断效果不佳的问题,提出一种多模态注意力卷积神经网络.该网络采用多个并行卷积层构建,并结合注意力机制,有效地提取了丰富的故障特征.然后提出了两种有限数据条件下的数据增强方法,解决了数据样本不足的问题.另外,将采集到的滚动轴承时域信号通过小波变换转换为时频图谱作为网络输入来提高数...

Research on Rolling Bearing Fault Diagnosis Under Variable Working Conditions Based on CNN

To solve the problem of poor fault diagnosis caused by the variable working environment of rolling bearings and insufficient data samples, a method called multi-modal attention convolutional neural network is proposed. This network is constructed by multiple parallel convolutional layers, and combines the attention mechanism to effectively extract rich fault features. Then, two data enhancement methods under limited data conditions are proposed to solve the problem of insufficient data samples. In addition, wavelet transform is used to convert the collected time-domain signals of rolling bearings into time-frequency spectrums as network input to improve data quality. The proposed method is experimentally analyzed by using fault data under multiple frequency conversions. The experimental results show that the method has high accuracy and obvious clustering effect under variable working conditions, which indicates that the method can effectively improve the accuracy of bearing fault diagnosis under variable working conditions and has a good application value.