基于多尺度CNN-BiLSTM的轴承故障诊断

Bearing fault diagnosis based on multi-scale CNN-BiLSTM

针对传统故障诊断方法提取特征不丰富、未充分利用时序特征的问题,提出了一种基于多尺度CNN和BiLSTM融合的滚动轴承故障诊断方法。首先设计多尺度CNN模型进行多尺度特征信息的提取;其次设计BiLSTM模型进行提取特征前后之间的内部关系;最后通过全连接层构建了特征信息与故障类型的映射,通过softmax分类器输出故障诊断结果。以准确率为评价指标,该方法在多负载场景下诊断准确率为99.2%,在变负载场景下诊断平均准确率为89.6%。实验结果表明,该方法具有良好的自适应工况的能力。

Aiming at the problem that the traditional fault diagnosis methods are not rich in feature extraction and the time series feature is not fully utilized,a rolling bearing fault diagnosis method based on multi-scale CNN and BiLSTM is proposed.First,design a multi-scale CNN model to extract multi-scale feature information;secondly,design a BiLSTM model to extract the internal relationship between before and after the feature;finally,build a mapping between feature information and fault types through a fully connected layer,and output fault diagnosis through the softmax classifier result.Taking the accuracy as an evaluation index,the method has a diagnostic accuracy rate of 99.2%in a multi-load scenario,and an average diagnostic accuracy rate of 89.6%in a variable load scenario.The experimental results show that this method has a good ability to adapt to working conditions.