基于EEMD-DCNN方法的电机轴承故障状态诊断研究

Research on Motor Bearing Fault Diagnosis based on EEMD-DCNN Method

为了提高电机轴承故障状态诊断精度,设计了一种基于EEMD和DCNN的融合方法,实现局部特征的提取和特征图的输出。DCNN从5个不同单元中提取特征,通过全连接层整合相关特征且利用Softmax层完成分类工作,实现对数据短时特征的提取。研究结果表明:采用的方式与BPNN、SVM、WDCNN相比,有着比较显著的优势,诊断效果较为平稳,且数次实验获得的指标标准差是0。该研究可以拓宽到其它的机械传动领域,具有很好的市场应用价值。

In order to improve the diagnosis accuracy of motor bearing fault state,a fusion method based on EEMD and DCNN is designed to extract local features and output feature maps.DCNN extracts the features from five different units,integrates the related features through the full connection layer and completes the classification work by using Softmax layer,so as to realize the short-time feature extraction of data.The research results show that the adopted method has the significant advantages compared with BPNN,SVM and WDCNN as well as the relatively stable diagnostic effect,and the indicator deviation of several experiments is 0.This research can be extended to other mechanical transmission fields,which has good value of market application.