Abstract: In view of the problem that the traditional fault diagnosis method of rolling bearing depends on the prior knowledge and expert experience too much,and the low recognition rate of some faults with single signal,a method of rolling bearing fault diagnosis based on information fusion under variable working condition was proposed.The test and multi-information acquisition system of rolling bearing was built.The metro traction motor bearing NU216 was selected as the research object and to prefabricate the defects,and the data of acoustic emission and vibration acceleration signals during the bearing test was acquired.The original signal was processed and extracted by wavelet packet decomposition,and the normalized feature information was fused by convolution neural network.The two-dimensional convolution neural network model was established to diagnose the bearing fault of metro traction motor under different conditions.The test results showed that the intelligent fault diagnosis method of subway traction motor bearing based on information fusion under variable working conditions could accurately identify the fault type of bearing while the load and speed change.When the neural network training set and the test set covered the same working conditions,the accuracy could reach 100%.

Key words: metro traction motor bearing, convolution neural network, information fusion, fault diagnosis

摘要： 针对传统滚动轴承故障诊断方法过于依赖先验知识和专家经验,以及单一信号对某些故障识别率偏低的问题,提出一种变工况下基于信息融合的地铁牵引电机轴承故障诊断方法。首先搭建滚动轴承试验与多信息采集系统;其次对地铁牵引电机轴承进行缺陷预制并采集轴承试验过程中的声发射和振动信号;然后用小波包分解对原始信号进行处理并提取特征,再用卷积神经网络对归一化后的特征信息进行融合;最后建立二维卷积神经网络模型,对不同工况下的地铁牵引电机轴承故障进行智能诊断。试验结果表明：变工况下基于信息融合的地铁牵引电机轴承故障智能诊断方法,可在载荷和转速变化的情况下准确识别轴承的故障类型,当神经网络训练集与测试集涵盖工况相同时,准确度可达100%。

关键词: 地铁牵引电机轴承, 卷积神经网络, 信息融合, 故障诊断