基于MTF-DenseNet的滚动轴承故障诊断

Fault Diagnosis of Rolling Bearing Based on MTF-DenseNet

针对滚动轴承振动信号的特征难以提取的问题,提出一种基于MTF-DenseNet的滚动轴承故障诊断方法。利用马尔可夫变迁场(MTF)将滚动轴承的一维时间序列信号进行编程,生成二维图像,从而很好地保留时间序列信号的时间依赖性和频率结构,之后将其输入到密集连接卷积网络(DenseNet)实现故障特征的提取,进而实现故障诊断。采用凯斯西储大学轴承数据集上数据进行实验,实验结果表明,该方法在数据集上的故障分类准确度达到99.35%,故障诊断效果良好。

Aiming at the problem that it is difficult to extract the characteristics of rolling bearing vibration signal,a rolling bearing fault diagnosis method based on MTF-DenseNet is proposed.Markov transition field(MTF)is used to program the one-dimensional time series signal of rolling bearing to generate two-dimensional image,so as to retain the time dependence and frequency structure of time series signal,and then input it to DenseNet to extract fault features,so as to realize fault diagnosis.The experimental results show that the fault classification accuracy of this method on the data set is 99.35%,and the fault diagnosis effect is good.