基于MSCNN与STFT的滚动轴承故障诊断研究

针对现有基于CNN(Convolution Neural Network)的滚动轴承故障诊断方法难以有效挖掘和利用数据中包含的多尺度信息问题,提出了一种多尺度卷积特征融合的滚动轴承故障诊断方法。加入上采样层,通过递归方式建立具有多尺度特征提取和融合能力的卷积神经网络MSCNN(MultiScale Convolution Neural Network)结构,提升模型对输入信号的理解能力。利用美国凯斯西储大学(CWRU)数据库对所提方法的有效性进行验证,采用短时傅里叶变换对滚动轴承信号进行频谱分析,将频谱样本输入到MSCNN网络中,数据分析表明,该方法能有效地提升故障的诊断精度。

Research of Fault Diagnosis of Rolling Bearing based on MSCNN and STFT

In view of the existing CNN(Convolution Neural Network)based rolling bearing fault diagnosis method,it is difficult to effectively excavate and utilize the multi-scale information contained in the data,a multi-scale convolution feature fusion method for rolling bearing fault diagnosis is proposed.A convolutional neural network(MSCNN)structure with multi-scale feature extraction and fusion ability is built by adding the upper sampling layer,which improves the understanding ability of the model to the input signal.CWRU database is used to verify the validity of the proposed method.The short-time Fourier transform is used to analyze the spectrum of rolling bearing signals.The spectrum samples are input into MSCNN network.The data analysis shows that the method can effectively improve the fault diagnosis accuracy.