基于多重同步挤压变换与深度脊波卷积自编码网络的滚动轴承故障诊断

Fault Diagnosis of Rolling Bearing Based on MSST and Deep Ridgelet Convolutional Auto-encoder Network

利用传统故障诊断方法对滚动轴承进行诊断时,存在故障特征提取困难以及故障模式难以辨识的问题。针对此问题,提出了一种基于多重同步挤压变换以及深度脊波卷积自编码网络的智能故障诊断方法。首先,利用多重同步挤压变换处理含噪信号能力强、具有优越的时频分解特性的特点,将采集的轴承故障信号进行MSST处理,得到分辨率较高的时频图像。然后,利用深度脊波卷积自编码网络自身泛化性能强、能够有效挖掘数据特征的特点,建立深度脊波卷积自编码网络识别模型。将降维至适当大小的时频图像输入到该模型系统中,进行自动特征提取和故障识别。实验结果表明,该方法提取故障特征信号能力较高,并能够有效地识别出不同的故障类型。

When using traditional fault diagnosis methods to diagnose rolling bearing,it is difficult to extract fault features and identify fault patterns.To solve this problem,an intelligent fault diagnosis method based on Multi synchrosqueezing transform(MSST)and Deep ridgelet convolutional autoencoder network(DRCAN)is proposed.Firstly,the MSST is used to process the noisy signal with strong ability and excellent time-frequency decomposition characteristics.The collected bearing fault signal is processed by MSST to obtain a high-resolution time-frequency image.Then the DRCAN has strong generalization performance,which can effectively mine the characteristics of data features,and establish the DRCAN recognition pattern.The time-frequency image reduced to the appropriate size is input into this model system for automatic feature extraction and fault recognition.Experimental results show that the method has high ability to extract fault feature signals and can effectively identify different fault types.