基于SWDAE-SVC的矿用齿轮箱自监督故障诊断方法  

The self-supervised fault diagnosis method for mine gearbox based on SWDAE-SVC

作　　者：李鑫 Li Xin(School of Mechatronic Engineering,China University of Mining and Technology,Jiangsu Xuzhou,221116,China)

机构地区：[1]中国矿业大学机电工程学院,江苏徐州221116

出　　处：《机械设计与制造工程》2023年第10期21-24,共4页Machine Design and Manufacturing Engineering

基　　金：中央高校基本科研业务费项目(20230N1048);江苏省自然科学基金(BK20231064)。

摘　　要：针对矿用齿轮箱振动数据易受噪声污染且故障类别标注困难问题,提出了一种基于栈式小波降噪自编码器(SWDAE)和支持向量聚类(SVC)的自监督故障诊断方法。首先,将小波映射函数引入栈式降噪自编码器(SDAE)模型,以实现强噪声下矿用齿轮箱的敏感故障特征提取。然后,利用所得高层抽象特征构建SVC模型,以实现无标签信息下的矿用齿轮箱故障诊断。实验结果表明,所提SWDAE-SVC方法具有优异的故障诊断性能。Aiming at the problems that the vibration data of mine gearboxes are easily polluted by noise and fault categories are difficult to label,a self-supervised fault diagnosis method based on stacked wavelet denoising autoencoder(SWDAE)and support vector clustering(SVC)is proposed.The wavelet mapping function is introduced into the stacked denoising autoencoder(SDAE)model to realize the sensitive fault feature extraction of the mine gearbox under strong noise.Then,the obtained high-level inherent features are used to construct the SVC model to achieve the fault diagnosis of mining gearboxes without label information.Experimental results show that the proposed SWDAE-SVC method has excellent fault diagnosis performance.

关 键 词：故障诊断 栈式降噪自编码器 小波映射函数 支持向量聚类 矿用齿轮箱 

分 类 号：TH17[机械工程—机械制造及自动化]