基于包络谱语义构建的零样本滚动轴承复合故障诊断方法  

Zero-shot rolling bearing compound fault diagnosis method based on envelope spectrum semantic construction

作　　者：甄冬[1,2] 孙赫明 冯国金 崔展博 田少宁 孔金震 ZHEN Dong;SUN Heming;FENG Guojin;CUI Zhanbo;TIAN Shaoning;KONG Jinzhen(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300401,China;Advanced Equipment Research Institute Co.,Ltd.of HEBUT,Tianjin 300401,China;Shijiazhuang Haishan Industrial Development Corporation,Shijiazhuang 050208,China)

机构地区：[1]河北工业大学机械工程学院,天津300401 [2]天津河工大先进装备研究院有限公司,天津300401 [3]石家庄海山实业发展总公司,石家庄050208

出　　处：《振动与冲击》2024年第14期189-200,283,共13页Journal of Vibration and Shock

基　　金：国家自然科学基金面上项目(52275101);天津市科技计划项目(21JCZDJC00720);2022年河北省春晖计划(E2022202047)。

摘　　要：深度学习算法在训练集完备的情况下可以实现较高的故障识别率,然而在真实工业场景中,滚动轴承的多种故障可能复合存在,通常难以获取充足的数据用于训练。为解决该问题,提出了一种基于包络谱语义构建的零样本复合故障诊断方法,在训练阶段使用单一故障数据构建了一个语义空间和一个特征空间,然后在识别阶段通过语义空间和特征空间的复合,实现对零样本情况下的复合故障识别。此外,考虑到包络谱能很好地表征滚动轴承故障特征,采用包络谱预处理故障信号以增强轴承故障的特征,并借助信号包络谱的物理含义来构建轴承单一故障和复合故障的语义。试验结果显示,所提模型在复合故障识别上取得了87.83%的准确率,优于对比模型。In real industrial environment,various compound faults may coexist in rolling bearings,and it is usually difficult to acquire sufficient sample data for training.To address this issue,a zero-shot compound fault diagnosis approach was proposed based on envelope spectrum semantic construction.During the training phase,a semantic space and a feature space were established using single fault data.Subsequently,during the recognition phase,compound fault recognition in zero-shot scenarios was realized through the combination of the semantic and feature spaces.Furthermore,recognizing the envelope spectrum’s capability in effectively characterizing rolling bearing fault features,the fault signals were preprocessed using envelope spectrum to enhance the bearing fault characteristics.The physical significance of the signal envelope spectrum was leveraged to construct the semantics for both single and compound bearing faults.The experimental results reveal that the proposed model achieves an accuracy of 87.83%in compound fault recognition,outperforming the compared models.

关 键 词：滚动轴承 复合故障诊断 零样本 包络谱 语义构建 

分 类 号：TH17[机械工程—机械制造及自动化] TH133