摘要
为解决变工况条件下源域样本和目标域样本分布差异导致不同域样本特征动态变化明显以及诊断准确率较低的问题,提出1种基于流形学习的测地流核函数(geodesic flow kernel,GFK)域自适应故障诊断方法,该方法在格拉斯曼流形中构造测地流式核,将频域特征转换为与故障信息相关的固有流形特征,减小源域与目标域在特征空间上的分布差异。研究结果表明:所提方法的平均诊断准确率较KNN、SAE、SA、CORAL方法在滚动轴承上分别提高24.98,12.18,8.14和6.36个百分点,在RV减速器行星轮上分别提高31.38,20.55,20.25和22.19个百分点,在齿轮上分别提高62.71,46.05,51.34和43.84个百分点,可有效提高不同旋转设备变工况条件下的诊断准确率。研究结果可为变工况条件下的旋转设备故障诊断研究提供参考。
In order to solve the problem that the distribution difference of samples in source domain and target domain under variable working conditions will lead to obvious dynamic change in the characteristics of samples in different domains and low diagnosis rate,an adaptive fault diagnosis method of geodesic flow kernel(GFK)function domain based on manifold learning was proposed.The geodesic flow kernel was constructed in the Glasman manifold,and the frequency domain features were converted into the inherent manifold features related to fault information,which reduced the distribution difference between the source domain and the target domain in the feature space.The results show that compared with KNN,SAE,SA and CORAL methods,the average accuracy of the proposed method is increased 24.98,12.18,8.14 and 6.36 percentage points respectively for rolling bearing,and 31.38,20.55,20.25 and 22.19 percentage points for RV reducer planetary wheel,respectively.In the gear,the improvement is 62.71,46.05,51.34 and 43.84 percentage points,respectively,and the diagnostic accuracy of rotating equipment under different working conditions is improved effectively.The research results can provide reference for the research on fault diagnosis of the rotating equipment under variable working conditions.
作者
普会杰
刘韬
刘畅
周俊
缪护
PU Huijie;LIU Tao;LIU Chang;ZHOU Jun;MIAO Hu(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming Yunnan 650500,China;Key Laboratory for Advanced Equipment Intelligent Manufacturing Technology of Yunnan Province,Kunming University of Science and Technology,Kunming Yunnan 650500,China;Kunming Yunnei Power Co.,Ltd.,Kunming Yunnan 650500,China)
出处
《中国安全生产科学技术》
CAS
CSCD
北大核心
2023年第8期209-216,共8页
Journal of Safety Science and Technology
基金
云南省科技厅重大科技专项计划项目(202202AC080008)
国家自然科学基金项目(52065030)。
关键词
故障诊断
设备安全
变工况
迁移学习
流形学习
fault diagnosis
equipment safety
variable working condition
transfer learning
manifold learning
