聚类量化在风机轴承退化评估中的应用  

Application of Cluster Quantization in Wind Turbine Bearing Degradation Assessment

作　　者：张磊[1] 陈长征[1] 周丽婷 杨明政 ZHANG Lei;CHEN Changzheng;ZHOU Liting;YANG Mingzheng(School of Mechanical Engineering,Shenyang University of Technology,Liaoning Shenyang 110870,China;School of Artificial Intelligence,Shenyang University of Technology,Liaoning Shenyang 110870,China)

机构地区：[1]沈阳工业大学机械工程学院,辽宁沈阳110870 [2]沈阳工业大学人工智能学院,辽宁沈阳110870

出　　处：《机械设计与制造》2024年第12期139-143,共5页Machinery Design & Manufacture

基　　金：基于结构噪声信息的大型风力机传动系统运行状态识别与故障诊断方法研究(51675350)。

摘　　要：为解决风机轴承退化指标单调性差、可解释性不足的问题,提出了一种基于t分布随机邻近嵌入(t-SNE)和聚类量化的风机轴承退化评估方法。该方法首先提取健康参考状态和任意时刻监测状态振动信号的时域、频域及时频域特征,并对其进行参照化特征融合;其次,为避免维数灾难利用t-SNE对高维数据进行降维;最后,选用聚类量化因子表征风机轴承的退化程度,设定自适应阈值,进而实现退化评估。经过与其他算法对比和实际信号验证,所建立的退化指标能够及时预警风机轴承早期故障,且单调性强、可有效减少误警率。In order to solve the problems of poor monotony and insufficient explanability of wind turbine bearing degradation indexes,a degradation assessment method for wind turbine bearing was proposed based on t-SNE and cluster quantization.Firstly,the time-domain,frequency-domain and time-frequency domain features were extracted from the vibration signals of the health reference state and the monitoring state at any time,and then the referential features were fused.Secondly,t-SNE is used to reduce the dimensionality of high-dimensional data.Finally,the clustering quantization factor was selected to characterize the degradation degree of the wind turbine bearing,and the adaptive threshold was set to realize the degradation assessment.Through the comparison of other algorithms and the verification of actual signals,the degradation index established in this paper can timely warn the early fault of the wind turbine bearing with strong monotony and can effectively reduce the false alarm rate.

关 键 词：t-SNE 聚类量化因子 自适应阈值 退化指标单调性 

分 类 号：TH16[机械工程—机械制造及自动化] TH17TH165.3