基于一维结构图熵的滚动轴承早期故障检测

打开文本图片集

关键词：滚动轴承；早期故障检测；图模型；一维结构图熵

中图分类号：TH133.33

DOI：10.3969/j.issn.1004-132X.2026.03.014 开放科学（资源服务）标识码（OSID）：

Abstract： To address the challenges of accurately identifying early faults in rolling bearings，a fault detection method was proposed based on one-dimensional structural graph entropy. A graph model was developed to transform time-series data into spatial structures，enabling efective extraction of bearing condition features. A complete graph model of signal short-time power spectrum was construtured，and the complexity changing rules of time-frequency energy distribution were captured. Leveraging the ability of entropy to describe signal nonlinearity，a one-dimensional structural graph entropy measure was defined to quantify the variations in complexity of model structure，whose mean value served as health indicator for assessing the condition of the bearings. Theoretical explanations and numerical analyses demonstrated the discriminative mechanism of health indicators regarding operating states. Additionally，an adaptive detection method was developed based on the characteristics of this health indicator. The method was experimentally validated on XJTU-SY， IMS， PHM，and pulp mill datasets. Results show that the method may accurately identify fault conditions without any parametric adjustments. When compared with methods such as mean square value，synchronized pseudo-velocity corrected mean square value，variance，and kurtosis，the pro posed health indicator shows superior robustness and trend -tracking performance.

Key words：rolling bearing; early fault detection；graph model; one-dimensional structure graph entropy

0 引言

滚动轴承作为风电机组、汽车变速箱、数控机床、航空发动机等诸多机械设备的核心零件之一，其健康状况会对设备的运行性能和使用寿命产生巨大影响。（剩余16850字）