大数据统计趋势分析和PCA的滚动轴承早期故障诊断

Big Data-driven Statistic Trend Analysis and PCA for Incipient Fault Diagnosis of Rolling Bearings

针对短时、小样本数据下提取的特征对早期故障敏感度和故障演化过程稳定度低、信息冗余的问题,提出了大数据统计趋势分析和核主元分析方法(Principal component and analysis,PCA)的滚动轴承故障演化特征提取和早期故障诊断方法。采集滚动轴承正常状态到完全失效状态的全寿命振动数据,计算原始数据中不同故障严重程度下的时频统计特征,建立各个统计特征描述的故障演化趋势,分析各个统计特征描述的故障演化特性,初步选择能够敏感且稳定感知故障演化过程的统计特征集,利用PCA分析初选结果中各个统计特征间的相关性和贡献度,进一步剔除冗余特征,最终得到能全面表征故障演化过程的特征。最后,使用滚动轴承全寿命振动数据验证本文所提方法的有效性。实验结果证明,标准差、均值频率、标准差频率等特征能敏感地检测滚动轴承早期内环故障并稳定跟踪其演化过程。

Aiming at those problems of weak sensitivity for incipient fault,low stability for fault evolution process and information redundancy with short time and small sample,a novel feature extraction and incipient fault diagnosis method for rolling bearings is proposed based on big data-driven statistic trend analysis for fault evolution process and principal component and analysis(PCA)in this paper.A life-time data of rolling bearings in whole lifecycle are collected by various sensors,then statistic characteristics of original vibration data at each echo are calculated,thus those fault evolution trend curves described by all statistic characteristics would be built.Based on fault evolution trend curves,the stability and sensitivity of each feature for fault evolution trends can be calculated,thus the primary features can be selected.The relevancy and contribution of each feature in the primary features by PCA was obtained,thus redundancy feature would be further removed,and those features which can fully reflect fault evolution process will be furthermore choose.Finally,the life-time data of rolling bearings were used to verify the effectiveness of proposed methods.The results have shown that the standard deviation,mean frequency,standard deviation frequency can more sensitively detect incipient inner race of rolling bearings and comprehensive track their evolution process.