中国机械工程Issue(13)：1760-1765,6.DOI:10.3969/j.issn.1004-132X.2014.13.012
多重分形去趋势波动分析在滚动轴承损伤程度识别中的应用
Application of Multifractal Detrended Fluctuation Analysis to Severity Identification of Rolling Bearing Damages
摘要
Abstract
The multifractal spectrum of bearing vibration data was estimated using MFDFA.As a result,the shapes and positions of the multifractal spectrum could be largely determined by the left-end,right-end and extreme points of the multifractal spectrum.Subsequently,coordinates of these characteristic points were used as characteristic parameters for describing dynamic properties of the bearings.MFDFA,together with four conventional temporal statistical parameters,wavelet transform (WT)and empirical mode decomposition(EMD),was exploited to recognize severity of damage of bearing balls and outer-races separately.Each of the Mahalanobis-distance(MD),BP neural network and support vector machine (SVM)algorithms was employed to classify the feature parameters de-rived from each of WT,EMD and MFDFA.Moreover,the effectiveness of these algorithms in severi-ty identification of bearing damage was compared.The results show that the methods associating MD with MFDFA and associating SVM with WT or EMD perform better than the others.The conclusions drawn in the early work seem to be further confirmed.关键词
多重分形去趋势波动分析/滚动轴承/损伤/程度识别Key words
multifractal detrended fluctuation analysis(MFDFA)/rolling bearing/damage/sever-ity identification分类
机械制造引用本文复制引用
林近山,陈前..多重分形去趋势波动分析在滚动轴承损伤程度识别中的应用[J].中国机械工程,2014,(13):1760-1765,6.基金项目
山东省自然科学基金资助项目(ZR2012EEL07) （ZR2012EEL07）
