基于谱质心直方图-SVM的滚动轴承故障诊断.pdfVIP

第34 卷 第5 期 噪 声 与 振 动 控 制 Vol 34 No.5 2014 年10月 NOISE AND VIBRATION CONTROL Oct. 2014 文章编号：1006-1355 2014 05-0187-05 基于谱质心直方图—SVM 的滚动轴承故障诊断 李军杰 ，马建仓，柏会宁 ，孟宏伟 （西北工业大学 电子信息学院，西安 710129 ） 摘 要：轴承早期故障信号的信噪比较低，较难识别。为提高轴承早期故障诊断的准确率，分析了现有方法和视觉 信息处理方式的相似点，从视觉信息处理的角度进行研究，提出一种基于谱质心直方图的滚动轴承故障特征提取方 法，将所提出方法用于滚动轴承不同运行状态下的故障特征提取，将所提取特征作为支持向量机的输入特征向量从而 实现对滚动轴承不同运行状态的智能分类。实验证明，谱质心直方图提高了轴承早期故障诊断准确率，准确率平均提 高了2 % ，优于Mel 频率倒谱系数特征。 关键词：振动与波；谱质心直方图；谱峭度；支持向量机；轴承早期故障诊断；视觉信息处理 中图分类号：TN911.7 文献标识码：A DOI 编码：10.3969/j.issn. 1006-1335.2014.05.041 FaultFault DiagnosisDiagnosis ofof RollingRolling BearingsBearings BasedBased onon SpectralSpectral CentroidCentroid Histograms-SVMHistograms-SVM LI Jun -j ie , MA Jian -cang , BAI Hui-ning , MENG Hong -wei School of Electronics Information, Northwestern Polytechnical University, Xi ’an, 710129, China AbstractAbstract : Early fault features of rolling bearings are usually immersed in heavy noise background. In other words, its signal-to-noise ratio SNR is too low to diagnose the fault. In this paper, the similarity of the existing methods and the way of visual information processing is analyzed in order to improve the diagnosis accuracy of the early fault diagnosis of the rolling bearings. From the viewpoint of visual information processing, a method of fault feature extraction of rolling bearings under different operation conditions based on spectral centroid histogram is proposed. Then, the extracted features are used as the input feature vectors of SVM to realize the intelligence diagnosis in different operation status. Experiments show that using the spectral centroid histogram can raise the diagnosis accuracy of the early fault diagnosis of the rolling bearings, and the accuracy is increased by 2 %, better than that of the MFCC features. KeyKey wordswords : vibration and wave ; spectral centroid histograms ; spectral kurtosis ; SVM ; early fault diagnosis ; visual information processing 滚动轴承广泛应用于旋转机械设备，轴承过大 因此，本文只对这3