基于谐波小波包和SVM的滚动轴承故障诊断方法

针对滚动轴承故障诊断问题开展研究,设计了基于谐波小波包和支持向量机（SVM）的新型诊断方法.与传统的时频特征提取方法相比,谐波小波包具有盒状频谱和无限细分的优势.首先对滚动轴承的振动数据进行谐波小波包分解,利用各频段的小波分解系数计算特征能量,归一化之后作为特征向量,为设计的多类SVM模型提供训练样本和测试样本.利用SVM的非线性映射能力,将三个二分类器相组合设计了基于二叉树的多类SVM模型,实现了对滚动轴承的故障诊断.最后,利用Case Western Reserve University电气工程实验室的滚动轴承试验台的振动数据对设计的诊断方法进行了验证.结果表明,设计的诊断方法比传统的方法具有更高的准确率.

Fault diagnosis of roller bearings with harmonic wavelet package and SVM

This paper focuses on roller bearing fault diagnosis and proposes a novel method with harmonic wavelet package and Support Vector Machine （SVM）. Compared with the conventional time-frequency fea- ture exaction method, harmonic wavelet package possesses better performances with box-shaped spectrum and unlimited subdivision. Firstly, the roller bearing vibration signals are decomposed with harmonic wavelet package and the feature energy is calculated with coefficients of each spectrum. After the feature energy has been normalized, the feature vector is available for the training samples or testing samples of the diagnosis model. Based on the multi-level binary tree, this paper establishes a multi-classification SVM model due to its superior nonlinear mapping capability. Three 2-classifications are incorporated to diagnose the roller bear- ing faults. Finally, the proposed method is tested with the vibration data from the roller bearing stand of the electric engineering lab in Case Western Reserve University. The experimental results show the proposed method possesses higher accuracy compared with conventional methods.