基于小波包的滚动轴承故障特征提取

Bearing Fault Feature Extraction Based on Wavelet Packet Transform

在深入分析离散小波包变换快速算法的基础上 ,给出了离散小波包变换快速算法中产生频率混淆的原因 ,即由正交镜像滤波器的非理想截止特性、隔点采样和隔点插零的特性共同作用产生的 ,提出了一种消除频率混淆的算法。利用该算法和原算法 ,分别对某型滚动轴承内环剥落故障的振动信号进行处理 ,提取其故障特征。结果表明 ,原算法由于存在频率混淆 ,可能掩盖故障特征 ,提出的新算法 ,由于很好地消除了频率混淆 ,能有效地提取滚动轴承局部故障的特征。

On the basis of analysis of the fast algorithm of discrete wavelet packet transform (the canonical algorithm), the causes of frequency aliasing that is inherent in the fast algorithm are given. The problem of frequency aliasing is caused by the frequency characteristics of quadrature mirror filters, the process of keeping one sample out of two and the process of putting one zero between each sample. An algorithm to avoid frequency aliasing for wavelet packet transform is presented. In order to compare with the canonical algorithm, both algorithms are applied to extract the fault feature of a ball bearing inner race defect. It is shown that the presented algorithm is effective to avoid frequency aliasing, and it is better than the canonical algorithm to extract the weak shock feature from a vibration signal related to the localized defects of a ball bearing