点击次数：

论文类型：会议论文

发表时间：2017-01-01

收录刊物：EI、CPCI-S、Scopus

页面范围：422-426

关键字：maximum correlation kurtosis de-convolution; wavelet packet binary tree; correlation kurtosis; rolling bearing; weak fault diagnosis

摘要：The traditional method of kurtosis is calculating the kurtosis value in the time domain. The larger the obtained value is, the stronger the corresponding impact characteristic. Afterward, in order to characterize the period of the signal, the correlation kurtosis is proposed by combining the correlation coefficient and kurtosis. In this paper, the correlation kurtosis is calculated in the frequency domain to select the optimal analysis frequency band. But it has a poor performance in the case of low signal-to-noise ratio. Therefore, the maximum correlation kurtosis de-convolution method is applied as a preprocessing method to enhance impact characteristic. The whole band is divided into multiple sub-bands based on the wavelet packet binary tree and the optimal band is corresponding to the sub-band for which correlation kurtosis value is max.