基于滑动峭度相关性准则的局部特征尺度分解分量筛选方法

A Method for Determining Effective Components of Local Characteristic—scale Decomposition Based on Sliding Kurtosis Correlation Coefficients

轴承故障振动信号具有非平稳、非线性特征,且可视为多个调幅一调频分量的叠加,单分量的包络蕴含了轴承的故障特征。局部特征尺度分解可将振动信号准确分解为多个内禀尺度分量之和,某些分量能清晰反映轴承的运行状态,根据包络谱可进行故障诊断;为了准确筛选有用分量,提出了基于滑动峭度相关性准则的分量筛选方法;首先,对信号进行局部特征尺度分解,得到若干个内察尺度分量;然后,对分量和原始信号分别计算滑动峭度,生成时间序列;最后,依据分量滑动峭度序列与原始信号滑动峭度序列的互相关系数筛选有用分量;通过轴承内圈故障数据分析发现:有用分量与非有用分量之间的滑动峭度互相关系数比互相关系数差异明显,区分度更大,有益于分量的分类、筛选。

The vibration signal of fault roller bearing is characterized by non--stationary and non--linear, and can be regarded as the su- perposition of some amplitude-- modulation components and frequency-- modulation components. The amplitude-- modulation components contains the fault characteristics of roller bearing. The vibration signal can be decomposed into several intrinsic mode components （ISC） by local characteristic--scale decomposition （LCD） accurately. Some components can clearly reflect the bearing running condition, the envelope spectral can be used to fault diagnosis. In order to process the roller bearing fault vibration signals, aiming at determining the effective com- ponents of LCD, a new method based on sliding kurtosis correlation coefficients was proposed. Firstly, the vibration signal was decomposed into some ilSC by LCD. Secondly, the sliding kurtosis of every ISC and the original signal were calculated. At last, the effective components were selected by the sliding kurtosis correlation coefficients. The analysis of the bearing fault data shows that the sliding kurtosis correlation coefficients can realize the determining effective components of LCD effectively, and has better classification ability than the correlation coefficients.