采用符号时间序列分析的轴承早期微弱故障预警
A Warning Method for Early Weak Failures of Bearings by Symbolic Time Series Analysis

针对轴承早期磨损故障预警难的问题,提出了一种采用符号时间序列分析的轴承早期微弱故障预警方法。首先对轴承振动信号进行相空间重构,其次利用聚类的思想对相空间进行分割,对分割的子区间赋予唯一的符号,从而将振动信号转化为符号时间序列,最后通过Lempel？？Ziv复杂度对符号时间序列进行定量分析。以Duffing方程为研究对象,验证了该方法对于动力学结构表征的精确性,表征精度可达95？？85％;与常规的符号时间序列分析方法相比,所提方法对动力学结构表征精度更高。利用该算法对于动力学结构变化的敏感性特点,将其应用于轴承状态监控中,试验结果显示算法可以发现轴承早期微弱磨损故障并实现全寿命性能衰退的监控。
A warning method for early weak failures of bearings is proposed based on symbolic time series analysis to address the problem in early warning of weak faults in bearings. The phase space of bearing vibration signals is reconstructed using C？？C method and then segmented by k？？means clustering. Then each segmented subinterval is given an unique symbol and the bearing vibration signals are transformed into a symbolic time series. Finally, a quantitative analysis of the symbolic time series is carried out through the Lempel？？Ziv complexity. The algorithm’s accuracy of the representation of the dynamic structure is verified using the Duffing equation as the study object, and its precision is up to 95？？85％. A comparison with the conventional symbolic time series analysis methods shows that the proposed method has a better accuracy. Since the proposed algorithm is sensitive to the dynamical changes, it is applied in the bearing state monitoring. Experimental results show that the algorithm can discover the early feeble wear？？out failures of bearings and realizes the monitoring of whole life performance degradation