基于高阶累积量的特征提取方法研究
马建峰,王信义
(北京理工大学 机械工程与自动化学院,北京 100081)
摘 要:提出一种基于高阶累积量和双谱估计的声发射信号特征生成方法,为铣刀的磨损提供新的征兆。实验表明,高阶累积量对铣刀的磨损非常敏感,可以有效地用于铣刀的磨损监测中。双谱估计可以用于铣刀磨损的诊断。
关键词:铣刀磨损;高阶累积量;实时在线监控;声发射监测
中图分类号:TB52 文献标识码:A 文章编号:1006-0316(2002)01-0030-02
Research on a new feature extraction algorithm based on high-order
cumulants and bispectrum on milling
MA Jian-feng, WANG Xin-yi
(School of Mechanical Engineering and Automation, Beijing Institute of Technology , Beijing 100081.China)
Abstract: This paper develops a new feature extraction algorthim based on high order cumulants. Experiments show that the 3rd. order cumulant and 4th - order cumulant are very sensitive to the flank wear. These two cumulants of the AE signals can be employedin pattern reognition for tool wear monitoring effectively . Bispectrum analysis can also be employed to monitor the tool state.
Key words: tool wear on milling; high-order cumulant; real- time and online monitoring;acoustic emission; feature ext raction
在铣刀的磨损监测中,选择合适的特征是至关重要的。如果所选的特征对刀具的状态不敏感,则很难将该特征应用于刀具的状态识别系统。特征生成主要采用的方法有:信号的时域及基于时序建模的特征生成、基于信号的常规谱分析的特征生成、基于小波变换(Wavelet Transform)的特征生成、分形与混沌特征生成[1]。作者探讨了一种新方法—基于高阶累积量的特征生成法。
基于高阶累积量的特征提取方法研究全文