基于连续高斯密度混合HMM的刀具磨损状态监测  

Tool Wear Condition Monitoring Based on Continuous Gaussian Mixture HMM

作　　者：王玫[1] 吕俊杰[1] 王杰[1] 

机构地区：[1]四川大学制造科学与工程学院,四川成都610065

出　　处：《四川大学学报（工程科学版）》2010年第3期240-245,共6页Journal of Sichuan University (Engineering Science Edition)

摘　　要：针对端面铣刀磨损状态的识别问题,根据隐式马尔可夫模型(HMM)具有良好的模式分类能力,提出了基于连续高斯密度混合HMM(CHMM)的刀具磨损状态监测系统。以铣削力作为监测信号,应用小波包理论对铣削力信号进行分析和消噪处理,并提取信号的能量特征作为CHMM的输入向量,训练CHMM模型,再用训练好的模型对未知的刀具磨损状态进行监测与识别,实验结果表明该模型可以对刀具磨损状态进行准确的识别,且所需训练样本数较少,对刀具状态的智能监测具有很好的实际意义。Aiming at the problem of the identification of cutting tool wear condition,and according to the merit of Hidden Markov Model （HMM） with perfect classifying ability,the system of the tool wear condition monitoring based on continuous Gaussian mixture HMM （CHMM） was presented.With the milling force as the monitoring signal,the wavelet package theory was adopted to analyze and de-noise the milling force signal,and extracts energy feature of the signal as import vectors of CHMM,trains CHMM.Then the unknown state of tool wear was monitored and identified using the trained model.The results of the experiment indicated that the model can exactly recognize the tool wear state and need small training samples,has significant realistic meaning to tool intelligent monitoring.

关 键 词：隐式马尔可夫模型 小波包分析 铣削力 刀具磨损 

分 类 号：TH164[机械工程—机械制造及自动化]