基于CHMM的齿轮箱状态识别研究  

Gearbox state recognition based on continuous hidden markov model

作　　者：滕红智[1,2] 赵建民[1] 贾希胜[1] 张星辉[1] 王正军[1] 

机构地区：[1]军械工程学院 [2]68129部队

出　　处：《振动与冲击》2012年第5期92-96,127,共6页Journal of Vibration and Shock

摘　　要：针对离散隐Markov模型(HMM)在状态识别中的不足,结合齿轮箱全寿命实验数据,研究了基于连续隐Markov模型(CHMM)的状态识别方法。建立了基于齿轮箱原始振动信号的CHMM状态识别框架,提出了基于K均值算法和交叉验证相结合的状态数优化方法,通过计算待确定观测数据的极大似然概率值来确定齿轮箱当前状态。结果表明,用原始振动信号作为CHMM的输入可以实现状态识别,验证了模型的有效性,为齿轮箱基于状态的维修提供了科学依据。Combined with full lifetime test data of gearbox, state recognition based on continuous hidden markov model （CHMM） was studied. The frame of state recognition based on CHMM using original vibration signal was established. Virtues and defects of existing classification methods classifying state in full life cycles were analyzed. State number optimization model was established based on K means and cross validation. Gearbox ＇s operating state was determined by calculating the maximum log-likelihood. The recognition results showed that the proposed method of state recognition based on CHMM using original vibration signals is feasible and effective.

关 键 词：连续隐马尔科夫模型 K均值 交叉验证 

分 类 号：TH17[机械工程—机械制造及自动化]