摘要
对于旋转机械启动过程的动态模式,提出了一种基于离散隐马尔可夫模型(DHMM)的旋转机械故障诊断新方法.该方法对旋转机械启动过程的局部振动信号进行FFT特征提取,然后利用自组织特征映射对提取的特征矢量进行预分类编码,把矢量编码作为观测序列引入到DHMM中进行机器学习和故障诊断实验.实验表明,提出的方法对旋转机械启动过程进行诊断是十分有效的.
A new method based on Discrete Hidden Markov Models （DHMM） is proposed for dynamic patterns recognition in running - up process of rotary machine. At first FFT features are extracted from local vibration signal of rotary machine in running - up stage, then coded FFT vectors into codes book of integer numbers by SOM, and introduced these codes book into DHMM for machine learning and classification. Finally, conclusions of fault diagnosis experiments are presented. Experiments which verified proposed method was effective.
出处
《大连民族学院学报》
CAS
2006年第3期12-15,27,共5页
Journal of Dalian Nationalities University
基金
国家自然科学基金研究资助项目(50405023)
关键词
离散隐马尔可夫模型
故障诊断
动态模式识别
旋转机械
矢量量化
discrete hidden Markov Model
fault diagnosis
dynamic pattern recognition
rotary machine
vector quantification
