基于隐马尔科夫模型的机床部件故障预警技术  

Components of Machine Tools's Failure Warning Based of Hidden Markov Model

作　　者：何邵灿[1] 高宏力[1] 许明恒[1] 

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《机械设计与制造》2012年第8期159-161,共3页Machinery Design & Manufacture

基　　金：国家重大科技专项资助项目(20092x04014-102);西南交通大学校基金资助项目(2008B13)

摘　　要：针对目前高档数控机床的故障只能在发生后被动维修,不能在故障发生前维修故障。提出了基于经验模态分解方法(EMD)和隐马尔科夫模型(HMM)的数控机床关键部件在线故障预警模型。通过采集分析部件运行时的信号,通过EMD进行信号分析选取合适的分量然后提取特征向量,当数据库收集500组数据时对预警模型进行一次迭代训练,如此循环。最终形成一个能比较好的表达该部件完整特性的故障预警模型。模型训练完毕后,部件的实时信号通过特征向量形式输入故障预警模型中进行计算,通过计算概率来判断部件当前所处的状态,在故障发生前进行预警。At present,the Computer Numerical Control（CNC）machine can only be passively repaired after the failure happens,and the failure cannot be warned before it happens.Because of this,propose the Critical Components Failure Warning Model for CNC machine based on Empirical Mode Decomposition（EMD）and Hidden Markov Model（HMM）in it.Specifically,this model collects and analyzes the signals when the component is running,extracting feature vectors through EMD signal analysis.When the datas collected in the database reach up to 500,we begin to train the warning model,and do this process iteratively.After training this model,the real-time signals of the component are inputted into the failure warning model through feature vectors and can be calculated.Then,by calculating the probabilities,the component＇s status can be judged,and thus the failure can be warned before it happens.

关 键 词：数控机床 关键部件 故障预警 EMD HMM 

分 类 号：TH16[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]