基于隐Markov模型的重型数控机床健康状态评估  

Hidden markov model based on the heavy-duty CNC health state estimate

作　　者：邓超[1] 孙耀宗[1] 李嵘 王远航[1] 熊尧[1] 

机构地区：[1]华中科技大学机械学院制造装备数字化国家工程中心,湖北武汉430074 [2]武汉重型机床集团有限公司,湖北武汉430070

出　　处：《计算机集成制造系统》2013年第3期552-558,共7页Computer Integrated Manufacturing Systems

基　　金：国家科技支撑计划资助项目(2012BAF08B00);国家973计划资助项目(2011CB706803)~~

摘　　要：为了辅助重型数控机床的综合健康状态评估,从性能劣化角度出发,建立基于多性能参数多观测序列的隐Markov健康状态评估模型,改进了以往基于单性能参数的隐Markov模型不能准确描述机床健康状态的问题。针对隐Markov模型的参数初始化难题,通过K-means方法进行参数聚类分析,使初始化参数趋向于全局最优解;由于单性能参数不能完全描述机床状态的隐含信息,提出一种基于多性能参数多观测序列值的隐Markov模型训练算法。通过某重型数控机床滚珠丝杠的健康状态评估实例,获取了滚珠丝杠的健康状态变化趋势,验证了方法的可行性和有效性。Heavy-duty CNC has the characteristics of diverse fault modes and causes, insufficient fault samples, which makes Health state assessment very difficult. Based on multi-capability parameter and multiple observation sequences, a HMM model was constructed which could reflect the performance degradation, and expressed the health state of Heavy-duty CNC clearly. Firstly, in order to solve the problem of parameter initialization, the effects of parameter on accuracy of model were resolved by K-means algorithm. Secondly, since single performance parame- ter was not sufficient for describing the health state of Heavy-duty CNC, the method further discussed the applica- tion of multiple observation sequences in modeling. Finally, the proposed health estimation model was validated by ball-screw of the Heavy-duty CNC and the result demonstrated its effectiveness.

关 键 词：重型数控机床 隐MARKOV模型 健康评估 状态劣化 

分 类 号：TH17[机械工程—机械制造及自动化] TP206[自动化与计算机技术—检测技术与自动化装置]