基于改进HSMM的设备故障预测方法研究  

Equipment fault prognosis based on improved HSMM

作　　者：夏震宇[1] 杨波[1] 

机构地区：[1]海军工程大学管理工程系,武汉430033

出　　处：《现代制造工程》2011年第8期118-122,共5页Modern Manufacturing Engineering

基　　金：国防十一五预研项目(1010503020303)

摘　　要：针对传统隐半马尔科夫模型(HSMM)在故障诊断和预测应用中存在的不足,对传统HSMM做了以下改进:一是将状态持续时间概率分布和监测值概率分布连续化,并假定其服从威布尔分布;二是基于状态开始时间的识别,提出了状态剩余持续时间;三是提出了时变转移概率的概念,给出了各时刻转移概率的计算方法。确立了基于改进HSMM的故障诊断和预测的方法体系,给出了故障诊断判据和设备剩余寿命的计算式。案例研究表明方法是合理有效的。To the four deficiencies of traditional HSMM applied in fault diagnosis and prognosis,the HSMM was improved as following.Firstly,the distribution of the state durations and detected value were continuous and parametric,the prior distribution was setted to Weibull distribution;secondly,the state residual durations was proposed based on the starting time of the state;thirdly,the time-varying transition probability was proposed and the formula of transition probability was given.The method of fault diagnosis and prognosis based on improved HSMM was setup,and the formulas of fault diagnosis criteria and Remaining Useful Life（RUL） of the equipment were presented.The rationality and the validity of this method was verified through an example.

关 键 词：故障预测 故障诊断 隐半马尔科夫模型 状态持续时间 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]