文档介绍：第18卷第8期 2o 2年8月计算机集成制造系统 Cbmputer Integrated 1Ⅵanufacturing Systems 012 文章编号:1006—5911(2012)08~1861—08 基于DDHSMM的设备运行状态识别与故障预测方法王宁1’2,孙树栋1’2,李淑敏1’2,蔡志强1’2 (,陕西西安710072 ,陕西西安710072) 摘要:针对设备运行状态识别与故障预测问题,提出一种基于时变转移概率的隐半Markov模型。该模型将设备历史运行信息融入Markov状态转移概率矩阵的估计过程中,使Markov状态转移概率矩阵具有时变特性。基于改进前向后向算法研究了相应的隐半Markov摸型参数估计方法,使其能够不断综合利用历史运行信息进行自我更新,以更加符合设备真实运行的过程。同时以该模型为基础,利用故障率方法建立了对设备剩余使用寿命进行预测的基本步骤。通过某滚动轴承运行状态识别实例演示了该模型的建模过程,证明了基于该模型的设备状态识别与预测方法比传统隐半Markov模型方法更为有效。关键词:时变转移概率;隐半Markov模型;故障率;状态识别;剩余有效寿命中图分类号:TP391 文献标志码:A Equipm哪tstate他∞gniti伽and fallItpr0孕I髑ti岱I眦thod based蚰D胁HSMM model wANG Ni,191’-,SUN吼扩如91伯,LJ S^“删i胛1仙,CAI (1-School ofMechamcal Engineering,Northwestem Polytechnical Unjversity,xi’an 710072,China; L丑boratoryofC0nt咖porary DleSignandIntegratedManufacturing Techn0109y,Ministry ofEducation, Nonhwestem Polytechmcal University,xi’an 710072,China) Abs嗽Ict:A赫119 thepmbl帆of equipment operation state identificationandfaultprognosis,a I)uratiomDepend— ent HiddenSerni—MarkovM0del(DDHS^dM)was proposed Inthism。del,the historicaloperation in如nnation was nlerged intoesti砌tion process of】Ⅵarkov statetmsition probability matrix,thusthe眦trix had variant char— ,the par锄eter estirrlationmethod ofHidden sefnj—MarkovM0del(HSMM)was studied based on irIlprovedforward_backward algo^th make self-renewal by using historicaloperation infomation The basic stepsforpredicting the UsefulLife(RUL)of equi肿nent、Ⅳas builtby using faultmte methocL Through a case of arolling bearing’s operation state todernonstratethemodeling process ofproposed model,and the resultshowed thattheproposed method was more effective than traditionalHSMM modeL Keywor凼:duration-dependent state tranSition pmbabilities;hidden serIli—MarkovⅡ10del;hazard rate;state recogni— tion;remailling useful 0 引言随着人们对设备安全性问题的日益关注,现有的设备维修理论方法已经不能满足可靠性的要求。预测性维修(Predictive Maintenance,PdM)作为一种新兴的维修方式已经成为维修界研究的热点。 PdM通过对设备状态进行监测和诊断,做出正确判断,从而制定出科学合理的维修策略,克服了传统