基于退化数据的最优预测维护决策支持方法  

A degradation measurement based decision support method for optimal predictive maintenance policy

作　　者：樊红东[1,2] 胡昌华[1] 陈茂银[2] 周东华[2] 

机构地区：[1]第二炮兵工程学院302教研室,陕西西安710025 [2]清华大学自动化系,北京100084

出　　处：《华中科技大学学报（自然科学版）》2009年第S1期45-48,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60736026)

摘　　要：提出了一种基于性能退化数据的最优预测维护决策支持方法.对历史性能退化数据进行分析,确定一种合适的非线性性能退化模型,并估计出其参数的先验分布,再根据此模型获得部件的剩余寿命分布,当新的数据来到时,根据贝叶斯理论对参数先验分布进行更新,从而达到对剩余寿命分布进行实时更新的目的.在此基础上,引入最优预测维护策略,给出了对应不同当前时刻的最优预测维护时间.用疲劳断裂数据进行了验证,结果表明了此方法的有效性.A decision support method for condition-based optimal predictive maintenance is proposed in this paper.First,a proper nonlinear degradation path model,as well as the prior distribution of its random parameters,is determined based on the analysis of historical degradation data.Then,the residual life distribution can be derived from the selected degradation path model.When a new degradation data is obtained,the prior distribution can be updated according to Bayesian formula.Thus,the residual life distribution can also be updated given the posterior distribution.Furthermore,a predictive maintenance policy is introduced which provides the optimal predictive maintenance time.At last,fatigue crack growth data are used to verify the effectiveness of the developed approach.

关 键 词：性能退化 剩余寿命分布 蒙特卡罗法 预测维护 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]