基于马氏链蒙特卡洛方法的数控系统可靠性评估  

The Reliability Evaluation of NC System Based on Markov Chain Monte Carlo Method

作　　者：李斌全[1] 戴怡[1] 

机构地区：[1]天津职业技术师范大学机械工程学院,天津300222

出　　处：《组合机床与自动化加工技术》2011年第10期69-71,共3页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金(50875186);国家数控机床科技重大专项(2009ZX04014-013)

摘　　要：数控系统属于高可靠性产品,应用贝叶斯理论,已经成为其可靠性评估的重要方法。由于数控系统寿命试验数据服从W eibull分布,导致后验分布出现数值大、高维复杂的情况,贝叶斯计算依靠的数值积分方法难以实施。根据马氏链蒙特卡洛(MCMC)方法思想,建立后验平稳分布的马尔科夫链,对分布参数的贝叶斯估计进行求解,解决了数值积分问题,保证可靠性评估的有效实施。通过与BUGS软件结果进行比较,表明提高了模型计算的稳健性、有效性及精度。NC system was high-reliability product,so the application of Bayesian theory had become an important means of its reliability evaluation.The failure data of NC system obeyed Weibull distribution which had complex forms.So the posterior distribution became more complicated and the numerical integration which Bayesian computing depended on was not available.The problem of numerical integration was solved by Markov chain Monte Carlo（MCMC） method.It ensures the effective implementation of reliability evaluation.Compared with BUGS package,the result of Bayesian estimation shows that it increases the robustness and effectiveness of the calculation.

关 键 词：数控系统 马氏链蒙特卡洛(MCMC)方法 可靠性评估 WEIBULL分布 贝叶斯理论 

分 类 号：TH17[机械工程—机械制造及自动化] TG65[金属学及工艺—金属切削加工及机床]