少样本故障数据数控机床的贝叶斯可靠性分析

Bayesian reliability analysis for numerical control machine tools with small-sized sample failure data

应用贝叶斯方法分析少样本故障数据数控机床的可靠性,给出2参数威布尔分布模型参数及数控机床可靠性指标的点估计和区间估计,通过马尔科夫链蒙特卡洛抽样解决了贝叶斯可靠性分析中求解复杂后验积分的难题。结合一具体实例,分析10台加工中心时间截尾的可靠性。计算结果表明:在充分利用先验信息的基础上,贝叶斯方法优于极大似然法和似然比检验法,适合于少样本数据的可靠性分析。

The reliability of numerical control（NC） machine tools with small-sized sample field data was analyzed using Bayesian method. Point and interval estimations of two-parameter Weibull distribution model and reliability indices of NC machine tools were presented. The problem of complex posterior integral in Bayesian reliability analysis was solved by Markov chain Monte Carlo（MCMC） sampling. Using a real field example, the reliability of 10 machining centers with time truncation was analyzed. The results show that based on the full use of priori information, Bayesian method is better than both maximum likelihood estimation method and likelihood ratio testing method, and suitable for reliability analyses of small-sized sample data.