基于贝叶斯法的定时截尾小样本指数型装备器材需求预测  

Demand Forecasting for Equipment Materials with Exponential Life Distribution Based on Bayesian Estimation under Type Ⅰ Censored Small Sample

作　　者：王铁宁[1] 吴龙涛[1] 杨帆[1] 

机构地区：[1]装甲兵工程学院技术保障工程系,北京100072

出　　处：《装甲兵工程学院学报》2017年第4期29-34,共6页Journal of Academy of Armored Force Engineering

基　　金：军队科研计划项目

摘　　要：针对高新装备器材故障数据少、需求规律不明确的问题,提出了一种定时截尾小样本条件下指数型装备器材的需求预测方法。基于贝叶斯法进行了装备器材寿命分布参数估计,讨论了先验分布和损失函数的选择问题。引入了K-S检验法对寿命分布模型进行拟合优度检验,并设计了寿命分布参数估计的卡方检验方法。综合考虑故障更换和定时更换,提出了部队装备器材年度需求预测方法,并采用蒙特卡洛法进行了仿真验证。结果表明:基于贝叶斯法的参数估计结果能够顺利通过检验,需求预测结果与仿真结果一致。In order to address the problem that the demand discipline of new and high-tech equipment materials cannot be mastered well because of lack of failure data, a method of demand forecasting is proposed for equipment materials with exponential distribution under small failure samples. In a view of small samples, the parameter of equipment material life distribution is estimated by Bayesian estimation, the option of prior distribution and loss function is discussed. Then, statistical tests for goodness-of-fit of the life distribution and parameter estimation are performed by K-S test and Chi-Square test respectively. In consideration of fault replacement and timely replacement, the annual demand forecasting method of army equipment is put forward and devised with a Monte-Carlo simulation test. As the example shows, the estimation result performs well in the test, and the forecasting value is identical to that from the simulation.

关 键 词：定时截尾小样本 装备器材 指数分布 需求预测 贝叶斯法 

分 类 号：E92[军事—军事装备学] N945.24[兵器科学与技术—武器系统与运用工程]