仿真结果统计检验方法  

STATISTICAL TEST METHOD FOR SI MULATION RESULTS

作　　者：傅惠民[1] 

机构地区：[1]北京航空航天大学固体力学研究所,北京100083

出　　处：《机械强度》2005年第5期598-603,共6页Journal of Mechanical Strength

基　　金：国防科技预研项目(413200204)资助。~~

摘　　要：提出一种仿真结果统计检验方法,包括多状态仿真结果检验方法和多因子统计分析方法。前者能对多个正态母体的仿真结果进行检验,以确定哪些状态的正态母体仿真结果正确,哪些状态的正态母体仿真结果错误;后者则能分析各个因素(因子)对仿真结果的影响,以确定哪些因素已被正确仿真,哪些因素尚未被正确模拟,从而指导仿真软件的编制。此外,文中还对指数分布截尾数据、位置—尺度分布族截尾数据和不完全数据情况下整体分布参数仿真结果的检验进行了讨论。A statistical test method for simulation results is presented, which includes the multi-state test and the multi-factor analysis method. By testing the simulation results of multiple normal populations, the former can confirm in which states they are correct and in which states they are not. The later can analyze the infection of each factor to the simulation results, and then points out which factors have been simulated correctly and which have not. Accordingly, it can direct the establishment of simulation software. Furthermore, the test method for the simulation results of the integral distribution parameters is also discussed with censored data in exponential distribution and incomplete data in location-scale distributions.

关 键 词：仿真 数字化设计 模型确认 方差分析 小样本 检验 

分 类 号：O241.7[理学—计算数学] TB115[理学—数学]