摘要
工程系统中不可避免地存在各种参数不确定性,利用数值计算模型对系统进行虚拟试验时应进行不确定性分析.大型耗时计算模型的不确定性分析将面临严重的的计算复杂性问题,为此,针对工程应用中耗时计算模型,提出一种基于贝叶斯预测模型的不确定性分析仿真方法,采用概率分布为参数不确定性建模,研究系统响应预测不确定性的概率特征.泰勒杆撞击实例验证了该方法的高效性.
Uncertainty analysis was necessary in virtual experiment ot engineering system since uncertainty of parameters in inevitable. Complexity problem was confronting uncertainty analysis of big-scale time-consuming calculation model. Here, a simulation approach based on Bayesian predictive model was proposed to predict the distribution function of prediction uncertainty, in which the parameter uncertainties were described in terms of probability density function and the expensive numerical model was approximated by its posterior distribution. A Taylor impact test demonstrates that the proposed method is efficient.
出处
《工程设计学报》
CSCD
北大核心
2008年第1期21-24,28,共5页
Chinese Journal of Engineering Design
基金
国防科工委基础科研项目资助(B0920060321)
关键词
不确定性分析
虚拟试验
预测模型
贝叶斯
uncertainty analysis
virtual experiment
predictive model
Bayesian
