摘要
提出了针对参数不确定性的综合贝叶斯的有限元模型修正方法。即在贝叶斯中融入用以代替原模型的随机模型和基于吉布斯抽样的蒙特卡罗马尔科夫链抽样算法,选定较敏感参数进行仿真获取模型参数的后验分布动态统计特征,进而评估参数不确定度。实例验证了文中方法的有效性,表明随机响应面模型可极大提高吉布斯抽样计算效率,尤其适用于含多维不确定性参数的复杂结构有限元模型不确定性修正问题。
The paper proposed a new synthesized bayesianian approach for structure finite element model updating by using parameters uncertainty estimation. It integrated stochastic models instead of the original model and markov chain Monte Carlo（MCMC） algorithm with gibbs sampling. By sensitivity analysis, sensitive parameters were used to simulate and obtain parameters dynamic posterior statistics characteristics, and then the parameters uncertainty was eatimated. Numerical simulation indicates that the approach is effective, It evidently improves the computational efficiency with SRSM, the method is suitable for finite element model updating of complex structure containing multi -parameters uncertainty.
出处
《机械科学与技术》
CSCD
北大核心
2014年第10期1545-1550,共6页
Mechanical Science and Technology for Aerospace Engineering
关键词
有限元方法
参数估计
随机模型
蒙特卡罗
马尔科夫过程
algorithms
computational efficiency
computer simulation
covariance matrix
efficiency
finite element method
flowcharting
Markov processes
mathematical models
Monte Carlo methods
parameter estimation
parameterization
sensitivity analysis
spot welding
statistics
stochastic models
