Uncertainty quantification of the finite element model of existing bridges for dynamic analysis
Uncertainty quantification of the finite element model of existing bridges for dynamic analysis
复制标题
DOI：
10.1007/s13349-012-0026-z
复制
发表时间：
2012-11
影响因子：
4.4
通讯作者：
M. Nishio;Juliette Marin;Y. Fujino
中科院分区：
文献类型：
--
作者：
M. Nishio;Juliette Marin;Y. Fujino
随着包括桥梁在内的老化土木结构越来越多，有限元分析在维护和维修设计以及结构健康监测中发挥着关键作用。尽管老化结构的动力特性由于老化或可能的损伤而与标称状态发生了变化，但用于动力分析的数值模型的验证和验证（V&V）通常受到的关注有限。本文通过将一系列V&V程序应用到桥梁有限元模型中，描述了系统V&V建模的一些实际考虑，特别是在既有桥梁和老化桥梁的建模中。应用V&V程序包括：(1)通过冲击试验获取实验数据；(2)使用网格收敛指数验证模型；(3)进行敏感性分析，提取对比较特征有影响的参数，即模态参数；(4)基于贝叶斯推理的不确定性量化。在不确定性量化过程中，成功地导出了有限元模型中不确定参数的后验概率分布，特别是梁和板的材料特性。本文还从复杂有限元模型的验证程序、比较特征的选择和适当的实验数据采集等方面讨论了改进现有结构建模中V&V实现的一些重要考虑因素。
With increasing number of aging civil structures including bridges, the finite element (FE) analysis assumes key roles in the use of the maintenance and repair design, and structural health monitoring. Despite the changes in dynamic characteristics of the aging structures from the nominal condition due to deteriorations or possible damages, limited attention has generally been paid to the verification and validation (V&V) of the numerical models used in the dynamic analysis. In this paper, we describe some practical considerations on the systematic V&V especially for the modeling of existing and aging bridges by applying a series of V&V procedure to the bridge FE model. The applied V&V procedure includes: (1) experimental data acquisition by impact tests, (2) model verification using grid convergence index, (3) sensitivity analysis to extract influential parameters to the comparative feature, namely the modal parameters, and (4) uncertainty quantification based on Bayesian inference. Posterior probability distributions of uncertain parameters in the FE model, especially material properties of the girders and slab, were then successfully derived in the uncertainty quantification procedure. The study also discusses some important considerations to improve implementation of V&V for the modeling of existing structures in terms of the verification procedure for complex FE models, the choice of comparative features, and the appropriate experimental data acquisition.