Application of rational methods for the selection of the most plausible FE models in structural analysis
conference paper
The accuracy of predictions of structural behavior by finite element analysis strongly depends on the formulation of the numerical model. Depending on the problem at hand, the analyst has to make a number of choices, which influence the numerical results to a greater or lesser extent. For complex cases, e.g. existing reinforced concrete structures with (hidden) damage due to load history or material deterioration, several alternative models can be defined and it is often not trivial to decide objectively which of the models is the most accurate one. Taking advantage of measurement data, the model parameter identification and the model selection techniques may be adopted to select the most plausible model and set of model parameters. Accordingly, such techniques enable to detect modeling errors and thus to reduce model uncertainty. This work explores the applicability of two rational methods, Bayesian Inference and the Error-Domain Model Falsification, in the context of model selection and model parameter identification for structural finite element analysis. The two probabilistic techniques utilize measurement data to select the most plausible finite element model within a set of finite element models or to falsify inadequate finite element models. The concept is explained by means of an example. In future research work, the benefits of using rational methods for finite element model selection will be demonstrated for the modeling of reinforced concrete members with (hidden) damage due to load history or reinforcement corrosion.
Topics
TNO Identifier
777325
ISBN
9783319594705
Publisher
Springer
Source title
2017 FIB Symposium - High Tech Concrete: Where Technology and Engineering Meet, 12-14 June 2017
Editor(s)
Lukovic, M.
Hordijk, D.A.
Hordijk, D.A.
Pages
1689-1698
Files
To receive the publication files, please send an e-mail request to TNO Repository.