.png){kind=icon}
.png){kind=icon}
Return
A probabilistic method for structural model updating using a model-data hybrid driven technique
.png){kind=icon}
PRE
AI
DOI：10.1016/j.istruc.2025.109057.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Nonlinear structural model updating is often hindered by various uncertainties such as measurement noise, modeling errors, and discretization effects. This paper develops a probabilistic method for nonlinear structural model updating using a model-data hybrid driven technique. Building on the Bayesian inference framework, the proposed method derives the posterior probability density functions (PDFs) of structural parameters based on the relationship between finite element (FE) model computations and measured responses. To address the complexities in parameter posterior distributions, which are often intractable to obtain through standard numerical integration, the Transitional Markov Chain Monte Carlo (TMCMC) algorithm is employed. Central to the proposed approach is the incorporation of two DenseNet-based surrogate models. The first model, FE-DenseNet, is designed to establish a mapping relationship between structural parameters and dynamic responses, serving as an effective substitute for the traditional structural FE model and significantly reducing computational costs. The second model, ME-DenseNet, addresses the challenge of modeling error representation, facilitating a comprehensive analysis of multi-source uncertainty. The applicability and accuracy of the proposed method are demonstrated through numerical investigations on a 3-story 4-span steel frame structure, complemented by a comparative analysis under various noise levels to affirm the method's robustness against noise. Further validation is achieved through experimental applications involving a shake table test of a scaled-down bridge tower, underscoring the method's potential for practical engineering applications.
Keywords:
Nonlinear model updating
Bayesian inference
Surrogate model
Modeling errors
Measurement errors