Improved efficiency and accuracy in Bayesian structural model updating via DREAM_CMAZS sampling and Kriging surrogates

verfasst von

Xinghua Chen, Ding Zhang, Michael Beer, Mohammad Aadil

Abstract

The inherent uncertainties in structural engineering systems cause inevitable discrepancies between actual states and idealized design models, posing challenges for accurate analysis and health monitoring. Traditional Bayesian model updating methods suffer from slow convergence and high computational costs when dealing with high-dimensional parameter spaces, mainly due to excessive finite element model evaluations. To overcome these limitations, this study proposes a novel and computationally efficient Bayesian model updating framework by integrating the DREAM_CMAZS sampling algorithm—which is a multi-chain differential evolution algorithm incorporating multiple strategic advantages, capable of accurately achieving convergence goals in high-dimensional parameter spaces under limited convergence conditions—with a Kriging surrogate model. The framework leverages the strengths of advanced Markov Chain Monte Carlo sampling—enhanced by Covariance Matrix Adaptation Evolution Strategies (CMA-ES), adaptive subspace sampling, and a survival of the fittest mechanism (SF)—and efficient surrogate modeling to enable scalable and robust inference for complex structural systems. Validation on three representative cases—a numerical simply supported beam, an experimental aluminum frame, and a laboratory-scale steel cantilever beam—demonstrates parameter identification errors within 0.001 %–4.259 %, a 61 % reduction in model evaluations, and up to a 77.1 % improvement in computational efficiency compared to conventional Bayesian approaches. This approach effectively addresses convergence instability and computational inefficiency in high-dimensional finite element model updating, providing a practical and scalable tool for structural health monitoring, damage detection, and risk-informed decision-making in civil infrastructure.

Organisationseinheit(en)

Institut für Risiko und Zuverlässigkeit

Externe Organisation(en)

China Three Gorges University
The University of Liverpool
Tongji University

Typ

Artikel

Journal

Structures

Band

80

ISSN

2352-0124

Publikationsdatum

10.2025

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Architektur, Tief- und Ingenieurbau, Bauwesen, Sicherheit, Risiko, Zuverlässigkeit und Qualität

Elektronische Version(en)

https://doi.org/10.1016/j.istruc.2025.109856 (Zugang: Geschlossen)