Intelligent optimization of finite element model parameters for large-span bridges based on MA-INFO algorithm

• Published in: Structures (Oxford) Vol. 70; p. 107617
• Main Authors: Ding, Jiaxuan, Gao, Liang, Shi, Shunwei, Zhang, Yanan, Yang, Mingmei
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2024
• Subjects: Finite element model updating; High-speed railway; INFO algorithm; Metamodel; Neural network; Suspension bridge; INFO algorithm; High-speed railway; Neural network; Suspension bridge; Metamodel; Finite element model updating
• ISSN: 2352-0124; 2352-0124
• DOI: 10.1016/j.istruc.2024.107617

This study proposes a bridge Finite Element (FE) model updating method based on the Metamodel Assisted weIghted meaN oF vectOrs (MA-INFO) algorithm, which improves the simulation accuracy for large-span bridge FE models. The first step of this method is to perform sensitivity analysis, with the aim of selecting the parameters involved in the update. Subsequently, a Radial Basis Function Neural Network (RBFNN) metamodel is constructed to replace time-consuming FE analysis. Combining the measurement data of static and dynamic responses, the INFO algorithm is introduced to update material and boundary condition parameters. Finally, A comparative analysis is conducted with other advanced optimization algorithms. Results demonstrate that the INFO outperforms multiple popular algorithms in terms of accuracy and convergence speed. After 18 iterations, the maximum relative error of natural frequency decreased from 12.8 % to 1.84 %, and the relative error of maximum deflection at mid-span decreased from 36.85 % to 3.39 %. Additionally, the updated model can accurately simulate static and dynamic responses, matching well with measurement results. The proposed MA-INFO offers a promising approach for accurate parameter estimation in large-span bridge FE models.