Surrogate-assisted finite element model updating for detecting scour depths in a continuous bridge

Yi He, Judy P. Yang*, Jie Yu

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

11 Scopus citations

Abstract

In this study, the Kriging surrogate-assisted method is proposed to improve the computational efficiency in finite element model updating (FEMU) based scour depth detection for the first time. Unlike the conventional methods of FEMU that update the structural parameters by the frequency surrogate models, the proposed method further takes the mode shapes into account by constructing the modal assurance criterion (MAC) surrogate models. The bi-objective functions of the frequency discrepancies and MAC values are directly optimized by the nondominated sorting genetic algorithm II to circumvent the specification of weighting factors as required in single-objective optimization. The preferred solution is selected from the nondominated solutions by the criterion of minimum distance from the equilibrium point. As the sensitivity analysis indicates that the transverse vibration modes are sensitive to the bridge scour, the transverse mode shapes measured on the bridge piers are adopted in scour detection. The illustrative examples show that the proposed method can greatly reduce the computational time with a competitive estimation of the scour depth vector. The parametric study further demonstrates the ability of the proposed method to yield satisfactory results even using the modal parameters of the first two modes with three measurement points.

Original languageEnglish
Article number101996
JournalJournal of Computational Science
Volume69
DOIs
StatePublished - May 2023

Keywords

  • Bridge scour
  • Finite element updating
  • Kriging surrogate model
  • Multi-objective optimization
  • NSGA-Ⅱ algorithm

Fingerprint

Dive into the research topics of 'Surrogate-assisted finite element model updating for detecting scour depths in a continuous bridge'. Together they form a unique fingerprint.

Cite this