Machine learning-aided prediction of windstorm-induced vibration responses of long-span suspension bridges

Aug 3, 2026

Upload by Sage

Long-span suspension bridges are highly vulnerable to windstorm-induced vibrations, creating major obstacles for field measurements due to multicollinearity and nonlinear relationships between wind features and dynamic bridge responses. To overcome these limitations, the study presents a machine-learning-assisted prediction framework that combines a predictor selector based on regularized neighborhood component analysis with kernel regression using a regularized support vector machine optimized via Bayesian hyperparameters. The Hardanger Bridge under different windstorms validates the approach and benchmarks it against state-of-the-art regression methods. Performance achieves R-squared values from 89% to 98%, demonstrating strong practicality and improved prediction of bridge dynamics across wind conditions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Structural health monitoring and machine learning in long-span bridges

Wind-induced effects and vibration mechanisms

Methodology and modeling (overview)

Received: 15 March 2024 Accepted: 16 November 2024 DOI: 10.1111/mice.13387 RESEARCH ARTICLE Machine learning-aided prediction of windstorm-induced vibration responses of long-span suspension bridges Alireza Entezami1,2 Hassan Sarmadi3 1 Department of Civil and Environmental Engineering, Politecnico di Milano, Milan, Italy 2Mitacs, Montréal, QC, Canada 3 Head of Research and Development, IPESFP Startup Company, Mashhad, Iran Correspondence Alireza Entezami, Department of Civil and Environmental Engineering, Politecnico di Milano, Milan, Italy. Email: [alireza.entezami@polimi.it](alireza.entezami@polimi.it) Abstract Long-span suspension bridges are significantly susceptible to windstorminduced vibrations, leading to critical challenges of field measurements along with multicollinearity and nonlinearity between wind features and bridge dynamic responses. To address these issues, this article proposes an innovative machine learning-assisted predictive method by integrating a predictor selector developed from regularized neighborhood components analysis and kernel regression modeling through a regularized support vector machine adjusted by Bayesian hyperparameter optimization. The crux of the proposed method lies in advanced machine learning algorithms including metric learning, kernel learning, and hybrid learning integrated in a regularized framework. Utilizing the Hardanger Bridge subjected to different windstorms, the performance of the proposed method is validated and then compared with state-of-the-art regression techniques. Results highlight the effectiveness and practicality of the proposed method with the minimum and maximum R-squared rates of 89% and 98%, respectively. It also surpasses the state-of-the-art regression techniques in predicting bridge dynamics under different windstorms. 1 INTRODUCTION Suspension bridges are complex, and critical civil structures designed to span long distances for vital transportation links and traffic flow management in areas where other bridge types may be impractical. These structures are often susceptible to natural and man-made risks such as windstorms, earthquakes, traffic overloads, etc., which can threaten their integrity and serviceability. Continuous assessment through structural health monitoring (SHM) is essential to ensure the safety and reliability of long-span suspension bridges, along with their critical components (Qarib & Adeli, 2014; Xu & Xia, 2012) . An SHM program involves sensor installation, data measurement, numerical model construction and updating, data analysis, and decision-making (H. Li et al., 2006) . This program is crucial for the development of modern and resilient structures, which are capable of selfdiagnosis and adaptation under both normal and extreme conditions, particularly in smart cities (Javadinasab Hormozabad et al., 2021) . For this program, it is feasible to take advantage of different sensing systems, which are critical in data collection for SHM (Amezquita-Sanchez et al., 2018; Pezeshki et al., 2023; Sarmadi et al., 2023) . From raw measured data, one can use different feature extractors (Amezquita-Sanchez & Adeli, 2015) to discover richer and more useful information such as acceleration root-mean-square (RMS), modal parameters (D. Kim et al., 2017; Perez-Ramirez et al., 2016), etc. Decision-making in SHM predominantly relies on diverse machine learning This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2024 The Author(s) . Computer-Aided Civil and Infrastructure Engineering published by Wiley Periodicals LLC on behalf of Editor. ENTEZAMI and SARMADI (ML) techniques encompassing classification, regression, and anomaly detection (Amezquita-Sanchez et al., 2020) . For SHM, these techniques are often applicable to visionbased damage classification (Jian

FAQs

What problem does the paper address for long-span suspension bridges?

It addresses the difficulty of accurately predicting windstorm-induced vibration responses, caused by multicollinearity and nonlinear dependencies between wind features and bridge dynamics.

How does the proposed method improve prediction accuracy?

It integrates a predictor selector from regularized neighborhood component analysis with kernel regression modeled by a regularized support vector machine whose hyperparameters are tuned using Bayesian optimization.

How was the method validated and how did it perform?

The approach was validated using the Hardanger Bridge subjected to different windstorms and compared with state-of-the-art regression techniques, achieving R-squared values between 89% and 98% and outperforming the benchmarks for predicting bridge dynamics under varying wind conditions.

Aug 3, 2026

Upload by Sage