Answers are grounded in this paper and cite their sources.
: Wind-induced vertical and torsional vibrations in long-span cable-supported bridges can cause excessive deck oscillations, twisting motions, and aerodynamic instabilities, potentially compromising structural safety and serviceability. Given the critical challenges of field measurements under strong wind loads such as data loss and sensor malfunction, bridge vibration prediction via machine learning provides substantial benefits for reliable monitoring and management. This study proposes a deep probabilistic regression method founded on Ensemble Bayesian Neural Networks (EBNN) to predict vibration behavior of long-span cable-supported bridges subjected to windstorms. The prediction framework lies in forecasting the root-mean-square values of vertical and torsional acceleration responses using some wind features extracted from wind data measurements. To manage predictive uncertainties caused by interdependencies among wind features, a statistical feature selection strategy is applied to enhance the reliability and stability of the proposed method. Probabilistic hyperparameter tuning based on Bayesian optimization is also implemented to fine-tune the EBNN regression model and enhance its generalization and predictive capabilities. The major contribution of this study lies in developing a deep learning-aided predictive method by integrating Bayesian inference, probabilistic modeling, and ensemble learning to accurately predict bridge dynamic behavior under strong wind loads. Monitoring data of a real-world long-span suspension bridge are used to validate the proposed method along with several comparative studies. Results confirm that EBNN achieves R 2 values ranging from 95%-98% and 93%-95% for the vertical and torsional responses in the training and testing phases, respectively.
No takes yet. Share an insight, caveat, or question.
Entezami et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: