Buy article PDF
The purchased file will be sent to you
via email after the payment is completed.
US$ 35
|
Smart Structures and Systems Volume 38, Number 2, August 2026 , pages 155-181 DOI: https://doi.org/10.12989/sss.2026.38.2.03 |
|
|
|
Uncertainty-aware hybrid model for predicting future traffic states of bridges under missing traffic data |
||
Zengpeng Zhang, Xiang-Yu Wang, Da-Wei Lin, Yuguang Fu, Fan Kong, Zhongxiang Liu, Zhen Sun
|
||
| Abstract | ||
| Accurately predicting future bridge traffic states is crucial for intelligent transportation systems and structural health monitoring. However, this task still faces two major challenges. First, traffic data often suffer from sparsity or missing values due to equipment failures, limited sampling frequency, and transmission interference. Second, most existing methods rely on deterministic point estimates, which fail to quantify the inherent uncertainty in traffic states. To address these issues, this paper proposes AMU-TrafficNet (Attention-enhanced multi-scale uncertainty-aware traffic state network), an uncertainty‑aware hybrid model designed to predict bridge traffic states under realistic missing traffic data conditions. The model innovatively integrates multi-scale feature extraction, attention mechanisms, and uncertainty quantification. By incorporating task‑specific loss functions, it achieves the dual capability of reconstructing missing traffic data and providing uncertainty-aware future traffic state predictions. A proof-of-concept case study on an in-service cable-stayed bridge is conducted to demonstrate the feasibility of the proposed method. The results show promising performance in reconstructing missing data and predicting future traffic states within the scope of this specific case. Furthermore, this study examines the effects of input types, data scales, and hyperparameters on the results, conducts ablation studies, and compares the proposed model with several conventional models to validate its effectiveness. | ||
| Key Words | ||
| bridges; future traffic state prediction; machine learning; missing traffic data reconstruction; uncertainty quantification | ||
| Address | ||
| (1) Zengpeng Zhang, Xiang-Yu Wang, Da-Wei Lin, Zhen Sun: School of Civil Engineering, Southeast University, Nanjing 211189, China; (2) Yuguang Fu: School of Civil and Environmental Engineering, Nanyang Technological University, 639798, Singapore; (3) Fan Kong: School of Civil Engineering, Hefei University of Technology, Hefei 230009, China; (4) Zhongxiang Liu: School of Transportation, Southeast University, Nanjing 210096, China; (5) Zhongxiang Liu, Zhen Sun: State Key Laboratory of Safety, Durability and Healthy Operation of Long Span Bridges, Nanjing 211189, China. | ||