Prestressed UHPC Beam Flexural Capacity Prediction Improvement with Constraint-Guided Neural Networks
International Journal of Bridge Engineering, Management and Research · 2025
DOI: 10.70465/ber.v2i3.31 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Bridge Engineering, Management and Research, 2025. Verfasst von Hailong Qu und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
This study develops a constraint-guided neural network (CGNN) approach to address prediction accuracy degradation in machine learning models when tested on unseen data after training with limited datasets, while simultaneously enhancing the generalization capabilities of conventional mechanical equations. The proposed model predicts flexural capacity of prestressed ultrahigh performance concrete (UHPC) beams using a small experimental dataset collected from literature. Physical constraints incorporated into the model include geometrical relationships, reinforcement ratio limitations, and strength correlations. Comprehensive performance comparisons demonstrate that the CGNN model outperforms traditional machine learning methods—including back propagation neural networks, categorical boosting, extreme gradient boosting, random forests, and decision trees—as well as conventional predictive equations. The CGNN exhibits superior performance across multiple metrics: mean ratio, coefficient of variation, mean relative error, coefficient of determination, maximum ratio, and minimum ratio, while maintaining consistent accuracy between training and testing phases. Model interpretability is achieved through SHapley Additive exPlanations (SHAP), and a user-friendly Graphical User Interface has been developed to facilitate adoption by UHPC design practitioners.
Zitieren
APA
Qu, H., Teng, L., Zhu, Y., Qu, Z., & Wu, H. (2025). Prestressed UHPC Beam Flexural Capacity Prediction Improvement with Constraint-Guided Neural Networks. International Journal of Bridge Engineering, Management and Research. https://doi.org/10.70465/ber.v2i3.31
Vancouver
Qu H, Teng L, Zhu Y, Qu Z, Wu H. Prestressed UHPC Beam Flexural Capacity Prediction Improvement with Constraint-Guided Neural Networks. International Journal of Bridge Engineering, Management and Research. 2025. doi:10.70465/ber.v2i3.31
BibTeX
@article{Qu2025,
author = {Qu, Hailong and Teng, Le and Zhu, Yanping and Qu, Zhongke and Wu, Hantao},
title = {Prestressed UHPC Beam Flexural Capacity Prediction Improvement with Constraint-Guided Neural Networks},
journal = {International Journal of Bridge Engineering, Management and Research},
year = {2025},
doi = {10.70465/ber.v2i3.31}
}
EUVIMED