From Borehole Data to Screening-Level Settlement Assessment: A Physics-Informed Machine Learning Framework for Highway Bridges in Data-Scarce Alluvial Corridors
Journal of Intelligent Decision Making and Information Science · 03.10.2026 · 3 (17s) · 225-250
DOI: 10.17762/jidmis.v3.5353 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 03.10.2026. Verfasst von Ramesh Kumar Singh. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Bridge foundation settlement prediction is essential for ensuring structural safety, serviceability, and long-term maintenance of transportation infrastructure. Although recent Machine Learning (ML) techniques have improved settlement estimation, most existing approaches primarily rely on data-driven learning and often lack physical interpretability, uncertainty quantification, and effective maintenance decision support. To address these limitations, this study proposes a Physics-Informed Extreme Gradient Boosting (Physics-Informed XGBoost) framework for accurate and interpretable bridge foundation settlement prediction. The proposed methodology integrates physics-based settlement formulation with data preprocessing, feature engineering, feature selection, grouped cross-validation, hyperparameter optimization, SHapley Additive exPlanations (SHAP)-based model interpretation, uncertainty quantification using a 95% Prediction Interval (PI), and bridge maintenance prioritization. The developed model achieved excellent predictive performance, with a Root Mean Square Error (RMSE) of 0.562 mm, Mean Absolute Error (MAE) of 0.474 mm, Mean Squared Error (MSE) of 0.316 mm², and a Coefficient of Determination (R²) of 0.9943. Furthermore, the proposed framework achieved a classification accuracy of 98.15%, while SHAP analysis identified the most influential geotechnical parameters governing settlement prediction. The uncertainty analysis further confirmed the reliability and robustness of the proposed framework. Overall, the developed approach provides an accurate, explainable, and practical solution for bridge settlement assessment and maintenance planning, supporting more resilient and sustainable transportation infrastructure management.
Zitieren
APA
Singh, R. K. (2026). From Borehole Data to Screening-Level Settlement Assessment: A Physics-Informed Machine Learning Framework for Highway Bridges in Data-Scarce Alluvial Corridors. Journal of Intelligent Decision Making and Information Science, 3(17s), 225-250. https://doi.org/10.17762/jidmis.v3.5353
Vancouver
Singh RK. From Borehole Data to Screening-Level Settlement Assessment: A Physics-Informed Machine Learning Framework for Highway Bridges in Data-Scarce Alluvial Corridors. Journal of Intelligent Decision Making and Information Science. 2026;3(17s):225-250. doi:10.17762/jidmis.v3.5353
BibTeX
@article{Singh2026,
author = {Singh, Ramesh Kumar},
title = {From Borehole Data to Screening-Level Settlement Assessment: A Physics-Informed Machine Learning Framework for Highway Bridges in Data-Scarce Alluvial Corridors},
journal = {Journal of Intelligent Decision Making and Information Science},
year = {2026},
volume = {3},
number = {17s},
pages = {225--250},
doi = {10.17762/jidmis.v3.5353}
}
EUVIMED