Scopus Indexed Publications

Paper Details


Title
Machine Learning Approaches for Evaluating Concrete Durability Under Tropical and Coastal Exposure in Bangladesh

Author
, Komol Bhowmik,

Email

Abstract

Concrete durability in tropical and coastal environments is severely challenged by high humidity, wet–dry cycles, chloride ingress, and sulfate-rich soils, which accelerate degradation and reduce service life. This study applies machine learning models to predict durability-related properties of concrete, focusing on rapid chloride penetration. A dataset of 501 experimental samples, including water–cement ratio, cement content, aggregate characteristics, curing age, and exposure conditions, was used. Three models—Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LGBM) and Artificial Neural Networks (ANN) - were evaluated using Root Mean Squared Error (RMSE), Mean absolute error (MAE) and Coefficient of determination (R2). LGBM achieved the highest predictive accuracy (RMSE = 0.298, R2 = 0.937). SHAP and partial dependency plots were used to interpret feature influence. The results indicate that the ensemble boosting models outperform neural networks, providing robust and reliable predictions for concrete durability under aggressive tropical and coastal conditions.


Keywords

Journal or Conference Name
2026 International Conference on Power, Electronics, Communications, Computing, and Intelligent Infrastructure, PECCII 2026

Publication Year
2026

Indexing
scopus