An explainable hybrid machine learning framework for predicting rapid chloride permeability in durable concrete

Chloride penetration is a critical factor influencing the durability of reinforced concrete structures because it initiates reinforcement corrosion and significantly shortens service life under aggressive environmental conditions. Although the Rapid Chloride Permeability Test (RCPT) is widely adopted to evaluate concrete resistance against chloride ingress, its experimental implementation is laborious, costly, and time-intensive, making it unsuitable for large-scale investigations. To provide an efficient alternative, this study presents an interpretable hybrid machine-learning framework for accurate prediction of RCPT values in concrete mixtures. A comprehensive dataset containing 483 concrete samples compiled from published literature was used for model development. The predictive capability of the models was enhanced by constructing ratio-based and interaction-derived features, followed by feature optimization using a Supervised Concrete Autoencoder (SCAE) and Variational Transformer-based Feature Selection (VTFS). Four hybrid learning models were subsequently established and comparatively assessed, including LightGBM–SSO, RF–OO, TabNet–MFO, and SAINT–PSO. Among the evaluated models, LightGBM–SSO exhibited the best predictive performance, achieving an R 2 of 0.9631, an RMSE of 46.75, and an MAE of 36.58. To improve model transparency, SHAP analysis was employed to quantify the contribution of individual variables, identifying the fly ash-to-water ratio (F/W) as the dominant factor governing chloride permeability. Residual-based bootstrap analysis provided a descriptive comparison of prediction-error dispersion and interval coverage, without establishing independent predictive calibration. The findings indicate that the developed framework provides an accurate, interpretable, and computationally efficient approach for estimating chloride permeability, offering valuable support for durability-based concrete mixture design and reducing reliance on extensive laboratory testing.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-19
DOI
https://doi.org/10.1038/s41598-026-80431-2
Primary Topic
Concrete Corrosion and Durability
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An explainable hybrid machine learning framework for predicting rapid chloride permeability in durable concrete

Hamid Reza Hanif, Mahdi Ghafourivayghan, Mostafa Kharkan Ghahjavarestani
Scientific Reports
Concrete Corrosion and Durability
article

An explainable hybrid machine learning framework for predicting rapid chloride permeability in durable concrete

Hamid Reza Hanif, Mahdi Ghafourivayghan, Mostafa Kharkan Ghahjavarestani
article en

Abstract

Chloride penetration is a critical factor influencing the durability of reinforced concrete structures because it initiates reinforcement corrosion and significantly shortens service life under aggressive environmental conditions. Although the Rapid Chloride Permeability Test (RCPT) is widely adopted to evaluate concrete resistance against chloride ingress, its experimental implementation is laborious, costly, and time-intensive, making it unsuitable for large-scale investigations. To provide an efficient alternative, this study presents an interpretable hybrid machine-learning framework for accurate prediction of RCPT values in concrete mixtures. A comprehensive dataset containing 483 concrete samples compiled from published literature was used for model development. The predictive capability of the models was enhanced by constructing ratio-based and interaction-derived features, followed by feature optimization using a Supervised Concrete Autoencoder (SCAE) and Variational Transformer-based Feature Selection (VTFS). Four hybrid learning models were subsequently established and comparatively assessed, including LightGBM–SSO, RF–OO, TabNet–MFO, and SAINT–PSO. Among the evaluated models, LightGBM–SSO exhibited the best predictive performance, achieving an R 2 of 0.9631, an RMSE of 46.75, and an MAE of 36.58. To improve model transparency, SHAP analysis was employed to quantify the contribution of individual variables, identifying the fly ash-to-water ratio (F/W) as the dominant factor governing chloride permeability. Residual-based bootstrap analysis provided a descriptive comparison of prediction-error dispersion and interval coverage, without establishing independent predictive calibration. The findings indicate that the developed framework provides an accurate, interpretable, and computationally efficient approach for estimating chloride permeability, offering valuable support for durability-based concrete mixture design and reducing reliance on extensive laboratory testing.

Scientific Reports
Islamic Azad University, Tehran (IR), Ural Federal University (RU), Iran University of Science and Technology (IR)
Concrete Corrosion and Durability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon