Journal of Materials and Engineering Structures
Volume 12, Numéro 4, Pages 513-534
2025-12-30
Prediction And Feature Analysis Of Self-compacting Concrete Strength Using Machine Learning With Recycled Coarse Aggregate And Supplementary Cementitious Materials
Authors : Pulkit Kavendra . Nigam Shivi .
Abstract
This study presents a machine learning based approach to predict the CS of SCC incorporating RCA and supplementary cementitious materials (SCMs), aimed at promoting sustainable construction practices. A dataset comprising 337 SCC mixes with varying input parameters such as cement (40–635 kg/m³), SCMs (0–592 kg/m³), water to binder ratio (0.25–0.45), and curing age (7–120 days) was utilized. Five regression models were evaluated- Extra Trees, AdaBoost, SVR Gradient Boosting, and ANN. Among them, Gradient Boosting achieved the highest accuracy with R² = 0.8876, RMSE = 4.032 MPa, and MAE=2.6382 MPa. Extra Trees followed closely with R²=0.8309 and RMSE=4.9444 MPa. SHAP interpretability analysis revealed that cement content and curing age were the most influential parameters in predicting CS. The study confirms the effectiveness of ML models in replacing time consuming lab tests and provides an interpretable, data driven framework for optimizing concrete mix design using industrial waste materials.
Keywords
Machine Learning ; Gradient Boosting ; K Fold Cross Validation ; Hyperparameter Tuning
Les articles similaires
Kumar Dinesh
.
Rao V.v.l. Kanta
.
Lakshmy P.
.
pages 155-166.
Dosho Yasuhiro
.
pages 395-402.
Djelloul Omar Kouider
.
Tahenni Touhami
.
pages 263-272.
Kumar Dinesh
.
Kanta Rao V.v.l
.
Lakshmy P.
.
pages 233-244.
Senhadji Y.
.
Benosman A.s.
.
Escadeillas G.
.
Mouli M.
.
Laoufi L.
.
Khelafi H.
.
pages 232-243.
