Machine Learning for Sustainable Concrete: Predictive Approaches Using Industrial Waste Materials

Fly ash (FA) has emerged as a promising additional cementitious ingredient in the concrete industry, with considerable benefits in terms of sustainability and performance of concrete. At the same time, the rapid increase in Marble Cutting Slurry Waste (MCSW) produced by ornamental stone firms has prompted major environmental issues about soil, air, and water contamination. The advantages of adding MCSW to concrete mixtures have been investigated in certain experiments by several researchers. Despite the known benefits of incorporating FA and MCSW in concrete, a data-driven understanding of its impact on compressive strength, particularly through Machine Learning (ML) models, remains underexplored. In this research work, an ML-based approach is used to forecast the compressive strength (CS) of concrete incorporating FA and MCSW, with the goal of improving resource use while addressing environmental concerns. A variety of machine learning techniques were applied, such as Decision Tree (DT), Random Forest (RF), Categorical Gradient Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM). The dataset comprises 270 points. These points were compiled from available literature sources. The performance of the developed models is assessed through different statistical indicators. All five ML models did well, but XGBoost was the best in predicting CS. Notably, XGBoost showed the highest coefficient of determination (R2) value of 0.97, signifying its greater prediction accuracy in contrast to DT, RF, LightGBM, and CatBoost. Moreover, the SHapley Additive exPlanations (SHAP) method is employed to interpret the predictions of the models. The findings provide a robust data-driven framework for optimizing eco-friendly concrete mix designs, contributing to sustainable construction practices and informed decision-making in material selection.