Eco-Friendly Self-Compacting Concrete Incorporating Fly Ash and Recycled Polyethylene Terephthalate (PET) Bottle Fibers

Abstract

Self-compacting concrete (SCC) has emerged as a sustainable construction material owing to its superior workability, reduced labor requirements, and enhanced durability. This study presents an integrated experimental and artificial intelligence (AI)-based framework for predicting the fresh and hardened properties of SCC incorporating Ground Calcium Carbonate (GCC), Class F fly ash, and recycled polyethylene terephthalate (PET) fibers. SCC mixtures were designed in accordance with IS 10262:2019, while fresh properties were evaluated following EFNARC (2005) guidelines and hardened properties were determined using the relevant Indian Standards. A comprehensive experimental database comprising 1,000 samples was developed, including input variables related to material proportions and output responses such as slump flow, T₅₀ flow time, V-funnel time, L-box ratio, compressive strength, flexural strength, and ultrasonic pulse velocity (UPV). The dataset was preprocessed using feature scaling, normalization, and an 80:20 train–test split with five-fold cross-validation before training six predictive models: Artificial Neural Network (ANN), Deep Neural Network (DNN), Extra Trees Regressor (ETR), Gradient Boosting Regressor (GBR), LightGBM, and CatBoost Regressor. Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and mean squared error (MSE). Explainable artificial intelligence techniques, including SHAP analysis and Partial Dependence Plots (PDP), were employed to interpret feature importance and nonlinear interactions. Among the investigated models, CatBoost Regressor demonstrated the highest predictive accuracy, exhibiting the strongest agreement between experimental and predicted values with the highest R² and the lowest prediction errors. The proposed AI-driven framework provides a reliable, interpretable, and computationally efficient approach for optimizing sustainable SCC mixtures while minimizing experimental effort, thereby supporting intelligent concrete mix design and promoting environmentally responsible construction practices.