Artificial Neural Network-Based Prediction of the Compressive Strength of Eco-Friendly Concrete Incorporating PET Granules

Abstract

The incorporation of polyethylene terephthalate (PET) granules as a partial replacement for traditional aggregates in concrete presents a promising strategy for addressing the pressing issues of plastic waste accumulation and environmental degradation. This study investigates the mechanical behavior of eco-friendly concrete mixtures containing PET granules and proposes an artificial neural network (ANN) model to predict their compressive strength with high accuracy. Experimental datasets from prior studies were used to train the ANN, with key input variables including water content, cement dosage, fine and coarse aggregates, and PET granule proportion.

The developed ANN model exhibited strong predictive performance, achieving a high correlation coefficient (R² = 0.9444), thereby confirming its robustness and reliability. Sensitivity analysis revealed that the water-to-cement ratio and PET granule content were the most influential factors affecting compressive strength. The findings indicate that PET granules can effectively replace natural aggregates up to an optimal threshold without compromising structural integrity, supporting their viability in sustainable concrete formulations.

This research underscores the dual environmental and engineering benefits of integrating recycled PET into concrete. The proposed ANN model offers a valuable tool for optimizing mix design while minimizing the need for extensive laboratory experimentation. Overall, the study contributes to sustainable construction practices by promoting the reuse of plastic waste and the development of greener building materials.