Concrete production imposes substantial environmental burdens, primarily through high carbon emissions and significant freshwater usage. This study addresses these challenges by developing a machine learning-based model to predict the compressive strength of concrete incorporating non-potable water, supporting sustainable construction practices. A comprehensive dataset of 1,056 samples was compiled from existing literature, encompassing key mix parameters such as fine and coarse aggregates, water-to-cement ratio, pH, and various supplementary cementitious materials. Multiple regression models were evaluated to predict compressive strength. Among these, the best-performing model achieved an R² of 0.98 and an RMSE of 1.45, demonstrating excellent predictive accuracy. Feature importance analysis identified the water-to-cement ratio, fine aggregate, and pH as the most influential variables affecting strength development. The study also applied explainable AI techniques to improve model interpretability and support informed engineering decisions. Sensitivity analysis confirmed model robustness across variable pH conditions, reinforcing its applicability to real-world wastewater variability. The results underscore the value of integrating non-potable water into concrete design and demonstrate the potential of optimized ML models to enhance resource efficiency, reduce environmental impact, and guide the development of greener infrastructure solutions.

Non-potable water, sustainable concrete, compressive strength, LSBoost, machine learning