Prediction of the Compressive Strength of Sustainable Concrete Produced with Powder Glass Using Standalone and Stack Machine Learning Methods

Waste glass forms a significant part of the solid waste stream globally. Since its consumption for manufacturing new glass is limited, it is mostly landfilled, which is not a sustainable mode of disposal and results in various environmental issues. Over the years, powder-grade waste glass has been used as a partial replacement for cement to produce eco-friendly concrete. This study aims to predict the compressive strength ofwaste glass concrete produced with 10, 20, and 25% wt.% replacement of cementwith powderwaste glass by employingArtificial Intelligence (AI). Specifically, the Stack (ensemble) machine learning approach, which combines multiple methods, including Adaptive Boosting (AdaBoost), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Decision Tree (DT), has been used to predict the 28 days compressive strength of concrete produced with various percentages of powder waste glass as partial replacement of cement. Comparison of the predicted compressive strengths with the laboratory test values shows that the employed machine learning (ML) models accurately predict the compressive strength of concrete mixtures that closely match the laboratory tested values. A comparison of the ML models' statistical performance data shows Stack's superior performance compared to other models.