Performance Evaluation of Sustainable Concrete Incorporating Mineral Fillers: Experimental Insights and Predictive Modeling Using Advanced Machine Learning

Concrete production significantly contributes to environmental concerns due to its high demand for natural resources and carbon emissions. To address these issues, sustainable alternatives such as incorporating quarry dust as a partial replacement for fine aggregates have gained attention. However, optimizing such mixtures while ensuring desirable engineering properties remains challenging. This study investigates the use of quarry dust in concrete through experimental testing and advanced machine learning (ML) modeling to evaluate its performance as an eco-friendly alternative. An experimental program was conducted for two grades of concrete (M15 and M20) to assess the fresh and hardened properties of concrete incorporating quarry dust at replacement levels ranging from 0% to 60%, along with a superplasticizer. Key properties such as workability and compressive, tensile and flexural strength were measured to evaluate the concrete’s performance. The experimental results were then used to develop ML models, specifically employing Gene-Expression Programming (GEP) and Random Forest (RF), to predict concrete mixtures' compressive, flexural and tensile strength based on various mix proportions. The models were validated using statistical performance metrics and sensitivity analysis to ensure their reliability. The predictive performance of RF slightly outperformed GEP, though GEP provided valuable insights through interpretable equations, making it an effective tool for optimizing sustainable concrete mixtures. Sensitivity analysis revealed that the results were well-aligned with existing literature and experimental findings, confirming the models' consistency and credibility. The study demonstrates the potential of soft computing techniques in enhancing the development of sustainable concrete.