Machine Learning Approach for Optimization of Concrete Mixes with Supplementary Materials

Abstract

Durability of concrete has always been an important topic since many concrete failures or signs of failure result from exposure to severe service or environmental conditions. The Gulf region has a harsh metrological environment and is currently witnessing a wide urbanization development. Billions of dollars are being spent on reinforced concrete construction projects and their repair in this region. These projects will not only exert pressure on resources but will also require high-performance materials for long-lasting durability and efficient repair systems. This study investigates the optimization of concrete mix designs using machine learning techniques, specifically focusing on the interplay between Ground Granulated Blast Furnace Slag (GGBS), Micro silica content, and their effects on concrete water permeability and cost. A Random Forest Regressor was employed to model the complex relationships between these variables, revealing key insights into how varying proportions of GGBS and Micro silica influence the overall performance and economic feasibility of concrete mixes. The analysis identified an optimal mix containing 61.1% GGBS and 14 kg/m³ micro silica, which achieves a favorable balance between low water permeability (1.37 mm) and cost efficiency (49.5 USD/m³). The insights gained from the model can inform better material selection for mix designs, contributing to more durable and cost-effective concrete structures, reducing the need for costly repairs, and minimizing resource consumption in urban development projects.