Improving Concrete Mix Design Predictions with Machine Learning Algorithms

The construction industry has a crucial impact on environmental pollution and on the phenomenon of climate change.One of the most impactful activities in this sector, is the concrete production. The concrete industry alone is a significant source of CO2 emissions, with concrete being the most widely used construction material. Concrete is a material mainly composed of cement, water, aggregates and additives. Despite the percentage of cement in concrete mixtures is comprised between 10% and 15% by weight, this component contributes up to 90% of the associated greenhouse gas emissions.The process involved in concrete production is complex and involve many uncertainties. To compensate for these unknowns and ensure structural safety, cement is frequently added in excess to improve mechanical performance. However, reducing the cement content in concrete formulations can have considerable environmental benefits. Therefore, enhancing the precision of predicting concrete’s mechanical characteristics is essential for lowering cement consumption without compromising safety.In this paper, the application of machine learning techniques to forecast the mechanical properties of concrete is presented. The main goal is to use the concrete mix design data to accurately predict the properties related to the resistance and the workability of concrete. To this extent, a dataset comprising roughly 1100 mix designs was utilized, combined with weather-related data from concrete batching plants. The study emphasizes regression and classification models to predict concrete properties, taking into account environmental and meteorological factors at the production site.