Prediction of fresh and hardened concrete properties using machine learning algorithms

Concrete stands as the paramount construction material today, renowned for its exceptional moldability, strength, and durability. Its composition primarily involves cement, water, fine and coarse aggregates, with occasional mineral and chemical admixtures tailored for specific applications. Fresh concrete's workability, typically measured as slump, contrasts with hardened concrete, predominantly evaluated through compressive strength at 7 and 28 days. These key indicators, slump, and compressive strength, pivot on the properties and proportions of diverse concrete ingredients. Achieving concrete with requisite properties necessitates computing trial mix proportions based on ingredient properties utilizing codal provisions for mix design. Subsequently, samples prepared with these trial mixes undergo laborious and time-consuming laboratory testing. Hence, predicting these test outcomes becomes imperative to attain targeted concrete properties with minimal trials efficiently. This study employs varied machine learning algorithms to forecast concrete slump and compressive strength on the 7th and 28th day, leveraging input parameters concerning the properties and proportions of concrete ingredients. The dataset utilized in this investigation encompasses test data spanning a five-year program, encompassing training, cross-validation, and testing phases. Diverse machine learning models are deployed and evaluated based on their prediction accuracy to optimize concrete mix design. Use of diverse machine learning algorithms to predict properties of concrete using extensive input parameters, based on extensive experimental dataset form major scientific prowess of the present study.