Bridging the gap in concrete slump prediction: Scalable, generalized, and deployable machine learning approach

Traditionally, slump cone‐based experimental testing has been the standard method for assessing concrete slump. In recent years, recognizing the advantages of data‐driven approaches, researchers have started exploring the use of machine learning (ML) for slump prediction of structural concrete. However, their proposed methodologies are limited in scalability, generalizability, and the practical deployment of ML‐based concrete slump prediction. This study aims to address these research gaps by presenting an efficient approach for slump prediction of industrial concrete. The scalability is enhanced through the analysis of a large dataset comprising 7497 observations from 38 distinct concrete designs. These designs correspond to various structural placements of concrete. Ensuring reliability and generalizability, a wide range of ML models are investigated using a systematic ML process that includes rigorous model selection and evaluation. Computational results show that the random forest outperforms other ML models with an accuracy of 89% and an area under the curve of 0.99. A user‐friendly interface is developed to apply ML with ease and enhanced usability for fresh concrete quality assessment. Focusing on slump prediction and performance assessment of structural concrete, this study includes valuable discussions on the concepts of robustness, scalability, generalization, and deployment of ML methods.