Explainable machine learning and ensemble models for predicting fresh properties of self consolidating concrete

Accurately predicting the fresh properties of self-consolidating concrete (SCC) is critical for enhancing construction efficiency and ensuring robust performance in complex and highly reinforced structures. Traditional experimental testing is time-consuming, costly, and prone to human error. In this study, over 2500 experimental data points were initially collected from 176 published studies to develop a comprehensive and reliable dataset. After rigorous data cleaning and filtering 348 SCC mix designs with complete rheological information were selected for model development. After rigorous data cleaning and filtering the dataset divided into 85% for training and 15% for testing. Five state-of-the-art machine learning (ML) models Gene Expression Programming (GEP), Deep Neural Networks (DNN), Decision Trees (DT), Support Vector Machines (SVM), and Random Forests (RF) were developed to predict slump flow (mm) and V-funnel time (s). Model interpretability was enhanced using Shapley Additive explanations (SHAP) and Partial Dependence Plots (PDP) to examine the influence of mix design variables. Among the models, GEP and DNN achieved the highest predictive accuracy with R² values up to 0.957 and 0.950 for V-funnel time (s) and 0.915 and 0.911 for slump flow (mm) respectively. The results highlight the strong potential of advanced ML methods to reliably forecast SCC's fresh properties, reduce reliance on extensive laboratory testing, and support rapid, data-driven optimization of concrete mixed designs in modern construction practice.