Data-driven machine learning approaches for predicting slump of fiber-reinforced concrete containing waste rubber and recycled aggregate

This research investigates the slump behavior of fiber-reinforced rubberized recycled aggregate concrete (FR3C) and its significance in the concrete industry. The fresh properties of FR3C are influenced by the variability of its constituents. The presence and proportions of waste rubber and fiber play crucial roles in the slump of fresh FR3C. Achieving the desired slump in FR3C is a critical challenge that relies on understanding the intricate internal relationships among its constituents. To address this challenge, a set of machine learning-based models is proposed to accurately predict the slump and uncover the internal dependencies within FR3C. The models are trained and tested using a comprehensive dataset comprising 464 experimental data points with varying mix proportions. Twelve machine learning models, including linear regression, ridge regression, lasso regression, support vector machine, k-nearest neighbors, decision tree, random forest, AdaBoost, Voting Regressor, Gradient Boost, CatBoost, and XGBoost, are employed in the analysis. The input characteristics considered in the models encompass nominal aggregate size, water-to-cement ratio (W/C), percentage of rubber, replacement level of recycled coarse aggregate (RCA), percentage of fiber and its type, use of plasticizer, and fly ash percentage. The results demonstrate that the XGBoost model outperforms the others in accurately predicting the slump of FR3C. It exhibits the highest coefficient of determination (R2), lowest root mean squared error (RMSE), and demonstrates strong performance on both training and testing data. The evaluation of feature importance emphasizes the critical influence of the W/C ratio, nominal aggregate size, and fiber percentage on the slump behavior of FR3C.