AI-Based Tool for Optimizing Crumb Rubber Concrete Mixes: A Practical Approach for Sustainable and Resilient Infrastructure

Crumb rubber concrete (CRC), produced by partially replacing fine aggregates with recycled tire rubber, enhances ductility, impact resistance, and energy absorption, making it attractive for sustainable construction. However, it is challenging to predict its fresh and hardened properties due to nonlinear interactions between rubber particles and other constituents. This study proposes an optimized back-propagation neural network (BPNN) model for accurately predicting CRC slump and 28-day compressive strength. A hybrid data set of 100 mix designs, compiled from laboratory tests and peer-reviewed literature, has been used to train and evaluate the model. The proposed 6-18-1 BPNN architecture, trained with the Levenberg–Marquardt algorithm, achieved strong predictive performance with R=0.9727 and root-mean square error (RMSE) = 1.0970 for slump, and R=0.9947 and RMSE=1.3132 for compressive strength. These results surpassed those obtained using support vector machine and multiple linear regression approaches. Robustness analysis through Monte Carlo simulations with ±5% input variation confirmed model stability, and experimental validation demonstrated close agreement with measured values. The developed artificial intelligence (AI) based framework offers a reliable and scalable decision-support tool for rapid CRC mix design optimization, reducing the need for time-intensive laboratory trials. From an industry perspective, the model enables faster design, supports resilient infrastructure planning, and contributes to sustainable construction practices.