The transition toward sustainable and resilient infrastructure requires both low-carbon construction materials and intelligent decision-making tools to support long-term asset performance. Recycled aggregate concrete (RAC), particularly incorporating waste brick aggregates, offers significant potential for reducing construction waste and embodied carbon, yet its structural performance remains highly variable due to material heterogeneity. This study presents an integrated experimental–data-driven framework to support sustainable infrastructure development through intelligent material design.
A series of laboratory tests were conducted to characterize the physical properties of blended recycled aggregates, including density, water absorption, and crushing resistance, and to evaluate their influence on the compressive strength of RAC. These experimental results were combined with curated data from the literature to establish a comprehensive database representing a wide range of recycled material qualities and mix proportions. Machine learning models were then developed to predict concrete performance based on aggregate characteristics and mixture parameters.
Comparative evaluation shows that a convolutional neural network provides superior prediction accuracy and generalization capability relative to conventional neural network approaches. The proposed framework enables rapid estimation of mechanical performance and supports optimized mix design without extensive trial batching. A demonstration case illustrates how the model can be used to identify suitable mix proportions that meet target strength requirements while maximizing recycled content.
The outcomes highlight the role of data-driven methods in advancing sustainable asset management and smart infrastructure systems, offering a practical pathway toward resilient, resource-efficient, and lower-carbon construction.