Machine learning-guided optimization of coarse aggregate mix proportion based on CO2 intensity index

Abstract

Aggregate accounts for 60‐80% volume fraction of concrete, which has a great influence on the CO2 emission and performance of concrete. Apart from natural coarse aggregate (NCA), recycled coarse aggregate (RCA) and carbonation recycled coarse aggregate (CRCA) are becoming an important component. This study established a database containing 925 experimental samples of compressive strength (CS) and CO2 emission, which including NCA, RCA, and CRCA concrete respectively. Additionally, the CO2 intensity index was introduced to evaluate the CS and CO2 emission. Machine learning (ML) methods were utilized to establish prediction models for CS, CO2 emissions, and CO2 intensity. The significance of features was analyzed through SHAP and PDP. For the optimization of coarse aggregate mix proportion, the GA and MOPSO algorithms were employed for single and bi-objective optimization designs, respectively. The results indicated that the optimization of coarse aggregate mix proportion can effectively reduce CO2 emission and CO2 intensity of concrete. A CRCA content of 30% is optimal for achieving both enhanced CS and reduced CO2 emissions. The carbonation treatment of RCA presents a viable approach for mitigating CO2 footprint and enhancing the mechanical properties of RCA concrete. The proposed optimization frame can facilitate appropriate decision making for low-carbon concrete design.

Keywords

• Carbonation
• CO2 utilization
• Recycled coarse aggregate
• Machine learning
• Computational design optimization