Sustainable Recycled Aggregate Concrete Design Using XGBoost, Data Augmentation, and Deterministic Multi-Objective Optimization

Biglarijoo, Nader; Kharazmi, Shahaboddin; Danaei, Mohammad Reza; Shams, Amin

doi:10.22075/jrce.2026.2709

Sustainable Recycled Aggregate Concrete Design Using XGBoost, Data Augmentation, and Deterministic Multi-Objective Optimization

Articles in Press

Document Type : Regular Paper

Authors

Nader Biglarijoo ¹

Shahaboddin Kharazmi ²

Mohammad Reza Danaei ³

Amin Shams ¹

¹ Assistant Professor, Faculty of Civil Engineering, Semnan University, Semnan, Iran

² Assistant Professor, Faculty of Mechanical Engineering, Semnan University, Semnan, Iran

³ M.Sc. Student, Faculty of Civil Engineering, Semnan University, Semnan, Iran

10.22075/jrce.2026.2709

Abstract

Construction and demolition waste and excessive natural aggregate use have increased demand for sustainable concrete technologies. This study proposes an AI-assisted framework for predicting and optimizing recycled aggregate concrete (RAC). Using a Central Composite Design, 20 concrete mixtures were produced based on parent concrete strength, recycled coarse aggregate (RCA) replacement ratio, and cement content. To address limited data, a surrogate-based augmentation method combining Gradient Boosting Regression and Gaussian noise injection expanded the dataset to 500 realistic samples. An XGBoost model predicted compressive strength, splitting tensile strength, and water absorption with high accuracy (R² = 0.991, 0.980, and 0.998). Feature analysis showed that parent concrete quality mainly controls mechanical performance, while RCA ratio strongly influences water absorption. Optimization of 125,000 virtual mixtures identified an optimal RAC mix containing about 61% RCA and 320 kg/m³ cement, achieving balanced strength, durability, sustainability, and reduced experimental workload. The findings show that RCA could replace natural aggregates to maintain sustainable development in the construction industry.

Keywords

Machine learning

Multi-objective optimization

Predictive model

Recycled concrete

Sustainable development

Subjects