A compressive strength prediction model for precast recycled aggregate concrete based on machine learning

Compared to conventional concrete, recycled aggregate concrete offers a more sustainable and environmentally friendly alternative by mitigating pollution and preserving natural resources. It provides an effective pathway for the construction industry to promote cleaner production and achieve long-term sustainable development. This study investigates the compressive strength of precast recycled aggregate concrete (PRAC), which incorporates recycled aggregates obtained from crushed precast concrete waste. A total of 153 experimental datasets were compiled, and four machine learning algorithms were employed to develop predictive models for the compressive strength of PRAC. Furthermore, SHapley Additive exPlanations were utilized to assess the relative importance of key influencing factors. The results reveal that the eXtreme Gradient Boosting model exhibits the highest predictive accuracy and robustness. Among all parameters, the water-to-cement ratio and fly ash content exert the most significant influence, both showing a negative correlation with the compressive strength of PRAC as their values increase.