Strength prediction of recycled aggregate concrete under sulfate attack using SVR–NSGA-II

This study investigates the prediction of residual compressive strength and mix proportion optimization of recycled aggregate concrete (RAC) under sulfate attack using machine learning algorithms. A database with 101 effective samples was used, considering 12 input parameters including raw materials, corrosive media, and exposure conditions and other relevant factors. A support vector regression model was developed to predict RAC strength, showing superior generalization and accuracy compared to traditional mathematical models. Additionally, a combination of nondominated sorting genetic algorithm II and ideal point method was employed to optimize RAC mix proportions, achieving both excellent sulfate resistance and cost efficiency. This research provides intelligent, efficient, and precise references for RAC application in engineering practice.