Prediction Model-based Multi-objective Optimization for Mix-ratio Design of Recycled Aggregate Concrete

The prediction model for mechanical properties of RAC was established through the Bayesian optimization-based Gaussian process regression (BO-GPR) method, where the input variables in BO-GPR model depend on the mix ratio of concrete. Then the compressive strength prediction model, the material cost, and environmental factors were simultaneously considered as objectives, while a multi-objective gray wolf optimization algorithm was developed for finding the optimal mix ratio. A total of 730 RAC datasets were used for training and testing the predication model, while the optimal design method for mix ratio was verified through RAC experiments. The experimental results show that the predicted, testing, and expected compressive strengths are nearly consistent, illustrating the effectiveness of the proposed method.