Research on Performance Prediction Model of Recycled Concrete Based on Machine Learning

This research seeks to utilize machine learning methods to forecast the behavior of recycled concrete, specifically emphasizing the evaluation of its compressive strength. By gathering and manipulating recycled concrete performance data, we identified significant factors such as water-cement ratio, aggregate replacement rate, and porosity index to support the model with reliable input. In the model selection stage, we compared multiple machine learning algorithms and finally selected support vector machine as the optimal model. The support vector machine achieved MSE of 0.020, MAE of 0.130, and R2 of 0.89 on the test set, demonstrating superior prediction capabilities. The findings demonstrate that machine learning techniques offer substantial benefits in understanding the intricate nonlinear connections in recycled concrete, offering fresh insights for the optimal utilization of recycled materials in design and engineering. This research confirms that machine learning is effective in materials science and shows its high potential in sustainable materials research. Experimental design and cross-validation guarantee the reliability and repeatability of results, offering a new method to evaluate the performance of recycled concrete with practical significance.