Research on Green Building Materials Performance Optimization and Prediction Based on Artificial Intelligence

This study proposes a CNN-RNN hybrid deep learning model for predicting and optimizing the mechanical properties of recycled concrete and bamboo fiber composites. Validation using the UCI Concrete Dataset (1,030 samples) demonstrates that the model achieves prediction accuracy with MAE=2.83MPa, RMSE=3.21MPa, and R²=0.951, outperforming existing methods such as Transformer and XGBoost. After 300 freeze-thaw cycles (ASTM C666 standard) validation, the optimized material maintains a relative dynamic elastic modulus retention rate of 90.8%, meeting engineering application requirements.