Machine Learning for predicting Compressive and Flexural strength of Fiber-reinforced Concrete

Abstract

This thesis explores the use of machine learning algorithms to predict the compressive and flexural strength of aluminum waste-based fiber-reinforced concrete as a green alternative for the development of mechanical properties at reduced environmental expense. Smart prediction models are to be created using three main algorithms: Artificial Neural Networks (ANN), Random Forest, and XGBoost. The procedure entailed experimental data collection, normalization as preprocessing, correlation analysis, and partitioning data into training, validation, and testing sets. Results indicated that XGBoost performed best with R² equal to approximately 0.80 for compressive strength and 0.97 for flexural strength. The study concludes that the use of AI in concrete mix design reduces reliance on costly experimental testing and sustainable building practices. Future studies are encouraged to expand data sources and explore hybrid model approaches for greater accuracy.