A comparative study of shear strength prediction models for SFRC deep beams without stirrups using Machine learning algorithms

Abstract

This study aims to evaluate the shear strength of stirrups-free Steel Fiber Reinforced Concrete (SFRC) deep beams and to predict their shear strength values using Machine Learning (ML) algorithms. A database of 172 tested SFRC deep beam specimens was compiled and processed, and the results were analyzed using Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost), and Gene Expression (GEP) algorithms. The LightGBM and XGBoost algorithms were the most accurate, with R2 values of 97.8% and 94.5%, respectively. The authors proposed a closed-form GEP-based model to predict the shear strength of SFRC deep beams, with an R2 value of 78.9%. The analysis results indicate that the GEP model accurately predicts the impacts of concrete strength, flexural steel percentage, and the ratio of shear span to beam depth. The findings of this study provide practitioners with a solid foundation for making accurate and practical assessments of shear strength in stirrups-free SFRC deep beams.