Investigation of artificial intelligence models in predicting the shear strength of FRP reinforced concrete beams

Reinforced concrete (RC) represents a fundamental medium in civil and transportation infrastructure. However, the integration of fiber-reinforced polymer (FRP) bars has emerged as a compelling solution to conventional steel for enhanced durability. Given the complex mechanical response of FRP-reinforced members, the precise quantification of their shear resistance is paramount for ensuring structural reliability. This study evaluates several artificial intelligence (AI) paradigms, including artificial neural networks (ANNs), support vector regression (SVR), random forest (RF), and M5Rules, to capture the non-linear relationships governing the shear strength of FRP-reinforced concrete beams. A primary contribution of this study is the integration of a hybrid SVR-GWO model, which integrates the grey wolf optimizer (GWO) to refine predictive fidelity. Empirical results demonstrated that the proposed SVR-GWO architecture outperformed standalone models, achieving the MAE of 6.98 kN and the MAPE of 11.01%. Furthermore, the hybrid approach facilitated significant error reductions in MAPE, ranging from 15.7% to 68.3% compared to benchmark AI techniques. By effectively capturing the non-linear relationships between input parameters and shear resistance, this study provides structural designers with a high-precision, AI-driven computational tool for the optimized design of FRP-RC beams.