ANN and Gradient Boosting-Based Predictive Techniques for Punching Shear Capacity of Concrete Slabs Reinforced with FRP Bars

This study represents a pioneering effort in using machine learning to enhance the design concrete slabs reinforced with fiber-reinforced polymer (FRP) bars, traditionally reliant on steel-based design equations. The research employed two distinct machine learning (ML) approaches: an artificial neural network (ANN) and Gradient Boosting, first introduced in this scenario. Both models developed from 99 experimental tests. The ANN model demonstrated a remarkable coefficient of determination (R²) of 0.96, significantly surpassing available standards such as ACI 440.1R-15, CAN/CSA S806-12, and JSCE recommendations and existing literature models. The introduction of Gradient Boosting is a novel aspect of this research. It gave better results with an R² of 0.98, surpassing the ANN’s 0.96. Therefore, its application marks a significant advancement in the field. This comparative analysis not only validates the superior predictive performance of the GB and ANN but also highlights the potential of Gradient Boosting in structural engineering applications. Through parametric studies, the investigation further analyzed the discrepancies between empirical and theoretical punching shear strengths, underlining the limitations of conventional design methods. The findings indicate the effectiveness of ML, particularly GB, in accurately forecasting the performance of FRP-reinforced slabs, thereby offering a substantial improvement over traditional design methodologies.