Explainable prediction model for punching shear strength of FRP-RC slabs based on kernel density estimation and XGBoost

Reinforced concrete (RC) slabs are widely used in modern building structures due to their superior properties and ease of construction. However, their mechanical properties are limited by their punching shear strength in the connection region with the columns. Researchers have attempted to add steel reinforcement in the form of studs and randomly distributed fibers to concrete slabs to improve the punching strength. An additional strengthening method that consists of the application is a Fiber-Reinforced Polymer (FRP). However, current codes poorly calculate the punching shear strength of FRP-RC slabs. The aim of this study is to create a robust model that can accurately predict its punching shear strength, thus improving the analysis and design of composite structures with FRP-RC slabs. In this study, 189 sets of experimental data were collected and expanded using kernel density estimation (KDE), considering the small amount of data. Secondly, a punching shear strength prediction model for FRP-RC panels was constructed using XGBoost and compared with the model modeled by codes and researchers. Finally, a model explainability study was conducted using SHapley additive exPlanations (SHAP). The results show that kernel density estimation significantly improves the robustness and accuracy of XGBoost. The R-squared, standard deviation, and root mean square error of XGBoost on the training set are 0.99, 0.001, and 0.001, respectively. On the test set, the R-squared, standard deviation, and root mean square error are 0.96, 62.687, and 67.484, respectively. The effective depth of the FRP-RC slabs is the most important and proportional to the punching shear strength. This study can provide guidance for the design of FRP-RC slabs.