Genetic programming-based model for estimating maximum pull load of fiber-reinforced polymer-to-concrete bond interfaces with graphical user interface implementation

This study presents a novel, interpretable machine learning framework for predicting the maximum pull load of fiber-reinforced polymer (FRP) bonded to concrete substrates. A comprehensive test database comprising 983 datasets was gathered from relevant existing studies. The datasets include key input parameters such as concrete compressive strength, bond length, width of FRP sheet, width of concrete block, FRP thickness, and elastic modulus of FRP sheets, with the maximum pull load as the output parameter. Utilizing this curated database, a symbolic regression model based on genetic programming (GP) was developed to uncover the nonlinear relationships among critical variables including axial stiffness of FRP, bond length, and concrete compressive strength. The model's predictive performance was evaluated using standard regression metrics, achieving mean absolute error (MAE) and root mean square error (RMSE) values below 5 kN, mean absolute percentage error (MAPE) slightly above 10%, and coefficient of determination (R2) exceeding 0.90 on both training and testing datasets. These results confirm the model's accuracy and generalizability. Unlike black-box models, symbolic regression offers an explicit mathematical expression, ensuring transparency and interpretability for engineering applications. To facilitate practical deployment, a user-friendly graphical user interface (GUI) named MaxPLoad-FRP-Concrete-GPaided-PredictionModel was developed, enabling practitioners to input key design parameters and obtain immediate, interpretable predictions. This tool serves as a valuable decision-support system in the structural design and quality control for FRP-strengthened concrete structures.