Predicting the strength of fiber reinforced polymer materials externally bonded to masonry using artificial intelligent techniques

Fiber reinforced polymer (FRP) retrofits are widely used to strengthen structures due to their advantages such as high strength-to-weight ratio and durability. However, the bond strength between FRP and masonry is crucial for the success of these retrofits. Limited data exists on the shear bond between FRP composites and masonry substrates, necessitating the development of accurate prediction models. This study aimed to create machine learning models based on 1583 tests from 56 different experiments on FRP-masonry bond strength. The researchers identified key factors influencing failure load and developed machine learning models using three algorithms. The proposed models outperformed an existing model with up to 97% accuracy in predicting shear bond strength. These findings have significant implications for designing safer and more effective FRP retrofits in masonry structures. The study also used Sobol sensitivity analysis and SHapley Additive exPlanations (SHAP) analysis to understand the machine learning models, identifying key input features and their importance in driving predictions. This enhances model transparency and reliability for practical use.