Machine learning based graphical interface for accurate estimation of FRP-concrete bond strength under diverse exposure conditions

Abstract

Predicting FRP-to-concrete bond strength (FRP-CBS) under diverse exposure conditions is an intricate task influenced by multiple variables. Yet, existing pertinent models have several limitations. Accordingly, this study proposes a novel data driven machine learning (ML) methodology to predict the FRP-CBS considering various exposure conditions. A comprehensive database on single and double lap-shear strength tests on concrete specimens was meticulously compiled. Twenty-seven analytical models were used to appraise the developed ML models. Feature importance analysis was conducted to ascertain the influence of input parameters on bond strength. The proposed data-driven ML models attained exceptional accuracy and superior performance compared to existing analytical models. To enhance the accuracy of bond strength estimation and simplify the process for practicing engineers and FRP applicators, a user-friendly graphical interface was developed. It could eliminate the need for complex design procedures, making it easier to accurately estimate the FRP-CBS, thus improving overall efficiency in engineering practice.

Keywords

• FRP
• Concrete
• Bond strength
• Durability
• Machine learning
• XGBoost