FRP bar-to-concrete connection durability in diverse environmental exposures: An optimal machine learning approach to predicting bond strength

The environmental durability of the bond between fibre-reinforced polymer (FRP) bar and concrete is a critical factor in studying the long-term performance of FRP-reinforced concrete structures. Accurate prediction of bond behaviour under harsh environmental conditions is essential to ensuring structural durability. This paper proposes a machine learning-based framework to model the durability of the FRP bar-to-concrete bond when exposed to diverse environmental conditions. A total of 1,244 durability test results are collected and analysed to establish a comprehensive database for predicting bond strength. A comprehensive range of parameters influencing bond durability, including the geometrical, material, mechanical, and environmental properties of both the FRP bars and concrete, are incorporated to develop accurate prediction models. The performance of the machine learning models is enhanced using Bayesian optimisation, which automatically tunes hyperparameter values during training. A detailed analysis is presented, including a comparative evaluation of machine learning models and a parametric study into the effects of environmental factors on bond durability. To demonstrate the practical value of machine learning, a closed-form equation is proposed and compared with existing formulations in design standards. The results demonstrate that the developed machine learning framework in this study is robust and capable of accurately predicting bond durability across a wide range of environmental conditions.