Comparative Study on Prediction of Interfacial Bond Strength of FRP with Concrete Using Machine Learning Methods

Abstract

As time elapses for a structure its strength decreases over some time but its utility keeps on growing this results in the same time demolition may cost more, for this problem rehabilitation is the solution. One rehabilitation material is Fiber Reinforced Polymer (FRP), binding on structural elements like beams and columns slabs strengthens the existing structure. So, it is necessary to know bond strength but it depends on several factors like FRP properties (Young modulus, thickness, bond length, tensile strength, width of FRP) and concrete block properties (compressive strength and width of concrete specimen). There are empirical equations to determine bond strength but in practice, these are very far from tested data. So, it is necessary to find the relation between FRP bond strength with respect to FRP and concrete properties. In present days, machine learning methods give good results when compared to conventional methods. So, it is necessary to use the best machine learning model to predict bond strength. This study aims to develop a comprehensive database of experimental results from direct shear specimens made of FRP concrete, and an assessment of the effectiveness of four machine learning algorithms including Support vector mechanism, Ridge regression, Lasso regression, and Elastic regression. The study will also develop a new equation for forecasting interfacial bond strength by considering the parameters discovered by the machine learning algorithm with interpretable physical meanings. The results of this study will provide valuable insights into the effectiveness of ML algorithms for predicting interfacial bond strength in FRP-concrete direct shear specimens and offer a new equation for forecasting interfacial bond strength with practical implications. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.