Data-Driven Prediction Models For Total Shear Strength of Reinforced Concrete Beams With Fiber Reinforced Polymers Using An Evolutionary Machine Learning Approach

Abstract

The strength of Reinforced Concrete (RC) structural elements may need to be improved due to building usage changes or damages that occurred after exposure to extreme loads. Fiber Reinforced Polymer (FRP) is commonly being used to enhance the performance of reinforced concrete beams due to several advantages such as having high strength and being lightweight. To perform the analysis and design of the members, there is a need for accurate models to determine the total shear strength of the structural elements strengthened with FRP sheets. In this paper, genetic programming has been successfully utilized to develop models to predict the total shear strength of the reinforced concrete beams. A strategy is adopted here to find a simple yet accurate formula to estimate the shear strength. These models can correlate the total shear strength of the beams reinforced with FRP sheets to the geometric and material properties of RC beams and FRP sheets, without the need for expensive laboratory tests. A compressive database of the total shear strength of the RC beams with FRP sheets was created from the literature. External validation and sensitivity analysis, using various statistical criteria, were conducted to assess the precision and validity of the proposed models. Based on 785 RC beams strengthened by externally bonded FRP sheets, tested between 1992 and 2022, two data-driven models were developed to predict the total shear strength of RC beams strengthened with FRP. The calculated correlations for Models I and II are 0.883 and 0.940, respectively. Superior performance was obtained compared to other models from the literature in accuracy. The proposed models can be utilized for design purposes and the development of structural solutions for existing structures.