Prediction of flexural strength in FRP bar reinforced concrete beams through a machine learning approach

Purpose The purpose of this study is to assess the incorporation of fiber reinforced polymer (FRP) bars in concrete as a reinforcement enhances the corrosion resistance in a concrete structure. However, FRP bars are not practically used due to a lack of standard codes. Various codes, including ACI-440-17 and CSA S806-12, have been established to provide guidelines for the incorporation of FRP bars in concrete as reinforcement. The application of these codes may result in over-reinforcement. Therefore, this research presents the use of a machine learning approach to predict the accurate flexural strength of the FRP beams with the use of 408 experimental results. Design/methodology/approach In this research, the input parameters are the width of the beam, effective depth of the beam, concrete compressive strength, FRP bar elastic modulus and FRP bar tensile strength. Three machine learning algorithms, namely, gene expression programming, multi-expression programming and artificial neural networks, are developed. The accuracy of the developed models was judged by R 2 , root means squared and mean absolute error. Finally, the study conducts prismatic analysis by considering different parameters. including depth and percentage of bottom reinforcement. Findings The artificial neural networks model result is the most accurate prediction (99%), with the lowest root mean squared error (2.66) and lowest mean absolute error (1.38). In addition, the result of SHapley Additive exPlanation analysis depicts that the effective depth and percentage of bottom reinforcement are the most influential parameters of FRP bars reinforced concrete beam. Therefore, the findings recommend that special attention should be given to the effective depth and percentage of bottom reinforcement. Originality/value Previous studies revealed that the flexural strength of concrete beams reinforced with FRP bars is significantly influenced by factors such as beam width, effective depth, concrete compressive strength, FRP bars’ elastic modulus and FRP bar tensile strength. Therefore, a substantial database comprising 408 experimental results considered for these parameters was compiled, and a simple and reliable model was proposed. The model developed in this research was compared with traditional codes, and it can be noted that the model developed in this study is much more accurate than the traditional codes.