Flexural strength prediction of concrete beams reinforced with hybrid FRP and steel bars based on machine learning

In the past decade, hybrid fiber-reinforced polymer (FRP) and steel reinforced concrete (hybrid FRP-steel RC) beams have attracted significant interest due to their relatively superior flexural behavior compared with the pure FRP bars reinforced concrete (FRP-RC) beams. The recent trend in research involves the utilization of machine learning (ML) tools to accurately predict the flexural behavior of RC beams. In this study, the two artificial intelligence technologies, namely artificial neural network (ANN) and gene expression programming (GEP), through a data-driven approach, were used to establish reliable and accurate ML models for predicting the flexural strength of hybrid FRP-steel RC beams. To build the ML models, a huge database of 134 data points was collected from the peer-reviewed published work. In both ANN and GEP models, eight main parameters were considered as input variables, including the beam width (b) b ) and depth (d), d ), steel and FRP reinforcement area (As A s and A f ), yield strength of steel bar (fy), f y ), concrete compressive strength (fc), f c ), the elasticity modulus of FRP bar (Ef) E f ) and the ultimate tensile strength of FRP bar (ffu). f f u ). The results demonstrated that the proposed ML models had lower mean absolute error (MAE), MAE ), mean absolute percentage error ( MAPE ) and root mean squared error ( RMSE ) as well as higher correlation coefficient (R2) R 2 ) than the ACI 440.11-22 equations. Finally, a parametric study was conducted using the suggested ML models to explore the effect of each input variable on the flexural strength of hybrid FRP-steel RC beams.