Modeling for torsional strength prediction of strengthened RC beams

This study investigates three models to predict the torsional strength of FRP-strengthened reinforced concrete (RC) structural beams: artificial neural network (ANN), nonlinear regression model (NLR), and linear regression model (LR). The researchers examined data from over 96 tested FRP-strengthened beams to develop these prediction models. The models account for 10 distinct variables (input parameters), such as the RC beam's width and height, the FRP sheet's thickness and elastic modulus, the yield stress of the longitudinal and transverse steels, the compressive strength of the concrete, the effective width of the FRP strips along the beam's length, the center-to-center spacing of the FRP strips, the angle of wrapping, and the number of FRP layers. The beam's torsional strength is an appropriate parameter. Several statistical measures, including correlation coefficient (R2), root mean squared error (RMSE), mean absolute error (MAE), Scatter Index (SI), and objective (OBJ) values, were employed to assess the efficiency of the presented models. With R2, RMSE, MAE, OBJ, and SI values of 0.99, 3.07 kN m, 2.41, 2.63, and 0.17 kN m, respectively, the results demonstrated that the ANN model outperformed the other models in predicting the ultimate torsional strength of strengthened RC beams. This study provides an important database that may be used as a benchmark for future predicts of the torsional strength of strengthened RC beams. The influence of each parameter on the torsional strength of these beams was further studied using sensitivity analysis. The results showed a highly accurate prediction of the torsional strength of FRP-strengthened RC beams.