Prediction of Shear Capacity of Reinforced Concrete Beams using Support Vector Regression and Adaptive Neuro-Fuzzy Inference Algorithms Optimized with Meta-Heuristic Algorithms

Considering the complexity of shear mechanisms of reinforced concrete beams and the effects of various parameters, creating a general model for the accurate estimation of the shear capacity is difficult. In addition, most guidelines for the determination of the shear capacity of reinforced concrete beams in empirical design codes have been obtained experimentally. Artificial intelligence algorithms have been widely used in this area in recent years. In this study, SVR, PANFIS, and GANFIS algorithms were used to predict the shear capacity of reinforced concrete beams. In this regard, the data of 175 experimental RC beam samples were collected. In these algorithms, values ​​of nine parameters affecting shear capacity were used as the input parameter and the shear capacity of the reinforced concrete beams as the output parameter. Using the Kfold validation method, training and test data were defined, and the predictions were performed accordingly. The results of predictions showed that the neuro-fuzzy inference system model with the genetic optimization algorithm had a higher accuracy than other algorithms with a second root mean square error of 0.06634 and a correlation coefficient of 0.996. Also, the grey system theory was used to determine the parametric sensitivity of the study variables on the shear capacity of reinforced concrete beams. The results showed that the mean coefficient of sensitivity analysis of the longitudinal rebar percentage parameter is greater than other parameters, indicating that the longitudinal rebar percentage parameter had more effects on shear capacity.