Prediction of Shear Strength of Reinforced Concrete Deep Beams Using Neuro-Fuzzy Inference System and Meta-Heuristic Algorithms

It is generally accepted that the shear strength of Reinforced Concrete (RC) deep beams depends on the mechanical and geometrical parameters of the beam. The accurate estimation of shear strength is a substantial problem in engineering design. However, the prediction of shear strength in this type of beams is not very accurate. One of the relatively accurate methods for estimating shear strength of beams is Artificial Intelligence (AI) methods. Adaptive Neuro-Fuzzy Inference System (ANFIS) was presented as an AI method. In this study, the efficiency of ANFIS incorporating meta-heuristic algorithms for predicting shear strength of RC beams was investigated. Meta-heuristic algorithms were used to determine the optimum parameters of ANFIS for providing the efficient models of the prediction of the RC beam shear strength. To evaluate the accuracy of the proposed method, its results were compared with those of other methods. For this purpose, the parameters of concrete compressive strength, cross-section width, effective depth, beam length, shear span-to-depth beam ratio (a/d), as well as percentage of longitudinal and transverse reinforcement were selected as input data, and the shear strength of reinforced concrete deep beam as the output data. Here, K-fold validation method with k = 10 was used to train and test the algorithms. The results showed that the proposed model with second root mean square error of 25.968 and correlation coefficient of 0.914 is more accurate than other methods. Therefore, neural fuzzy inference system with meta-heuristic algorithms can be adopted as an efficient tool in the prediction of the shear strength of deep beams.