prediction of shear capacity of rc deep beams via a soft computing method

It is well known that the shear capacity of rc deep beams is affected by many mechanical and geometric parameters. the accurate prediction of the shear capacity still stands out as one of the major stumbling blocks in structural engineering practice. traditional prediction methods have often proven less than precise. on the other hand, artificial intelligence-based methods, particularly those represented by svms, have presented themselves as a promising alternative. this research employed an enhanced machine learning technique, known as wls-svm, to estimate the shear capacity of reinforced concrete deep beams. in assembling a comprehensive dataset, 214 experimental results are obtained from the literature. from selected inputs and outputs, under the supervision of a teaching-learning type approach, a predictive model is derived via wls-svm. this model is compared with other ai-based methods and codified design procedures. it presented the best accuracy, with major statistical indicators, including an r² of 0.9804, showing the superiority of the wls-svm approach when compared to other methods. therefore, the study's results reveal wls-svm as a very accurate and viable option for the structural calculation and design of reinforced concrete deep beams.