Modeling shear strength of medium- to ultra-high-strength concrete beams with stirrups using SVR and genetic algorithm

This paper presents a data-driven machine learning approach of support vector regression (SVR) with genetic algorithm (GA) optimization called SVR-GA for predicting the shear strength capacity medium- to ultra-high concrete beams longitudinal reinforcement and vertical stirrups. One hundred forty eight experimental samples collected different geometric, material physical factors from literature were utilized fivefold cross validation. Shear influence such as stirrup spacing, beam width, span-to-depth ratio, effective depth beam, compressive tensile strength, product ratio yield served input variables. The simulation results show that model can achieve highest accuracy in prediction based on testing set coefficient determination (R2) 0.9642, root mean squared error 1.4685 absolute 1.0216 superior traditional SVR 0.9379, 2.0375 1.4917, which both perform better than multiple linear ACI-318. Furthermore, sensitivity analysis reveals most important variables affecting result are strength. Three-dimensional input/output maps employed reflect nonlinear variation two coupling All all, proposed excellent stirrups comparison obtained by SVR, MLP