Shear Strength Prediction of SFRC Beams Using Machine Learning

Resumo

Concrete is one of the most used structural materials in the world, but it has some limitations such as
brittle behavior and low deformation capacity when being tensioned, which makes it susceptible to the appearance
of cracks that occur in its interior. The steel fiber reinforced concrete (SFRC) improves the mechanical behavior
of concrete structures allowing, for example, the partial or total replacement of the stirrup reinforcement in
structural elements. However, due to the complex shear behavior, it is necessary to understand the contribution of
each composite parameter to the behavior of the structural member in order to better design new structures with
this material. Therefore, the present work uses Machine Learning algorithms to predict the shear strength of SFRC
beams from a survey of experimental data found in the literature. Different input parameters are considered for
this study, such as the fiber volume fraction, fiber aspect ratio, among others. The results demonstrated a high
accuracy of the ML algorithm, reaching a R2 = 0.96. Moreover, a comparison of the shear prediction capacity
between ML approach and the analytical models available in literature is performed. This research aims to
contribute to a better understanding of the shear phenomenon of SFRC beams, taking into account different aspects
that can assist and aid design engineers in the conception and execution of structural projects.