Assessment of flexural and splitting strength of steel fiber reinforced concrete using automated neural network search

Flexural and splitting strength behavior of conventional concrete can significantly be improved by incorporating the fibers in it. A significant number of research studies have been conducted on various types of fibers and their influence on the tensile capacity of concrete. However, as an important property, tensile capacity of fiber reinforced concrete (FRC) is not modelled properly. Therefore, this paper intends to formulate a model based on experiments that show the relationship between the fiber properties such as the aspect ratio (length/diameter), fiber content, compressive strength, flexural strength and splitting strength of FRC. For the purpose of modeling, various FRC mixes only with steel fiber are adopted from the existing research papers. Automated neural network search (ANS) is then developed and used to investigate the effect of input parameters such as fiber content, aspect ratio and compressive strength to the output parameters of flexural and splitting strength of FRC. It is found that the ANS model can be used to predict the flexural and splitting strength of FRC in a sensible precision.