Title: Modeling Tensile Strength of Concrete Using Support Vector Regression
Author(s): J. A. Guzmán-Torres, F. J. Domínguez-Mota, E. M. Alonso-Guzmán, and W. Martínez-Molina
Publication: Materials Journal
Appears on pages(s): 25-37
Keywords: electrical resistivity; machine learning (ML); support vector regression; tensile strength
Innovations are being developed with different additives in concrete mixtures to enhance their performance under certain conditions. This work models the split tensile strength through the electrical resistivity using a machine learning (ML) algorithm in different concrete mixtures. One of the most frequent nondestructive tests used on concrete is electrical resistivity (Er) for the simplicity of taking measurements on concrete elements. ML involves methods that provide solutions to advanced problems employing computational algorithms. This research employs a support vector regression algorithm, which can predict the split tensile strength value using a nondestructive test. The outcomes of this research exhibit a high correlation between electrical concrete resistivity and split tensile strength over 90%. Support vector regression algorithm
achieves accuracy over 93% in the forecasting task. Of note is that the modified concrete contains starch as an addition to prove its performance compared to conventional concrete.