PREDICTION OF REINFORCED CONCRETE COLUMNS LIMIT STATES USING MACHINE LEARNING ALGORITHM

Abstract

The assessment methods for estimating the behavior of the complex mechanics of reinforced concrete structural elements were primarily based on experimental investigation, followed by collective evaluation of experimental databases from the available experimental literature. There is still a lot of uncertainty today about the strength and deformability criteria that have been derived from tests due to the differences in the experimental test setups of the individual research studies that fed into the databases. Following these investigations, the regulatory methods of seismic assessment were developed. The topic covered in this research is the effect of test setup on the derived criteria, and the second-order effects that the test setups have introduced into the behavior of structural elements. The research focuses on elements that exhibit pronounced strength degradation with plastic deformation and brittle failure characteristics. The shear strength reduction that has been attributed to the magnitude of the imposed ductility is investigated, and it is determined how much of this degradation is recognizable, i.e., how much is a consequence of the experimental setup nonlinearity. While the available methods of assessing shear strength differ because they are all empirical, they all acknowledge the contribution of individual resistance mechanisms, such as concrete, transverse reinforcement, and axial load. The experimental setup nonlinearity has an impact on the last of these three contributions. In this work, the experimental results are correlated with a revised formulation of column shear strength after the values have been corrected, with a special focus on elements with inadequate structural detailing configuration. Finally, through the use of machine learning algorithms the development of an improved formula for predicting the shear capacity of reinfroced concrete columns is performed.