Shear transfer strength estimation of concrete elements using generalized artificial neural network models

Abstract

Abstract Based on published test findings, this article outlines the use of artificial neural networks (ANNs) to forecast the efficiency factor of shear transfer strength in concrete. Backpropagation neural networks with feed-forward have been employed. The ANN model was created by incorporating a huge experimental database and carefully selecting the architecture and training procedure. The presented ANN model offered a more accurate tool to compute R (where R is a measure of the closeness of association of the points in a scatterplot to a linear regression line based on those points) and capture the impacts of five primary parameters: concrete compressive strength, steel reinforcement ratio, steel yield strength, fiber volumetric ration, and steel fiber aspect ratio are given from experimental data. The obtained results reveal that the first important parameter is concrete compressive strength. In addition, ρ y f y parameter represents the normalized tensile force in steel reinforcements of section, whereas the smallest importance parameter L/D is aspect ratio of steel fibers. Also, the current study illustrated the facilities of using generalized artificial neural networks on predicting the shear transfer strength across the concrete sections, whether they are fibrous or not. From the results, the correlation factor (R 2) is estimated to be about 83%, which means it had a good correlation within the input parameters. In addition, the mean absolute percentage error was 2.06.