Estimating Ultimate Moment Capacity of Spirally Reinforced Concrete Columns Using Various Artificial Neural Networks

Abstract

Over the last few decades, intensive investigations on the artificial neural network capabilities for addressing structural engineering problems have been concluded in the literature. Multiple models for predicting the load-bearing capacity and failure mode have been developed in this regard. However, most of the studies on the capabilities of artificial neural networks for estimating the ultimate moment capacity were focused on the feedforward backpropagation approach. As a result, this research aims to investigate the performance of using different artificial neural network approaches to forecast the ultimate moment capacity of spiral RC columns. As a part of the study, the performance of feedforward backpropagation, cascade-forward neural networks, and generalized regression neural networks will be compared and evaluated against experimental and traditional results. The findings demonstrated that artificial neural networks provide a reliable method for forecasting the spiral RC columns' moment capacity, and they can outweigh code-based empirical formulation.