Evolution of nonlinear analysis of RC sections: From iterative computations to rapid moment–curvature prediction using ANN models

Abstract

Nonlinear moment–curvature (M–φ) analysis is essential for evaluating the strength and ductility of reinforced concrete (RC) structures. Traditional numerical methods, such as iterative fiber analysis, provide high accuracy but are computationally intensive. This paper presents the evolution of approaches – from empirical and classical numerical methods to modern artificial intelligence (AI)-based models. Examples of applying artificial neural networks (ANN) and hybrid ANFIS models for predicting the nonlinear response of beams, columns, and beam–column joints are analyzed. Special attention is given to model architecture, input–output parameters, and prediction accuracy. The paper discusses differences between pointwise and parametric models, as well as recent approaches enhancing ANN interpretability. The study provides a critical review of ANN applications in defining moment–curvature relationships and identifies directions for future research in structural engineering.