Second-order Analysis of Slender GFRP Reinforced Concrete Columns Using Artificial Neural Network

Abstract

Analysis of slender concrete columns reinforced with glass fiber-reinforced polymer (GFRP) bars is required for design and optimization purposes. Finite element, finite difference, and analytical-numerical tools were developed in literature to accurately predict the response of slender columns reinforced with GFRP bars. The primary challenge with using these tools is the high computational cost required to accurately predict the response of the slender columns due to the material and geometric nonlinearities. Design codes and standards provide a simplified second-order analysis method, called the moment magnification method, to avoid the high computational cost associated with conducting complex nonlinear second-order analyses. The accuracy of the moment magnification method is often compromised when used in analyzing slender elements. Therefore, there is a need for developing efficient and accurate methods of analyzing slender columns, particularly when a large number of analyses is required (e.g., reliability analysis and optimization applications). The objective of this study is to propose an efficient regression-based method to predict the second-order effects on GFRP reinforced concrete (RC) columns by utilizing the artificial neural network (ANN) approach. Nonlinear finite-difference analysis of slender GFRP-RC columns was utilized to generate a training dataset including multiple eccentricity ratios, slenderness ratios, reinforcement ratios, section aspect ratios, and material properties to train the ANN. Preliminary results indicate an average error of less than 1 kN for the second-order analysis based on the trained ANN model. Preliminary reliability analysis of slender GFRP-RC columns indicates that using the developed ANN method yields a significant reduction in the computational cost as compared with existing finite difference methods.