Prediction of axial structural performance of CFRP-wrapped concrete compressive members using machine learning tools

Umara Nasir , Nejib Ghazouani , Nabil Ben Kahla , Aqeel Ur Rehman

Abstract

This study presents a data-driven approach for predicting the axial load-carrying strength (ALCS) of fiber-reinforced polymer (FRP) confined concrete columns using artificial neural networks (ANN). A comprehensive experimental database of 265 FRP-confined columns with varying geometries, material properties, and confinement characteristics was developed. Initially, 14 existing empirical models were evaluated, and a modified version of the Teng et al. model was proposed, achieving an R2 of 0.9219. To enhance predictive accuracy, a multilayer feedforward backpropagation ANN model was trained and validated using 66% and 33% of the dataset, respectively. The optimal ANN architecture, consisting of 9 neurons in the first hidden layer and 5 in the second, yielded superior prediction performance with a correlation coefficient R2=0.9956, a mean absolute error (MAE) of 1.43%, and a predicted average ALCS of 4383.82 kN compared to the experimental mean of 4309.60 kN. These results demonstrate that the proposed ANN model offers a highly reliable and practical tool for estimating the axial strength of FRP-confined concrete columns, outperforming traditional analytical models and supporting the advancement of intelligent structural design methods.

Key Words

ANN models; columns; experimental database; FRP confined concrete; MAE; parametric study

Address

Umara Nasir, Aqeel Ur Rehman — Department of Civil Engineering, University of Engineering and Technology Taxila, 47050, Pakistan
Nejib Ghazouani — Mining Research Center, Northern Border university, Arar 73213, Arar, Saudi Arabia
Nabil Ben Kahla — 1) Department of Civil Engineering, College of Engineering, King Khalid University, PO Box 394, Abha 61411, Saudi Arabia, 2) Center for Engineering and Technology Innovations, King Khalid University, Abha 61421, Saudi Arabia

PDF Viewer

Preview is limited to the first 3 pages. Sign in to access the full PDF.

← Back to Volume 37, No. 2