Development of Machine Learning Algorithms to Predict the Ultimate Axial Capacity of Fire Damaged Circular Columns Repaired with CFRP Composites

Salman, Muhammad, Yaqub, Muhammad, Noman, Muhammad, Faizan, Muhammad, Javaid, Muhammad Ahmed et Maqsood, Saad (2026). Development of Machine Learning Algorithms to Predict the Ultimate Axial Capacity of Fire Damaged Circular Columns Repaired with CFRP Composites. Journal of Civil Engineering Frontiers, 7 (1).

PDF Licence d'utilisation : CC BY-NC-SA 4.0. Télécharger (1MB) | Prévisualisation

Résumé

This paper presents a study that extends the application of carbon fiber reinforced polymer (CFRP) composite confinement technology to strengthen circular concrete columns damaged by fire. This study utilized data from 125 column specimens sourced from the literature. It examined ten parameters: column diameter, height, initial compressive strength of concrete, initial tensile strength of steel, longitudinal reinforcement ratio, fire temperature, exposure time, number of CFRP layers, CFRP thickness, and CFRP tensile modulus, which were used as inputs for the model. The objective was to predict the ultimate axial strength of fire-damaged circular columns repaired with CFRP composites. This study employs both multiple regression analysis and a deep neural network (DNN) to predict the structural behavior of reinforced concrete (RC) columns and accurately forecast their repaired axial capacity. The proposed deep neural network (DNN) model demonstrated a robust agreement with experimental investigations, boasting an overall correlation factor (R) of 0.99852. Deep neural networks outperformed multiple regression analysis in predicting axial strength, with predictions closely matching experimental results from previous studies. The work also presents a parametric study to examine the effect of different input parameters on the axial strength of RC columns. Parametric analysis indicates that the repaired axial strength increases with higher concrete initial compressive strength, greater CFRP thickness and tensile modulus, and more CFRP layers, whereas it decreases with higher fire temperatures, longer exposure durations, and larger column diameters.

Type de document :	Article
Validation par les pairs :	Oui
Mots-clés :	Carbon fiber reinforced polymer ; Deep Neural Network (DNN) ; Heat damaged ; Circular concrete columns / Polymère renforcé de fibres de carbone ; Réseau neuronal profond (DNN) ; Dommages thermiques ; Poteau circulaire béton armé
Départements et unités départementales :	Département de mathématiques, informatique et génie
Date de dépôt :	01 oct. 2026 16:28
Dernière modification :	01 oct. 2026 19:53
URI :	https://semaphore.uqar.ca/id/eprint/3672