REVIEW OF MACHINE LEARNING APPLICATIONS FOR PREDICTING ULTIMATE LOAD CAPACITY OF REINFORCED CONCRETE COLUMNS CONFINED WITH FIBER REINFORCED POLYMER (FRP) JACKETS UNDER AXIAL COMPRESSION

Abstract

The prediction of ultimate load capacity in fiber-reinforced polymer (FRP) confined reinforced concrete columns is a complex engineering challenge that has garnered significant research attention over the past two decades. Traditional empirical models and finite element analysis approaches have limitations in capturing the intricate nonlinear behavior of FRP-confined concrete under axial compression. Machine learning algorithms offer promising alternatives for improving prediction accuracy and efficiency. This review systematically examines the application of various machine learning techniques, including artificial neural networks, support vector machines, random forest, gradient boosting, and ensemble methods for predicting the ultimate load capacity of FRP-confined reinforced concrete columns. The paper analyzes 70+ peer-reviewed publications, evaluates model performance metrics, compares different machine learning approaches with conventional empirical methods, and identifies emerging trends in deep learning applications. Key findings indicate that machine learning models achieve superior prediction accuracy (R² values ranging from 0.92-0.95) compared to traditional empirical codes (R² values around 0.82-0.87). The review also presents a comprehensive bibliometric analysis of the research landscape and discusses future research directions including hybrid models, real-time prediction systems, and interpretability enhancements for machine learning-based prediction frameworks.