What does this research mean for the field?

An integrated framework combining optimized Random Forest models and reliability analysis accurately predicts the structural performance of GFRP-confined concrete columns, demonstrating that while FRP confinement enhances safety margins, load eccentricity significantly reduces axial capacity. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The aim is to evaluate the structural performance of GFRP confined concrete columns under different loading scenarios using machine learning and reliability analysis.

June 15, 2026Discover Materials0 citationsOpen Access

Performance evaluation of GFRP confined concrete columns under axial and eccentric loading using experimental machine learning and reliability analysis

AGAbdurra’uf M. GoraUniversity of Nottingham Malaysia Campus JJJ. JayaprakashVellore Institute of Technology University MAM P AnwarUniversity of Nottingham Malaysia Campus

Key Points

The aim is to evaluate the structural performance of GFRP confined concrete columns under different loading scenarios using machine learning and reliability analysis.
Developed predictive models using 124 data samples from literature and experiments.
Implemented machine learning techniques including Random Forest optimized with Particle Swarm Optimization, Whale Optimization Algorithm, and Long Short-Term Memory.
Conducted statistical validation to ensure reliability and robustness of the models.
Optimized Random Forest models outperformed LSTM in predictive accuracy and stability.
Increased load eccentricity significantly reduced axial capacity and reliability of columns.
The proposed framework effectively predicts performance, suggesting improved design strategies for GFRP-confined columns.

Abstract

This study presents an extensive evaluation of the structural performance of concrete columns confined with glass fiber-reinforced polymer (GFRP) under both concentric and eccentric loading scenarios. An integrated framework that combines experimental investigation, machine learning (ML) modelling, and reliability analysis is proposed to improve predictive accuracy and structural safety assessment. The predictive models were developed using 124 data samples, which included 112 data points obtained from existing literature and 12 experimental results from the current investigation. Three machine learning models were implemented and assessed: Random Forest optimized using Particle Swarm Optimization (RF–PSO), Random Forest optimized using Whale Optimization Algorithm (RF–WOA), and Long Short-Term Memory (LSTM). When compared to LSTM, the optimized RF-based models showed better predictive performance with greater accuracy and stability, demonstrating the applicability of ensemble learning approaches for tabular structural datasets. The robustness of the dataset and the reliability of the model were verified by statistical validation, which included multicollinearity tests, correlation analysis, and distribution evaluation. The predicted axial load capacities were further integrated into a reliability framework built based on the First-Order Reliability Method (FORM) that evaluates structural safety in terms of reliability index ( β ) and probability of failure ( Pf ). The findings show that while improved FRP confinement enhances structural performance and safety margins, increasing load eccentricity considerably reduces axial capacity and reliability. In order to properly capture geometric impacts on column behavior, the study also includes a physically significant parameter, the column aspect ratio. Overall, the proposed hybrid framework demonstrates a reliable and interpretable approach for predicting and evaluating the performance of GFRP-confined concrete columns, providing significant insights for reliability-based design and optimization of FRP-confined structural systems.

Ask AI

Helpful

Bookmark

View Full Paper