Evaluation of machine learning techniques for capacity prediction of cold-formed steel beams subjected to bending

Aug 2, 2026

Upload by Sage

Stiffened and unstiffened cold-formed steel beams often show complex nonlinear bending behavior due to flexure effects and excessive slenderness, which can limit the accuracy of conventional design approaches. This study evaluates six classical machine learning algorithms and four ensemble models for predicting bending capacity using verified finite element analyses. A unified capacity prediction method is developed, including a stacking ensemble that combines six selectively chosen base models to improve performance. Ensemble methods, especially AdaBoost, Gradient Boosting, Random Forest, and Extra Trees, outperform single models, supporting reliable and cost-effective structural design decisions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

Introduction

Cold-Formed Steel sections and failure modes

Analytical design methods (DSM and EWM)

Machine learning for capacity prediction

Summary and findings

Rakenteiden Mekaniikka (Journal of Structural Mechanics) Vol. 57, No. 2, 2024, pp. 43–64 [http://rakenteidenmekaniikka.journal.fi](http://rakenteidenmekaniikka.journal.fi) [https://doi.org/10.23998/rm.144743](https://doi.org/10.23998/rm.144743) © 2024 The Authors Open access under the license CC BY 4.0 Evaluation of machine learning techniques for capacity prediction of cold-formed steel beams subjected to bending Ayman R. Hamdallah 1, Antti H. Niemi and Ahmad Abdullah Summary Stiffened Cold-Formed Steel (CFS) sections often exhibit intricate nonlinear behaviors attributable to factors such as flexure effects and excessive slenderness. Traditional design methodologies, including the direct stiffness method, may inadequately capture these subtleties, potentially resulting in conservative or suboptimal designs. This study aimed to evaluate the performance of various machine learning algorithms, including simple and ensemble models, to predict the bending capacity of stiffened and unstiffened cold-formed beams in pure bending. A parametric study was conducted based on verified finite element analysis, and the machine learning algorithms were utilized to develop a unified capacity prediction method. The performance of six classical machine learning algorithms and four ensemble models were compared. The findings demonstrate that ensemble models, including AdaBoost, Gradient Boosting, Random Forest, and Extra Trees, outperform simple machine learning models in predicting the bending capacity of CFS beams. Moreover, introducing the stacking ensemble technique, using six different base models selectively, resulted in better performance than the individual baseline models. The approach addressed the nonlinearity pattern in the dataset caused by the flexure effect and excessive slenderness. The study suggests that adopting the proposed numerical and machine learning techniques could be a reliable method for predicting the structural behaviour and conducting cost-effective design of CFS beams, compared to the traditional analytical methods. Keywords: cold-formed section, stiffeners, finite element analyses, machine learning, bending Received: 5 April 2024. Accepted: 31 July 2024. Published online: 9 August 2024. Introduction Cold-Formed Steel (CFS) cross sections are used effectively and extensively in the construction industry and many structural applications as secondary load-carrying elements, such as roof purlins sections and transmission line towers. Most commonly, CFS cross sections are C-sections and Z-sections. These sections are manufactured by bending flat sheets with a thickness range from 0.4mm to 6.4mm, according to Eurocode 3 [1] and North America’s typical thickness ranges [2] . CFS has attractive advantages such as ease of installation and prefabrication, high strength-to-weight ratios, and high structural efficiency. Furthermore, the inherent versatility of the manufacturing process facilitates the creation of diverse geometries, thereby offering significant potential for the optimization of CFS sections to align with specific structural objectives. This adaptability holds promise for enhancing manufacturing practices and realizing optimized structural designs. CFS sections typically consist of plate elements with a significant width-to-thickness ratio. As a result, local buckling and distortional buckling are the primary failure modes for cold-formed steel members. These geometric failure modes hinder the efficient utilization of material, which can be resisted with numerous techniques. In plate mechanics, incorporating stiffeners can enhance the section's strength by providing outof-plane support to the flat plate elements. Recently, the structural behavior of CFS stiffened sections has been investigated. For instance, complex edge stiffener, simple lips, perpendicular or inclined to flanges profile, have been used to improve the structural behavior of the channel-section columns against the exp

FAQs

Why is bending capacity prediction for cold-formed steel beams challenging?

The beams can exhibit intricate nonlinear behavior from flexure effects and excessive slenderness, and traditional design methods may not capture these details accurately.

Which machine learning approaches are compared in the study?

Six classical machine learning algorithms and four ensemble models are compared, and a stacking ensemble is additionally used to combine selected base models.

What is the main conclusion about model performance?

Ensemble models such as AdaBoost, Gradient Boosting, Random Forest, and Extra Trees outperform simple models, and the stacking ensemble yields better results than individual baseline models.

Aug 2, 2026

Upload by Sage