City Research Online

Abstract

This study presents a machine learning framework for predicting the lateral-distortional buckling resistance of steel-concrete composite beams with optimised elliptically-based web openings. Current design codes, such as EN 1993-1-13, do not address perforated beams with vertical elliptical web openings while there is a growing demand for vertical ellipses as they overperform in several failure modes. To overcome this limitation, a high-fidelity numerical database comprising 360 finite element simulations was developed, incorporating material and geometric nonlinearities and validated against experimental results. An artificial neural network combined with particle swarm optimization was trained using eleven geometric and mechanical input parameters to estimate the lateral-distortional buckling resistance. The model achieved high accuracy and outperformed standard code-based predictions. To enhance engineering interpretability, SHapley Additive exPlanations were used to identify the most influential parameters. A user-accessible web-based tool was developed for practical applications. A reliability analysis based on Annex D of EN 1990 (2002) was performed to evaluate the proposed design method and recommend a partial safety factor for the lateral–distortional buckling resistance of steel-concrete composite beams with elliptically-based web openings. The proposed framework provides a robust and accurate alternative for evaluating the lateraldistortional buckling resistance of composite perforated beams.

Publication Type:	Article
Publisher Keywords:	steel-concrete composite beams; elliptically-based web openings; lateral distortional buckling; machine learning
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TA Engineering (General). Civil engineering (General) T Technology > TH Building construction
Departments:	School of Science & Technology School of Science & Technology > Department of Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads