Work

Explained machine learning models for the fire resistance of trusses made of L, T and X sections

Year: 2025

Type: article

Abstract: Trusses very often consist of monosymmetric and built‐up compressed members, which may fail in torsional or flexural‐torsional buckling, potentially leading to the collapse of structures. Previous studies have demonstrated that these phenomena may be relevant in case of fire and are not adequately considered in design standards, and improved design equations were proposed to increase the safety of these members at elevated temperatures. As a further step, machine learning (ML) models are developed in this study, providing accurate yet fast calculation models for compressed members belonging to truss systems. This work describes the development of Artificial neural networks, Random Forests and Support Vector Regression, building upon a numerical dataset of the compressive capacity of members with angles, tee or cruciform sections and different lengths and sectional dimensions. The machine learning models are explained using a combination of partial dependence plots and SHapley Additive exPlanations. The accuracy versus safety trade‐off is discussed for a better‐informed model selection. It is shown that ML models are more accurate than the analytical method.

Source: ce/papers

DOI: 10.1002/cepa.70085

Authors: Luca Possidente, Carlos Couto

Institutions: University College London, University of Aveiro

Language: English

Topic: Structural Load-Bearing Analysis

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $3,140

Funder: Fundação para a Ciência e a Tecnologia