Predicting The Capacity of Cold-Formed Steel Hollow Sections under Elevated Temperatures Using Deep Learning | ||||
| Port-Said Engineering Research Journal | ||||
| Articles in Press, Accepted Manuscript, Available Online from 19 June 2025 | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/pserj.2025.360999.1394 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Mosub Gamal Lawash  1; Ashraf El-Sabbagh1; Mohamed Elghandour2; Tarek Sharaf3
| ||||
| 1Department of engineering, port-said university, Port-said, Egypt | ||||
| 2Faculty of Engineering, Civil Engineering Department, Port Said University | ||||
| 3Faculty of Engineering, Port Said University | ||||
| Abstract | ||||
| Because of the complex interaction between local and global instability modes at high temperatures, predicting the strength of thin-walled columns in fire situations is challenging. This study explores the use of machine learning to assess the fire resistance of steel hollow sections with rectangular and square shapes. Initially, a reliable finite element model is employed to assess column behavior, generating a comprehensive dataset that encompasses a variety of cross-sections, slenderness ratios, and temperatures. This dataset forms the backbone for training and evaluating machine learning models, specifically Deep Neural Network (DNN), Extreme Gradient Boosting XGBoost, and Support Vector Regression (SVR). Among these, the DNN model showcased pronounced accuracy, particularly when tested on previously unseen data; notably, its application for predicting the buckling capacity of these specific hollow sections under elevated temperatures represents a pioneering approach not previously investigated. This deep learning methodology offers a significant advantage by drastically reducing the computational effort required to determine buckling capacities compared to traditional, time-intensive finite element modelling. An in-depth error analysis further elucidated the origins of prediction discrepancies, underscoring the necessity of recognizing the model's boundaries. | ||||
| Keywords | ||||
| Cold-formed steel column; Axial capacity; Elevated temperature; Hollow section; Deep Learning | ||||
|
Statistics Article View: 86 |
||||
