Predicting post-earthquake fire behaviour of steel moment frames using different classification models
نویسندگان
1 Associate Professor, Faculty of Engineering, University of Mohaghegh Ardabili, Ardabil, Iran
2 Faculty Member, Department of Civil Engineering, Germi Branch, Islamic Azad University, Germi, Iran
3 Professor, Faculty of Engineering, University of Mohaghegh Ardabili, Ardabil, Iran
doi
10.61882/NMCE.2512.1112چکیده
Post-earthquake fire (PEF), often triggered by damage to electrical and gas systems, poses a severe threat to steel structures, especially when seismic damage disables fire protection systems. Accurately predicting structural behavior under PEF is critical for emergency response. This study investigates the behavior of steel Moment-Resisting Frames (MRFs) under PEF through a two-phase approach. In the first phase, finite element (FE) simulations of 4- and 8-story intermediate and special MRFs were conducted under various fire scenarios (single-bay, two-bay, and entire-floor) following different seismic intensity levels (without earthquake, with design base earthquake and with maximum considered earthquake). Then the structural response of different models was investigated and compared in terms of failure modes (local, partial, global) and fire resistance times, revealing significant insights into the combined effects of building height, frame type, fire location, and prior seismic damage. In the second phase, the study employs Machine Learning (ML) techniques to predict the stability of steel MRFs under PEF. Using the dataset generated from FE simulations, five common ML models were trained and tested. The results demonstrate that ML techniques are highly effective tools for predicting both the failure mode and collapse time of steel MRFs subjected to PEF, providing a rapid assessment methodology that complements detailed numerical analysis for emergency response scenarios.
