Fire Resistance of Steel Beams with Intumescent Coating Exposed to Fire Using ANSYS and Machine Learning
The thermal conductivity of steel is high compared to other materials such as concrete or timber. Therefore, fire protection measures are applied to prolong the duration between the onset of fire exposure and the final loss of load-bearing function of a steel structure. The most common passive fire...
שמור ב:
| Main Authors: | , , |
|---|---|
| פורמט: | Article |
| שפה: | אנגלית |
| יצא לאור: |
MDPI AG
2025-07-01
|
| סדרה: | Buildings |
| נושאים: | |
| גישה מקוונת: | https://www.mdpi.com/2075-5309/15/13/2334 |
| תגים: |
הוספת תג
אין תגיות, היה/י הראשונ/ה לתייג את הרשומה!
|
| סיכום: | The thermal conductivity of steel is high compared to other materials such as concrete or timber. Therefore, fire protection measures are applied to prolong the duration between the onset of fire exposure and the final loss of load-bearing function of a steel structure. The most common passive fire protection measure is the application of intumescent coating (IC), a thin film that expands at elevated temperatures and forms an insulating char layer of lower thermal conductivity. This paper focuses on structural steel beams with IPE open-section profiles protected by a water-based IC and subjected to static and standard fire loading. ANSYS 16.0 is used to simulate heat transfer, with thermal conductivity function described by standard multivariate linear regression analysis, followed by mechanical analysis considering degradation of material mechanical properties at elevated temperatures. Simulations are conducted for all IPE profile sizes, with varying initial degrees of utilisation, beam lengths, and coating thicknesses. Results indicated fire resistance times ranging from 24 to 53.5 min, demonstrating a relatively good level of fire resistance even with the minimal IC thickness. Furthermore, artificial neural networks were developed to predict the fire resistance time of steel members with IC using varying numbers of hidden neurons and subset ratios. The model achieved a predictability level of 99.9% upon evaluation. |
|---|---|
| ISSN: | 2075-5309 |