Measuring and modelling fire behaviour in timber compartments using Artificial Intelligence
File(s)
Author(s)
Amin, Rikesh
Type
Thesis
Abstract
The building construction industry contributes 40% of global carbon emissions, emphasising the need for sustainable materials like mass timber over steel and concrete. Mass timber offers a higher strength-to-weight ratio, aesthetic appeal, lower cost, and greater sustainability. However, timber’s inherent flammability poses fire hazards that require different fire protection measures. Current methods to understand fire behaviour to inform fire engineering are expensive and time-consuming. With an increase in high-rise mass timber buildings, there is a widening research gap in understanding compartment fires in such structures. This thesis addresses this, offering a historical review of how artificial intelligence (AI) is used alongside traditional and advanced computational tools to support engineering design. The thesis introduces the CodeRed compartment fire experiment series covering a 352 m2 floor area, with timber ceilings and columns. The research develops data-driven models to predict timber charring rates, outperforming Eurocode-5. External flaming is also examined, where limited research exists on flame shape and size. A novel AI diagnostic tool was developed to instantly measure flames under crosswind conditions in both outdoor and laboratory environments. Results showed that timber impacts fire dynamics. The fully exposed cross-laminated timber (CLT) ceiling produced an average external flame height of 2.7 m compared to 1.4 m for the half-encapsulated ceiling. Ventilation also critically influences the duration and height of external flaming. Measurements from the diagnostic tool were used to calculate heat release rates (HRR) of external flames in CodeRed. Fully exposed CLT reached a peak of 43 MW with an average of 10 MW, contributing 36% to the compartment’s total HRR. By integrating AI with fire science, this thesis advances understanding of timber as a construction material and its fire behaviour and demonstrates AI’s role in fire safety. Ultimately, the findings contribute to fire engineering and a more sustainable built environment.
Date Issued
2024-10-01
Date Awarded
01/03/2025
Advisor
Rein, Guillermo
Sponsor
Engineering and Physical Sciences Research Council
Grant Number
EP/V519534/1/2434442
Publisher Department
Department of Mechanical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)