In this video Stephen describes how the complexity of fire development undermines the predictive capabilities of even the most advanced simulation models.
 
 
Background:

My research focuses around the development of computer simulations for modelling fires, so as to ultimately predict their impact on structures and human occupants. These are important tools to facilitate the further development and appropriate application of “performance-based” methods for fire safety engineering – which allow more flexibility in design and indeed may be the only suitable approaches for demonstrating safety objectives can be met in more complex scenarios.

I’m interested in all aspects of the fire problem, including fire growth and development, toxic species generation and thermo-mechanical response of relevant structures. My work advocates innovative and generalised approaches to modelling which reflect the complexities of the real fire scenarios, where there are often great uncertainties in the conditions, i.e. what items are burning, the nature of the combustibles, the ventilations conditions (with hazardous conditions resulting from “under-ventilated” combustion in enclosed spaces), the fire protection materials, etc all of which can have a drastic effect on fire development and structural response – not to mention the interaction with human beings, both those fleeing the fire “egress” and emergency responder intervening to suppress it.

Models for fire, whether simple zonal approaches or based on computational fluid dynamics (CFD), have traditionally been restricted to study of smoke movement. Though useful in an assumed steady phase of a fire, considering the complexities of fire growth we expect deterministic models to be very poor at representing the early phase of a fire. The challenge is all the greater when we factor in the requirement of determining what the structure or the human occupants may do, and how that may interact with the fire (e.g. when doors are opened for egress or fire service access, or glazing falls out changing the ventilation conditions, etc).

But the modern built environment is sensor rich, and we have a vision that sensor measurements of fire conditions can be exploited to steer fire simulations and provide emergency responders with an overview of possible hazard evolution. This novel approach is named FireGrid, it is a vision we are pursuing in multi-disciplinary collaboration involving computation via the grid using high performance computing (HPC), wireless sensor technologies and interpretation by knowledge-based reasoning to provide decision support information which meets the needs of emergency responders. The technology has been demonstrated in live demonstration in full-scale fire tests. Further work is needed but we hope ultimately to contribute to the development of safer environments.

Find out more:

School of Engineering profile: http://www.eng.ed.ac.uk/about/people/dr-stephen-welch

Edinburgh Research Explorer: http://www.research.ed.ac.uk/portal/swelch