Characterising natural fires in large compartments – revisiting an early travelling fire test (BST/FRS 1993) with CFD

Xu Dai, Stephen Welch, David Rush, Marion Charlier, Johan Anderson

Research output: Contribution to conferencePaperpeer-review


This paper presents a careful assessment of fire conditions in a long enclosure, open only at one end, which contained a regularly spaced timber crib fire load and is dominated by under-ventilated combustion. Though the geometrical arrangement, with fully enclosed side walls, differs from many more recent “travelling fire” tests, the essential fire behaviour shows a very clear progression of the main burning zone, driven by a combination of fire spread, ventilation and fuel burn out. By contrast many other travelling fire tests have been designed to be fuel-controlled, in much more open structures. Moreover, due to the enclosed nature of the test, it was observed to result in very high temperatures in the protected and unprotected steel beam members, especially during the phase when the fire travelled back from the opening to the ignition location at the rear of the compartment. The investigated test in this paper is Test number 2, which is one of nine tests carried out at the BRE Cardington laboratory in the UK, led by British Steel Technical (BST) and hosted by the Fire Research Station (FRS).

CFD modelling, with NIST’s Fire Dynamics Simulator (FDS), is invoked to assist in exploration and interpretation of the test results, and to evaluate model capabilities for this complex fire scenario. For simplicity, instead of modelling the complex pyrolysis and combustion of timber on a stick-by-stick basis, the entire wood cribs are represented in the FDS model as single objects. Thus ignition and burning are represented in a simplified manner at the scale of the individual crib faces, with idealised mass loss curves derived from measurements in selected rows. Though tied to empirical mass loss, the model is stretched in representing heavily under-ventilated conditions occurring in certain domains. Despite these simplifications and challenges the model does prove capable of representing the qualitative behaviours to a satisfactory level. In particular, with appropriate parameter choice for reaction-to-fire properties, the FDS models are able to represent the fire spread rates for the two distinct stages of the fire: the fire travelling from rear to the front of the compartment in search of oxygen; and the fire travelling back from the opening to the rear as fuel is consumed. The results suggest the potential value of such simplified representations of cribs in FDS for travelling fire scenarios.

In the calibrated model, the magnitude of the thermocouple (TC) temperatures at three locations (ignition location, centre location, and opening location) shows generally good agreement between the FDS model and the test at the stage when the fire travels to the opening. However, significantly higher discrepancy appears at the stage when the fire travels back. In addition to the highly simplified representation of crib burning, another likely reason for the latter may be the malfunction of the mass loss measurement in several wood cribs at this stage, which led to significant uncertainties in prescribed burning rates. Another large source of uncertainty relates to residual heat in compartment boundaries and fuel embers, the latter not being easy to represent in CFD models. These results suggest avenues to explore in future model development.
Original languageEnglish
Publication statusPublished - 3 Jul 2019
Event15th International Conference and Exhibition on Fire Science & Engineering
- Royal Holloway College, London, United Kingdom
Duration: 1 Jul 20193 Jul 2019


Conference15th International Conference and Exhibition on Fire Science & Engineering
Abbreviated titleInterflam 2019
Country/TerritoryUnited Kingdom
Internet address


Dive into the research topics of 'Characterising natural fires in large compartments – revisiting an early travelling fire test (BST/FRS 1993) with CFD'. Together they form a unique fingerprint.

Cite this