View the PDF here

“Scaling-up” Fire Spread on Wood Cribs using CFD

By: Xu Dai[*], a, Naveed Alam b, Chang Liu c, Ali Nadjai b, David Rush c, Stephen Welch *,c 

a School of Engineering, The University of Liverpool, United Kingdom

b School of Built Environment, Ulster University, United Kingdom

c School of Engineering, The University of Edinburgh, United Kingdom

This article is the short version of the published paper: “Scaling-up” Fire Spread on Wood Cribs to Predict a Large-Scale Travelling Fire Test Using CFD.

Dai, X., Alam, N., Nadjai, A., Liu, C., Rush, D., Welch, S. (2024). ““Scaling-up” fire spread on wood cribs to predict a large-scale travelling fire test using CFD”, Advances in Engineering Software, Elsevier Ltd, 189:103589.

Research Background

The compartment fire model stands as a cornerstone in fire engineering, offering insight into fires contained within enclosures. Evolving from this framework, travelling fires have gained prominence as a significant fire scenario in large compartments, spurred on by research over the past decade aimed at furnishing streamlined design tools for performance-based structural fire design [1, 2].

Utilizing simulation-based methods to analyse the fire dynamics of travelling fires presents a promising adjunct to experimental studies. These approaches offer valuable perspectives on the evolving boundary conditions influencing structural responses and ultimately failures. While they enable systematic exploration of parameters independent of experimental uncertainties, the demonstration of suitably comprehensive models has hitherto been lacking.

Here, we explore the potential for “scaling-up” a “stick-by-stick” CFD model [3] which had been carefully calibrated for the case of an isolated crib, of 2.8 m diameter, to a uniformly distributed fuel bed of extent 4.2 × 14.0 m located within an open compartment 9 × 15 m in plan, with internal height 2.9 m [4-6], see Figure 1 from our original paper [7]. Note that the Fire Dynamics Simulator (FDS) 6.7.0 [8] was used as the CFD tool in this study. 

Throughout the “scaling-up” process, all parameters in the CFD model setup [3], such as wood properties and characteristics, remain unchanged, except for adjustments to the plan area of the crib to simulate an extensive uniform fuel bed, and alterations to the geometry of the enclosure  (2.46 m ceiling with 0.35 m down stands above isolated crib and large compartment of plan 9.0 × 15.0 by 2.9 m in height). The method’s credibility is subsequently evaluated by comparing the predictions of the scaled-up CFD model with measurements from large-scale compartment fire tests.


Figure 1. Research method [7].

Evaluated parameters encompass fire spread, burn-away, heat release rate, gas phase temperatures, and incident radiant heat flux on the fuel bed. Successful demonstration of predictive capability in fire modelling suggests the potential for conducting “numerical experiments” in the future to assess the underlying physical mechanisms, which are currently beyond the scope of traditional large-scale onsite experiments.

Research Findings 

The outcomes regarding fire spread and burnout predictions are encouraging, as depicted in Figure 2, while the evolution of heat release rate also aligns well with experimental data. Additionally, there is a relatively close consistency between predicted and measured incident radiant fluxes during fire spread on the wood cribs, as illustrated in Figure 3. Discrepancies in predicted post fire fluxes and gas phase temperatures can be attributed to the wind effects (not modelled) on the fire plume and deficiencies in representing heat transfer from glowing embers, as noted in [7]. These elements are expected to minimally affect the prediction of fire spread on a horizontally-oriented flat fuel bed, which is the primary focus of the present study.

Figure 2. Scaled-up CFD model predicted fire spread comparison with the test, at 20 mins, 40 mins, 60 mins, and 80 mins (full video animation can be accessed via [7]).

Figure 3. Comparison on incident radiant heat fluxes from thin skin calorimeters (TSC) at fuel bed top level centreline along fire trajectory (TSCF-3 failed during test data acquisition after 30 mins) [7].


The fire spread leading edge on the 14.0 m × 4.2 m continuous fuel bed is presented in Figure 4. It demonstrates the fire initially developed as a “ring” like shape with a slower pace due to lower HRR in the initial 20 mins, then gradually accelerated with its leading edge becoming more of a “line” like shape, with the growing HRR, and finally reaching its peak spread rate when the fire reached the fuel bed end at around 70 mins.

Figure 4. Fire spread development with 5 mins intervals, interpreted from the model [7].


Figure 5 illustrates the distribution of incident radiant heat flux every 10 minutes at the 14.0 m × 4.2 m fuel bed level. Additionally, the fire's leading and trailing edges are depicted to aid in understanding the mechanism of fire spread and its correlation with the magnitude of the incident radiative flux. It was observed that the heat flux near the leading edge was approximately 12 kW/m², indicating that sensitivities to various pre-heating trends are relatively minor in this scenario. It is interesting that the critical incident heat flux for the same wood sample, as determined by the cone calorimeter test, was 13.5 kW/m² [3]. The area of incident heat flux ranging from 3 kW/m² to 12 kW/m² ahead of the leading edge gradually expanded during the test due to the increasing fire size and the accumulation of the hot smoke layer, which progressively enhanced radiation for pre-heating (as seen, for instance, from 20 minutes to 50 minutes).

In contrast, the incident heat flux close to the trailing edge was much higher than the heat flux close to the leading edge, exceeding that by an order of magnitude later in the test and reaching very high values of around 150 kW/m2 at 60 mins and 70 mins. This discrepancy is likely due to the "burnt" solid wood sticks in the model continuing to receive incident radiant heat flux not only from the fire plume and smoke layer but also from the surrounding compartment elements, including the heated concrete ceiling (following direct fire plume impingement as the fire moved away) and the fireboard downstands. This highlights a notable deficiency in current travelling fire models [9] which typically overlook the potentially high intensity of incident heat flux from glowing embers. Consequently, it suggests that subsequent structural responses using such travelling fire models are likely to inadequately represent the "localized" cooling effect through numerical simulations [10, 11].


Figure 5. Incident heat flux on the top layer of wood sticks with 10 mins intervals (heat flux unit kW/m2) [7].


The predictions achieved show that the now established “numerical simulator” looks to have good potential as a tool to explore and characterise the behaviour of travelling fires subject to different compartment boundary conditions, including, but not limited to, the effects of: 1) compartment geometry, opening locations and sizes, 2) extent and location of fuel bed, and 3) thermal properties of the boundaries. The equivalent experimental studies of the full range of these parameters would be infeasible, and would also be subject to significant uncertainties, including the effects of varying ambient conditions (temperature and wind) as well as differences in fuel moisture, which are standardised in the simulations.


[1] Dai, X., Welch, S., Usmani, A. (2017). A critical review of ‘travelling fire’ scenarios for performance-based structural engineering, Fire Saf J, 91:568–578. 

[2] Stern-Gottfried, J., Rein, G. (2012). Travelling fires for structural design–Part I: Literature review, Fire Saf J, 54:74-85. 

[3] Dai, X., Gamba, A., Liu, C., Anderson, J., Charlier, M., Rush, D., Welch, S. (2022). An engineering CFD model for fire spread on wood cribs for travelling fires, Advances in Eng Soft, 173:103213. 

[4] Nadjai, A., Alam, N., Charlier, M., Vassart, O., Dai, X., Franssen, J-M., Sjöström, J. (2022). Travelling fire in full scale experimental building subjected to open ventilation conditions, J Structural Fire Eng, 14(2):149-166. 

[5] Nadjai, A., Alam, N., Charlier, M., Vassart, O., Welch, S., Glorieux, A., Sjöström, J. (2022). Large scale fire test: The development of a travelling fire in open ventilation conditions and its influence on the surrounding steel structure, Fire Saf J, 130:103575.  

[6] Alam, N., Nadjai, A., Charlier, M., Vassart, O., Welch, S., Sjöström, J., Dai, X. (2022). Large scale travelling fire tests with open ventilation conditions and their effect on the surrounding steel structure – The second fire test, J Constr Steel Research, 188:107032. 

[7] Dai, X., Alam, N., Nadjai, A., Liu, C., Rush, D., Welch, S. (2024). “Scaling-up” fire spread on wood cribs to predict a large-scale travelling fire test using CFD, Advances in Engineering Software, Elsevier Ltd, 189:103589.

[8] McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Vanella, M., Weinschenk, C., Overholt, K. (2017). Fire Dynamics Simulator, Technical Reference Guide Volume 1: Mathematical Model, NIST Special Publication 1018-1, 6th ed., National Institute of Standards and Technology (NIST), Gaithersburg, Maryland, USA, and VTT Technical Research Centre of Finland, Espoo, Finland.

[9] Dai, X., Welch, S., Vassart, O., Cábová, K., Jiang, L., Maclean, J., Clifton, C., Usmani, A. (2020). An extended travelling fire method framework for performance-based structural design, Fire and Materials, 44.  pp. 437-457, 

[10] Nan, Z., Dai, X., Chen, H., Welch, S., Usmani, A. (2022). A numerical investigation of 3D structural behaviour for steel-composite structures under various travelling fire scenarios, Engineering Structures, 267:114587. 

[11] Jiang, J., Lu, Y.L., Dai, X., Li, G.Q., Chen, W. J., Ye, H. (2021). Disproportionate collapse of steel-framed gravity buildings under travelling fires, Engineering Structures, 245:112799,

[*] Corresponding authors. E-mail address:,