Fire test

6 February 2019

Computer simulation has been used to model the impact of fire on a transport package. Alex Bond, Darrell Egarr, Lucy Horton and Yang Sui explain how a CFD model compares with simpler analysis approaches.

AS A PART OF CANADA’S long-term plan to move all used nuclear fuel to a central underground repository, the Nuclear Waste Management Organisation (NWMO) is developing a new package designated the basket transportation package (BTP).

The BTP is to be designated as a Type B (U) package and must be certified by the Canadian Nuclear Safety Commission (CNSC) as meeting its Packaging and Transport of Nuclear Substances Regulations (PTNSR).1 The PTNSR heavily references the International Atomic Energy Agency’s (IAEA) Regulations for the Safe Transport of Radioactive Material, 2012 Edition.2 These regulations require the robustness of the package to be demonstrated through a number of tests that simulate bounding conditions in severe transportation accidents, including impact, thermal and pressure categories.

The thermal test subjects the package to a hypothetical, fully engulfing hydrocarbon fire, with a nominal free stream/ambient fire temperature of 800°C for 30 minutes, following which the package is allowed to cool in quiescent ambient conditions.

In computer simulations of the IAEA scenario, the fire is generally simulated implicitly, using an incident heat flux based on the Stefan-Boltzmann black-body radiation law, with incident flux proportional to the fourth power of temperature differential and a uniform flame temperature of 800°C. However, this is a significant simplification of real-world test conditions, and therefore an investigation was undertaken to explicitly simulate the pool fire to allow a comparison of the results against other approaches. This was undertaken in two phases as follows:

1. Undertake a validation and verification case for the combustion model.

2. Deploy the combustion model on BTP transportation package.

Model geometry

The validation case that was used was a half-scale fire test of the road cask for 10-year irradiated fuel (see ‘Fire Testing Of Half-Scale Road Cask For 10 Year Cooled Irradiated Fuel’, Ontario Hydro Report No 88-127-K)3. The model geometry is presented in Figure 1. The cask was placed on a support stand at the centre of a pool of kerosene contained within a shallow pan.

The cask and cask lid are made of stainless steel (304L). The impact limiter is made of redwood encased in stainless steel sheet metal. In the model, the stainless-steel skin was ignored, as the thermal resistance of the thin steel skin is low relative to the redwood core. Material properties for stainless steel were taken from BS EN 1993 4 and for Redwood from the Wood Handbook 5. As the impact limiter is heated by the fire, it is likely that the redwood would char, resulting in a change in thermal properties. With time the depth of this char increases, gradually insulating the contents from any external heat flux (the fire). This effect was not included in the model. A second effect relating to char is soot forming on the surface which will result in a high thermal emissivity (increasing the rate of radiative heating of the package); this effect is taken into account. This is a conservative assumption, and is used because of the uncertainty over the rate of charring, and the thermal properties associated with varying degrees of char at the surface. As the actual Basket Transfer Flask (the new design) does not have the wooden impact limiter, it was considered unnecessary to model the wood burn in explicit detail.

One of the corners of the impact limiter was damaged during drop tests and this was included by simply slicing the corner off, rather than allowing for changes in material properties. Additionally, the geometry was simplified, and the seals, vent and drain were not resolved. In the real fire test, there were steel structures inside the cask which would add additional thermal mass and inertia, which were not included in the CFD model.

The support was made of steel posts filled with concrete. These were wrapped with a Fiberfrax blanket to reduce heat transfer into the support. Therefore, in the model the support was modelled as adiabatic, and heat transfer between the stand and box was not included in the calculation.

Figure 2 presents the geometry of the BTP. Under Normal Conditions of Transport (NCT) for the initial condition, it was assumed the flask was vertical. However, it was assumed that the flask horizontal during the fire and cooldown (following an accident). An impact to the flask was likely to knock it into a horizontal orientation. Sensitivity to the position of the flask was not undertaken.

The fuel baskets were modelled as three homogenous regions, with a density of 4169kg/m3; the total thermal output was 675W (225W per basket) and temperature- dependent properties were used for the thermal conductivity and specific heat capacity. The surface emissivity of the external surfaces of the BTP was set to 0.9, as in the validation case.

The diameter of the pool was scaled from the validation case to give the same ratio between flask length and pool diameter, giving a pool diameter of 7m. The edge of the pool was 2.4-2.7m from the external surfaces of the flask, compliant with the condition2 that the edge of the pool should be between 1m and 3m from the external surfaces of the specimen. The height of the cask above the fuel surface was also scaled from the validation study, to try and ensure that the cask is fully engulfed in the fire. The distance between the fuel surface and the underside of the cask body was set to 1.5m.

Boundary conditions

For the validation case, the initial temperature of the flask was assumed to be a uniform 60°C, for consistency with the experimental conditions, as the flask was initially pre-heated3. A review of the temperature data shows that in practice the initial temperature was not uniform, but the effect of this discrepancy is expected to be small.

For the BTP, the initial temperature distribution was determined from a steady-state calculation using conservative solar radiation flux values from IAEA guidance.6 As the solar insolation values are specified as heat flux values, perfect absorptivity for solar radiation is assumed. The surface emissivity on internal and external surfaces was taken to be 0.21.

The ambient air temperature in the physical test was assumed to be 10°C, whereas, for the BTP assessment, an ambient air temperature of 38°C was used, in accordance with IAEA guidance.6 The air cavity within the flask was represented as a conducting solid, with temperature- dependent properties. Radiation and convection were not solved in this region. During the fire, it was assumed that there was no wind, which is consistent with the IAEA guidance that “wind speeds of less than about 2m/s should not detract from the test”.

Fire calculation

During the test, kerosene was supplied to the pan so that the pool did not burn dry, although the supply rate of fuel is not known. 

The burning rate (and hence vaporisation rate) was estimated from correlations in the literature. However, this resulted in flame heights that were too high when compared to reference 3. Additionally, the temperature of the fuel is not known, and is likely to change with time, due to heat transfer from the fire. For the validation case, a value of 10°C (based on ambient temperature) was chosen. Sensitivity tests to the fuel temperature did not result in a noticeable change in flame temperature. The black body temperature at the pool for radiation was set to 800°C. The thermal properties of kerosene and the properties of the products of combustion (carbon dioxide, carbon monoxide, water vapour, and soot) were the default values in Fluent v17.1.

The approach taken for the vaporisation rate was to iterate the vaporisation rate until the iso-surface of 800°C engulfed the flask, consistent with IAEA guidance.4 The resulting vaporisation rate, used in the validation case analysis, was 0.16kg/s. This is a little lower than the theoretical calculated value of 0.26kg/s using the methodology of Zabetakis and Burgess7 but is still considered representative of a hydrocarbon pool fire. For the BTP assessment, the burning rate used was 1.5kg/s which was equal to the rate calculated following reference 7 with constants from Babrauskas8. The evaporation rate from the pool was set to a constant across the pool surface, so the local concentrations of oxygen and fuel may not be representative, and this may cause differences in the reaction zone and hence radiative fluxes.

Cool-down phase

During the cool-down period, heat transfer from the flask was modelled using algebraic expressions for radiation and natural convection9. Temperature-dependent heat transfer coefficients were specified.

Ambient air temperature was taken as noted above, and in the case of the BTP flask solar insolation was included as defined in BS EN 19934. The simulation time after the fire was around 6 hours and therefore insolation was constant during cooldown. The surface emissivity of the external surfaces of the BTP was kept at 0.9 for charred surfaces, as it was during the fire and the absorptivity for solar radiation was taken to be1.

Physical models

Initial condition

The standard k-ε turbulence model was selected based on experience that this model was expected to give reasonable results.

The eddy dissipation combustion model was used as it is known to be a robust model in predicting turbulent reacting flows. The products of combustion were based on the yields given for well-ventilated fires for kerosene.10 However, due to the diameter of the pool fire (3m), it is plausible that the fire is not ‘well ventilated’, particularly towards the centre of the fire, where entrainment and mixing of the fuel with air may be reduced relative to the outside of the pool. Thus, it was possible that the yield of soot may be higher than for a well-ventilated condition. As the yield for soot is an unknown quantity, and significantly influences the absorption coefficient for radiation and radiative heat flux, the yield was increased by a factor of three compared to that reported for well-ventilated fires. (The soot yield used in the assessment was 0.126kg/kg). The yield of soot, however, was still considered to be within physical bounds based on a calculation for the soot yield under poorly ventilated conditions, again using data presented in reference 10.

The Discrete Ordinates (DO) and Weighted Sum of Gray Gas Model was used. The DO model is valid throughout the entire range of optical thicknesses and was therefore considered to be an appropriate model. The model included the effect of radiation from soot by augmenting the absorption coefficients with the empirical correlation by Sazhin11. 


Validation case

Figure 3 presents contours of the gas temperature during the fire and the extent of the 800°C surface, demonstrating that the cask is fully engulfed above 800°C, and hence consistent with the IAEA scenario.

Figure 4 presents temperatures of the thermocouples calculated by the CFD model, compared to those measured during the fire test. The thermocouples that were used for the comparison were those on the bottom and middle parts of the main steel cask. Results from the bottom of the flask are presented. In most cases, the temperatures compare reasonably well.

These comparison points were selected as the BTP flask does not include the impact limiter. Due to potential charring of the wood in the impact limiter and hence changes in material properties, in addition to the resolution of the damage to the lid, it was felt that there was more uncertainty in the model with respect to the temperatures measured near the impact limiter.

Based on judgement and experience with similar CFD analysis, acceptance criteria of 15% for internal temperatures and 30% for external temperatures were used. Peak temperatures were compared and the maximum difference for internal surfaces was 14%. For two of the external thermocouples (59 and 64), the difference in peak temperature was greater than 30%. Of the remaining external thermocouples, the maximum difference in peak temperatures was 27% at thermocouple 62.

The temperature at the centre-bottom (thermocouple 64) was lower in the CFD results, possibly due to the combustion model, which calculates relatively lower gas temperatures below the cask. This could be due to this region being slightly depleted of oxygen as it is at the centre of the pool. The measured temperature of thermocouple 59 was significantly higher than the other thermocouple temperatures. This could be due to a slight crosswind during the physical fire test, which is not replicated in the computer simulation.

BTP Assessment

Figure 5 presents contours of the gas temperature during the fire and the extent of the 800°C surface, which fully engulfs the BTP. Compared to the validation study, temperatures in the vicinity of the cask, particularly underneath, are up to 200°C higher. This could be due to scaling for the larger cask, resulting in a higher fuel mass flow and therefore a hotter fire. Flame temperatures adjacent to the package are in excess of 1200°C. The radiative heat flux impinging on the package, determined from the CFD analysis, is 1548kW, whereas the equivalent radiation heat flux determined using the Stephan Boltzmann equation for an engulfing flame temperature of 800°C is 937kW.


Explicit modelling of the combustion process is complex, with numerous physical processes occurring. When using empirical correlations in literature for the burning rate of fuel, it was found that the flame height was over- predicted. As the burning rate was not measured, there is uncertainty whether this is due to the accuracy of the CFD model, or practical difficulties in measuring the flame height. Therefore, in order to achieve representative results in a validation case, the fuel supply rate and the soot content of the products of combustion had to be tuned to achieve a thermal response representative of the half scale flask.

However, once it had been tuned, the combustion model gave a good correlation between the physical test and the analytical findings.

Results inside the package were within 15% of the measured values, while external results were generally within 30% of the measured values. These discrepancies were partially due to variations introduced in the real world testing.

The tuned model was then successfully used to assess the BTP. It was noted that the flame temperature calculated is locally much higher than 800°C (>1200°C) and the simulation might therefore be overly conservative. It is considered that this may be due to the burning rate of fuel not scaling linearly as the pool size increases, and the soot content may not be constant when scaling the pool fire due to variation in completeness of combustion.

Without further empirical data, the accuracy of the model when applied to a full-scale flask is uncertain, although the heat fluxes that have been calculated are conservative relative to the IAEA guidance.

At the current time, it is concluded that the use of CFD simulation for accident conditions of transport should adhere to the IAEA guidance.  

Author information: Alex Bond, Principal consultant, MMI-Thornton Tomasetti; Darrell Egarr, CFD specialist at Dyson – formerly senior consultant MMI-Thornton Tomasetti; Lucy Horton, Research engineer at Dyson – formerly project engineer at MMI-Thornton Tomasetti; Yang Sui, Design engineer, used fuel transportation package at NWMO 


  1. Canadian Nuclear Safety Commission (CNSC),“Packaging and Transport of Nuclear Substances Regulations (PTNSR),” 2015.
  2. International Atomic Energy Agency (IAEA), “Advisory Material for the IAEA Regulations for the Safe Transport of Radioactive Material (2012 Edition).”, No. SSG-26, Vienna, Austria, 2014.
  3. “Fire Testing Of Half-Scale Road Cask For 10 Year Cooled Irradiated Fuel”, Ontario Hydro Report No 88-127-K.
  4. BS EN 1993 Part 1-2 , 2005, Eurocode 3: Design of steel structures - Part 1-2: General rules - Structural fire design.
  5. “Wood Handbook”, FPL-GTR-190: Forest Products Laboratory, 2010.
  6. International Atomic Energy Agency (IAEA), “Advisory Material for the IAEA Regulations for the Safe Transport of Radioactive Material (2012 Edition).”, No. SSG-26, Vienna, Austria, 2014.
  7. Zabetakis, M. G. and Burgess, D. S. (1961). “Research on the hazards associated with the production and handling of liquid hydrogen”. US Bureau of Mines RI5707, Pittsburgh, PA.
  8. Babrauskas, V. (1983). “Estimating large pool fire burning rates”. Fire Technology, 19, 251-261.
  9. Bejan, A., Kraus, A.D., Heat Transfer Textbook, Wiley.
  10. SPFE Handbook of Fire Protection Engineering, 4th Edition, SFPE & NFPA.
  11. Sazhin, S.S., Sazhina, E.M., (1996), “The effective-emissivity approximation for the thermal radiation transfer problem”, Fuel, Vol 75, No. 14, pp 1646-1654. 

Figure 3. Gas temperature contours (above) and 800°C contour (right) for the validation model
Figure 1. Geometry for half-scale cask model
Figure 2. Geometry for Basket Transportation Package model
Figure 4. Half scale validation model – temperatures at inner bottom thermocouples
Figure 4. Half scale validation model – temperatures at inner bottom thermocouples
Figure 5. Gas temperature contours (left) and 800°C contour (right) for the full scale BTP

Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.