CFD Simulations of Pool Fires in a Confined and Ventilated Enclosure Using the Peatross–Beyler Correlation to Calculate the Mass Loss Rate

The aim of this paper is to analyse the application of the Peatross–Beyler (P&B) correlation (Fire and safety science—Proceedings of the fifth international symposium, 1997) to calculate the Mass loss rate (MLR) for a pool fire in a confined and ventilated enclosure for a range of conditions.The experimental references considered are the PRISME-SI-D1, D2 and D6 tests (Prétrel et al. in 9th International seminar on fire safety in nuclear power plants and installations, 2005), conducted by the Institut de Radioprotection et de Sûreté Nucléaire.The dimensions of the enclosure used in the experiment are 5 m in length, 6 m in width and 4 m in height. A pool fire, \(0.4\,\hbox {m}^2\) hydrogenated tetrapropylene (TPH, \(\hbox {C}_{12}\hbox {H}_{26}\)), is located in the center of the room. The compartment is connected to the outside through an inlet and outlet. The P&B correlation is implemented as a boundary condition in computational fluid dynamics calculations. The MLR outcome depends on the average \(\hbox {O}_2\) concentration in a predefined volume and the characteristics of the fuel. The influence of the size and location of the predefined volume, the ventilation branch position (at 1 m and 3.65 m from the floor) and the Renewal Rate (\(R_r\)) (\(4.7\,\hbox {h}^{-1}\) and \(8.4\,\hbox {h}^{-1}\)) are studied. Two types of \(\hbox {O}_2\) predefined volumes have been tested: layer and ring. The layer volume is located in the low part of the compartment, the base of the volume spanning the compartment floor. The ring is a volume around the pit. It is assumed for both approaches that the measured \(\hbox {O}_2\) is representative of the available oxygen for the flame. The P&B boundary condition predicts the influence of the vitiation on the MLR. Other methods, such as imposing the MLR expected in open condition, overpredict the amount of injected fuel. The results show that the P&B correlation provides good agreement with the experimental data. The deviation between experimental data and numerical prediction for the average MLR in the best case is ?5.0% with absolute values of 0.004 kg/s and 0.0038 kg/s for the experiment and the simulation respectively for case PRS-SI-D1. The MLR calculation is influenced by the position of the ventilation opening and the \(R_r\). The temperature and \(\hbox {O}_2\) concentration profiles are significantly influenced by the ventilation configuration. These differences are related to the transport of the injected air from the inlet to the floor by density difference.     

Fire Whirl Experimental Facility with No Enclosure of Solid Walls: Design and Validation

In this work, a fire whirl facility with no enclosure of solid walls was designed, for experimental simulation of fire whirls in open field. Air curtains were used to produce the generating eddy of fire whirl. Tests were conducted to evaluate the capability of the new facility. It was found that there was an optimal tilt angle for the air curtains to produce stable fire whirls. Experiments of fire whirls under different pool diameters showed that the mass loss rate \( \dot{m} \) depended on the circulation \( \varGamma \) by \( \dot{m}\sim \varGamma^{1.18} \) and the dimensionless flame height satisfied \( H^{*} \sim \varGamma^{*0.71} \) (where the asterisks denote dimensionless variables). It was also found that the ratio of the continuous flame height to the whole flame height for different pool diameters was 0.67. The centerline temperatures varied with the normalized height \( Z \) by \( Z^{0.08} \), \( Z^{ - 1.30} \) and \( Z^{ - 2.18} \), respectively in the regions of the continuous flame, intermittent flame and plume. The radiative fraction was calculated to be close to 44%. All these results fit well with literature data, thereby the capability of the new facility for producing stable fire whirls was fully validated. Some potential applications of the new facility were discussed in detail. Especially, instruments such as infrared camera can be used to investigate the flame radiation of fire whirls by the new facility.     

A Finite Element Model for the Simulation of the UL-94 Burning Test

The tendency of the polymers to melt and drip when they are exposed to external heat source play a very important role in the ignition and the spread of fire. Numerical simulation is a promising methodology for predicting this behaviour. In this paper, a computational procedure that aims at analyzing the combustion, melting and flame spread of polymer is presented. The method models the polymer using a Lagrangian framework adopting the particle finite element method framework while the surrounding air is solved on a fixed Eulerian mesh. This approach allows to treat naturally the polymer shape deformations and to solve the thermo-mechanical problem in a staggered fashion. The problems are coupled using an embedded Dirichlet–Neumann scheme. A simple combustion model and a radiation modeling strategy are included in the air domain. With this strategy the burning of a polypropylene specimen under UL-94 vertical test conditions is simulated. Input parameters for the modelling (density, specific heat, conductivity and viscosity) and results for the validation of the numerical model has been obtained from different literature sources and by IMDEA burning a specimen of dimensions of \(148 \times 13 \times 3.2\,{\mathrm {mm}}^3\). Temperature measurements in the polymer have been recorder by means of three thermocouples exceeding the 1000 K. Simultaneously a digital camera was used to record the burning process. In addition, thermal decomposition of the material (Arrhenius coefficient \({\mathrm {A}}=7.14 \times 10^{16}\,{\mathrm {min}}^{-1}\) and activation energy \({\mathrm {E}}=240.67\,{\mathrm {kJ/mol}}\)) as and changes in viscosity (\(\mu \)) as a function of temperature were obtained. Finally, a good agreement between the experimental and the numerical can be seen in terms of shape of the polymer as well as in the temperature evolution inside the polymer.     

Numerical Investigation of Back-Layering Length and Critical Velocity in Curved Subway Tunnels with Different Turning Radius

Curved tunnels are inevitable subjected to the city underground geological conditions. Due to the catastrophic consequence of tunnel fires with high population density, the related researches on fire safety of curved tunnel are full of significance. Therefore, a series of curved subway tunnels with turning radius of 300–1000 m were investigated numerically by FDS 5.5.3 in terms of the smoke back-layering length and critical ventilation velocity under the heat release rate of 5–10 MW. Theoretical analysis shows that the curved tunnel with the local resistance has an advantage of preventing smoke spreading compared with straight tunnel. The simulation results also indicated that both the smoke back-layering length and the critical ventilation velocity increased with the rising turning radius, and the straight tunnel has the largest values. In fact, the local resistance impact factor for the smoke back-layering length in the curved tunnel, \( k_{f} \), was controlled by turning radius \( R \) and ventilation velocity \( V \). The dimensionless critical velocity increased slightly from \( 0.638Q^{*1/3} \) to \( 0.669Q^{*1/3} \) when the turning radius increased from 300 m to 1000 m. Without considering the influence of turning radius (local resistance), previous models cannot be applied to the curved tunnel. The improved prediction models about smoke back-layering length and critical velocity with the factor of turning radius could provide a technical guideline for the tunnel ventilation designs.     

A Quantitative Infrared Imaging System for In Situ Characterization of Composite Materials in Fire Tests

A novel non-intrusive measurement system based on quantitative infrared imaging has been designed and developed specifically for the study of composite plates submitted to fire. The system consists of two synchronized infrared cameras that image both sides of the sample during a fire test, providing surface temperature maps spatially corregistered. Flame effects on measured temperature are minimized through selection of a spectral band with near negligible infrared absorption-emission (wavelength centre 9585 nm, full width at half maximum 135 nm), as well as software post-processing. An ad hoc experiment has shown that this procedure retrieves surface temperatures with an uncertainty of \(\pm 5\) K, compared to a systematic error larger than 60 K for a classic thermographic measurement. Surface emissivities of both sides of the sample are measured and included in the retrieval procedure. By adding a flash lamp, the system implements an adaptation of the classical Parker’s flash method to thermally thick samples, providing also a map of thermal diffusivities along the sample both before and after the burning. In the region most degraded by fire, the effective thermal diffusivity is reduced approximately one order of magnitude as compared to the pre-test value (from 5.9 × 10^?7 m² s^?1 to 0.5 × 10^?7 m² s^?1). Several composite samples have been analysed while exposed to fire in different conditions, showing that thermal diffusivity after the burning shows a strong correlation with the local maximum temperature reached during the test. More precisely, in the temperature range between \(\sim \)325 and 350\(^{\circ }\)C a drastic change in diffusivity seems to takes place, in a way that suggest a phase change.     

Effect of Wind on Smoldering Combustion Limits of Moist Pine Needle Beds

We studied moist pine needle beds burning under the effect of wind, in order to determine the upper moisture limit for which there is fire propagation for different wind velocities. For this purpose we built a wind tunnel that allowed us to burn a 600 mm by 150 mm by 40 mm bed under wind velocities between 0.5 m/s and 5.0 m/s and controlled air temperature. Results show an increase in moisture limit from 54% to approximately 140%, for the velocity range indicated. Combustion at limiting conditions proceeds mainly by smoldering with some periods of flaming combustion. It was observed that, for conditions close to extinction, the smoldering front is not quenched at the surface. Additionally, it was also observed that a strong flow of hot gases exit from the fuel bed at the free surface. These two observations lead to the conclusion that the main heat sink is moisture evaporation and that heat losses to the surroundings is reduced by the blowing effect of the hot gases coming off the bed. A dimensional analysis suggests a correlation between moisture limit and wind velocity of the form M = A ? B/ \( {v} \) _w ² , where M is moisture limit for fire propagation, A and B are constants, and \( {v} \) _w is wind speed. Two dimensionless numbers helped to plot the smoldering temperature and fire propagation velocity in a more meaningful way. They are \( \Uppi_{1} = {{T_{sml} c_{p,g} } \mathord{\left/ {\vphantom {{T_{sml} c_{p,g} } {v_{w}^{2} }}} \right. \kern-0pt} {v_{w}^{2} }} \) and \( \Uppi_{2} = {{Mh_{fg} } \mathord{\left/ {\vphantom {{Mh_{fg} } {v_{w}^{2} }}} \right. \kern-0pt} {v_{w}^{2} }} \) , where T _sml is smoldering temperature, c _p,g is the gas specific heat, M is fuel moisture content and h _fg is the latent heat of water evaporation. A relatively high moisture limit at 5 m/s wind velocity is possible due to the relatively high air flow into the smoldering front and the efficient heat feedback produced in forward smoldering.     

Failure of Fractional Horsepower Ventilation Fan Motors

Fractional horsepower motors are used to power exhaust fans such as those found in bathrooms and oven hoods. There was an increase in fires attributed to these motors shortly after the year 2000. Still, many argue that these motors cannot cause fires because they contain a thermal cutoff (TCO) unit that shuts off electrical current when the TCO reaches a specific temperature. In this paper we describe an unsafe failure mode that can occur after extended use. As fans age, the bearings wear out, resulting in heating of the coil wires. Excessive heating results in break down of the wire insulation, which can lead to a short between wires and create an autotransformer configuration. An autotransformer can support high current in only part of the wiring, resulting in resistive heating sufficient for ignition. With the low thermal conductivity of the coil (\(0.0162\,W/(\hbox {cm}\,{^{\circ }}\hbox {C})\)) and the poor performance of the TCO (no cutoff up to \(260{^{\circ }}\hbox {C}\)), a large temperature gradient is possible. The gradient allows for ignition in one part of the motor without activating the TCO. We experimentally validated the proposed mechanism in an isolated motor and a complete fan assembly resulting in ignition after 18 s and 23 s. The assembled fan burned for more than 6 minutes before we extinguished the flames for safety reasons. Our results are consistent with field results from actual fires.     

Combination of Kriging methods and multi-fractal analysis for estimating spatial distribution of geotechnical parameters

Spatial variability (randomness, correlation, and singularity) within the geotechnical parameters of complicated geological movements influences the estimation quality that depends on how well mathematical tools can account for variability through limited observations of a spatial field. Classical statistical methods depict randomness well, but cannot account for the problems associated with spatial correlations. Geostatistical methods such as ordinary Kriging (OK), universal Kriging (UK), and co-Kriging (CK) can produce predictions based on spatial auto-correlation and cross-correlation, but are always accompanied by average smoothing effects; a local singularity created by nonlinear geo-processes, therefore, requires special methods to be properly evaluated. In this study, a shallow load-bearing stratum of silt clay (length = 525 m, width = 80 m) at the former 2010 Expo Park in Shanghai was explored by performing 42 borehole laboratory experiments, which provided the key geotechnical parameters: the cohesion coefficient (\( C \), in kPa), the friction angle (\( \varphi \), in ^o), and the compression modulus (\( E_{\text{S}} \), in MPa). First, Kriging methods such as OK, UK, and CK estimated these geotechnical parameters, then a multi-fractal analysis was employed to measure the local singularity. Cross-validation illustrates that multi-fractal analysis has the ability to depict a local anomaly, and further that the auxiliary information utilized in CK improves spatial estimation accuracy.     

Computational modeling of natural ventilation in low-rise non-rectangular floor-plan buildings

Natural ventilation (NV) is a relevant passive strategy for the design of buildings in seek of energy savings and the improvement of the indoor air quality and the thermal comfort. The main aim of this work is to present a comprehensive NV modeling study of a non-rectangular floor-plan dwelling. Given the arbitrary shape of the building, recourse is made to computational fluid dynamics (CFD) to determine the surface-averaged pressure coefficients (\(\overline {{C_p}} \)). The CFD model was calibrated to match experimental data from an extensive wind tunnel database for low-rise buildings. Then, \(\overline {{C_p}} \) computation via CFD is used to feed the building performance simulation software EnergyPlus, in replacement of the built-in Swami and Chandra parametric model that is only valid for estimating \(\overline {{C_p}} \) in rectangular floor-plan buildings. This computational tool is used to investigate the effect of NV on the thermal performance and the airflow rate in a social housing located in the Argentine Littoral region. Simulation results of the considered building show that NV enables to reduce even more than 65% of the cooling degree-hours. Furthermore, regarding to the \(\overline {{C_p}} \) source (either CFD or Swami and Chandra’s), it is also found that this data has a considerable effect on the airflow rates, but a little effect on the thermal performance.     

Study of the Ultimate Load Capacity of K-Type Tube-Gusset Plate Connections

In order to investigate the ultimate load capacity of K-type tube-gusset plate connections with stiffened plate, the static tests of five full-scale specimens were conducted in this study. The results indicate that the end stiffened plate is critical for improving the load capacity of the connections. In addition, the parametric nonlinear finite element analysis of the K-type tube-gusset plate specimens was performed with account of such non-dimensional parameters as chord diameter-to-thickness ratio (γ), plate width-to-chord diameter ratio (α), plate thickness-to-chord thickness ratio (\(\tau_{1}\)), stiffened plate thickness (t_d), and nominal-to-yield stress ratio (η). The above analysis implies that the ultimate load capacity decreases with the increment of γ and increases with the increment of α and \(\tau_{1}\), while it is only slightly affected by the stiffened plate thickness. Compare the results of the finite element analysis with assessment by design guides existing. Based on the former results, an equation for estimating the load capacity of K-type tube-stiffened gusset plate is proposed.     

Radiant Ignition of Polyurethane Foam: The Effect of Sample Size

Although smouldering ignition of upholstery items remains a leading cause of residential fire deaths, relatively little research is conducted on the topic. An experimental investigation of the effect of sample size on the ignition and spread of smouldering and flaming in polyurethane foam under natural flow conditions is reported here. Polyurethane foam samples are used because this is a common material in modern, residential environments and one for which there exists significant quantities of previous experimental data in the literature. Samples of different square cross-section size and a fixed height of 150 mm are insulated on all sides except the top which is exposed to a radiant heat flux and is open to the air. Samples with side lengths of 50 mm, 100 mm, and 140 mm are studied. Ignition and spread dynamics are diagnosed using thirteen thermocouples located along the vertical centre line. The onset of smouldering ignition (13  $\hbox{kW}\, \hbox{m}^{-2}$ , 8  $\hbox{kW}\, \hbox{m}^{-2}$ and $7\,\hbox{kW}\, \hbox{m}^{-2}$ for 50 mm, 100 mm and 140 mm sample sizes respectively) is observed at significantly lower heat fluxes that flaming (45  $\hbox{kW}\,\hbox{m}^{-2}$ , 32  $\hbox{kW}\,\hbox{m}^{-2}$ and $30\,\hbox{kW}\,\hbox{m}^{-2}$ respectively). Critical heat fluxes for smouldering and flaming ignition increase with decreasing sample size, with smouldering ignition being significantly more sensitive to sample size than flaming ignition under the size range studied. Smouldering spread rates are measured in the range from 3  $\hbox{mm}\, \hbox{min}^{-1}$ to $25\,\hbox{mm}\, \hbox{min}^{-1}$ and found to be a strong function of the heat flux and depth of the smoulder front. The effect of sample size on smouldering has been theoretically proposed before but this is the first time that this effect has been demonstrated experimentally for ignition. The fact that large samples result in the lowest critical heat flux could have implications for testing procedures and translation of results from small-scale testing to real-scale in the built environment.     

Engineering Approach for Designing a Thermal Test of Real-Scale Steel Beam Exposed to Localized Fire

This paper reports the design and results of a thermal test on heating of a 6 m long steel W beam subjected to a localized fire conducted at the National Fire Research Laboratory of the National Institute of Standards and Technology. A engineering approach was proposed to determine the heat release rate of the test fire. By the approach, a recently developed simple fire model was first used to approximately calculate the heat release rate and then a sophisticated model was used to check/refine the calculation. The concept of adiabatic surface temperature was used in the sophisticated model to represent the thermal boundary conditions at exposed surfaces in fire. The proposed approach successfully predicted the critical value of heat release rate of 500 kW to reach a target temperature of \(500^{\circ }\hbox {C}\) in the test specimen. A calibration test was also conducted to understand the difference between the predicted and measured steel temperatures in the investigated test, and found that the sophisticated model over-predict the adiabatic surface temperatures which would contribute to the over-prediction of the steel temperatures. The error of the predicted maximum steel temperature in the test specimen was within 10%. The study reported here is not necessarily a validation of the sophisticated model, rather the study provides a successful case study using current knowledge and tools to design realistic and controlled fire tests.     

Floatation extraction of thinly-emulsified tributyl phosphate stabilized by silicic acid

We have studied regularities of floatation extraction of fine-emulsion tributyl phosphate stabilized by silicic acid. Most completely the floatation process occurs at pH 5–6. At the concentration of silicic acid 500–550 mg/dm³ the degree of floatation extraction of tributyl phosphate (α) reaches 70% (floatation time of tributyl phosphate is 10 min). In the presence of HNO₃ (1–5 mol/dm³) the value α passes through the maximum at \(C_{HNO_3 } \)= 3 mol/dm³. A conclusion was made about an important role in the studied floatation process of not only electrical, but also structural-mechanical properties of the surface of the air drops and bubbles.     

The socially optimal and equilibrium locations of two stores or libraries with consumer search

This paper examines the socially optimal (and also equilibrium) locations of two stores or libraries on a linear market of unit length. If each consumer has probability \(w\) of finding a desired product at each store, then the socially optimal locations are never completely centralized for full information, but are completely centralized for \(w \le 0.5\) when costly visit search is necessary. The Nash equilibrium locations of two stores, and various alternative models for the socially optimal locations of two stores, are also examined.     

A note on the null distribution of the local spatial heteroscedasticity (LOSH) statistic

Recently, Ord and Getis (Ann Reg Sci 48:529–539, 2012) developed a local statistic \(H_i\) , called local spatial heteroscedasticity statistic, to identify boundaries of clusters and to describe the nature of heteroscedasticity within clusters. Furthermore, in order to implement the hypothesis testing, Ord and Getis suggested a chi-square approximation method to approximate the null distribution of \(H_i\) , but they said that the validity of the chi-square approximation remains to be investigated and some other approximation methods are still worthy of being developed. Motivated by this suggestion, we propose in this paper a bootstrap procedure to approximate the null distribution of \(H_i\) and conduct some simulation to empirically assess the validity of the bootstrap and chi-square methods. The results demonstrate that the bootstrap method can provide a more accurate approximation than the chi-square method at the cost of more computation time. Moreover, the power of \(H_i\) in identifying boundaries of clusters is empirically examined using the proposed bootstrap method to compute \(p\)  values of the tests, and the multiple comparison issue is also discussed.     

Examination on Fatigue Limit and Crack Growth Characteristic of SBHS700 Base Metal

The object of this study is to clarify fatigue limit and fatigue crack growth characteristic of SBHS700 base metal which is 780 MPa class steel. This study carried out the fatigue tests of SBHS700 base metal containing different defect size, and the fatigue limit was compared with that of the conventional steel. Test results indicate that the fatigue limit increased with decrease initial defect size, and can be precisely evaluated by using \(\sqrt {area}\) parameter model. This paper also presents that crack growth characteristic was almost the same as that of the conventional steel from the observation of striations by using Scanning electron microscope and length of beach marks.     

The neighborhood or the region? Reassessing the density–wage relationship using geocoded data

I analyze the effects of sub-city-level density of economic activity on wages. Using a geocoded dataset on employment and wages in the city areas of Sweden, the analysis is based on squares representing “neighborhoods” ( $0.0625\,\hbox {km}^{2})$ , “districts” ( $1\,\hbox {km}^{2})$ , and “agglomerations” ( $10\,\hbox {km}^{2})$ . The wage-density elasticity depends on spatial resolution, with the elasticity being highest in neighborhood squares, where a doubling of density is associated with wage increases of 1.2 %, or roughly the size of the elasticity for region density. Moving from a mean-density neighborhood to the densest neighborhood would on average increase wages by 9 %. The results are consistent with (i) the existence of a localized density spillover effect and (ii) quite sharp attenuation of human capital spillovers. An implication of the findings is that if the data source is not sufficiently disaggregated, analyses of the density–wage link risk understating the benefits of working in dense parts of regions, such as the central business districts.     

Hydrogeochemie von Eisenausfällungen an Grundwasseraustritten der Abraumkippe Berrenrath (Rheinisches Braunkohlenrevier)

Groundwater at the abandoned lignite mine in Berrenrath (Germany) is acidic and highly mineralised. Average iron concentrations of 588 mg/l indicate a high potential for acidification. Furthermore, investigations revealed a complete depletion of the carbonate buffer capacity (CaCO₃ and $\mathrm{HCO}_{3}^{-}$ ) during chemical groundwater evolution. Over the next few decades, no decrease of dissolved pyrite oxidation products is expected. On the contrary, within the drainage ditches at the edges of the dump, an intense precipitation of iron oxides is occurring. However, despite oxygen saturation, precipitation of iron oxides is incomplete due to strong acidification of the surface water (pH<3.0). The main contamination of the iron sludge is caused by high nickel concentrations. This element is only loosely bound to the iron oxides and is easily mobilised. Because of high nickel- and sulphate-concentrations, the excavated material has to be deposited in a controlled landfill (German dump category of 2 to 3).     

The Use of Fractional Factorial Design for Atrium Fires Prediction

In order to mitigate the excessive computational cost of atrium fire simulations, a novel methodology based on the use of the Fractional Factorial Design technique to obtain an experimental validated tool, in the form of a surface response model, capable to predict fire induced conditions is proposed. This methodology is supported by results from a Design of Experiments benchmark, which consists of a set of FDS simulations in the present work. Specifically, a \(2^{6-2}_{IV}\) approach has been considered and applied to a 20 m cubic atrium. Thus, six factors have been considered, namely the fire Heat Release Rate (HRR) and location, the exhaust flow rate, the exhaust location and activation time, and the inlet vents area. Furthermore, the smoke temperature at the roof and 15 m high and the smoke layer height have been considered the variables of interest. Subsequently, a multiple linear regression analysis has been performed to predict and compare the steady and non-steady temperature profiles and the smoke layer drop with six novel full-scale atrium fire tests, and also with specific adjusted FDS models. In addition, this methodology has been extended successfully to predict the non-steady behaviour of the fire tests. At the steady state, the HRR and the exhaust flow rate have been found to be the most relevant factors. The results obtained with the proposed methodology show a good fit both with the fire tests and with the adjusted FDS models, with discrepancies mostly below 14%. For non-steady conditions, a time analysis of the influence of the six factors has been carried out. Again, remarkable good agreement with the time-dependent experimental results is achieved, with average discrepancies below 12%, being the larger differences found in the prediction of local effects, such as the smoke ceiling jet, for high HRR or when the make-up air influence is significant. The results turn this methodology into a powerful and useful tool for fire safety designs.     

Experiments and Numerical Simulations of Pressure Effects in Apartment Fires

The fire induced pressure and its influence on ventilation flows within a compartment have not been studied in detail previously. In this research work, we have investigated the development of gas pressure and the resulting flows in compartment fires first experimentally, by burning a series of heptane pool and polyurethane mattress fires inside a real, 58.6 m\(^2\) by 2.57 m high, apartment and then by carrying out numerical simulations of the experiments with the FDS code. The experiments were conducted with three different ventilation duct configurations to simulate three different airtightness conditions. The peak heat release rates were less than 1 MW and the burning times were about 180 s. The experimental results indicate that the gas pressure in relatively closed apartment can become high enough to revert the flows of the ventilation system, prevent escape through inwards-opening doors, and even break some structures. The peak gas temperatures under the ceiling of the burn room were about 300°C. The pool fires remained well-ventilated. The pressure ranges encountered in the experiments were between 100 Pa to 1650 Pa and the pressure occured within 50 s of ignition. We also report the FDS validation for this type of simulations and discuss the process of modelling the ventilation system and leakages.