Fire Australia 2017 Fire Safety Engineering Stream ... · accuracy of the calculation results would be of the ... and the laws of buoyancy and applied them to floating and ... the

Page 1 of 24

Fire Australia 2017

Fire Safety Engineering Stream - Quantification of Fire Safety

AS/NZS 1668.1:2015 – IS IT GOOD ENOUGH FOR PERFORMANCE-BASED SMOKE EXHAUST SYSTEM DESIGN

M.C. HUI and M. KHEDHER

BCA Logic Pty Ltd

Level 6, 210 Clarence Street, Sydney 2000

ABSTRACT

This paper examines the shortcoming of the methodology to determine smoke exhaust rates as specified in section 9 ‘Hot layer smoke control systems’ of AS/NZS 1668.1-2015 by referring to the fundamental principles in various plume theories currently used by fire safety engineering practitioners locally and overseas, and by carrying out numerical experiments to investigate the performance of a mechanical smoke extraction system that operates at the exhaust rates determined by using the methodology in the joint Australian/New Zealand Standard for four design fires with a steady state heat release rate of 1.5, 5, 10 and 15MW respectively.

1 INTRODUCTION

In Australia, smoke exhaust systems are commonly used to delay the onset of untenable conditions in performance based fire safety designs. There are two options to determine the smoke extraction rate of a smoke exhaust system, that is, either to follow the Deemed-to-Satisfy Provisions as stipulated in Specification E2.2b of the National Construction Code (Volume One – Building Code of Australia Class 2 to Class 9 buildings) or to calculate the exhaust rate from first principles as a performance-based design. Calculations from first principles could be in the form of hand calculations based on empirical formulae in fire engineering literature, by using zone models, or by using field models.

The recently published Australian Standard AS/NZS 1668.1:2015 introduces a new section that sets out design requirements for hot layer smoke control systems. The performance criteria of the systems are that on initiation of fire mode, the hot layer smoke control systems shall operate to exhaust smoke from the hot smoke layer at a rate as determined in accordance with Clause 9.4.3 of the Standard.

Clause 9.4.3 states that the required exhaust airflow rate shall be determined by:

• using Figure 9.2(a) for general cases, which is identical to Figure 2 in Specification E2.2b of the Building Code of Australia (BCA), or

• using Figure 9.2(b) for atrium which is similar to Figure 3.4 in Specification G3.8 of the BCA, but the smoke exhaust rate required in the Australian Standard is consistently lower than that in the BCA, or

• using the equation, EAC = 2.6864Q + 0.1513 FP SLH1.5 (Equation 1)

where EAC = exhaust air capacity (m³/s)

Q = fire size (MW)

FP = fire perimeter (m)

SLH = height to underside of smoke layer (m)

Page 2 of 24

The exhaust rate equation in AS/NZS 1668.1:2015 with four decimal places suggest that the

accuracy of the calculation results would be of the order of 0.0001m³/s but there is no information on how the equation was derived nor reference to any engineering literature regarding the source of the equation. There is also no indication in the Standard on which type of fire plumes to which the equation is applicable. Currently the type of fire plumes can be categorised as follows:

• Point source axisymmetric plumes

• Finite area source axisymmetric plumes

• Wall plumes

• Corner plumes

• Line plumes

• Window plumes

• Balcony spill plumes

Furthermore, the equation does not provide any information on or method to estimate the temperature of the hot smoke layer, thus the designer would not be able to determine the likely time of failure of the smoke exhaust fans as well as the likelihood of flashover in the enclosure of fire origin.

In addition, Clause 9.9.2 of the Standard prescribes the make-up airflow rates (in m³/s) which are the exhaust rates determined from Equation 1 minus 4, 13, 27 and 40m³/s for the associated design fires with steady state heat release rates of 1.5, 5, 10 and 15MW respectively, irrespective of the type of fire plumes and smoke layer temperatures.

This paper provides a review of the fundamental principles in various plume theories currently used by fire safety engineering practitioners (locally and overseas), and reports the results of a number of numerical experiments. The experiments were carried out by using zone modelling and field modelling techniques to determine the performance of a smoke extraction system that operates at the smoke exhaust rates determined from Equation 1 for the four design fires mentioned above.

2 PLUME THEORIES

Observations and studies on fluid flow could date back to ancient times because such subject was closely related to daily lives such as irrigation and sailing. Bahrami (2011) stated that Archimedes (285 – 212 B.C.) postulated the parallelogram law for addition of vectors and the laws of buoyancy and applied them to floating and submerged objects.

In regard to the studies of fire plumes which formed the foundation of a number of plume theories, attribution may be made to the work of Yih (1950), Rouse et al. (1952), Morton et al. (1956) and Yokoi (1960).

The properties of fire plumes are important in the field of fire safety engineering as such properties relate to fire heating of building structures, smoke filling rates, fire venting and the design of fire detection and fire suppression systems.

Of particular note is that the mass flow at a particular elevation in a fire plume is almost completely attributable to air entrained by the plume at lower elevations (Heskestad 2016).

2.1 Large fire plume (UK)

As a result of a few large fires around the late 1950’s and early 1960’s in single-storey industrial buildings that could not be sub-divided into smaller fire compartments due to operational requirements, Thomas et al. (1963) investigated theoretically and experimentally the use of roof vents to exhaust the hot gases produced by a fire in a large single-storey

Page 3 of 24

building. In particular, a ‘large fire’ equation was derived from simple assumptions of entrainment, constant and upward velocity in the flame zone, based on the works of others such as Rouse et al. (1952) and Simms et al. (1960), and took the following form:

M = 0.096P(H – d)3/2o(gs/o)½ (Equation 2)

where

M is the mass rate of entrainment of air

P is the perimeter of the fire

H is the ceiling height measured above the fire source

d is the depth of the hot smoke layer under the ceiling

o is the density of ambient air

s is the mean density of the hot smoke layer

g is the gravitational constant

Equation 1 was based on the assumption that Af > ½(H – d) and hf > (H – d) where Af is the area of the fire and hf is the height of flames above floor level.

Assuming the smoke layer temperature is that of the flame as 1500°F (816°C) used in (Thomas et al. 1963), Equation 2 in SI units presented in modern fire engineering literature such as (CIBSE 2010) takes the following form:

M = 0.188Pz3/2 (kg/s) (Equation 3)

where z is the distance between the bottom of the hot smoke layer and the top of the fire bed.

2.2 Large fire plume (US)

The prediction of smoke mass flow rate in NFPA 92 (2012) for ‘large fire’ similar to that described in section 2.1 above follows the following procedures.

Firstly the limiting flame height needs to be estimated by the following equation:

zl = 0.166Qc2/5

where

zl is the limiting flame height (m)

Qc is the convective portion of the heat release rate (kW)

Secondly if the distance between the base of the fire and the smoke layer interface (Note: this is an input parameter for smoke exhaust system designs) is less than or equal to the

limiting flame height, i.e. z zl, then the mass flow rate in the plume at the height z can be determined by:

M = 0.032Qc3/5z (kg/s) (Equation 4)

The form of Equation 4 is significantly different from that reported by Heskestad (1982, 2016) where the mass flow rate in the plume at the height z, for fire source diameters of 0.3m and greater, is predicted by:

M = 0.0058Qc(z/zl) (Equation 5)

However, investigation of the differences between the two equations is beyond the scope of this paper, and the NFPA 92 (2012) equation, i.e. equation 4, has been adopted for this study.

Page 4 of 24

2.3 Point source axisymmetric plume (strong plume)

For small fires that can be approximated to a point source, Thomas et al. (1963) developed the following equation based on the work of Yokoi (1960) and Yih (1952) for the mass rate of entrainment of air into the fire plume.

M = 0.153c[Qcg/(cCpTo)]⅓z5/3 (Equation 6)

where

M is the mass rate of entrainment of air

c is the mean density of the hot smoke layer

o is the density of ambient air

Qc is the convective heat release rate of the fire

g is the gravitational constant

Cp is the specific heat capacity of air

To is the ambient temperature

z is the distance between the effective point source and the bottom of the smoke layer, i.e. z = zo + h - ds

where zo is the distance between the effective point source and the floor (in modern text this is referred to the point of virtual origin), h is the ceiling height, and ds is the depth of the smoke layer.

The significance of equation 6 is that the mass entrainment rate is much more influenced by the vertical distance available for entrainment (power of 5/3) than the heat release rate (power of 1/3).

The modern form of Equation 6 in SI units was developed by Heskastad (1982) and included in NFPA92B (1991) since its first edition as follows.

M = 0.071Qc⅓z5/3 + 0.0018Qc when z > zl (Equation 7)

where zl = 0.166Qc2/5

3 NUMERICAL EXPERIMENTS

In order to compare the predictive capabilities of the equations in AS/NZS 1668.1:2015 for smoke exhaust rate and make-up air rate with the equations for the plume theories discussed above, numerical simulations using zone and field modelling techniques have been carried out for a hypothetical building that has nominal dimensions of 30 m x 30 m x 9 m (1) and is served by a mechanical smoke extraction system.

Such building could be seen as a representative of an atrium or sports halls and multi-function spaces for holding a variety of events.

In the numerical experiments, the design fire sizes follow that in the BCA Specification E2.2b, i.e. having a steady state heat release rate (HRR) of 1.5, 5, 10 and 15 MW, assuming that the fire bed is located at the floor level and the convective heat release rate is 70% of the total heat release rate.

The HRRs of the four design fires have been assumed to grow at ‘fast’ t-squared growth rate for the unsteady state period until peak HRR is reached. A t-squared ‘fast’ fire would reach 1055 kW within 150 s after it started. This assumption is in line with the recommended

1 The height of the building in the field model numerical analysis could be about 2.2% higher than 9 m due to mesh cell

size configuration. However, such increase of building height does not affect the comparisons that are based on the conditions at the steady state of the design fire.

Page 5 of 24

growth rate for retail occupancies and in stadia as specified in Table 7-2 of Engineers Australia practice note for design fires published in 2012 by the Society of Fire Safety.

The fire safety design aims to maintain a smoke layer interface height of 6.5m so that the estimated smoke layer temperature will hopefully be below the temperatures (500°C to 600°C) that are typically considered as an indication that flashover is to occur, noting that there are criteria and have no physical content but are of considerable practical usefulness (Thomas 1995). This is because the zone and field models selected for the numerical experiments were not developed for fully developed fires.

Using equation 1 and clause 9.9.2 of AS/NZS 1668.1:2015, the smoke exhaust rates and make-up air flow rates for the four design fires selected have been determined and shown in Table 1 below.

Table 1: Smoke exhaust & make-up air flow rates determined in accordance with AS/NZS 1668.1:2015

Heat release rate (HRR) (MW) 1.5 5 10 15

Smoke exhaust rate (m³ s-1) 19.1 43.5 67.0 90.4

Make-up air flow rate (m³ s-1) 15.1 30.5 40.0 50.4

The smoke exhaust rates outlined in Table 1 above have been adopted in the numerical experiments, with the exhaust being provided via four smoke exhaust outlets that are evenly distributed in the enclosure to avoid plugholing.

In the simulations, it was assumed that the smoke exhaust fans would start after activation of a smoke detector in the building. The detector activation time was calculated by using the temperature equivalence approach (Heskestad 1981) which assumes the smoke detector to be a sensitive heat detector that would actuate at 13ºC above ambient temperature (IFEG 2005). It is recognised that there are more sophisticated methods to estimate the actuation time of smoke detectors by other means (e.g. Geiman and Gottuk 2006, Mowrer et al. 2008). However, as this study is on comparison of smoke exhaust rates after the design fires have reached their steady state, the actual time of smoke detector actuation is immaterial for the study.

Once a fan has started on activation of the smoke detector, it was assumed that it would take 30 seconds to ramp up to its 100% capacity. The activation time of the smoke detector was estimated to be 120 seconds for the four design fires. Thus, the smoke exhaust fans have been specified in the simulations to have zero flow rate at 120 seconds and 100% flow rate at 150 seconds with a linear ramp-up function.

Make-up air is provided via four openings with each one located at the centre of each external wall, instead of using a mechanically assisted make-up air system. The opening areas for make-up air for each of the four design fires were specified to be 2.4 m x 1.6 m, 5.2 m x 1.6 m, 4.2 m x 1.8 m, and 6 m x 6 m respectively, based on an exhaust area to make-up air area ratio of 2 for design fires 10 and 15 MW and a ratio of 3 for design fires 1.5 and 5 MW. Narayanan (1993) suggests that the make-up air openings must be located no higher than the lower edge of the smoke reservoir and the NFPA 92B (1991) also specifies that the make-up air opening should be located below the smoke layer interface height. In the field modelling study, the location of make-up air opening was specified to be one mesh cell size (0.6 m) above the floor level for design fires of 10 & 15 MW, and two mesh cell sizes (0.8 m) above the floor level for design fires of 1.5 & 5 MW.

As a comparison, the smoke exhaust rates predicted by the three plume theories outlined above have been calculated by using equations 3, 4 and 7. As these equations provide estimates of the mass flow rates rather than volumetric flow rates, the temperature of the smoke layer is required to be estimated for determining the density of smoke. Such estimate is typically based on the conservation of energy principle and assuming adiabatic boundary conditions; the average temperature of the smoke layer can thus be given by:

Page 6 of 24

Ts – To = Qc/(MCp) (Equation 8)

where

Ts is the temperature of the smoke layer (°C or K)

To is the ambient temperature (°C or K)

The smoke exhaust rates required to maintain a smoke layer interface height of 6.5m as predicted by equations 3, 4 and 7 in conjunction with equation 8 are summarised in Table 2, and the predicted smoke layer temperatures are provided in Table 3.

Table 2: Smoke exhaust rates and make-up air flow rates determined in accordance with equations 3, 4, 7 and 8 in this paper for smoke layer interface height of 6.5m

Total HRR (MW) 1.5 5 10 15

Smoke exhaust rate (m³ s-1) by equation 3 (UK large fire plume)

18.6 41.1 61.4 81.8

Smoke exhaust rate (m³ s-1) by equation 4 (US large fire plume)

17.9 39.5 58.9 78.4

Smoke exhaust rate (m³ s-1) by equation 7 (point source axisymmetric plume)

18.5 36.2 57.0 76.2

Smoke exhaust rate (m³ s-1) by equation 1 (AS/NZS 1668.1:2015)

19.1 43.5 67.0 90.4

Table 3: Predicted smoke layer temperature for smoke layer interface height of 6.5m

Total HRR (MW) 1.5 5 10 15

Smoke layer temperature (°C) by equation 3 (UK large fire plume)

37 76 132 189

Smoke layer temperature (°C) by equation 4 (US large fire plume)

98 146 186 215

Smoke layer temperature (°C) by equation 7 (point source axisymmetric plume)

78 134 181 215

3.1 Zone Modelling

The smoke layer interface heights in the hypothetical building for the four design fire sizes at the steady state condition have been estimated by using the zone model CFAST Version 7.2.1 (Peacock et al. 2016b) developed by NIST.

Whilst the initial release of the CFAST model in the HAZARD I package (Peacock et al. 1991) had a limitation in scope to one- and two- family residential structures, the model has been applied by various researchers to other types of buildings such as Duong (1990) and Chow (1994).

Recently data from a number of experiments, including that in two large hangars with dimensions of 97.8m x 73.8m x 15m high and 45.7m x 73.8m x 22m high, were used to compare with CFAST predictions on smoke layer temperature and interface height for verification and validation (V&V) purposes (Peacock et al. 2016a). The conclusions drawn from the V&V exercise were that the use of the model is appropriate over a range of conditions for a variety of fire scenarios and although differences between the model and the experiments were evident in these studies, most differences can be explained by limitations of the model and the experiments.

Moreover, a large fire in a building would create a two-zone environment, as evident in the FCRC Project 6 shopping centre fire tests (Bennetts et al. 1988) conducted in the “Burn Hall”

Page 7 of 24

at the Scientific Services Laboratories in Port Melbourne which has a ceiling height of 9m and a volume of 12,500m³, and observed by one of the authors of this paper personally during one of the tests.

The key assumptions in the input data include adiabatic boundary conditions, 30% radiative fraction for the design fire, and a limiting oxygen concentration of 15%.

3.2 Field Modelling

3.2.1 Introduction

Fire Dynamics Simulator (FDS), a computational fluid dynamics (CFD) model of fire-driven fluid flow, was used to estimate the smoke layer interface heights by using the smoke exhaust rates determined from the equations in AS/NZS 1668.1:2015 for the four design fires.

The FDS software is the product of a collaborative effort led by the National Institute of Standards and Technology (NIST) of the United States Department of Commerce and VTT Technical Research Centre of Finland. FDS solves numerically a form of Navier-Stokes equations appropriate for low-speed (Ma < 0.3) thermally-driven flow with an emphasis on smoke and heat transport from fire (McGrattan et al. 2016). In this study, the latest version of FDS, Version 6.5.3, available to the authors at the time of preparation of this paper has been used.

Smokeview is a visualisation program that can be used to read and view graphically the output of FDS (Forney 2016).

Typical output measurements of FDS for the gas phase include gas temperature, gas velocity, gas density, smoke concentration and visibility, pressure and heat flux per unit volume. For the solid phase FDS predicts surface and interior temperature, heat flux of both radiative and convective and burning rate.

FDS requires the user to specify multiple parameters such as computational domain (building geometry, mesh and cell size), boundary conditions, initial conditions, and characteristics of the design fire (chemical composition of fuel, mass fraction of species, heat of combustion, and radiative fraction). An overview of the computational domain and key FDS user input parameters are provided below:

3.2.2 Overview of simulation environment

Front and top 3D views of the hypothetical building are shown in Figure 1. Figure 2 contains two 2D views of the building; (a) is a top view illustrating the location of fire and smoke exhaust outlets in the roof, and (b) is a side view showing make-up air opening in the external wall and location and of measuring devices within the building. Figure 3 shows 3D views of the building; (a) is a side view showing measuring devices, make-up air openings, exhaust outlets, smoke layer interface devices, and location of fire, and (b) is a side view showing the configuration of the multiple meshes.

Page 8 of 24

1(a)

1(b)

Figure 1: Simulated building; (a) 3D front view: outline of building, fire location, and computational domain extension (blue lines), (b) 3D top view of building (solid)

2(a)

2(b)

Figure 2: (a) 2D top view showing smoke exhaust outlets on the roof, make-up air inlets in external walls, fire at the centre of the building, and domain extension (outer white square) (b) 2D side view showing make-up air

opening in the external wall, and measuring devices in the building

3(a)

3(b)

Figure 3: 3D view of simulated building; (a) 3D side view showing measuring devices, roof vents and make-up air openings, and (b) 3D top view showing multiple meshes in the computational domain

Page 9 of 24

3.2.3 FDS user input parameters

3.2.3.1 Design fires

The heat release rate (HRR) is an important parameter of the design fire and an FDS user input, which is usually specified as a function of time (Engineers Australia 2012).

Table 4 lists the characteristics of the four design fires. In each simulation the fire was placed at the centre of the building and was specified to be located at one cell above the floor level.

The simulation period for each run was specified to finish approximately 10 minutes after the design fire HRR has reached the peak heat release rate in order to allow for the smoke layer, temperature and other conditions within the building to stabilise.

The chemical composition, heat of combustion and yield of species of the burning fuel was assumed to be 50% wood and 50% polyurethane (flexible) foam which is the average of the associated data for GM21, GM23, GM25 and GM27 outlined in Table 3-4.16 and Table 3-4.18 of the SFPE Handbook 4th Edition. The user defined soot yield is one-third (0.034 g/g) of the estimated value of 0.101 g/g because it has been reported that FDS over-predicts smoke concentration by a factor of 2 to 5 times the measured values, as the dynamic aerosol behaviours of soot, i.e. deposition, settling and agglomeration, have not yet been accounted for by the model (Gottuk et al. 2009).

Table 4: Characteristics of selected design fires

Design fires

Total heat

release rate

(MW)

Fire perimeter (AS/NZS

1668.1:2015) (m)

Fire bed area (m2)

Fire intensity (kW/m2)

Characteristic fire diameter

D* (m)

Time to reach peak

HRR (s)

FDS simulation duration

(s)

1 1.5 6 2.25 667 1.13 179 800

2 5 12 9 556 1.83 327 950

3 10 16 16 625 2.41 462 1080

4 15 20 25 600 2.83 566 1180

3.2.3.2 Computation domain, meshes and cell sizes

Mesh resolution is another important parameter in field modelling that uses large eddy simulation techniques. FDS uses second-order accurate approximations of both the temporal and spatial derivatives of the Navier-Stokes equations, meaning that the discretization error is proportional to the square of the time step or cell size. In theory, reducing the grid cell size by a factor of two reduces the discretization error by a factor of four, however, it also increases the computing time by a factor of 16 (McGrattan et al. 2016).

With any grid resolution study, a point of diminishing returns is reached when the improvement in the quality of the results outweighed by the cost of the computation. The US Nuclear Regulatory Commission (2007) study identified that refining the gird cell size beyond

the ratio of characteristic fire diameter to computational cell size (D*/x) of 4 to 16 did not make a noticeable difference in simulation results. For fire applications, the accuracy of the model depends on the characteristic fire diameter D*, which is defined as (Zhou et al. 2010):

D* = [Q / (o Cp To g0.5)]2/5

where:

D* = characteristics fire diameter (m)

Q = total heat release rate (kW)

Page 10 of 24

o = density of air at ambient temperature (kg m-3)

Cp = specific heat (kJ kg-1 K-1)

To = temperature of air (K)

g = gravity acceleration (9.81 m s-2)

Petterson (2002) cited in (Zhou et al. 2010) suggested that the cell size near the fire should be 0.05-0.1D*, and further away from the fire location the largest cell size in the computational domain should not exceed 0.5D*.

The computational domain in this study was divided into five meshes in order to specify a finer grid for the fire mesh and a coarser grid for areas further away from the fire, and to allow for a reduced CPU time through utilisation of the multi-core processor (Intel i7-4770K, 3.50GHz) on the PC’s. The boundary of the computational domain extends from -1.2 m to 37.2 m in the x-axis, -1.2 m to 37.2 m in the y-axis, and 0 m to 9.6 m in the z-axis. The building geometry extends from 2.4 m to 33.6 m in the x-axis, 2.4 m to 33.6 m in the y-axis and 0 m to 9.0 m in the z-axis.

In the grid sensitivity study, all input parameters remained the same except the fire region mesh cell size. Table 5 details the mesh cell sizes that have been analysed.

Note: for ease of reference to the simulation runs throughout this paper, the naming of simulation ID was based on the heat release rate and the fire region cell size modelled, e.g. for design fire of 1.5 MW and a fire mesh cell size of 200 mm the simulation ID is referred to as 1.5-200.

Table 5: Sensitivity analysis mesh sizes

Fire total heat

release rate (MW)

Fire region

grid cell size (cm)

Bounding meshes grid size

(cm)

Simulation run

scenario ID

Total smoke

exhaust outlet

area (m2)

Total make-up air inlet

area (m2)

Total number of cells

CPU time

(hours)

1.5

0.1 x 0.1 x 0.1

0.4 x 0.4 x 0.4

1.5-100 5.12

(1.6 m x 0.8 m for each of 4 outlets)

15.36 (2.4 m x 1.6 m for each of 4

inlets)

587,288 79.9

0.2 x 0.2 x 0.2

0.4 x 0.4 x 0.4

1.5-200 257,560 10.7

5

0.1 x 0.1 x 0.1

0.4 x 0.4 x 0.4

5-100 10.24

(1.6 m x 1.6 m for each of 4 outlets)

33.28 (5.2 m x 1.6 m for each of 4

inlets)

795,800 125.1

0.2 x 0.2 x 0.2

0.4 x 0.4 x 0.4

5-200 283,800 23.5

10

0.15 x 0.15 x 0.15

0.6 x 0.6 x 0.6

10-150 14.4

(3 m x 1.2 m for each of 4 outlets)

30.24 (4.2 m x 1.8 m for each of 4

inlets)

292,336 40.3

0.3 x 0.3 x 0.3

0.6 x 0.6 x 0.6

10-300 90,736 18.4

15

0.15 x 0.15 x 0.15

0.6 x 0.6 x 0.6

15-150 23.04

(2.4 m x 2.4 m for each of 4 outlets)

43.2 (6 m x

1.8 m for each of 4

inlets)

356,848 92.3

0.3 x 0.3 x 0.3

0.6 x 0.6 x 0.6

15-300 97,904 6.3

Apart from grid independence, there are other “tricks of the trade” items (Hui and Mendham 2015) such as computational domain extension that could significantly and adversely influence the simulation results. Zhang et al. (2010) studied the effects of computational

Page 11 of 24

domain on numerical simulation of building fire results and identified the need to extend the domain boundary beyond the physical boundaries of vertical vents opening of the enclosure in order to take into account the interactions of inflow and outflow properly. The study compared experimental data with the numerical results of a domain with and without extended boundary and analysed domain extension in x, y and z directions under different heat release rates. It was reported in Zhang et al. (2010) that the difference in results diminishes as the domain is extended gradually. The extension in the perpendicular directions to the opening was identified to have the most effect on the results. The (Zhang et al. 2010) study also indicated the results of the extended computational domain yielded better agreement with experimental data.

The computational domain for the FDS simulations presented in this paper was thus extended by 3.6 m, i.e. one hydraulic diameter of the make-up air inlet opening, according to the findings in Zhang et al. (2010), beyond the building physical geometry in the positive and negative x-axis and y-axis. The x-axis and y-axis boundary planes were specified to be open vents to eliminate the effect of physical / domain boundaries on the results.

3.2.4 FDS Output parameters

Table 6 details the output parameters recorded and the location of the measuring devices. The output parameters write intervals were specified to be 10 s for devices, Plot3D and slice files. Knowledge of visibility and smoke layer interface is required so that the exhaust capacity modelled could be related to the smoke layer interface.

Table 6: Output parameters collected

Output parameters Locations of measuring devices

Velocity (m s-1) Velocity measuring device positioned at the centreline of each exhaust outlet and make-up air opening

Temperature (oC)

Temperature measuring devices positioned at the centre of the fire and at the centre of each exhaust outlet. These devices are placed between the roof and 1.8 m above floor level.

Slices positioned to cross through the centre of the fire, exhaust outlets and make-up air openings in the x-z plane and y-z plane.

Visibility (m) Slices positioned to cross through the centre of the fire, exhaust outlets and make-up air openings in x-z plane and y-z plane

Smoke layer interface height above floor level (m)

Four ‘Layer Height’ devices for measuring the layer height positioned to measure the smoke layer interface between the floor and roof

Volumetric flow rate (m3 s-1) Volumetric flow measuring devices for measuring of the flow through each exhaust outlet and make-up air opening

4 SIMULATION RESULTS

4.1 Zone modelling

The modelled heat release rates, predicted smoke layer interface heights and temperatures determined by CFAST 7.2.1 are shown in Figures 4 - 6 respectively.

Figure 4 indicates that the heat release rates in the simulations followed the prescribed design fire values.

Figure 5 presents an anomaly that the smoke layer interface heights for design fires of 5, 10 and 15MW could be maintained at about 8.7m from floor level with the smoke exhaust rates determined from AS/NZS 1668.1:2015, which is not what is expected. For the design fire of

Page 12 of 24

1.5MW, the simulation indicates that the smoke exhaust rate determined from the joint Standard is not adequate to maintain the smoke layer interface height at 6.5m.

Figure 6 depicts the predicted smoke layer temperatures which at steady states are higher than the temperatures predicted by all three plume theories presented in section 3.

Figure 4: Modelled Design Fire Heat Release Rates

Figure 5: Predicted Smoke Layer Interface Heights

Figure 6: Predicted Smoke Layer Temperatures

0

2000

4000

6000

8000

10000

12000

14000

16000

0 200 400 600 800 1000 1200 1400 1600 1800

HR

R (

kW)

Time (s)

HRR-1.5 HRR-5

HRR-10 HRR-15

5.0

5.5

6.0

6.5

7.0

7.5

8.0

8.5

9.0

0 200 400 600 800 1000 1200 1400 1600 1800

Hei

ght

(m)

Time (s)

HGT-1.5 HGT-5

HGT-10 HGT-15

Design

0

50

100

150

200

250

300

0 200 400 600 800 1000 1200 1400 1600 1800

Tem

per

atu

re (

°C)

Time (s)

TEMP-1.5 TEMP-5

TEMP-10 TEMP-15

Page 13 of 24

4.2 Field modelling

4.2.1 Grid independence study

A grid independence study aimed to determine a mesh cell size that will produce accurate results without inefficient use of resources and unnecessary increase of computational time was performed.

Figure 7 shows the FDS predicted smoke layer interface heights and upper layer temperatures as a result of varying the fire region mesh cell size for each design fire. The deviations between the output data are tabulated in Table 7. It should be noted that the computation of percentage deviations was based on the output data after the fire has reached steady state.

(a) 1.5 MW & 5 MW– temperature

(b) 10 MW & 15 MW – temperature

(c) 1.5 MW & 5 MW – layer interface height

(d) 5 MW & 15 MW – layer interface height

Figure 7: Grid independence study results

The results presented in Table 7 show that a coarser mesh for the fire region for design fires of 5, 10 and 15 MW predicted a higher upper layer temperature and a lower smoke layer interface height than that with a finer mesh. For the 1.5MW design fire, the difference in the results between the coarser grid and the finer grid is minimal with respect to upper layer temperature and smoke layer interface height.

The CPU time with a coarser mesh for the fire region was approximately 6.5, 4.3, 1.2 and 13.7 times shorter than the CPU time with a finer mesh for the 1.5, 5, 10 and 15 MW design fires respectively (Table 5). As the FDS results with a coarser mesh in the fire region were found to be more conservative, the results of simulations with a coarser mesh in the fire region have been adopted for the following analyses.

Page 14 of 24

Table 7: FDS output parameters percentage deviation

Fire total heat

release rate

(MW)

Fire region mesh cell

sizes

Percentage deviation

HRR Average make-up

air velocity

Smoke layer interface height

Upper layer temperature

1.5 100 mm vs

200 mm 3.82 0.66

1.60 (minimal difference)

0.22 (minimal difference)

5 100 mm vs

200 mm 3.31 1.46

7.16 (coarser mesh shows lower layer

height)

3.74 (coarser mesh shows higher temperature)

10 150 mm vs

300 mm 2.38 0.59

9.32 (coarser mesh shows lower layer

height)

0.32 (minimal difference)

15 150 mm vs

300 mm 3.87 2.80

9.97 (coarser mesh shows lower layer

height)

4.74 (coarser mesh shows higher temperature)

4.2.2 Simulation results

Selected relevant results predicted by FDS for the four design fires are discussed within this section.

4.2.2.1 Heat release rate (HRR)

Figure 8 shows that steady state HRRs were reached as specified and maintained for the specified remaining duration of the simulations which means the fires were well ventilated.

Figure 8: FDS simulated HRRs of the four design fires

4.2.2.2 Velocity profile through exhaust outlets and make-up air openings

Velocity-measuring devices were placed at the centre of each exhaust outlet and make-up air opening. The FDS predicated velocities at each exhaust outlet were averaged to calculate the exhaust velocity. The same method was used to calculate the averaged make-up air velocity.

Figure 9 and 10 show the averaged smoke exhaust velocities and make-up air velocities for the four design fires respectively, and steady state conditions were achieved after the heat release rates of the design fires reached their respective peak value.

Page 15 of 24

(a) 1.5 MW

(b) 5 MW

(c) 10 MW

(d) 15 MW

Figure 9: Averaged velocity at the smoke exhaust outlets for simulated design fires; (a) 1.5 MW, (b) 5 MW, (c) 10 MW, and (d) 15 MW

Figure 10: Averaged make-up air velocity for simulated design fires

4.2.2.3 Volumetric flow rate

Figure 11 shows the predicted smoke exhaust and make-up air volumetric flow rates for the four design fires. The data shows the make-up air volumetric flow rate predicted by FDS is approximately 5 to 10% more than that determined from the AS/NZS 1668.1:2015 equation. This means the Australian Standard equation underestimates the quantity of make-up air required. However, this underestimate is not considered to be detrimental in this study as natural make-up air is provided.

Page 16 of 24

However, the velocity of make-up air at the perimeter of large open space needs to be limited to sufficient low value so as to not deflect the fire plume significantly, increase the fire plume entrainment rate, and disturb the smoke layer interface (NFPA 92 2012). The NFPA 92 (2012) recommends a velocity of approximately 1 m/s based on the flame deflection data (Mudan and Croce 1988).

(a) 1.5MW

(b) 5.0MW

(c) 10MW

(d) 15MW

Figure 11: Simulated smoke exhaust and make-up air volumetric flow rates for design fires of 1.5, 5, 10 and 15MW by FDS

4.2.2.4 Smoke layer interface height and visibility

Four ‘Layer Height’ devices positioned at 5.1m from the centre of the fire along the positive and negative x-axis as well as y-axis were specified to measure the smoke layer interface.

Figure 12 shows the height of the smoke layer interface for each design fire which is the mean value of the layer heights predicted by FDS from the four ‘Layer height’ devices. Figure 12 illustrates that the smoke layer interface stabilised at 4.57 m, 4.31 m, 3.77 m and 3.31 m during the steady state period of the 1.5, 5, 10 and 15 MW design fires respectively.

The formation of a constant smoke layer interface suggests that the mass of smoke exhausted equalled the mass of air entrained into the plume.

The results in Figure 12 show the smoke exhaust rates determined in accordance with AS/NZS 1668.1:2015 are not sufficient to maintain the design smoke layer interface height of 6.5m.

Page 17 of 24

Figure 12: Smoke layer interface height predicted by FDS

Figure 13 and Figure 14 show the y-z plane images of visibility after steady state conditions were reached. These images show that the depth of the smoke layer decreases with an increase in HRR. The visibility within the lower section of the building for the duration of the simulation remained above 10 m which is considered a suitable tenability limit for large enclosures to assist occupants escape (Engineers Australia 2014).

(a) 1.5 MW at 700 s

(b) 5 MW at 850 s

Figure 13: Smoke layer interface heights for 1.5 & 5 MW design fires

(a) 10 MW at 1000 s

(b) 15 MW at 1100 s

Figure 14: Smoke layer interface heights for 10 & 15 MW design fires

Page 18 of 24

4.2.2.5 Temperature profile

Five thermocouple trees were included in the simulations to measure the air temperatures; one each at the centreline of each exhaust outlet and one at the centreline of the fire. Each tree has seven thermocouples that were positioned between 1.6 m above the floor level and the roof for design fires 1.5 MW and 5 MW, and between 1.8 m above the floor level and the roof for design fires 10 MW and 15 MW. The vertical distance between each thermocouple was set to be 1.2 m. The four temperatures measured (excluding the thermocouples at the centreline of the fire) at each of the seven elevations were averaged to calculate the mean air temperature at that corresponding elevation.

Figure 15 shows the averaged air temperature at each elevation for the modelled scenarios. The results at 9 m for design fires 10 MW and 15 MW being 20ºC were discarded because the devices snapped to the obstruction which is the roof and recorded the temperature of the object rather than the temperature of the gas.

(a) 1.5 MW

(b) 5.0 MW

(c) 10 MW

(d) 15 MW

Figure 15: Averaged temperature measurements at multiple elevations within the building for design fires: (a) 1.5 MW, (b) 5 MW (c) 10 MW and (d) 15 MW

5 DISCUSSIONS

The zone modelling results as shown in Figure 5 indicate that the smoke exhaust rate determined by using the AS/NZS 1668.1:2015 equation for 1.5MW design fire (19.1m³/s) is not sufficient to maintain a smoke layer interface height of 6.5m; instead, it would require 20.4m³/s from CFAST (trial and error) simulation.

Other results in Figure 5 indicate that the opposite observation applies to design fires of 5, 10 and 15MW steady state heat release rate, where smoke exhaust rates determined by the

Page 19 of 24

joint Standard were found to be more than sufficient to maintain a 6.5m interface height. Additional (trial and error) CFAST simulations indicate that the minimum smoke exhaust rates to maintain 6.5m interface height are 37.8, 60.0 and 80.4m³/s only for design fires of 5, 10 and 15MW respectively, as opposed to 43.5, 67.0 and 90.4m³/s respectively determined by the joint Standard.

Overprediction of the smoke layer interface height by CFAST may be due to the mass flow rate across the smoke layer interface not being accounted for in the zone model. Chow et al. (2006) suggested that such mass flow rate could be up to 30% of the smoke mass exhaust rate based on their experimental data from full-scale burning tests on a mechanical smoke exhaust system in a big hall.

The field modelling results on predicted smoke layer interface heights presented in Figure 12 show that all four smoke exhaust rates determined by the joint Standard could not maintain an interface height of 6.5m for the four design fires, but at 3.2m to 4.2m only.

It should be noted that the smoke layer interface heights in Figure 12 were not determined by visually analysing the slice files at different elevations by the modeller nor by the N-percentage rule (Cooper et al. 1982) which are both considered to be subjective, but by the FDS code internally (see page 225 in McGrattan et al. 2016) using the integral ratio method (He et al. 1998).

Coyle and Novozhilov (2007) compared the predictions by FDS Version 4.0.5 with the results of full-scale experiments carried out by Yamana and Tanaka (1985) with respect to mechanical extraction of smoke (exhaust rate of 6.0 m³ s-1 with natural make-up air) in the BRI test facility of 30m x 24m x 26.3m high with a 1,300kW steady state fire.

Their FDS simulation run (1A) with 500 x 500 x 500mm grid cell size produced rather scattered results at the initial stage and predicted strong downward propagation of smoke contradicting with common sense expectations and experimental results, whereas simulation run (1B) with 400 x 400 x 400mm grid cell size overpredicted smoke descent rate initially and slightly underpredicted at later stages (Figure 16). Coyle and Novozhilov (2007) reported that the overprediction error is approximately 12% at 50s whereas the underprediction error is approximately 35% at 400s.

Figure 16: Comparison of FDS predicted smoke layer interface height with experimental data. Smoke

extraction of 6.0 m³ s-1 in BRI test facility (adopted from Fig. 7 in Coyle and Novozhilov (2007))

Page 20 of 24

Coyle and Novozhilov (2007) also compared the predictions by FDS Version 4.0.5 with other experiments with lower exhaust rates of 4.5 and 3.2 m³ s-1 where the make-up air inlet was deliberately closed in the experiments to investigate how well the mechanical extraction system would perform without replacement air. Coyle and Novozhilov (2007) reported that in the 4.5 m³ s-1 exhaust rate scenario, FDS overpredicted the smoke filling rate from a height of 26.3m to a height of 16m by approximately 8% and thereafter it underpredicted the smoke filling rate by approximately 34%. In the 3.2 m³ s-1 exhaust rate scenario, FDS overpredicted the smoke filling rate by approximately 40% for the initial 130s and thereafter underpredicted the filling rate which resulted in a final layer height error (the predicted height was higher than the actual height measured in the experiment) of 42% at 400s.

It can be deduced from the discussions above that whilst simulations by using field modelling techniques could provide much more detailed information on parameters such as temperatures, velocities, radiant heat flux and the like of the smoke plume and smoke layer spatially than by using zone modelling techniques, the accuracy of prediction of smoke layer interface height from field modelling is not necessarily higher than that from zone modelling per se.

A comparison (Table 8) of the smoke layer temperatures predicted by zone modelling (Figure 6) and field modelling (Figure 15) with that by the three plume theories (Table 3) indicates that there is a general over-prediction of smoke layer temperature by the plume theories and CFAST. This may be due to the conservative assumption in Equation 8 of this paper that there is no heat loss from the smoke layer but in reality a significant amount of heat has been lost through the smoke exhaust outlets. Additional cooling of the fire plume by the make-up air is considered to be unlikely because the make-up air velocities recorded in FDS simulations range from 1.5 to 2.5ms-1 only (Figure 10).

It is well known that CFAST overpredicts the upper layer temperature (Peacock et al. 2016a) and in some experiments the deviations could be between 10 to 50% (Luo et al. 1996/1997). Peacock et al. (2016a) states that the most common cause of over prediction of upper layer temperature by CFAST can be attributed to not knowing the species production (soot) and relative absorption of radiation by the gas layers which highlights the importance of scenario specification.

A brief literature review indicates that FDS also overpredicts the upper layer temperature. The study of Coyle and Novozhilov (2007) with respect to the 6.0 m³ s-1 exhaust rate scenario reports that the maximum temperature error is approximately 21% at 100s. In the work of Beji et al. (2014) that has the objective to evaluate the capabilities of FDS version 5.5.3 with a multi-compartment experimental set-up, general overprediction of the upper layer and lower layer temperatures were observed in the rooms adjacent and connected to the burn room as well as the corridor that connected to the burn room.

Table 8: Comparison of predicted smoke layer temperatures between plume theories and CFAST & FDS simulations for smoke layer interface height of 6.5m

Smoke layer temperature (°C) \ Total HRR (MW) 1.5 5 10 15

UK large fire plume prediction 37 76 132 189

US large fire plume prediction 98 146 186 215

Point source axisymmetric plume prediction 78 134 181 215

Maximum smoke layer temperature predicted in CFAST simulations

~108 ~158 ~230 ~278

Maximum temperature of smoke layer near the roof recorded in the FDS simulations

~51 ~78 ~108 ~122

Page 21 of 24

As the smoke exhaust rates determined by the joint Standard are higher than the predictions from the three selected plume theories (see Table 2), it implies that using the plume theories would also overestimate the smoke layer interface height if the field modelling simulation results are to be used as the benchmark.

Since these plume theories have been widely used in overseas and documented in national engineering standards or fire engineering literature over at least the last three decades without causing significant and adverse consequences or receiving criticism from the fire engineering fraternity or the approval authorities, further research is required to examine whether there are other input parameters and assumptions in the FDS simulations which could influence the output as well as to compare with available experimental data such as that used in the study by Coyle and Novozhilov (2007).

6 CONCLUSION

The methodology to determine smoke exhaust rates specified in section 9 ‘Hot layer smoke control systems’ of AS/NZS 1668.1-2015 has been examined. It was found that the joint Standard has not indicated which types of fire plume are intended to be covered by the methodology, has not provided guidance on how the temperature of the upper smoke layer can be estimated, and has not disclosed the technical basis of the methodology. A significant revision of section 9 of the joint Standard would be required for it to be useful as a tool for performance-based smoke exhaust system design.

Numerical experiments were conducted in the study by using zone modelling and field modelling techniques. Based on the assumption that the smoke exhaust rates predicted by the joint Standard and plume theories in Table 2 are reasonably accurate, it was found that the CFAST zone model (V7.2.1) underpredicts the smoke filling rates by 11 to 13%, whereas the FDS field model (V6.5.3) overpredicts the smoke filling rates by about 60% according to the results presented in Figure 12. These observations demonstrate that whilst field modelling has the capability to yield more detailed information about the smoke properties spatially and temporally, it does not infer the predictions by using field modelling techniques must be more accurate than by simpler methods such as zone modelling and empirical equations.

Additional research is required to examine the predictive capabilities of FDS V6.5.3 by comparing with available experimental data and to identify which user inputs have significant influence on the simulation results.

REFERENCES

Australian Building Codes Board, International Fire Engineering Guidelines (IFEG), Edition 2005.

Bahrami, M. (2011). Introduction, Notes for ENSC 283 Introduction to Fluid Mechanics. http://www.sfu.ca/~mbahrami/ENSC%20283/Notes/Introduction.pdf, Simon Fraser University [Accessed 23 February 2017].

Beji, T., Bonte, F. and Merci, B. (2014). Numerical Simulations of a Mechanically-Ventilated Multi-Compartment Fire. In: Fire Safety Science – Proceedings of the Eleventh International Symposium. International Association for fire Safety Science.

Bennetts, I.D., Poh, K.W., Poon, S.L., Thomas, I.R., Lee, A.C., Beever, P.F., Ramsay, G.C. and Timms, G.R. (1988). Fire Safety in Shopping Centres. Final Research Report Project 6, Fire Code Reform Research Program.

Chow, W.K. (1994). A Short Note on the Simulation of the Atrium Smoke Filling Process Using Fire Zone Models. J. of Fire Sciences, 12: 516-528.

Page 22 of 24

Chow, W.K., Yi, L., Shi, C.L., Li, Y.Z. and Huo, R. (2006). Mass flow rates across layer interface in a two-layer zone model in an atrium with mechanical exhaust system. Building and Environment, 41:1198-1202.

CIBSE Guide E (2010). Fire Safety Engineering. 3rd ed. The Chartered Institution of Building Services Engineers, London.

Cooper, L.Y., Harkleroad, M., Quintiere, J. and Rinkinen, W. (1982). An Experimental Study of Upper Hot Layer Stratification in Full-Scale Multiroom Fire Scenarios. Journal of Heat Transfer, 104:741-749.

Coyle, P. and Novozhilov, V. (2007). Further Validation of Fire Dynamics Simulator Using Smoke Management Studies. International Journal on Engineering Performance-Based Fire Codes, 9(1):7-30.

Duong, D.Q. (1990). The Accuracy of Computer Fire Models: Some Comparisons with Experimental Data from Australia. Fire Safety Journal, 16:415-431.

Engineers Australia, 2012, Practice note for design fires, Version 1.2. Society of Fire Safety, New South Wales, Australia.

Engineers Australia, 2014, Practice note for tenability criteria in building fires, Version 2, Society of Fire Safety, New South Wales, Australia.

Fire Dynamic Simulator, 2016, Version 6.5, National Institute of Standards and Technology (NIST).

Forney, G.P. (2016). Smokeview, A Tool for Visualizing Fire Dynamics Simulation Data Volume II: Technical Reference Guide, NIST Special Publication 1017-2, Sixth Edition. National Institute of Standards and Technology, US Department of Commerce, USA.

Geiman, J.P. and Gottuk, D.T. (2006). Evaluation of Smoke Detector Response Estimation Methods: Part I – Optical Density, Temperature Rise and Velocity at Alarm. Journal of Fire Protection Engineering, 16(4):251-268.

Gottuk, D., Mealy, C. and Floyd, J. (2009). Smoke Transport and FDS Validation. Fire Safety Science – Proceedings of the Ninth International Symposium, pp. 129-140.

He, Y.P., Fernando, A. and Luo, M.C. (1998). Determination of interface height from measured parameter profile in enclosure fire experiment. Fire Safety Journal, 31:19-38.

Heskestad, G. (1982). Engineering relations for fire plumes. SFPE TR 82-8, Society of Fire Protection Engineers.

Heskestad, G. (2016). Fire Plumes, Flame Height, and Air Entrainment. In: Hurley, M.J., ed. SFPE Handbook of Fire Protection Engineering, 5th ed. Chapter 13, pp. 396-428, Society of Fire Protection Engineers, Springers New York.

Hinkley, P.L. (1986). Rates of ‘production’ of hot gases in roof venting experiments. Fire Safety Journal, Vol. 10, pp. 57-65.

Hui, M.C. and Mendham, F. (2015). Is the use of CFD in developing Alternative Solutions regarded as a professional engineering service and how could its quality be assured? Presentation at the IFE Australia Conference, Fire Safety – So much more than compliance with the NCC-BCA, Canberra, 8th and 9th October 2015.

Luo, M., He, Y. and Beck, V. (1996/1997). Comparison of Existing Fire Model Predictions With Experimental Results From Real Fire Scenarios. Journal of Applied Fire Science, 6(4):357.

McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C. and Overholt, K. (2016). Fire Dynamics Simulator User’s Guide, FDS Version 6.5.0, NIST Special Publication 1019 Sixth Edition, National Institute of Standards and Technology, US Department of Commerce, USA.

Page 23 of 24

Morton, B.R., Taylor, G. and Turner, J.S. (1956). Turbulent gravitational convection from maintained and instantaneous sources. Proc. Roy. Soc. A., 234:1-23.

Mowrer, F.W., Milke, J.A. and Gandhi, P. (2008). Validation of a Smoke Detection Performance Prediction Methodology Volume 4. Evaluation of FDS Smoke Detection Prediction Methodology. Fire Protection Research Foundation, Quincy, Massachusetts, USA.

Mudan, K. S. and Croce, P. A. (1988). “Fire Hazard Calculations for Large Open Hydrocarbon Fires,” SFPE Handbook of Fire Protection Engineering, 4th edition, National Fire Protection Association, Quincy, MA.

Narayanan, P. (1993). Smoke control in multi-storey buildings. Building Research Association of New Zealand (BRANZ), study report SR 50, Judgeford, New Zealand.

NFPA 92 (2012). Standard for Smoke Control Systems, National Fire Protection Association, Massachusetts, pp. 12.

NFPA 92B (1991). Guide for smoke management systems in malls, atria and large areas. National Fire Protection Association, Massachusetts, pp. 31.

Peacock, R.D., Forney, G.P. and Reneke, P.A. (2016a). CFAST – Consolidated Model of Fire Growth and Smoke Transport (Version 7) Volume 3: Verification and Validation Guide, NIST Technical Note 1889v3. National Institute of Standards and Technology, USA.

Peacock, R.D., Jones, W.W., Bukowski, R.W. and Forney, C.L. (1991). Technical Reference Guide for the HAZARD I Fire Hazard Assessment Method Version 1.1, NIST Handbook 146, Volume II, Building and Fire Research Laboratory, Gaithersburg, MD, USA.

Peacock, R.D., McGrattan, K.B., Forney, G.P. and Reneke, P.A. (2016b). CFAST – Consolidated Model of Fire Growth and Smoke Transport (Version 7) Volume 1: Technical Reference Guide, NIST Technical Note 1889v1, National Institute of Standards and Technology, USA.

Petterson, N.M. (2002). Assessing the Feasibility of Reducing the Grid Resolution in FDS Field Modelling. Fire Engineering Research Report, School of Engineering, University of Canterbury, Christchurch, New Zealand.

Rouse, H., Yih, C.S. and Humphreys, H.W. (1952). Gravitational convection from a boundary source. Tellus, 4:201-210.

Simms, D.L., Hird, D. and Wraight, H.G.H. (1960). The temperature and duration of fires: Part I: Some experiments with models with restricted ventilation. Fire Research Note No. 412, Fire Research Station, Boreham Wood, Herts, UK.

Tewarson, A, (2008). Generation of Heat and Gaseous, Liquid, and Solid Products in Fires. In DiNenno, P.J., ed., The SFPE Handbook of Fire Protection Engineering, 4th Edition, National Fire Protection Association, Chapter 3-4, Table 3-4.16 and Table 3-4.18.

Thomas, P.H. (1995). The Growth of Fire – Ignition to Full Involvement. In Cox, G., ed., Combustion Fundamentals of Fire, Academic Press Limited, London.

Thomas, P.H., Hinkley, P.L., Theobald, C.R. and Simms, D.L. (1963). Investigations into the flow of hot gases in roof venting. Fire Research Technical Paper No. 7, HMSO, London.

U.S. Nuclear Regulatory Commission and Electric Power Research Institute. (2007). Verification and validation of selected fire models for nuclear power plant applications. NUREG-1824, Volume 7: Fire Dynamics Simulator, Final Report, USA.

Yamana, T. and Tanaka, T. (1985). Smoke control in large scale spaces, part 2. Building Research Institute, Ministry of Construction, Japan.

Yih, C.S. (1952). Free convection due to a point source of heat. Proceedings of the First U.S. National Congress of Applied Mechanics, Chicago, pp. 941-947.

Page 24 of 24

Yokoi, S. (1960). Study on the prevention of fire-spread caused by hot upward current. Japanese Ministry of Construction Building Research Institute Report No. 34, Tokyo.

Zhang, X., Yang, M., Wang, J. and He, Y. (2010). Effects of Computational Domain on Numerical Simulation of Building Fires, Journal of Fire Protection Engineering, 20(4):225-251.

Zhou, D., Wang J. and He, Y. (2010). Numerical simulation study of smoke exhaust efficiency in an atrium. Journal of Fire Protection Engineering, 20(2):117-142.