Fire Safety Journal 91 (2017) 389398

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/resaf

IAFSS 12th Symposium 2017

Experimental and numerical evaluation of the inuence of the soot yield on MARK
the visibility in smoke in CFD analysis

Wojciech Wgrzyskia, , Gabriele Vigneb,c
a
Building Research Institute, Fire Research Department, Filtrowa 1 St., 00-611 Warsaw, Poland
b
JVVA Fire & Risk, Velazquez 157, 28002 Madrid, Spain
c
University of Jan, Fluid Mechanics Department, Spain

A R T I C L E I N F O A BS T RAC T

Keywords: Experimental and numerical analysis have been performed to evaluate the inuence of the soot yield parameter
Fire on the results of advanced engineering analysis, in regards to visibility. After identifying soot yield as the most
Smoke inuential factor on the results, fuels with various values of Ys have been analysed in a re chamber and then
CFD compared to numerical data. The numerical analysis has been performed using two dierent CFD packages,
Smoke management
ANSYS Fluent, and Fire Dynamics Simulator. The numerical analysis itself show an apparent hyperbolic trend
Visibility
of the visibility when changing the soot yield with clear consequences on the ASET (Available Safe Egress Time).
Soot yield
Below a cut-o point, that exists at a soot yield value close to Ys =0,10 g/g, a small change in the parameter
causes a substantial shift in the results (visibility or ASET time), while above this value an increase to soot yield
does barely inuence the results. Qualitative assessment of the results shows a need for use of conservative
values of Ys in engineering analysis if detailed and precise material data is not available. Additionally to the full-
scale experiments, a real case study has been included to show how this research can be translated into the Fire
Safety Engineering design process. In this study, change of Ys value below 0,10 g/g caused a signicant change
of the qualitative assessment of the results of CFD.

1. Introduction performed in the Building Research Institute (ITB) in Warsaw, Poland.

Many parameters exist that can be used to describe the smoke
properties, e.g. temperature, mass density of products, obscuration
density, transmittance, toxic gasses concentration, etc. Among these
numerous variables, when conducting Fire Safety Engineering (FSE)
analysis, one of the most important is the visibility in smoke. It is
very common in FSE study, to see the visibility to be the rst parameter
that meets its critical value (tenability criteria). Visibility in numerical
modeling is a result of modeling the transport of combustion products
within the model, especially the soot aerosol. The soot is introduced
into numerical domain through a source model, and its amount is
directly aected by the eective heat of combustion of the fuel (Hc,e)
and the soot yield (Ys). The visibility sub model is also dependent on
the visibility factor (K) and the mass extinction coecient of the smoke
(). Previous research [1] has demonstrated, that among these para-
meters, the soot yield (Ys) has the greatest inuence on the value of
visibility, while observation of engineering projects shown that is the
one that is most liberally chosen.
The aim of this study was to further investigate the impact of soot
yield on visibility. This goal was pursued through full-scale re tests
Dierent res have been examined, using fuels with a high diversity of
soot yield values, varying from Ys =0.001 g/g to Ys =0.178 g/g. The
density of the smoke layer has been measured with an optical
densitometer.
The results have been compared to parametric numerical analysis,
performed with ANSYS Fluent and Fire Dynamics Simulator. The
numerical analysis shows a hyperbolic trend of the Visibility when
increasing the soot yield with a cut-o point around 0.10 g/g, where a
lower value can produce a relevant change, while a higher value a
negligible change in the Visibility and therefore to the ASET (Available
Safe Egress Time). Ultimately, a case study has been included to show
how critical the soot yield is in the determination of the ASET in a real
building.
2. Modeling visibility in smoke
2.1. Optical properties of smoke
The optical properties of smoke aerosol in the air are comparable
with other dispersive systems. The intensity of light passing through

Corresponding author.

http://dx.doi.org/10.1016/j.resaf.2017.03.053
Received 14 February 2017; Accepted 27 March 2017
Available online 07 April 2017
0379-7112/ 2017 Elsevier Ltd. All rights reserved.
W. Wgrzyski, G. Vigne
Nomenclature
E Illuminance [lx]
K Visibility factor [-]
Hce Eective heat of combustion [MJ/kg]
Q Heat Release Rate [kW]
Km Mass extinction coecient [m2]
ms mass ux of smoke [kg/s]
l Length of light path [m]
cp Specic heat [kJ/kg K]
T Temperature [C or K]
I Luminous intensity [cd]
Ys Soot yield [g/g]
z Height in a re plume [m]
the smoke depends on scattering, absorption, diraction and other
smaller eects, which are dependent on the parameters of the smoke
and the light. Black smoke is highly absorbing, while white smoke is
highly scattering the light. As smoke contains condensed water
particles, additional light refraction is observed. As it is not possible
to measure the absorption and scattering eects separately, their
combined damping eect is measured. In engineering calculations,
these eects are simplied even further, as the optical properties of
smoke and their inuence on the light are not modeled at all, but rather
evaluated with simplied mathematical models of visibility within the
smoke in function of the local smoke concentration.
Fig. 1 presents a typical method of displaying the results of visibility
in smoke in a CFD analysis. Points (a), (b) and (c) are relatively close to
each other but have radically dierent values of visibility. This does not
mean that a subject in point (a) would see at 30 m distance, while one
at (c) only at 5 m. The physical meaning is that an object would be
visible from 30 m, 20 m or 5 m if a whole room was lled with a
uniform smoke layer, as dense as in points respectively, (a), (b) or (c).
To realistically determine the visibility through a non-uniform smoke,
more advanced methods such as ray-tracing should be used.
2.2. Introduction of the smoke into the numerical model
In complex models, the re is usually represented by one of three
typical approaches, each of them can be implemented either as a 2D
source (surface) or a 3D source (volume), depending on the solver:
a) xed source of heat and smoke, which is described by its volume
(area), and the amount of energy and mass of combustion products
emitted within. The yields and their change in time are pre-dened,
and chemistry models are not used. Complex phenomena such as
Fig. 1. The local visibility range plot (most left, range from 0 to 30 m and more, for K =3) is created as an array of visibility values from individual cells (middle clips). Value within each
of the cells represents the distance, from which a certain object (eg. sign, light) would be seen, in a room (right side drawings) with uniform smoke corresponding to the mass
concentration of the smoke within that cell.
390
Fire Safety Journal 91 (2017) 389398
D Fire diameter [m]
Greek
density [kg/m3]
Wavelength [nm]
Specic mass extinction coecient [m2/g]
subscripts
p into the plume
0 ambient
ame
pyrolysis, self-extinguishing or under-ventilated combustion are not
explicitly modeled;
b) source term that emits fuel or a mixture of fuel and oxygen, which is
further burned using simple chemistry models (e.g. 2-eq. Arrhenius
models [2], or pre-mixed burning model [3]). The yields are
dependent on the chemistry and local oxygen concentration, which
allows the inclusion of self-extinguishing or under-ventilated burn-
ing phenomena, but due to predened yield, time to burnout is not
explicitly modeled;
c) model of materials that have a mass, and are the source of fuel
through pyrolysis or evaporation. The chemistry of the combustion
model is similar to model b). This approach allows modeling of the
fuel depletion, the spread of the re, but is the most computation-
ally expensive and does not have sucient validation for complex
materials, thus is rarely used in practical engineering applications.
Models a) and b) are a prescript representation of a re, which can
be considered explanatory, but not predictive. Once the smoke is
released into the model, to estimate the smoke density in particular
control volumes (cells) of a CFD model, continuity, momentum, and
mass transport equations, along with turbulent ow sub-model are
solved. The smoke is generated within the source of re, represented as
soot particles introduced to the convective stream of air also produced
by the source. This representation heavily relies on the main para-
meters relevant to the soot production eective heat of combustion
(Hc,e), heat release rate (Q) and the soot yield parameter (Ys). The
mass of smoke introduced into the model can be presented as:
Q
ms = Ys
Hc, eff (1)
The amount of the soot often referred as the mass smoke
W. Wgrzyski, G. Vigne
concentration, is known in every control volume of the model, in every
discrete time-step.
A detailed description of CFD modeling is not presented in this
paper; please refer to [46] for information on the use of CFD in Fire
Safety Engineering. These documents provide a description of the
dierential equations used in the models, along with sub-model of the
mass transport. For the description of the applicability of CFD models,
the reader may refer to [7,8].
2.3. Light dampening by the smoke
The basic interaction between light and an aerosol is described by
the law of Boguer-Lambert-Beer, Eq. (2), which represents the relation
between light dampening and extinction coecient (Km) [9].
I = I0 eKm l (2)
The application of the Boguer-Lambert-Beer law for smoke has the
following limitations:
a) the smoke layer should be homogenous. If visibility is estimated for
heterogeneous mixture it must be referred as local visibility
parameter and related separately for each control volume, for
which the calculation was performed.
b) the mechanism of light dampening is absorption and not scattering
while in re smoke, both phenomena occur at the same time;
c) a single light source exists, or light sources are parallel - while in
buildings, there are multiple sources of light;
d) the source of light is monochromatic - while in buildings both
natural and articial light sources are polychromatic.
Authors agree with Mayerhfer et al. [10], that taking simply the
negative decadic logarithm of a singular transmittance measurement
does not compute the absorbance and whatever results cannot be
quantitatively evaluated and will not linearly depend on concentration
or thickness. However, some solutions for optimizing the use of
Lamberts-Beer law exists, and are used for computing photometrical
data in the context of fog and other atmospheric phenomena [1113].
Despite its limitations, the Boguer-Lambert-Beer law is still the
most common approximation of the eect of smoke on the light passing
it. To describe the smoke layer with a mathematical model, obscuration
density parameter, also known as the optical density of the smoke, is
introduced. This parameter relies on the size of smoke particles, their
distribution, diameter, absorption and scattering coecients and the
length of lightwave. As a simplication of this, a specic mass
extinction coecient can be introduced (), which was discussed in
depth in [14,15]. Introducing this coecient into the Boguer-Lambert-
Beer equation it results in:
I = I0 eml (3)
The values of mass extinction coecients for various ammable
materials can be found in the literature, ranging from 600 to 9800 m2/
g, with the standard value implemented e.g. in FDS v.6 of 8700 m2/g. A
comprehensive discussion on the values of mass extinction coecient
can be found in [14,15]. Detailed information on how this parameter is
derived from thermal decomposition and combustion of polyester
materials is presented in [16,17]. A dierent approach, not directly
connected to the material but the light is presented by Widmann [9]. A
power-law correlation is introduced for the mass specic extinction
coecient dependent on the wavelength of the light (). This correla-
tion allows estimating the Km coecient for a well-described source of
light with constant wavelength.
= = 48081.0088 (4)
The Boguer-Lambert-Beer law can answer the question if the light
391
Fire Safety Journal 91 (2017) 389398
can pass through the layer of smoke, but it is not yet the indicator of a
visibility of an exit door or an evacuation sign.
2.4. Modeling the visibility in smoke
The human eye can distinguish a shape (such as an evacuation sign)
when the dierence between the luminance of the object and its
background exceeds the critical contrast value, at which the object is
visible. This critical contrast value will rely on the light intensity. As an
example for a mist conditions this value is c =0.02 [18], while some
sources present this value as a range from 0.01 to 0.05 [19]. This
parameter is also obviously dependent on the individual eyesight
characteristic of a person.
L t Lb
> c
Lt (5)
By combining this theory with the mathematical model of light
dampening eect, Jin derived a mathematical model that allows
estimation of visibility through a homogenous layer of smoke. [20,21].
1 L
V= loge t
Cs c E (6)
Both brightness of the sign (Lt) and its illuminance from all
directions (E) will inuence the visibility value. To simplify this
equation, based on a test program in a small chamber, lled with
white smoke produced by pyrolysis of wood, most of the variables were
combined into single coecient K is known as:
1
V = K
Cs (7)
The widely used values of K coecient are 3 for light reecting
signs and 8 for light emitting signs, although, in original work of Jin
(1970) [18,20,22], this was a variable dependent on the dimensionless
brightness of a sign (Lt/E), Fig. 2. In the SFPE Handbook of Fire
Protection Engineering (3rd edition, 2002), this value is presented as a
range between 2 and 4 for light reecting signs and 510 for light
emitting signs [19]. In the last edition of the SFPE Handbook of Fire
Protection Engineering (5th edition, 2016), a comment was included in
chapter relevant to visibility in smoke [23]:
It should be noted that the constant k in Equation 61.4 tends to be
larger than the previous data indicated by Jin ().
When applying the theory of Jin into modern CFD models, the user
has to be advised that this theory was derived by Jin in early 1970's,
before the development of two-zone models [24], ASET/RSET concept
[25] or early CFD models [26] developed for Fire Safety Engineering
Fig. 2. Dimensionless number K in function of dimensionless brightness of an
evacuation sign, as printed in original paper by Jin [20].
W. Wgrzyski, G. Vigne
applications. This means that the theory may be expanded beyond its
original area of application, and as such, it has to be taken with
necessary caution. Application of the theory of Jin directly into modern
CFD models, for each control volume, has taken the form of:
K
V=
ms
3. Full-scale experiments
3.1. Murcia experiment
In 2015, Vigne et al. [1] conducted a research study with the
objective to evaluate the sensitivity of visibility when varying certain
critical parameters that have a direct inuence on it. Such parameters
were the mass extinction coecient (), the production rate of soot (Ys)
and the visibility factor (K).
The analysis consisted of an initial statistical analysis performed in
order to determine the most relevant parameters. Specic values
(ranges) were dened for the input variables, Ys: 0.015 g/g, 0.120 g/g
and 0.230 g/g; : 7600 m2/kg, 8700 m2/kg and 9800 m2/kg; K: 3 and
8. This approach resulted in 18 possible combinations of these factors.
The simulations were based on full-scale tests i.e. the same test set up
was used as model setup (ventilation conditions, re sizes, the
geometry of volume, etc.). The full-scale tests had been performed at
an earlier stage in the Metal Technology Center (CTM) in Murcia [1]
and were used for validation of numerical analysis. The facility has a
cubic form, dimensions of 20 m x 20 m x 20 m and is equipped with
both natural and mechanical ventilation systems to control the test
environment. The next step was to simulate 18 models with FDS (Fire
Dynamics Simulator, v.5.3.1 [2]).
Further analysis was then performed to determine the dependency
between input data with the simulated output data. The results
obtained from the simulations were to some degree also compared
with the results obtained in obtained the full-scale tests, although it
was not possible to directly check the visibility since there was no
instrumentation dedicated to measuring visibility when the tests were
performed.
The results of numerical analysis have shown that the Soot Yield is
the most inuential variable, and there exists a cut-o value i.e. the
actual eect on the output factor (visibility) was very dependent on if
the parameter value was above or below this cut-o limit. The variation
in the rate of the soot yield (Ys) to values lower than 0.120 g/g
signicantly and inversely aects the visibility. A decrease in Ys to
13% resulted in a decrease of visibility of about 700%. The situation is
dierent for of Ys greater than 0.120 g/g, where an increase of 83% in
Ys caused a 30% decrease in visibility. Other parameters (specic
extinction coecient - and visibility factor - K) shown a linear
relationship with visibility. After this study, the Soot Yield (Ys)
Fig. 3. Detail of the ame (left) and photo of the experiment with detail of the smoke layer (right) [1].
Fire Safety Journal 91 (2017) 389398
parameter was chosen as the most inuential for further studies
(Fig. 3).
3.2. Warsaw experiment
To better estimate the impact of soot yield on visibility, full-scale
(8)
re tests have been performed in the Building Research Institute (ITB)
in Warsaw, Poland. The test facility includes a room with dimensions of
1010 m and a height of 4 m, equipped with a mechanical smoke
exhaust system. Previous experiments have proven that a state of
equilibrium is achieved in the chamber for res with power ranging
from 130 to 160 kW and approximately 2.00 m diameter. This means,
that same amount of smoke is exhausted, as enters the smoke layer,
and the height of layer is stable. This phenomenon is used in
advantage, to measure the obscuration within the smoke layer in
longer time-span (close to a steady-state measurement in a homo-
genous smoke layer). Dierent res have been examined, using fuels
with a high diversity of soot yield values, varying from Ys =0.001 g/g to
Ys =0.178 g/g. The density of the smoke layer has been measured with
optical densitometer that is a part of the test chamber. Information
about this experiment can be found in [27,28].
The following materials were chosen for assessment [29,30]:
methyl alcohol, Ys =0.001 g/g
propane alcohol, Ys =0.015 g/g
heptane, Ys =0.035 g/g
Toluene, Ys =0.178 g/g
The cross-section of the test chamber is presented in Fig. 4, and the
images from performed tests are presented in Fig. 5. Results of the
experiments are presented further in the paper, along with the results
of numerical studies.
4. Validation and verication of numerical model of visibility
in CFD
The Warsaw ITB experiment was recreated in numerical models
ANSYS Fluent (v.17.0.0) and Fire Dynamics Simulator (v.6.4.0) to
validate these models for further parametric studies.
Two dierent approaches were used, with following sub-models:
a) ANSYS Fluent package, second order numerical scheme (segre-
gate-type numerical scheme), with Realizable RANS k- turbulence
model (full buoyancy and enhanced wall treatment sub-models)
and P1 radiation model. Solution convergence was measured at
each discrete time step. The re was modeled as a volumetric source
of heat and smoke [3].
b) FDS package, Low Mach, explicit, second-order solver. LES turbu-
lence model and DO gray gas radiation model. Structured uniform
392
W. Wgrzyski, G. Vigne Fire Safety Journal 91 (2017) 389398

Fig. 4. Conguration of the test facility.

Cartesian mesh. Eddy dissipation model (fast chemistry) for the
single-step reaction between fuel and oxidizer [5].
Both models represented the same numerical domain of the test
chamber. The size of the domain was 10,0010,00 m and height of
4,00 m. Inlets and outlets were modeled, as in the physical experiment,
Fig. 4. The mesh size in FDS model was 10 cm (Cartesian structured
mesh), and in ANSYS ranging from 10 to 40 cm (unstructured
tetrahedral mesh, growth ratio 1:1,20).
In case of FDS the re was modeled as a burner with
dimensions of 0,50,5 m, while in ANSYS the re was represented as
a release of energy and smoke-particles in a pre-dened volume with
dimensions of 0,50,5 m and height of 1 m (simplied approach, no
chemistry sub-model). The pre-dened HRR of burner (or energy
release in the volume) was dened based on the burning time and mass
loss rate of the experiments. Authors are aware of the existence of
rene sub-models which allow to predict the evaporation of the fuel
and therefore evaluate the HRR; however, predicting the HRR in such
way would add a source of uncertainty that would move the focus of
this study. Previous studies performed by the authors have shown a
high level of uncertainty comparing real scale tests and predicted HRR
in several heptane pool res [31]. Soot yield value of the fuel was a
subject of the parametric study, while heat of combustion was chosen
as corresponding to heptane (Hc =44,4 MJ/kg).
In order to check the performance of the numerical models used, a
preliminary validation analysis has been performed comparing the
measured and predicted the temperature in test data obtained from
dierent tests (in particular, tests with Heptane, Propanol, Toluene and
Methanol). The results show a relatively good agreement between
modeled and experimental temperature data for Heptane and Propanol
with a maximum discrepancy of about 20%. Higher discrepancies have
been found for Methanol and Toluene, which can be attributed to the
uncertainty of HRR measurement in the model. The resulting smoke
layer, Fig. 7, in CFD analysis was close to homogenous and stable, as
observed in the full-scale experiments Fig. 5.
The following graphs show the comparison between experimental
and numerical data (FDS and Fluent) data for a specic thermocouple
(T30, in Fig. 6 above) which was reasonably away from the re plume
and enough immersed in the smoke layer. Several Thermocouple has
been installed in the chamber and dened in the model, but only this
was used in this validation since it was in a relevant position and was
not aected by the uctuation of the ame (Figs. 79).
5. Parametric study #1: soot yield vs visibility
Twenty (20) models based on the Warsaw experiment have been

Methanol, Ys = 0.001 g/g Propanol, Ys = 0.015 g/g

Heptane, Ys = 0.035 g/g Toluene, Ys = 0.178 g/g

Fig. 5. Pictures of the ITB test chamber during each of the test performed.

393
W. Wgrzyski, G. Vigne
Fig. 6. Position of the Thermocouple T30 employed in the analysis (FDS model), next to
the optical densitometer, used as the presented reference for model validation.
processed both in Fluent and FDS [28] with a soot yield range between
0.01 g/g and 0.2 g/g using the Heptane reaction as a base. Despite the
possible error with the use of Boguer-Lambert-Beer law, the same
mathematical model was also applied to the transmittance measure-
ments of the Warsaw experiment and presented in the plots along
with the results of the numerical study. The results for both Fluent and
FDS (and the experiments) show a linear trend for obscuration density
(which is a representation of the mass concentration of smoke, Fig. 10)
and a hyperbolic trend of the visibility when increasing the soot yield,
conrming the previous work undertaken by the authors.
As expected, the increase of the amount of soot introduced to the
model increases the amount of the soot within the steady state smoke
layer. For the visibility in smoke, a cut-o soot yield value exists at
about 0.100.12 g/g. Before that value, visibility changes considerably
while after that value, the variation can almost insignicant. The
graphs below show the visibility trend against soot yield calculated
with Fluent and FDS both for reecting and emitting signs. It can also
be noted, that the measurements of the obscuration density (and the
visibility) dier between experiment and both numerical models
substantially. The authors address this dierence to the uncertainty
of the choice of mass extinction coecient () which was taken directly
from the literature. Authors were unable to measure this value for the
smoke generated in the experiment, as this requires dedicated,
sophisticated apparatus. The use of Widmann's Eq. (4) provided with
better correlation, but for the purpose of this study, Authors used the
values that are default for the CFD software, to illustrate the practical
aspects of the change of soot yield.
6. Parametric study #2: soot yield vs aset
Using previously described numerical experiment the results were
investigated at the time, after which 10 m visibility criterion was
exceeded in the model, within the homogenous smoke layer (for
visibility factor K=3). The 10 m visibility criterion was chosen as a
representative due to its popularity among Fire Safety Engineers, and it
is mentioned in various guidelines in this area: [23,3234]. It must be
Fig. 7. Sample results of reference studies performed in ANSYS Fluent (left) and FDS (right). Besides the plume, ceiling jet and zone transition regions, the layer can be considered
homogenous.
394
Fire Safety Journal 91 (2017) 389398
noted that some sources give higher values than 10 m (e.g. 16 m in
[35], 13 m for unfamiliar users [22]).
For the time to reach visibility of 10 m, similar cut-o value existed,
at approximately 0.10 and 0.12 g/g. For soot yield values below the cut-
o point, the time to reach 10 m visibility was signicantly shorter. For
soot yield values above this point, the time to reach 10 m visibility did
not dier more, than the resolution of time measurement (for
subsequent steps in soot yield value). It is important to note, that with
the conservative assumption, the time to reach 10 m visibility was close
to 1015 s, while for values under 0,03 it was more than one minute
(Figs. 11 and 12).
7. Parametric study #3: large volume compartment
A real case has been included to show how observations from
studies #1 and #2 can be translated into the Fire Safety Engineering
design process. This exercise concerns in determining the concentra-
tion of smoke, and the visibility in smoke in a real building through a
series of numerical simulations using ANSYS Fluent software. The
analysis was conducted for a compartment with dimensions of 20.00 m
x 25.00 m x 5.00 m (W x L x H), and a volume of 5000 m. The
compartment was connected with an adjacent space (a corridor of a
shopping mall) through opening dimensioned 8.00 m x 3.50 m. Smoke
and heat exhaust system in the compartment had a volumetric capacity
of 60000 m/h, and the air was exhausted through two ducts with 11
inlets on each of them (22 total). The re was modeled as a frustum
with a height of 1.75 m and the total volume of 3.2 m. The total heat
release rate of the re was limited to 2.50 MW, and it was described
with t2 growth function, with =0.0469 kW/s2, Fig. 13. Hc,e was
equal to 20.00 MJ/kg, and the soot yield was a variable ranging from
0.02 g/g to 0.16 g/g (0.02 step). This set of boundary conditions and
mathematical models was previously validated using the experiments
shown previously in this paper.
The following physical sub-models were used:
Segregate numerical scheme, second order solver;
Realizable RANS k- turbulence model;
P1 radiation model;
Wall boundaries gypsum plasterboard, a third type of b-c.
The results were assessed with the respect to:
The mass density of smoke and visibility in smoke (K=3), in point in
a middle of the room (x=12.5 m, y=10 m, z=2.00 m) Fig. 14;
The mass density of smoke and visibility in smoke (K=3) in a plot
(axis y=9.00 m) through the re, in the 420th second of the analysis,
Fig. 15.
Quantitative assessment of the results shows, that the mass density
W. Wgrzyski, G. Vigne Fire Safety Journal 91 (2017) 389398

Fig. 8. Experimental validation temperature comparison in the optical densitometer proximity, ANSYS Fluent.

Fig. 9. Experimental validation temperature comparison in the optical densitometer proximity, FDS.

Fig. 10. Obscuration density vs. Soot Yield.

of smoke between subsequent models diered by a close to constant
value (i.e., the dierence between Ys =0.02 g/g and Ys =0.04 g/g was
approximately the same, as between Ys =0.04 g/g and Ys =0.06 g/g).
The only signicant dierences were observed in the late stage of the
simulation when the ows within the compartment diered between
the scenarios. Despite the fact, that the mass concentration of smoke
within the room was proportional to the change of Soot Yield value, the
visibility in smoke diered substantially. The time to reach 10 m
visibility in the chosen point was: 410 s for Ys =0.02 g/g; 312 s for Ys
=0.04 g/g; 285 s for Ys =0.06 g/g; 260 s for Ys =0.08 g/g; 227 s to 237 s
for Ys =0.100.16 g/g. It can be noted, that with Ys between 0.02
0.10 g/g, the time to reach 10 m visibility did range from 410 s to
260 s, while for Ys between 0.100.16 g/g, this time, varied between
227 s and 237 s. This shows, that if the user chooses the Ys=0.10 g/g, it
can be considered as a conservative assumption, that will not negatively
inuence the qualitative analysis of the results.

395
W. Wgrzyski, G. Vigne Fire Safety Journal 91 (2017) 389398

Fig. 11. Visibility range vs. Soot Yield for K=3 and K=8.

Fig. 12. Time to reach 10 m visibility (K =3) in experiment, ANSYS Fluent, and FDS.

Fig. 13. Numerical model of the volume compartment (5000 m3, left) and the HRR of the re used (right).

The qualitative assessment of the results diers substantially
between ve cases with Ys < 0.10 g/g, while for cases with Ys
0.10 g/g the results were similar. In the two lowest soot yield value
cases (Ys 0.02 and 0.04 g/g) the smoke exhaust system was eective,
while for cases with Ys between 0.040.08 g/g the system did not
manage a smoke-free layer in the compartment. For soot yield values
below 0.06 g/g, a smoke lodging in the bottom layer of the room is not
observed (especially in the visibility plots), while for higher soot yield
the smoke is observed in the whole height of the room. Additionally,
with the Ys lower than 0.08 g/g, the conditions outside the compart-
ment can be considered tenable, while at higher Ys they are not. The
dierences between the scenarios have severe consequences on the

396
W. Wgrzyski, G. Vigne
Fig. 14. Plot of the mass density of smoke (left) and visibility in smoke (right) in a point in the middle of the compartment.
Fig. 15. Mass density of smoke and visibility in smoke in a plot through the compartment, t=420 s.
assessment for Ys < 0.04 g/g the system could be considered correct,
at 0.04 0.08 g/g the conditions inside the room were untenable, but
outside it was correct and at Ys 0.10 g/g large amount of the smoke
was observed in and out of the room. If the CFD user relied on a low Ys
value (Ys < 0.10 g/g), his assessment may be invalid, and potential
problems with the smoke control system can be overlooked.
8. Conclusions and future works
Four experimental tests have been performed in the Building
Research Institute (ITB) in Warsaw, Poland, using dierent fuels with
a broad range of soot yield (from 0.001 g/g for methanol to 0.178 g/g
for toluene). The results have been compared to four (4) numerical
models processed with both ANSYS Fluent and Fire Dynamics
Simulator. The numerical models were over predicting the tempera-
ture, which may be connected to the non-perfect modeling of the HRR
the burning process was non-uniform, while in CFD authors used a
397
Fire Safety Journal 91 (2017) 389398
steady state heat and smoke generation. After the validation, a
parametric numerical study was performed looking at 20 dierent
cases, each one with a soot yield dierence of 0.01 g/g (from 0.01 g/g
to 0.20 g/g).
It is dicult to compare optical densitometer measurements
(experimental) with measurements in CFD software. The main reason
behind it is that transmittance measured with a monochromatic light
source of optical densitometer does follow the limitations of Lambert-
Beer law, described in Chapter 2. In numerical modeling, this value is
not a result of light ray-tracing, but a mathematical description of the
phenomena using multiple approximations (e.g. specic extinction
coecient, K value). Despite this, for the purpose of this study the
change of visibility caused by a change in fuel soot yield could be
assessed, as subsequent cases were estimated with the same mathe-
matical approach, and the only variable was the soot yield value. This
approach was chosen for further parametric analysis.
The numerical analysis, both for ANSYS Fluent and FDS, show an
W. Wgrzyski, G. Vigne
hyperbolic trend of the Visibility when increasing the soot yield with a
cut-o point around 0.10 g/g where a variation of the soot yield can
produce a relevant (below 0.10 g/g) or a negligible (above 0.10 g/g)
change in the Visibility.
One of the major ndings of this study is that unless the materials
(and thus the chemical reactions) in the building to be assessed by a
Fire Engineering analysis are well known, the value of soot yield below
approximately 0.10 g/g should be used with extremely cautiousness
when performing an ASET/RSET exercise.
References
[1] G. Vigne, To what extent can we rely upon the results obtained from advanced
smoke modelling, SFPE Cph. (2015).
[2] K. McGrattan, S. Hostikka, R. McDermott, J. Floyd, C. Weinschenk, K. Overholt,
Fire Dynamics Simulator Users Guide, Sixth Edition. http://dx.doi.org/10.6028/
NIST.SP.1019, 2016.
[3] ANSYS, ANSYS Fluent 14.5.0 - Technical Documentation (2014).
[4] K. McGrattan, S. Miles, Modeling Fires Using Computational Fluid Dynamics
(CFD)SFPE Handb. Fire Prot. Eng., Springer, New York, New York, NY, 2016, pp.
10341065. http://dx.doi.org/10.1007/978-1-4939-2565-0_32.
[5] K. McGrattan, R. McDermott, J. Floyd, S. Hostikka, G. Forney, H. Baum,
Computational uid dynamics modelling of re, Int. J. Comut. Fluid Dyn. 26
(2012) 349361. http://dx.doi.org/10.1080/10618562.2012.659663.
[6] B. Merci, Introduction to Fluid MechanicsSFPE Handb. Fire Prot. Eng., Springer
New York, New York, NY, 2016, pp. 124. http://dx.doi.org/10.1007/978-1-4939-
2565-0_1.
[7] P. Too, W. Wgrzyski, R. Porowski, Hand Calculations, Zone Models and CFD
Areas of Disagreement and Limits of Application in Practical Fire Protection
Engineering, in: Proceedings of the 11th Conference Performance-Based Codes Fire
Saf. Des. Methods, SFPE, 2016.
[8] Society of Fire Protection Engineers, SFPE Eng. Guide to Performance-Based Fire
Protection, 2nd Ed., 2007.
[9] J.F. Widmann, Evaluation of the planck mean absorption coecients for radiation
transport through smoke, Combust. Sci. Technol. 175 (2003) 22992308. http://
dx.doi.org/10.1080/714923279.
[10] T.G. Mayerhfer, H. Mutschke, J. Popp, Employing theories far beyond their limits
- the case of the (Boguer-)Beer-Lambert law, ChemPhysChem (2016). http://
dx.doi.org/10.1002/cphc.201600114.
[11] M. Marchetti, V. Boucher, J. Dumoulin, M. Colomb, Retrieving visibility distance in
fog combining infrared thermography, Principal Components Analysis and Partial
Least-Square regression, Infrared Phys. Technol. 71 (2015) 289297. http://
dx.doi.org/10.1016/j.infrared.2015.05.002.
[12] R. Nebuloni, Empirical relationships between extinction coecient and visibility in
fog, Appl. Opt. 44 (2005) 37953804. http://dx.doi.org/10.1364/AO.44.003795.
[13] F. Nadeem, T. Javornik, E. Leitgeb, V. Kvicera, G. Kandus, Continental fog
attenuation empirical relationship from measured visibility data, Radioengineering
398
Fire Safety Journal 91 (2017) 389398
19 (2010) 596600.
[14] G.W. Mulholland, C. Croarkin, Specic extinction coecient of ame generated
smoke, Fire Mater. 24 (2000) 227230. http://dx.doi.org/10.1002/1099-
1018(200009/10)24:5 < 227::AID-FAM742 > 3.0.CO;2-9.
[15] G.W. Mulholland, Smoke Production and PropertiesSFPE Handb. Fire Prot.
Eng.Fourth edi, NFPA & SFPE, Quincy, MA, 2008.
[16] M. Konecki, M. Pka, Calculation of visibility in smoke generated during thermal
decomposition and combustion of polyester materials, in: Mod. build. Mater.
struct. tech. MBMST 2010, Vilnius (2010).
[17] M. Konecki, M. Pka, Simple re model for comparative studies of critical
conditions during combustion of chosen polymer materials, J. Civ. Eng. Manag. 15
(2009) 247257. http://dx.doi.org/10.3846/1392-3730.2009.15.247-257.
[18] T. Jin, Visibility through re smoke (II), Bull. Fire Prev. Soc. Jpn. 21 (1971).
[19] T. Jin, Visibility and Human Behavior in Fire Smoke, in: SFPE Handb. Fire Prot.
Eng., 3rd Editio: pp. 2-422-53, 2002.
[20] T. Jin, Visibility through re smoke (I), Bull. Fire Prev. Soc. Jpn. 19 (1970).
[21] T. Jin, Visibility through re smoke, J. Fire Flammabl. 9 (1978) 135155.
[22] T. Jin, Visibility through re smoke (III), Bull. Fire Prev. Soc. Jpn. 22 (1972).
[23] T. Yamada, Y. Akizuki, Visibility and Human Behavior in Fire SmokeSFPE Handb.
Fire Prot. Eng., Springer, New York, New York, NY, 2016, pp. 21812206. http://
dx.doi.org/10.1007/978-1-4939-2565-0_61.
[24] W.D. Walton, D.J. Carpenter, C.B. Wood, Zone Computer Fire Models for
EnclosuresSFPE Handb. Fire Prot. Eng., Springer New York, New York, NY, 2016,
pp. 10241033. http://dx.doi.org/10.1007/978-1-4939-2565-0_31.
[25] L.Y. Cooper, A Mathematical Model for Estimating Available Safe Egress Time in
Fires, Fire Mater. 6 (1982) 135144 (doi:0.1002/fam.810060307).
[26] H.W. Emmons, The prediction of res in buildings, Symp. Combust., 17 , pp.
11011111. doi:http://dx.doi.org/10.1016/S0082-0784(79)80105-8, 1979.
[27] G. Vigne, W. Wgrzyski, Experimental and numerical analysis of the inuence of
the soot yield when conducting Cfd analysis for the estimation of the available safe
evacuation time, in: Interam, 2016, 2016.
[28] G. Vigne, W. Wgrzyski, Inuence of Variability of Soot Yield Parameter in
Assessing the Safe Evacuation Conditions in Advanced Modeling Analysis. Results
of Physical and Numerical Modeling Comparison, in: Proceedings of the 11th
Conference Performance-Based Codes Fire Saf. Des. Methods, SFPE, 2016.
[29] D.J. Rasbash, D.D. Drysdale, Fundamentals of smoke production, Fire Saf. J. 5
(1982) 7786. http://dx.doi.org/10.1016/0379-7112(82)90008-X.
[30] W. Wgrzyski, G. Krajewski, P. Suchy, Zmienno cech materiaowych a szaco-
wany zasig widzialnoci w analizach CFD, in: spraw. Rocz. - NP-105, Instytut
Techniki Budowlanej, 2015.
[31] G. Vigne, J. Jnsson, J. Stern-Gottfried, The Use of CFD Modelling to Predict
Smoke Movement in Tunnels, in: International Congr. Smoke Control Build.
Tunnels, Santander, 2007.
[32] W. Poh, Tenability criteria for design of smoke hazard management systems,
Ecolibrium 8 (2011).
[33] W. Poh, Tenability in building res: limits and design, Fire Aust. (2010).
[34] W. Wgrzyski, G. Krajewski, Systemy wentylacji poarowej garay. Projektowanie,
ocena, odbir, 493/2015, Instytut Techniki Budowlanej, 2015.
[35] P. Rubini, Q. Zhang, J. Moss, Simulation of visibility in smoke laden environments,
Fire Sci. Eng. Univ. (2007) 35 http://citeseerx.ist.psu.edu/viewdoc/download?
doi=10.1.1.106.8079 & rep=rep1 & type=pdf.