Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

Fire Technology, 54, 1067–1091, 2018

Ó 2018 Springer Science+Business Media, LLC, part of Springer Nature.


Manufactured in The United States
https://doi.org/10.1007/s10694-018-0725-z

Factors Affecting the Make-Up Air


and Their Influence on the Dynamics
of Atrium Fires

Pablo Ayala, Institute for Research in Technology, Escuela Te´cnica Superior de


Ingenierı´a, Universidad Pontificia Comillas, C/ Santa Cruz de Marcenado
26, 28015 Madrid, Spain
Alexis Cantizano* , Institute for Research in Technology, Escuela Te´cnica
Superior de Ingenierı´a, Universidad Pontificia Comillas, C/ Santa Cruz de
Marcenado 26, 28015 Madrid, Spain
Guillermo Rein, Department of Mechanical Engineering, Imperial College
London, London, UK
Cándido Gutie´rrez-Montes, Fluid Dynamics Division of the Department of
Mining and Mechanical Engineering, Universidad de Jae´n, Jae´n, Spain

Received: 10 November 2017/Accepted: 18 April 2018

Abstract. In case of fire, constructive features of typical atria could favor the spread
of smoke. This makes the design of their smoke control and management systems a
challenging task. Five full-scale fire experiments in the literature have been analyzed
and numerically compared in FDS v6 to explore the influence of the make-up air.
However, these fire experiments cover only a limited number of set-ups and condi-
tions, and require further numerical modeling to obtain a deeper understanding of
the makeup air influence. Subsequently, 84 simulations with FDS v6 have been car-
ried out, considering different vent areas (air velocity from 0.4 to 5.3 m/s) and config-
urations, two heat release rates (2.5 and 5 MW), and two pan locations. It is
demonstrated that make-up air velocities lower than the prescribed limit of 1 m/s, by
the international codes, may induce adverse conditions. Based on our results, we rec-
ommended fire engineers to numerically assess the fire scenario with even lower
velocity values. The results also show that asymmetric configurations are prone to
induce circulation around the flame which can contribute to the formation of longer
flames and fire whirls. Thus, this numerical study links two fire types allowing the
connection of pool fires to fire whirls, which completely differ in behaviour and
smoke filling, for the sake of design of fire safety.

Keywords: Full-scale fire experiment, Fire dynamics simulator (FDS), Make-up air, Smoke manage-
ment, Fire whirls

* Correspondence should be addressed to: Alexis Cantizano, E-mail: alexis.cantizano@comillas.edu

1
1068 Fire Technology 2018

1. Introduction
Nowadays it is very common to find atria, large volume spaces, or big construc-
tions in modern buildings. In case of fire, catastrophic casualties and loss of prop-
erty may be caused if not detected before growing into an uncontrollable size,
being the toxic gases in the smoke one of the major causes of fatalities [1]. There-
fore, the smoke filling process has been widely investigated to analyze smoke ven-
tilation performance in large atria [2–7]. More specifically, Klote [8] described that
make-up air, fresh air mechanically or naturally supplied through the openings, is
an important factor in atrium mechanical smoke extraction system. In this regard,
the main international codes limit the make-up air velocity towards the fire to a
value of 1 m/s to avoid flame disturbances unless a higher velocity could be vali-
dated by an engineering analysis [9, 10].
Hadjisophocleous and Zhou [11] found numerically that make-up air velocities
around 1 m/s may deflect the flame, increasing the smoke production in the fire
plume. This phenomenon was reported to be more noticeable for atria with
heights below 20 m. In relation to this study, Zhou and Hadjisophocleous [12]
performed a numerical research on the parameters that have an impact on the fire
plume, showing that velocities of 1 m/s can tilt the flame of even 5 MW fires.
Moreover, the influence of the opening distributions and their heights with
respect to the smoke layer interface, as well as the make-up air velocity on the
fire-induced conditions in atria, was studied by Hadjisophocleous and Lougheed
[13], Yi et al. [14] and Kerber and Milke [15]. These works recommend to locate
the openings below the smoke layer interface at the steady state in order to avoid
mixing as well as to introduce clean air directly to the smoke layer. Moreover,
symmetric make-up air distribution is advisable with velocities below 1 m/s in
order to avoid flame deflections that entail larger smoke production. Finally, they
proposed to diffuse the make-up air to reduce its velocity in the vicinity of the
flame, and then, to diminish its effect. More recently, Pongratz et al. [16] con-
cluded that make-up air injected at lower elevations creates a more extreme smoke
condition within the plume and smoke layer in smaller atria, mainly with heights
lower than 10 m.
The use of computational models (CFD) to predict fire dynamics and smoke
movement has increased importantly, being Fire Dynamic Simulator (FDS) the
most used one [17]. In fact, these numerical models allow detailed analysis which
may verify exceptions to the inlet velocity limitation or smoke layer height
requirements. However, most of the previous studies did not consider the domain
extension of the openings that later Zhang et al. [18] and Gutiérrez et al. [19]
proved to be crucial in a small enclosure facility and in an atrium, respectively.
Both studies included comparisons between numerical and experimental data.
Under the circumstances aforementioned, the main aim of this paper is to per-
form a comprehensive study on the effect of different venting conditions on the
smoke behaviour in atria. Five full-scale tests are presented and analysed in
Sect. 2. A numerical model is described and validated with the experimental mea-
surements. Subsequently, in Sect. 3, a numerical study is carried out to explore the
influence of the make-up air velocity, the pan location, and the inlet vents distri-
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1069

bution. The results achieved are discussed in Sect. 4, focused on the effect of dif-
ferent parameters on the fire induced inner thermal and flow fields. Finally, the
last section is devoted to the conclusions.

2. Fire Tests and Numerical Model Validation


In this section, we will first use the experimental data to both validate the numeri-
cal model and roughly analyze the influence of different parameters affecting the
make-up air inlet conditions. Thus, the results from five full-scale atrium experi-
ments have been used to assess the make-up air influence on the fire induced inner
conditions (Table 1).The experiments have been conducted in the same experimen-
tal facility, the Fire Atrium of Murcia, Spain [20–24]. For a complete description
of the experimental set-up, see Ayala et al. [20].
Basically, the atrium is a 19.5 m 9 19.5 m 9 19.5 m facility (Fig. 1), with a
pyramidal shaped roof of thin steel walls and ground made of concrete. Four fans
at the roof, each of them of two speeds, can generate different exhaust rates and
the air inlet vents at the atrium base can be partially opened. The facility was
instrumented with 69 bare thermocouples. Mainly the time evolution of the tem-
perature at the near field, i.e. the fire plume, the far field, as well as the smoke
layer height, through a thermocouple tree, were registered. The smoke layer height
was characterized with the thermocouple tree measurements, by means of the
least-square method [14, 25] as well as the n-percent (N=30%) method [15, 26].
The fire consisted of a heptane pool-fire of different diameters, thus with different
values of HRR that were evaluated by the fuel mass loss rate measured with load
cells. The pan was placed at the centre (C) or at a diagonal of the base (D).

2.1. Numerical Model


As it has been previously commented, Fire Dynamic Simulator (version 6, Release
Candidate 3) has been used to carry out the numerical simulations. The models used
to account for combustion, turbulence and radiation are the Eddy Dissipation Con-

Table 1
Summary of Experimental Conditions During the Fire Tests.
Table Based on Data from [20]

Fire tests: #1 #2 #3 #4 #5

HRR (MW) 2.5 2.6 1.8 2.6 2.9


Pan location C D C C C
Burning time (s) 647 625 708 646 524
Inlet Configurations Sym.1 Asym.2 Asym. Sym. Sym.
Opening Area (m2 ) 48.8 24.4 24.4 24.4 12.2
Flow rate (m3 =s) 18.3 18.3 18.3 32.1 27.5
Air velocity (m/s) 0.4 0.75 0.75 1.3 2.25

1
A1, A3, C1, C2 open
2
A1, C1 open
1070 Fire Technology 2018

Figure 1. Fire Atrium sketch.

cept (EDC) model with a thermal extinction model, the Deardorff model (Cv = 0.1)
and the radiation transport equation with 100 radiation angles, respectively.
The atrium geometry has been reproduced with the make-up air inlets as open
vents. The fans have been modelled imposing the flow rate curves corresponding
to the exhaust values and the outer atmospheric conditions are set considering
quiescent atmosphere. The grid size has to be small enough to properly model the
turbulence effects and, for the LES method, a spatial resolution between
1=4 > R > 1=16 is recommended [27], where the spatial resolution is defined as
R ¼ D=D , being D the element size, and D the characteristic diameter of the
plume, [28]. The element size chosen for all the simulations performed was equal
to 13 cm, which corresponds to a spatial resolution from 1/11 (for a fire of 2.6
MW) to 1/9 (for a fire of 1.7 MW) within the recommended range. The walls and
the roof were modelled as 6 mm walls of thermally thick galvanized steel; the
ground floor was simulated as a thermally thick layer of concrete [20]. The pool
fires were simulated by introducing the estimated experimental HRR curves, based
on the measured mass loss rate, as an input (Fig. 2). As can be seen, test #3 has a
particular behaviour which will be explained in Sect. 2.2. Finally, the radiation
fraction for the heptane has been considered 0.35, [22].

2.2. Numerical Model Validation


In the present subsection, the proper simulation of the five tests is pursued. First,
the configuration of an appropriate computational domain is addressed as well as
the description of the different induced flow patterns.
The inclusion of domain extensions in fire modelling plays an important role
when the fire conditions or the smoke generated are clearly influenced by the
boundary conditions. The lack of domain extensions at the openings to the atmo-
sphere may eliminate physical fluid phenomena such as the vena contracta effect,
modifying the inflow patterns which directly influence the flame, [19]. Moreover,
flame deflections produce a wider fire plume and a higher degree of turbulence,
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1071

7.5

7 test #1
6.5 test #2
test #3
6 test #4
5.5 test #5

5
Heat Release Rate (MW)

4.5

3.5

2.5

1.5

0.5

0
0 100 200 300 400 500 600 700
t (s)

Figure 2. Experimental HRR curves. Figure based on data from [20].

with the subsequent larger production of smoke, which has a direct impact on the
smoke layer interface. Therefore, a wrong numerical implementation can provide
with unrealistic fire scenarios precluding from a proper hazard evaluation.
The length of the extended domains in the numerical model of the atrium, which
are located at the openings, is large enough to accomplish the minimum extension
proposed by Gutiérrez et al. [19], L=Deq  0.72, where L is the domain extension
length and Deq is the equivalent diameter of the vent. The same length value (L =
3.25 m) has been used for the extensions of the different simulations herein reported.
As a consequence of the latter, Fig. 3 shows the characteristic flow patterns, at
a height h = 1.3 m and time t = 300 s, obtained for the five test models, under

Test #1 Test #2 Test #3 Test #4 Test #5

V [m/s]

1.5

0.5

Figure 3. In-plane velocity vectors 1.3 m high. Top figures are with
domain extensions and bottom figures are without domain
extensiones. From left to right: tests #1, #2, #3, #4, #5, at t = 300 s.
1072 Fire Technology 2018

different symmetric and asymmetric inlet distributions as well as with different


inlet velocities ranged from 0.4 to 2.25 m/s. As it can be appreciated in the differ-
ent flow patterns, the simulations with the domain extensions present higher, but
more realistic, velocities, which cause greater flame perturbations. This has a
direct impact on the results as it will be later shown. In case of test #1, which has
a symmetrical configuration and a low make-up air velocity (vinlet = 0.4 m/s), the
domain extension effect is hardly noticeable, but as the inlet velocity increases for
asymmetric inlet configurations, or when the complete fire scenario is not symmet-
ric, the domain extensions have a greater impact on the results obtained.This can
be clearly appreciated when the pan is located at the diagonal and close to an
opening (test #2) and for asymmetric configurations (test #3).
The last test, test #3, is of special interest as the simulation performed with the
extended domain permits capturing the flame whirls observed experimentally
(Fig. 2). An extensive review on the formation of these whirls can be found in
Gollner et al. [29]. In this case, a circular flow pattern appears at the early stages
of the fire, increasing the azimuthal velocity around the flame, and reaching even-
tually a maximum value of near 5 m/s from which it remains stable during the
rest of the simulation. On the contrary, the no inclusion of the extensions pro-
vides a hardly perturbed in-plane velocity field, with no significant effect on the
fire. As a result, when including the extended domain is possible to predict the
experimentally observed smoke layer flow perturbation and mixing caused by the
flame self-rotation.
Finally, the tests with symmetric make-up air distributions, and with velocities
larger than 1 m/s (test #4 and test #5), present a particular behaviour around the
flame when the domain extensions are included, i.e. the inlet air flows from the
confronted vents collide, generating large flame disturbances which are not pre-
dicted with the model without outer domain. On the contrary, the large momen-
tum of the air streams impedes the formation of a flow around the flame, which
would eventually lead to the formation of a fire whirl.
In short, quite different flow patterns have been observed when the numerical
domain is extended. As it will be shown below when comparing with the experi-
mental data, the simulations with extensions agree better with the measurements,
which indicates that the inclusion of extensions provides more realistic and physi-
cal solutions (Fig. 5). In fact, these domain extensions are required when the inlet
velocity is above the limit prescribed by the international codes and when the
openings distribution may cause circular flow patterns, even with inlet velocities
below 1 m/s. These different flow patterns affect directly the smoke production
throughout their influence on the flame as well as on the fire plume. Therefore,
the temperature profiles, as well as the smoke layer drop, are also altered.
In this context, once the importance of including the outer domain and its influ-
ence on the induced flow patterns and the possible formation of a flame whirl
have been commented, an experimental-numerical comparison of different make-
up air configurations (symmetric and asymmetric) and inlet velocities (varying the
cross-sectional area of the openings) is briefly performed in the next subsections,
considering the fire induced inner conditions as well as the smoke production, i.e.
the smoke layer interface.
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1073

(a) (b) (c)


300 80 20
Exp. Exp. L-S Exp.
250 Sim. Sim. n-% Exp.
Temperature (°C)

Temperature (° C)
Sim. ext. 60 Sim. ext. 15 L-S Sim.
200 L-S Sim. Ext.

Height (m)
150 40 10

100
20 5
50

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 4. Predictions of test #1 without and with domain extensions.


Figure based on data from [20]. (a) Temperature in the fire plume at
h = 6.62 m; (b) Temperature in the wall A at h = 5 m; (c) Smoke layer
drop.

2.2.1. Inlet Vents Distribution The temperatures at the fire plume, at a height of h
= 6.62 m, and in the far field, close to the wall A, at a height of h = 5 m in the
vicinity of the smoke layer position during the final stages are compared in Fig. 4,
for test #1. This test evolved as a pool fire in open air which can be clearly appre-
ciated at the fire plume, Fig. 4 a. Thus, the smoke layer forms and growths pro-
gressively without any perturbation affecting it. In this case, as it can be observed,
the domain extensions have a small impact, due to the symmetric venting configu-
ration and the low inlet velocities (vinlet = 0.4 m/s). In general, both simulations
agree well with the experimental data. As for the smoke layer drop, again good
agreement is achieved, with both simulations presenting the same behavior for
both the n-percent and the least square methods.
Moreover, the influence of the domain extensions in tests #2 and #3, Fig. 5, in
which the make-up air distribution is asymmetric and the inlet velocity is rela-
tively low (vinlet = 0.75 m/s) becomes more significant and necessary.As the diffu-
sion of the make-up air is lower inside the atrium with the domain extensions, the
make-up air velocity increases in the surroundings of the vents, as it can be
observed in the high values of air velocity through these openings in Fig. 3. In
both tests, the flow patterns at the bottom of the atrium obtained with and with-
out extensions differ significantly from the one observed in test #1, with larger dis-
turbances on the flame and then more smoke production. The impact of the
proper inlet air conditions is significant as it can be noticed in the temperature
field and smoke layer height time evolutions. In test #2, the simulations show a
more accurate prediction with the domain extended for the temperatures at the
wall close to the fire, Fig. 5a. In this case, the larger reach of the inlet air stream
impinging the fire pool increases the mixing of smoke across the smoke layer
interface, with the consequent temperature increase at the lower heights. It is also
specially relevant how the inclination of the flame is properly well predicted with
1074 Fire Technology 2018

(a) (b) (c)


80 80 20
Exp. Exp. L-S Exp.
Sim. Sim.
Temperature (°C)

Temperature (°C)
n-% Exp.
60 60 15
Sim. ext. Sim. ext. L-S Sim.

Height (m)
L-S Sim. Ext.
40 40 10

20 20 5

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(d) (e) (f)


300 80 20
Exp. Exp. L-S Exp.
250
Sim. Sim.
Temperature (°C)

Temperature (°C)

n-% Exp.
60 15
Sim. ext. Sim. ext. L-S Sim.

Height (m)
200
L-S Sim. Ext.
150 40 10

100
20 5
50

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 5. Predictions of tests #2 (a–c) and #3 (d–f) without and with


domain extensions. Figure based on data from [20]. (a) Temperature
in the wall C at h = 5 m; (b) Temperature in the wall A at h = 7.5 m;
(c) Smoke layer drop; (d) Temperature in fire plume at h = 6.62 m;
(e) Temperature in the wall A at h = 5 m; (f) Smoke layer drop.

the domain extensions in test #3 (Fig. 5d), in which the vents layout induce the
appearance of fire whirls, generating consequently more smoke, as can be seen in
Fig. 5e, f. In this case, the thermocouple placed at the fire plume presents at the
early states, e.g. t ¼ 150 s, a sudden temperature drop, caused by the flame
revolving around the central axis of the pool fire. Then, once the circular flow
around the fire induces the fire whirls, large and fast temperature increments are
observed, corresponding to the fire whirls formed [20]. The effect on the smoke
temperature field and the mixing across the smoke interface are also predicted in
this case, Fig. 5e, f. However, it has to be noticed that due to the enhanced mix-
ing, specially for the asymmetric configuration, the smoke layer height estimation
presents larger uncertainty.

2.2.2. Inlet Velocity Tests #4 and #5, which were conducted with a symmetric
make-up air distribution but with higher values of velocities (1.3 and 2.25 m/s,
respectively), present also the necessity of the use of the extended domain, Fig. 6.
In these cases, the diffusion of the make-up air is again important. Therefore,
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1075

(a) (b) (c)


300 60 60

250 50 50
Temperature (°C)

Temperature (° C)

Temperature (° C)
200 40 40

150 30 30

100 20 20
Exp
FDS
50 10 10
FDS ext
0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(d) (e) (f)


300 60 60
Exp
250 50 50
FDS
Temperature (° C)

Temperature (° C)

Temperature (° C)
FDS ext
200 40 40

150 30 30

100 20 20

50 10 10

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 6. Predictions of tests #4 (a–c) and #5 (d–f) without and with


domain extensions. Figure based on data from [24]. (a) and (d)
Temperature in the fire plume at h = 6.62 m; (b) and (e) Temperature
in the far field at h = 6.0 m; (c) and (f) Temperature in the far field at
h = 6.5 m.

when the make-up air velocity is high and the openings configuration is symmet-
ric, the confronted make-up air flows reach each other, and consequently, the
flame is disturbed with the following increment in the smoke production. Those
effects can be appreciated in the temperature profile over the fire plume (Fig. 6a
and d), where the temperature predicted is lower due to flame deflections, and
more importantly on the far field (Fig. 6b, c, e and f), where the temperature is
affected by the perturbations on the flame and the boosted mixing across the
smoke interface as a consequence of the flame tilt.

3. Numerical Study on the Make-Up Air Supply


The main international codes limit the make-up air velocity to 1 m/s, recom-
mending also a non-direct stream which can deflect the flame [9, 10]. Moreover,
the openings should be located below the smoke layer interface, otherwise the
make-up air enters the smoke layer, increasing the quantity of smoke and the mix-
ing in the interface of the smoke layer.
1076 Fire Technology 2018

Figure 7. Different make-up air distributions.

Additionally, there can be other parameters that may entail a risk, such as the
opening distribution and the pan location, [13, 15, 23]. In this regard, this section
presents a numerical study of the make-up air supply in order to study the influ-
ence of the inlet velocity, the pan location and the inlet distribution. In all, a
parametric study consisting of 84 different simulations has been performed. The
simulations are carried out based on the numerical model previously described
and validated, including extensions in the domain at the openings. In this case the
HRR of the fire has been modelled as a piecewise linear function, i.e. starting with
a value of 0, at 10.7 s of simulation the HRR is equal to the 62.4% of its nominal
value, at 81.5 s the 89.1%, and at 200 s the 100%, remaining constant until the
end of the fire test, [21, 22, 26].
In the first set of simulations, symmetric and asymmetric opening configurations
are assessed with the pool fire located at the centre of the atrium, as it can be
observed in Fig. 7. On the one hand, three symmetrical make-up air configura-
tions, with the vents located at the bottom of the atrium, (sym. 1, 2 and 3) have
been considered, displacing progressively the openings towards the centre of the
atrium, respectively. The sym. 1 configuration corresponds to tests #1, #4, and #5.
On the other hand, two asymmetric configurations (asym. 1 and 2) have been
assessed, being the main difference between them the distance from the centre of
the atrium, as in the previous symmetric configurations. The asym. 1 configura-
tion corresponds to test #3. Additionally, these vents have been modelled fully
opened, and partially opened (50% and 25%) in both configurations (symmetric
and asymmetric). All of these configurations are carried out with two different
values of HRR (2.5 and 5 MW) and two exhaust flow rates (V_ ¼18.3 and
V_ ¼32.1 m/s3 ). In short, a total of 60 simulations have been run, which are shown
in tables 2 and 3.
In the second set of simulations, tests 61–84 in table 4, a total of 24 simulations
are presented, with fire conditions similar to those of the first set, but with the dif-
ference that the pan is located at the diagonal, as in test #2. In summary, sym. 1
and asym. 1 configurations are assessed with fully opened and partially opened
vents (50% and 25%) under two HRRs (2.5 and 5 MW) and with two exhaust
flow rates (V_ ¼18.3 and V_ ¼32.1 m=s3 ).
Tables 2-4 also includes the average results of the smoke layer height (SL),
assessed by FDS [30], the fire plume temperature at h = 7.25 m and the wall A
temperature at h = 5 m.
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1077

Table 2
Results of Symmetrical Models

No. HRR (MW) Open vents vinlet (m/s) V_ (m3 =s) SL (m) TFP ( C) TWA ( C)

1 2.5 Sym.1 (100%) 0.7 32.1 7.9 128 26


2 5.0 Sym.1 (100%) 0.7 32.1 6.7 243 42
3 2.5 Sym.1 (100%) 0.4 18.3 5.2 150 42
4 5.0 Sym.1 (100%) 0.4 18.3 4.1 297 113
5 2.5 Sym.1 (50%) 1.3 32.1 6.8 90 28
6 5.0 Sym.1 (50%) 1.3 32.1 5.4 175 60
7 2.5 Sym.1 (50%) 0.7 18.3 5.1 151 49
8 5.0 Sym.1 (50%) 0.7 18.3 3.4 275 118
9 2.5 Sym.1 (25%) 2.6 32.1 5.6 66 35
10 5.0 Sym.1 (25%) 2.6 32.1 4.3 135 90
11 2.5 Sym.1 (25%) 1.5 18.3 4.4 126 63
12 5.0 Sym.1 (25%) 1.5 18.3 3.3 279 120
13 2.5 Sym.2 (100%) 0.7 32.1 8.5 167 27
14 5.0 Sym.2 (100%) 0.7 32.1 7.0 284 43
15 2.5 Sym.2 (100%) 0.4 18.3 5.9 182 39
16 5.0 Sym.2 (100%) 0.4 18.3 3.7 301 118
17 2.5 Sym.2 (50%) 1.3 32.1 5.6 87 34
18 5.0 Sym.2 (50%) 1.3 32.1 6.7 238 43
19 2.5 Sym.2 (50%) 0.7 18.3 5.6 160 40
20 5.0 Sym.2 (50%) 0.7 18.3 3.6 289 118
21 2.5 Sym.2 (25%) 2.6 32.1 4.5 66 48
22 5.0 Sym.2 (25%) 2.6 32.1 5.4 195 55
23 2.5 Sym.2 (25%) 1.5 18.3 4.8 136 58
24 5.0 Sym.2 (25%) 1.5 18.3 3.6 300 118
25 2.5 Sym.3 (100%) 0.7 32.1 8.4 150 26
26 5.0 Sym.3 (100%) 0.7 32.1 6.5 255 44
27 2.5 Sym.3 (100%) 0.4 18.3 5.5 169 40
28 5.0 Sym.3 (100%) 0.4 18.3 3.5 293 115
29 2.5 Sym.3 (50%) 1.3 32.1 6.9 104 32
30 5.0 Sym.3 (50%) 1.3 32.1 5.7 217 45
31 2.5 Sym.3 (50%) 0.7 18.3 5.8 170 39
32 5.0 Sym.3 (50%) 0.7 18.3 3.6 293 105
33 2.5 Sym.3 (25%) 2.6 32.1 6.1 54 45
34 5.0 Sym.3 (25%) 2.6 32.1 4.0 89 76
35 2.5 Sym.3 (25%) 1.5 18.3 5.2 153 49
36 5.0 Sym.3 (25%) 1.5 18.3 2.7 256 109


Averaged value during the last 100 s

Averaged value after 200 s until the end of the simulation

4. Results of the Numerical Analysis


Based on the aforementioned three series of simulations, the three parameters that
influence most the smoke control and management of an atrium fire, regarding
the make-up air inlet conditions, are independently assessed: the inlet velocity, the
pan location and the inlet distribution. In order to compare them, showing the
most significant varying figures in the sake of clarity and simplicity, the tempera-
ture profiles as well as the smoke layer drop are next assessed.
1078 Fire Technology 2018

4.1. Inlet Velocity


In the set of simulations, inlet velocities from 0.4 to 5.3 m/s have been considered
by means of the variation of the opening cross-sectional area, the number of
openings, and the exhaust flow rate. The lower velocities are induced with larger
areas under symmetric configurations for the lowest exhaust rate, whereas the
higher velocities are reached with asymmetric ventilation, small opening cross-sec-
tional areas, and the largest exhaust rate. To better study the influence of the inlet
velocity, the simulations of the different configurations, i.e. the simulations of
symmetric and asymmetric configurations should be analysed separately.
Figure 8a–i show the temperature at h = 7.25 m in the fire plume (Fig. 8a, d,
g), at h = 5 m high close to the wall A (Fig. 8b, e, h), and the smoke layer drop
(Fig.c, f, i) for the make-up air configurations sym. 1, sym. 3 and asym. 1, with an
exhaust flow rate of 32.1 m3 =s and a HRR of 5 MW.
Different behaviours can be identified. First of all, the sym. 1 configuration is
assessed with the opening sections of 100, 50 and 25% (simulations 2, 6 and 10)
which correspond to make-up air velocities of 0.7, 1.3 and 2.6 m/s, respectively,
i.e. large inlet velocity variation. As it can be observed, at the fire plume tempera-
ture, Fig. 8a, the temperature decreases with the reduction of the opening areas,
i.e. increase of the inlet velocity. The lower predicted temperature obtained, with
higher inlet velocities, is caused by the enhancement of flame perturbations, which
at the same time produces a larger amount of smoke. The latter can be appreci-
ated at the far field, at h = 5 m close to the wall A, Fig. 8b, where an increment
on the inlet velocity is associated with a higher temperature due to the increment
in the aforementioned smoke production. Finally, this behaviour is verified on the
smoke layer drop, Fig. 8c, being more noticeable when the inlet velocity varies
from 0.7 to 1.3 m/s with differences of 1.4 m. Nevertheless, when the velocity
changes from 1.3 to 2.6 m/s the difference reached is 1.1 m, thus as recommended
larger inlet velocities may induce flame perturbations and enhance the smoke pro-
duction, inducing consequently more hazardous scenarios.
Similarly, in case the HRR of sym. 1 configuration is reduced to 2.5 MW (sim-
ulations 1, 5 and 9), the effect of the increment on the make-up air velocity pre-
sents the same trends. However, at the steady state, a difference of 1.1 m is
obtained when increasing the make-up air velocity from 0.7 to 1.3 m/s and of 1.2
m when changing from 1.3 to 2.6 m/s, Table 2.
Secondly, Fig. 8d–f show the influence of the make-up air velocity on the sym.
3 configuration, i.e. symmetric configuration with the openings located at the cen-
tre of the walls, Fig. 7, (simulations 26, 30, and 34). It can be observed that the
flame in this configuration is remarkably affected by the highest make-up air
velocity, Fig. 8d. In the case of the highest velocity, the strong jets directed to the
fire deflect it, which induce important instabilities with direct impact on the smoke
layer height. For the remaining two make-up air velocities, 0.7 and 1.3 m/s, smal-
ler temperature variations are predicted overall, and as in sym. 1 configuration, an
increment of the make-up air velocity entails flame deflections, being observable as
a reduction in the fire plume temperature. The smoke layer interface reached at
the steady state is lower with the higher velocity, being this difference of 0.8 m.
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1079

(a) (b) (c)


400 140 20

120 100%, 0.7 m/s


Temperature (°C)

Temperature (°C)
50%, 1.3 m/s
300 15
100 25%, 2.6 m/s

Height (m)
80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(d) (e) (f)


400 140 20

120 100%, 0.7 m/s


Temperature (°C)

Temperature (°C)

50%, 1.3 m/s


300 15
100 25%, 2.6 m/s

Height (m)
80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(g) (h) (i)


400 140 20

120 100%, 1.3 m/s


Temperature (°C)

Temperature (°C)

50%, 2.6 m/s


300 15
100 25%, 5.3 m/s
Height (m)

80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 8. Symmetric 1, symmetric 3 and asymmetric 1 configurations


with an exhaust flow rate of 32.1 m3 =s and a 5 MW. (a), (d) and (g)
Fire plume temperature at h = 7.25 m; (b), (e) and (h) wall A
temperature at h = 5 m; and (c), (f) and (i) smoke layer drop.

The latter differences are more noticeable when the HRR decreases thus becom-
ing more sensitive to any outer phenomena (simulations 25, 29, and 33). The
smoke layer drop presents a maximum difference of 1.5 m.
Differently, when the exhaust flow rate is reduced to 18.3 m3 =s, being the inlet
velocities equal to 0.4, 0.7, and 1.5 m/s, i.e. small inlet vent variation, the results
slightly differ confirming that the limit value of 1 m/s depends also on the configu-
ration and on the value of the HRR. Thus, for a HRR of 2.5 MW, the value of
1.5 m/s clearly generates more smoke than with the lower values, except for sym.
1080 Fire Technology 2018

Table 3
Results of Asymmetrical Models

No. HRR (MW) Open Vents vinlet (m/s) V_ (m3 =s) SL (m) TFP ( C) TWA ( C)

37 2.5 Asym.1 (100%) 1.3 32.1 5.4 70 31


38 5.0 Asym.1 (100%) 1.3 32.1 4.0 95 91
39 2.5 Asym.1 (100%) 0.7 18.3 4.7 95 57
40 5.0 Asym.1 (100%) 0.7 18.3 2.6 123 123
41 2.5 Asym.1 (50%) 2.6 32.1 6.3 49 26
42 5.0 Asym.1 (50%) 2.6 32.1 4.3 93 61
43 2.5 Asym.1 (50%) 1.5 18.3 4.0 77 66
44 5.0 Asym.1 (50%) 1.5 18.3 2.8 123 120
45 2.5 Asym.1 (25%) 5.3 32.1 8.0 53 30
46 5.0 Asym.1 (25%) 5.3 32.1 6.7 89 41
47 2.5 Asym.1 (25%) 3.0 18.3 4.4 75 44
48 5.0 Asym.1 (25%) 3.0 18.3 3.2 128 121
49 2.5 Asym.2 (100%) 1.3 32.1 5.6 45 36
50 5.0 Asym.2 (100%) 1.3 32.1 4.0 106 95
51 2.5 Asym.2 (100%) 0.7 18.3 3.1 72 71
52 5.0 Asym.2 (100%) 0.7 18.3 2.6 130 127
53 2.5 Asym.2 (50%) 2.6 32.1 5.7 49 31
54 5.0 Asym.2 (50%) 2.6 32.1 4.7 93 83
55 2.5 Asym.2 (50%) 1.5 18.3 3.6 76 68
56 5.0 Asym.2 (50%) 1.5 18.3 2.6 121 121
57 2.5 Asym.2 (25%) 5.3 32.1 12.8 154 28
58 5.0 Asym.2 (25%) 5.3 32.1 4.9 90 47
59 2.5 Asym.2 (25%) 3.0 18.3 4.4 76 67
60 5.0 Asym.2 (25%) 3.0 18.3 2.9 128 123


Averaged value during the last 100 s

Averaged value after 200 s until the end of the simulation

3 where no relevant differences can be found. However, in the case of 5 MW, the
highest value of inlet velocity clearly produces more smoke only for sym. 3, that
is, when the streams impinge directly on the area of the flame, whereas no differ-
ence is observed for the remaining inlet velocities of 0.4 and 0.7 m/s.
Thirdly, figures 8g–i show the influence of the make-up air velocity with asym.
1 configuration and velocities from 1.3 to 5.3 m/s (simulations 38, 42, and 46,
Table 3). It can be observed how the flame is completely inclined and rotates
around the pool, Fig. 8g. The smoke layer height is observed to decrease as the
opening area increases. However, a high degree of mixing is generated due to the
large disturbances induced as the make-up air velocity increases, Fig. 8i. This phe-
nomenon is much more noticeable with the highest make-up air velocity, ascend-
ing the smoke layer height 2.6 m with respect to the lowest make-up air velocity.
Regarding the same configuration with similar make-up air velocities but with
HRR of 2.5 MW (simulations 37, 41 and 45), the smoke layer interface is higher
as the inlet velocity is increased, with a maximum difference of 2.7 m.
The simulations with an exhaust flow rate of 18.3 m3 =s with the same opening
configuration, asym. 1, and both values of HRR (simulations 39, 40, 43, 44, 47,
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1081

Table 4
Results with the Pan Located at the Diagonal

No. HRR (MW) Open vents vinlet (m/s) V_ (m3 =s) SL (m) TWA ( C)

61 2.5 Sym.1 (100%) 0.7 32.1 6.3 28


62 5.0 Sym.1 (100%) 0.7 32.1 5.3 61
63 2.5 Sym.1 (100%) 0.4 18.3 4.1 65
64 5.0 Sym.1 (100%) 0.4 18.3 2.7 116
65 2.5 Sym.1 (50%) 1.3 32.1 5.5 32
66 5.0 Sym.1 (50%) 1.3 32.1 4.1 76
67 2.5 Sym.1 (50%) 0.7 18.3 3.3 63
68 5.0 Sym.1 (50%) 0.7 18.3 2.6 116
69 2.5 Sym.1 (25%) 2.6 32.1 5.7 33
70 5.0 Sym.1 (25%) 2.6 32.1 4.2 78
71 2.5 Sym.1 (25%) 1.5 18.3 3.6 62
72 5.0 Sym.1 (25%) 1.5 18.3 2.6 119
73 2.5 Asym.1 (100%) 1.3 32.1 5.4 24
74 5.0 Asym.1 (100%) 1.3 32.1 4.3 74
75 2.5 Asym.1 (100%) 0.7 18.3 3.9 58
76 5.0 Asym.1 (100%) 0.7 18.3 2.8 106
77 2.5 Asym.1 (50%) 2.5 32.1 9.1 27
78 5.0 Asym.1 (50%) 2.5 32.1 6.1 45
79 2.5 Asym.1 (50%) 1.5 18.3 4.3 57
80 5.0 Asym.1 (50%) 1.5 18.3 3.0 107
81 2.5 Asym.1 (25%) 5.3 32.1 4.2 33
82 5.0 Asym.1 (25%) 5.3 32.1 7.5 42
83 2.5 Asym.1 (25%) 3.0 18.3 4.4 56
84 5.0 Asym.1 (25%) 3.0 18.3 3.4 106


Averaged value during the last 100 s

and 48) present the same behaviour. According to these results, it becomes crucial
to avoid any undesirable circulation of air on the vicinity of the flame, regardless
the inlet velocity, as the formation of fire whirls has been observed also for inlet
velocities lower than 1 m/s.
From the previous results, it is important to note that the influence of the inlet
velocity is highly related both to the value of the HRR of the fire and the inlet
configuration. In the symmetric configuration, the smoke production is enhanced
when the inlet velocity increases for values above the limit of 1 m/s. Furthermore,
for values near but below this limit, the smoke behaviour depends on the whole
fire scenario without a clear tendency based only on the inlet velocity being, there-
fore, necessary a deeper engineering analysis also in this case. In asymmetric open-
ing configurations, increasing the make-up air velocity enhances the mixing in the
smoke layer height generating a wide blurred smoke layer interface. In addition,
dangerous phenomena like fire whirls might appear, causing violent flame rota-
tions and unexpected high values of the flame height.
1082 Fire Technology 2018

(a) (b)
20 20
100%, 0.7 m/s
50%, 1.3 m/s
15 15
25%, 2.6 m/s
Height (m)

Height (m)
10 10

5 5

0 0
0 200 400 600 0 200 400 600
t (s) t (s)

(c) (d)
20 20
100%, 1.3 m/s
50%, 2.6 m/s
15 15
25%, 5.3 m/s
Height (m)

Height (m)

10 10

5 5

0 0
0 200 400 600 0 200 400 600
t (s) t (s)

Figure 9. Smoke layer drop for symmetric 1 and asymmetric 1


configurations, with an exhaust flow rate of 32.1 m3 =s, a HRR of
5 MW, and different openings areas. (a) and (c) Pan at the centre; (b)
and (d) Pan at the diagonal.

4.2. Pan Location


The influence of the pan location on the smoke behaviour is assessed by compar-
ing the smoke production for the pool fire located at the centre and at the diago-
nal of the atrium. In both fire scenarios, the fire plume can be considered
axisymmetric, because the distance from the walls to the fire is large enough (more
than 5 m), and thus, the smoke production is expected to be the same.
Firstly, the case of large HRR is analyzed. Thus, Fig. 9 shows the smoke layer
drop for a HRR of 5 MW, with the pool fire located at the centre (Fig. 9a, c) and
at the diagonal (Fig. 9b, d) for the make-up air configurations sym1 and asym1.
The simulations herein shown considered the vents fully opened (100%) and par-
tially opened (50 and 25%) with an exhaust flow rate of 32.1 m3 =s.
On the one hand, the results of sym. 1 configuration with the pan at the centre
(Fig. 9a, simulations 2, 6, and 10) show a higher smoke layer height than with the
pan located at one diagonal (Fig. 9b, simulations 62, 66, and 70). When the inlet
velocity is too large, i.e. the vents are 25% opened, the flame is significantly and
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1083

equally disturbed at both locations, thus the smoke production in both cases is
high. However, when the inlet velocity is lower, even to the limit of 1 m/s, the
inlet flow provokes high flame disturbances, and consequently a higher smoke
production, only on the pan located closer to the opening
As for the same comparison for a HRR of 2.5 MW (simulations 1, 5 and 9,
and 61, 65 and 69), the smoke behaviour predicted according to the pan location
is similar to that of the previous cases. Moreover, regarding the cases with the
lower exhaust flow rate (18.3 m3 =s) under this configuration (sym. 1), the influence
of the pan location is again remarkable, obtaining always more smoke produced
when placed at the diagonal. In these cases, it might be preferable to inject the
make-up air far away from the fire since even at inlet velocities larger than 1 m/s,
as it would not directly affect the surroundings of the fire, as shown next.
It is relevant to highlight that when the pan is located at the diagonal, no
appreciable differences on the smoke layer height are observable, between the val-
ues of the inlet velocities tested when they are above the limit 1 m/s.
On the other hand, with the asym. 1 configuration and the pan at the centre
(Fig. 9c, simulations 38, 42, and 46) the smoke layer height is lower and more
defined than with the pan at the diagonal (Fig. 9d, simulations 74, 78, and 82). In
the former configuration, some mixing can be only observed with the highest
velocity whereas in the latter configuration, the mixing can be appreciable when
the vents are always partially opened (50 and 25%). The smoke layer height
increases with the pan located at the diagonal, but due to the higher disturbances
generated at the flame, the smoke spreads more. This can be corroborated from
the vertical temperature distributions in Fig. 10, where high temperatures are pre-
dicted close to the ground as well as a not well defined smoke layer (Fig. 10d).
This might affect the prediction of the smoke layer interface [31].
In addition for a lower HRR (2.5 MW), the smoke layer interface becomes
much more unstable, specially when the pan is located at the diagonal position.
Also, the simulations with an exhaust flow rate of 18.3 m3 =s show that the smoke
fills the atria because of the high temperatures at the base, providing with inaccu-
rate values for the smoke layer height, as commented above. In these cases, the
values of the inlet velocity do not affect the smoke layer behavior.

4.3. Inlet Distribution


The influence of the inlet vents distribution is studied by means of two configura-
tions (symmetric and asymmetric) that create two completely different make-up
air patterns around the flame. The symmetric configurations create flow patterns
which tend to minimize their impact on the flame, whereas the asymmetric config-
urations provoke circular flow patterns that can easily generate flame swirls. A
total of five configurations are assessed: sym. 1, sym. 2, sym. 3, asym. 1 and asym.
2. The comparisons are made between the configurations that have the same
make-up air velocity in order to guarantee that the only influence on the scenario
is the location of the openings.
Figure 11 shows the temperature at h = 7.25 m in the fire plume (Fig. 11a, d),
at h = 5 m high close to the wall A (Fig. 11b, e), and the smoke layer height
1084 Fire Technology 2018

(a) (b)
20 20
0s
100 s
15 15
200 s
Height (m)

Height (m)
300 s
400 s
10 10
500 s
600 s
5 5

0 0
0 50 100 0 50 100
Temperature (°C) Temperature (°C)

(c) (d)
20 20
0s
100 s
15 15
200 s
Height (m)

Height (m)

300 s
400 s
10 10
500 s
600 s
5 5

0 0
0 50 100 0 50 100
Temperature (°C) Temperature (°C)

Figure 10. Vertical temperature distributions for symmetric 1 and


asymmetric 1 configurations, with an exhaust flow rate of 32.1 m3 =s,
a HRR of 5 MW, and different openings areas. (a) and (c) Pan at the
centre; (b) and (d) Pan at the diagonal.

(Fig. 11c, f) of sym. 1 and asym. 1 configurations both with a make-up air veloc-
ity of 1.3 m/s with 2.5 MW, and 5 MW fires (simulations 5, 6, 37, and 38).
Simulations with 2.5 MW, Fig. 11a, show that the predicted temperature in the
fire plume at h = 7.25 m is mostly higher in case of the symmetric configuration.
Nevertheless, there are some temperature peaks in the asym. 1 ventilation configu-
ration. These peaks may correspond to flame swirls as a consequence of the make-
up air induced flow pattern. It has to be also considered that the same HRR
curve has been used in all cases, when it is well known that the induction of a fire
involves the sudden increase in HRR which would contribute to further worsen
the inner conditions. Moreover, Fig. 11b, which presents the temperature at 5 m
high close to the wall A, shows the temperature of the asym. 1 configuration is
generally higher, which indicates a larger smoke production. It can be also appre-
ciated that, when there is a temperature peak in Fig. 11a, there is a drop on the
temperature close to the wall A, as it happens in the time interval of 400  t 
500. This phenomenon can be also observed in the smoke layer drop, in which an
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1085

(a) (b) (c)


300 80 20
Sym.
250
Temperature (°C)

Temperature (° C)
Asym.
60 15

Height (m)
200

150 40 10

100
20 5
50

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(d) (e) (f)


400 140 20

120 Sym.
Temperature (° C)

Temperature (° C)

Asym.
300 15
100

Height (m)
80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 11. Symmetric 1 and asymmetric 1 configurations with a


make-up air velocity of 1.3 m/s, and HRRs of 2.5 and 5 MW,
respectively. (a) and (d) Fire plume temperature at h = 7.25 m; (b)
and (e) wall A temperature at h = 5 m; (c) and (f) smoke layer drop.

increment on the smoke layer interface occurs at the same time interval. Addition-
ally, Fig. 11c shows that the smoke layer drop presents a more unstable behaviour
in case of asymmetric ventilation. A similar behaviour of the smoke can be appre-
ciated in the temperature profiles as well as in the smoke layer drop for the 5 MW
fire simulations (Fig. 11d–e). For example, the asym. 1 simulations present tem-
perature peaks in the fire plume after 200 s (Fig. 11d), which at the same time
produces a decrement on the temperature at 5 m high close to the wall (Fig. 11e)
and an ascent on the smoke layer interface (Fig. 11f). In short, similar conclusions
can be achieved from the results from both HRRs, with the only difference that
the smoke production is higher for simulations with the 5 MW fire, which entails
higher temperature at 5 m high close to the wall A with respect to the simulations
of 2.5 MW fire. It has to be mentioned that for the asymmetric cases the evalua-
tion method of the smoke layer height is not very accurate because of the small
gradient of temperature close to the ground, as commented above (Fig. 10).
Considering again the sym. 1 and the asym. 1 configurations but with a make-
up air velocity of 0.7 m/s (simulations 7, 8, 39, and 40), the smoke behaviour is
similar to that of the previous results with a velocity of 1.3 m/s.
1086 Fire Technology 2018

And if the inlet velocity in both configurations is increased up to 2.6 m/s (simu-
lations and 19–20), the smoke layer interface obtained for both configurations is
similar, Table 2. Again, the asymmetric configurations show an important amount
of mixing in the smoke layer due to the high value of the make-up air velocity.
Additionally, the influence of the relative position of the opening vents with
respect to the centre is studied for the different symmetric configurations (sym. 1,
sym. 2 and sym. 3) and asymmetric configurations (asym. 1 and asym. 2), Fig. 7.
Figure 12a–d show the comparisons of the three symmetric configurations (simu-
lations 6, 18 and 30) and the two asymmetric configurations (simulations 38 and
50) with a make-up air velocity of 1.3 m/s, respectively. It has been assessed the
temperature at h = 7.25 m over the fire (Fig. 12a, d), at h = 5 m close to the
wall A (Fig. 12b, e) and the smoke layer drop (Fig. 12c, f) obtained for 5 MW
fires with an exhaust flow rate of 32.1 m3 =s, and with partially opened vents
(50%).
On the one hand, the influence of the three symmetric configurations on the fire
plume temperature profiles, Fig. 12a, shows that with sym. 2 configuration, the
flame disturbances are lower, reaching higher temperatures. As the openings are

(a) (b) (c)


400 140 20

120 Sym. 1
Sym. 2
Temperature (°C)

Temperature (°C)

300 15
100 Sym. 3
Height (m)

80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

(d) (e) (f)


400 140 20

120 Asym. 1
Temperature (°C)

Temperature (°C)

Asym. 2
300 15
100
Height (m)

80
200 10
60

40
100 5
20

0 0 0
0 200 400 600 0 200 400 600 0 200 400 600
t (s) t (s) t (s)

Figure 12. Symmetric and asymmetric configurations with a make-up


air velocity of 1.3 m/s and with a HRR of 5 MW. (a) and (d) Fire
plume temperature at h = 7.25 m; (b) and (e) wall A temperature at h
= 5 m; (c) and (f) smoke layer drop.
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1087

moved away from the fire, i.e. sim. 1, the flame disturbances become more impor-
tant. These flame perturbations provoke a larger smoke production, which can be
observed at 5 m high close to the wall A, Fig. 12b, where the temperature rises
only in the sym. 1 configuration. Regarding the behaviour of the smoke layer
drop in sym. 1 and sym.3 configurations, Fig. 12c, it can be observed that lower
heights are reached than with sym. 2 configuration. Therefore, it seems to exist an
optimum injection configuration which strongly depends on the fire location,
among others, such as the fire power, indicating that a flexible make-up multi-in-
jection system would be advisable to provide with a more secure design.
As for the symmetric configurations, with a HRR of 2.5 MW (simulations 5,
17, and 29), the predictions related to sym. 1 and sym. 3 configurations are again
similar. Nevertheless, with the sym. 2 configuration the smoke layer height is
below the values obtained with the other configurations. This behaviour confirms
once again the combined action of the different parameters of the fire scenario,
even for symmetric configurations.
On the other hand, the comparison of the asymmetric configurations is shown
in Fig. 12d–f. The fire plume temperature, Fig. 12d, shows that in both configura-
tions the flame is deflected and some possible swirls appear at certain instants dur-
ing the fire, which are identified by temperature peaks, such as in asym. 1
configuration after 200 s and in asym. 2 configuration before 400 s. These swirl
formations indicate that the flow pattern is circular in both cases. In the far field,
Fig. 12e, sudden strong changes in the inner conditions can be observed at 5 m
high close to the wall A, at 400 s and for the asym. 2 configuration. This tempera-
ture reduction is higher in asym. 2 configuration, which indicates that the smoke
production is more sensitive as the openings are closer to the centre, and then, the
circular flow pattern performs faster in this configuration. Finally, the smoke layer
drops are similar in both configurations as it can be observed in Fig. 12f, which
agrees with the fact that both present circular flow patterns. Again, the swirl for-
mation shows a slight increase in the smoke layer interface, which can be observed
around 400 s in asym. 2 configuration.
The same behaviours of the smoke and flame are observed when the previous
comparison is carried out for a HRR of 2.5 MW (simulations 13 and 33).

5. Conclusions
The present paper considers the effect of different make-up air venting conditions
on the smoke behaviour and the dynamics of atrium fires. Thus, five full-scale fire
experiments have been first conceived to explore the influence of the make-up air
and the layout of vents. In this regard, three tests present similar geometrical con-
figuration, but different make-up air inlet velocity (from 0.4 to 2.25 m/s). The
venting layout is varied from symmetric to asymmetric in two tests with inlet
velocities below 1 m/s. Additionally, the fire location is varied in two tests with
the same asymmetric venting layout, thus, considering different venting configura-
tions relative to the fire.
1088 Fire Technology 2018

Numerical simulations of these tests have been conducted, stressing the impor-
tance of the use of domain extensions when modelling the inlet conditions, which
have a deep impact on the dynamics of atrium fires. The inclusion of these exten-
sions at the openings is required to properly predict the inner fire induced condi-
tions when the inlet distribution creates circular flow patterns or the inlet velocity
is above the recommended limit of 1 m/s. These two particular conditions should
be considered during the smoke control and management system design due to the
increase of smoke hazard, as under these circumstances the smoke production is
enhanced because of their direct influence on the flame behaviour. Furthermore,
the numerical model has been validated to capture dangerous phenomena like
flame deflections or even fire whirls.
Subsequently, an extensive numerical study has been carried out, by means of
84 simulations, using different vent cross-sectional areas and configurations (3
symmetric and 2 asymmetric), two values of HRRs (2.5 and 5 MW), and two dif-
ferent pan locations. A range of make-up air velocity from 0.4 to 5.3 m/s has been
explored.
The make-up air velocity assessment shows that the influence of the inlet veloc-
ity is related to the value of the HRR of the fire and the inlet configuration,
specifically, when the make-up air velocity is close to, but below, the prescribed
limit of 1 m/s. This involves that for very low inlet values, when increasing the
velocity, the smoke production is enhanced and then a lower smoke layer height is
obtained. Furthermore, for very high values, the mixing in the smoke layer height
is so relevant that a not well defined layer is obtained. However, for values closer
to this prescribed limit, a not clear tendency is observable meaning that a deeper
analysis of the complete fire scenario is always necessary, even for values below
the limit of 1 m/s.
When the pan location influence is analyzed, the make-up air velocity limit
should be again carefully considered. It is noticeable that when the pan is placed
at the diagonal, and a HRR of 2.5 MW, i.e. a low HRR fire, no relevant differ-
ences are appreciable when the make-up air velocity is below or above 1 m/s.
Generally, with the fire located at the diagonal, the opening is closer to it, being
thus more affected by the make-up air in the case of symmetric venting configura-
tions. In the case of asymmetric configurations, the pan located at the centre pre-
sents circular flow patterns which produces a large amount of smoke. Besides,
when the pan is moved to the diagonal, a higher degree of turbulence is generated
by the flame, the smoke is more spread and a high degree of mixing is present.
Again, it has been observed how the temperature difference between the hot
smoke layer and the cold free smoke layer is not high enough to accurately pre-
dict the smoke layer height.
In this regard, the analysis carried out herein confirms that asymmetric configu-
rations are described as more hazardous than the symmetric ones because the
smoke production is larger. Besides, the results indicate that asymmetric configu-
rations may induce circular flow patterns around the flame which can generate
violent flame rotations and even possible fire whirls, with a considerable increment
on the flame heights, HRR and on the potential threat to the occupants and
structure.
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1089

In summary, the make-up air should be considered carefully in the smoke con-
trol and management systems design, being recommendable to avoid asymmetric
configurations because they can provoke dangerous fire scenarios, in which the
flame reaches unpredictable heights and the smoke layer descends more than the
expected height. In addition, it is recommended to assess the complete fire sce-
nario when the make-up air velocities approach the limit of 1 m/s because, in
terms of smoke production, adverse, hazardous phenomena have been observed to
be generated for values below such prescribed threshold. Finally, the relevance of
this numerical study also lies in the analysis of the cases where a transition from a
pool fire to a fire whirl is possible. Both generic fire types completely differ in
behaviour and this type of numerical research can be used to obtain precise
results in the design of fire safety systems.

Acknowledgements
This research was supported by Fundación Mapfre, Institute for Research in
Technology (IIT) of Universidad Pontificia Comillas, and the Spanish MINECO
(Subdirección General de Gestión de Ayudas a la Investigación) under Project
DPI2014-59292-C3-3. Moreover, these authors want to acknowledge the compa-
nies Sodeca, Xtralis and Gabriele Vigne of JVVA for the support during the full-
scale fire experiments. Additionally, it is important to thank National Institute of
Standard and Technology for making FDS available.

References
1. Alarie Y (2002) Toxicity of fire smoke. Crit Rev Toxicol 32(4):259. https://doi.org/
10.1080/20024091064246
2. Chow W (1997) On the use of time constants for specifying the smoke filling process in
atrium halls. Fire Saf J 28(2):165. https://doi.org/10.1016/S0379-7112(96)00071-9
3. Mowrer FW (2002) Enclosure smoke filling and management with mechanical ventila-
tion. Fire Technol 38(1):33. https://doi.org/10.1023/A:1013476730932
4. Shi C, Lu W, Chow W, Huo R (2007) An investigation on spill plume development
and natural filling in large full-scale atrium under retail shop fire. Int J Heat Mass
Transf 50(3):513. https://doi.org/10.1016/j.ijheatmasstransfer.2006.07.020
5. Capote JA, Alvear D, Abreu OV, Lázaro M, Espina P (2009) Scale tests of smoke fill-
ing in large atria. Fire Technol 45(2):201. https://doi.org/10.1007/s10694-008-0074-4
6. Qin T, Guo Y, Chan C, Lin W (2009) Numerical simulation of the spread of smoke in
an atrium under fire scenario. Build Environ 44(1):56. https://doi.org/10.1016/j.buil-
denv.2008.01.014
7. Hao-wei Y, Wen-li D, Dong L, Rogner A, Jing-wei L (2011) Procedia Engineering
11(Supplement C), 608. In: The 5th Conference on Performance-based Fire and Fire
Protection Engineering. https://doi.org/10.1016/j.proeng.2011.04.703
8. Klote JH (1994) Method of predicting smoke movement in atria with application to
smoke management. NIST, Gaithersburg
1090 Fire Technology 2018

9. NFPA 92, (2015) Standard for smoke control systems (National Fire Protection Associ-
ation, Quincy, USA, 2015)
10. Council I.B (2009) International Building Code (International Building Council, 2009)
11. Hadjisophocleous G, Zhou JJ (2008) Evaluation of atrium smoke exhaust make-up air
velocity. ASHRAE Trans 114(1):147
12. Zhou J, Hadjisophocleous G (2008) Parameters affecting fire plumes. ASHRAE Trans
114(1):140
13. Hadjisophocleous GV, Lougheed GD (1999) Experimental and numerical study of
smoke conditions in an atrium with mechanical exhaust. Int J Eng Perform Based Fire
Codes 1(3):183
14. Yi L, Chow W, Li Y, Huo R (2005) A simple two-layer zone model on mechanical
exhaust in an atrium. Build Environ 40(7):869. https://doi.org/10.1016/j.buil-
denv.2004.08.018
15. Kerber S, Milke JA (2007) Using FDS to simulate smoke layer interface height in a
simple atrium. Fire Technol 43(1):45. https://doi.org/10.1007/s10694-007-0007-7
16. Pongratz C, Milke JA, Trouve A (2016) A CFD study to identify methods to increase
maximum velocity of makeup air for atrium smoke control. ASHRAE Trans 122(2):10
17. Floyd J, Forney G, Hostikka S, Korhonen T, McDermott R, McGrattan K (2013) Fire
dynamics simulator (version 6) users guide, 1st edn. National Institute of Standard and
Technology, Gaithersburg
18. Zhang X, Yang M, Wang J, He Y (2010) Effects of computational domain on numeri-
cal simulation of building fires. J Fire Prot Eng 20(4):225. https://doi.org/10.1177/
1042391510367349
19. Gutirrez-Montes C, Sanmiguel-Rojas E, Burgos MA, Viedma A (2012) On the fluid
dynamics of the make-up inlet air and the prediction of anomalous fire dynamics in a
large-scale facility. Fire Saf J 51:27. https://doi.org/10.1016/j.firesaf.2012.02.007
20. Ayala P, Cantizano A, Rein G, Vigne G, Gutiérrez-Montes C (2016) Fire experiments
and simulations in a full-scale atrium under transient and asymmetric venting condi-
tions. Fire Technol 52(1):51. https://doi.org/10.1007/s10694-015-0487-9
21. Gutirrez-Montes C, Sanmiguel-Rojas E, Kaiser AS, Viedma A (2008) Numerical model
and validation experiments of atrium enclosure fire in a new fire test facility. Build
Environ 43(11):1912. https://doi.org/10.1016/j.buildenv.2007.11.010
22. Gutirrez-Montes C, Sanmiguel-Rojas E, Viedma A, Rein G (2009) Experimental data
and numerical modelling of 1.3 and 2.3 MW fires in a 20 m cubic atrium. Build Envi-
ron 44(9):1827. https://doi.org/10.1016/j.buildenv.12.010
23. Gutirrez-Montes C, Sanmiguel-Rojas E, Viedma A (2010) Influence of different make-
up air configurations on the fire-induced conditions in an atrium. Build Environ
45(11):2458. https://doi.org/10.1016/j.buildenv.2010.05.006
24. Ayala P, Cantizano A, Sánchez-Úbeda EF, Gutiérrez-Montes C (2017) The use of frac-
tional factorial design for atrium fires prediction. Fire Technol 53(2):893. https://
doi.org/10.1007/s10694-016-0609-z
25. Beji T, Verstockt S, Van de Walle R, Merci B (2012) Prediction of smoke filling in large
volumes by means of data assimilation-based numerical simulations. J Fire Sci
30(4):300. https://doi.org/10.1177/0734904112437845
26. Hostikka S, Kokkala M, Vaari J (2001) NFSC2 Test Series, VTT Research Notes 2104;
Technical Research Centre of Finland: Finland
27. Hill K, Dreisbach J, Joglar F, Najafi B, McGrattan K, Peacock R, Hamins A (2007)
Verification and validation of selected fire models for nuclear power plant applications.
Tech. Rep. NUREG 1824, United States Nuclear Regulatory Commission, Washing-
ton, DC
Factors Affecting the Make-Up Air and Their Influence on the Dynamics 1091

28. SFPE (2002) Handbook of fire protection engineering, 3rd edn. National Fire Protec-
tion Association, Quincy
29. Tohidi A, Gollner MJ, Xiao H (2018) Fire Whirls. Ann Rev Fluid Mech. https://
doi.org/10.1146/annurev-fluid-122316-045209
30. Janssens M, Tran HC (1992) Data reduction of room tests for zone model validation. J
Fire Sci 10(6):528. https://doi.org/10.1177/073490419201000604
31. He Y, Fernando A, Luo M (1998) Determination of interface height from measured
parameter profile in enclosure fire experiment. Fire Saf J 31(1):19. https://doi.org/
10.1016/S0379-7112(97)00064-7

You might also like