i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5

Available at www.sciencedirect.com

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

CFD simulations on small hydrogen releases inside

a ventilated facility and assessment of ventilation efficiency

E. Papanikolaou a,c,*, A.G. Venetsanos a, G.M. Cerchiara b, M. Carcassi b, N. Markatos c

a
Environmental Research Laboratory, National Centre for Scientific Research Demokritos 15310, Aghia Paraskevi Attikis, Greece
b
Università degli Studi di Pisa, Dipartimento di Ingegneria Meccanica Nucleare e della Produzione, via Diotisalvi 2, 56126 Pisa, Italy
c
National Technical University of Athens, School of Chemical Engineering, Department of Process Analysis and Plant Design, Heroon
Polytechniou 9, 15780 Zografou, Greece

article info abstract

Article history: The use of stationary H2 and fuel cell systems is expected to increase rapidly in the future.
Received 19 January 2010 In order to facilitate the safe introduction of this new technology, the HyPer project, funded
Received in revised form by the EC, developed a public harmonized Installation Permitting Guidance (IPG) document
18 April 2010 for the installation of small stationary H2 and fuel cell systems for use in various envi-
Accepted 20 April 2010 ronments. The present contribution focuses on the safety assessment of a facility, inside
Available online 20 May 2010 which a small H2 fuel cell system (4.8 kWe) is installed and operated. Dispersion experi-
ments were designed and performed by partner UNIPI. The scenarios considered cover
Keywords: releases occurring inside the fuel cell at the valve of the inlet gas pipeline just before the
Hydrogen pressure regulator, which controls the H2 flow to the fuel cell system. H2 was expected to
Fuel cell leak out of the fuel cell into the facility and then outdoors through the ventilation system.
Ventilation assessment The initial leakage diameter was chosen based on the Italian technical guidelines for the
ATEX enforcement of the ATEX European directive. Several natural ventilation configurations
Experiments were examined. The performed tests were simulated by NCSRD using the ADREA-HF code.
CFD The numerical analysis took into account the full interior of the fuel cell, in order to
Enclosure investigate for any potential accumulation effects. Comparisons between predicted and
experimental H2 concentrations at 4 sensor locations inside the facility are reported.
Finally, an overall assessment of the ventilation efficiency was made based on the simu-
lations and experiments.
ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.

1. Introduction and presents assessments of current knowledge on installa-

The HyPer project [1] was aimed at providing a comprehensive
agreed installation permitting process for developers, design
engineers, manufacturers, installers and authorities having
jurisdiction across the European Union. The work within this
scope is reported in the Installation Permitting Guidance
document (IPG) [2] which provides recommendations for the
safe installation of small stationary H2 and fuel cell systems
tion requirements and case studies of representative instal-
lations. A combing experimental and modelling work was
carried out to address topics relevant to small stationary H2
and fuel cell installations covering scenarios of high-pressure
releases, small foreseeable releases, catastrophic releases and
the effects of walls and barriers. A summary of the modelling
and experimental programme in the EC FP6 project HYPER is
given in [3].

* Corresponding author. Present address: European Commission DG-JRC, Institute for Energy, Cleaner Energy Unit, 1755 ZG Petten,
The Netherlands.
E-mail addresses: efthymia.papanikolaou@ec.europa.eu, ep04@ipta.demokritos.gr (E. Papanikolaou).
0360-3199/$ e see front matter ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2010.04.119
2598 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5
Nomenclature
c volume fraction (v/v)
f stratification factor (VH2/Vtotal)
q H2 mass fraction
Specifically for scenarios on small foreseeable releases, the
aim was to determine the ventilation requirements in enclo-
sures containing fuel cells, such that in the event of a fore-
seeable leak, the concentration of H2 in air will not exceed 50%
of the Lower Flammability Limit (LFL) which allows the
enclosure not to be classified as zone 2 based on the Norm EN-
60079-10 [4]. A modelling and experimental work was carried
out within this scope. Experiments with natural and forced
ventilation were carried out by partner UNIPI. This paper
presents the modelling work on these experiments performed
by NCSRD and an overall assessment of the ventilation effi-
ciency of the enclosure.
2. Description of the UNIPI experiments
The dimensions of the CVE facility which was used to test
the scenarios on small foreseeable releases were 2753 mm,
3233 mm and 2814 mm providing an internal volume of
25 m3 (Fig. 1). The dimensions of the fuel cell were 800 mm,
688 mm and 1024 mm and its location inside the CVE is
shown in Fig. 2.
The leak of H2 was assumed at the valve of the inlet gas
pipeline just before the pressure reducer. The internal pres-
sure of the fuel cell was considered between 2 and 5 bars
whereas after the pressure reducer the pressure value was
350 mbar. In order to take into account the worst probable H2
loss, the highest pressure value was chosen, i.e. 5 bars. The
calculations of the H2 flow were performed with EFFECT-SGIS
Fig. 1 e CVE Experimental facility at UNIPI.
Qaw,Qin volumetric air flow in (m3/s)
Qair volumetric air flow (m3/s)
QH2 volumetric H2 release rate (m3/s)
r density (kg/m3)
V volume (m3)
7.3. The diameter of the leakage was chosen based on the
Italian technical guidelines for the enforcement of ATEX
European directive [5,6]. The Italian guidelines, taken into
consideration for the scenarios investigated in this work, refer
to small accidental leakages from pipelines with release
diameters up to 150 mm. Specifically, the CEI 31-35 H2 loss
from a valve is analogous to a flux from a hole with area
0.25 mm2. Calculations of the H2 flow were also done for
release areas of 0.5 mm2 and 1 mm2 to study more dangerous
cases then the one according to the CEI 31-35 guidelines [6].
The purpose of the experiments was to determine the
necessary ventilation requirements, either natural or forced,
for the enclosure not to classified as zone 2 in case of a small
accidental leakage of the fuel cell. In this case the limit of H2 in
air should be less than 50% of LFL, i.e. less than 2% H2 by
volume. Five sensors to measure H2 concentration were
available in order to define the volume around the fuel cell
where this limit is not overtaken. Fig. 2 shows the location of
the sensors inside the facility.
Concerning the natural ventilation experiments, the
ventilation was provided by openings on the two opposite
walls of the facility. The area of Vent 1 could be fixed from
a minimum value of 0.35 m2 (Vent 1) to a maximum of 0.70
(0.35 þ 0.35) m2 (Vent 1 þ Vent 1_2), the area of Vent 3 was
0.35 m2 and the areas of Vent 2 and Vent 4 were 0.14 m2 each
(Fig. 3). The several ventilation configurations are given in
Fig. 5. Concerning the forced ventilation experiments, either
a small air flow (0.14 m3/s) or a large air flow (0.28 m3/s) from
the fan was investigated. The forced ventilation experiments
were performed for the cases that natural ventilation failed to
provide adequate dilution of the released H2 with air for the
limit of 2% H2 to be respected.
Fig. 2 e Simplified drawing of CVE facility and position of
the fuel cell and sensors.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5 2599

Fig. 3 e Calculation domain showing vent locations of the

Fig. 4 e Interior of fuel cell.
CVE facility (created using DELTA- code).

UNIPI performed 31 experiments with natural ventilation,

1 experiment with the facility fully closed and 10 experiments
with forced ventilation. Several natural ventilation configu-
rations were examined whereas four different H2 release
locations were chosen. Two of them were inside the fuel cell,
one horizontal and one vertical. The horizontal release was
close to the back fan of the fuel cell directed to the outside
environment and the vertical was close to the upper part of
the fuel cell. The other two release locations were outside the
fuel cell with directions either horizontal or vertical. The H2
flow rate was approximately 40 l/min. As mentioned before,
two higher release flow rates were examined (90 l/min and
180 l/min) to study more dangerous cases than the one spec-
ified based on the CEI 31-35. Finally, wind conditions were
measured for some tests. The wind was measured far from
the facility at 5 m height in some cases. The experimentalists
reported that for most cases the wind was rather weak. Details
of the experimental work can be found in [7,8].
3. CFD simulations of the UNIPI experiments
NCSRD simulated 10 of the UNIPI natural ventilation experi-
ments. These covered scenarios with different ventilation
configurations, two release directions (horizontal and vertical)
and two release flow rates (40 t/min and 90 t/min). The hori-
zontal release was close to the back fan of the fuel cell directed
outside (y_out release) and the vertical upwards was close to
the upper part of the fuel cell (z_up release). The effect of the
wind on the selected cases was considered negligible since its
reported velocity was 1 m/s. Details of the simulated cases are
given in Table 1. The interior of the fuel cell was included in
the simulations to study possible accumulation of H2 (Fig. 4).
Details on the geometry of the fuel cell and the CVE facility can
be found in Refs. [9,10]
3.1. Modelling strategy
The dispersion calculations were performed using the
ADREA-HF CFD Code [11] Validation studies of the ADREA-HF
code for gaseous H2 release and dispersion and an overview
of the ADREA-HF code for H2 applications can be found in
Refs. [12e15]
Turbulence was modelled using the two equation standard
k-epsilon model of Launder and Spalding [16,17] modified for
buoyancy effects [18,19]. The mixing of H2 with air was
calculated by solving the three dimensional transient, fully

Table 1 e Specifications of simulated experiments.

Test Temp (K) H2 flow Vent area (m2) Release time Leak Hole diameter Pipe diameter
(l/min) (s) direction (mm) (mm)

3 288.65 40 0.35 (up) 0.14 (down) 1200 y_out 1.00 1.00

6 283.15 40 0.35 (up) 0.35 (up) 1200 y_out 1.00 1.00
14 290.15 40 0.35 (up) 0.14 (down) 1200 z_up 1.00 6.00
11 290.85 90 0.35 (up) 0.14 (down) 420 y_out 6.00 6.00
12 292.35 90 0.35 (up) 0.35 (up) 0.14 (down) 373 y_out 6.00 6.00
18 298.15 90 0.35 (up) 0.35 (up) 0.35 (down) 900 z_up 1.00 6.00
2600 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5

Fig. 5 e Natural ventilation configuration of the experimental tests.

compressible conservation equations for mixture mass
(continuity equation), mixture momentum (for the three
velocity components) and the H2 vapour mass fraction
transport equation. The numerical options used were the first
order implicit scheme for time integration and the first order
upwind scheme for the discretization of the convective terms
[20] An automatic time step increase/decrease mechanism
was applied with 103 s initial time step with maximum CFL
number equal to 10.
The computational domain extended the experimental
facility by 3 or 4 computational cells in the y direction when the
wall in the x-z plane had ventilation openings. The minimum
cell size was 0.012 m for the majority of the simulated tests. An
aspect ratio of 1.12 was used to increase the cell size far from
the source. Volume porosity and area permeability approach
was used to model the complex geometrical layout with
Cartesian grid. The total number of active cells ranged from
130,000 to 240,000 depending on each test. The geometrical pre-
processing was performed with the DELTA-B Code [21]
The initial conditions for all tests were atmospheric pres-
sure and temperature depending on each test as given by
UNIPI. The conditions at the source depended on each test
whereas the velocity was assigned a value to give the indi-
cated by the experiments mass flow-rate.
3.2. Results and discussion
Due to page limitations, this work presents 6 of the 10 simu-
lated scenarios which the authors consider to be the most
indicative. Fig. 6e11 show a comparison between the H2
volumetric concentration of sensors 2, 3, 4, and 5 of UNIPI
experiments and NCSRD simulations for Tests 3, 6, 11, 12, 14

Fig. 6 e Comparison of experimental results with simulated for Test 3.

 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5 2601

Fig. 7 e Comparison of experimental results with simulated for Test 6.

and 18. It can be seen that generally there is a good agreement
between the experiments and the simulations. Sensor 1 (the
one located close to the release point) is not shown in the
figures since this type of sensors couldn’t measure values
higher than 20% of H2 concentration and therefore in most
cases this specific sensor was saturated.
For the tests with 40 lt/min H2 release flow rate (tests 3, 6 and
14) neither the experiments nor the simulations showed a H2
concentration reaching 2% which allows the enclosure not to be
classified as zone 2. For the tests with 90 lt/min H2 release flow
rate (tests 11, 12 and 18) the H2 concentration exceeded the limit
of 2% only in some sensors. Specifically, for test 18 only the
readings of sensor 3 showed a H2 concentration exceeding 2%.
For test 11 both experimental results and simulation showed a H2
concentration exceeding the limit of 2% in sensor 3 but again the
values were less then 2.5%. The simulation also showed sensor 5
to slightly exceed 2%. For test 12, both the experimental and
simulated value of sensor 3 slightly exceeded 2%.

Fig. 8 e Comparison of experimental results with Fig. 9 e Comparison of experimental results with
simulated for Test 14. simulated for Test 11.
2602 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5

Fig. 10 e Comparison of experimental results with

Fig. 11 e Comparison of experimental results with
simulated for Test 12.
simulated for Test 18.

Comparison of the experimental results with the simu-

lations for test 3, showed that the maximum difference
between them was in sensors 2 and 3 and did not exceed
0.6%. For test 6, the maximum difference was in sensor 4
and did not exceed 0.35%. For the rest of the sensors, the
agreement was very good. For test 11, the experiment
showed a slight excess of 2% of H2 concentration for sensor
3 while sensor 5 almost reached that value. In simulations
sensors 3 and 5 exceeded that value although following the
trend of the experimental results. Both simulation and
experiment show a decline in the H2 concentration in all
sensors after approximately 360 s. This is due to the profile
of the release which started to decrease at 360 s and reached
a zero value at 420 s. For test 12, the maximum difference
between the experiment and the simulation was found in
sensor 4 and did not exceed 0.7% whereas for test 14 it was
in sensor 2 and did not exceed 0.96%. The experimental
values of sensors 2 and 4 of test 14 were almost zero during
the release time. For test 18 the maximum difference
between the experimental and simulated values were in
sensors 2, 3 and 4 and did not exceed 0.5%. The differences
between the results can be attributed to the sensitivity of
the experimental measurements as it was reported that the
sensors had a precision of 0.2% H2 concentration. Further-
more, temperature could have also influenced the results as
the roof and one side of the facility were entirely covered
with panels of glass.
4. Overall assessment of ventilation
efficiency
4.1. Normalized ventilation correlation
The simplified model considers a H2 leak within an enclosure
with two vents, one near the top and one close to the bottom at
the opposite wall. Steady-state and isothermal conditions were
assumed. The following relationships were used: (a) H2 mass
conservation ðrH2 QH2 ¼ qout rout Qout Þ, (b) total mass conserva-
tion ðrair Qin þ rH2 QH2 ¼ rout Qout Þ and (c) mixture density:
1=rout ¼ qout =rH2 þ ð1  qout Þ=rair to derive the relationship
cout ¼ QH2 =ðQin þ QH2 Þ. In the case of stratification inside the
enclosure, the assumption of a homogeneous mixture only in

Fig. 12 e Schematic of simplified model, homogeneous mixture covering whole enclosure (left), homogeneous mixture
covering upper part of enclosure (right).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5 2603

Fig. 13 e Qaw/QH2 (based on CEI 31-35) versus H2 concentration experimental readings of sensor 5.

a part of the enclosure is made, were f ¼ Vupperpart =Vtotal thus 4.3. NCSRD calculations
cout ¼ VH2 =Vupperpart ¼ VH2 =f $Vtotal ¼ c=f and finally:
cout ¼ 1=fQH2 =Qin þ QH2 (Fig. 12). Non-dimensional ventilation rates versus concentration at the
top opening, based on the NCSRD calculations are shown in
Fig. 14. Qaw was now calculated from the simulations as the sum
4.2. UNIPI tests
of air flowing inside the facility from any ventilation opening.
The simulations revealed that the part of the facility from
Non-dimensional ventilation rates versus concentration at the
the ground to the height of the release, which is almost 1/3 of
top opening of a residential naturally ventilated garage were
the facility’s volume, contains minimum or zero H2.
addressed by Barley [22] Following this work, the ratio Qaw/QH2
Unlike Fig. 13, Fig. 14 shows that the simulated tests lie
versus H2 concentration (v/v) at the top vent (sensor 5), based
between the lines with correlation f ¼ 1/2 and f ¼ 1. Further-
on the UNIPI experimental results, is shown in Fig. 13. Here QH2
more, the values of Qaw/QH2 for the tests with the same release
is the measured H2 volumetric release rate and Qaw is the
flow rate are now different for different ventilation configu-
volumetric air flow in, estimated based on the CEI 31-35
rations. Additionally, the values of Qaw of tests with the same
formulas provided by UNIPI (Table 2). In the calculation of Qaw,
ventilation configuration but with different release flow rates
wind was assumed to have 1 m/s velocity for all cases.
were different. The CFD predictions of the values of Qaw take
Fig. 13 shows the experimental results as squares, where
into account not only the ventilation configuration but also
the number of each square is the number of the corresponding
the presence of H2 release. Fig. 15 shows a comparison
UNIPI experiment. The solid lines in Fig. 13 represent the
between the calculated Qaw from simulations and the Qaw
Qaw =QH2 ¼ f =C  1 correlation, which is described in Section
based on CEI 31-35. In most of the tests, Qaw based on CEI 31-35
4.1. The factor f is the stratification factor as H2 is not homo-
is under-predicted. The under-prediction is higher in cases
geneously distributed within the enclosure but occupies
were all openings are at the same wall (tests 5, 6, 15 and 18). In
a volume close to the ceiling. Factor f is the ratio between the
test 14, which has one opening close to the ceiling and the
H2-air layer volume divided by the volume of the enclosure.
other close to the ground at opposite wall, the calculated Qaw
Based on Fig. 13, the correlation f ¼ 1/3 seems to be a good
is very close to the CEI 31-35 one.
fit of the experimental data except for tests 16, 18, 4, 13, 14, 17
and 23. The graph shows that the values of Qaw are somehow
insensitive to the different ventilation configurations. Specif-
ically, the value of Qaw/QH2 is the same for tests 16, 9, 10, 3, 14,
Table 2 e Qaw values for UNIPI tests (based on CEI 31-35).
4, 17, 13 which have the same release conditions (same QH2)
but test 16 has different ventilation configuration (see Fig. 5). Unipi Test 1,2,5 3, 4, 9, 10, 11, 6, 7, 8 15 16
The same holds for tests 8, 7, 6 and 15 which have the same 13, 14, 17, 23
QH2 but test 15 has different ventilation configuration. Addi- Qaw (103) 8.75 37.78 17.5 17.5 37.78
tionally, close values of Qaw/QH2 are shown in the graph for the Unipi test 12 18 19 20, 22 21
group of tests 20, 21 and 12, 23, 19, 11. However tests 20, 21 and
Qaw (103) 39.9 26.25 39.9 78.64 75.57
tests 12, 11 have different ventilation configurations.
2604 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5

Fig. 14 e Qaw/QH2 (NCSRD simulations) versus H2 concentration simulation results at sensor 5.

measurements and the outdoor location of the facility. For the

cases with 40 l/min release flow rate, classified as small fore-
seeable release, neither the experiments nor the simulations
showed a H2 concentration reaching 2% which allows the
enclosure not to be classified as zone 2. For the cases with
higher release rate (90 lt/min) both experiments and simula-
tions showed a H2 concentration exceeding 2% in some
sensors. However the values were less the 2.5%.
Additionally, an assessment of the ventilation efficiency
based on the CEI 31-35 formulas and the NCSRD simulations
was made. It was found that the air flow in (Qaw), calculated
based on the CEI 31-35 formulas, is somehow insensitive to
the different ventilation configurations. CFD predictions of the
air flow in were sensitive to both ventilation configurations
and the presence and strength of the H2 release. In some of the
Fig. 15 e Comparison between Qaw calculated from
tests, Qaw based on CEI 31-35 is under-predicted as compared
simulations and CEI 31-35 formulas.
with the one calculated form the simulations. Furthermore,
the under-prediction is higher for the cases where all open-
ings were at the same wall. It is believed that the under-
predictions would lead into too conservative ventilation
5. Conclusions requirements in enclosures containing fuel cells.

This paper presents a part of the modelling work performed by

NCSRD on small foreseeable H2 release experiments within the
Acknowledgements
EC FP6 HyPer project. The main goal of this work was to inves-
tigate the ventilation requirements in order not to classify the
The authors would like to thank the European Commission for
enclosure as zone 2 in the event of a small foreseeable leak. The
co-funding of this work in the framework of the FP6 STREP
concentration of H2 in this case, should not exceed the limit of 2%
HYPER project [1].
H2. The experiments with natural and forced ventilation were
carried out by partner UNIPI whereas simulations of the natural
ventilation configurations were performed by partner NCSRD. references
In general good agreement was found between predicted
and experimentally measured concentration time histories
for all simulated cases. The differences between the results [1] http://www.hyperproject.eu/, http://epshypp.web.its.
can be attributed to the sensitivity of the experimental manchester.ac.uk/.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 6 ( 2 0 1 1 ) 2 5 9 7 e2 6 0 5
[2] HYPER installation permitting guidance for hydrogen and
fuel cells stationary applications. Available at: http://www.
hyperproject.eu/; January 2009.
[3] Brennan S, Bengaouer A, Carcassi M, Cerchiara G, Evans G,
Friedrich A, et al. Hydrogen and fuel cell stationary
applications: key findings of modelling and experimental
work in the HYPER project. In: Third international
conference on hydrogen safety, Ajaccio, Corsica, France,
16e18 September, 2009.
[4] CEI EN-60079-10. Electrical apparatus for explosive
atmospheres e part 10: classification of hazardous areas, 2nd
ed.; 2004.
[5] Directive 1999/92/EC of the European Parliament and of the
Council of 16 December 1999. Minimum requirements for
improving the safety and health protection of workers
potentially at risk from explosive atmospheres (15th
Individual Directive within the meaning of Article 16(1) of
Directive 89/391/EEC). Official Journal L 2000;023.
[6] CEI 31-35. Electrical apparatus for explosive atmosphere,
guide for classification of hazardous areas, 3rd ed.; 2007.
[7] Technical report by partner UNIPI. Available: http://www.
hyperproject.eu/ [accessed 25.06.08].
[8] Cerchiara GM. Test results, UNIPI. Available: http://www.
hyperproject.eu/; May 2008.
[9] Technical report by partner NCSRD. Available: http://www.
hyperproject.eu/ [accessed 12.10.07].
[10] NCSRDeUNIPI Program. Available: http://www.hyperproject.
eu/ [accessed 04.07.07].
[11] Bartzis JG. ADREA-HF: a three-dimensional finite volume
code for vapour cloud dispersion in complex terrain, EUR
Report 13580 EN.
[12] Venetsanos AG, Huld T, Adams P, Bartzis JG. Source,
dispersion and combustion modelling of an accidental
release of hydrogen in an urban environment. J Hazard Mater
2003;A105:1e25.
2605
[13] Gallego E, Migoya E, Martin-Valdepenas JM, Crespo A, Garcia
J, Venetsanos A, et al. An intercomparison exercise on the
capabilities of CFD models to predict distribution and mixing
of H2 in a closed vessel. In: First international conference on
hydrogen safety, Pisa, Italy; 2005.
[14] Papanikolaou EA, Venetsanos AG. CFD modelling for helium
releases in a private garage without forced ventilation. In:
First international conference on hydrogen safety, Pisa, Italy;
2005.
[15] Venetsanos AG, Papanikolaou E, Bartzis JG. The ADREA-HF
CFD code for consequence assessment of hydrogen
applications. In: Third international conference on hydrogen
safety, Ajaccio, Corsica, France, 16e18 September, 2009.
[16] Launder BE, Spalding DB. The numerical computation of
turbulent flows. Comput Methods Appl Mech Eng 1974;3(2):
269e89.
[17] Markatos NC. Mathematical modelling of turbulent flows.
Appl Math Modell 1986;10(3):190e220.
[18] Markatos NC, Malin MR, Cox G. Mathematical modelling of
buoyancy-induced smoke flow in enclosures. Int J Heat Mass
Transfer 1982;25(1):63e75;
Correspondence. Int J Heat Mass Transfer 1982;25(11):
1777e8.
[19] Markatos NC, Pericleous KA. Laminar and turbulent natural
convection in an enclosed cavity. Int J Heat Mass Transfer
1984;27(5):755e72.
[20] Patankar SV. Numerical heat transfer and fluid flow.
Hemisphere Corporation; 1980.
[21] Venetsanos AG, Catsaros N, Wurtz J, Bartzis JG. The DELTA-B
code: a computer code for the simulation of the geometry of
three-dimensional buildings. Code structure and users
manual, EUR Report 16326 EN.
[22] Barley CD, Gawlik K, Ohi J, Hewett R. Analysis of buoyancy-
driven ventilation of hydrogen from buildings. In: Second
ICHS, San Sebastian, Spain, 2007.