See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/277561864

Modelling heat and mass transfer of a broiler

house using computational fluid dynamics

ARTICLE in BIOSYSTEMS ENGINEERING · AUGUST 2015

Impact Factor: 1.62 · DOI: 10.1016/j.biosystemseng.2015.05.004

READS

137

6 AUTHORS, INCLUDING:

Pierre-Emmanuel Bournet Mélynda Hassouna

Agrocampus Ouest, Angers, France. Resear… French National Institute for Agricultural R…
57 PUBLICATIONS 269 CITATIONS 97 PUBLICATIONS 336 CITATIONS

SEE PROFILE SEE PROFILE

Paul Robin Murat Kacira

French National Institute for Agricultural R… The University of Arizona
102 PUBLICATIONS 618 CITATIONS 67 PUBLICATIONS 475 CITATIONS

SEE PROFILE SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Murat Kacira
letting you access and read them immediately. Retrieved on: 22 February 2016
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/issn/15375110

Research Paper

Modelling heat and mass transfer of a broiler house

using computational fluid dynamics

Fernando Rojano a, Pierre-Emmanuel Bournet a,*, Melynda Hassouna b,

Paul Robin b, Murat Kacira c, Christopher Y. Choi d
a
EPHor, Environmental Physics and Horticulture Research Unit, Agrocampus Ouest, Centre d'Angers, 2,
rue Le N^otre, 49045 Angers, France
b
SAS Lab, INRA, 65 Rue de Saint-Brieuc, 35042 Rennes, France
c
Agricultural and Biosystems Engineering, The University of Arizona, Tucson, AZ 85721, USA
d
Biological Systems Engineering, University of Wisconsin, Madison, WI 53706, USA

article info
Improvements to the living conditions in semi-enclosed spaces such as broiler houses can
Article history: be achieved by better control of the heat and mass transport that occur in climate and air
Received 27 January 2015 quality. This study shows that computer-aided modelling, and in particular computational
Received in revised form fluid dynamics (CFD), can provide to researchers the ability to integrate the primary forces
16 April 2015 that interact at the interior environment. A two dimensional CFD model was used to assess
Accepted 6 May 2015 the dynamics of a broiler house by investigating sensible and latent heat, as well as mass
Published online 31 May 2015 transport and radiative transfer energy, as these relate to the environment of the broiler
house. Validation data related to temperature, absolute humidity and CO2 were collected
Keywords: both inside and outside of a naturally ventilated broiler house. Inside data was logged at
Henhouse various locations to identify the degree of homogeneity throughout space. The CFD model
Natural ventilation replicated two contrasting cases: an early stage and a late stage of production. The pre-
Sensible and latent heat dicted values for temperature, absolute humidity and CO2 were in good agreement with
CO2 experimental data. For instance, the first case had a ventilation rate of 10 air changes h
1,
Livestock building and obtained a root-mean-square error (RMSE) of 1.0  C, 0.3 g [H2O] kg
1 [dry air] and
CFD modelling 134 ppm for temperature, absolute humidity and CO2, respectively. The second case had
ventilation rates of 25 air changes h
1, and obtained a RMSE of 0.9  C and 0.48 g [H2O] kg
1
[dry air] for temperature and absolute humidity, respectively.
© 2015 IAgrE. Published by Elsevier Ltd. All rights reserved.

Raising animals in buildings has also enabled producers to

1. Introduction
The shift from traditional to modern animal production has
helped to meet the increasing demand for meat whilst also
minimising production costs and incrementing productivity.
locate operations on smaller plots of land closer to large
markets. Also, livestock buildings solve a number of the
living-condition issues associated with high-density, acceler-
ated weight gain and reduced mortality rates. All these

* Corresponding author. Tel.: þ33 (0)2 41 22 55 04; fax: þ33 (0)2 41 22 55 53.
E-mail address: Pierre-Emmanuel.Bournet@agrocampus-ouest.fr (P.-E. Bournet).
http://dx.doi.org/10.1016/j.biosystemseng.2015.05.004
1537-5110/© 2015 IAgrE. Published by Elsevier Ltd. All rights reserved.
26 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

cleaning purposes, in contrast to the operation of livestock

Nomenclature buildings for pigs or cows where a significant amount of water
is required.
a absorption coefficient, m
1
AH total animal heat, W
1.1. Environment control in broiler houses
AHL latent heat from animal, W
AHS sensible heat from animal, W
Operating a broiler house efficiently depends on controlling
C gas concentration, ppm
the climate at an early stage of production because the young
d nth day of production cycle, day
broilers are susceptible to temperature changes. Typically, in
e water vapour pressure, Pa
mild climates, heaters burning propane are used to maintain
H characteristic height of the broiler house, m
the required interior temperatures during the early stages. As
h absolute humidity, g[H2O] kg
1 [dry air]
the broilers gain weight, they become more resistant to cli-
HHL latent heat from heater, W
matic changes and produce more heat, which helps to regu-
HHS sensible heat from heater, W
late the temperature inside the building and thereby reducing
HFfo heat flux from outdoor floor, W m
2
the heat provided by the heaters.
I radiance, W m
3 sr
1
Although temperature is the main climatic parameter, the
kz turbulence kinetic energy, m2 s
2
overall climate indoors is determined by the dynamics of the
LHL latent heat from litter, W
sensible and latent heat present in the building. Conse-
LHS sensible heat from litter, W
quently, the water vapour released from various sources such
m mass of hen, kg
as the broilers, the rearing system, the heaters and the degree
N ventilation rate decay, h
1
of natural ventilation should all be included when calculating
n refractive index
heat and mass balances. Also gases derived from the dy-
po atmospheric pressure, Pa
namics of heat and water vapour generation such as CO2
q atmospheric radiative flux, W m
2
should be incorporated since they impact on the environment
r Position vector, m
and animal welfare.
RMSE root-mean-square error
Furthermore, the climatic conditions that must be pro-
s unitary vector along the propagation of
vided in order to keep the broilers within the particular ranges
radiation
of temperature and humidity that enhance rapid weight gain
s0 scattering direction vector
and optimal welfare also provide favourable conditions for the
SR solar radiation, W m
2
bacteria found in the rearing system to increase the emissions
Sf source term (i.e. buoyancy effects)
of heat, humidity and various gases. As more humidity and
T air temperature,  C
gases are generated, the air quality deteriorates unless
Tfi temperature from indoor floor,  C
appropriate ventilation is maintained.
Ti indoor temperature,  C
To outdoor temperature,  C
1.2. Use of CFD modelling to assess livestock building
t time, s
environment
uj velocity in direction j (i.e. for two-dimensional
domain j ¼ 1, 2), m s
1
Broiler-house environmental conditions can be studied using
uo wind velocity, m s
1
modelling techniques that handle heat and mass transport
xj coordinate in direction j, m
phenomena. The availability of advanced modelling tools,
a wind direction, 
based on numerical solutions such as computational fluid
3 emissivity [from 0 to 1]
dynamics (CFD), analyses heat and mass sources (i.e. gases)
3z turbulence energy dissipation, m2 s
3
simultaneously, as these sources fluctuate due to the ex-
f denotes velocity, temperature, turbulent
change of indoor and outdoor air that is occurring by natural
kinetic energy, dissipation kinetic energy and
ventilation.
mass transport
The quality of the CFD model generated from data taken
Gf diffusion coefficient of the variable f
inside a livestock building depends on features such as the
s StefaneBoltzmann constant
accuracy of the numerical solution obtained, which in turn
(5.67  10
8 W m
2 C
1)
depends on choosing the proper turbulence model, turbulence
ss scattering coefficient, m
1
intensity, grid quality and discretisation scheme, and
F Diffusion phase function
convergence settings of the numerical solution. These fea-
Ut solid angle, sr
tures can be achieved using the method deployed by Ramponi
and Blocken (2012) that involves scaling down the livestock
building and relies on particle image velocimetry to validate
changes make the understanding of the climate and air the velocity field.
quality conditions within the buildings an imperative. To Various CFD studies validated by field experiments have
analyse the living-conditions, we chose to investigate broiler analysed methods of modifying the interior of livestock
production because the system has a high ratio (2:1) for con- buildings to improve their atmospheric environment. Specif-
verting feed mass into weight gain (Gerber et al., 2013). Also, ically, CFD modelling has been used to consider air properties,
broiler houses can be operated in a way that saves water for airflow conditions and the physical features of particular
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
livestock buildings in order to accurately simulate realistic
conditions. A number of case studies and their corresponding
operating conditions can be compared on the basis of air
renewal rates in livestock buildings that use either mechani-
cal or natural ventilation. To do so, CFD models of naturally
ventilated broiler houses must include climate instabilities in
the experimental validation process since there is no air-flow
control. These challenges were addressed in the studies pre-
sented by Lee, Sase, and Sung (2007) whose aim was to accu-
rately simulate air currents under steady-state conditions. Seo
et al. (2009) developed a CFD model that included animal heat
production and served to analyse energy saving strategies.
Although the CFD predictions were acceptable for the given
operating conditions, the incorporation of sensible and latent
heat was not investigated.
Other studies were focused on identifying the thermal and
mass gradients that cause airflow wakes and stagnant regions
within the broiler house (Mogharbel, Ghali, Ghaddar, & Abiad,
2014; Mostafa et al., 2012; Song-Ming, Pei-Feng, Nan, Jin-Song,
& Zhang-Ying, 2012). However, such studies have only
considered temperature and tracer gases such as CO2 and
NH3. Consequently, the CFD model they created served only to
simulate energy balance and air renewal rates. Humidity was
investigated by Rojano et al. (2014), because its inclusion in the
analysis can increase the accuracy of predictions; also, esti-
mating air-renewal rates could improve energy conservation
strategies and assist in the evaluation of air quality during
winter production periods.
Moreover, CFD models can be enhanced by incorporating
the influence of the animals, and by using standardised
sensitivity models and meshing techniques such as recom-
mended by Bjerg et al. (2013). Thus, this work investigates the
spatial and temporal variability of the indoor temperature,
humidity and CO2 levels that occur under typical operating
conditions in order to reinforce the validation and accuracy of
a CFD model. Subsequently, the predictions generated by the
CFD model could be utilised when studying the transport of
gases and in particle tracking and virus dispersion, as was
done by Souris et al. (2014).
1.3. Purpose of the present study
This study uses CFD to model climate and air quality in
naturally ventilated broiler houses. Our investigation merges
climate and air quality issues by proposing a two dimensional
CFD model that addresses the following aims: 1) incorpora-
tion in a distributed climate of the heat, water vapour and CO2
Fig. 1 e (a) Broiler house dimensions and (b) experimental setup.
27
generated by animals as well as gases emitted by combustion
of propane associated with the heating system; 2) inclusion of
short- and long-wavelength radiation with a discrete ordi-
nate model; and 3) validation of the CFD model by deter-
mining its accuracy to replicate the spatiotemporal variability
of temperature and humidity measured within a broiler
house.
Two cases were analysed: one case at the beginning of the
growing cycle when the broilers were young (about 0.5 kg
mass) to assess the influence of the heaters and animals
under low ventilation rates; and another case at the end of the
growing cycle when the broilers were mature (about 3 kg
mass) to assess the influence of different densities of animals
under high ventilation rates and heaters turned off. The
heater was operated using propane and perfect combustion
was assumed in order to compute by-products. Spatial dis-
tribution of temperature and relative humidity sensors
assisted to validate variability of climate and determine ac-
curacy of the CFD model. Since previous research has not
considered together all the sub-models, a two dimensional
approach was used to incorporate, test and validate all the
sub-models using experimental data taken from a cross-view
with respect to the ridgeline of the building.
2. Materials and methods
2.1. Experimental setup
A cycle of organic production (84 d) in a broiler house at the
experimental station “Le Magneraud” located in western
France (46.15 N,
0.69 W), was selected for this study. The
broiler house had a volume of 158.4 m3 and was naturally
ventilated through lateral curtains with a maximum opening
of 0.53 m. The house had an eave height of 1.8 m. The ridge
height (H), equal to 2.6 m, was taken as a reference measure in
order to determine the dimensions of the whole domain under
study; the remainder specifications can be found in Fig. 1a.
Polystyrene was the main material used in all the walls and
roof, but an additional layer of steel panel was added to the
exterior of the roof. The physical properties of these materials
are indicated in Table 1. The two heaters located in the house
were uniformly distributed along the length, and each had a
heating capacity of 4.8 kW. The broilers were kept indoors
during the first 35 d; then two traps (length and height of 2 m
and 0.53 m respectively) were opened to give the chickens
access to a back yard during the day (9:00 h to 17:00 h).
28 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
Table 1 e Material specifications.
Material Density Specific heat
(kg m
3) (J kg
1 C
1)
Wall (polystyrene) 50 1300
Wall (curtain) 920 2100
Roof (metal sheet) 6700 490
Fig. 2 e Broiler geometry and sensor location.
Air velocity and its direction, air temperature, relative hu-
midity and global solar radiation data were collected by a
weather station (AWS310, Vaisala, France) placed 30 m away
of the broiler house. Inside the broiler house, a vertical and a
horizontal plane were established to monitor climate and air
quality. The vertical plane was defined based on three mea-
surement lines (B, C and D) placed at 0.7 m, 1.4 m and 2 m from
the floor. In each measurement line, three temperature
(±0.1  C) and relative humidity (±3%) sensors were placed as
indicated in Fig. 2 and data stored using a portable data logger
(DL-101T USB, Voltcraft, France). The horizontal plane
considered three measurement lines (A, C and E) at 1.4 m from
the floor and a distance between them of 2.38 m; using similar
devices (Fig. 2). Two temperature sensors were used to collect
data from the roof, other two in the ceiling and other two in
the floor under the bedding material; all of them type J and
calibrated. Data from the exterior and interior of the broiler
house were collected with a time step of 10 min for a whole
cycle of production.
Gases measurements were made at six locations on the
vertical plane. The sampling points were located 2.1 m away of
the walls, as indicated in Fig. 2. Additionally, there was a
sampling point in each opening of the curtain at 2.2 m height
above the floor and 0.2 m away from the curtain. Since it was
not possible to take simultaneous measurements at each
sampling point (due to the capabilities of the gas measurement
system) data were recorded serially for each location. Outdoor
and indoor CO2 concentrations were measured by using photo-
acoustic infrared spectrometry (INNOVA 1312, LumaSense
Tech., Germany) coupled with a multipoint sampler and doser
(INNOVA 1303, LumaSense Tech., Germany).
Thermal conductivity Thickness
(W m
1 C
1) (mm)
0.05 40
0.33 0.5
43 0.5
2.2. Modelling settings
Data were collected from 21/April/2014 to 13/July/2014. Mete-
orological data were filtered, excluding all but those cases
when the wind blew perpendicular to the ridgeline (with a
range ±14.4 ), the broiler house having an orientation of 22
NE. These conditions occurred during several short periods
(approximately 1 h), two of them were chosen to represent the
entire cycle that could be replicated in a cross-view, two-
dimensional model of the broiler house. The first period (Case
I) corresponded to the early stage of production (19th day). The
second period (Case II) corresponded to the late stage of pro-
duction (69th day). The CFD model evaluated three scenarios
of Case II: 33%, 66% and 100% of 10 animals m
2. The animal
area was assumed to be a region located at floor level with a
height of 200 mm (Fig. 1a) where the heat, water vapour and
CO2 produced by the animals and litter were assigned.
Because air moved perpendicular to the ridgeline in the
two cases, homogeneity was assumed along the length of the
building (i.e. uniform distribution of the animals and litter),
favouring a two-dimensional modelling approach. The cross-
view plane was located at 3.6 m from one lateral wall, and this
plane included a heater and a trap (Fig. 2). The heater was
placed at the centre and 1.7 m above the floor with 0.05 m and
0.4 m for height and length, respectively; in which the gen-
eration of volumetric heat, water vapour and CO2 were set
according to the heater's specifications; whereas the trap was
located at the bottom section of the wall. The assumption of
homogeneity along the ridgeline reduced by half the size of
the region in which data was collected and enabled us to
locate that area away of the main entrance in order to
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
minimise the disturbances occurring near the opening during
feeding, supervision and animal management.
2.2.1. Governing equations
The modelling stage was based on a finite-volume approach
involving the NaviereStokes equations that follow the
Reynolds-average method and rely on two additional equa-
tions to determine the turbulence quantities (kinetic energy k
and dissipation rate 3). Ansys Fluent™ (2014) was used to find
the numerical solutions, and the turbulence ke3 realizable
model was chosen since it has proved to be consistent in
various similar studies dedicated to livestock buildings (Bjerg
et al., 2013; Norton, Grant, Fallon, & Sun, 2010; Ramponi &
Blocken, 2012).


vf vuj f v vf
þ ¼ Gf þ Sf
vt vxj vxj vxj
where f is the variable of interest (velocity in the j direction,
temperature, turbulent kinetic energy, dissipation rate of
turbulent kinetic energy and mass fraction), and t, Gf, Sf
represent time, diffusion coefficient, and source term,
respectively. Gravitational effects were included in the source
term of the momentum equations.
The solar radiation and thermal radiative energy were
evaluated by means of the discrete ordinate model (DOM). The
DOM is based on the radiative transfer equation (Eq. (2)) which
has been used in several studies of greenhouses, such as the
work of Bournet, Khaoua, and Boulard (2007) where accurate
predictions were obtained:
Z4p
dIðr; sÞ sT4 ss
þ ða þ ss ÞIðr; sÞ ¼ an2 þ Iðr; s0 ÞFðs; s0 ÞdU0
ds ap 4p
0 from field measurements such as wind speed (Case I: 3.1 ms
1;
Eq. (2) includes the variation of the luminance I, the radiation
decay, and gain of luminance by emission and diffusion.
2.2.2. Animal and litter sources
The heat, water vapour and CO2 emitted by the animals were
estimated according to CIGR (2002), which recommends a
basal heat metabolism empirically related to the weight of the
animal. Eqs. (3)e(6) compute the total sensible and latent heat,
respectively. Total heat was calculated by applying Eq. (3):


20ð20
Ti Þ
AH ¼ 10:62m0:75  1 þ
1000
where m is animal mass. Ti is average indoor temperature. AH
is the total heat generated. Animal mass was estimated based
on a curve identified in the work of Rojano et al. (2014):
2 variability during the time considered for the two cases. The
m ¼ 0:0003346d þ 0:0236706d
0:024548
The sensible heat AHS. was estimated with Eq. (5) and
latent heat deduced from Eqs. (3) and (5):


0:228  
AHS ¼ 0:61  AH
 T2i
1000
AHL ¼ AH
AHS
The amount of CO2 produced by the animals was calcu-
lated as proportional to the total amount of heat produced by
29
the animals; for 1 kW of heat of the broilers also produced
0.185 m3 [CO2] h
1.
Additionally, the estimations of sensible and latent heat
produced by the litter were based on previous experimental
results that established a correlation between total heat pro-
duced by the animal and that produced by the litter. Meda
(2011) found that 10% of the total heat produced by the ani-
mal would be the litter heat and that amount of heat could be
split into 60% and 40% for sensible and latent heat,
respectively.
2.2.3. Heater
To accomplish temperature requirements, the model consid-
ered the heater to be the main heat source during the early
stage of animal production. The amount of heat was assumed
(1) to be constant, and due to the combustion of propane there
were by-products, such as water vapour and CO2. A perfect
combustion was assumed, as was a perfect rate of production
for these two gases, following the stoichiometry equation:
C3 H8 þ 5O2 / 3CO2 þ 4H2 O (7)

1
The amount of 0.015 g [C3H8] s (propane density equal to
493 kg m
3) used by the heater yielded an estimated amount of
0.043 g [CO2] s
1 and 0.023 g [H2O] s
1. From the total amount
of heat emitted by the heater, 677.6 W and 52.4 W were sen-
sible (HHS) and latent (HHL) heat, respectively.
2.3. Initial and boundary conditions
The simulation set the initial conditions based on the out-
(2) comes of the steady state conditions that used information
Case II: 1.75 ms
1), air temperature (Case I: 15.9  C; Case II:
25.4  C), absolute humidity (Case I: 6.3 g [H2O] kg
1 [dry air];
Case II: 10.2 g [H2O] kg
1 [dry air]) and CO2 concentration (Case
I: 417 ppm; Case II: 380 ppm). In these settings temperatures
from indoor floor and solar radiation were set according to the
measurements observed in the broiler house.
The left side of the domain (Fig. 3) was set as the inlet
(ventilation inlet) whereas the right side of the domain as the
outlet (pressure outlet) at atmospheric pressure (po). of
100.4 kPa (i.e. 80 m above sea level). The air temperature, ab-
solute humidity and CO2 were set at the inlet boundary ac-
(3)
cording to the field measurements. The top side of the domain
was set as a wall with slip conditions. The top, inlet and outlet
boundaries were designated as short and long wave radiation
emitters (Table 4) where the values were set as constants
since field measurements of solar radiation had less than 5%
(4)
internal floor was set with a constant temperature observed in
field measurements and external floor of the poultry house
was set as adiabatic. All the walls related to the broiler house
were treated using the enhanced wall treatment according to
(5) Ansys Fluent™ (2014) in order to improve calculations of the
thermal and convective boundary layer. The simulation
considered an incompressible ideal gas under the influence of
(6)
the gravitational forces. The numerical iterations were ini-
tialised for the entire domain with input values indicated in
Table 2. The wind profile, kinetic energy and turbulence
30 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

Fig. 3 e Boundary conditions of the domain studied.

Table 2 e Boundary conditions and sources.

Boundary Value Reference
Inlet Wind velocity, uo(x2) Richards and Hoxey (1993)
Turbulence kinetic energy, kz Richards and Hoxey (1993)
Turbulence energy dissipation, 3z Richards and Hoxey (1993)
Outdoor temperature, To Experimental data
Absolute humidity, h Experimental data
Air-CO2 Experimental data
Outlet Atmospheric pressure, po Experimental data
Top Shortwave radiation, SR Experimental data
and longwave radiation, q
Floor indoor Indoor floor temperature, Tfi Experimental data
Floor outdoor Outdoor floor heat flux, HFfo Fixed
Animal heat Sensible heat from animal, AHS CIGR (2002)
Latent heat from animal, AHL CIGR (2002)
Animal-CO2 CIGR (2002)
Litter heat Sensible heat from litter, LHS Meda (2011)
Latent heat from litter, LHL Meda (2011)
Litter-CO2 CIGR (2002)
Heater Sensible heat from heater, HHS Stoichiometric equation
Latent heat from heater, HHL Stoichiometric equation
Heater-CO2 Stoichiometric equation

energy dissipation were calculated based on the equations where N is the decay rate starting with the concentration C(0),
recommended by Richards and Hoxey (1993). and the value N denotes the efficiency of the ventilation.

2.4. Simulation of the ventilation rates

3. Results
To estimate ventilation rates using a virtual tracer gas, a two-
step numerical procedure developed by Ould Khaoua,
3.1. Experimental data
Bournet, Migeon, Boulard, and Chasse riaux (2006) was used.
Firstly, a numerical solution was found at every time step, and
A summary of the climate data for the whole growing period is
the tracer gas was incorporated into the species equation.
provided in Fig. 4 including wind velocity measurements and
Secondly, all the cells within the broiler house were set to an
their corresponding temperature and humidity values within
arbitrary concentration equal to 1 and all cells outdoors were
the psychrometric chart. Wind velocity is emphasised since in
set to a concentration equal to 0. Because the concentration of
conjunction with buoyancy forces it can be the driving force to
the tracer gas decreases based on wind velocity, the tracer
change the internal climate and air quality conditions.
concentration (C) can be estimated as a function of time (t). An
Furthermore, according to the coefficient of variation, wind
exponential decay over time can be identified:
velocity was identified as the variable with the second highest
CðtÞ ¼ Cð0Þe
Nt (8) dispersed values (after solar radiation) among four climate
factors logged (Table 3).
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8 31

Fig. 4 e Outdoor climate data every 10 min in 84 days.

Table 3 e Outdoor climate summary.

Wind velocity, uo Wind direction, a Outdoor temp, To Absolute humidity, h Solar radiation, SR
Average 1.3 160.1 15.6 10.2 364.5
Standard deviation 0.9 77.3 5.3 2.87 309
Coefficient of variation 0.7 0.48 0.34 0.27 0.84

The two cases (I and II) described in Table 4 as represen- 3.2. CFD model
tative of the contrasting stages of the production cycle, had a
time interval of at least 1:30 h (Fig. 5); such data had stability in The domain indicated in Fig. 3 was used to create the grid.
wind direction but stability was not guaranteed in tempera- Three independent grid densities were tested following the
ture, absolute humidity and CO2 concentration. Over the same grid convergence index (GCI) criterion to mesh the calculation
time interval, spatially distributed sensors measured indoor domain. In each grid, the same procedure related to the cell
temperature, absolute humidity and CO2 (as shown in Fig. 2). distribution was applied; a higher density was imposed near
Solar radiation measurements were collected in order to any wall which means, in the present case, a neighbouring cell
incorporate the effects of thermal radiation. Because the size decreased by 25% if it was closer to the wall. Also, the CFD
simulated period was short, the sun was assumed to be fixed. model in each of the grids included the specifications of wall,
The azimuth (Case I: 207.7 ; Case II: 235.6 ) and elevation (Case roof and curtain thicknesses and their corresponding thermal
I: 57.3 ; Case II: 57.5 ) of sun for the simulated period were both properties as indicated in Table 1.
required to find the angle of irradiation. Radiation produced by Evaluation of the grid performance was made using input
long waves was included considering the effective sky tem- data taken in Case I under steady-state conditions with out-
perature (Tsky ¼ 0.0552T1.5 o ) according to Duffie and Beckman door climate at 14:40 h. The DOM was implemented using 15
(1991) assuming a grey body emitter; then the radiative flux iterations of the energy equation and the number of angles
was calculated using q ¼ 3sT4sky with 3 ¼ 1.24(10  e/ was increased up to 7 where results became stabilised. The
To þ 273.15)1/7 (Brutsaert, 1975). solver (Fluent, 2014) found the numerical solution by following
the method SIMPLEC with a skewness correction equal to two
and a second-order upwind discretisation scheme. The
Table 4 e Cases simulated in CFD. convergence had to meet residuals <10
6 for the continuity, x
and y velocity, kinetic energy and turbulence energy
Case I Case II
Day of production 19 69
Time 14:40e16:20 15:40e17:00
Time steps 10 9
Curtain opening 60 mm 300 mm
Traps Closed Open
Heater On Off
Estimated hen weight 0.5 kg 2.9 kg
Density of hens 10 hens m
2 Variable
Other input data Average Average
Outdoor temperature 16.3 (±0.6)  C 25.6 (±0.7)  C
Floor indoor temperature 34 (±1)  C 24 (±1.2)  C
Shortwave radiation 743 (±69) W m
2 835 (±120) W m
2
Longwave radiation 199(±5) W m
2 243 (±7) W m
2
CO2 417 ppm 380 ppm
Fig. 5 e Outdoor climate for Case I and Case II.
32 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

dissipation; and <10
12 for water vapour, discrete ordinates,

energy and CO2. Since gravity caused instability in the
convergence, the solver found a solution by gradually incre-
menting its value from 0 to 9.81 m s
2.
From the numerical solution found in each grid, the ve-
locity profile was computed in the vertical axis (dimension H
in Fig. 2), and this profile was used to calculate the GCI
following the methodology used by Roache (1994). Upon the
velocity profile, ratios of cell refinement (r) and GCI with a Fig. 6 e Observed and predicted indoor temperature across
safety factor equal to 1.25 were calculated between grids the vertical plane specified in Fig. 2.
(Table 5). The GCI provided the expected percentage of relative
error between the grids, assuming the finer grid produces the
outcomes closer to the exact solution. For instance, GCI be-
tween grid 2 and 3 did not improve significantly; thus grid 2
was chosen for the CFD model having a maximum aspect ratio
of 65, a minimum orthogonality of 0.5 and yþ values were less
than 300 near the walls. The same grid was chosen for Case II
with the only modifications being in the opening of the cur-
tains and traps, which afforded a higher ventilation rate and
thus a faster convergence.
The initial conditions of the CFD model for Case I were Fig. 7 e Observed and predicted indoor temperature across
defined by the numerical solution found during the bench- the horizontal plane specified in Fig. 2.
marking of grid 2, whereas the initial conditions for Case II
were defined using the numerical solution under steady-state
conditions and following the corresponding input data related
to the heat and mass sources, as indicated in Fig. 5 and Table and E. Field observations along measurement lines A and E
4. Then the two cases, using the same settings, were consid- showed that a homogeneity of temperature prevailed along
ered under unsteady-state conditions with the same time step the length of the broiler house as the effects of the heater
as the experimental data and modifications in the conver- diminished (Fig. 7). Also, the assumption that the heat emitted
gence settings as adaptive. Case I and II involved 10 and 9 time by animals and litter was distributed along the entire indoor
steps, respectively; solutions were found between 2500 and bottom area was deemed acceptable because experiments
5500 iterations. found a minimum of temperature differences between mea-
surement lines A and E.
3.3. Case I The same sensor array used to collect temperature data
was also used to collect data on absolute humidity as shown in
The interactions occurring among the different heat and mass Figs. 8 and 9 aiding in the analysis of latent heat. Those related
sources (i.e. water vapour and CO2 produced by the heater, to the vertical plane resulted in an RMSE of 0.26 g [H2O] kg
1
animals and litter) shaped, via air currents, the thermal and [dry air]. The absolute humidity data across the horizontal
mass gradients that were identified using sensors placed over plane produced similar prevailing values along measurement
the vertical and horizontal planes (Fig. 2). The experimental lines A and E, which were lower than the values along the
data sets were compared with the predictions of the CFD measurement line C as a consequence of the water vapour
model, both for temperature and absolute humidity along emitted by the heater. Both experimental and simulated data
every measurement line. Temperature on the vertical plane showed almost homogeneous conditions across both planes;
(Fig. 6) showed good agreement (RMSE ¼ 1.0  C); the lowest
experimental values were observed at the upper part (mea-
surement line B) and the homogeneous area between mea-
surement lines C and D since air entering and leaving the
upper part of the broiler house dissipated heat and gases. Data
recorded on the horizontal plane showed higher temperatures
along measurement line C, due to the proximity of the heater,
compared to the temperatures along measurement lines A

Table 5 e Grids evaluated.

Grid Number of cells Ratio GCI
(r) (%)
1 136,200 e e
2 185,220 1.35 0.75 Fig. 8 e Observed and predicted indoor absolute humidity
3 229,000 1.23 0.09
across the vertical plane specified in Fig. 2.
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
Fig. 9 e Observed and predicted indoor absolute humidity
across the horizontal plane specified in Fig. 2.
however, the field measurements across vertical plane were
more heterogeneous than the values produced by the CFD
model. A comparison of the average overall external and in-
ternal absolute humidity indicated that the animals, litter and
heater could increase the internal absolute humidity by 25%
following to the operating conditions of Case I.
The overall conditions established for the broiler house
implied that homogeneity was prevalent indoors, except
within the zones near the heater and the air entrance and exit.
The effects of the heater created primarily the gradient for
temperature and humidity. The contributions from the ani-
mals and litter were scarcely perceptible at one of the bottom
corners (Fig. 10). They represented approximately 20% of the
total indoor generated heat and were distributed across the
bottom area. The water vapour was presented as relative
humidity gradients in order to show that there was no
condensation during the simulated time period; given that
condensation could be a significant source error. The CO2
gradient was estimated by assuming the exterior and heater
contributions to be constant amounts and only the CO2
amount produced by the animals and rearing system to be
variable. Those conditions were chosen in order to replicate in
Fig. 10 e CFD results for Case I at 15:30 h (on 5/04/2014): (a) temperature, (b) relative humidity, (c) CO2 concentration and (d)
stream lines.
33
Table 6 e Water vapour and carbon dioxide for Case I.
Absolute humidity,a CO2,b ppm
g [H2O] kg
1 [dry air]
Exp CFD model Exp CFD model
Average indoors 8.1 7.9 929 852
Standard deviation 0.23 0.15 102 38
a
RMSE 0.292.
b
RMSE 134.
the CFD model the expected gradients of CO2 (in ppm), as
indicated in Fig. 10.
Average indoor conditions were compared by selecting 7
and 6 locations for logging data associated with water vapour
and CO2, respectively, as indicated in Fig. 2, for a 100-min in-
terval. Results of experimental and simulated data are pre-
sented in Table 6 with their corresponding standard
deviations. The average indoor conditions shown by the
experimental data were found to be in good agreement with
those of the CFD model. However, the CFD model produced
prevalent underestimations which could be originated by the
process of simplification used to represent the amount of heat
from heater in a two dimensional approach.
3.4. Case II
The heater being turned off and significant amount of venti-
lation caused thermal gradients to be attenuated making in-
door climate and air quality conditions with almost similar to
those conditions presented in the exterior. Increasing the air
circulation accelerated the removal of heat and gases gener-
ated indoors, and with the traps opened, not all the animals
remained inside the broiler house (according to the field ob-
servations, approximately 75% of all the animals did remain
indoors; given that they could leave the building and return
freely and the percentage could have changed during the 90-
min simulation). Also, as indicated by the experimental data
and the CFD model, the heating of the interior could have
induced more broilers to move outside. These factors
34 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

exacerbated the difficulty of accurately testing the effects that

heating may have had on the animals. In order to overcome
this problem, the CFD model evaluated three animal densities:
33%, 66% and 100% of 10 animals m
2. The latent and sensible
heat generated by the litter was considered fixed (also
assuming all the animals stayed indoors), and thus quanti-
fying the heat generated solely by the broilers. At this stage of
production, each animal weighed approximately 3 kg and the
total heat contribution was approximately 23 W animal
1 (in
contrast to Case I, wherein animals contributed by approxi-
mately 7 W animal
1).
The experimental data showed that either plane (vertical
or horizontal, Fig. 2) could be used to represent the average Fig. 12 e Observed and predicted indoor temperature
indoor climatic conditions since the field observations for across the horizontal plane specified in Fig. 2.
each plane were nearly identical due to the heater being
turned off and high ventilation rates caused by both curtains
opened 300 mm. The evolution of the field measurements of
temperature and absolute humidity across both the vertical
and horizontal plane was simplified to show only the average
values with their corresponding standard deviations. The
heating process experienced in the experiment was found to
be equal in both planes as a result of a decrement in wind
velocity and an increment in outdoor temperature; in
contrast, absolute humidity was found to fluctuate between 10
and 11 g [H2O] kg
1 [dry air]. Predicting the temperature across
the horizontal plane could provide the best results if all the
animals were assumed to be indoors since predicting tem-
perature would result in an overestimation across the vertical
plane (Figs. 11 and 12). Experimental data of absolute hu-
midity falls between the scenarios of 66% and 100% for the
vertical plane and between 33% and 66% for the horizontal Fig. 13 e Observed and predicted indoor absolute humidity
plane (Figs. 13 and 14). Standard deviations of the predicted across the vertical plane specified in Fig. 2.
temperature were 0.67  C, 0.77  C and 0.96  C for the scenarios
33%, 66% and 100%, respectively, whereas for predicted ab-
solute humidity, the values were 0.22 g [H2O] kg
1 [dry air],
0.30 g [H2O] kg
1 [dry air] and 0.39 g [H2O] kg
1 [dry air] for the
scenarios 33%, 66% and 100%, respectively.
The comprehensive analysis of the experimental and
simulated temperature and absolute humidity data is sum-
marized in Figs. 15 and 16, respectively. A comparison of each
sampling point along the vertical plane (7 sensors) and the
evolution along the 9 time steps enabled the calculation of the
RMSE for temperature and absolute humidity. The CFD
model's temperature prediction improved as the number of

Fig. 14 e Observed and predicted indoor absolute humidity

across the horizontal plane specified in Fig. 2.

animals indoors increased, and a best RMSE (equal to 0.9  C)

Fig. 11 e Observed and predicted indoor temperature
across the vertical plane specified in Fig. 2.
was achieved when assuming all animals were indoors; in
contrast, the best prediction for absolute humidity was found
when assuming only 66% of the animals were indoors
(RMSE ¼ 0.37 g [H2O] kg
1 [dry air]). These results were
acceptable approximations of the field observations made
that indicated for about 75% of the animals were indoors.
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
Fig. 15 e Observed and predicted indoor temperature
prediction for 33%, 66% and 100% of the 10 animals m¡2.
Fig. 16 e Observed and predicted indoor absolute humidity
prediction for 33%, 66% and 100% of the 10 animals m¡2.
The stability of input data for Case II allowed simulating
the three scenarios (33%, 66% and 100%), emphasizing that
animal's heat could increase the average indoor temperature
and absolute humidity. Since the best prediction fell within
66% and 100%, gradients of temperature, relative humidity,
Fig. 17 e CFD results for Case II at time 15:30 h (29/06/2014) with 100% of the animals indoors (a) temperature, (b) relative
humidity, (c) CO2 concentration and (d) stream lines.
35
CO2 and air currents are presented in Fig. 17 for the scenario of
100%.
Animals within the broiler house contributed to the main
indoor heat source and created thermal gradients. The heat
from the animals accumulated near the bottom since that
area prevailed as stagnant zone (Fig. 17). Relative humidity
and CO2 were, in general, homogeneous with high and low
values located at the stagnant area and air entrance, respec-
tively. Homogeneity was possible due to the distributed heat
and gas sources caused only by the animals and litter. Air
circulation created a whirl with a centre near the heater; the
outlets at the curtains and traps were balanced with the ca-
pacity of the opening of the curtains at the air entrance.
4. Discussion
To assess the heterogeneity of temperature and absolute hu-
midity distributions provided by the vertical and horizontal
planes, the experimental and simulated data were compared
on the basis of the coefficient of variation. The values for Case
I and II are presented in Tables 7 and 8, respectively. In Case I,
the experimental and simulated data showed more hetero-
geneity in temperature than absolute humidity. Nevertheless,
more homogeneity was found within the simulated data than
within the experimental data. That is, the ideal conditions
assumed for computational simulations may have caused
homogeneous outcomes. In contrast, the experimental data
was collected under outdoor conditions mainly influenced by
the natural ventilation, the heater, and the animal's activity
and also depended on the accurate real-time readings of the
sensors in response to those changes.
The experimental data, collected in Case II, showed a
similar homogeneity across both the vertical and horizontal
planes. The simulated data indicated that the heterogeneity
increased as the animal density increased. The homogeneity
of the temperature measured across the vertical plane
matched more closely with the simulated data between the
scenarios of 66% and 100% of the 10 animals m
2. With respect
to absolute humidity, the experimental data across the verti-
cal plane matched more closely with simulated data of the
scenario of 66%. Also, the absolute humidity readings across
36 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8

Table 7 e Coefficient of variation for the temperature of the broiler house climate in Case I.
Vertical plane Horizontal plane
Average Standard deviation Average Standard deviation
Temperature CFD 0.0020 0.00009
EXP 0.0056 0.00028 0.0073 0.00017
Absolute humidity CFD 0.0004 0.00001
EXP 0.0018 0.00035 0.0018 0.00032

Table 8 e Coefficient of variation for the absolute humidity of the broiler house climate in Case II.
Vertical plane Horizontal plane
Average Standard deviation Average Standard deviation
Temperature CFD 100% 0.0022 0.00015
CFD 66% 0.0015 0.00011
CFD 33% 0.0007 0.00005
EXP 0.0008 0.00040 0.0006 0.00045
Absolute humidity CFD 100% 0.0275 0.00183
CFD 66% 0.0204 0.00170
CFD 33% 0.0105 0.00110
EXP 0.0154 0.00433 0.0272 0.00572

the horizontal plane proved to be more heterogeneous than
the absolute humidity readings across the vertical plane; most
likely, the opened traps, which were in line with the vertical
plane, increased the air circulation and created this effect.
A modification of the heater representation in the 2D CFD
model, given that the heater had in fact a cylindrical shape
instead of the prisma could have induced errors. However, the
total amount of heat supplied by the heater that corresponded
to a section 1 m in depth was followed. Thus, thermal and
mass gradients obtained from the CFD model could more
precisely approximate the average indoor climatic conditions
and gas concentrations.
Radiative heating in the CFD model was mainly influenced
by the solar radiation and the different temperatures gener-
ated inside the broiler house. The highest interior tempera-
tures occurred at the heater and the bottom area, whilst the
lowest temperatures occurred along the walls; all these
different temperatures determining the average indoor tem-
perature. The irradiative energy from the sun was signifi-
cantly attenuated by the layer of polystyrene that had the
effect of insulation on the roof; the amount of heat flow
transferred through the roof was determined by placing two
temperature sensors on the roof and two others at the ceiling.
The experimental data estimated a heat flow lost through the
roof with respect to the total heat generated inside (all ani-
mals and heater) of approximately 2% and
20% for Case I and
II, respectively. Nonetheless, the solar radiation that pene-
trated through the lateral curtains added heat to the indoor
climate that was not possible to prove experimentally.
In case I, 34  C, which was the average experimental indoor
floor temperature, used as the thermal boundary condition
also matched average temperatures found at the broilers'
feather coat (Zhao, Xin, & Dong, 2013). This particular tem-
perature also facilitated bacterial activity in the litter and
straw that was replaced periodically to cover the litter and
provided clean bedding. In contrast, in Case II the average
indoor floor temperature was significantly lower (24  C), due to
the increased ventilation and because the animals could leave
the broiler house.
Given that exterior climate and features of building were
properly represented in the 2D CFD model, the difference be-
tween the experimental and simulated data for the two cases
could have been mainly caused by the restrained capability of
the governing equations used to represent the actual dy-
namics of heat and mass generated by the animals and rear-
ing system. For instance, moisture content in the rearing
system, as an element not considered in the governing
equation, can modify the amount of the heat produced.
Furthermore, two studies (Meda, 2011; Pedersen et al., 2008)
have suggested several factors that may also cause the vari-
ability of the heat and mass emissions from the rearing sys-
tem. The more important factors are the average interior
temperature and moisture, the carbon content in litter and
bedding, and the degradation time.
Based on the outcomes of computational simulations, the
ventilation rates were calculated. In Case I, accuracy was
easily accomplished because only the opening of the curtains
determined the ventilation rates. In Case II, opening the traps
caused a misrepresentation of the real conditions because the
traps were located at specific points along the length of the
broiler house. In general, air changes per hour increased
approximately 2.4 times in Case II, even though in Case II the
mean outside air velocity decreased gradually from 1.8 m s
1
to 1.1 m s
1 (Fig. 18).
5. Conclusions
The improved CFD model designed to simulate the climate
inside a broiler house takes into account the main heat and
mass sources (heater, animals and litter). Experimental data,
obtained by means of temperature and humidity sensors
spatially distributed, was used to determine the accuracy of
the CFD model. Two dissimilar cases were used to evaluate
 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
Fig. 18 e Estimation of the ventilation rates through CFD.
the response of the CFD model to the heat and mass produced
under different operating conditions. These two cases were
chosen from a whole cycle of production due to the stability of
the field measurements when air blew perpendicular to the
ridgeline of the broiler house.
The operating conditions of the first case demonstrated the
reliability over time and space of the CFD model to predict
temperature, water vapour and CO2. The second case exam-
ined the effects of the sensible and latent heat in the broiler
house simulating three different densities of animals through
the CFD model. In both cases, temperature and absolute hu-
midity gradients were reviewed in a vertical and horizontal
plane by means of the coefficient of variation. Based on the
accuracy reached for climate in each of the two cases; the
same CFD model was used to predict the air exchange rates.
Since the data collected in the field pertained to the abso-
lute humidity and CO2 emitted by the animals and the litter
together, it cannot be claimed that errors could have been
caused by the estimations that CIGR (2002) or Meda (2011)
recommended. The inherently high variability of gas
amounts emitted could have been caused by the manner in
which the broiler house was managed and the rearing system
operated. Nevertheless, the CFD model was able to demon-
strate that the animals, the litter and the by-products gener-
ated by the heater must be considered in order to improve the
accuracy of the calculations used to balance heat and mass
adequately. Hence, the approach used in this study should
also serve to establish future methods for estimating air
renewal rates together with the gas emissions occurring in
naturally ventilated broiler houses. Since governing equations
incorporated in the 2D CFD model had acceptable results, the
same equations can be used for improvement of the model-
ling conducted in three-dimensional studies.
Acknowledgements
We appreciate the support of the staff at INRA-Le Magneraud
and INRA-Rennes for assisting in the installation and super-
vision of the Data Acquisition System.
37
references
Bjerg, B., Cascone, G., Lee, I.-B., Bartzanas, T., Norton, T., Hong, S.-
W., et al. (2013). Special issue: emissions from naturally
ventilated livestock buildings review modelling of ammonia
emissions from naturally ventilated livestock buildings. Part 3:
CFD. Biosystems Engineering, 116(3), 259e275.
Bournet, P. E., Ould Khaoua, S. A., & Boulard, T. (2007). Numerical
prediction of the effect of vent arrangements on the
ventilation and energy transfer in a multi-span glasshouse
using a bi-band radiation model. Biosystems Engineering, 98(2),
224e234.
Brutsaert, W. (1975). On a derivable formula for long-wave
radiation from clear skies. Water Resources Research, 11(5),
742e744.
CIGR. (2002). Heat and moisture production at animal and house
level. In S. Pedersen, & K. Sallvik (Eds.), 4th Report of working
group on climatization of animal houses. Horsens, Denmark:
CIGR. Published by Danish Institute of Agricultural Sciences.
Duffie, J., & Beckman, W. A. (1991). Solar engineering of thermal
processes (2nd ed.). New York, USA: John Wiley & Sons.
Fluent. (2014). Ansys Fluent user's guide. Canonsburg, Pa.: ANSYS,
Inc.
Gerber, P. J., Steinfeld, H., Henderson, B., Mottet, A., Opio, C.,
Dijkman, J., et al. (2013). Tackling climate change through
livestock: A global assessment of emissions and mitigation
opportunities. Rome: FAO.
Lee, I., Sase, S., & Sung, S. (2007). Evaluation of CFD accuracy for
the ventilation study of a naturally ventilated broiler house.
JARQ, 41, 53e64.
Meda, B. (2011). Une approche dynamique des flux d'elements et
d'energie des ateliers de production avicole avec ou sans parcours:
Conception et Application du Modele MOLDAVI (PhD thesis).
Rennes, France: Agrocampus Ouest.
Mogharbel, O. E., Ghali, K., Ghaddar, N., & Abiad, M. G. (2014).
Simulation of a localized heating system for broiler brooding
to improve energy performance. International Journal of Energy
Research, 38, 125e138.
Mostafa, E., Lee, I.-B., Song, S.-H., Kwon, K.-S., Seo, I.-H., Hong, S.-
W., et al. (2012). Computational fluid dynamics simulation of
air temperature distribution inside broiler building fitted with
duct ventilation system. Biosystems Engineering, 112(4),
293e303.
Norton, T., Grant, J., Fallon, R., & Sun, D.-W. (2010). Optimising the
ventilation configuration of naturally ventilated livestock
buildings for improved indoor environmental homogeneity.
Building and Environment, 45(4), 983e995.
Ould Khaoua, S. A., Bournet, P. E., Migeon, C., Boulard, T., &
Chasse riaux, G. (2006). Analysis of greenhouse ventilation
efficiency based on computational fluid dynamics. Biosystems
Engineering, 95(1), 83e98.
Pedersen, S., Blanes-Vidal, V., Joergensen, H., Chwalibog, A.,
Haeussermann, A., Heetkamp, M. J. W., et al. (2008). Carbon
dioxide production in animal houses: a literature review.
Agricultural Engineering International, CIGR Journal, X, 1e19.
Ramponi, R., & Blocken, B. (2012). CFD simulation of cross-
ventilation for a generic isolated building: Impact of
computational parameters. Building and Environment, 53,
34e48.
Richards, P. J., & Hoxey, R. P. (1993). Appropriate boundary
conditions for computational wind engineering models using
the ke3 turbulence model. Journal of Wind Engineering and
Industrial Aerodynamics, 47, 145e153.
Roache, P. J. (1994). Perspective: a method for uniform reporting of
grid refinement studies. Journal of Fluids Engineering, 116(3),
405e413.
38 b i o s y s t e m s e n g i n e e r i n g 1 3 6 ( 2 0 1 5 ) 2 5 e3 8
Rojano, F., Bournet, P.-E., Robin, P., Hassouna, M., Choi, C. Y., &
Kacira, M. (2014). Test of two different schemes through CFD
to include heat and mass transfer induced by animals inside a
broiler house. In International conference of agricultural
engineering (Zurich, Switzerland, July 6e10).
Seo, I.-H., Lee, I.-B., Moon, O.-K., Kim, H.-T., Hwang, H.-S.,
Hong, S.-W., et al. (2009). Improvement of the ventilation
system of a naturally ventilated broiler house in the cold
season using computational simulations. Biosystems
Engineering, 104(1), 106e117.
Song-Ming, Z., Pei-Feng, L., Nan, J., Jin-Song, H., & Zhang-Ying, Y.
(2012). Optimization of local ventilation system for gaseous
pollutants removal in broiler house using CFD simulation. In
International conference of agricultural engineering CIGR-AgeEng
2012: Agriculture and engineering for a healthier life (Valencia,
Spain July 8e12).
Souris, M., Selenic, D., Khaklang, S., Ninphanomchai, S., Minet, G.,
Gonzalez, J.-P., et al. (2014). Poultry farm vulnerability and risk
of avian influenza re-emergence in Thailand. International
Journal of Environmental Research and Public Health, 11(1),
934e951.
Zhao, Y., Xin, H., & Dong, B. (2013). Use of infrared thermography
to assess laying-hen feather coverage. Poultry Science, 92(2),
295e302.