b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5

Available online at www.sciencedirect.com

ScienceDirect

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

Research Paper

Evaluation of a precision air-supply system in

naturally ventilated freestall dairy barns

Xiaoshuai Wang a,b, Guoqiang Zhang a,*, Christopher Y. Choi c

a
Department of Engineering, Aarhus University, Blichers All
e 20, P.O. Box 50, 8830, Tjele, Denmark
b
College of Biosystems Engineering and Food Science, Zhejiang University, 866 Yuhangtang Road, Hangzhou,
310058, Zhejiang, China
c
Department of Biological Systems Engineering, University of Wisconsin-Madison, 460 Henry Mall, Madison, 53706,
WI, United States

article info
A good ventilation system is critical to the success of any commercial cattle raising
Article history: operation. Most dairies rely on natural ventilation systems, with supplemental cooling
Received 18 January 2018 systems (e.g. cooling fans, ceiling ventilation, polytube cooling, etc.) to help mitigate the
Received in revised form heat stress imposed on the cows when natural ventilation proves insufficient, usually
4 August 2018 during hot, humid and windless situations. However, the effects of these supplemental
Accepted 21 August 2018 cooling are insufficient in many cases. Therefore, the current study assessed an alternative
supplemental cooling system, referred to as a precision air supply system (PASS), with
respect to how effectively it could remove excess heat from cows inside barns. The effects
Keywords: of three key parameters (i.e. supply-air speed, distribution tube diameter, and air-supply
Freestall dairy barn angle) on the performance of PASS were evaluated with respect to both a standing cow
Supplemental cooling system and a reclining cow, by means of a response surface methodology and a series of properly
Precision air supply system validated computational fluid dynamics (CFD) simulations. The results showed that PASS
Heat stress mitigation performed better when the bulk air was still and that all three investigated parameters
Numerical simulation helped to improve heat dissipation. A perturbation plotting indicated that adjusting the air-
supply angle may well be the best first option when seeking to improve the cooling per-
formance of PASS, followed by increasing tube diameter as the next best option. When
serving as a supplemental cooling system deployed in a freestall, naturally ventilated barn,
PASS should work well to mitigate heat stress when the temperature and humidity index
(THI) increases beyond the threshold value (i.e., THI > 68).
© 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.

stressed cow will eat less and produce less milk (Schneider,
1. Introduction Beede, & Wilcox, 1988; West, 2003). The US dairy industry
alone suffers a tremendous economic loss due to heat stress
In many regions, heat stress is a common threat to profitable (annual losses averaged $897 million for dairy) (St-Pierre,
dairy cattle farming during warm periods because a heat

* Corresponding author.
E-mail address: guoqiang.zhang@eng.au.dk (G. Zhang).
https://doi.org/10.1016/j.biosystemseng.2018.08.005
1537-5110/© 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.
2 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5

overly chilly downdraft). Most systems using this concept for

Nomenclature summer cooling employ both low-volume, high-speed and
high-volume, low-speed supplemental cooling fans that cool
Roman letters symbols the cows when they are confined for extended periods (usu-
D distribution-tube diameter (m) ally in holding pens, feeding alleys, and freestall areas
K number of investigated parameters (Holmes et al., 2013)) and whenever the cows voluntarily
C specified number of replicated central points spend considerable amount of time bunched closely together.
U velocity (ms
1) This practice has proven to be effective regardless of which
S source term type of fan is operated (Kammel, Raabe, & Kappelman, 2003;
SA skin surface area (m2) Worley & Bernard, 2005, 2008); however, such hybrid sys-
W body weight (kg) tems generally increase the speed of the air passing over the
Vfan airflow rate provided by fan cows; thus, their cooling capacity declines as indoor temper-
DP differential pressure (Pa) ature and humidity levels rise, because such rises hinder the
T time (s) heat loss that results from both cutaneous convection and
Tt tissue temperature ( C) evaporation. In another words, the supplemental cooling fans
V supply-air velocity (ms
1) can fail to cool a cow when the indoor air is hot and humid.
N number of tests Also, these supplemental, fan-driven systems cannot guar-
antee that every cow will receive enough airflow. Cows not in
Acronyms
the system's target areas will often receive an insufficient
PASS Precision air supply system
cooling benefit.
AOZ Animal occupant zone
Since the advent of Zhang, Morsing and Strøm's early de-
RSM Response surface methodology
signs, a type of dairy house known as a “Saudi barn” has
CFD computational fluid dynamics
become a popular ventilation solution when combined with
RANS Reynolds-averaged Navier-Stokes
roof-mounted evaporative coolers that blow cooled air
SST shear stress transport
downward into the bedding area. Most recently, Mondaca and
THI Temperature humidity index
Choi (2016a) proposed a system based on the concept of
Greek letters channelling fresh air to the animal-occupied zone (AOZ)
Р density (kgm
3) through large-diameter polycarbonate plastic tubes and then
F variables jetting it downward, through smaller-diameter holes, directly
ds thickness of skin layer (mm) onto AOZ. The computational and experimental results ob-
dh depth of hair coat layer (mm) tained from tests of this “polytube ventilation system,”
b coefficient showed that the system could achieve a satisfactory cooling
n kinematic viscosity (m2s
1) effect. Both of these approaches are similar to the ceiling jet
q air-supply angle ( ) ventilation system used in pig houses (Zong, Li, & Zhang,
2015), which relies on diffuse ceiling inlets most of time to
target animals during warm summer periods, and all of the
Cobanov, & Schnitkey, 2003). Increasing airflow in the animal described systems can to some degree overcome the decline
occupied zone (AOZ) by means of a properly designed and in cooling performance that occurs when the indoor air tem-
controlled cooling system can usually alleviate the threat of perature increases; however, none can ensure that every an-
heat stress, improve thermal comfort in AOZ, and ensure a imal will receive a sufficient amount of fresh air that it will be
healthy and productive herd. Natural ventilation systems are properly cooled. For this reason, a cooling system that could
used widely because they can produce an adequate ventila- provide either fresh or pre-cooled air even more precisely to
tion rate cheaply, but they cannot be precisely controlled and each cow could significantly benefit the industry.
are highly dependent on exterior wind conditions. In contrast, Developing such a cooling system for specific use in dairy
mechanical ventilation systems are relatively easy to control barns should first entail a thoughtful consideration of bovine
and far less dependent on exterior conditions. A properly behaviour and in particular the behaviour of dairy cows,
designed and operated mechanical ventilation system can because the modern dairy cow spends so much time indoors
usually provide a desired airflow pattern and adequate in her stall (often over 13 h a day) (Cook, Mentink, Bennett, &
ventilation rate even during hot and humid conditions. Burgi, 2007). What is more, so long as a cow remains in her
However, since these systems generally require more energy thermal comfort zone, she will spend as many as 12e13 h a
and tend to be noisy, the optimal solution would seem to be a day resting in a reclining posture (Cook, Bennett, & Nordlund,
system that combines the advantages of natural and me- 2004; Jensen, Pedersen, & Munksgaard, 2005), whereas, if she
chanical ventilation (Hoff et al., 2009). suffers heat stress, she will stand for longer periods to expose
Many hybrid ventilation systems have been developed and more skin area and thereby enhance heat dissipation
tested over the last several decades. Two early and widely (Hillman, Lee, & Willard, 2005). However, long standing times
used designs, developed by Strøm and Zhang (1989) and have been associated with higher incidences of such health
Zhang, Morsing, and Strøm (1992), rely on natural ventilation issues as lameness (Whay & Shearer, 2017), while, conversely,
most of the time, with the mechanical component switching increasing lying time may well benefit a cow's health and
on only when the difference between the indoor and outdoor welfare. Given such findings, it seems reasonable to conclude
temperatures exceeds the system's set point (thus avoiding an that dairy producers could improve the health of their herds if
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
they could induce their animals to spend more time reclining
when the temperature-humidity index (THI) increases beyond
the heat stress threshold value; i.e., THI > 68 (Zimbelman
et al., 2009). Hence, any cooling-system design that induces
cows to recline longer should place special emphasis on
improving the heat-dissipation capacity of a reclining cow,
and since that capacity is largely dependent on the amount of
cooling air that can be passed over the animal, we propose a
design, named the “precision air supply system” (PASS), which
is based on the concept of providing precise amounts of either
fresh air, or possibly cooled air, to each cow in her stall. Since
we could find no applicable assessment of any system com-
parable to the one we propose, we assessed the cooling per-
formance of PASS with respect to both a reclining and
standing cow and in regards to the effects that the design and
a set of prescribed operation parameters (including the air-
supply opening size, the supplied air velocity, and the air-
supply angle) would have on the convective heat-dispersion
rate.
In short, this study was designed to achieve the following
objectives: 1) to develop mathematical models that could
predict the cooling performance of PASS in a naturally venti-
lated cattle barn; 2) to evaluate the effects that the design and
operation parameters would have on the system's perfor-
mance; and 3) to obtain an understanding of the concept's
applicability, one that would facilitate the system's design and
control. To simplify the complexity associated with using an
entire dairy barn housing hundreds of cows in identical stalls,
our research was conducted using one representative stall
unit and a virtual cow in both a standing and a reclining
posture.
2. Methodology and materials
2.1. Description of precision air supply system in cattle
barn
A properly designed PASS should be able to provide a better
local thermal environment for any cows during hot, humid
and windless periods when the sole natural ventilation sys-
tem is insufficient to mitigate heat stress. The system (PASS)
consists of a main duct with several subsidiary distribution
tubes connected as shown in Fig. 1a and b. Using this ducting
network, PASS can entrain either fresh outdoor air (and
potentially cooler and drier air), or else pre-cooled breeze in
conjunction with other pre-cooling systems (e.g. geothermal
Fig. 1 e The concept of precision air supply system (PASS) to cool cows in head-to-head stalls (a), and PASS with cows in
both reclining and standing postures (b).
3
heat exchanger (Krommweh, Ro € smann, & Büscher, 2014),
direct evaporative cooling (evaporative cooling pad (Rong,
Pedersen, Jensen, Morsing, & Zhang, 2017)), etc.). In order to
guarantee that a desired amount of a high-speed air stream
will be supplied to each cow in her stall, each subsidiary dis-
tribution tube was deployed at a certain angle in relation to a
particular cow. It should be noted that, in order to avoid dis-
rupting the usual motions of either a standing or reclining
cow, the height of the air supply opening was fixed at a dis-
tance of 2 m away from the bedding floor. At this distance,
even a standing cow cannot damage the supply tube because
it is restricted by the stall's neck rail.
In addition to these measures, PASS can be integrated with
a control system that includes wireless motion sensors and
real-time readings of ambient thermal environment,
including the ambient air velocity, air temperature, and hu-
midity, although the implementation of such a system was
beyond the scope of the present study. Thus, it is possible to
operate the proposed system in an energy- and performance-
efficient way.
2.2. Methodology in general
To predict the cooling performance of PASS and reveal the
effects that it can achieve, two mathematical models were
built, each based on key parameters, one to represent a
standing cow and the other a reclining cow. Because mathe-
matical modelling normally requires an appropriate valida-
tion procedure by a series of experiments that is difficult to
conduct (and therefore time consuming and costly), a multi-
variate statistical technique called Response Surface Meth-
odology (RSM) was used as a way to reduce the number of
required experiments while maintaining an acceptable degree
of accuracy (Shen, Zhang, & Bjerg, 2013). Even so, the RSM
approach still requires a number of tests. Since both field and
laboratory experiments involve a variety of limitations and
difficulties (such as maintaining stable micro-environmental
operating conditions and the confinement of live animals), a
less costly way to produce the data needed for mathematical
modelling should be employed. Consequently, a computa-
tional fluid dynamics (CFD) approach that utilises numerical
simulations to solve heat and mass transfer problems was
employed in this investigation. CFD has been used widely in
agricultural and biological engineering (Bjerg et al., 2013;
Mondaca & Choi, 2016b; Mondaca, Rojano, Choi, &
Gebremedhin, 2013; Rong, Liu, Pedersen, & Zhang, 2015;
Wang, Wang, & Wu, 2014; Zong & Zhang, 2014).
4 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
2.3. Laboratory experiment
2.3.1. Experiment setup
To generate data for CFD model validation, a laboratory
experiment was conducted in a chamber located in Building
L38, Air Physics Lab, Aarhus University, Foulum, Denmark.
The dimension of the chamber was 4.00 m  1.39 m  1.55 m
in length  width  height (Fig. 2). It was composed of two
sidewalls and a floor. To represent a cow, a hollow cylinder
0.5 m in diameter and 1.055 m in length was suspended above
the floor by two holders at an axis height of 0.375 m. To
maintain the constant heat generation rate of the cylinder,
four heat lamps, each with a power capacity of 100 W, were
installed equidistant inside the cylinder. During the experi-
ment, the four lamps were turned on, and in total they pro-
duced a heat generation rate of 400 W inside the cylinder. Note
that, this amount of heat is clearly less than the amount
produced by a 625 kg lactating cow (CIGR, 2002); however, by
controlling the surface temperature of cylinder similar to skin
temperature of a cow and considering the convective heat
removing process, this available laboratory facility was cho-
sen to present the force heat convection from a cow.
A fan with a full-run capacity of 450 m3 h
1 was used to
blow air to the cylinder via a tube (0.25 m in diameter). A flow
meter (FMU/FMDRU 100-80, Lindab, Denmark) based on the
differential pressure tapping principle was connected to the
fan. After using a pressure probe to measure the drop in
pressure (differential pressure) as the air passed through the
orifice (Model 9596, TSI, USA), the airflow rate could be
calculated using Eq. (1):
pffiffiffiffiffiffi
Vfan ¼ 7:32 DP
where Vfan denotes the airflow rate provided by the fan, l$s ;
the DP denotes the differential pressure between both sides of
an orifice, Pa (Lindab A/S, 2017). In the experiment, the air flow
rate was set at 322.7 m3 h
1.
Fig. 2 e Experimental setup and measurement locations. Black dots represent the air temperature and air velocity
measurement points; gray dots represent the surface temperature measurement locations on cylinder, walls and floor;
triangle is the measurement location of supplying air temperature; 5-point star is the location of orifice and the place to
measure the pressure difference.
2.3.2. Measurement procedure
Air temperature measurements were conducted at 7 locations
on the leeward side of the cylinder using 7 thermocouples
(Type J) simultaneously. All of the seven thermocouples were
mounted on a stand. The seven thermocouples were arrayed
at heights ranging from 20 cm to 80 cm with an interval of
10 cm between each as shown in Fig. 2. The measuring loca-
tions were fixed at horizontal distances of 42 cm from the right
side wall and 25 cm from the end of the cylinder. Air speeds
were measured at the same locations with the air temperature
measurements, and recorded simultaneously by 7 omnidi-
rectional air velocity sensors (Model 8475, TSI, USA).
The air-temperature and air-speed measurements were
taken separately, and each measurement period lasted
30 min. During the air-speed measurement period, the
surface temperatures of the cylinder were measured at 6
locations using 6 thermocouples attached to the surface as
shown in Fig. 2. In addition, during the entire experi-
mental period, the surface temperatures of the sidewalls
and floor of the chamber, as well as the air temperatures
at the jet opening, were measured using thermocouples.
In this laboratory experiment, all the temperature data
were documented by a data logger (1000 Series, Eltek, En-
gland) every 30s and all the air-speed data were docu-
mented by a data logger (CR1000, Campbell Scientific, USA)
every 10s.
2.4. Numerical simulation
2.4.1. Governing equations
The steady Reynolds-averaged Navier-Stokes (RANS)
(1) approach was applied because it has proved effective and

1 accurate in simulations related to heat release and air move-
ment (Rong, Nielsen, Bjerg, & Zhang, 2016). The general gov-
erning equation for a steady case can be expressed as Eq. (2)
(Patankar, 1980):
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
   
vðrFÞ v rFuj v vF
þ ¼ GF þ SF
vt vxj vxj vxj
where r is density, kg$m
3; uj is the velocity component in j
direction, m$s
1; F represents the common variables of in-
terest, i.e., three velocity components, m$s
1, temperature, K,
species (such as moisture), turbulent kinetic energy, m2$s
2,
and its dissipation rate, m2$s
3; GF is the transport coefficient
dependent on F; and SF is the source term dependent on F.
2.4.2. Cow geometric models
In this study, the geometric models for a cow in both a
standing and a reclining posture were drawn in Rhinoceros 5
(Robert McNeel & Associates, Seattle, WA, USA) with reference
to actual cows, as shown in Fig. 3. Both the reclining and
standing cow models were legless because including legs in
the CFD geometric models would have introduced numerous
sharp angles, the presence of which would have increased the
grid number and worsened grid quality. The legs and hooves
were excluded also because the total surface area of a cow's
legs and hooves only accounts for 16% of the animal's total
skin area (Minagawa, 1994); thus, it can be said that the ma-
jority of body heat dissipates from a cow's head, neck and
torso. Also, the body length of the reclining cow model was set
at 10% shorter than that of the standing cow model because an
actual reclining cow normally curls up her body (as reported
by McFarland (2003)). The ratio of 10% was determined in
accordance with a photographic illustration showing the ris-
ing motion of a reclining cow (Irish & Merrill, 1986).
Creating a useful geometric model of a cow depends on a
careful measuring of certain parts of a typical cow's body, such
as withers height, body length, and total skin surface. These
parameters can and should be rendered in the same scale by
referring to reliable illustrations depicting a real cow. The in-
formation pertaining to withers height and body length for
both postures is shown in Fig. 3. Once the body weight of a cow
is known, the surface area can then be determined by
applying Eq. (3), which is used for estimating the surface area
of a typical Holstein (Berman, 2003).
SA ¼ 0:14W0:57
2
where SA is the skin area of a cow, m ; and W is the body
weight of a cow, kg.
In this study, a 625 kg Holstein cow was modelled. Thus,
the corresponding skin area was 5.49 m2 (calculated from Eq.
(3)). With respect to the legless model, the surface areas of skin
that are in contact with air were set as 4.61 m2 (84% of the total
Fig. 3 e Simplification of geometrical models of a relining cow (a) and standing cow (b).
5
(2)
Fig. 4 e The structure of skin and hair coat of the cow used
in this study.
skin area) and 3.46 m2 (acquired from meshing software) for a
standing cow and a reclining cow, respectively.
2.4.3. Sub-model of skin and hair coat
Because the skin and hair coat of a cow greatly affects the rate
of heat exchange occurring between the cow and the ambient
environment (Bertipaglia, da Silva, Cardoso, & Fries, 2007), it is
necessary to properly model them as a sub-model in the CFD
model. Skin is generally composed of three layers: epidermis,
dermis, and hypodermis, as shown in Fig. 4 (Xu, Lu, Seffen, &
Ng, 2009), and beneath the hypodermis, there is body tissue. In
this study, the temperature of the body tissue (Tt) was regar-
ded as equal to the body core temperature, which was set at
38.7  C (Turner, Chastain, Hemken, Gates, & Crist, 1992).
The thickness of the skin varies between breeds; for
instance, the average thickness of skin is 8.15 mm in a Devon
cow, 6.7 mm in a Hereford, 6.0 mm in a Holstein Friesian, 5.77
in a Zebu, and 5.46 in a Jersey (Dowling, 1955). In this study,
the thickness of the skin (ds) was fixed at 6 mm. Also, since
cows are commonly shaved in summer to improve heat
dissipation, the depth of the hair coat (dh) was set as 3 mm.
Both the skin and the hair coat were treated as solid
conductive shells in the CFD model. Such treatment of a hair
(3) coat as a conductive shell with a certain thermal conductivity
has proved to be an acceptable approach for investigating
convective heat transfer occurring on the surface of a cow
(Mondaca et al., 2016b). Due to a dearth of literature on the
thermal properties of a cow's skin, the average thermal con-
ductivity of the skin of a pig (with the value of 0.41 W m
1 K
1)
was used. The effective thermal conductivity of the hair coat
6 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
was set at 62.98 mW m
1 K
1 (Maia, da Silva, Souza Junior, da
Silva, & Domingos, 2009).
2.4.4. Computational domain, mesh and boundary conditions
In the achieving of a numerical simulation, the size of the
computational domain plays an important role. Many simu-
lation cases show that an inappropriate computational
domain can significantly impair the accuracy of simulation
results (Bjerg et al., 2013; Lee, Sase, & Sung, 2007). Blockage
effects produced by the lateral boundaries can be disregarded
when the blockage ratio (defined as the ratio of the projected
area of the cow in airflow direction to the inlet area) is less
than 3% (Franke, Hellsten, Schlunzen, & Carissimo, 2011).
Thus, in order to eliminate the blockage effects produced by
the boundaries established in this study, the dimensions of
the computational domain were set at 6 m  6 m  18 m, as
shown in Fig. 5.
Due to the complexity of a cow's geometry, the computa-
tional domain was divided into two parts, “domain 1” and
“domain 2”, as shown in Fig. 5. Structured and unstructured
meshes were adopted in both parts, respectively, and to make
the data flow, the interface method was used to couple both.
The treatment for each boundary is given in Table 1.
2.4.5. Simulation scheme
The SIMPLE (Semi-Implicit Method for Pressure-Linked Equa-
tions) scheme was used to discretise the governing equations.
Second order upwind discretisation was selected for mo-
mentum, turbulent kinetic energy, specific dissipation rate
and energy in order to improve the accuracy of the final so-
lution. The iteration was considered converged only when the
monitored values (the volume-averaged velocity of sub-
domain and surface-averaged heat transfer coefficient
occurring on the cow-body surface) were stable. In addition,
Fig. 5 e Computational domain and setup of boundary conditions: (a) the computational domain for a reclining cow and the
boundary condition setup; and (b) the computational domain for a standing cow and its dimensional information.
the net mass flow rate of all inlets and outlets in the whole
computational domain was less than 10
4 kg s
1 (Shen, Zhang,
& Bjerg, 2012).
The Shear-Stress Transport (SST) keu turbulence model
(Menter, 1992) was chosen because it has proven to be a reli-
able computational turbulence model and offers the advan-
tages of efficiency and accuracy (Rong et al., 2016). For the SST
k-u turbulence model, the accuracy of the prediction depends
to a great extent on the yþ value, which should be smaller than
1. Therefore, in this study, to maintain an average yþ value
smaller than 1 on the cow's surface, the height of the first
boundary layer was set at 0.2 mm.
2.4.6. Mesh convergence test
A good mesh accelerates iterating speed, reduces convergence
time, and provides acceptable predictions. The optimum mesh
distribution, regarded as an adequate compromise between
computational accuracy and efficiency (Norton, Grant, Fallon,
& Sun, 2009), was determined by means of a mesh convergence
test. Four different mesh profiles in different mesh densities
were simulated using the same boundary conditions, and the
results were subsequently compared. Mesh convergence was
evaluated using the volume-averaged velocities of domain 2
(shown in Fig. 5) in different mesh cases. Table 2 shows the
results of comparing the different mesh levels. The volume-
averaged air velocity in case 1 was regarded as the accurate
one. The relative difference in average velocity, shown in Table
2, indicated that the meshing criterion in case 3 could be
applied in the following meshing work.
2.5. Response surface methodology modelling
To investigate the effects of the design and operation pa-
rameters on the cooling performance of PASS, response
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5 7

Note that, normally, a correctly adjusted neck rail will

Table 1 e Boundary conditions setup in numerical
restrict a standing cow to step as deeply as a reclining cow
simulations.
does. Hence, a horizontal distance of 10 cm between the
Item Treatment
standing cow and the jet opening was settled on in this study,
Inlet Velocity inlet, air temperature was 28  C; as shown in both Figs. 1b and 6b. For the reclining cow, the
air velocity was 0.2 m s
1 or 0.5 m s
1 or horizontal distance was fixed as zero.
1.0 m s
1 to simulate three ambient
Two three-dimensional experimental domains were built
airflow conditions (the magnitude of
this air velocity was deemed as Bulk Air
up for both the reclining cow and the standing cow
Speed) (maximum value and minimum values of each variable are
Side walls Symmetry listed in Table 3).
Ceiling and floor No slip wall, no heat flux To assess the cooling performance of PASS in relation to
Precision air supply Velocity inlet, supply air temperature heat-stress mitigation, the convective heat flux dissipating
was 22  C, supply air velocity varied in
from a cow's surface was set as the response in both RSM
different case
models (because this variable will indicate the dispersion of
Outlet Pressure outlet, reverse flow
temperature was 28  C heat from the cow into the ambient air).

2.5.2. Experimental design for response surface methodology

Table 2 e Cells number and mesh convergence.
Items Case 1 Case 2 Case 3
Total number of grid 9,747,660 3,876,229 2,970,174
Ratio of grid 1.000 0.398 0.305
Average velocity of 0.214 0.214 0.214
domain 2
Relative difference 1.000 0.999 0.999
of average velocity
surface methodology (RSM), which is a combination of
mathematical and statistical techniques, was selected (since
RSM could explore the relationship between explanatory
variables and targeted responses and thereby obtain the
optimal responses) (Whitcomb & Anderson, 2004). RSM is
widely applied in not only the model development phase but
also during model improvement and model optimisation
(Myers, Montgomery, & Anderson-Cook, 2016).
2.5.1. Experimental domain and the response in response
surface methodology model
The experimental domain, one of the most important com-
ponents in RSM modelling, is defined by the minimum and
maximum limits of the experimental variables. In this study,
the three design and operation parameters of PASS, i.e.,
supply-air velocity (v), distribution-tube diameter (d) and air-
supply angle (q), were selected as the experimental vari-
ables. All of the three investigated parameters were evenly
divided into three levels (low level, medium level, and high
level).
The supply-air velocity was selected according to the rec-
ommended air-flow speed around a cow (2.5 m s
1 e 3 m s
1)
housed in a dairy barn equipped with a tunnel ventilation
system (Tyson, 2010). Due to jet decay, we set the supply-air
velocity at three different levels (2 m s
1, 3 m s
1, 4 m s
1).
The distribution-tube diameter was set at 0.1 m, 0.15 m and
0.2 m. The levels of air-supply angle were determined ac-
cording to the length of the cow body (from rump to neck) as
shown in Fig. 6. The air-supply angle ranged from 40 to 70 for
the reclining cow case and from 15 to 40 for the standing cow
case.
model
While many different experimental designs have been used
Case 4 to conduct RSM modelling (such as the central composite
1,925,220 design, the full factorial design, and the Box-Behnken design)
0.198 (Shen et al., 2013), in this study, the Box-Behnken design
0.213 (defined as a rotatable or nearly rotatable second order
design) (Ferreira et al., 2007), was adopted because it requires
0.997
fewer tests compared to other experimental designs (Shen
et al., 2012). The Design Experts (Version 8.0, Stat-Ease, Inc,
Minneapolis, USA) was used to perform the Box-Behnken
experimental designs for both postures and subsequent
data analysis.
The number of tests (N) required (in the Box-Behnken
design) to build the RSM model can be calculated according
to Eq. (4):
N ¼ 2kðk
1Þ þ c (4)
where k is number of investigated parameters; and c is the
specified number of replicated central points.
Normally, the testing of the central point requires several
replications in order to minimise model error. Yet, in the
simulation procedure, replication could be omitted because in
an accurate CFD simulation, results obtained using the same
boundary conditions should be the same. Therefore, thirteen
simulation runs were needed to model each case (one for each
posture).
2.5.3. Establishment of response surface methodology models
A polynomial based on the Taylor series expansion is a very
common model expression when using the RSM method.
Hence, the quadratic polynomial model was used to establish
the RSM models associated with the standing and reclining
cow:
X
3 X
3 X
3 X
3
y ¼ b0 þ bi xi þ bii x2i þ bij xi xj (5)
i¼1 i¼1 i¼1 j>i
where y is the predicted response; b0 is the model constant; bi
are linear coefficients; bii are the quadratic coefficients; bij are
the cross product coefficients; and xi and xj are independent
variables.
To ensure that the response data was approximately
normal in its distribution, Box-Cox transform analysis was to
8 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5

Fig. 6 e Determination of ranges of the jet angle of PASS according to a reclining cow (a) and a standing cow (b).

Table 3 e Levels of investigated parameters, supply-air velocity (v), distribution-tube diameter (d) and air-supply angle (q).
Simulation case Parameter Unit Low level (
1) Medium level (0) High level (þ1)
Reclining cow d m 0.1 0.15 0.2
v m$s
1 2 3 4

q 40 55 70
Standing cow d M 0.1 0.15 0.2
v m$s
1 2 3 4

q 15 27.5 40

test the data (and transform any that were not distributed
normally in Design Experts). The fitting quality of the RSM
models was evaluated by calculating the F-value and p-value
using ANOVA. The F-value is defined as the ratio of residual
variances in a model both with and without the predictor. A
larger F-value and a smaller p-value indicate the higher sig-
nificance of the corresponding term. Terms that corresponded
with p values lower than 0.10 were kept in the RSM models
while the rest (the terms corresponding with a p value greater
than 0.10) were removed. Hence, the final RSM model con-
sisted only of the terms that corresponded with a p value of
less than 0.10. To enhance the quality of model fit, a trans-
formation of the response is commonly applied, especially
when the data are not normally distributed. In addition, the
fitting performance of the RSM model can be evaluated by
applying adjusted R2 and predicted R2. Adjusted R2 is a
modified version of R2, one that has been adjusted to account
for the total number of design points. Predicted R2 is a mea-
sure of how well the model predicts a response. In general, the
higher the adjusted R2 and predicted R2, the better the model
fits the data and the more accurately the model predicts
(Myers et al., 2016).
back flow vortex, along with the jet passing over the cylinder
end, was generated beyond the sharp edge at the end of the
cylinder (such a vortex flow can be unsteady). Whatever the
case, it presented a challenge that the measured air speed at
that point could not be represented as an averaged value.
Such phenomena cannot be reflected in CFD simulations,
because the airflow was treated as steady state. Regardless of
the differences in the air velocity at the height of 0.5 m, the
trend in air velocity was consistent. Table 4 shows a com-
parison of the temperature profiles obtained by experiment
versus simulation with the largest relative error being
smaller than 0.76%. Due to the agreement between the
experimental data and the simulated data in terms of both
air velocity and air temperature (except for the air velocity
occurring at a height of 0.5 m), the numerical model was
deemed valid, and, hence, the simulated results could be
used in the following simulations.

3. Results and discussion

3.1. Computational fluid dynamics model validation

The comparison of the air velocity profiles acquired from

laboratorial experiments and CFD simulations is presented in
Fig. 7. As can be observed, most of the predicted air velocities
were proximate to the corresponding measured air velocities
(except for the air velocity at a height of 0.5 m above the
floor, where the simulated velocity was larger). Possibly, a
Fig. 7 e Comparison of air velocity between measured
experiment and simulation at different measurement
heights.
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5 9

Table 4 e Comparison of air temperature between experiment and simulation at different measurement heights.
Measurement height 0.2 m 0.3 m 0.4 m 0.5 m 0.6 m 0.7 m 0.8 m
Measured temperature,  C 23.49 23.40 23.48 23.57 23.76 23.95 23.67
Simulated temperature,  C 23.40 23.40 23.40 23.49 23.66 23.80 23.85
Relative difference, % 0.38 0.00 0.34 0.34 0.42 0.63 0.76

Fig. 8 e Streamline plot and skin surface temperature of relining cow (a) and standing cow (b).

3.2. Assessment of precision air supply system cooling stress mitigation effect. Besides, as Fig. 8 shows, the skin
performance surface temperature on the back of the reclining cow was
relatively higher than that on the back of the standing cow,
To assess the ability of PASS to cool a cow in either posture, implying that more heat was removed from the back of the
two operation scenarios were conceived; in the one involving standing cow due to the distance of the air outlet to the target
a reclining cow, the supply-air velocity, the diameter of the skin surface.
distribution tube, and the air-supply angle were set at 3 m s
1, The heat fluxes dissipating from both the reclining cow
0.15 m, and 55 , respectively, and in the scenario involving a and the standing cow, cooled with and without PASS, are lis-
standing cow, the same variables were set at 3 m s
1, 0.15 m, ted in Table 5, which shows that, with the application of PASS
and 27.5 , respectively. Figure 8 shows the streamlines of the under the given setup, heat-flux increases of 25.60 W m
2 and
air-supply jets and the surface-temperature contour of the 18.64 W m
2 were observed emanating from the surface of the
cow in both postures. reclining cow and the standing cow, respectively. The in-
According to the streamline plots, the velocity of the crease in heat lost from the reclining cow and the standing
impinging jet decreased gradually, from 3 m s
1, as the jet's cow were 89 W and the 86 W, respectively. Such sizable im-
travelling distance increased. The airspeed of about 1.5 m s
1 provements in the amount of heat flux and heat loss indicated
recorded close to the reclining cow was less than the one that the application of PASS under the given setup condition
recorded close to the standing cow (about 2 m s
1), probably could be a good option for alleviating heat stress.
because the distance from the distribution-tube opening to As is well known, when a cow feels heat stress, she will
the reclining cow was shorter. shift from lying down to standing up (Atkins, Cook, Mondaca,
Compared with the surface temperature profiles of both & Choi, 2018, in review) because doing so increases the total
the reclining cow and the standing cow when neither was surface area exposed to air and allow more heat to dissipate
subjected to PASS, the skin-surface temperature in the cases more quickly (Wang, Zhang, & Choi, 2018). However, when
involving PASS obviously decreased along the back area PASS was in operation, the heat lost from the reclining cow
covered by the impinging jet. Such a decrease in surface was even greater than that lost from the standing cow not
temperature indicated that PASS produced a significant heat- subjected to PASS (according to Table 5), which means that,
with the help of PASS, there is no need for a reclining cow to
stand up to increase heat dissipation and maintain thermal
Table 5 e Heat flux and total rate of heat dissipation from comfort. In another words, the time spent reclining could
cow skin surface (inlet air velocity was 0.2 ms¡1). potentially be prolonged when the PASS is used, and the cow's
Area of skin with/without Heat flux Heat health would thus benefit.
Cow exposing to air PASS W$m
2 loss
posture m2 W 3.3. Effect of different bulk air speed
Reclining 3.46 Without 26.00 90
With 51.60 179 In order to check the feasibility of PASS under different indoor
Standing 4.61 Without 25.30 117 air speeds, we conducted four more simulations involving
With 43.94 203 bulk air speeds of 0.5 m s
1 and 1.0 m s
1 for both postures, as
10 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
Fig. 9 e Variations of heat loss from a cow with and
without PASS in both reclining and standing postures
under three different bulk air speeds (0.2 m s¡1, 0.5 m s¡1,
and 1.0 m s¡1). The PASS supply air velocity is kept at
3 m s¡1, the tube diameter is 0.15 m, and the air supply
angle is 55 for reclining cow and 27.5 for standing cow.
stated in Table 1. All of the system's design and operation
parameters were kept the same as in the cases that involved a
bulk air speed of 0.2 m s
1. Thus, the effect of bulk air speed on
the system's cooling performance could be assessed. The
variations in the amount of heat lost from the skin surface of a
cow in either posture, reclining or standing, are displayed in
Fig. 9.
As can be seen in Fig. 9, for the cases that did not involve
PASS, the heat lost from a cow increased as the bulk air speed
increased from 0.2 m s
1 to 0.5 m s
1, and then to 1.0 m s
1,
indicating that increasing air speed could increase the heat
flux and amount of heat released from a cow, because a higher
air speed will result a larger convective heat transfer coeffi-
cient on the surface of the cow (Wang et al., 2018).
In the reclining-cow case, the convective heat loss ach-
ieved using PASS under the given setup decreased as the bulk
air speed increased from 0.2 m s
1 to 0.5 m s
1, and then
increased again, with the bulk air speed continuously
increasing to 1.0 m s
1. The overlap of curves indicated that
the cooling effect achieved using PASS was close to zero. Ac-
cording to Fig. 9, the increased heat loss experienced by the
reclining cow dropped rapidly from 89 W to around zero W as
the bulk air speed increased from 0.2 m s
1 to 0.5 m s
1, and
then remained stable, with the bulk air speed continuously
increasing. Such variation in the increased rate of heat loss
associated with the reclining cow indicated that the best
cooling performance to be achieved using PASS under given
setup occurred when the bulk air speed was 0.2 m s
1, while
its cooling performance dropped rapidly when bulk air speed
increased. When the bulk air speed rose above 0.5 m s
1, the
cooling performance associated with PASS under the given
setup became very limited (and could be neglected). The
reason was that the impinging jet was so greatly affected by
the bulk air movement that it failed to reach the surface of the
reclining cow. Besides, it can be assumed that the heat loss
experienced by the reclining cow at 0.2 m s
1 was even larger
than at 1 m s
1, indicating that the application of PASS when
the bulk air is still could cool reclining cows with considerable
efficiency.
In the standing-cow case, the convective heat loss ach-
ieved using PASS under the given setup increased gradually as
the bulk air speed increased from 0.2 m s
1 to 1.0 m s
1.
However, the increase in heat loss experienced by the stand-
ing cow under the given setup and using PASS decreased
gradually from 86 W to 57 W, and to 10 W as the bulk air speed
increased from 0.2 m s
1 to 0.5 m s
1 and then to 1.0 m s
1, all
of which indicated that the cooling performance of PASS
declined as the bulk air speed increased, with the best per-
formance associated with a bulk air speed of 0.2 m s
1. This
also implied that PASS, when operated under a given setup,
could still cool the cow if the bulk air speed was maintained at
around 1.0 m s
1, although the increase in heat loss would be
very limited. When the bulk air speed rises to over 1.0 m s
1,
the cooling performance of PASS will be negligent. The rate of
heat loss declines because, as the speed of the bulk air in-
creases, less and less of the standing cow's skin surface is
struck directly by the jet of air.
Although our findings were obtained based on PASS used
according to the given setups and restricted to limited bulk air
speeds, reasonable conclusions can still be drawn: 1) bulk air
speed significantly affects the cooling performance of PASS; 2)
during the hot and windless conditions, PASS can serve as an
effective cooling method; 3) however, if the indoor air is
moving quickly, the air jet supplied by PASS will be swept
away due its limited flow rate and momentum, and the sys-
tem will become less effective than when indoor-air move-
ment is slow. Thus, PASS should be operated in accordance
with the prevailing indoor climate conditions (using the real-
time readings of indoor air speed, temperature and relative
humidity) and cow's presence and posture in each stall (using
a motion detector to identify cow's standing or reclining
posture). That is, PASS can be integrated with rapidly
advancing latest technologies of wireless, ultra-low-power,
low-cost, and real-time Internet of Things (IoT) devices to
measure microclimate conditions and cow's location and
posture.
3.4. Response surface methodology model
3.4.1. Model fitting
To investigate the effects of the design and operation pa-
rameters, an RSM model was used. The Box-Behnken designs
of both postures and the corresponding simulated results
pertaining to the convective heat transfer of the cow are listed
in Table 6. As can be seen, the convective heat flux ranged
from 20.75 W m
2 to 78.41 W m
2 for the reclining cow and
from 27.93 W m
2 to 81.93 W m
2 for the standing cow. These
results indicated that all the three investigated factors could
affect the cooling performance of PASS.
Based on the Box-Cox transformation analysis in Design-
Expert (Stat-Ease, Inc., Minneapolis, MN, USA), both a
reverse square-root transformation of the response data
associated with reclining cow case (from HFr to HF
0.5
r ) and a
base-10-log transformation of the response data associated
with standing cow case (from HFs to log10HFs) were con-
ducted. The ANOVA results for both postures were listed in
Table 7. As can be seen, the p-values of the three terms (v$q,
d$q, and d2) used in the case involving the reclining cow and
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5 11

reduction saves more than half the experimental time

Table 6 e The Box-Behnken experimental design
required by a full factorial design; however, model uncertainty
arrangement and responses for both standing and
reclining cow case. q, v, and d represent air-supply angle, might be introduced (Shen et al., 2012). Both established RSM
supply-air velocity, and distribution-tube diameter. models were therefore verified using two random cases for
each. All of the parameter values were selected randomly
Reclining case Standing case
 
within the experimental space (Table 9). Ideally, the rela-
q, v, d,m Heat flux, q, v, d, m Heat flux,
tionship between the results predicted by the RSM model and
m$s
1 W$m
2 m$s
1 W$m
2
the CFD simulated results will be the same, and the less dif-
70 4 0.15 68.29 27.5 4 0.2 62.23 ference between the simulated and predicted data will indi-
55 2 0.1 27.06 40 3 0.2 81.93
cate a better predictive capacity. A comparison of the results
55 4 0.1 46.87 40 2 0.15 49.27
70 3 0.2 73.62 27.5 2 0.1 32.14
predicted by RSM model and the CFD simulations are listed in
40 2 0.15 26.61 15 3 0.1 27.93 Table 9. As can be seen, the relative errors were all less than
55 2 0.2 20.75 40 4 0.15 76.79 5%, indicating that both RSM models could be deemed
70 3 0.1 42.01 27.5 2 0.2 46.03 acceptable and could be applied in the following parametric
40 4 0.15 36.66 15 3 0.2 32.48 studies. These two models could also be used in future studies
55 3 0.15 51.60 40 3 0.1 43.74
if the PASS setups, cow sizes, and air movement are all similar
40 3 0.2 39.31 27.5 4 0.1 39.43
to those used in this case.
70 2 0.15 39.92 15 4 0.15 34.59
55 4 0.2 78.41 27.5 3 0.15 43.94
40 3 0.1 26.53 15 2 0.15 28.58 3.5. Sensitivity analysis

Sensitivity analysis, which reveals each factor's sensitivity to

the four terms (v$d, v2, d2 and q2) in the case involving the
the responses, can be achieved by applying any of several
standing cow were larger than 0.1, indicating that these
methods. In this study, perturbation plotting was used to
terms were not significant to the corresponding transformed
reveal how a certain factor responds as the level of the factor
responses, and thus these terms were deemed meaningless
changes when the other factors are fixed at their medium
and not included.
levels (Oh, Rheem, Sim, Kim, & Baek, 1995). In perturbation
The RSM model equations for both cases are listed in Table
plotting, a steep slope or curvature indicates the factor's
8. As can be observed, the values of the adjusted R2 and the
sensitivity to the response.
predicted R2 for both cases were very close to 1, implying that
In this study, two perturbation figures (Fig. 10a and b), one
the fit quality and predicted capacity of both RSM models
representing a reclining cow and the other standing cow, were
could be deemed acceptable. The predicted plot, as opposed to
plotted in order to evaluate the effect that each experimental
the actual plot for both postures, shows that the values pre-
factor might have on the heat dissipate rate. As can be seen, in
dicted by the RSM models were similar to the corresponding
both cases, the response (heat flux) increased as supply air
simulated values. Theoretically, the data points should be
velocity, tube diameter and jet angle increased. All three
split evenly by the 45-degree line, and the closer it lies to the
curves intersect at the reference point.
solid line, the better the model's capacity to predict. According
Figure 10a illustrates the perturbations that the investi-
to the two plots shown in Table 8, it can be safely concluded
gated parameters had on the amount of heat flux associated
that the transformed model can precisely predict the relation
with the reclining cow, based on the reference point at which
between the investigated variables and the response.
all three investigated factors were at their medium levels. As a
consequence of increasing the tube diameter from 0.1 m to
3.4.2. Response surface methodology model verification
0.2 m, the heat flux increased from 38.16 W m
2 to
With the help of the RSM and Box-Behnken designs, the
66.61 W m
2. A rise in heat flux of 22.53 W m
2 (from
number of experimental runs required could be greatly
36.91 W m
2 to 59.44 W m
2) was observed when the supply-
reduced (to 13 for each posture case). No doubt such a
air velocity was incrementally increased from 2 m s
1 to
4 m s
1. Adjusting the air-supply angle from 40 to 70 also
positively affected the cooling performance of PASS, with the
Table 7 e ANOVA of the transformed responses. q, v, and
d represent air-supply angle, supply-air velocity, and heat flux increasing from 37.86 W m
2 to 67.32 W m
2. As
distribution-tube diameter. shown in Fig. 10a, the sensitivities of all three investigated
parameters were similar when they were within a span that
Source Reclining cow case Standing cow case
ranged from low to medium level. With the value of the
F-value p-value F-value p-value
investigated parameters continuously increasing, the sensi-
Model 111.32 0.0013 207.66 0.0005 tivities of air-supply angle and tube diameter remained
v 258.25 0.0005 205.3 0.0007 similar, while the sensitivity of air-supply velocity decreased
d 328.37 0.0004 401.05 0.0003
gradually (because the curvature of the arc associated with
q 350.31 0.0003 1163.94 <0.0001
v$d 9.3 0.0554 2.99 0.1825
air-supply velocity decreased). This decrease indicates that
v$q 1.93 0.2586 20.2 0.0206 increasing supply-air velocity is not a good way to increase the
d$q 0.25 0.6507 71.81 0.0035 heat flux when it is already in the higher half of the range for a
v2 8.95 0.0581 0.59 0.4989 reclining cow.
d2 1.66 0.2877 1.26 0.344 Figure 10b illustrates the perturbations that the investi-
q2 49.52 0.0059 0.52 0.5246
gated parameters had on the amount of heat flux dissipating
 12
Table 8 e RSM model equations for both investigated cases. The adjusted and predicted R2 were also listed.
Case RSM model Adjusted R2 Predicted R2 Predicted vs. Actual plot

0:5
3
3 2
5 2
Reclining cow HFr ¼ 0:59186
0:06121v
0:67441d
7:79  10 q þ 0:09365v  d þ 4:95  10 v þ 5:85  10 q 0.9864 0.9620

b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
Standing cow log10 HFs ¼ 1:39085 þ 1:49  10
3 v
0:5779d
7:23  10
3 q þ 2:20  10
3 v  q þ 0:08286d  q2 0.9912 0.9799

where, HFs and HFr represent the heat flux on the surface of standing and reclining cow, respectively, W$m
2; v is the velocity of supply air, m$s
1; d is the diameter of the distribute tube, m; q is the air
supply angle,  .
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5 13

Table 9 e Verification of both RSM model equations with random values of parameters in corresponding RSM
computational domains.
Case air supply Air velocity, Diameter, Predicted HF, Simulated HF, Relative
angle,  M s
1 m W m
2 W m
2 error, %
Reclining cow 50 3.50 0.14 46.65 48.74 4.3
Standing cow 32 2.80 0.16 49.75 50.14 0.8

Fig. 10 e Perturbation plots of the heat flux with the variations of parameters associated with the reclining cow (a) and the
standing cow (b). For the reclining cow, the reference point is supply air velocity at 3 m·s¡1, tube diameter at 0.15 m, and air
supply angle at 55 ; and for the standing case, the reference point is supply air velocity at 3 m·s¡1, tube diameter at 0.15 m,
and air supply angle at 27.5 . Coded unit means the levels of experimental parameters where ¡1 represents the low level, 1
represents the high level, and 0 is the medium level.

from the standing cow's surface, based on a reference point at
which all three investigated factors were at their medium
levels. As can be seen, the heat flux increased steadily from
37.63 W m
2 to 50.05 W m
2 as the air velocity increased from
2 m s
1 to 4 m s
1, and from 35.56 W m
2 to 52.97 W m
2 as the
tube diameter was enlarged from 0.1 m to 0.2 m, and from
30.91 W m
2 to 60.94 W m
2 when the air-supply angle was
changed from 15 to 40 . The increases in heat flux associated
with all the three investigated parameters indicate that all
these three parameters have positive effects on the heat flux.
In addition, the curvature of the air-supply angle was steepest,
followed by the one related to tube diameter, with the one
related to supply-air velocity being the smallest, which in-
dicates that the air-supply angle is the most sensitive of the
investigated parameters.
Based on the sensitivity analysis, the jet's angle has the
most significant effect on the cooling performance of PASS,
regardless of posture. Note that, although all the three
investigated parameters positively affected the performance
of PASS, changing the jet angle did not increase of airflow rate.
Hence, adjusting the air-supply angle to the optimal position
should be the first step when seeking to enhance the cow's
heat dissipation rate. Once the optimal air supply angle has
been achieved, tube diameter becomes the best second choice,
leaving air-velocity increase as the final option.
4. Conclusions
This work assessed the cooling performance of PASS with
respect to both a cow standing inside a stall and a cow
reclining in the same stall. The authors developed two CFD
models to simulate the effects achievable using PASS to help
cool a cow in both a reclining and standing posture. The CFD
model was validated with the experimental data and was
further used to assess the system's performance and to
conduct a parametric analysis. Based on the results obtained,
the following conclusions could be drawn:
1) In the given investigation setups and the system config-
uration, PASS could provide a better local thermal envi-
ronment for cows and remove more heat from the
animals. The system's best cooling performance occurs
when the indoor air is still, and the more quickly the in-
door air moves the less effectively the system performs.
2) RSM, in conjunction with the Box-Behnken design, re-
duces the number of experimental runs required. Two
mathematical models associated with a cow in a
standing posture and a reclining posture were devel-
oped and properly verified.
3) All of the investigated parameters (supply air velocity,
tube diameter, and air supply angle) of the PASS
installation had positive effects on the rate of the
convective heat transfer associated with the simulated
cows. Among the three investigated parameters, the
best first option for achieving a better cooling perfor-
mance proved to be adjusting the air supply angle, with
increasing the tube diameter being the best second
option. Increasing the supply-air velocity should thus be
the last option.
In summary, unlike other hybrid ventilation systems, PASS
not only overcomes the decline in cooling performance that
occurs as the indoor air temperature increases, but also en-
sures that every cow in its stall can receive a sufficient amount
of fresh air. Therefore, PASS can be used to supplement the
14 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
effects produced by natural ventilation in a freestall barn
because it performs best during periods when natural venti-
lation fails to adequately cool the animals.
Acknowledgements
This work was funded mainly by the “Optimized animal-
specific barn climatization facing temperature rise and
increased climate variability” (Optibarn) research grant, pro-
vided by the EU-FACCE-JPI project consortium, DK-Grant No.
4215-00004B, Innovation Foundation Denmark. The China
Scholarship Council, China (Grant number: 201406320185) and
S.C. VAN FONDEN, Denmark (Grant number: 1626) supported
the lead author's study in Denmark.
references
Atkins, I. K., Cook, N. B., Mondaca, M. D., & Choi, C. Y. (2018).
Continuous respiration rate measurement of heat stressed
dairy cows and relations to environment, body temperature,
and lying behavior. Transactions of ASABE (in review).
Berman, A. (2003). Effects of body surface area estimates on
predicted energy requirements and heat stress. Journal of Dairy
Science, 86(11), 3605e3610.
Bertipaglia, E. C. A., da Silva, R. G., Cardoso, V., & Fries, L. A. (2007).
Hair coat characteristics and sweating rate of Braford cows in
Brazil. Livestock Science, 112(1e2), 99e108.
Bjerg, B., Cascone, G., Lee, I.-B., Bartzanas, T., Norton, T., Hong, S.-
W., et al. (2013). Modelling of ammonia emissions from
naturally ventilated livestock buildings. Part 3: CFD modelling.
Biosystems Engineering, 116(3), 259e275.
CIGR. (2002). Climatization of animal houses. Heat and moisture
production at animal and house levels. 4th Report of Working Group.
Horsens (Denmark): Research Centre Bygholm.
Cook, N. B., Bennett, T. B., & Nordlund, K. V. (2004). Effect of free
stall surface on daily activity patterns in dairy cows with
relevance to lameness prevalence. Journal of Dairy Science,
87(9), 2912e2922.
Cook, N. B., Mentink, R. L., Bennett, T. B., & Burgi, K. (2007). The
effect of heat stress and lameness on time budgets of lactating
dairy cows. Journal of Dairy Science, 90(4), 1674e1682.
Dowling, D. (1955). The thickness of cattle skin. Australian Journal
of Agricultural Research, 6(5), 776e785.
Ferreira, S. L. C., Bruns, R. E., Ferreira, H. S., Matos, G. D.,
David, J. M., Branda ~ o, G. C., et al. (2007). Box-Behnken design:
An alternative for the optimization of analytical methods.
Analytica Chimica Acta, 597(2), 179e186.
Franke, J., Hellsten, A., Schlunzen, K. H., & Carissimo, B. (2011).
The COST 732 best practice guideline for CFD simulation of
flows in the urban environment: A summary. International
Journal of Environment and Pollution, 44(1e4), 419e427.
Hillman, P., Lee, C., & Willard, S. (2005). Thermoregulatory
responses associated with lying and standing in heat-stressed
dairy cows. Transactions of the ASAE, 48(2), 795e801.
Hoff, S. J., Harmon, J. D., Chen, L., Janni, K. A., Schmidt, D. R.,
Nicolai, R. E., et al. (2009). Partial biofiltration of exhaust air
from a hybrid ventilated deep-pit swine finisher barn. Applied
Engineering in Agriculture, 25(2), 269.
Holmes, B., Cook, N. B., Funk, T., Graves, R., Kammel, D. W.,
Reinemann, D. J., et al. (2013). Dairy freestall housing and
equipment. Ames, Iowa: MidWest Plan Service: Iowa State
University.
Irish, W. W., & Merrill, W. G. (1986). Design parameters for
freestalls. In Paper presented at the the dairy freestall housing
symposium, Harrisburg, Pennsylvania.
Jensen, M. B., Pedersen, L. J., & Munksgaard, L. (2005). The effect of
reward duration on demand functions for rest in dairy heifers
and lying requirements as measured by demand functions.
Applied Animal Behaviour Science, 90(3e4), 207e217.
Kammel, D. W., Raabe, M. E., & Kappelman, J. J. (2003). Design of
high volume low speed fan supplemental cooling system in
dairy free stall barns. In Paper presented at the fifth international
dairy housing conference, Fort Worth, Texas. http://elibrary.asabe.
org/abstract.asp?aid¼11628&t¼1.
Krommweh, M. S., Ro € smann, P., & Büscher, W. (2014).
Investigation of heating and cooling potential of a modular
housing system for fattening pigs with integrated geothermal
heat exchanger. Biosystems Engineering, 121, 118e129.
Lee, I.-b., Sase, S., & Sung, S.-h. (2007). Evaluation of CFD accuracy
for the ventilation study of a naturally ventilated broiler
house. Japan Agricultural Research Quarterly, 41(1), 53e64.
Lindab A/S. (2017). Lindab dampers assembly instructions.
Maia, A. S. C., da Silva, R. G., Souza Junior, J. B. F. d., da Silva, R. B.,
& Domingos, H. G. T. (2009). Effective thermal conductivity of
the hair coat of Holstein cows in a tropical environment.
Revista Brasileira de Zootecnia, 38(11), 2218e2223.
McFarland, D. F. (2003). Fresstall design: Cow recommended
refinements. In Paper presented at the fifth international dairy
housing Fort Worth, Texas. http://elibrary.asabe.org/abstract.
asp?aid¼11613&t¼1.
Menter, F. R. (1992). Improved two-equation k-omega turbulence
models for aerodynamic flows. NASA Ames Research Center.
Minagawa, H. (1994). Surface area, volume, and projected area of
Japanese-shorthorn cattle measured by stereo
photogrammetry using non-metric cameras. Journal of
Agricultural Meteorology, 50(1), 17e22.
Mondaca, M. D., & Choi, C. Y. (2016a). A computational fluid
dynamics model of a perforated polyethylene tube ventilation
system for dairy operations. Transactions of the ASABE, 59(6),
1585e1594.
Mondaca, M. D., & Choi, C. Y. (2016b). An evaluation of simplifying
assumptions in dairy cow computational fluid dynamics
models. Transactions of the ASABE, 59(6), 1575e1584.
Mondaca, M. R., Rojano, F., Choi, C. Y., & Gebremedhin, K. G.
(2013). A conjugate heat and mass transfer model to evaluate
the efficiency of conductive cooling for dairy cattle.
Transactions of the ASABE, 56(6), 1471e1482.
Myers, R. H., Montgomery, D. C., & Anderson-Cook, C. M. (2016).
Response surface methodology: Process and product optimization
using designed experiments. John Wiley & Sons.
Norton, T., Grant, J., Fallon, R., & Sun, D.-W. (2009). Assessing the
ventilation effectiveness of naturally ventilated livestock
buildings under wind dominated conditions using
computational fluid dynamics. Biosystems Engineering, 103(1),
78e99.
Oh, S., Rheem, S., Sim, J., Kim, S., & Baek, Y. (1995). Optimizing
conditions for the growth of Lactobacillus casei YIT 9018 in
tryptone-yeast extract-glucose medium by using response
surface methodology. Applied and Environmental Microbiology,
61(11), 3809e3814.
Patankar, S. (1980). Numerical heat transfer and fluid flow.
Washington, DC: Hemisphere Publishing Corp.
Rong, L., Liu, D., Pedersen, E. F., & Zhang, G. (2015). The effect of
wind speed and direction and surrounding maize on hybrid
ventilation in a dairy cow building in Denmark. Energy and
Buildings, 86(0), 25e34.
Rong, L., Nielsen, P. V., Bjerg, B., & Zhang, G. (2016). Summary of
best guidelines and validation of CFD modeling in livestock
buildings to ensure prediction quality. Computers and
Electronics in Agriculture, 121, 180e190.
 b i o s y s t e m s e n g i n e e r i n g 1 7 5 ( 2 0 1 8 ) 1 e1 5
Rong, L., Pedersen, P., Jensen, T. L., Morsing, S., & Zhang, G. (2017).
Dynamic performance of an evaporative cooling pad
investigated in a wind tunnel for application in hot and arid
climate. Biosystems Engineering, 156(Supplement C), 173e182.
Schneider, P., Beede, D., & Wilcox, C. (1988). Nycterohemeral
patterns of acid-base status, mineral concentrations and
digestive function of lactating cows in natural or chamber heat
stress environments. Journal of Animal Science, 66(1), 112e125.
Shen, X., Zhang, G., & Bjerg, B. (2012). Investigation of response
surface methodology for modelling ventilation rate of a
naturally ventilated building. Building and Environment, 54,
174e185.
Shen, X., Zhang, G., & Bjerg, B. (2013). Assessments of
experimental designs in response surface modelling process:
Estimating ventilation rate in naturally ventilated livestock
buildings. Energy and Buildings, 62(0), 570e580.
St-Pierre, N. R., Cobanov, B., & Schnitkey, G. (2003). Economic
losses from heat stress by US livestock industries. Journal of
Dairy Science, 86(Supplement), E52eE77.
Strøm, J. S., & Zhang, G. (1989). Thermal control in animal
buildings. In Paper presented at the 11th International Congress on
Agricultural Engineering, Dublin.
Turner, L. W., Chastain, J. P., Hemken, R. W., Gates, R. S., &
Crist, W. L. (1992). Reducing heat stress in dairy cows through
sprinkler and fan cooling. Applied Engineering in Agriculture,
8(2), 251e256.
Tyson, J. T. (2010). Dairy heat abatement system selection tool. In
Paper presented at the 2010 ASABE Annual International Meeting,
Pittsburgh, Pennsylvania.
Wang, K., Wang, X., & Wu, B. (2014). Assessment of hygrothermal
conditions in a farrowing room with a wet-pad cooling system
based on CFD simulation and field measurements.
Transactions of the ASABE, 57(5), 1493e1500.
Wang, X., Zhang, G., & Choi, C. Y. (2018). Effect of airflow speed
and direction on convective heat transfer of standing and
reclining cows. Biosystems Engineering, 167, 87e98.
West, J. W. (2003). Effects of heat-stress on production in dairy
cattle. Journal of Dairy Science, 86(6), 2131e2144.
15
Whay, H. R., & Shearer, J. K. (2017). The impact of lameness on
welfare of the dairy cow. Veterinary Clinics: Food Animal Practice,
33(2), 153e164.
Whitcomb, P. J., & Anderson, M. J. (2004). RSM simplified: Optimizing
processes using response surface methods for design of experiments.
CRC press.
Worley, J. W., & Bernard, J. K. (2005). Use of high volume low
speed (HVLS) fans for cooling dairy cows in a free stall barn in
a hot humid climate. In Paper presented at the ASAE annual
meeting, Tampa, Florida. http://elibrary.asabe.org/abstract.asp?
aid¼19511&t¼5.
Worley, J. W., & Bernard, J. K. (2008). Cooling effectiveness of high-
volume low-speed fans versus conventional fans in a free-
stall dairy barn in hot, humid Conditions11Mention of a trade
name does not imply an endorsement. The Professional Animal
Scientist, 24(1), 23e28.
Xu, F., Lu, T. J., Seffen, K. A., & Ng, E. Y. K. (2009). Mathematical
modeling of skin bioheat transfer. Applied Mechanics Reviews,
62(5), 801e835.
Zhang, G., Morsing, S., & Strøm, J. S. (1992). Design principle and
control strategy for an automatically controlled hybrid
ventilation system. In Paper presented at the AgEng 92
International Conference on Agriculture Engineering Uppsala.
Zimbelman, R., Rhoads, R., Rhoads, M., Duff, G., Baumgard, L., &
Collier, R. (2009). A re-evaluation of the impact of temperature
humidity index (THI) and black globe humidity index (BGHI)
on milk production in high producing dairy cows. In Paper
presented at the Proceedings of the southwest nutrition and
management conference, Tempe. Arizona, February.
Zong, C., Li, H., & Zhang, G. (2015). Ammonia and greenhouse gas
emissions from fattening pig house with two types of partial
pit ventilation systems. Agriculture, Ecosystems & Environment,
208, 94e105.
Zong, C., & Zhang, G. (2014). Numerical modelling of airflow and
gas dispersion in the pit headspace via slatted floor:
Comparison of two modelling approaches. Computers and
Electronics in Agriculture, 109, 200e211.