Upload
duongphuc
View
224
Download
2
Embed Size (px)
Citation preview
TechnicalPapers
-51-
Keihin Technical Review Vol.6 (2017)
Investigation of CFD Temperature Prediction for
HVAC Units※
Technical Paper
1. Introduction
In recent years , due to h igh per formance
requirements of air conditioning systems, the
c o m p l ex i t y o f H VAC i n t e r n a l f l ow p a t h i s
progressing, i.e. leaf/right independent process
control type system for automobile, and rear air
conditioning for the rear seat passengers, etc.,
furthermore, despite speeding up in development,
the demand for quality assurance, performance
improvement and cost reduction is increasing.
Therefore, in the development project of HVAC
units, it is necessary to ensure target performance
at an early stage and to make efforts to improve
quality.
In o rder to ach ieve the demand ta rge t in
performance improvement at an early stage of
development, i t is indispensable to grasp the
information on flow and temperature field using CFD
simulation. However, in the case where warm air
and cool air mixed inside HVAC units, if utilization
of a default simulation condition in the commercial
CFD software to predict temperature at duct outlet,
in which the default setting of Prandtl number is
usually set to about 0.7 to 0.9 as laminar airflow(1),
there is a big difference between the simulation
result and the measured data, and the quantitative
evaluation is very difficult and a hard task, because
the airflow inside HVAC units has a very complex
turbulence vortex structure(2), and moreover the
structure of the turbulence airflow changes with
the variation of the gap between the temperature
control damper and the associated shut rib, due to
the various air mixing opening degrees. Therefore,
the determination of the turbulent Prandtl number
is a key point for improvement of CFD simulation
accuracy.
According to our previous s tudies , i t has
been confirmed that some simulation results for
temperature are close to the measured valves
by adjusting the turbulent Prandtl number to an
arbitrary constant of 0.7 or less, and it is also known
that the precision of simulation results deteriorates
depending on the position of measurement point, the
reason for this is that the turbulent Prandtl number
is not a constant value but is considered to be
change with variation of the flow and temperature
field. For instance, Kitada et al.(3) determined a
function of the turbulent Prandtl number with eddy
viscosity coefficient and the temperature gradient by
an experimental studies of air mix basic model, and
applied in CFD simulation to improve temperature
prediction accuracy.
In the study, a functionalized Prt is proposed by
theoretical analysis using the numerical results of
velocity and the measured data of temperature in a
Key Words: Air conditioning system, HVAC unit, CFD, Turbulent prandtl number
Koei ITO*1 Jun LIU*1 Jun ONODERA*1
*1 Air-conditioning System Development Department, Air-conditioning System Business Operations
※ Received 19 June 2017, Reprinted with permission from paper No. 20175127, proceedings of technical session No. 25-17, 2017 JSAE Annual Congress (Spring). Further use or distribution is not permitted without permission from JSAE.
-52-
Investigation of CFD Temperature Prediction for HVAC Units
transport terms (u'v', T'v') and utilization of the
definition of the eddy diffusivity, which gives
dy
duε
ρτ t = u'v' = εm (5)
dy
dTρq t = T'v' = εh (6)
then substituting the Eqs. (5) and (6) into Eq. (1)
yields
dU
dT
dU
dy
dy
dT
T'v'
u'v'
T'v'
u'v'Prt = = (7)
thus, if the velocity and temperature fields, and the
turbulent fluctuating terms are known, the Prt can be
evaluated.
The latter method will be used here.
3. Investigation Using an Air Mix Model
3.1. Experimental Verification
As shown in Fig. 1, create a simple experimental
model to imitate the air mix condition inside
the HVAC units, two heat exchangers are placed
in model, one heat exchanger with warm water
generates a warm air flow, and mixing with a cold
air flow is performed by a guide near the confluence
on the downstream, mult iple sensor pins are
arranged in lattice pattern in air mix region.
T Sensor V Sensor
Sensor Pin
IN-Flow
Out-Flow
Observation area
Heat Exchanger
Warm water
h
b
t
Air mix Guide
Fig. 1 Air mix model for Experiment
simple air mix model. In addition, the functionalized
Prt was applied in CFD simulation to predict the
temperature of the air flow inside HVAC units,
and the simulation accuracy will be verified by the
comparison between the experimental data and the
simulated result.
2. Turbulent Prandtl Number
2.1. Definition of Turbulent Prandtl Number
The turbulent Prandtl number is a classical
approach for the heat transfer problem in turbulence,
similarly as in molecular Prandtl number, which is
defined as the ratio between the momentum eddy
diffusivity εm and the heat transfer eddy diffusivity
εh, which is
Prt = εh
εm (1)
using Boussinesq relation, the shear stress τ and
heat flux q can be written as follows(4).
dy
du = (ν + εm)
ρτ
(2)
dy
dT = (α + εh) ρCp
q (3)
There are basically two methods by which Prt
may be evaluated experimentally.
In one method, utilizing experimental time-
average velocity and temperature profiles together
with the integrated Reynolds equations. From
equations (2) and (3) to obtain
du – 1
dyρτ
dy
dT
ρCp
q – 1
Prt = (4)
thus, if the momentum and heat flux variation with
the distance from the wall are known, εm, εh, and
hence Prt can be evaluated.
Alternatively, direct measurement of the Reynolds
TechnicalPapers
-53-
Keihin Technical Review Vol.6 (2017)
The experiment was carried out on a test bench
that can be controlled the flow rate of air and
temperature of warm water, the temperature of
warm water is fixed and the inlet air volume is
set in 3 cases. A multi-channel wind velocity and
wind temperature sensor system is fixed on multiple
sensor pins to measure the velocity and temperature
field of the air mix region.
3.2. Validation by CFD
Since the wind velocity sensor is not able to
measure Reynolds stress u'v' of the air flow field,
after clarifying how much the CFD velocity result
correlates with the measured data, the CFD results
will be attempted to use instead of the measured
data.
3.2.1. Simulation Conditions
The 3D-CFD model completely reflects the
experimental model with the multiple sensor pins (see
Fig. 2). The heat exchanger is defined as a porous
body, and the permeability coefficient is calculated
from the experimental data of air flow resistance.
Additionally, the amount of heat generation per unit
volume is inputted as a quadratic function of air
velocity in the heat exchanger, and the maximum
heating temperature is limited so as to be equal
to the warm water temperature at inlet of heat
exchanger.
In this research, the CFD simulation is carried
out by commercial CFD software, and the simulation
conditions are listed in Table 1.
Fluid model
State
Mesh Structure
Turbulence model
Difference Scheme
Inlet volume flux [m3/h]
Air mix model
Unsteady
All Polyhedral
DES (SST k-ω)
2nd up wind scheme
200, 300, 400
Table 1 Simulation conditions
12
0 5
50V [m/s] T [°C]
Fig. 3 Velocity and temperature of CFD at 400m3/h
IN-Flow
Out-FlowAir mix Guide
PorousGeneration of heat
Porous
Sensor Pin
Fig. 2 Air mix model for CFD
ExperimentSimulation
20
15
10
5
0
Vel
ocity
[m/s
]
Measurement Point
200m3/h 300m3/h 400m3/h
Fig. 4 Correlation between experiment and CFD part1
3.2.2. A Comparison of Simulation and Experiment
Figure 3 present the flow field and temperature
f i e ld o f CFD resu l t s i n case o f 400m 3/h , i t
indicates that the mixing region has a non-uniform
temperature field, and at each measurement point in
grid pattern, there are temperature and flow fields
with different gradients, respectively.
A comparisons of velocity at each measured
point between simulat ion and experiment are
shown in Fig. 4-5, they show that the CFD results
of ve loc i ty g ive a c lose agreement wi th the
experimental data, the coefficient of determination
R2 equals 0.89 and the standard error of the mean is
-54-
Investigation of CFD Temperature Prediction for HVAC Units
17.8%, in the study, the error rate level is acceptable
for determination of Prt.
coefficient, respectively, they can be determined
using separation of variables Re and Pr by simply
taking the natural logarithm to Eq. (8).
Equation (8) is applied to CFD simulation of the
simple air mix model, and compared to the case of
Prt = 0.9, maximum error decreased from 10.8°C to
8.0°C within the interval ±2σ as shown in Fig. 8,
and the probability of occurrence within the interval
V [m/s]
12
0
Velocity FieldExperiment
CFD
02468
1012141618
0 2 4 6 8 10 12 14 16 18
Vel
ocity
on
CFD
[m
/s]
Velocity on Experiment [m/s]
Fig. 5 Correlation between experiment and CFD part2
4. Determination of Functionalized Prt
As shown in Fig. 6, a new orthogonal coordinate
system (xi*, yi
*) which the xi* axis fixed along the
direction of main flow U at measurement point is
introduced, where subscript i denotes the number
of measurement point, and the coordinate system
(xi, yi) is the reference coordinate system in CFD
simulation.
In the coordinate system (xi*, yi
*), the turbulent
fluctuating term (u'v', T'v') and spatial derivative
of velocity and temperature fields (dU, dT) at
each measurement point are calculated using CFD
numerical results of velocity and measured data of
temperature, respectively. Thereupon, according to
Eq. (7), Prt can be obtained with known Reynolds
transport terms and the velocity and temperature
fields.
H e r e , t h e r e l a t i o n s h i p o f P r t w i t h t w o
dimensionless numbers Reynolds number (Re) and
molecular Prandtl number (Pr) is plotted in Fig. 7,
and the Prt is assumed as a function of both Re and
Pr as follows
Prt = aReb Pr c (8)
where, the coefficients a, b and c are constant
Fig. 6 Components of velocity
Fig. 7 Plotting Prt with variation of Re and Pr
uxi
uyi
uxi*
=
uy
ux
cosθisinθi
cosθi − sinθiuxi*
uyi*
uyi*
θ
Prt
Prt
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
Re HighLow Pr HighLow
50
5
Experiment Prt = Function Prt =0.9T [°C]
Average Error
10.8°C
8.0°C
2σ (Prt = 0.9)
2σ (Prt = Function)
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Error [°C]
0
5
10
15
20
Freq
uenc
y
Prt = 0.9Prt = Function
Fig. 8 Comparison of temperature fields
TechnicalPapers
-55-
Keihin Technical Review Vol.6 (2017)
±5°C was upgraded 15%, which improves from
68.2% for Prt = 0.9 to 83.6% for functionalized Prt.
Consequently, utilization of the functionalized Prt is
effective for improving the temperature simulation
accuracy of air mix model.
5. Applied on CFD Simulation of HVAC Units
T h e f u n c t i o n a l i z e d P r t wa s e m p l oy e d t o
CFD simulation of HVAC units (see Fig. 9) for
verifying its simulation accuracy, and Table 2
shows the conditions of CFD simulation. Here, the
dimensionless temperature Td is defined by dividing
the air temperature T by the heater warm water
temperature TH.
F i g u r e 10 p r e s e n t t h e c h a r a c t e r i s t i c s o f
temperature varied with respect to damper open
degree changes in the Bi-Level (B/L) and the
Heat Differential (H/D) modes of HVAC units,
where solid line indicates measured data and dash
line indicates simulation results, respectively. It
shows that the inclination of the thermal control
characteristics at each blowing outlet are obtained,
and the tendency is approximately in agreement with
experimental data.
In regards to the probability distribution of errors,
in comparison to a default case of constant Prt = 0.9,
maximum error decreased from 11.2°C to 7.9°C
within the interval ±2σ as shown in Fig. 11, and the
probability of occurrence within the interval ±5°C
is 67.4% for Prt = 0.9 and 80.4% for functionalized
Prt, which was upgraded 13.4%, this demonstrates
1. VENT Dr2. VENT As3. HEAT Dr4. HEAT As5. R-HEAT Dr6. R-HEAT As7. DEF Dr8. DEF As
2
1
87
3
4
65
Fig. 9 HVAC model
00.10.20.30.40.50.60.70.80.9
1
0 10 20 30 40 50 60 70 80 90 100
Td
Air mix position [%]
00.10.20.30.40.50.60.70.80.9
1
0 10 20 30 40 50 60 70 80 90 100
Td
Air mix position [%]
B/L H/D
B/L H/D
ExperimentSimulation
ExperimentSimulation
HEAT DrHEAT As
VENT DrVENT AsR-HEAT DrR-HEAT AsDEF DrDEF As
Table 2 Simulation conditions
Fig. 10 Thermal control characteristics
Fluid model
State
Mesh Structure
Turbulence model
Difference Scheme
Semi center HVAC
Steady
All Polyhedral
Realizable k-ε
2nd up wind scheme
Fig. 11 Comparison of temperature
-7.9°C2σ (Prt = Function)
0
10
20
30
40
50
60
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14
Freq
uenc
y
Prt = 0.9Prt = Function
2σ (Prt = 0.9)
Average Error
Error [°C]
11.2°C
-56-
Investigation of CFD Temperature Prediction for HVAC Units
that the functionalized Prt has been effective for
improving temperature prediction accuracy in CFD
simulation of the HVAC units.
On the one hand, in the opening degree 70% of
the H/D mode, although the simulation temperature
(Td) of DEF was only 0.05 lower than the measured
data and it showed a relatively good results, but on
the other hand, in the opening degree 100%, the
simulation temperature of DEF and R-HEAT were
higher than the measured data, and that of HEAT
showed a tendency to be lower conversely, it is
considered that the correction of functionalized Prt
is insufficient for variation of thermal transfer and
diffusion, due to the reason that Re number of air
toward the DEF becomes high and exceeds the range
as shown in Fig. 12. This is probably due to the
fact that the air mix model which used in the study,
is relatively simple mixing with only one air mix
guide, and the application range of functionalized Prt
is the same length as the Prt distribution range (see
Fig. 7).
In order to improve adaptability to different
complex flow fields in each mode and opening
degree of HVAC units, the coefficients a, b and c in
Eq. (8) are considered to vary at a critical value of
Re number, and further verification is necessary.
and numerical results for velocity, and the turbulent
Prandtl number (Prt) can be expressed as a function
of both Reynolds number (Re) and molecular Prandtl
number (Pr).
In addition, the functionalized Prt was employed
to predict the temperature of the air flow at the
outlet with CFD simulation of HVAC units. As a
result, simulation results of the temperature field
were in relatively good agreement with experimental
data. This demonstrates that the functionalized
Prt has been effective for improving temperature
prediction accuracy in CFD simulation of the HVAC
units.
7. Future Tasks
As mentioned in chapter 5, since only one simple
flow field is discussed, and the applicability range to
CFD simulation of HVAC units was limited, then in
the future tasks, the applicability can be improved
with increasing of discussion cases, such as the
number of guides and variation of the temperature
difference between air and warm water.
Moreover, i t just set the wall heat transfer
coefficient as a constant in CFD simulation in the
study, however, the certainty of the constant has not
been clarified. Therefore, the other future challenge
is to elucidate the physical phenomenon concerning
heat exchange between the external environment
and HVAC case or the duct, and to identify the wall
heat transfer coefficient for further improvement in
simulation accuracy.
References
(1) Warren h. Giedt: Principles of Engineering Heat
Transfer, translation by Suzumu Yokohori, etc.,
Maruzen Press., p. 333 (1960)
(2) Jun Onodera, etc.: Numerical Simulation of Flow-
Induced Noise in Automotive HVAC Systems,
Fig. 12 Flow field of HVAC unit in H/D mode
Td Re PrtA/M
Position
70%
100%
1
0
High
Low 0
1
6. Conclusion
In the study, a functionalized Prt is obtained from
a combination of experimental data for temperature
TechnicalPapers
-57-
Keihin Technical Review Vol.6 (2017)
Authors
K. ITO J. LIU J. ONODERA
In the paper, a theoret ical-based approach
to identifying turbulent Prandtl number (Prt) in
air mixing is challenged, and a function of non-
unity Prt number is proposed. In addition, we are
also planning to improve the function of Prt with
increment of experimental conditions in the next
step, and to expand its application range on CFD
simulation of HVAC units. Finally, I would like to
thank everyone for their guidance and cooperation
during the research. (ITO)
Keihin Technical Review, Vol. 1, p. 19-24 (2012)
(3) M o t o h i r o K i t a d a , e t c . : D ev e l o p m e n t o f
Automotive A/C System Basic Performance
Simulator-Accuracy of Temperature Calculation
Technique Development, Proceeding of 2000
JSAE congress (Autumn), No. 169 (2000)
(4) Basim O. Hasan: Turbulent Prandtl Number and
its Use in Prediction of Heat Transfer Coefficient
for Liquids, Nahrain University, College of
Engineering Journal (NUCEJ) Vol. 10, No. 1
(2007)
(5) Shinobu Toshima: Yosetsu Hinshitsu Kanri, Japan
Standards Association, p. 13-23 (1967)