Documente Academic
Documente Profesional
Documente Cultură
COOLING COILS
Sponsored by
ASHRAE, Inc.
ASHRAE 1194-RP
NOVEMBER 2005
ii
TABLE OF CONTENTS
Page
LIST OF TABLES............................................................................................................. iv
LIST OF FIGURES ........................................................................................................... vi
NOMENCLATURE ........................................................................................................... x
ABSTRACT..................................................................................................................... xiii
1. INTRODUCTION .......................................................................................................... 1
1.1 Background................................................................................................................ 1
1.2 Literature Review ...................................................................................................... 2
1.3 Objectives and Methodology..................................................................................... 5
1.4 Report Organization .................................................................................................. 6
2. EXPERIMENTAL TEST FACILITY ............................................................................ 8
2.1 Description ................................................................................................................ 8
2.2 Instrumentation.......................................................................................................... 9
2.3 Heat Exchangers...................................................................................................... 11
2.4 Test Procedure ......................................................................................................... 15
2.5 Uncertainty Analysis ............................................................................................... 16
2.6 Heat Transfer Coefficients ...................................................................................... 21
3. ANALYSIS OF MODELING ASSUMPTIONS.......................................................... 25
3.1 Introduction ............................................................................................................. 25
3.2 Steady-State Modeling of Single Straight Finned Tube.......................................... 27
3.3 Investigation of Water Condensate Effect............................................................... 29
3.4 Dynamic Modeling of Single Straight Finned Tube ............................................... 35
3.5 Investigation of Assumption of Lewis Number of Unity........................................ 36
3.6 Investigation of Application of Fin Efficiency in Dynamics .................................. 41
3.7 Improvement of Heat Transfer Fin Efficiency under Dehumidification................. 45
4. DYNAMIC FORWARD MODELING ........................................................................ 50
4.1 Model Development ................................................................................................ 50
4.1.1 Detailed Model.................................................................................................. 50
4.1.2 Counter-Flow Model......................................................................................... 55
4.1.3 Water-Side ε-NTU Model................................................................................. 57
4.2 Solution Techniques ................................................................................................ 58
iii
Page
4.2.1 Detailed Model.................................................................................................. 58
4.2.2 Simplified Models............................................................................................. 64
4.3 Numerical Studies ................................................................................................... 65
4.3.1 Detailed Model.................................................................................................. 65
4.3.2 Simplified Models............................................................................................. 71
4.4 Experimental Validation.......................................................................................... 78
4.5 Computational Speed and Implementation ........................................................... 101
4.6 Comparisons with Existing Simplified Models..................................................... 102
5. DYNAMIC INVERSE MODELING ......................................................................... 110
5.1 Model Description and Implementation................................................................ 110
5.2 Parameter Estimation............................................................................................. 111
5.3 Experimental Validation........................................................................................ 118
6. CONCLUSIONS AND RECOMMENDATIONS ..................................................... 123
6.1 Conclusions ........................................................................................................... 123
6.2 Recommendations ................................................................................................. 124
LIST OF REFERENCES................................................................................................ 125
APPENDICES
Appendix A: Experimental Sample Data of a 4-row Coil with Louvered Fins .......... 127
Appendix B: Experimental Validation Results of a 4-row Coil With Louvered Fins. 128
Appendix C: Model User Manual ............................................................................... 137
C.1 Forward Model Example ................................................................................... 137
C.2 Inverse Model Example ..................................................................................... 141
iv
LIST OF TABLES
Table Page
2.1: Characteristics of test coils. ..................................................................................... 14
2.2: Dynamic coil test conditions.................................................................................... 16
2.3: Uncertainties of direct measured values. ................................................................. 17
2.4: Uncertainties of coil inlet conditions. ...................................................................... 18
2.5: Uncertainties of coil outlet conditions. .................................................................... 19
2.6: Uncertainties of coil heat transfer rates. .................................................................. 20
2.7: Uncertainties of coil mass flow rates of water condensate...................................... 21
3.1: Characteristics of a single straight finned tube........................................................ 31
3.2: Inlet conditions for investigation of water condensate. ........................................... 32
3.3: Outlet conditions and corresponding information of water condensate for a single
straight finned tube. ................................................................................................. 33
3.4: Outlet conditions and corresponding information of water condensate for an 8-
row coil. ................................................................................................................... 34
3.5: Thermal and vapor diffusivity data for dry and saturated moist air. ....................... 37
3.6: Inlet conditions for investigation of Lewis number................................................. 38
3.7: Inlet conditions for investigation of fin efficiencies................................................ 43
4.1: Inlet conditions for numerical studies of forward models. ...................................... 66
4.2: Inlet conditions for changes in air flow rate of an 8-row coil.................................. 79
4.3: Inlet conditions for changes in water flow rate of an 8-row coil............................. 85
4.4: Inlet conditions for simulated feedback control of fluid flow rates of an 8-row
coil. .......................................................................................................................... 89
4.5: Inlet conditions for changes in air inlet temperature and humidity, and water inlet
temperature of an 8-row coil.................................................................................... 93
4.6: Inlet conditions for simulated feedback control of fluid flow rates of a 4-row coil.97
4.7: Computational speeds for different solution methods. .......................................... 101
4.8: Inlet conditions for model comparisons................................................................. 103
v
Table Page
5.1: Steady-state experimental training data of an 8-row coil for heat conductance
regression. .............................................................................................................. 112
5.2: Transient experimental training data of an 8-row coil for heat capacitance
regression. .............................................................................................................. 113
Appendix Table
A.1: Steady-state experimental data of a 4-row coil for air-side convection coefficient
calculation.............................................................................................................. 127
B.1: Transient inlet conditions of a 4-row Coil. ............................................................ 128
vi
LIST OF FIGURES
Figure Page
2.1: Air-side arrangement of PACT facility. .................................................................... 8
2.2: Water-side arrangement of PACT facility. ................................................................ 9
2.3: Schematic illustrations of coil flow circuiting......................................................... 12
2.4: Physical images of test coils. ................................................................................... 13
2.5: Comparisons of air-side convection coefficient for an 8-row coil and a 4-row coil.
................................................................................................................................. 23
3.1: Schematic illustration of a single straight finned tube............................................. 26
3.2: Laminar condensation on a straight fin surface. ...................................................... 30
3.3: Schematic of condensate film thickness along fin height direction between fins
of an 8-row coil........................................................................................................ 35
3.4: Transient comparison results for Le = 1.0 and Le = 0.9 a step change in air flow
rate. .......................................................................................................................... 40
3.5: Transient comparison results for simplified model and distributed model for a
dry fin....................................................................................................................... 43
3.6: Transient comparison results for simplified model and distributed model for a
wet fin. ..................................................................................................................... 44
3.7: Schematic illustration of uniform cross section of a straight fin. ............................ 46
3.8: Improved transient results for simplified model for a wet fin. ................................ 49
4.1: Schematic of a full-row serpentine cooling coil. ..................................................... 51
4.2: Solution procedure for detailed steady-state model................................................. 62
4.3: Solution procedure for detailed dynamic model...................................................... 64
4.4: Transient comparison results for detailed model at different values of N for an
8-row dry coil (∆t = 1 second)................................................................................. 67
4.5: Transient comparison results for detailed model at different values of N for an
8-row wet coil (∆t = 1 second). ............................................................................... 68
4.6: Transient comparison results for detailed model at different values of ∆t for an
8-row dry coil (N = 80)............................................................................................ 69
vii
Figure Page
4.7: Transient comparison results for detailed model at different values of ∆t for an
8-row wet coil (N = 80). .......................................................................................... 70
4.8: Transient comparison results for detailed model and counter-flow model for an
8-row dry coil........................................................................................................... 72
4.9: Transient comparison results for detailed model and counter-flow model for an
8-row wet coil. ......................................................................................................... 73
4.10: Transient comparison results for detailed model and counter-flow model for a
4-row dry coil........................................................................................................... 74
4.11: Transient comparison results for detailed model and counter-flow model for a
4-row wet coil. ......................................................................................................... 75
4.12: Transient comparison results for detailed model and water-side ε-NTU model for
an 8-row dry coil...................................................................................................... 76
4.13: Transient comparison results for counter-flow model and water-side ε-NTU
model for an 8-row wet coil..................................................................................... 77
4.14: Transient responses to increase in air flow rate for an 8-row fully dry coil. ........... 80
4.15: Transient responses to increase in air flow rate for an 8-row partially dry and
partially wet coil. ..................................................................................................... 81
4.16: Transient responses to increase in air flow rate for an 8-row fully wet coil............ 82
4.17: Transient responses to increase in air flow rate at low water flow rates for an
8-row partially dry and partially wet coil. ............................................................... 83
4.18: Transient responses to decrease in air flow rate for an 8-row partially dry and
partially wet coil. ..................................................................................................... 84
4.19: Transient responses to increase in water flow rate for an 8-row partially dry and
partially wet coil. ..................................................................................................... 86
4.20: Transient responses to increase in water flow rate at high air inlet humidities for
an 8-row partially dry and partially wet coil............................................................ 87
4.21: Transient responses to increase in water flow rate at low air flow rates for an
8-row partially dry and partially wet coil. ............................................................... 88
4.22: Simulated feedback control of air flow rate for an 8-row coil................................. 90
4.23: Simulated feedback control of water flow rate for an 8-row coil............................ 90
4.24: Transient responses to simulated feedback control of air flow rate for an 8-row
partially dry and partially wet coil........................................................................... 91
4.25: Transient responses to simulated feedback control of water flow rate for an
8-row partially dry and partially wet coil. ............................................................... 92
viii
Figure Page
4.26: Transient responses to decrease in air inlet temperature for an 8-row partially dry
and partially wet coil................................................................................................ 94
4.27: Transient responses to increase in air inlet humidity for an 8-row partially dry
and partially wet coil................................................................................................ 95
4.28: Transient responses to increase in water inlet temperature for an 8-row partially
dry and partially wet coil. ........................................................................................ 96
4.29: Simulated feedback control of air flow rate for a 4-row coil................................... 98
4.30: Simulated feedback control of water flow rate for a 4-row coil.............................. 98
4.31: Transient responses to simulated feedback control of air flow rate for a 4-row
partially dry and partially wet coil........................................................................... 99
4.32: Transient responses to simulated feedback control of water flow rate for a 4-row
partially dry and partially wet coil......................................................................... 100
4.33: Step change in air flow rate. .................................................................................. 103
4.34: Step change in water flow rate............................................................................... 104
4.35: Sinusoidal variation in air flow rate....................................................................... 104
4.36: Sinusoidal variation in water flow rate. ................................................................. 104
4.37: Transient comparison results to step change in air flow rate for different models.105
4.38: Transient comparison results to step change in water flow rate for different
models.................................................................................................................... 106
4.39: Transient comparison results to sinusoidal variation in air flow rate for different
models.................................................................................................................... 107
4.40: Transient comparison results to sinusoidal variation in water flow rate for
different models. .................................................................................................... 108
5.1: Comparisons for training data and regression results under steady-state
conditions............................................................................................................... 116
5.2: Heat conductance comparisons for empirical correlations and regression results 117
5.3: Comparisons for training data and regression results under transient conditions . 118
5.4: Inverse model validation under steady-state conditions ........................................ 119
5.5: Inverse model validation under an increase in air flow rate for an 8-row partially
dry and partially wet coil. ...................................................................................... 120
5.6: Inverse model validation under simulated feedback control of air flow rate for
an 8-row partially dry and partially wet coil.......................................................... 121
ix
NOMENCLATURE
Symbols
A area
C heat capacitance
Ċ heat capacitance rate
CF correction factor
Cp specific heat
D diameter; water vapor diffusivity in air
df fin thickness
f friction factor
g gravitational acceleration
h convection coefficient; specific enthalpy
H height
hD mass transfer coefficient
k heat conductivity
L length
Le Lewis number
M mass
M mass flow rate
NTU number of heat transfer unit
Nu Nusselt number
P perimeter
Pr Prandtl number
q heat transfer rate
R heat resistance
xi
Symbols
Re Reynolds number
RH relative humidity
t time variable
T temperature
U uncertainty
V velocity
x space variable
y space variable
Greek Letters
α thermal diffusivity
δ water condensate film thickness
Γ mass flow rate of water condensate
ε effectiveness
η fin efficiency
Θ dimensionless temperature
µ viscosity
ρ density
Ψ dimensionless humidity
ω humidity ratio
Subscripts
a air
b fin base
c coil material; cross section
f fin
in inlet
out outlet
s saturation
xii
Symbols
t tube
w water
Superscripts
’ per unit length
” per unit area
* heat and mass transfer
0 previous time
xiii
ABSTRACT
developed on the basis of the forward model and a two-step approach for parameter
estimation was presented. Both models were coded in C++ and implemented as a DLL
file that can be called in MatLab.
1
1. INTRODUCTION
1.1 Background
A chilled water cooling coil can operate under either steady-state or dynamic
conditions at different times. Under steady-state conditions, coil inlet and outlet
parameters are constant for a sufficient time period so that the coil states do not change
appreciably. However, a coil often operates with dynamic changes in inlet conditions
(e.g., fluid temperatures and flow rates). This can happen when a cooling system starts or
is shut down or when there is a sudden disturbance introduced by feedback control in
response to a change in outdoor conditions, load requirements, set point temperature, etc.
With a single disturbance, a coil attempts to move from one equilibrium state to another.
Dynamic models can predict the operating states of the coil during these transitions as
opposed to steady-state models, which predict equilibrium states before and after such a
change.
Dynamic models are useful in early development stages for new feedback
controllers. For chilled water cooling coils, supply air set point temperature leaving a
coil is typically maintained by adjusting water flow rate using a feedback controller to
open and close a valve. New controller concepts can be tested extensively using
simulation more cost-effectively than using experimental setups. In addition, dynamic
models could be embedded within “emulators” for testing actual control hardware. Good
controller performance not only reduces energy costs but also improves comfort and
reduces wear and tear on actuators, and therefore increases equipment lifetime.
Dynamic models are also useful in testing automated fault detection and
diagnostic (FDD) algorithms that are being developed. FDD algorithms typically utilize
steady-state models and incorporate steady-state detectors to filter out dynamic data.
2
Dynamic models that allow introduction of faults could be used to test these steady-state
FDD methods much faster and at much lower cost than performing experiments.
It is important to have fully-validated dynamic models for development and
evaluation of new controllers and FDD algorithms. In this project, the focus is to develop
dynamic models for chilled water cooling coils, which is a key component in commercial
cooling systems.
An extensive literature review was conducted for the purpose of this project. In
the last 40 years or so, a number of cooling coil models have been developed. Some of
them are quasi-static models that may be appropriate for transient simulations if the
dynamic variations of the inputs are slower than the response of the cooling coil. The
model developed by Braun et al. (1989) is an example of simplified quasi-static models,
which utilizes effectiveness relationships for heat transfer only, and heat and mass
transfer. This lumped model was validated by detailed numerical solutions and
manufacture’s catalog data to work well in predicting air and water outlet states under
steady-state conditions.
For transient models in the HVAC field, Bourdouxhe et al. (1996) developed an
annotated reference guide through a research project (RP-738) sponsored by ASHRAE.
It was found that there is a lot of literature on transient modeling of heat exchangers, but
there appears to be relatively little work on transient modeling of cooling and
dehumidifying coils.
For heat exchangers without dehumidification, Gartner and Harrison (1963)
developed an analytical model for cross-flow heat exchangers suitable for the purpose of
control engineering. They derived three partial differential equations, which describe
energy storage in the primary fluid (air), heat exchanger material, and secondary fluid
(water), with the assumption of steady flow for both flow streams. The solutions were
presented in terms of transfer functions that relate outlet air and water temperatures to
variations of inlet air and water temperatures under the condition of constant fluid flow
3
rates. The validity of the model was demonstrated through comparisons with transient
experimental data. This model has been the basis of many other transient heat exchanger
models, including Gartner and Harrison (1965), Gartner and Daane (1969), Tamm
(1969), Tamm and Green (1973), Bhargava et al. (1975), and Jawadi (1988). Gartner and
Daane (1969) developed a mathematical model for serpentine cross-flow heat exchangers
considering the changes of fluid velocities, while Jawadi (1988) simplified the model by
neglecting transients associated with air.
Dehumidification adds a significant complication. A small number of papers
were found for dynamic modeling of cooling and dehumidifying coils. Based on the
level of detail, the models are categorized into reference model, lumped model, and filter
model as proposed by Bourdouxhe et al. (1996). A reference model is based on a fairly
detailed understanding of physical phenomena and usually involves partial differential
equations. A lumped model is also based on a physical representation, but with certain
simplifications, so that the mathematical expressions are reduced to a small number of
ordinary differential and algebraic equations. A filter model is based on the use of a time
constant added to the outputs of a steady-state model. The physical significance of a
filter model may be very poor.
McCullagh et al. (1969) presented a row-by-row reference model that considered
the dynamic behavior of chilled water cooling and dehumidifying coils. A coil was
discretized into a series of rows. For each row, two lumped approximations of tube and
fin were assumed, which led to uniform temperature distributions in tube and fin. It was
assumed that all the heat was transferred from air to fin, and from fin to tube, that air and
water had constant velocities, and air was completely mixed between rows. The energy
balance applied to the control volumes of air, tube, fin, and water led to five partial
differential equations for each row. A finite-difference numerical solution was used to
determine transient responses of air outlet temperature and humidity, and water outlet
temperature to a step change in water inlet temperature. Steady-state predictions of the
model were compared to experimental results under wet conditions and showed
reasonably good agreement for both sensible and latent heat transfer, except for two-row
4
coils where the assumption of complete mixing of air between rows was poor. The
transient response of the model was not validated using experiments.
Clark (1985) developed a filter model for circular or continuous finned serpentine
cooling and dehumidifying coils with four or more rows in a counter cross-flow
configuration. The model was based on the steady-state model presented by Elmahdy
and Mitalas (1977). In dry conditions, the heat flow rate between moist air and water was
calculated on the basis of logarithmic mean temperature difference (LMTD); while in wet
cases logarithmic mean enthalpy difference (LMHD) was used to calculate the heat flow
rate. The coil dynamics were modeled very simply by adding a single time constant to
the steady-state air outlet temperature and humidity, and water outlet temperature. This
time constant was a function of heat capacitance associated with coil material and overall
heat transfer coefficient. Use of the model to represent the dynamics of coils with fewer
than four rows was not recommended. The dynamics of the coil model were investigated
experimentally only for dry conditions with changes in water flow rate and water inlet
temperature. The agreement between model and experiment was reasonably good for the
conditions considered.
Ding et al. (1990) developed a lumped model applied to finned serpentine cooling
and dehumidifying coils. The dynamic behavior of a coil was described by starting with
a steady-state model. It was assumed that a first-order model was sufficiently accurate to
reproduce the response of a coil to changes in water inlet temperature or water flow rate.
A single capacitance represented the total energy storage of the coil metal material and
water. The energy transfer rates were estimated using effectiveness models similar to
those presented by Braun et al. (1989). Air dry-bulb temperatures were used as driving
potentials for dry cases, and air wet-bulb temperatures were used for wet conditions. The
air dry bulb temperature at the outlet of coil was calculated based on the “contact”
method, which considered a uniform external coil surface temperature. As commonly
done in cooling and dehumidifying coil models, the energy storage in condensed water on
the external surface of coil was neglected. The model was validated experimentally only
for dry coils, and showed that the agreement between model prediction and experimental
measurement was good for air outlet temperature, but poor for water outlet temperature.
5
dehumidifying coil, using physical parameters that characterize the coil and are provided
by a user. The inverse model is typically a simpler formulation where lumped parameters
are determined from regression using experimental training data or data generated by a
detailed forward model. For either type of model, the goal is to achieve the simplest
model that represents a good compromise between accuracy and computational
requirements, since a coil model is likely used in combination with other transient models
of HVAC&R equipment.
The general approach to meet the objectives starts with the analysis of dynamic
modeling assumptions. As the first step, a reference model for a single straight finned
tube with simplified geometry was developed, which allows evaluation of the effects of
several important simplifications. After the identification of reasonable assumptions, a
detailed forward model for representative cooling coil geometries was developed, along
with some simplified models. A numerical study was performed to evaluate the best
formulation, integration method, and space and time discretization to achieve desired
accuracy and computational speed. The model performance was fully validated using
experimental data from two different coils of varying dimensions. A simpler inverse
model was developed adopting the forward modeling approach and non-linear regression
technique was used for parameter estimation to minimize errors between model
predictions and training data.
demonstrated. More validation results can be found in Appendices B. Model results are
also compared with results for existing models from the literature. Also described are the
execution speed and model implementation. A detailed user manual is available in
Appendix C. A method for estimating parameters is the main focus in Chapter 5 for
dynamic inverse modeling. Conclusions are summarized in the last chapter and
recommendations are proposed for further research.
8
2.1 Description
The Purdue Air Coil Test (PACT) facility, which is located at the Ray W. Herrick
Laboratories at Purdue University, was used to perform a variety of transient experiments
for the purpose of model validation. Figure 2.1 and Figure 2.2 illustrate schematically
the air-side and water-side arrangements of the PACT facility, respectively.
The air-side arrangement is a psychrometric wind tunnel that allows control of air
temperature, humidity, and flow rate to the inlet of the test section. Room air is drawn by
a variable-speed fan through a precooler and electric heater, and then passes through a
steam humidifier and air velocity measurement section that consists of an air straightener
and a Pitot tube array. After that, air enters the insulated test section with a test coil
between the upstream and downstream air mixers, thermocouple grids, and humidity
sensors. At the outlet, air is discharged to outdoors. Air inlet temperature and humidity
9
can be varied using the preconditioning equipment, while air flow rate can be changed by
adjusting the frequency of the variable-speed fan.
Test Coil
For the water-side arrangement, a tank is utilized to store chilled water coming
from the chiller and to supply chilled water to a test coil. Water temperature entering the
coil is controlled by mixing chilled water and building hot water through an electric
three-way mixing valve. Water flow rate can be adjusted using a variable-speed pump.
Thermocouples and a coriolis-type flow meter are installed for measurements.
2.2 Instrumentation
Air dry bulb temperatures upstream and downstream of the test section are
measured individually using nine thermocouples located on a grid, which is divided into
nine equal sections. Prior to each grid, air passes through an air mixer, which is used to
decrease or eliminate stratification in air stream either entering or leaving the test section.
10
The thermocouples are T type with a fine diameter of 0.0032×0.0254 m. The accuracy
after laboratory calibration is ±0.2 °C. The response time is 1 second in still air and
expected to be faster in moving air. The mean temperature from the nine thermocouples
is considered as air temperature, because the temperature distribution over the cross
section of the air duct is fairly uniform. The maximum difference among the nine
temperatures on each thermocouple grid was found to be less than 0.4 °C, which is
similar to the uncertainty of a single thermocouple.
Air humidities are measured using relative humidity sensors, which are mounted
beside the thermocouple grids behind the air stream mixers. The accuracy is ±1 % of the
measured value in the range of 0 to 90 % R.H., and ±2 % in the range of 90 to 100 %
R.H. The response time is 5 seconds in still air and expected to be faster than 5 seconds
in a convective flow. The humidity distribution over the cross section of the air duct was
investigated by measuring the humidities at the nine-thermocouple positions on each grid.
It was found that the variations of humidity at both upstream and downstream cross
sections were less than 2 % R.H., which means that the air humidity distribution is quite
uniform over the cross section. A single sensor is used at the inlet or outlet of the test
section for humidity measurement.
Air velocity is measured with a Pitot tube array, which consists of five Pitot tubes
installed evenly along the diagonal of the cross section of the air duct. The velocity
distribution was investigated by measuring the velocities on a uniform 5×5 matrix over
the cross section. The individual velocity varies within ±10 % of the average and the
distribution is considered to be uniform. The average velocity from the five Pitot tubes is
regarded as air velocity. The Pitot tube array is connected to a low differential pressure
transducer with a range of 0 to 0.00254 m H2O and an accuracy of ±0.25 % of the full
scale.
Water temperatures are measured by T type thermocouples that are submerged in
water flow at the inlet and outlet of a test coil. The accuracy of the thermocouples is
within ±0.2 ºC of the indicated value after calibration. The response time is 0.04 second
in still water. Water flow rate is determined by using a mass flow meter. The nominal
11
flow rate of the flow meter is 0 to 4.5 kg/s and the accuracy is ±0.5 % of the indicated
value.
It is believed that all the sensors can respond fast enough to capture transient
outlet changes associated with transient changes in inlet conditions of a test coil that can
be implemented within the facility.
A Hewlett Packard data acquisition system (Model 75000 Series B) is used to
collect data from the instrumentation. The sampling time interval is 1 second. A Dell
Pentium personal computer is used for further data analysis using the HP VEE (Visual
Engineering Environment) software.
Two different coils were tested in order to validate model performance. These
two coils represent two fin geometries of wavy and louvered fins, and two combinations
of coil depth and fin spacing that are 8 rows deep with 8 fins per inch and 4 rows deep
with 12 fins per inch. The arrangements of flow circuiting for the 8-row and 4-row coils
are illustrated schematically in Figure 2.3 and the real pictures are shown in Figure 2.4.
The characteristics of the coils are described in detail in Table 2.1.
12
8-row coil
4-row coil
Coil depth
0.264 0.132
(m)
Tube thickness
0.0004 0.0004
(m)
Fin thickness
0.0002 0.0002
(m)
15
A number of transient tests were performed where the coil inlet conditions were
varied and the responses of the outlet conditions were measured. The tests were run from
one steady-state condition to another. The inlet conditions include air flow rate, air inlet
temperature and humidity, and water flow rate and water inlet temperature. The outlet
conditions include air outlet temperature and humidity, and water outlet temperature.
The variations in inlet conditions include step change, ramp change, and change
associated with feedback control of air or water flow rate. The variations in air flow rate
were controlled by the variable-speed fan. The fan responded relatively fast, so a change
in air flow rate was achieved within 5 seconds. Similarly, the water flow rate was varied
by the variable-speed pump and a change occurred also within 5 seconds. The air inlet
temperature was controlled by the electric heater, which had heat capacitance itself, so
the temperature could not be changed abruptly. The water inlet temperature was adjusted
by mixing cold water with hot water through the electric 3-way mixing valve, which took
time to achieve a balance. For both temperatures, it took about 100 seconds to
accomplish a variation. The air inlet humidity was controlled by the steam humidifier
and a variation happened within 20 seconds. Table 2.2 summarizes the test conditions
briefly.
16
During each transient test, one of the inlet parameters was varied while others
were kept constant. The coils were tested under a range of air flow rates and water flow
rates, and conditions where the coils were fully dry, partially dry and partially wet, and
completely wet.
U R = ⎢∑ ⎜ ⋅ U Xi ⎟ ⎥ (2.1)
⎢⎣ i=1 ⎝ ∂X i ⎠ ⎥⎦
Measurement Uncertainty
Temperature
±0.2
(°C)
Humidity 0 - 90: ±1
(%) 90 - 100: ±2
Timer
±1
(s)
Scale
±0.1
(g)
18
*
Air temperature is measured by a nine-thermocouple grid. The uncertainty refers to that
for the average of nine thermocouple measurements.
20
[1]
TEST Calculation means the product of air mass flow rate and the difference of air
inlet and outlet humidity ratio.
[2]
Collected means the result of condensed water mass over the collection time interval.
For the heat transfer rate, the uncertainty analysis in Table 2.6 shows that the
water-side uncertainty is less than the air side uncertainty in most cases, which implies
that the heat transfer rate calculated from water-side measurement is more reliable than
that from the air side and can be considered as the total heat transfer rate of the coil under
steady-state conditions. Overall the heat balances and mass balances are less than 6 % in
most cases. The test data are reliable for the purpose of model validation.
Air-side and water-side convection heat transfer coefficients are needed for the
forward models. Water-side convection coefficient is determined using available
correlations based on the value of water-side Reynolds number, Rew. For laminar flow
(Rew ≤ 2300), the Sieder-Tate correlation is used; for fully turbulent flow (Rew ≥ 10,000),
the Dittus-Boelter equation is used; and for transition flow (2300 ≤ Rew ≤ 10,000), the
22
2,300 ≤ Re w ≤ 10,000, Nu w =
( f w 8) ⋅ ( Re w -1000 ) ⋅ Prw
(2.4)
1+12.7 ⋅ ( f w 8 ) ⋅ ( Prw 2 3 -1)
12
f w = ( 0.790 ⋅ ln ( Re w ) -1.64 )
-2
(2.5)
and
Re w ≥ 11,000, Nu w =0.023 ⋅ Re w 4 5 ⋅ Prw 0.4 (2.6)
where Nuw and Prw are the water-side Nusselt number and Prandtl number, respectively.
Air-side convection coefficient depends on various factors, such as fin and tube
geometries, tube circuiting arrangement, and air-side dry or wet surface condition. There
are some available correlations in the literature, but all of them have certain limitations
with respect to coil parameters and are not completely suitable for the test coils. For this
study, specific correlations are determined for each test coil using experimental data
under steady-state operating conditions.
Table 2.4 and Table 2.5 give steady-state experimental data for the 8-row coil
with wavy fins. With these data, air-side convection coefficients at different face
velocities were calculated by trial and error using the detailed steady-state coil model
developed in Chapter 4 for both dry and wet cases. Over the range of air velocities
considered in this study, a linear correlation works extremely well. The correlations for
the 8-row coil are
h a =10.41⋅ Va +19.43 (2.7)
for dry cases, and
h *a =10.05 ⋅ Va +24.32 (2.8)
for wet cases.
23
Here Va is the air-side face velocity in units of m/s, and ha and h*a are the air
convection coefficients without or with dehumidification and in units of W/m2·K. These
correlations match the experimental values perfectly (R2 = 0.999).
By the same approach, correlations for the 4-row coil with louvered fins were
obtained based on the steady-state data given in Table A.1 in Appendix A. For dry cases,
the expression is
h a =24.89 ⋅ Va +17.72 (2.9)
and for wet cases, it is
h *a =26.53 ⋅ Va +27.26 (2.10)
Figure 2.5 shows comparisons of air-side convection coefficient for the 8-row coil
and the 4-row coil for the experimental data along with the linear relations. The
uncertainties for ha of the 8-row coil are illustrated at the experimental data points, and
they are less than l3 %. The uncertainties for other convection coefficients are of the
same order.
100 120
80 100
ha (W/m -K)
ha (W/m -K)
80
60
2
2
60
40
*
40
20 20
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Va (m/s) Va (m/s)
8-row Exp. 4-row Exp. 8-row Exp. 4-row Exp.
8-row Linear 4-row Linear 8-row Linear 4-row Linear
Figure 2.5: Comparisons of air-side convection coefficient for an 8-row coil and a 4-row
coil.
24
Under both dry and wet cases, the convection coefficients for the 4-row coil are
higher than the 8-row coil. This is expected since the 4-row coil has 1.5 times fin density
of the 8-row coil, which results in higher local velocities around the tubes and fins. Also,
louvered fins have a stronger effect in breaking up the boundary layer and enhancing
turbulence compared with wavy fins.
considered for the 8-row coil and also describe the experimental data very well.
However, a linear relation is even better within a small range of air velocities. This has
been verified using data from Kays and London (1984).
25
3.1 Introduction
fin on tubes along the air flow direction within a counter cross-flow coil, and the air-side
heat and mass transfer is analogous for both geometries. The single straight finned tube
is illustrated in Figure 3.1 schematically.
Lf df
Tf
Air y dy Air
dx Tt Hf
x
Water Water Fin
Tw Tube
Air Air
It is a counter flow heat exchanger with a straight fin attached to a tube. Air flows
over the fin surface and water flows through the tube. The tube outer surface is assumed
to be insulated from the air to simplify the modeling. A two-dimensional heat conduction
problem within the fin is considered by neglecting the heat resistance along the fin
thickness direction, since the fin thickness, df, is rather small compared to the fin length,
Lf, and the fin height, Hf. Conduction within the fin in the flow direction was included in
this analysis for completeness and because it was relatively easy to include. However, it
is not an important effect.
Both a steady-state and dynamic model were developed for this simplified
geometry. The steady-state model is used to investigate the thickness of water
condensate film and evaluate its potential impact on the dynamics. The dynamic model
is used to evaluate the assumption of a Lewis number of unity and the use of fin
efficiency in dynamic coil modeling.
27
Assuming constant heat conductivity for the fin material, kf, the heat equation for
the fin control volume, dx×dy, from a steady-state energy balance is expressed as
∂ 2 Tf ∂ 2 Tf
kf ⋅ ⋅ df+k f ⋅ ⋅ df+q 'a-f =0 (3.1)
∂x 2
∂y 2
where Tf is the local fin temperature, and q’a-f is the heat transfer rate from the air flow
stream to the fin per unit length in the y direction.
Assuming adiabatic conditions along the fin left, right, and tip boundaries, and
neglecting the heat resistance of the tube attached to the fin base, the boundary conditions
are
∂Tf
x=0, =0 (3.2)
∂x
∂Tf
x=L f , =0 (3.3)
∂x
∂Tf
y=0, k f ⋅ ⋅ df+q 'w-t =0 (3.4)
∂y
∂Tf
y=H f , =0 (3.5)
∂y
where q 'w-t is the heat transfer rate from the water flow stream to the tube per unit length
in the x direction.
The air-side heat transfer, q 'a-f , is determined using the effectiveness-NTU
method with the assumption of uniform fin temperature over a control volume. When the
fin is dry and only heat transfer occurs, q 'a-f has the form of
' ⋅ ( T -T )
q 'a-f =ε a ⋅ C (3.6)
a a,in f
and
ε a =1-e-NTUa (3.7)
h a ⋅ dx
NTU a = ' (3.8)
C a
28
where Ta,in is the local air inlet temperature, ha is the convection heat transfer coefficient
' is the air heat capacitance rate per unit length in the y
between air and fin, and C a
direction.
The local air outlet temperature is calculated as
Ta,out =Ta,in -ε a ⋅ ( Ta,in -Tf ) (3.9)
When the fin temperature is below the dew point temperature of the inlet air
locally, dehumidification occurs and the fin is wet. Assuming the heat of vaporization of
water is constant within a small temperature range, then q 'a-f is the overall heat and mass
transfer rate and determined by (Kuehn et al., 1998)
h *a ⋅ dx
NTU*a = ' (3.12)
C a
where ha,in is the local air inlet enthalpy, hs,f is the saturation air enthalpy at Tf, h *a is the
' is the air
convection coefficient for air-side heat transfer under dehumidification, M a
mass flow rate per unit length in the y direction, and Le is the Lewis number.
The local air outlet enthalpy is determined as
q 'a-f
h a,out =h a,in - (3.13)
M'
a
The local air outlet temperature is still obtained from Equation (3.9) based on a heat
transfer analysis.
The heat transfer rate between water and tube, q 'w-t , is
the fin base temperature, and Pt is the tube inner perimeter. The energy equation
associated with the water flow stream is written as
⋅ ∂Tw -q ' =0
C (3.15)
w w-t
∂x
where Ċw is the water heat capacitance rate.
The two-dimensional heat conduction problem for steady state was discretized
using a second-order centered finite-difference method (Incropera and DeWitt, 1996).
The discretized equations were solved using a line-by-line Tri-Diagonal Matrix
Algorithm (TDMA) described by Murthy and Mathur (1998). The temperature
distribution within the fin can be obtained, together with the air temperature and
humidity, and water temperature distributions.
When dehumidification occurs, water condensate is retained on the fin and tube
surfaces, and the amount can be estimated using a steady-state model for steady-state
conditions. Water condensate has a heat capacitance and may have an effect on the
dynamic behavior of the finned tube. This effect was investigated initially by comparing
the heat capacitance associated with water condensate to the total heat capacitances of fin
and tube and water residing in the tube. If the heat capacitance of water condensate
accounts for a relatively small percentage of the overall heat capacitance at extremely wet
cases, then the effect of water condensate can be considered negligible. Otherwise, a
detailed transient analysis of energy transfer and mass transfer of water condensate would
be applied. The one exception would be the case where conditions change such that
retained water condensate would re-evaporate back into the air stream. This occurs
within direct expansion systems with on/off cycling of the compressor. However, the
current study is restricted to water coils that operate continuously, and where re-
evaporation rarely occurs.
In order to simplify the calculation of the amount of water condensate on the fin
and tube surfaces, film-wise condensation is assumed to prevail on the fin surface, and
30
drop-wise condensation is neglected. Also, the impact of the tube is neglected for the
distribution of water condensate on the fin surface. Other assumptions include: constant
viscosity and density for the water condensate film, negligible momentum and energy
transfers by advection in the condensate film, negligible shear stress at the liquid gas
interface, and laminar flow. These assumptions are conservative in that they would lead
to an over prediction of the condensate retained on the fin surface.
The schematic illustration is shown in Figure 3.2.
y y ωa(x,y*)
δ(x,y)
’a
M ’a
M
Fin Surface
Γ’(x,y)
Γ’(x,y)
dx
Fin length
0.3
(m)
Fin height
0.02
(m)
Fin thickness
0.0002
(m)
The parameters in Table 3.1 are representative of a single fin within the 8-row test
coil with wavy fins and were chosen so as to provide a good match to the steady-state and
32
dynamic performance for that coil. The ratio of air-side and water-side heat transfer
resistances of the straight finned tube is equal to that of the 8-row coil. The ratios of fin,
tube, and water heat capacitances of the straight finned tube are equal to those of the 8-
row coil. The 8-row coil can be considered to be composed of 8×24×16 of these straight
finned tubes, and the coil outlet conditions can be represented by those of the single
straight finned tube.
Table 3.2 describes the inlet conditions for the simulation results. The air-side
convection coefficients are 45.9 and 49.8 W/m2·K for dry and wet fin at the air mass flow
rate of 0.36 g/s, respectively, which are the same as the air-side convection coefficients of
the 8-row coil at a typical air flow rate of 1.11 kg/s. The water-side heat resistance per
unit length is 0.30 m·K/W at the water mass flow rate of 0.16 g/s, which is the same as
the water-side heat resistance of the 8-row coil at a typical water flow rate of 0.49 kg/s.
The tube geometry, which is not directly related to the model calculation, is not specified
here. Outlet conditions, along with the information about water condensate on the fin
surface, are given in Table 3.3 for the inlet conditions of Table 3.2
Table 3.3: Outlet conditions and corresponding information of water condensate for a
single straight finned tube.
In Table 3.3 MW.C. represents the mass of water condensate on the fin surface, and
CW.C. represents the percentage of heat capacitance associated with the water condensate
occupying the overall heat capacitances of the single straight finned tube. From case 1 to
case 5, the air inlet humidity increases from 30 % to 70 % with other identical inlet
conditions and the fin surface changes from fully dry to fully wet. At the wettest case of
air inlet humidity at 70 % in case 5, the heat capacitance associated with water
condensate accounts for 7.55 % of the overall heat capacitance. Thus, the energy storage
associated with the condensate layer is relatively small and its impact on the coil
dynamics would not be significant enough to justify the added complexity of including
transients associated with the condensate.
It should be noted that this example only considered one portion of a continuous
fin that would stretch across several tubes and extend from top to bottom of a real cooling
coil. However, the model was extended to estimate the condensate on the full-size 8-row
test coil and the capacitance of the water layer was estimated to be less than 8.3 % of the
total coil material and water capacitances. The results are summarized in Table 3.4.
34
Table 3.4: Outlet conditions and corresponding information of water condensate for an 8-
row coil.
Figure 3.3 depicts the film thickness at the air outlet as it grows along the fin
height direction compared to the spacing between fins for Case 5.
35
2 × 10-4 m 3 × 10-3 m
Condensate
Film
0.61 m
Fin
3.3 × 10-5 m
Figure 3.3: Schematic of condensate film thickness along fin height direction between
fins of an 8-row coil.
The film thickness is less than 2.2 % of the fin spacing, which can hardly affect
the air flow pattern. Also noted is that the model is conservative for estimating the film
thickness and the effects should be less than predicted. Therefore, it is believed that the
effect of water condensate on coil dynamics can be neglected under wet operating
conditions.
For the dynamic model, the energy storage associated with the fin, tube, and water
flow stream was considered, while the energy storage within the air was neglected. In
addition, the effect of water condensate retained on fin and tube surfaces was neglected
too. Under dynamic conditions, Equation (3.1) for the heat conduction within the fin is
modified by adding an energy storage term for the fin so that
36
∂Tf ⎛ ∂ 2T ∂ 2T ⎞
C"f ⋅ - ⎜ k f ⋅ 2f ⋅ df+k f ⋅ 2f ⋅ df+q 'a-f ⎟ =0 (3.18)
∂t ⎝ ∂x ∂y ⎠
where C”f is the fin heat capacitance per unit area of fin surface.
Considering the energy storage of the tube, one of the boundary conditions,
Equation (3.4) is modified accordingly as
∂Tt ⎛ ∂T ⎞
y=0, C't ⋅ - ⎜ k f ⋅ f ⋅ df+q 'w-t ⎟ =0 (3.19)
∂t ⎝ ∂y ⎠
where C’t is the tube heat capacitance per unit length in the x direction. The other three
boundary conditions remain unchanged.
The air and water-side heat transfer rates, q’a-f and q’w-t, are calculated using the
same approach as for the steady-state model. The transient energy equation for the water
flow stream is rewritten as
∂Tw ⎛ ∂T ⎞
C'w ⋅ - ⎜ C w ⋅ w -q 'w-t ⎟ =0 (3.20)
∂t ⎝ ∂x ⎠
by adding the energy storage of the water flow stream. Here, C’w is the water heat
capacitance per unit length in the x direction.
Similar to the steady-state model, the dynamic model was discretized using a
second-order centered finite-difference method combined with an implicit integration
algorithm (Incropera and DeWitt, 1996). The discretized equations were also solved
again using the line-by-line Tri-Diagonal Matrix Algorithm (TDMA). This model is used
as a baseline to evaluate other modeling assumptions in the next sections.
Lewis number is an important modeling parameter for a heat and mass transfer
process, which represents the Lewis relation of
h *a
Le= (3.21)
h D ⋅ Cpa
37
where hD is the diffusion coefficient for air-side water vapor transfer, and Cpa is the air
specific heat. For turbulent air flow, the value of Le is approximately 1.0; for laminar
flow, it can be calculated as
23
⎛α⎞
Le= ⎜ ⎟ (3.22)
⎝D⎠
where α is the thermal diffusivity and D is the water vapor diffusivity in air.
Thermal and water vapor diffusivity data for dry and saturated moist air, which
are used to estimate the value of Lewis number for laminar flow, are summarized in
Table 3.5 (Kuehn et al., 1998).
Table 3.5: Thermal and vapor diffusivity data for dry and saturated moist air.
Temperature Degree of α D
α/D Le
(°C) Saturation (m2/s × 10-5) (m2/s × 10-5)
From Table 3.5, it is seen that α/D varies in a small range of 0.855 - 0.85 for dry
air, and 0.854 - 0.843 for saturated moist air at temperatures within 10 ºC and 37.8 ºC.
Based on the correlation of Le and α/D in Equation (3.22), Lewis number is
approximately 0.9 for all representative laminar flow operating conditions for cooling
38
coils, which is reasonably close to 1.0. If Lewis number could be assumed as unity, the
air to fin energy transfer can be simplified significantly as
' ⋅ ( h -h )
q 'a-f =ε*a ⋅ M (3.23)
a a,in s,f
The dynamic model for the simplified geometry described in Table 3.1 was used
to investigate differences in model predictions for Lewis numbers of 0.9 and 1.0. Five
transient conditions were simulated for the inlet conditions given in Table 3.6. All the
inlet variables experienced step changes from one value to another.
Figure 3.4 gives comparison results for transient responses of the air outlet
averaged temperature and humidity, and water outlet temperature to the step change of air
flow rate (Case 1) as an example. The table along with the plots summarizes the outlet
parameters at the initial and final steady-state conditions. The air-side convection
39
coefficients were 32.7 and 37.1 W/m2·K for a dry and wet fin, respectively, at an air mass
flow rate of 0.18 g/s.
40
15.0 10.5
14.5
10.0
14.0
T a,out (deg C) 13.5 9.5
ω a,out (g/kg)
13.0 9.0
12.5
12.0 8.5
11.5 8.0
11.0
7.5
10.5
10.0 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Le = 1.0 Le = 0.9 Le = 1.0 Le = 0.9
15.5
15.0
T w,out (deg C)
14.5
14.0
13.5
13.0
12.5
0 25 50 75 100
time (s)
Le = 1.0 Le = 0.9
Figure 3.4: Transient comparison results for Le = 1.0 and Le = 0.9 a step change in air
flow rate.
41
From these comparisons, the model predictions of air outlet temperature are very
close for Le = 0.9 and Le = 1.0 during the transient period and at steady states. The air
outlet humidity and water outlet temperature show similar trends. Other transient cases
in Table 3.6 were checked and confirmed this conclusion. A Lewis number of unity is a
reasonable assumption and is adopted throughout the rest of this work.
∂Tw ⎛ ∂Tw ⎞
C'w ⋅ -⎜ C w ⋅ ⋅ dx-q 'w-t ⎟ = 0 (3.26)
∂t ⎝ ∂x ⎠
Here, for a dry fin, q 'a-f is calculated as
42
' ⋅ ( T -T )
q 'a-f =ε a ⋅ C (3.27)
a a,in t
with
ηf ⋅ h a ⋅ dx
NTU a = (3.28)
'
C a
where ηf is the heat transfer only fin efficiency. The effectiveness for dry heat transfer,
εa, is determined with Equation (3.7).
For a wet fin, q 'a-f is determined by
' ⋅ ( h -h )
q 'a-f =ε*a ⋅ M (3.29)
a a,in s,t
with
η*f ⋅ h *a ⋅ dx
NTU*a = ' (3.30)
C a
where η*f is the fin efficiency for heat and mass transfer. The effectiveness for heat and
and the local air outlet enthalpy under wet conditions is expressed as
h a,out =h a,in -ε*a ⋅ ( h a,in -h s,t ) (3.32)
11.5 13.0
11.0
12.5
10.5
10.0 12.0
T a,out (deg C)
T w,out (deg C)
9.5 11.5
9.0
11.0
8.5
8.0 10.5
7.5
10.0
7.0
6.5 9.5
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Simplified Model Distributed Model Simplified Model Distributed Model
Simplified
11.08 6.98 12.71 9.66
Model
Distributed
11.12 7.01 12.69 9.66
Model
Figure 3.5: Transient comparison results for simplified model and distributed model for a
dry fin.
44
15 10.5
14 10.0
T a,out (deg C) 9.5
ω a,out (g/kg)
13
9.0
12
8.5
11
8.0
10 7.5
9 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Simplified Model Distributed Model Simplified Model Distributed Model
15.5
15.0
T w,out (deg C)
14.5
14.0
13.5
13.0
12.5
0 25 50 75 100
time (s)
Simplified Model Distributed Model
Simplified
13.99 9.48 10.31 7.62 15.13 12.73
Model
Distributed
14.71 10.13 10.11 7.51 15 12.64
Model
Figure 3.6: Transient comparison results for simplified model and distributed model for a
wet fin.
45
It is seen from Figure 3.5 that the predictions of transient responses of the air and
water outlet temperatures are fairly close for the simplified fin efficiency approach and
the distributed model. Since there is no dehumidification, the air outlet humidity ratio is
the same as the inlet humidity ratio and the results are not shown here. The steady-state
results in the attached table also match each other almost exactly. This was expected
since the fin efficiency results from an analytical solution to the fin heat transfer problem
at steady state. Cases other than air flow rate change were considered for dry fins and
gave similar results. Therefore, for dry fins, the fin efficiency approach works well
within the transient model.
Figure 3.6 gives results for a wet fin. Here, the predictions of transient responses
of the air outlet humidity and water outlet temperature are also very close for the
simplified fin efficiency approach and the distributed model. However, there is a
relatively large discrepancy in air outlet temperature between the two approaches.
Similar results were observed for other wet operating conditions. This issue was
investigated and the discrepancy was found to be a result of the method that is used to
calculate the heat transfer fin efficiency for wet fins. An improved method is developed
in the next section.
Usually a heat transfer fin efficiency is applied to the calculation of sensible heat
transfer regardless of whether condensation occurs or not. However, it was found that a
relatively large discrepancy could be caused by using the heat transfer only fin efficiency
developed for a dry fin to calculate sensible heat transfer for a wet fin. The heat transfer
only fin efficiency is derived based on the temperature distribution for a dry fin, but when
dehumidification occurs, the temperature distribution on a wet fin is different from that of
a dry fin.
A modified fin efficiency for heat transfer under dehumidification was developed
starting from the derivations of heat transfer only and heat and mass transfer fin
efficiencies for a classic straight fin depicted in Figure 3.7.
46
Tf Tf
y hs,f y
Hf Hf
Water layer
Tb Tb, hs,b
with
h a ⋅ Pc
m= (3.35)
k f ⋅ Ac
where Tb is the fin base temperature. By definition, the fin efficiency for heat transfer
only is determined as
47
tanh ( m ⋅ H f )
ηf = (3.36)
m ⋅ Hf
When the fin is wet, heat and mass transfer occurs simultaneously. For a Lewis
number of unity, the energy balance is modified as
d 2 Tf h *a
kf ⋅ ⋅ A c + ⋅ ( h a,in -h s,f ) ⋅ Pc =0 (3.37)
dy 2 Cpa
In order to simplify this equation, a saturation specific heat of air (Braun, et al.,
1989) is defined by
dh s,f
Cs = (3.38)
dTf
It is reasonable to assume Cs as a constant within a small temperature range of a
fin, and it is often evaluated at the fin base temperature. By eliminating Tf, Equation
(3.37) could be simplified as
2
k f d h s,f h *a
⋅ ⋅ A c + ⋅ ( h a,in -h s,f ) ⋅ Pc =0 (3.39)
Cs dy 2 Cpa
By analogy to the solution for fin heat transfer, the profile of hs,f is given by
with
h *a Cp,a ⋅ Pc
m* = (3.41)
k f Cs ⋅ A c
where hs,b is the saturated air enthalpy at the fin base temperature, Tb. The fin efficiency
for heat and mass transfer together can be calculated as
*
tanh ( m* ⋅ H f )
η =
f (3.42)
m* ⋅ H f
which is of identical form to the heat transfer only fin efficiency in Equation (3.36) with
modified parameters.
48
However, a temperature profile for a wet fin is not identical to that for a dry fin.
The profile of Tf depends on the profile of hs,f through the relationship of Cs, which can
be approximated as
h s,f ( y ) -h s,b
Cs = (3.43)
Tf ( y ) -Tb
Tf ( y ) -Ta,in
=
h s,f ( y ) -h a,in
⋅
(h s,b -h a,in )
-⎜
⎛ ( h s,b -h a,in ) ⎞
-1⎟ (3.44)
Tb -Ta,in h s,b -h a,in Cs ⋅ ( Tb -Ta,in ) ⎜⎝ Cs ⋅ ( Tb -Ta,in ) ⎟⎠
CFη =
(h s,b -h a,in )
(3.45)
Cs ⋅ ( Tb -Ta,in )
Tf ( y ) -Ta,in cosh ( m* ⋅ ( H f -y ) )
= ⋅ CFη - ( CFη -1) (3.46)
Tb -Ta,in cosh ( m* ⋅ H f )
Finally, the heat transfer fin efficiency under dehumidification is determined from the
heat and mass transfer fin efficiency and the correction factor according to
ηfM =η*f ⋅ CFη - ( CFη -1) (3.47)
Apparently, the introduction of a constant Cs along the fin height direction causes
some error in the overall heat and mass transfer. This results in a small difference in
water outlet temperature between the simplified fin efficiency model and the distributed
analysis shown in Figure 3.6. However, the error is very small and considered to be
acceptable. On the other hand, the difference in air outlet temperature between the two
methods is significant, and mainly caused by the heat transfer fin efficiency. In the fin
efficiency approach, ηf in Equation (3.36) was used for air temperature calculation when
the fin is wet. By using the modified fin efficiency ηfM given in Equation (3.47), the
results for air outlet temperature under Case 2 in Table 3.7 are improved significantly,
which are shown in Figure 3.8.
49
15 10.5
14 10.0
T a,out (deg C) 9.5
ω a,out (g/kg)
13
9.0
12
8.5
11
8.0
10 7.5
9 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Simplified Model Distributed Model Simplified Model Distributed Model
Simplified
14.69 10.01 10.03 7.41 15.13 12.73
Model
Distributed
14.71 10.13 10.11 7.51 15 12.64
Model
Figure 3.8: Improved transient results for simplified model for a wet fin.
Other cases for wet fins were investigated and gave similar results. It can be
concluded that the fin efficiency approach works well for both dry and wet fins under
transient operating conditions.
50
Chilled water cooling coils are usually counter cross-flow finned tube heat
exchangers with various arrangements. Figure 4.1 shows one classical full-row
serpentine circuiting arrangement. A single cross-flow finned tube can be considered as a
basic element of a coil. The detailed forward model was developed starting from a
control volume on a portion of a single finned tube.
51
Ta,in , h a,in , M a
Control Volume
Tw, out , M w
Tw,in , M w
Ta, out , h a, out , M a
In this geometry, air flows over a finned tube and is cooled and possibly
dehumidified due to contact with cold surfaces. The dimension x is measured in the
direction of water flow, whereas the dimension y is measured in the direction of air flow,
which is normal to the water flow direction. The model includes the following
assumptions: water is incompressible, ideal gas mixture for air and water vapor, constant
densities and specific heats for air, water, and fin and tube material, negligible conduction
in air and water flow directions for both fluids, negligible conduction in the x direction
for fin and tube material, water and air velocities are uniform along both flow directions,
negligible energy storage within air, a Lewis number of unity for wet coils, negligible
52
effect of water condensate, and the temperature profile in the y direction (within water,
tube and fin material) follows the steady-state profile.
The steady-state forward model was developed first. For a dry coil without
dehumidification, the governing equations for a control volume are given by
⋅ dTw + 1 ⋅ ( T -T ) =0
C w
dx R 'w
w c (4.1)
1 1
' ( c
⋅ T -Ta,in ) + ' ⋅ ( Tc -Tw ) =0 (4.2)
Ra Rw
where Tw and Tc are the local “bulk” temperatures of water and coil material, Ta,in is the
air inlet temperature locally, Ċw is the heat capacitance rate associated with the water
flow stream, and R 'w and R 'a are the heat resistances for heat transfer per unit length in
the x direction between water and coil material, and coil material and air, respectively.
An approach that is consistent with the use of air inlet temperature in Equation
(4.2) is to use the effectiveness-NTU method for the air-side heat transfer calculation.
With this approach, the heat resistances become
1
R 'w = (4.3)
h w ⋅ Pt
and
1
R 'a = ' (4.4)
εa ⋅ C a
where hw is the convection heat transfer coefficient for water flow through the tube, Pt is
' is the heat
the tube inner perimeter, εa is the air-side heat transfer effectiveness, and C a
capacitance rate of the air flow stream per unit length in the x direction. Furthermore, the
air-side heat transfer effectiveness is expressed as
ε a =1-e-NTUa (4.5)
where the air-side NTUa is calculated by
ηa ⋅ h a ⋅ A 'a
NTU a = ' (4.6)
C a
53
and where ηa is the overall fin efficiency for heat transfer only, ha is the convection
coefficient for air-side heat transfer, and A 'a is the air-side surface area per unit length in
the x direction.
The local outlet air temperature can be determined by
Ta,out =Ta,in +ε a ⋅ ( Tc -Ta,in ) (4.7)
which serves as the local inlet air temperature for the corresponding control volume on
the next row, if there is any in the air flow direction. The average temperature for all the
control volumes on the last row is considered as the air outlet temperature for the coil
When a surface is wet, the driving potential for heat and mass transfer is primarily
the difference between air enthalpy and saturation air enthalpy at coil surface
temperature. With the assumption of a Lewis Number of unity, Equation (4.2) is
replaced for the wet case by
1 1
*' ( s,c
⋅ h -h a,in ) + ' ⋅ ( Tc -Tw ) =0 (4.8)
Ra Rw
while Equation (4.1) remains unchanged. In Equation (4.8), ha,in is the air inlet enthalpy
locally, hs,c is the saturation air enthalpy at Tc, and R *'a is the resistance for heat and mass
transfer per unit length in the x direction between coil material and air, which can be
expressed as
1
R *'a = (4.9)
ε ⋅M
* '
a a
where in this case ε*a is an effectiveness for combined heat and mass transfer determined
from
*
ε*a =1-e-NTUa (4.10)
' is the air mass flow rate per unit length in the x direction, and
and where M a
η*a ⋅ h *a ⋅ A 'a
NTU*a = ' (4.11)
C a
Here η*a is the overall fin efficiency for both air-side heat and mass transfer (different
from ηa, which is for heat transfer only), and h*a is the convection coefficient for air-side
CFη =
(h s,c -h a,in )
(4.13)
Cs,c ⋅ ( Tc -Ta,in )
dTc 1 1
C'c ⋅ + ⋅ ( Tc -Ta,in ) + ' ⋅ ( Tc -Tw ) =0 (4.15)
dt R 'a Rw
For the wet case, with negligible effect of water condensate on coil outer surfaces,
Equation (4.15) is replaced by
dTc 1 1
C'c ⋅ + *' ⋅ ( h s,c -h a,in ) + ' ⋅ ( Tc -Tw ) =0 (4.16)
dt R a Rw
55
where C'w and C'c are the heat capacitances of water and coil material per unit length in
the x direction.
Since energy storage within air is neglected and a steady-state energy balance is
applicable, Equation (4.7) and Equation (4.12) can still be used to find the local air outlet
temperature and enthalpy. The local air inlet condition is determined in the same way for
the steady-state model.
1 1
⋅ ( Tc -Ta,in ) + ⋅ ( Tc -Tw ) =0 (4.18)
Ra Rw
where Tw,in is the local water inlet temperature, and Rw and Ra are the heat resistances for
heat transfer between water and coil material, and coil material and air of the control
volume.
The water-side heat resistance, Rw , in this case can be calculated by
N
Rw = (4.19)
h w ⋅ A w,tot
where Aw,tot is the water-side total surface area of heat transfer, and N is the arbitrary
number of control volumes for the coil.
56
where Ċa is the heat capacitance rate associated with air flow stream of the coil. Equation
(4.5) is still used to determine εa with NTUa obtained by
ηa ⋅ h a ⋅ A a,tot
NTU a = (4.21)
C ⋅N
a
where R *a is the resistance for heat and mass transfer between coil material and air of the
control volume, which can be expressed by
1
R *a = (4.23)
ε ⋅M
*
a a
is the air mass flow rate of the coil. To obtain ε* , Equation (4.10) is used
where M a a
with
η*a ⋅ h *a ⋅ A a,tot
NTU*a = (4.24)
C ⋅N
a
The dynamic model was developed based on the steady-state model by adding
energy storage terms associated with water residing in the tubes and coil metal material.
The governing equations are rewritten as
dTw 1
Cw ⋅ +Cw ⋅ ( Tw -Tw,in ) + ⋅ ( Tw -Tc ) =0 (4.25)
dt Rw
dTc 1 1
Cc ⋅ + ⋅ ( Tc -Ta,in ) + ⋅ ( Tc -Tw ) =0 (4.26)
dt R a Rw
for the dry case, and for the wet case, Equation (4.26) needs to be replaced by
dTc 1 1
Cc ⋅ + * ⋅ ( h s,c -h a,in ) + ⋅ ( Tc -Tw ) =0 (4.27)
dt R a Rw
57
where Cw and Cc are the heat capacitances of water and coil material for the control
volume. Cw and Cc can be determined by
Cw,tot
Cw = (4.28)
N
and
Cc,tot
Cc = (4.29)
N
where Cw,tot and Cc,tot are the total heat capacitances of water and metal material of the
coil,
To determine the local air outlet parameters, the counter-flow forward model
adopts the approach used in the detailed forward model, which is described in the last
section.
One way to reduce the number of required control volumes and speed up the
computation is to assume a steady-state water temperature profile within each control
volume instead of uniform temperature distribution. With this assumption, the water-side
heat transfer rate can be evaluated by the effectiveness-NTU method.
Based on the counter-flow model, under steady-state conditions, the governing
equations of (4.17) and (4.18) for the dry case are modified as
⋅ ( T -T ) + 1 ⋅ ( T -T ) =0
C w w,out w,in
Rw
w,in c (4.30)
1 1
⋅ ( Tc -Ta,in ) + ⋅ ( Tc -Tw,in ) =0 (4.31)
Ra Rw
and where εw is the water-side heat transfer effectiveness of a control volume expressed
as
ε w =1-e-NTU w (4.34)
and
h w ⋅ A w,tot
NTU w = (4.35)
⋅N
C w
Under dynamic conditions, the rate of water temperature change within a control volume
is assumed to be represented by that at the outlet of the same control volume or
dTw dTw,out
= (4.36)
dt dt
where Tw,out is the local water outlet temperature. The transient governing equations are
updated as
dTw,out ⋅ ( T -T ) + 1 ⋅ ( T -T ) =0
Cw ⋅ +C w w,out w,in w,in c (4.37)
dt Rw
dTc 1 1
Cc ⋅ + ⋅ ( Tc -Ta,in ) + ⋅ ( Tc -Tw,in ) =0 (4.38)
dt R a Rw
for the dry case and for the wet case, Equation (4.38) is replaced with
dTc 1 1
Cc ⋅ + * ⋅ ( h s,c -h a,in ) + ⋅ ( Tc -Tw,in ) =0 (4.39)
dt R a Rw
The air-side modeling approach remains the same as described in the last sections.
A numerical technique was utilized to solve the detailed forward model to get the
spatial and time distributions of air outlet temperature and humidity, and water outlet
temperature of a coil. In order to handle partially dry and partially wet coils, it is most
appropriate to discretize the governing differential equations spatially. For each control
volume, it is necessary to evaluate whether moisture condenses or not. This is
accomplished by comparing the value of coil material temperature to the dew point
59
temperature of inlet air locally. If the coil material temperature is below the dew point
temperature, dehumidification occurs and the heat and mass transfer equations apply;
otherwise the heat transfer equations apply.
The steady-state model is considered first. Applying the upwind differencing
scheme in space, the governing equations of (4.1) is discretized as
⋅(
Tw -Tw,in ) 1
C w + ⋅ ( Tw -Tc ) =0 (4.40)
∆x R 'w
where ∆x is the space step along the water flow direction. Along with Equation (4.2) for
a dry control volume, the two equations can be solved directly to get Tw and Tc. For a
wet control volume, Equation (4.8) is applied, where Tc and hs,c are coupled. Newton’s
method is utilized to solve this equation together with Equation (4.40).
There are two boundary conditions for the model. One is the air inlet temperature
and humidity, or enthalpy; the other is the water inlet temperature. For multiple row
coils, the overall coil model involves using the coil circuiting and the tube element model
to determine the air inlet and water inlet states for other tube elements. Gauss-Seidel
iteration along the x direction is used to solve the resulting model for counter cross-flow
coils. The air outlet states for each control volume are set to certain guess values
initially, so the air inlet states are known based on coil circuiting. Then, the dry analysis
along the x direction is applied to predict the coil material temperature for each control
volume. This coil temperature is compared with the dew point temperature of inlet air
locally to determine if the dry analysis is valid or condensation occurs and the wet
analysis should be utilized instead. However, the wet analysis could also be used to
predict the coil temperature for evaluation of dry or wet conditions. Both approaches
were compared and gave very close results. The dry analysis is adopted since it requires
less calculation than the wet analysis. After this step, all the local outlet conditions are
calculated and the air-side outlet states are checked for convergence using a relative
tolerance. If the maximum of the relative differences between the new values and the
initial guess values for all the control volumes is less than the tolerance, then the iteration
is stopped and the coil states are output to a file. Otherwise, the guess values of the air
60
inlet states are updated by the new values and the above procedure is repeated until
converged. The solution procedure is depicted by the flow chart shown in Figure 4.2.
61
Parameters
Dry Analysis
Is temperature of
Yes
Update:
Air outlet states
No
Wet Analysis
Calculate:
Local outlet conditions
Is convergence
No tolerance satisfied?
Yes
Output:
Outlet conditions, coil states
Stop
62
It should be noted that the boundary for the onset of condensation on coil outer
surfaces can not be identified before the model is solved. Therefore, it is difficult to
utilize direct solution techniques to solve the governing equations of all the control
volumes simultaneously, since the basic equations are different for dry and wet control
volumes. Instead, the iterative approach is more appropriate and efficient.
The solution of the dynamic model is presented next. Based on the steady-state
discretization, the implicit time differencing scheme is used. The governing differential
equations of (4.14), (4.15), and (4.16) become
C '
⋅
(T w -Tw0 )
⋅(
+C
Tw -Tw,in )
+
1
⋅ ( Tw -Tc ) =0 (4.41)
w w
∆t ∆x R 'w
( T -T ) + 0
1 1
' ( c
⋅ T -Ta,in ) + ' ⋅ ( Tc -Tw ) =0
c c
C '
c ⋅ (4.42)
∆t Ra Rw
and
( T -T ) + 0
1 1
*' ( s,c
⋅ h -h a,in ) + ' ⋅ ( Tc -Tw ) =0
c c
C '
c ⋅ (4.43)
∆t Ra Rw
where ∆t is the time step, and the superscript 0 represents the value at previous time.
First, the steady-state model is used to provide the initial coil states for the initial
inlet conditions. After that, at each time step, the solution techniques for the dynamic
model are basically the same as for the steady-state model. The calculation continues
until the end of the dynamic process. The solution procedure is described in Figure 4.3.
63
Parameters
Dry Analysis
Is temperature of Yes
Update: No
Air outlet states
Wet Analysis
Calculate:
Local outlet conditions
Is convergence
No tolerance satisfied?
Yes
Output:
Outlet conditions, coil states
Yes
Stop
64
For the counter-flow model, the steady-state governing equations (4.17), (4.18),
and (4.22) are already in a discretized format. In the dynamic model, to discretize the
energy storage terms, the implicit differencing scheme applies to the governing equations
(4.25), (4.26), and (4.27), leading to
Cw ⋅
(T w -Tw0 )
⋅ ( T -T ) + 1 ⋅ ( T -T ) =0
+C (4.44)
w w w,in w c
∆t Rw
Cc ⋅
( T -T ) +
c
0
c 1
⋅ ( Tc -Ta,in ) +
1
⋅ ( Tc -Tw ) =0 (4.45)
∆t Ra Rw
and
( T -T ) + 0
1 1
* ( s,c
⋅ h -h a,in ) + ⋅ ( Tc -Tw ) =0
c c
Cc ⋅ (4.46)
∆t Ra Rw
The main difference between the counter-flow model and the detailed model is
the description of coil circuiting and thus the bookkeeping for local air inlet conditions.
Once the coil models are created, the structures of governing equations are basically the
same. So the solution techniques for the detailed model are inherited to solve the
counter-flow model.
Similarly, for the water-side ε-NTU model, the governing equations (4.30),
(4.31), and (4.32) describing the steady-state model are ready to be solved. The
governing equations (4.37), (4.38), and (4.39) of the dynamic model can be discretized as
Cw ⋅
(T w,out
0
-Tw,out ) ⋅ ( T -T ) + 1 ⋅ ( T -T ) =0
+C (4.47)
w w,out w,in w,in c
∆t Rw
Cc ⋅
( T -T ) +
c
0
c 1
⋅ ( Tc -Ta,in ) +
1
⋅ ( Tc -Tw,in ) =0 (4.48)
∆t Ra Rw
and
65
( T -T ) + 0
1 1
* ( s,c
⋅ h -h a,in ) + ⋅ ( Tc -Tw,in ) =0
c c
Cc ⋅ (4.49)
∆t Ra Rw
by applying the implicit scheme. Again, the solution procedures follow those for the
detailed model.
For this model, the explicit time differencing scheme was also investigated. The
discretized governing equations for the dynamic model become
Cw ⋅
(T w,out
0
-Tw,out ) ⋅ ( T 0 -T 0 ) + 1 ⋅ ( T 0 -T 0 ) =0
+C (4.50)
w w,out w,in w,in c
∆t Rw
( T -T ) + 0
1
⋅ ( Tc0 -Ta,in
1
Cc ⋅
c
∆t
c
Ra
0
) +
Rw
⋅ ( Tc0 -Tw,in
0
) =0 (4.51)
and
( T -T ) + 0
1 1
* ( s,c ) ⋅ ( Tc0 -Tw,in ) =0
c c
Cc ⋅ ⋅ h 0 -h a,in
0
+ 0
(4.52)
∆t Ra Rw
So far three forward models from detailed to simplified have been developed.
Numerical experiments were performed to identify the most appropriate dynamic model
for cooling coils. Numerical study was first applied to the detailed model. The main
purpose was to investigate the effects of different values of N, number of control
volumes, and ∆t, time step, on the accuracy of model predictions and determine the
appropriate combination of N and ∆t to get desired accuracy with minimum
computational time. Then, the results of the counter-flow model were compared with
those of the detailed model to identify if the counter-flow assumption was valid. And the
66
water-side ε-NTU model was studied to verify its accuracy and potential for reducing the
number of control volumes.
Several representative dynamic operating conditions were utilized for the
following numerical studies. They include step changes of air flow rate and water flow
rate, and step changes of air inlet temperature and humidity, and water inlet temperature.
The transient results illustrated in this section are under step changes of air flow rate for
dry and wet coils. The inlet conditions are given in Table 4.1.
The detailed forward model was applied to the 8-row test coil with its physical
configuration described in Figure 2.3 and Table 2.1. Four different values of N were
investigated, which are 8, 40, 80, and 160 (1, 5, 10, and 20 nodes for each row), at a
uniform ∆t of 1 second that is proved to be an appropriate value in the next part.
Comparison results for coil outlet air temperature and humidity, and water temperature
are shown in Figure 4.4 and Figure 4.5 for both dry and wet cases, respectively.
67
13 14
12
13
T a,out (deg C)
T w,out (deg C)
11
12
10
11
9
8 10
7 9
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
N=8 N = 40 N = 80 N = 160 N=8 N = 40 N = 80 N = 160
Figure 4.4: Transient comparison results for detailed model at different values of N for an
8-row dry coil (∆t = 1 second).
68
16 11.0
10.5
15
T a,out (deg C) 10.0
14
Wa,out (g/kg)
9.5
13 9.0
8.5
12
8.0
11
7.5
10 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
N=8 N = 40 N = 80 N = 160 N=8 N = 40 N = 80 N = 160
16.0
15.5
T w,out (deg C)
15.0
14.5
14.0
13.5
13.0
0 25 50 75 100
time (s)
N=8 N = 40 N = 80 N = 160
Figure 4.5: Transient comparison results for detailed model at different values of N for an
8-row wet coil (∆t = 1 second).
69
Both Figure 4.4 and Figure 4.5 show that there exist differences between the
simulation results using 8 nodes and 40 nodes. The differences in water-side outlet
temperature are greater for the wet coil, because the model with more control volumes
can more accurately determine the transition from dry to wet conditions. The results for
80 and 160 nodes are almost identical to those for 40 nodes.
The effect of time step on transient simulation results corresponding to the coil
inlet conditions given in Table 4.1 are shown in Figure 4.6 and Figure 4.7 for N = 80.
13 14
12
13
T a,out (deg C)
T w,out (deg C)
11
12
10
11
9
8 10
7 9
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
∆t = 0.1 ∆t = 1 ∆t = 2 ∆t = 5 ∆t = 0.1 ∆t = 1 ∆t = 2 ∆t = 5
Figure 4.6: Transient comparison results for detailed model at different values of ∆t for
an 8-row dry coil (N = 80).
70
16 11.0
10.5
15
T a,out (deg C) 10.0
14
Wa,out (g/kg)
9.5
13 9.0
8.5
12
8.0
11
7.5
10 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
∆t = 0.1 ∆t = 1 ∆t = 2 ∆t = 5 ∆t = 0.1 ∆t = 1 ∆t = 2 ∆t = 5
16.0
15.5
T w,out (deg C)
15.0
14.5
14.0
13.5
13.0
0 25 50 75 100
time (s)
∆t = 0.1 ∆t = 1 ∆t = 2 ∆t = 5
Figure 4.7: Transient comparison results for detailed model at different values of ∆t for
an 8-row wet coil (N = 80).
71
A 5-second time step gives a slower response than a 2-second time step and gives
less accurate transient results. But decreasing the time step further to 1 or 0.1 second
does not give much improvement. The detailed model using 80 nodes and a 1-second
time step is considered as a benchmark to investigate the accuracy of other simplified
models.
The prediction results obtained from the counter-flow model are compared to
those from the detailed model according to the inlet conditions in Table 4.1, as shown in
Figure 4.8 and Figure 4.9, respectively. 80 control volumes and a 1-second time step
were utilized for both models.
72
13 14
12 13
T w,out (deg C)
T a,out (deg C)
11
12
10
11
9
8 10
7 9
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (80CVs) Counter-flow (80CVs) Detailed (80CVs) Counter-flow (80CVs)
Detailed
7.47 11.73 10.07 13.23
Model
Counter-flow
7.5 11.75 10.06 13.18
Model
Figure 4.8: Transient comparison results for detailed model and counter-flow model for
an 8-row dry coil.
73
16 11.0
10.5
15
T a,out (deg C) 10.0
Wa,out (g/kg)
14
9.5
13 9.0
8.5
12
8.0
11
7.5
10 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (80CVs) Counter-flow (80CVs) Detailed (80CVs) Counter-flow (80CVs)
16.0
15.5
T w,out (deg C)
15.0
14.5
14.0
13.5
13.0
0 25 50 75 100
time (s)
Detailed (80CVs) Counter-flow (80CVs)
Detailed
10.65 15.23 7.71 10.36 13.21 15.5
Model
Counter-flow
10.68 15.23 7.74 10.35 13.18 15.48
Model
Figure 4.9: Transient comparison results for detailed model and counter-flow model for
an 8-row wet coil.
74
It is seen from Figure 4.8 and Figure 4.9 that the predictions of transient responses
of the coil outlet parameters are very close for the detailed model and the counter-flow
model for the same number of control volumes. Therefore, the counter-flow assumption
is a valid approximation for a counter cross-flow coil with 8 rows. Comparison results
for the 4-row coil were also obtained. Two sample results are shown in Figure 4.10 and
Figure 4.11. The agreement is slightly worse than for the 8-row coil but the counter-flow
model is still appropriate.
15 13.0
12.5
14
12.0
T w,out (deg C)
T a,out (deg C)
13
11.5
12 11.0
10.5
11
10.0
10
9.5
9 9.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (40CVs) Counter-flow (40CVs) Detailed (40CVs) Counter-flow (40CVs)
Detailed
9.42 13.76 9.5 12.05
Model
Counter-flow
9.5 13.84 9.48 11.96
Model
Figure 4.10: Transient comparison results for detailed model and counter-flow model for
a 4-row dry coil.
75
18 12.0
11.5
17
11.0
T a,out (deg C)
Wa,out (g/kg)
16
10.5
15 10.0
9.5
14
9.0
13
8.5
12 8.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (40CVs) Counter-flow (40CVs) Detailed (40CVs) Counter-flow (40CVs)
14.0
13.5
T w,out (deg C)
13.0
12.5
12.0
11.5
11.0
0 25 50 75 100
time (s)
Detailed (40CVs) Counter-flow (40CVs)
Detailed
13.11 16.64 8.9 11.36 11.63 13.24
Model
Counter-flow
13.17 16.67 8.94 11.33 11.58 13.19
Model
Figure 4.11: Transient comparison results for detailed model and counter-flow model for
a 4-row wet coil.
76
Comparison results for the detailed model and the water-side ε-NTU model (with
much fewer control volumes) are illustrated in Figure 4.12 and Figure 4.13 for the 8-row
coil, respectively. The implicit solution was applied to the water-side ε-NTU model, and
8 control volumes and a 1-second time step were utilized.
13 14
12 13
T w,out (deg C)
T a,out (deg C)
11
12
10
11
9
8 10
7 9
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (80CVs) Water-side ε-NTU (8CVs) Detailed (80CVs) Water-side ε-NTU (8CVs)
Detailed
7.47 11.73 10.07 13.23
Model
Water-side
7.5 11.73 10.06 13.2
ε-NTU Model
Figure 4.12: Transient comparison results for detailed model and water-side ε-NTU
model for an 8-row dry coil.
77
16 11.0
10.5
15
T a,out (deg C) 10.0
Wa,out (g/kg)
14
9.5
13 9.0
8.5
12
8.0
11
7.5
10 7.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Detailed (80CVs) Water-side ε-NTU (8CVs) Detailed (80CVs) Water-side ε-NTU (8CVs)
16.0
15.5
T w,out (deg C)
15.0
14.5
14.0
13.5
13.0
0 25 50 75 100
time (s)
Detailed (80CVs) Water-side ε-NTU (8CVs)
Detailed
10.65 15.23 7.71 10.36 13.21 15.5
Model
Water-side
10.67 15.19 7.72 10.31 13.2 15.56
ε-NTU Model
Figure 4.13: Transient comparison results for counter-flow model and water-side ε-NTU
model for an 8-row wet coil.
78
The comparison results under either dry or wet conditions show that the water-
side ε-NTU model can achieve the same accuracy with 10 times fewer control volumes
compared to the detailed model. Comparison results for the 4-row coil are quite similar
in this respect. The assumption of a steady-state water temperature profile within each
control volume works well and can reduce the number of control volumes significantly.
Numerical transient investigations for changes other than air flow rate were performed to
confirm the conclusion.
Based on the above numerical studies, it can be concluded that the water-side ε-
NTU model is the most appropriate dynamic forward model for cooling coils of those
considered. It treats counter cross-flow coils as counter-flow coils and does not need
detailed information about coil circuiting and allows large control volumes, e.g., one row.
Computational speed is an important issue when the model is coupled to other equipment
and used in long-term (e.g., one year) simulations. The model performance will be
validated by dynamic experimental data extensively in the next section.
The dynamic forward model can predict various transient responses of coil outlet
conditions caused by variations of coil inlet conditions. In order to validate the model
performance, here specifically referring to the water-side ε-NTU model, a number of
transient experiments under different transient operating conditions were performed, and
the test results were compared with the model predictions. Steady-state results are used
to normalize transient test and model results as follows
Ta,out -Ta,out,ini
Θa,out = (4.53)
Ta,out,fin -Ta,out,ini
and
ωa,out -ωa,out,ini
Ψ a,out = (4.54)
ωa,out,fin -ωa,out,ini
Tw,out -Tw,out,ini
Θ w,out = (4.55)
Tw,out,fin -Tw,out,ini
for water outlet temperature, where the subscripts ini and fin represent values at initial
and final steady states.
Transient responses of the coil outlet conditions to step changes or fast ramp
changes (less than 5 seconds) of air flow rate were investigated first. Table 4.2
summarizes the coil inlet conditions in detail for the 8-row coil with wavy fins.
Table 4.2: Inlet conditions for changes in air flow rate of an 8-row coil.
Cases 1, 2, and 3 are for fully dry, partially dry and partially wet, and fully wet
coil conditions when the air flow rate changed from a low to high level at high water flow
rates. Case 4 is for a partially dry and partially wet coil condition at low water flow rates,
and Case 5 is for a change in air flow rate from a high to low level. Transient results for
experimental measurements and model predictions are given in Figure 4.14, Figure 4.15,
Figure 4.16, Figure 4.17, and Figure 4.18.
80
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Θw,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
Figure 4.14: Transient responses to increase in air flow rate for an 8-row fully dry coil.
81
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.15: Transient responses to increase in air flow rate for an 8-row partially dry and
partially wet coil.
82
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.16: Transient responses to increase in air flow rate for an 8-row fully wet coil.
83
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.17: Transient responses to increase in air flow rate at low water flow rates for an
8-row partially dry and partially wet coil.
84
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.18: Transient responses to decrease in air flow rate for an 8-row partially dry
and partially wet coil.
85
From the transient curves presented in these figures, it is seen that the dynamic
forward model predictions agree well with the experimental data under a wide range of
scenarios involving air flow changes. The initial and final steady-state parameters before
and after each transient process are given in the tables attached to each figure. The
steady-state predictions also agree well with the experimental results.
The results of transient responses to step changes or fast ramp changes (less than
10 seconds) in water flow rate were investigated as well. Table 4.3 gives the coil inlet
parameters under these transient conditions.
Table 4.3: Inlet conditions for changes in water flow rate of an 8-row coil.
Cases 6 and 7 are both for partially dry and partially wet coils when the water
flow rate changed from a low to high level under high air flow rates, while Case 8 is
under low air flow rate. Transient comparison results for experimental measurements
and model predictions are given in Figure 4.19, Figure 4.20, and Figure 4.21.
86
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.19: Transient responses to increase in water flow rate for an 8-row partially dry
and partially wet coil.
87
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.20: Transient responses to increase in water flow rate at high air inlet humidities
for an 8-row partially dry and partially wet coil.
88
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.21: Transient responses to increase in water flow rate at low air flow rates for an
8-row partially dry and partially wet coil.
89
Again, the transients caused by changes of water flow rate are predicted well by
the model for different scenarios. It is noticed that the water outlet temperature increases
a little before it starts decreasing during the initial transients for both the model and
experimental results. This can be explained by the ramp change in water flow rate.
Initially, an increase in water flow rate enhances the water-side heat transfer rate, which
counters the effect of an increased capacitance rate and causes the increase of water outlet
temperature. With a further increase in water flow rate, the increase of water-side heat
flow rate is dominated by the increased capacitance rate effect and the water outlet
temperature decreases.
Besides single step change or fast ramp change of air or water flow rate, transient
responses to simulated feedback control of both air and water flow rates were
investigated. The simulated feedback control of air flow rate was achieved by adjusting
the speed of the variable speed blower randomly, and likewise for water flow rate the
variable speed pump was adjusted to simulate feedback control. The purpose was to
evaluate if the dynamic model could predict transients well when fluid flow rates
changed continuously, so how they were really controlled was not the concern here.
Table 4.4 describes the coil inlet conditions and Figure 4.22 and Figure 4.23 show the
changes of air or water flow rate for each transient experiment.
Table 4.4: Inlet conditions for simulated feedback control of fluid flow rates of an 8-row
coil.
Case 9 is for a partially dry and partially wet coil when the air flow rate was
controlled as shown in Figure 4.22, while Case 10 is for the water flow rate controlled as
90
shown in Figure 4.23. Transient comparison results for experimental measurements and
model predictions are illustrated in Figure 4.24 and Figure 4.25.
1.4
1.3
M dot,a (kg/s)
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0 50 100 150 200
time (s)
Figure 4.22: Simulated feedback control of air flow rate for an 8-row coil.
0.50
0.45
M dot,w (kg/s)
0.40
0.35
0.30
0.25
0.20
0 50 100 150 200
time (s)
Figure 4.23: Simulated feedback control of water flow rate for an 8-row coil.
91
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 50 100 150 200 0 50 100 150 200
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200
time (s)
TEST MODEL
Figure 4.24: Transient responses to simulated feedback control of air flow rate for an 8-
row partially dry and partially wet coil.
92
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 50 100 150 200 0 50 100 150 200
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200
time (s)
TEST MODEL
Figure 4.25: Transient responses to simulated feedback control of water flow rate for an
8-row partially dry and partially wet coil.
93
It can be seen from Figure 4.24 and Figure 4.25 that the model predictions follow
the test measurements reasonably well for air outlet temperature and humidity, and water
outlet temperature. The tables attached to the figures give the parameters used to
normalize the transient results. They are the minimum and maximum values for each
outlet parameter during the transient processes, and the test and model values are close to
each other except for the minimum air outlet humidity in Figure 4.24. This relatively
large difference could be caused by the transient measurement error associated with the
humidity sensor. However, this error is difficult to identify in this transient process, since
the tests involved multiple changes in air flow rate continuously. Overall, the model can
handle transient conditions stimulated by feedback control of fluid flow rates.
In addition to transients associated with changes in fluid flow rates, transients
caused by changes in air inlet temperature and humidity, and water inlet temperature
were considered. Table 4.5 summarizes the inlet parameters considered. Both
temperature changes occurred slowly within 100 seconds while the air humidity change
was relatively fast and achieved within 20 seconds.
Table 4.5: Inlet conditions for changes in air inlet temperature and humidity, and water
inlet temperature of an 8-row coil.
The three cases are all for partially dry and partially wet coils at high fluid flow
rates. Transient comparison results for experimental measurements and model
predictions are given in Figure 4.26, Figure 4.27, and Figure 4.28.
94
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.26: Transient responses to decrease in air inlet temperature for an 8-row partially
dry and partially wet coil.
95
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.27: Transient responses to increase in air inlet humidity for an 8-row partially
dry and partially wet coil.
96
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure 4.28: Transient responses to increase in water inlet temperature for an 8-row
partially dry and partially wet coil.
97
Figure 4.26 and Figure 4.27 show that the model does a good job in predicting
transient coil operation under varying air inlet temperature and humidity. The
fluctuations in Figure 4.26 appear to be large because the overall changes of outlet
parameters during the transient process are relatively small; the results presented in the
normalized format enlarge the fluctuations. From Figure 4.28, the transients to changes
in water inlet temperature are simulated well by the model as well. The initial decrease
in water outlet temperature is caused by a slight increase in water flow rate when the
water inlet temperature was increased.
The 4-row coil with louvered fins was also tested under similar transient operating
conditions. Two sample results are presented here, which are the transient responses to
the simulated feedback control of air and water flow rates, respectively. Table 4.6
describes the inlet conditions, and Figure 4.31 and Figure 4.32 show comparison results
for experimental measurements and model simulations for the outlet conditions. More
transient results can be found in Appendix B. These results show that the dynamic model
works well for this coil geometry as well.
Table 4.6: Inlet conditions for simulated feedback control of fluid flow rates of a 4-row
coil.
1.4
1.3
1.2
M dot,a (kg/s)
1.1
1.0
0.9
0.8
0.7
0.6
0 50 100 150 200
time (s)
Figure 4.29: Simulated feedback control of air flow rate for a 4-row coil.
0.50
0.45
M dot,w (kg/s)
0.40
0.35
0.30
0.25
0.20
0 50 100 150 200
time (s)
Figure 4.30: Simulated feedback control of water flow rate for a 4-row coil.
99
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 50 100 150 200 0 50 100 150 200
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200
time (s)
TEST MODEL
Figure 4.31: Transient responses to simulated feedback control of air flow rate for a 4-
row partially dry and partially wet coil.
100
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 50 100 150 200 0 50 100 150 200
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200
time (s)
TEST MODEL
Figure 4.32: Transient responses to simulated feedback control of water flow rate for a 4-
row partially dry and partially wet coil.
101
Computational speed
(Real time/computational time)
Condition
Implicit Explicit
Table 4.7 shows that for the same accuracy, the implicit method runs 10000 times
faster than real time under fully dry conditions and 750 times faster than real time under
fully wet conditions. For partially dry and partially wet coils, the computational speed is
between those of the fully dry and wet coils. The explicit method does not provide any
advantage in accelerating the model calculation for a dry coil because although no
iteration is needed when solving the model, a much smaller time step is required to
achieve the desired accuracy. The time step for the implicit method is 1 second while it
102
was reduced to 0.1 second for the explicit method. However for the wet coil the explicit
method was 3.33 times faster than the implicit method because the advantage of no
iteration counters the disadvantage of a small time step (More iterations are needed to
solve the wet coil model than the dry coil model when using the implicit method).
Overall, the water-side ε-NTU model using the explicit solution approach is
recommended.
The dynamic forward model was coded in C++ and implemented as a DLL
(Dynamic Link Library) file that is callable from a variety of software under the
Windows environment, e.g., MatLab. The steady-state model, which is used to prepare
initial coil states for the dynamic model, was also coded in C++ and implemented in the
DLL file. A detailed user manual is provided in Appendix C, including an example
calling the model in MatLab.
The simplified distributed model developed in this study was compared with
existing simplified models developed by Braun et al. (1989), Clark (1985), and Ding et
al. (1990). All three of these existing models are lumped models and do not account for
the distributed nature of the transient terms. Furthermore, the model of Braun et al. is
quasi-static. The purpose of the comparisons was to identify if any of these lumped
models could simulate coil transients reasonably well with less computational
requirements. The models presented by McCullagh et al. (1969) and Chow (1997) were
not compared, since these two models have the same order of complexity as the current
model and do not have an advantage in execution speed.
Four representative transient conditions were investigated for the 8-row coil with
wavy fins, which are step changes and sinusoidal variations in air flow rates and water
flow rates. The inlet conditions are summarized in Table 4.8 and the corresponding
comparison results are given in Figure 4.37, Figure 4.38, Figure 4.39, and Figure 4.40,
respectively.
103
1.4
1.2
M dot,a (kg/s)
1.0
0.8
0.6
0.4
0.2
0.0
0 25 50 75 100
time (s)
0.6
0.5
M dot,w (kg/s)
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
1.4
1.2
M dot,a (kg/s)
1.0
0.8
0.6
0.4
0.2
0.0
0 25 50 75 100
time (s)
0.6
0.5
M dot,w (kg/s)
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Zhou Braun Clark Ding Zhou Braun Clark Ding
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
Zhou Braun Clark Ding
Figure 4.37: Transient comparison results to step change in air flow rate for different
models.
106
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Zhou Braun Clark Ding Zhou Braun Clark Ding
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
Zhou Braun Clark Ding
Figure 4.38: Transient comparison results to step change in water flow rate for different
models.
107
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Zhou Braun Clark Ding Zhou Braun Clark Ding
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
Zhou Braun Clark Ding
Figure 4.39: Transient comparison results to sinusoidal variation in air flow rate for
different models.
108
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
Zhou Braun Clark Ding Zhou Braun Clark Ding
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
Zhou Braun Clark Ding
Figure 4.40: Transient comparison results to sinusoidal variation in water flow rate for
different models.
109
The model developed by Braun et al. predicts step changes in coil outlet
parameters in response to step changes in air and water flow rates, which is expected
since this model is quasi-static and contains no transient terms. For the sinusoidal
variations in both fluid flow rates, the quasi-static results responded faster compared with
the transient models, especially the water-side outlet temperatures. Also, this model over
predicts the amplitude of the transient responses for outlet parameters. Therefore, the
comparison results confirm that a transient model is required to simulate transient
performance for cooling coils.
The Clark model improves the transient predictions significantly by simply
adding a single time constant to the quasi-static outlet parameters. However, the air-side
outlet parameters responded more slowly and the water-side outlet temperature responded
faster in comparison with the current model. The Clark model gives reasonable transient
results, but a single time constant can not precisely describe transients of cooling coils
because the time constants for the water-side and air-side are different.
The model developed by Ding et al. showed initial “jumps” in transient responses
of water-side outlet temperatures when the air and water flow rates were changed by
steps. The main reason is that in the Ding model, the water-side heat resistance is
approximated by the difference of the overall heat resistance and the air-side heat
resistance, which is only valid under steady-state conditions. The Ding model did a
better job in simulating coil transients for sinusoidal variations in fluid flow rates. From
the tables attached to the figures, it is seen that there exist relatively large differences
between the Zhou and Ding model for the outlet parameters used to normalize the
transient results. The Ding models considers a coil as either fully dry or fully wet by
comparing the equivalent coil surface temperature to the air inlet dewpoint temperature,
and wet-bulb temperature differences are used as driving potentials instead of enthalpy
differences in the wet regime. Both approximations can cause error in predicting the total
energy transfer. When calculating the air-side outlet temperature, the air-side
effectiveness for total energy transfer is applied rather than the sensible effectiveness,
which leads to errors in the air outlet parameters.
110
When using the simplified dynamic forward model developed in the last chapter,
detailed information for the coil is needed in order to calculate heat conductances and
capacitances. However, this information is not always available; instead certain
experimental data can be obtained for the coil. The purpose of inverse modeling is to
develop a model where overall coil parameters can be estimated from experimental
training data without knowledge of the coil physical characteristics.
The inverse model adopts the water-side ε-NTU forward modeling approach,
which assumes a counter-flow arrangement and uses the effectiveness-NTU method to
calculate heat transfer rates for both air side and water side. There are basically five
parameters, hwAw,tot, ηahaAa,tot, η*ah*aAa,tot, and Cw,tot, Cc,tot, which need to be determined
based on training data from a coil before the inverse model can be used. hwAw,tot is a
water-side heat conductance combining convection heat transfer coefficient with total
heat transfer area, and ηahaAa,tot and η*ah*aAa,tot are air-side heat conductances combining
total heat transfer area with overall fin efficiency and convection coefficient for dry and
wet conditions. Cw,tot and Cc,tot are total heat capacitances associated with water in tubes
and coil fin and tube material.
The inverse model also adopts the solution techniques used for the water-side ε-
NTU forward model, that is, Gauss-Seidel iteration for steady-state solution and an
explicit algorithm for dynamic solution. Similar to the forward model , the models
dealing with steady-state and dynamic operating conditions were coded in C++ and
implemented in a DLL file. More information about model usage can be found in
Appendix C.
111
where Tw,out,c and Ta,out,c are water and air outlet temperatures of a coil, and the
superscripts M and T represents values from model prediction and training data from
experimental measurement.
The dependence of hwAw,tot and ηahaAa,tot on fluid flow rates can be correlated
assuming two standard functional forms of
h w A w,tot =f w ( M
) =C ⋅ ( M
w 1
) C2
w (5.2)
ηa h a A a,tot =f a ( M
) =C ⋅ ( M
a 3
)C4
a (5.3)
in units of kW/K. Here C1, C2, C3 and C4 are unknowns to be determined by regression.
η*ah*aAa,tot is assumed to be equal to ηahaAa,tot to simplify the regression process.
There are two steps to accomplish the parameter estimation procedure. The first
step is to use steady-state training data to estimate C1, C2, C3 and C4, the coefficients
associated with coil heat conductances. Then, transient data is used to estimate Cw,tot and
Cc,tot, the heat capacitances of a coil. Sample data for training are given in Table 5.1 and
Table 5.2, which were collected from experiments for the 8-row test coil with wavy fins.
Table 5.1 gives coil inlet and outlet parameters under steady-state conditions with
variations of air flow rate and water flow rate for dry and wet cases. Table 5.2 gives the
coil data in detail under a transient condition of increasing air flow rate, which is Case 1
in Table 4.2.
112
Table 5.1: Steady-state experimental training data of an 8-row coil for heat conductance
regression.
Table 5.2: Transient experimental training data of an 8-row coil for heat capacitance
regression.
Table 5.2: Transient experimental training data of an 8-row coil for heat capacitance
regression (Cont’d).
The steady-state data were investigated and it was found that using purely dry
data did not result in good estimates of C1, C2, C3 and C4. Because it is the overall heat
resistance of a coil that determines the coil performance and outlet parameters under
steady-state dry conditions, hwAw,tot and ηahaAa,tot could not be estimated separately. Wet
data, which contain information for both total and sensible energy transfers, should be
used.
The parameter estimation can be done using an embedded function of least
squares non-linear regression in MatLab. Based on the data in Table 5.1, the four
115
coefficients of C1, C2, C3 and C4 were calculated as 7.2, 0.93, 3.7, and 0.9, respectively.
Equation (5.2) and Equation (5.3) become
h w A w,tot =7.2 ⋅ ( M
)0.93 (kW/K)
w (5.4)
ηa h a A a,tot =3.7 ⋅ ( M
)0.9 (kW/K)
a (5.5)
30 1.0
25 0.9
qregression (kW)
0.8
SHRregression
20
0.7
15
0.6
10 0.5
5 0.4
5 10 15 20 25 30 0.4 0.5 0.6 0.7 0.8 0.9 1.0
qdata (kW) SHRdata
Figure 5.1: Comparisons for training data and regression results under steady-state
conditions.
The left figure shows that qregression agrees with qdata well, and the maximum
relative difference among them is 9 %. In the right figure, SHRregression matches SHRdata
within a relative difference of 7.3 %. However, the regression produced a bias of over
prediction for SHR. This is probably caused by the use of a single heat conductance for
air-side heat and mass transfer, and sensible heat transfer under wet conditions in the
inverse model.
The regression correlations were also compared to the empirical correlations used
in the forward model, which are illustrated in Figure 5.2.
117
6 5
5
4
ηahaAa,tot (kW/K)
hw Aw,tot (kW/K)
4
3
3
2
2
1
1
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.4 0.6 0.8 1 1.2 1.4
(kg/s)
M
M (kg/s)
w a
Empirical Correlation Regression Empirical Correlation Regression
Figure 5.2: Heat conductance comparisons for empirical correlations and regression
results
The left figure shows the water-side heat conductance comparison results. At
lower water flow rates, the regression results agree with the empirical correlation well.
However, the regression results were underestimated at higher water flow rates and the
underestimation increases with an increase in water flow rate. The right figure shows the
air-side heat conductance comparison results. The regression results were overestimated
and the overestimation increases as the air flow rate increases. These differences were
mainly caused by limited training data. There were basically four flow rate levels tested
for both water and air within the ranges considered in the training data, which were not
enough to get accurate regression results. In order to improve the accuracy of the heat
conductance regression, more training data are needed at more flow rate levels. Although
there are differences between the regression and the empirical correlations for both heat
conductances, the steady-state experimental validation in the next section shows that
these regression results work well for the model predictions.
After the correlations of coil heat conductances have been determined, Cw,tot and
Cc,tot, the heat capacitances need to be estimated. Using the transient training data given
in Table 5.2 and least squares non-linear regression from MatLab, Cw,tot and Cc,tot were
118
determined as 27.04 kJ/K and 21.91 kJ/K, respectively. Comparisons for the training
data and the regression results are shown in Figure 5.3.
10 13
9
T a,out,regression (deg C)
12
T w,out,regression (deg C)
8 11
7 10
6 9
5 8
5 6 7 8 9 10 8 9 10 11 12 13
T a,out,data (deg C) T w,out,data (deg C)
Figure 5.3: Comparisons for training data and regression results under transient
conditions
The regressed heat capacitances replicate the training data very well (RMS = 0.14
˚C). It is known that Cw,tot and Cc,tot are 36.49 kJ/K and 20.44 kJ/K from the coil physical
description. The water heat capacitance obtained by regression was underestimated by
26 %, and the coil material heat capacitance from regression was overestimated by 7 %.
With one transient training data set, although the regression itself is very good, the results
can have a relatively large deviation.
Similar to the forward model, the inverse model can predict various operating
conditions for air to water heat exchangers. The performance of the inverse model with
the regression results was validated by the steady-state and transient experimental data
from the 8-row test coil with wavy fins.
119
First, the steady-state data in Table 2.4 and Table 2.5 were investigated. Figure
5.4 shows comparison results for the coil outlet temperatures from the experimental
measurements and the inverse model predictions.
19 19
17 17
T a,out,inverse (deg C)
T w,out,inverse (deg C)
15 15
13 13
11 11
9 9
7 7
5 5
5 7 9 11 13 15 17 19 5 7 9 11 13 15 17 19
T a,out,test (deg C) T w,out,test (deg C)
For the air-side outlet temperatures, the maximum relative difference between test
and model is 9.44 %, and it is 4.87 % for the water-side outlet temperatures. The inverse
model can simulate steady-state conditions well.
Case 2 in Table 4.2 and Case 9 in Table 4.4 were used to validate the transient
performance of the inverse model. Comparison results are shown in Figure 5.5 and
Figure 5.6.
120
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST INVERSE MODEL TEST INVERSE MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST INVERSE MODEL
INVERSE
8.49 12.23 6.93 8.77 10.37 13.22
MODEL
Figure 5.5: Inverse model validation under an increase in air flow rate for an 8-row
partially dry and partially wet coil.
121
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 50 100 150 200 0 50 100 150 200
time (s) time (s)
TEST INVERSE MODEL TEST INVERSE MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100 150 200
time (s)
TEST INVERSE MODEL
INVERSE
7.66 9.97 6.35 7.05 9.55 11.65
MODEL
Figure 5.6: Inverse model validation under simulated feedback control of air flow rate for
an 8-row partially dry and partially wet coil.
122
6.1 Conclusions
Dynamic models of chilled water cooling coils are valuable tools for developers
of system control and FDD. Simulation models can speed up the development and
evaluation of new controllers and FDD algorithms by generating data more efficiently
compared to running real experiments.
In this study, a simplified distributed coil model was developed, which considers
counter cross-flow coils as pure counter-flow arrangements and utilizes an effectiveness-
NTU approach for both air-side and water-side energy transfer calculations. The model
is believed to be the first that has been extensively validated by experimental data under
fully dry, partially dry and partially wet, and fully wet conditions with various transients.
This study was also the first to investigate numerical solution techniques and
computational speeds for transient cooling coil models. It was found that large control
volumes (e.g., one row) are allowed, when using both air and water-side effectiveness
relations to characterize state variable distributions in the flow directions. Also, an
explicit integration scheme is more appropriate for both dry and wet operating conditions
than an implicit scheme, while both methods run much faster than real time. An inverse
coil modeling approach was also developed that uses both steady-state and transient
experimental training data in a two-step approach. Both the forward and inverse models
were coded in C++ and implemented as a DLL file, and can be used in MatLab.
As part of development process, the lumped fin efficiency approach was validated
to be applicable for simulating transient performance for finned-tube coils. In addition, a
modified fin efficiency for sensible heat transfer was developed for dehumidifying
conditions to replace the heat transfer only fin efficiency. This is a valuable contribution
that provides improved predictions of sensible heat transfer. The water condensate effect
124
on coil transients and a Lewis number of unity were also proved to be reasonable
assumptions.
6.2 Recommendations
LIST OF REFERENCES
Bhargava, S.C., McQuiston, F.C., and Zirkle, L.D., 1975, “Transfer Functions for
Crossflow Multirow Heat Exchangers,” ASHRAE Transactions, Vol. 81, Part 2, pp. 294-
314.
Bourdouxhe, J.P., Grodent, M., and Lebrun, J., 1996, Reference Guide for Dynamic
Models of HVAC Equipment, American Society of Heating, Refrigerating, and Air-
Conditioning Engineers, Atlanta, GA.
Braun, J.E., Klein, S.A., and Mitchell, J.W., 1989, “Effectiveness Models for Cooling
Towers and Cooling Coils,” ASHRAE Transactions, Vol. 95, Part 2, pp. 164-174.
Chow, T.T., 1997, “Chilled Water Cooling Coil Models from Empirical to Fundamental,”
Numerical Heat Transfer Part A – Applications, Vol. 32, No. 1, pp. 63-83.
Clark, D.R., 1985, “Type 12: Cooling or Dehumidifying Coil,” HVACSIM+ Reference
Manual, National Bureau of Standards, pp. 63-68.
Ding, X., Eppe, J.P., Lebrun, J., and Wasacz, M., 1990, “Cooling Coil Model to be Used
in Transient and /or Wet Regimes. Theoretical Analysis and Experimental Validation.”
Proceedings of the Third International Conference on System Simulation in Building,
Liège, Belgium, pp. 405-441.
Elmahdy, A.H., and Mitalas, G.P., 1977, “A Simple Model for Cooling and
Dehumidifying Coils for Use in Calculating Energy Requirements for Buildings,”
ASHRAE Transactions, Vol. 83, Part 2, pp. 103-117.
Gartner, J.R., and Harrison, H.L., 1963, “Frequency Response Transfer Function for a
Tube in Crossflow,” ASHRAE Transactions, Vol. 69, pp. 323-330.
Gartner, J.R., and Daane, L.E., 1969, “Dynamic Response Relations for a Serpentine
Crossflow Heat Exchanger with Water Velocity Disturbance,” ASHRAE Transactions,
Vol. 75, Part 2, pp. 53-68.
126
Holmes, M.J., 1988, HVAC Component Specification: Heating and Cooling Coils, in
Energy Conservation in Building and Community Systems Programme, Annex X: System
Simulation, International Energy Agency.
Incropera, F.P., and DeWitt, D.P., 1996, Fundamentals of Heat and Mass Transfer, John
Wiley & Sons, New York, NY.
Jawadi, Z., 1988, “A Simple Transient Heating Coil Model. Analysis of Time Dependent
Thermal Systems,” Winter Annual Meeting of ASME, pp. 63-69.
Jota, F.G., and Dexter, A.L., 1986, “Self-tuning Control of a Cooling-coil: Simulated and
Experimental Assessment,” Proceedings of the International Conference on System
Simulation in Buildings, Liège, Belgium, pp. 577-598.
Kays, W.M., and London, A.L., 1984, Compact Heat Exchangers, McGraw-Hill, New
York, NY
Kuehn, T.H., Ramsey, J.W., and Threlkeld, J.L., 1998, Thermal Environmental
Engineering, Prentice-Hall, Upper Saddle River, NJ
McCullagh, K.R., Green, G.H., and Chandra, S.S., 1969, “An Analysis of Chill Water
Cooling Dehumidifying Coils Using Dynamic Relationships,” ASHRAE Transactions,
Vol. 75, Part 2, pp. 200-209.
Murthy, J.Y., and Mathur, S.R., 1998, Numerical Methods in Heat, Mass, and Momentum
Transfer.
Stoecker, W.F., 1975, Procedures for Simulating the Performance of Components and
Systems for Energy Calculations, ASHRAE, Atlanta, GA.
Tamm, H., 1969, “Dynamic Response Relations for Multi-Row Crossflow Heat
Exchangers,” ASHRAE Transactions, Vol. 75, Part 1, pp. 69-80.
Tamm, H., and Green, G.H., 1973, “Experimental Multi-Row Crossflow Heat Exchanger
Dynamics,” ASHRAE Transactions, Vol. 79, Part 2, pp. 9-18.
127
Table A.1: Steady-state experimental data of a 4-row coil for air-side convection
coefficient calculation.
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure B.1: Transient responses to increase in air flow rate for a 4-row partially dry and
partially wet coil.
130
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure B.2: Transient responses to increase in air flow rate at low water flow rates for a
4-row partially dry and partially wet coil.
131
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure B.3: Transient responses to decrease in air flow rate for a 4-row partially dry and
partially wet coil.
132
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure B.4: Transient responses to increase in water flow rate for a 4-row partially dry
and partially wet coil.
133
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure B.5: Transient responses to decrease in water flow rate at low air flow rates for a
4-row partially dry and partially wet coil.
134
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure B.6: Transient responses to decrease in air inlet temperature for a 4-row partially
dry and partially wet coil.
135
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
Ψa,out
Θa,out
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-0.1 -0.1
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.0
0.9
0.8
0.7
0.6
Θw,out
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
0 25 50 75 100
time (s)
TEST MODEL
Figure B.7: Transient responses to decrease in air inlet humidity for a 4-row partially dry
and partially wet coil.
136
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Ψa,out
Θa,out
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 0 25 50 75 100
time (s) time (s)
TEST MODEL TEST MODEL
1.1
1.0
0.9
0.8
0.7
Θw,out
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 25 50 75 100
time (s)
TEST MODEL
Figure B.8: Transient responses to increase in water inlet temperature for a 4-row
partially dry and partially wet coil.
137
Two files of “coilmodel.dll” and “coilmodel.h” are copied into the directory of
“d:\coilmodel\”. Four functions in the model package are “steadyFor”, “dynamicFor”,
“steadyInv”, and “dynamicInv”, which represent steady-state forward model, dynamic
forward model, steady-state inverse model, and dynamic inverse model. The format of
these functions is
• steadyFor (Par, Inlet, pOutlet)
• dynamicFor (Time, Par, Inlet, pOutlet)
• steadyInv (Par, Con, Inlet, pOutlet)
• dynamicInv (Time, Par, Con, Cap, Inlet, pOutlet)
Model usage in MatLab is demonstrated by two examples, of which one is using
the forward model to simulate Case 9 in Table 4.4, and the other is using the inverse
model to simulate the same case.
% Call “steadyFor” to get the initial coil states under the initial inlet condition.
i = 1;
Inlet = Minlet(:,i);
% Inlet(1): air inlet temperature.
% Inlet(2): air inlet relative humidity.
% Inlet(3): air mass flow rate.
% Inlet(4): water inlet temperature.
% Inlet(5): water mass flow rate.
Outlet = [0 0 0 0 0 0];
pOutlet = libpointer('doublePtr',Outlet);
calllib('coilmodel','steadyFor',Par,Inlet,pOutlet);
Outlet = get(pOutlet,'Value');
% Outlet(1): air outlet temperature.
% Outlet(2): air outlet absolute humidity.
% Outlet(3): water outlet temperature.
% Outlet(4): heat transfer rate.
% Outlet(5): sensible heat ratio.
% Outlet(6): fraction of dry coil.
% Coil outlet states are written into the text file of “outlet.txt”.
fid = fopen('d:\coilmodel\outlet.txt');
M = fscanf(fid,'%f',[7,inf]);
fclose(fid);
M = M';
% M(:,1): time.
% M(:,2): air outlet temperature.
% M(:,3): air outlet absolute humidity.
% M(:,4): water outlet temperature.
% M(:,5): air side heat transfer rate.
% M(:,6): sensible heat ratio.
% M(:,7): fraction of dry coil surface.
addpath('d:\coilmodel');
loadlibrary coilmodel.dll coilmodel.h;
libfunctions coilmodel;
% Different from the forward model, much fewer parameters are needed by the inverse
model, so their values are simply set here.
N = 8;
% N: number of control volumes.
Tol = 1e-5;
% Tol: relative convergence tolerance.
B = 101325.0;
142
% B: atmospheric pressure.
Cpa = 1.027;
% Cpa: air specific heat.
Cpw = 4.202;
% Cpw: water specific heat.
Par = [N,Tol,B,Cpa,Cpw];
C3 = 3.7;
C4 = 0.9;
% ηa h a Aa,tot =C3 ⋅ ( M
) C4
a
C5 = C3;
C6 = C4;
% η*a h*a A a,tot =C5 ⋅ ( M
) =C ⋅ ( M
)
C6 C4
a 3 a
Con = [C1,C2,C3,C4,C5,C6];
Cw = 27.04;
% Cw: water heat capacitance.
Cc = 21.91;
% Cc: coil material heat capacitance.
Cap = [Cw,Cc];
fid = fopen('d:\coilmodel\inlet.txt');
Minlet = fscanf(fid,'%f',[5,inf]);
fclose(fid);
143
% Call “steadyInv”.
i = 1;
Inlet = Minlet(:,i);
Outlet = [0 0 0 0 0 0];
pOutlet = libpointer('doublePtr',Outlet);
calllib('coilmodel','steadyInv',Par,Con,Inlet,pOutlet);
Outlet = get(pOutlet,'Value');
% Call “dynamicInv”.
DSTART = 0;
DEND = 200;
interval = 1;
tstart = DSTART;
tend = tstart;
dt = 1;
while tend < DEND
tstart = tend;
tend = tstart+interval;
time = [tstart,tend,dt];
i = i + 1;
Inlet = Minlet(:,i);
calllib('coilmodel','dynamicInv',time,Par,Con,Cap,Inlet,pOutlet);
Outlet = get(pOutlet,'Value');
end
unloadlibrary coilmodel;
fid = fopen('d:\coilmodel\outlet.txt');
M = fscanf(fid,'%f',[7,inf]);
fclose(fid);
144
M = M';
dt = M(:,1);
Tao = M(:,2);
plot(dt,Tao);
xlabel('time(s)');
ylabel('Tao(C)');