Assessment of Factors Contributing To Refrigerator Cycling Losses
Assessment of Factors Contributing To Refrigerator Cycling Losses
Assessment of Factors Contributing To Refrigerator Cycling Losses
Air Conditioning and Refrigeration Center A National Science Foundation/University Cooperative Research Center
iii
Table of Contents
Page
Abstract.........................................................................................................................................iii
3.5 Refrigerant Movement from Compressors with High Side Oil Sumps.................................. 16
4.5 Energy Penalty Resulting From Migration for 90 °F Ambient Case ...................................... 28
4.7 Energy Implications Resulting From Migration for 60 °F Ambient Case ............................... 29
4.8 Conclusions......................................................................................................................... 30
iv
5.1 Introduction ......................................................................................................................... 31
5.6 Conclusions......................................................................................................................... 41
6.4 Conclusions......................................................................................................................... 45
References.................................................................................................................................50
v
List of Figures
Page
Figure 2.1 Compressor heat transfer coefficient...........................................................................................................................5
Figure 2.2 Isentropic compressor efficiency during cycling for 90 °F ambient case...............................................................6
Figure 2.3 Isentropic compressor efficiency during cycling for 60 °F ambient case...............................................................7
Figure 2.4 Cabinet thermal mass effect on evaporator inlet air temperature...........................................................................10
Figure 3.1 Pressure variation in compressor sump during cycling..........................................................................................13
Figure 3.2 Compressor temperature during cycling ...................................................................................................................13
Figure 3.3 Average compressor shell temperature during cycling ..........................................................................................13
Figure 3.4 Solubility of R-12 in 13 ounces of naphthene...........................................................................................................15
Figure 3.5 Mass of refrigerant in compressor oil during cycling .............................................................................................15
Figure 3.6 Pressure variation on compressor high side.............................................................................................................17
Figure 3.7 Refrigerant movement with high side sump ..............................................................................................................18
Figure 4.1 Liquid migration ............................................................................................................................................................19
Figure 4.2 Vapor migration.............................................................................................................................................................20
Figure 4.3 Evaporator inlet temperatures for 90 °F ambient case.............................................................................................21
Figure 4.4 Condenser and evaporator off-cycle pressures for 90 °F ambient case...............................................................22
Figure 4.5 Condenser off-cycle temperatures for 90 °F ambient case.....................................................................................23
Figure 4.6 Predicted and measured condenser pressure during off-cycle..............................................................................27
Figure 4.7 Evaporator inlet temperatures for 60 °F ambient case.............................................................................................29
Figure 4.8 Condenser and evaporator off-cycle pressures for 60 °F ambient case...............................................................29
Figure 5.1 Evaporator exit temperature at beginning of on-cycle in a 90°F room..................................................................32
Figure 5.2 Evaporator capacity for 90 °F ambient case..............................................................................................................34
Figure 5.3 Evaporator air temperatures for a 90°F ambient case..............................................................................................35
Figure 5.4 Total system power for 90 °F ambient case..............................................................................................................36
Figure 5.5 COP for 90 °F ambient case.........................................................................................................................................36
Figure 5.6 Evaporator exit temperature at beginning of on-cycle in a 60°F room..................................................................37
Figure 5.7 Evaporator capacity for 60 °F ambient case..............................................................................................................38
Figure 5.8 Evaporator air temperatures for a 60°F ambient case..............................................................................................39
Figure 5.9 Condenser mass flow rates for 60 °F ambient case..................................................................................................39
Figure 5.10 Condenser subcooling for 60 °F ambient case.......................................................................................................40
Figure 5.11 COP for 60 °F ambient case.......................................................................................................................................41
Figure 6.1 Superheat and subcooling during cycling................................................................................................................42
Figure 6.2 Effect of condenser subcooling on capillary tube mass flow rate.........................................................................43
Figure 6.3 Effect of evaporator superheat on capillary tube mass flow rate ..........................................................................44
Figure A.1 Data acquisition system block diagram....................................................................................................................52
Figure A.2 Brass instrumentation block......................................................................................................................................53
Figure A.3 New thermocouple instrumentation method ...........................................................................................................53
Figure A.4 New pressure tap instrumentation technique .........................................................................................................53
Figure A.5 Surface thermocouple .................................................................................................................................................54
Figure A.6 Steady state comparison of surface and immersion thermocouples....................................................................54
vi
Figure A.7 Surface thermocouple model......................................................................................................................................56
Figure A.8 Surface thermocouple correlation at compressor exit ............................................................................................57
Figure A.9 Surface thermocouple correlation at evaporator exit..............................................................................................57
Figure A.10 Transient comparison of surface and immersion thermocouples at compressor exit......................................58
Figure A.11 Transient comparison of surface and immersion thermocouples at evaporator inlet .....................................58
Figure A.12 Surface and immersion temperatures at compressor exit during cycling ..........................................................59
Figure A.13 Surface and immersion temperatures at evaporator inlet during cycling..........................................................59
Figure A.14 Comparison of Predicted and measured refrigerant temperatures at compressor exit ....................................60
Figure A.15 Comparison of predicted and measured refrigerant temperatures at evaporator inlet....................................60
Figure A.16 Comparison of PC519E and Valhalla Scientific 2100 power transducers...........................................................62
Figure B.1 Fan location...................................................................................................................................................................65
Figure B.2 Freezer Energy Balance ...............................................................................................................................................65
Figure B.3 Refrigerator Energy Balance.......................................................................................................................................66
Figure B.4 Parameter estimation minimum...................................................................................................................................67
Figure E.1 Measured vs. predicted compressor power for fall 1992 data ...............................................................................78
Figure E.2 Measured vs. predicted compressor power for spring 1993 data .........................................................................78
vii
List of Tables
Page
Table 2.1 Steady state and cycling performance..........................................................................................................................4
Table 2.2 Heat exchanger parameters .............................................................................................................................................9
Table 3.1 R-12/naphthene constants for Grebner-Crawford model.........................................................................................14
Table A.1 Accuracy of data acquisition system........................................................................................................................51
Table A.2 Steady state comparison of surface and immersion thermocouples .....................................................................55
Table A.3 Surface thermocouple k values ...................................................................................................................................56
Table A. Comparison of surface and immersion thermocouples during cycling...................................................................58
Table A.5 Surface thermocouple error summary for transient data.........................................................................................60
Table A.6 Power transducer application and accuracy.............................................................................................................61
Table B.1 Summary of UA testing methods................................................................................................................................63
Table D.1 Dittus-Boelter parameter values..................................................................................................................................76
viii
Chapter 1: Introduction and Literature Review
1.1 Purpose
The purpose of this research project is to obtain a detailed understanding of the cycling performance of
refrigerator-freezers by identifying and quantifying the individual phenomena which degrade energy efficiency,
relative to its steady state level. This is important because the Department of Energy uses a steady state simulation
model as a basis for refrigerator standard setting, but tests for compliance under cycling conditions. Another
objective of this research is to develop some insights about the need for a transient model, or whether a quasi-steady
model could suffice with minor corrections. These goals are accomplished through the analysis of experimental data
obtained with a typical household refrigerator (Amana model TC18MBL). It is a single evaporator, reciprocating
compressor unit with a top mounted freezer compartment. Improvements made on the existing instrumentation will
also be discussed.
1
The thermal mass of the system components has also been identified as a possible source of cycling losses.
Staley et al (1992) state that the thermal mass of the refrigerator cabinet is not likely to be an important contributing
factor to cycling losses, but the thermal mass of the compressor may improve transient performance by keeping
discharge temperatures below steady state levels throughout the cycle.
2
Chapter 2: Thermal Mass Effects
2.1 Introduction
The purpose of this chapter is to begin an investigation of the transient behavior of a refrigerator, with an
analysis of the thermal mass of the system components. Data for this analysis were taken in a 90 °F ambient room,
with a charge of 12.5 ounces of R-12. The refrigerator was overcharged from its recommended 8 ounces for other
testing, however qualitative results should be similar. The temperature control was set in its middle position to
produce 60-minute cycles.
3
Table 2.1 Steady state and cycling performance
Instantaneous
Measured Quantity Steady State Average Cycling
Cycling
TA:Evaporator Inlet (°F) 3.4 3.1 3.1
TA:Ambient (°F) 90.0 90.0
TA:Condenser Inlet (°F) 96.4 96.0
TA:Mid (°F) 101.6 100.7
TR:Evaporator Exit (°F) -11.3 -13.3
TR:Evaporator Inlet (°F) -12.9 -14.7
TR:Condenser Exit (°F) 103.7 102.3
Condenser Exit Subcooling (°F) 17.6 17.8
TR:Condenser Inlet (°F) 198.6 190.8
Condenser Inlet Superheat (°F) 77.0 70.4
TR:Compressor Inlet (°F) 77.1 76.6
TS:Compressor Shell (°F) 151.5 146.2
Compressor Inlet Pressure (psia) 18.0 17.6
Compressor Exit Pressure (psia) 176.0 173.3
Mass Flow Rate at Cond. Exit (lbs/hr) 13.2 12.9
Q Condenser (Btu/hr) 941 907
Q Compressor (Btu/hr) 474 432
Q Heaters (Btu/hr) 260 0 0
Q Evaporator (Btu/hr) 751 738 459
System Power (Btu/hr) 818 793 502
COP 0.92 0.93 0.91
dU comp
Wa = m
Ý∆h + Q comp +
dt 2.1
i ∆h is the actual enthalpy change of the refrigerant across the compressor, Qcomp is the heat rejected by the
where m
compressor, and Ucomp is the energy stored in the compressor shell. The relative magnitudes of the changes in these
terms determine whether W a increases or decreases during cycling.
At the instantaneous data point 23 minutes into the on-cycle, the suction inlet conditions were nearly
identical. However, the discharge gas was cooler, possibly due to the compressor thermal mass. This reduced the
first term in Equation 2.1 by about 16 Btu/hr.
The second term, Qcomp, can be calculated using Equation 2.2,
4
Qcomp = h comp∆T
2.2
where h comp is the overall compressor convective heat transfer coefficient, and ∆T is the temperature difference
between the compressor shell and the surrounding air. The compressor heat transfer coefficient is calculated using
26 steady state data points taken at various ambient temperatures (for which Qcomp is known from a refrigerant side
energy balance) as shown in Figure 2.1. The instantaneous value for Qcomp is 432 Btu/hr during cycling, or 42 Btu/hr
less than the 474 Btu/hr observed during steady state operation. The compressor shell temperatures were 146.2 °F
and 151.5 °F respectively, as shown in Table 2.1.
The thermal storage term can be evaluated using Equation 2.3,
dUcomp dT
= mc p
dt dt 2.3
in which the thermal mass of the compressor shell, mc p, is estimated to be 3 Btu/°F, and T is the compressor shell
temperature. By measuring the rate of increase in the compressor shell temperature during cycling, the magnitude of
the storage term is estimated to be 36 Btu/hr.
600
Qcomp
500
400
Btu/hr
300
100
0
0 10 20 30 40 50 60
TS:Compressor Shell - TA:Mid (°F)
Summing the changes in all of the terms in Equation 2.4, W a decreases by 22 Btu/hr for the cycling case,
which nearly accounts for the 25 Btu/hr difference in measured power. Most of the error is attributed to the
uncertainty of the thermal mass of the compressor, and in the power measurement.
A Second Law analysis can be used to determine the effect of thermal mass on compressor efficiency. The
isentropic efficiency of the compressor, as defined by Equation 2.4,
Ws
ηs =
Wa 2.4
can be calculated for all points during the on-cycle, and then compared to the steady state value corresponding to
the same inlet air temperatures to the evaporator and condenser. The isentropic efficiency, ηs, is just the isentropic
5
work of compression, W s, divided by the measured compressor power, W a . The measured compressor power is
compared to the power predicted by the compressor map in Appendix E. The isentropic work of compression is
calculated using Equation 2.5,
Ý(hin − h out,s )
Ws = m
2.5
in which m
i is the mass flow rate given by the compressor map (Admiraal, 1993), h in is the enthalpy of the measured
inlet state, and h out,s is the hypothetical enthalpy that would be obtained by isentropically compressing to the
measured exit pressure. The turbine flow meter at the condenser exit cannot be used to measure the mass flow rate
through the compressor during cycling; the mass flow rate through the compressor can be up to 10 percent greater
than the mass flow rate at the condenser exit during the beginning of the on-cycle as charge collects in the condenser
(see Chapter 5). The compressor efficiency for the on-cycle is plotted in Figure 2.2.
0.5
0.4
Compressor Efficiency
0.3
Cycling (Wa)
Cycling (Wmap)
Steady State
0.2
0.1
0
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
Figure 2.2 Isentropic compressor efficiency during cycling for 90 °F ambient case
During the first minute of the on-cycle, the mass flow rate through the compressor is not known with
certainty because of the passage of liquid slugs, thus ηs is not shown. The instantaneous efficiency rose from 0.41
to 0.44 in the first 5 minutes, and then remained nearly constant for the remainder of the on-cycle. The corresponding
steady state efficiency observed at the same evaporator and condenser air inlet temperatures, ηss, was 0.439. This is
the same as the measured efficiency for most of the on-cycle. One reason for the initial increase in ηs is the decrease
in compressor power, which occurs as a result of the decreasing suction pressure.
Also shown in Figure 2.2 is the isentropic compressor efficiency calculated using the compressor map
power, ηmap, instead of the measured power. The compressor map power, W map , is a function of the measured
compressor inlet and exit pressures, and was found to agree very well with W a for steady state data taken in a 90 °F
ambient room (see Appendix E). Therefore, W map indicates how much power would be required by the compressor
for the measured inlet and exit pressures, if the compressor were running at steady state. Any difference between
W map and W a should therefore be attributable to the lower shell temperature during cycling. The steady state
6
efficiency, ηss is higher than ηmap at the point 23 minutes into the on-cycle because the inlet and exit pressures are
slightly different as indicated in Table 2.1.
Since the compressor shell temperature increases throughout the on-cycle, approaching the steady state
temperature, ηmap should approach ηs rather than diverge as shown in Figure 2.2. One explanation for this anomaly is
instrumentation error. W map is very sensitive to the compressor inlet pressure; a constant 0.5 psia decrease in the
pressure measurement would shift ηmap upward by 0.01, placing it above ηs. Since the compressor inlet pressure is
measured using the sum of three separate transducer readings with a combined uncertainty of ±0.8psia, a 0.5 psia
offset is entirely possible. Also, W a is determined by subtracting the evaporator and condenser fan powers from the
total system power. The overall uncertainty in this measurement is ±4.1W (14 Btu/hr), which could also account for
the difference between ηmap and ηs.
A second set of cycling data was examined to see if the effect of the compressor thermal mass was similar.
Figure 2.3 compares ηs and ηmap for data taken in a 60 °F ambient room. This time, ηs was lower than ηmap , and they
converged throughout the cycle as expected. It appears that about half of the difference between ηs and ηmap is
accounted for by the 9W steady state offset in W map as detailed in Appendix E. The compressor map, created from
data obtained in 90 °F ambient conditions, under-predicts the compressor power at 60 °F ambient conditions by
about 9W, independent of the pressure ratio. The compressor efficiency using the corrected map is also shown in
Figure 2.3. Our data sets do not contain any steady state points that have similar inlet and exit pressures for
comparison.
0.5
0.4
Compressor Efficiency
0.3
0.2
Cycling (Wa)
Cycling (Wmap)
0.1 Cycling (Wmap corrected)
0
0:00 2:00 4:00 6:00 8:00 10:00 12:00
time (m:s)
Figure 2.3 Isentropic compressor efficiency during cycling for 60 °F ambient case
Several other factors could contribute to the disagreement of ηmap and ηs in Figures 2.2 and 2.3. At lower
shell temperatures, the lubricating oil becomes more viscous, thus increasing friction between moving parts. This
would tend to reduce ηs below ηmap . Also, as the shell temperature varies, the mechanical clearances between parts
changes, which could increase or decrease frictional losses.
One difference between the 60 °F and 90 °F ambient cases is that the compressor efficiency is much lower
for the 60 °F case. This trend was apparent from our steady state data as shown in Appendix E, and can be explained
7
by the lower pressure ratios (which decrease the required compressor power) that are characteristic of lower ambient
temperatures. An electric motor is usually less efficient when operating under a reduced load, thus the overall
compressor efficiency is lower. Also, at lower temperatures, the lubricating oil is more viscous, thus frictional losses
are increased.
In summary, the uncertainty in pressure and power measurements make it difficult to compare the observed
cycling efficiency with the predicted steady state efficiency for the same inlet and exit pressures. More detailed tests
are being planned to measure this effect.
A potentially more important factor is the effect of the compressor thermal mass on discharge temperature.
This may have significant "downstream effects" on overall cycle performance under some operating conditions. For
a case in which the condenser exit is saturated, a lower discharge temperature would decrease the condenser
desuperheating area, allowing a larger two-phase area. For a given heat transfer, this would be accompanied with a
decrease in condensing temperature, hence a higher COP. Unfortunately, our data sets do not include a cycling case
with a two-phase condenser exit to verify this hypothesis. For cases like the one analyzed in this chapter with a
subcooled condenser exit, the smaller desuperheating area is offset with a larger subcooled area, and the heat transfer
rate and condensing area remain unchanged. This was analyzed using the ACRC2 steady state simulation model
(Porter and Bullard, 1992). To simulate the effect of lower discharge temperatures caused by the compressor thermal
mass, h comp was varied for a case with 11 °F of condenser subcooling and 13 °F of evaporator superheat. The
increasing compressor heat rejection (to the air and shell combined) was accompanied by an equivalent decrease in
condenser heat transfer, which resulted in a nearly constant evaporator load. The model predicted that the COP and
the total energy usage would remain unchanged despite the varying discharge temperature. This is consistent with
the data presented here.
It appears from the linearity of Figure 2.1 that Qcomp can in fact be modeled as a function of compressor shell
temp erature. If this linearity holds for compressors in general, the parameter h comp could be estimated using the dome
temperature supplied with the compressor calorimetry data at the standard rating condition.
h L
Bi = r c
k 2.6
The characteristic length for either heat exchanger, Lc , is just the tube wall thickness, h r is the convection
heat transfer coefficient (conservatively chosen as the refrigerant side coefficient), and k is the thermal conductivity
of the heat exchanger material. The values used for this calculation are found in Table 2.2. In both cases, the Biot
8
number is far less than 0.1, hence the lumped capacitance assumptions are valid. With the lumped capacitance
method, the importance of the thermal mass of the heat exchangers can be evaluated using their thermal time
constants. Assuming a low Biot number, a solid that is subjected to a sudden temperature change will undergo
approximately 2/3 of that temperature change in one time constant. After five time constants, the solid is considered
to have reached equilibrium with its new surroundings. If the heat exchanger time constants are small compared to
the cycling period of the refrigerator, then thermal mass effects can be neglected. The time constant for a heat
exchanger is determined from Equation 2.7. The terms on the left side represent the energy entering the control
volume from the refrigerant and air sides respectively. The term on the right side represents the energy stored in the
heat exchanger itself. The subscripts r, a, and s refer to refrigerant side, air side, and surface properties respectively.
The convection coefficients, h r and h a , are estimated using least squares techniques by Admiraal and Bullard (1993),
and Cavallaro (1993). The heat exchanger areas, A r and A s, and the mass of the heat exchanger, m, are calculated from
manufacturer's data. The specific heat of the heat exchanger material is cp, and temperatures are denoted with T's.
The solution of this differential equation results in a block of constants called the thermal time constant, τ, as given
by Equation 2.8.
h r A r (T r − Ts ) + ha A a (Ta − T s ) = mc p
dT s
dt 2.7
3600mc p
τ=
h r Ar + h a Aa 2.8
This equation is solved for both the evaporator and the condenser using the values in Table 2.2. Even with
conservative values chosen for the convection coefficients, the time constants are very small compared to a typical
30 minute on-cycle. From this calculation, it appears that the heat exchangers can respond rapidly to changes in their
environment, and that their thermal mass has negligible effect on the refrigerator's performance.
9
percent run-time and energy usage are unaffected. The EER averaged over a continuous quasi-steady pulldown is
approximately equal to the EER at the midpoint. Thus, from a design standpoint, the thermal mass of the refrigerator
cabinet is unimportant.
From a modeling standpoint, the cabinet thermal mass is very important in determining the length of the on-
and off-cycles. As air circulates within the cabinet, it exchanges heat with the walls and shelves, thus the evaporator
must also cool the cabinet. To illustrate the importance of this effect, the measured decline in evaporator air inlet
temperature is compared with the temperature decline that would occur with a massless cabinet in Figure 2.4.
This hypothetical air temperature is obtained from the solution of Equation 2.9,
dT Ý
mc v = Q air
dt 2.9
where m is the mass of the air inside the cabinet (1.4 lbs), cv is the specific heat of air (0.173 Btu/lb-°F), T is the
evaporator. UA frig∆Tfrig and UA frez∆Tfrez represent the heat that leaks through the cabinet walls from the ambient air,
the cabinet loads during the beginning part of the on-cycle (see Chapter 5). Without the added thermal mass of the
cabinet, the evaporator could have achieved its temperature set point in under 4 minutes, compared with 30 minutes
for the actual on-cycle.
25
TA:EvapIn (measured)
20 TA:EvapIn (massless cabinet)
TA:EvapIn (with cabinet)
15
Temperature (°F)
10
-5
-10
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
Figure 2.4 Cabinet thermal mass effect on evaporator inlet air temperature
To model the actual evaporator air inlet temperature decline, a cabinet thermal mass term is added to
Equation 2.9. To simplify the calculations, the interior cabinet walls are assumed to follow the internal air
temperature, and the outside cabinet walls remain at the ambient temperature. The temperature profile of the wall is
10
assumed to be linear, thus the average temperature of the cabinet is the average of the inside and outside air
temperatures (Sugalski et al, 1991). With these simplifying assumptions, Equation 2.11
mc v
dT 1
+ mc p
dt 2
( ) dT Ý
cab dt
= Q air
2.11
is used to model the cabinet, in which (mc p)cab represents the thermal mass of the cabinet. Setting (mc p)cab to 11
Btu/°F for this particular cabinet, the predicted evaporator air inlet temperature agrees with the measured evaporator
air inlet temperature to within 2 °F, as shown in Figure 2.3. This calculation shows that the cabinet can be modeled
using a simple lumped capacitance equation. Improvements can be made by realizing that some parts of the cabinet,
for example the shelves, will undergo a larger temperature swing than the walls. They, like the various materials in the
walls, can be modeled as a separate lumps for improved accuracy.
A major portion of the evaporator load for both steady state and cycling cases is the heat that leaks in
through the cabinet walls from the ambient air. Therefore, it is important to investigate the effects of cycling on
cabinet conductance. Both the fresh food and freezer compartment cabinet conductance's, UA frig and UA frez
respectively, are usually determined using a reverse heat leak test, as detailed in Appendix B.
A common assumption of quasi-steady state models that predict transient performance is that the cabinet
conductances are constant throughout the on- and off-cycles. During the off-cycle, the evaporator fan shuts off,
and there is no longer forced convection in the freezer compartment. This leads to a decrease in UA frez. UA frig is less
affected, since the fresh food compartment air flow rate is typically only 15 percent of the total evaporator fan air flow
rate.
The affect on UA frig and UA frez of varying air flow rates was investigated by changing the air circulation fans
used in the reverse heat leak tests. Using a 105 cfm muffin fan in the fresh food compartment resulted in a UA frig of
1.011 W/°F. Switching to a 15 cfm fan dropped UA frig by 11 percent to 0.898 W/°F. The results in the freezer
compartment were more pronounced. Using the 105 cfm muffin fan in the freezer compartment (with the evaporator
fan off) resulted in a UA frez of 0.433 W/°F, and using the 45 cfm evaporator fan yielded a UA frez of 0.530 W/°F. This
22 percent increase is attributed to the different air flow patterns behind the evaporator. This experiment shows that
the internal air flow rates significantly affect the cabinet UAs. The decrease in cabinet conductance during the off-
cycle may be as high as 20 percent, as evidenced by this experiment. To properly account for the cabinet load in a
cycling model, different values for UA frig and UA frez are required depending on whether or not the evaporator fan is
running.
11
Chapter 3: Refrigerant Solubility in Compressor Oil
3.1 Introduction
The purpose of this chapter is to estimate the magnitude of the movement of refrigerant to and from the oil
in the compressor sump during transient operation. Refrigerant vapor enters the compressor shell and interacts
freely with the oil in the sump before being compressed and expelled to the condenser. The amount of refrigerant
dissolved in the oil changes as the pressure and temperature vary, thereby varying the amount of charge available to
the rest of the system. The Grebner-Crawford model is used to estimate the solubility of the refrigerant in the oil for
typical operating conditions. Following is an analysis for a case in which the refrigerator was charged with 12.5
ounces (360g) of R-12, and the ambient temperature was 75°F. This particular data is from an experiment in which the
refrigerator was intentionally overcharged.
12
70
60
50
Pressure (psia)
40
30
20
10
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
160
140
120
100
80
60
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
140
130
120
Temperature (°F)
110
100
90
80
70
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
13
3.3 Grebner-Crawford Refrigerant/Oil Mixture Model
The Grebner-Crawford equations (1992) empirically model an equilibrium mixture of refrigerant and oil. They
are based on measurements taken at the ACRC for mixtures of R-12 and mineral oil over a range of temperatures and
pressures which include refrigerator operating conditions. The equations, 3.1 through 3.4, take the form:
x2
A = x1+ 3.3
w1/2
x4 x5 x6 x7
B = x3 + + + + 3.4
w1/2 w w3/2 w2
where T*, the nondimensional superheat for the mixture, is a function of the mixture temperature, T[°R], and the
saturation temperature, Tsat(P) [°R], of the pure refrigerant at the mixture vapor pressure, P [psia]. The liquid
refrigerant mass fraction, w, is the mass of liquid refrigerant divided by the total mass of the liquid mixture. The
constants x1 through x7 were determined by Grebner and Crawford from experimental data using least squares curve
fits. For a mixture of R-12 and Naphthenic oil, the constants are given in Table 3.1.
The model is intended to predict equilibrium temperature (knowing w and P) or equilibrium pressure
(knowing w and T). Using the model to predict pressure in an R-12/Naphthene mixture, for example, results in a root
mean square error of only 2.23 psia over a range of 0 to 550 psia. The model can also be used to predict liquid
refrigerant fraction (knowing T and P). Unfortunately, the accuracy here slips to ± 10% (the mass fraction is known
to within ±0.10). Despite this, it should still be able to predict trends in refrigerant solubility in the compressor oil.
The compressor used in this analysis was a Tecums eh model AE1390V. It was lubricated by 13 ounces of
grade 32 Naphthenic mineral oil (Sanvordenker, 1992). Knowing the mass of the oil and the mass fraction of
refrigerant in the oil, the amount of refrigerant dissolved in the oil can be estimated using the Grebner-Crawford
model. This is plotted for the various temperatures and pressures normally found in the sump in Figure 3.4. The plot
looks nearly identical for R134a/ester oil mixtures.
14
15 psia
25 psia
35 psia
20 45 psia
55 psia
65 psia
75 psia
Mass of R-12 in Oil (ounces)
15 85 psia
95 psia
105 psia
115 psia
125 psia
10 135 psia
145 psia
155 psia
0
75 95 115 135 155 175
Temperature (°F)
When at rest, the pressure and temperature were measured to be roughly 67 psia and 75°F. This
corresponds to about 9 ounces of refrigerant, out of a total of 12.5 ounces, dissolved into the oil. During regular
cycling, the pressure varied from roughly 15 to 30 psia and the temperature varied from roughly 110 to 130°F. This
corresponds to about 1 ounce of refrigerant in the oil, thus 8 ounces of refrigerant boiled out of the oil during initial
pulldown. During regular cycling, the pressure and temperature fluctuations were much smaller than during
pulldown, so the amount of refrigerant entering and leaving the oil is not as large.
10
Mass of R-12 in Oil (ounces)
0
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
15
At the equilibrium initial conditions, the Grebner-Crawford model predicts that more than 9 ounces of
refrigerant are dissolved in the oil. During the initial pulldown, nearly all of the dissolved refrigerant boils out of the
oil, and only 0.3 ounces remain. The bulk of this fluid motion is caused by the temperature change as the compressor
warms up, although the dropping pressure also contributes. This is easily shown by checking the model in Figure
3.4. During the following off-cycle, 0.4 ounces of refrigerant dissolve back into the oil. Again checking with Figure
3.4, it is clear that the solubility during regular cycling is primarily a function of pressure fluctuations, as the
temperature is relatively stable. During the following cycles, the model predicts that 0.4 ounces of refrigerant move to
and from the compressor oil.
3.5 Refrigerant Movement from Compressors with High Side Oil Sumps
Some compressors, including newer rotary models, have the oil sump on the high side. The oil sump in this
case sees the compressor's discharge pressure, which may lead to different trends in refrigerant migration. Currently
our data on a compressor with a high side sump is incomplete. The following analysis considers a hypothetical
rotary compressor and uses the same data obtained from the low side compressor, with one exception. Since rotary
compressors typically use about 25 % less oil than their rotary counterparts, the amount of oil is scaled down to 10
ounces. It is important to realize that the Grebner-Crawford model predicts liquid refrigerant mass fraction; changing
the amount of oil just scales the mass of dissolved refrigerant in Figure 3.4 proportionally. This hypothetical
configuration can be modeled using a similar analysis to determine if the magnitude of the movement is similar.
Since the evaporating and condensing temperatures of most refrigerators are similar, the pressures on either
side of the compressor will also be similar. The variation of pressure on the compressor high side with time is
measured and plotted in Figure 3.6. In comparison to the low side pressure plotted in Figure 3.1, the pressure swings
on the high side are much larger. This should lead to larger movements of refrigerant into and out of the oil.
The temperature fluctuation of the oil is likely to be similar to the case when the oil sump is on the low side.
As an approximation, the average compressor shell temperature, as plotted in Figure 3.3, can once again be taken as
the oil's temperature. Previously, with a low side oil sump, temperature fluctuations had little effect on solubility due
to the low pressures; however, the conditions at the high side of the compressor shell are such that the solubility is a
strong function of both temperature and pressure (see Figure 3.4). It is now important to get an accurate
measurement of the oil temperature to correctly model the solubility behavior.
16
180
160
Pressure (psia) 140
120
100
80
60
40
20
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
Using the temperature and pressure fluctuations outlined above, the predicted mass of refrigerant dissolved
in 10 ounces of oil is shown in Figure 3.7. During the initial pulldown, some of the refrigerant boils out of the oil as
before, because of the temperature change. This time however, the pressure driving force opposes the temperature
driving force, and less refrigerant is boiled out of the oil.
There is a strong discontinuity in the model's predictions just after the compressor turns on. The Grebner-
Crawford equilibrium correlation yields an absurd result: more than 100% of the refrigerant is in the oil! This indicates
that the system does not move immediately to an equilibrium state upon startup. As the compressor warms up, it
tends to d rive refrigerant out of the oil, and after 91 minutes the system reaches a near-equilibrium state in which
about 6.8 ounces of refrigerant remain dissolved in the oil. The pressure at the end of this first on-cycle is 148.3 psia
and the temperature is 135.5 °F. This can be compared with steady state data for which the pressure is 147.1 psia and
the temperature is 133.3 °F. Since the steady state condition is by definition an equilibrium condition, we can accept
the model's prediction that 7.3 ounces of refrigerant are dissolved in the oil. Perhaps this is why systems with rotary
compressors require a larger refrigerant charge. When the compressor cuts off, there is again a sharp discontinuity,
and about 10 minutes into the 25 minute off-cycle, the pressure and temperature approach an equilibrium that
corresponds to about 0.5 ounce of refrigerant remaining in solution.
17
12
10
Mass of R-12 in Oil (ounces)
8
0
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
During the subsequent cycles, there are several immediate differences in the solubility trends. First, the
direction of refrigerant movement is in the opposite direction for the two types of compressors. Second, with a high
side sump, the refrigerant/oil mixture never reaches a steady state pressure and temperature during the on-cycle.
Because of the compressor's thermal mass, the sump oil is still getting warmer at the end of the on-cycle. Thus it is
not possible to know from available data with certainty whether the oil and refrigerant have reached equilibrium
during the on-cycle. It is therefore possible that the actual movement of refrigerant to and from the oil is much
smaller than the model's predictions, which represent the maximum possible magnitude. There could be a substantial
difference in the amount of refrigerant dissolved in the oil for steady state conditions versus cycling conditions.
Additional experiments are being designed to test this.
The worst case would occur if the adjustment to equilibrium is rapid. For the high side sump, as much as 6
ounces of charge move into and out of the oil during regular cycling, compared to 0.4 ounces for the low side sump.
This is strongly related to the temperature of the oil. If the oil temperature were higher, the magnitude of refrigerant
movement would be much smaller.
It is important to realize that the Grebner-Crawford model applies to equilibrium conditions. It is not
intended to predict transient behavior of mixtures, which explains the large predicted mass swings immediately
following compressor turn-on or turn-off in Figure 3.7. The mixture has not reached an equilibrium state and the
model predictions are invalid. At the end of on or off cycles, it is believed that the mixture is approaching equilibrium,
and that the model predictions, especially for the case of a low side sump, are correct. A future calculation to verify
equilibrium conditions will commence shortly. Using void fraction formulae, the mass of refrigerant in all system
components can be estimated. These will be summed with the refrigerant/oil model's predictions to get a total
refrigerant mass. Of course this mass should be a constant, and its deviation would indicate nonequilibrium
conditions.
18
Chapter 4: Refrigerant Migration
4.1 Introduction
When a refrigerator shuts off during its cycle, refrigerant migrates through the capillary tube from the
condenser to the evaporator until the pressures equalize. This chapter will outline two possible modes of refrigerant
migration and discuss the resulting heat transfer. Evidence will be presented to match data from an actual refrigerator
to one of these modes. For this analysis, the refrigerator was intentionally overcharged from 8.5 to 11.5 ounces,
however qualitative results should be similar for an optimally charged refrigerator. For the first data set, the
temperature control was set to produce 60 minute cycles in a 90 °F ambient room. A second data set was obtained in
a 60 °F ambient room using an 80W heat load to maintain a 60 minute cycle length.
200
Evaporator
Pressure (psia)
a
150
a'
100
Condenser 50
b'
b
0
0 20 40 60 80 100
Enthalpy (BTU/lb)
19
The second migration mode is shown in Figure 4.2. Rather than maintaining a liquid seal at the condenser
exit, vapor enters the capillary tube and migrates into the evaporator. The refrigerant initially follows path a-b in the
accompanying pressure-enthalpy diagram, and at a later time step, path a'-b'. Clearly, the enthalpy of the refrigerant
entering the evaporator is higher during vapor migration than it is for liquid migration. The energy implications of
this difference will be analyzed in subsequent sections of this chapter.
The thermal mass of the suction line/capillary tube heat exchanger is estimated to be only 0.033 Btu/°R.
During the off-cycle, the temperature of the evaporator end of the heat exchanger rises from 40 to 65 °F while the
condenser end stays at a constant 90 °F. This corresponds to only 0.4 Btu change of energy stored in the heat
exchanger. Some of this energy comes from the surrounding room air, and the remainder from the refrigerant flowing
in the capillary tube. This is unlikely to cool the migrating refrigerant substantially, thus the assumption of an
adiabatic capillary tube for the migration model is valid.
250
Evaporator 200
Pressure (psia)
150 a
100 a'
50
Condenser b'
b
0
0 20 40 60 80 100
Enthalpy (BTU/lb)
20
is initially cooler than refrigerant in the capillary tube. Also, taking into account the acceleration of the refrigerant
through the capillary tube would further decrease the temperature of the entering refrigerant.
Freezer Air
40 Measured Evaporator Inlet
Vapor Migration (max)
30 Liquid Migration (max)
Two-phase
Temperature (°F)
20
10
-10
-20
-10:00 0:00 10:00 20:00 30:00 40:00
Time(m:s)
It is interesting to note that the maximum entrance temperature for the liquid migration case is exactly the
same as the two-phase temperature at the evaporator pressure. As saturated liquid migrates through the capillary
tube, it remains two-phase, hence its temperature is uniquely defined by the pressure.
Almost immediately, the evaporator inlet and the two-phase temperatures become warmer than the
surrounding freezer air, and stay that way for the rest of the off-cycle. Although they are not plotted, the air
temperatures measured just above and below the evaporator are also lower than the refrigerant temperature at the
evaporator inlet. During most of the off-cycle, these air temperatures are within 2 °F of the average freezer air
temperature. Also not plotted is the evaporator exit temperature, which, like the inlet temperature, is warmer than the
surrounding air temperature for most of the off- cycle. All of this indicates a reversal of heat transfer direction;
during the off-cycle heat is transferred from the evaporator to the freezer air. This is consistent with the
measurements made by Staley et al (1992) on a General Electric refrigerator.
Figure 4.4 shows the measured evaporator and condenser pressures during the same off-cycle. Since the
pressures equalize about 8 minutes into the off-cycle, migration must be completed by this time. An interesting
feature of the evaporator pressure is that it first increases as expected during the migration, but then drops slightly
after the pressures have equalized. Since the mass in the evaporator is essentially fixed once the pressures have
equalized, it is likely that this small drop in pressure is caused by the heat transfer from the evaporator to the freezer
air.
21
50
45 Condenser
Evaporator
Pressure (psia) 40
35
30
25
20
15
10
-10:00 0:00 10:00 20:00 30:00 40:00
Time (m:s)
Figure 4.4 Condenser and evaporator off-cycle pressures for 90 °F ambient case
For the initial two minutes of the off-cycle, the data are consistent with the hypothesis of liquid migration,
since the maximum possible entering refrigerant temperature is slightly greater than the measured evaporator inlet
temperature. For the next six minutes only vapor migration is possible, because the maximum refrigerant temperature
for liquid migration is below the measured inlet temperature. After this point, the pressures in the evaporator and
condenser have essentially equalized, and migration ceases.
Figure 4.5 shows condenser temperatures during the off-cycle. The air temperature is measured near the
middle of the condenser at the rear of the refrigerator. Air temperatures are measured at other locations near the
condenser, and the trends are similar. The measurement of the refrigerant temperature at the condenser inlet is
heavily influenced by the compressor shell temperature because of the proximity of the thermocouple. The
compressor shell (and hence the refrigerant at the condenser inlet) is warmer than the surrounding air throughout the
off-cycle. This results in heat transfer from the compressor and this part of the condenser to the air. At the exit the
opposite is true. The air is warmer than the condenser, so heat is transferred to this section of the condenser. Also
plotted in Figure 4.5 is the saturation temperature at the condenser pressure. Once the compressor stops, the
contents of the condenser are believed to be two-phase until a majority of the charge has migrated out. After this
point, the remaining refrigerant is superheated vapor. Within two minutes of the start of the off-cycle, the two-phase
temperature falls below the surrounding air temperature. This indicates that heat is being transferred from the air to
the refrigerant as it boils. The temperature in no part of the condenser falls as low as the two-phase line in Figure 4.5
indicates, because the condenser becomes superheated before the pressure equalizes with that of the evaporator.
22
Air
Condenser Inlet
180 Condenser Exit
Two-phase
135
Temperature (°F)
90
45
0
-10:00 0:00 10:00 20:00 30:00 40:00
Time (m:s)
It is difficult to describe the temperature of the condenser during the off-cycle because it is non-uniform.
The coolest part is probably the bottom set of tubes, where the liquid part of the two-phase refrigerant drains.
Unfortunately, no thermocouples exist in this location to verify this. The thermal mass of the compressor keeps the
inlet the warmest part of the entire condenser, and the thermal mass of the condenser itself is probably what keeps
the exit at a nearly constant 100 °F. Migrating vapor traveling through the condenser coil has a low heat capacity,
and is unlikely to alter the coil temperature significantly. Instead, the vapor temperature rises from its saturation
temperature to match the coil temperature at the exit. This was verified through a surface temperature measurement at
the condenser exit.
These observations provide solid evidence for vapor migration during the period from two to eight minutes
into the on-cycle, but don't indicate which mode(s) of migration occur for the first two minutes. At least initially, it is
likely that liquid migration occurs because the condenser exit is subcooled at the end of the on-cycle. From
geometrical considerations, it is estimated that as much 1 ounce of liquid migrates before the liquid seal is broken,
and vapor migration takes over. A computer simulation has been written to model the migration process, and it is
presented in the next section.
23
during regular refrigerator operation, and is initially true during the off-cycle migration process. Under choked flow
conditions, the receiver (evaporator) pressure has no effect on the mass flow rate. For a given capillary tube
geometry, the mass flow rate becomes solely a function of the condenser pressure, as given by Equation 4.1,
γ −1 2
1+
1 1 1 γ + 1 M2out
M out
fL γ + 1
= ln 2 − − − ln
D 2γ 1 + γ − 1 M 2 γ M 2out M 2in 2γ M2in
2
in
4.2.a
γ
γ −1 2 γ −1
P cond = P in 1+ Min 4.2.b
2
1
γ −1 2 2
Pout Min 1 + 2 M in
=
Pin M out 1+ γ − 1 M 2
2 out
4.2.c
Pout = Pevap 4.2.d
capillary tube, L, is measured to be 125.5 inches long, and its diameter, D, is given by the manufacturer to be 0.033
inches. The friction factor, f, is estimated from a Moody diagram to be 0.03 throughout most of the flow conditions.
The specific heat ratio, γ, is assumed to be a constant 1.104 for all calculations. Throughout this discussion, the
subscript "in" represents the inlet to the capillary tube, likewise the subscript "out" represents the exit. The
subscript "cond" refers to conditions within the condenser where the fluid velocity is neglected, and "evap" refers to
similar conditions in the evaporator. Finally, Mach numbers are designated with M, absolute pressures with P, and
absolute temperatures with T. Equations 4.2.a and 4.2.c are Fanno flow relations, and Equation 4.2.b is an isentropic
relation. Equation 4.2.d represents a condition that must be met in order to have subsonic (unchoked operation) flow
throughout the capillary tube.
24
Equations 4.3 - 4.5 can now be solved sequentially to determine the mass flow rate.
Tcond
= 1+ γ −1 M 2in 4.3
Tin 2
c in = γRT in 4.4
ρ M c A
Ývap = in in in
m 4.5
3600
Equation 4.3 is an isentropic relation which provides the temperature at the inlet to the capillary tube. This
is used in Equation 4.4 to calculate the speed of sound, cin, of the refrigerant vapor. The ideal gas constant for R-12,
R, is 411 ft 2/sec 2°R. Finally, the mass flow rate of the refrigerant vapor, m
i vap, is calculated in Equation 4.5 in lbs/hr.
In this equation, the density of refrigerant vapor entering the capillary tube, ρ in, is a thermodynamic function of
temperature and pressure, which have already been determined. The cross-sectional area of the capillary tube
entrance in ft 2 is denoted by A.
A better estimate of the refrigerant temperature entering the evaporator can be obtained from Equation 4.6,
which is also a Fanno relation.
γ −1 2
Tout 1+ 2 Min
=
Tin 1 + γ −1 M2
2
out
4.6
Once liquid and vapor mass flow rates are known functions of temperature and pressure, a model can be
developed to study refrigerant migration. Using the basic schematic presented in section 4.2, an unsteady energy
balance, Equation 4.7, can be written for the condenser control volume. The internal energy of the control volume,
Econd cv, can be written as shown in Equation 4.8. The condenser in this particular refrigerator is made entirely of steel,
and all of its
dEcond cv
= QÝcond cv − m
Ýhout
dt 4.7
Econd cv = msteel cpTcond + mrefu ref 4.8
dimensions are known from the manufacturer. This makes it a simple matter to calculate its thermal mass, msteel cp,
which in this case is 0.435 Btu/°F. Tcond is the temperature of the saturated refrigerant, and also represents the
temperature of the condenser itself. Even though the condenser temp erature is clearly non-uniform, using Tcond to
represent both the refrigerant and coil temperatures significantly simplifies the calculations. The mass of refrigerant
in the condenser, mref, is initially 0.375 lbs., and it decreases as the refrigerant migrates out of the condenser. The
internal energy of the refrigerant, u ref, is a thermodynamic function of temperature and quality, x, which is in turn
defined by equation 4.9,
V - m ref vliq
x=
(
m ref vvap − vliq ) 4.9
25
which is derived from the definition of quality. The volume of the vessel (in this case the condenser), V, is estimated
from manufacturer specifications to be 0.007666 ft 3, and the specific volumes of saturated liquid and vapor
refrigerants, v liq and v vap respectively, are thermodynamic functions of temperature and pressure. The rate of heat
? cond cv, which is given by Equation 4.10. Since the condenser fan is not
transfer to the condenser is represented by Q
dEevap cv
= −Q
Ý
evap cv + m
Ýh in
dt 4.11
The terms in this equation are defined in a similar manner as those used in the condenser control volume. It
is unimportant to know whether the energy added to the evaporator control volume (m
i h in) is stored within
? evap cv) or transferred to the surrounding air (Q? evap cv). Figure 4.3 shows that the evaporator is essentially
evaporator (E
in thermal equilibrium with the surrounding air by the end of the off-cycle. This means that the heat transfer between
the evaporator and the air is negligible. Therefore, the energy penalty associated with migration is just the integral
over time of m
i h in.
All of these equations were solved using a program which is listed in Appendix C. All derivatives are taken
numerically, and all thermodynamic properties are curve fits of table data written in the form of functions. The block
of simultaneous equations describing Fanno flow were solved using a Newton-Raphson subroutine. The initial
values for temperature and pressure were those measured at the end of an on-cycle. It is assumed that 1 ounce of
liquid migrates initially, and that all subsequent migration is vapor. A plot of the model's predicted condenser
pressure variation with time is compared with the measured pressure in Figure 4.6.
26
180
Measured
160 Predicted
Liquid Seal
Pressure (psia) 140
120
100
80
60
40
20
0 100 200 300 400 500
Time (sec)
Also shown in Figure 4.6 is the hypothetical case for which a liquid seal is maintained as long as possible at
the condenser exit.
The migration model agrees very well for the beginning of the off-cycle, however it does not predict the
transition from a two-phase to a superheated condenser well. The model shows this transition as a distinct change of
slope at 375 seconds. In the actual condenser this transition may not be as sharp. While the model assumes that the
entire condenser is of uniform temperature, it actually varies in different locations as discussed previously. Although
no thermocouples exist near the middle of the condenser, it is likely that the temperature here drops substantially
lower than in the rest of the condenser as the refrigerant inside boils off. This non-uniformity is difficult to model,
and is likely to be a major source of error in the model. Also, once the liquid in the condenser is gone, the pressure is
determined using a constant compressibility factor in conjunction with the ideal gas law. The compressibility factor
used in the calculations is 0.83, and is obtained from a generalized compressibility chart for temperatures and
pressures typically measured at the condenser exit during the off-cycle. This is likely to introduce significant error in
the pressure calculation, since the compressibility factor varies substantially with superheat near the saturation
dome. A better curve fit for pressure as a function of density and pressure could alleviate this shortcoming. There is
excellent agreement during the vapor migration portion of the model while the condenser is clearly two-phase. This
helps to verify the estimated mass flow rate, and the total mass that has migrated, which is important in determining
the energy losses associated with the migration process. Also, the model indicates that vapor migration is likely to
begin very early in the off-cycle.
If a liquid seal is maintained as long as possible, the migration process takes place much faster because the
mass flow rate of liquid is much higher than the mass flow rate of vapor. To model this case, liquid migration is
assumed until only 1 ounce of refrigerant remains in the condenser. At this point, the liquid seal is broken and vapor
migration takes over, as indicated by the change in slope 100 seconds into the off-cycle. After about 40 seconds of
vapor migration, the condenser is no longer two-phase, and the pressure drops rapidly to its equilibrium value. A
perfect liquid seal would allow all but 0.15 ounces to migrate as liquid, but this is probably unattainable in practice.
The horizontal tubing used in the condenser is likely to allow refrigerant vapor to migrate well before all the liquid is
27
gone. If the exit was at the bottom of a vertical tube, it might be possible to maintain the liquid seal for almost the
entire off-cycle.
28
Freezer Air
40 Measured Evaporator Inlet
Vapor Migration (max)
Liquid Migration (max)
30 Two-phase
Temperature (°F)
20
10
-10
-20
-10:00 0:00 10:00 20:00 30:00
Time (m:s)
50
45 PCondOut(psia)
40 PEvapOut(psia)
Pressure (psia)
35
30
25
20
15
10
-10:00 0:00 10:00 20:00 30:00
Time (m:s)
Figure 4.8 Condenser and evaporator off-cycle pressures for 60 °F ambient case
An important difference between the two data sets is the direction of heat transfer during the off-cycle.
Figure 4.7 shows that the freezer air temperatures is at all times greater than the evaporator temperature. This results
in heat transfer from the air to the refrigerant in the evaporator, producing a beneficial cooling effect. Since the
evaporator inlet refrigerant temperature is near the saturation temperature for the second half of the off-cycle, it is
likely that the refrigerant is boiling during this time.
29
vapor, all of the stored energy plus the energy energy added to the refrigerant from the freezer air is removed at the
same time. This additional refrigerating effect results in an efficiency increase. The refrigerating effect, ∆Qrefrig, can
be estimated using Equation 4.12, where UA e is the natural convection coefficient
(
∆Q refrig = ∫ UA e T air − Tevap dt ) 4.12
of the evaporator. Tair and Tevap are the air and evaporator temperatures, and t is the time during which heat transfer
occurs. Using a rough estimate of 2 Btu/h °R for UA e , the total refrigerating effect is only 3 Btu, which is negligible
compared to the 362 Btu cabinet load for the entire cycle. Thus, the migration process itself does not impose an
additional load on the refrigerator for the 60 °F ambient case.
4.8 Conclusions
The migration process itself does not appear to have much impact on the overall evaporator load during
cycling. However, having the surplus charge in the evaporator at the beginning of the on-cycle may slow the return
to a quasi-steady operating condition. The next chapter will focus on the phenomena associated with startup. It is
believed that the evaporator is nearly evacuated by the compressor as it first starts, and that the starved evaporator
causes a loss of efficiency until charge is redistributed through the capillary tube. If this is true, having extra charge
in the evaporator at the beginning of the on-cycle may reduce this starving effect. Conversely, if a surplus of charge
remains in the evaporator during the initial part of the on-cycle, it will take longer for the evaporating temperature to
drop. This too will decrease efficiency as it takes longer for the evaporator heat transfer rate to reach its quasi-steady
value.
30
Chapter 5: Charge Redistribution
5.1 Introduction
Immediately following the compressor startup, several transients keep the refrigerator from operating at
steady state. One reason for this is the maldistribution of charge throughout the system. By the end of the off-cycle,
the evaporator may contain up to 6.5 ounces of saturated refrigerant (see Chapter 4). About 5.5 ounces of this must
be redistributed to the other components in order to achieve quasi-steady state operation. As the compressor starts
up again, this charge may begin to leave the evaporator gradually as a vapor, or as slugs of saturated liquid. If slugs
of liquid refrigerant are drawn out of the evaporator into the compressor, the evaporator may reach its steady state
charge rather quickly. Furthermore, if the slugs vaporized upon entering the warm compressor, the resulting lower
discharge/suction pressure ratio would increase the mass flow rate through the compressor, thus further accelerating
the redistribution of charge to the condenser. However, if the slug flow from the evaporator is too vigorous, it could
starve the evaporator of refrigerant. This would result in high superheat levels and an accompanying loss of
capacity until a steady state distribution is reached.
If there is an accumulator at the evaporator exit and it is oriented properly, much of the liquid refrigerant
could be retained in the evaporator, and only vapor would exit. In this case the liquid would have to be boiled off
resulting in heat transfer from the freezer air to the refrigerant. The cooling effect, ∆Q, can be calculated using
Equation 5.1. The mass of the liquid refrigerant
∆Q = mh fg 5.1
in pounds is given by m, and h fg is the enthalpy of vaporization at the evaporating temperature, in this case 68 Btu/lb.
If, for instance, 5 ounces of refrigerant are boiled off during the first part of the on-cycle, the cooling effect is 21 Btu.
During that time, however, there will only be vapor leaving the evaporator, and it will take longer for the saturation
pressure and temperature to drop, and for the rate of heat transfer from the freezer air to increase.
The instantaneous amount of charge in the evaporator is determined by the difference between the flow
rates into and out of the evaporator. The flow rate into the evaporator from the capillary tube is initially much lower
than its steady state value, because the exit of the condenser contains superheated vapor at the beginning of the on-
cycle. The vapor flow rate is approximately one fourth the liquid flow rate through the same pressure difference (see
Chapter 4). As the charge builds up in the condenser, the exit quality drops and the mass flow rate through the
capillary tube increases.
This chapter examines the redistribution of charge and its effects on refrigerator performance. For this
experiment, the refrigerator was intentionally overcharged from 8.5 to 11.5 ounces; however, qualitative results
should be similar for an optimally charged refrigerator. Data were taken in both a 90 °F and a 60 °F ambient room, and
analyzed separately.
31
two-phase exit. The data show that it is unlikely that all of the charge slugged out as a liquid. This would have
starved the evaporator of refrigerant, and forced it to operate with a high level of superheat until more charge was
brought in through the capillary tube.
Figure 5.1 shows the refrigerant temperature at the evaporator exit, as well as the saturation temperature at
the measured exit pressure, for an on-cycle beginning at time 0:00. If the measured evaporator exit temperature
differed from the saturation temperature, it would indicate that either superheated vapor or subcooled liquid must
have flowed out of the evaporator. Except for the pressure spike about 10 seconds into the on-cycle, the refrigerant
temperature closely followed the saturation temperature. Unfortunately, this does not indicate whether liquid or
vapor exited the evaporator. However, it is likely that the pressure spike was caused by a slug of liquid passing the
pressure tap. During a subsequent cycle, a pressure spike was also observed, but this time it was in the opposite
direction. The data acquisition system used in this study was limited to a ten-second sampling interval. For future
work it will need to be modified to obtain a finer resolution to better observe pressure spikes and other sharp
transients.
20
Evaporator Exit
15 Compressor Inlet
Saturation Temperature
Temperature (°F)
10
-5
-10
-1:00 0:00 1:00 2:00 3:00 4:00 5:00 6:00
Time (m:s)
In order to gain additional insights, a small microphone was placed near the accumulator exit to check if
liquid slugs could be heard passing it. The microphone was embedded in a small hole in a piece of foam insulation
such that its face was pressed against the tubing. This assembly was wrapped with a second layer of foam to
provide further insulation from outside noise. The microphone was fed into a standard audio tape recorder which
was turned on just before the beginning of an on-cycle.
The dominant sounds on the tape are the compressor and fan noises; however, with the aid of an audio
equalizer, much of this can be filtered out. During the beginning portion of the off-cycle, liquid refrigerant can be
heard splashing inside the evaporator. This activity ceases approximately midway into the 25 minute off-cycle.
During the first ten seconds of the on-cycle, several quick gurgling noises can be heard which probably indicate the
movement of liquid slugs. Between 40 and 60 seconds into the on-cycle, sharp pinging or knocking sounds are
heard, possibly the result of liquid entering the compressor. This is supported by Figure 5.1, which shows that the
32
refrigerant at the compressor inlet is saturated at that time. The information obtained by the microphone supports the
hypothesis that several liquid slugs exited the evaporator, although it does not confirm it, nor does it give any
indication of how much liquid might have slugged out before vapor began to flow.
The refrigerant temperature at the compressor inlet dropped from nearly 100 °F to 5 °F during the first 30
seconds of the on-cycle. The refrigerant became saturated and remained that way until the capillary tube/suction line
heat exchanger (interchanger) began to work (about 70 seconds into the on-cycle), after which the refrigerant in the
suction line was clearly superheated. During the first 60 seconds of the on-cycle, the suction line surface
temperature dropped nearly linearly from 60 °F to 5 °F at the evaporator end, and from 90 °F to 40 °F at the
compressor end. The energy stored in the interchanger was primarily removed by the refrigerant flowing through the
suction line, since the mass flow rate through the capillary tube is initially very low.
During the first 30 seconds of the on-cycle, both the refrigerant and the suction line surface temperatures
changed rapidly. Much of the refrigerant that exited the evaporator must have been vapor since its temperature rose
as it cooled the suction line. The thermal mass of the interchanger, 0.033 Btu/°F, is estimated by multiplying its mass
by the specific heat of copper. In order for the abrupt temperature change to have occurred, the refrigerant must
have absorbed about 1 Btu from the interchanger during the first 30 seconds of the on-cycle. Assuming a typical
mass flow rate of 15 lbs/hr (based on the compressor map at the measured evaporating and condensing pressures),
about 2 ounces of saturated refrigerant vapor could have flowed from the evaporator. In order for this vapor to
absorb the 1 Btu of energy stored in the interchanger, it would have to undergo an average temperature change of 55
°F. This can be compared to the time-averaged temperature change that was measured during the first 30 seconds of
the on-cycle, which for this case was 47 °F. Since the measured temperature change is very close to the temperature
change that would have to occur for vapor-only flow, it is believed that most of the refrigerant exited the evaporator
as saturated vapor. This is further confirmed by comparing the observed refrigerant-side convective heat transfer
coeficient with the theoretical vapor-only coefficient obtained from the Dittus-Boelter equation. The details of this
calculation are found in Appendix D.
Between 30 and 60 seconds into the on-cycle, the refrigerant temperature at the compressor inlet stabilized
to the saturation pressure, while the suction line surface temperature continued to drop rapidly as it rejected another
1 Btu of stored energy. During this period, liquid slugs evidently exited the evaporator and vaporized as they
absorbed energy from the suction line. If saturated vapor were exiting the evaporator, it would have superheated and
continued to rise in temperature, which is inconsistent with the data in Figure 5.1.
Evidently, during the first 30 seconds of the on-cycle, most of the charge leaving the evaporator was
saturated vapor. For the next 30 seconds, predominantly liquid slugs exited the evaporator, and they vaporized as
they continued to remove heat from the suction line. One explanation for this is that initially the liquid slugs were
trapped by the accumulator, and mainly saturated vapor flowed through the suction line. After the accumulator filled
up, liquid began to spill out, and several slugs flowed into the suction line. About 70 seconds into the on-cycle, the
interchanger began to work, and any refrigerant flowing through the suction line was superheated before it could
reach the compressor.
33
5.3 Evaporator Performance for 90 °F Ambient Case
? evap, is most easily calculated using an
For a transient analysis, the evaporator heat transfer rate in Btu/hr, Q
air side energy balance as indicated by Equation 5.2. The volumetric flow
60VÝc p air ∆T
Qevap =
Ý
v 5.2
? , is known to be a constant 68 cfm (Admiraal and Bullard, 1993), and the specific volume
rate of the evaporator fan, V
of air, v, is calculated using the ideal gas law. The specific heat of air, cp air , is taken at the average air temperature
around the evaporator, and ∆T is the measured air temperature difference through the evaporator in °F.
? evap climbed
The evaporator heat transfer rate is plotted for the first fifteen minutes of the on-cycle in Figure 5.2. Q
rapidly for the first four minutes before it approached its average quasi-steady value of about 552 Btu/hr. The
surplus of charge in the evaporator caused it to operate with a saturated exit throughout the duration of the on-cycle.
800
700
600
QEvap (Btu/hr)
500
400
300
200
100 QEvap (air side)
0
0:00 5:00 10:00 15:00
Time (m:s)
Initially, the evaporator was in thermal equilibrium with the surrounding air as shown in Figure 5.3. With no
temperature difference, no heat transfer can occur. This explains the low evaporator capacity at the beginning of the
on-cycle. As the surplus charge slowly boiled off, the pressure and the saturation temperature of the evaporator
declined, and heat transfer between it and the air increased. Warm air drawn in from the fresh food compartment
caused the initial upward spike in evaporator inlet air temperature. Thereafter, the air temperatures declined
monotonically for the rest of the on-cycle.
34
Evaporator Inlet Air
20 Evaporator Exit Air
Saturation Temperature
15
10
Temperature (°F)
0
-5
-10
-15
-20
-5:00 0:00 5:00 10:00 15:00
Time (m:s)
The evaporator heat transfer rate became nearly constant within four minutes of startup, despite the
declining air temperature. This represents the balance points of the system, which are defined by the governing set
of simultaneous equations. From this point forward, the temp erature difference between the incoming air and the
evaporator remained nearly constant. The heat transfer coefficient was also constant, since the entire evaporator was
always flooded. The system maintained a constant evaporator heat transfer rate by dropping the evaporating
temperature.
The loss in capacity can be quantified by comparing the average evaporator heat transfer rate with the
"steady state" value. Over the entire on-cycle, the evaporator removed heat at an average rate of 539 Btu/hr. By
ignoring the first four minutes of the on-cycle during which the evaporator capacity was rising rapidly, the average
? evap was 552 Btu/hr. This indicates a 2.3 % loss in capacity as a result of the startup transient.
quasi-steady value of Q
The lower initial capacity would not be considered an efficiency loss if the system power was correspondingly lower
during the beginning part of the on-cycle. However, Figure 5.4 shows that the opposite was true. The compressor
power declined sharply at the beginning of the on-cycle because the evaporating pressure was dropping. Although
the condensing pressure was increasing during this time, which by itself would have increased power consumption,
the low-side pressure had a stronger influence on compressor power than the high-side pressure. This is verified by
the compressor calorimetry supplied by the manufacturer.
35
1200
1000
Power (Btu/hr)
800
600
400
Total System Power
200
0
0 5 10 15 20 25 30
Time (m:s)
The COP during the on-cycle is plotted in Figure 5.5. The plot shows that nearly all of the degradation of
COP occurred during the first four minutes of the on-cycle as the charge was being redistributed.
1
COP
0.8
0.6
COP
0.4
0.2
0
0:00 5:00 10:00 15:00
Time (m:s)
36
of the particular arrangement of liquid and vapor within the evaporator. Since the freezer air temperature was warmer
than the evaporator during the off-cycle, heat transfer from the air evidently superheated pockets of vapor near the
exit of the evaporator (see Chapter 4). At the start of the on-cycle, the exit immediately became two-phase as it did in
the previous case. This in itself cannot confirm whether liquid, vapor, or both flowed from the evaporator upon
startup.
20
Evaporator Exit
15 Compressor Inlet
Saturation Temperature
10
Temperature (°F)
0
-5
-10
-15
-20
-1:00 0:00 1:00 2:00 3:00
Time (m:s)
Once again a microphone was employed to gather additional insight about refrigerant flow throughout the
cycle. This time, the refrigerant was heard splashing inside the evaporator throughout the entire 20 minute off-cycle.
Since the evaporator and condenser pressures equalized, and the contents of the condenser became entirely
superheated (both within 10 minutes), it is likely that the off-cycle migration process was completed by the time the
30 minute on-cycle began. The charge that had migrated was evidently boiling and rearranging itself within the
evaporator coils. A different tape recorder was used this time to record the microphone signal, and unfortunately the
fidelity suffered. It became much more difficult to pick out the sound of flowing refrigerant even with the aid of an
audio filter. Once the compressor started, faint gurgling sounds were heard to indicate that liquid slugs were leaving
the evaporator. The sharp tapping sounds heard in the 90 °F ambient case were no longer discernible. Future tests
should employ a very high quality audio tape recorder with a noise filter to remove some of the microphone hiss. An
audio equalizer with 10 or more bands would be helpful to filter out compressor noise, which dominated the tape. It is
unlikely that further insulating the microphone would reduce the noise level, since the sound is transmitted directly
through the metal tubing.
As it did before, the refrigerant at the compressor inlet became saturated, although for a shorter time than in
the 90 °F case. This is another indication that slugs of liquid may have been drawn out of the evaporator.
37
5.8 shows the evaporator air and refrigerant temperatures for this particular cycle. The refrigerant temperature at the
evaporator inlet was the same as the saturation temperature throughout the on-cycle, indicating that at no time was
the entire evaporator filled with superheated vapor; there was always sufficient charge to maintain a two-phase inlet.
It is likely that as the compressor started, it drew out enough charge to cause the exit to become superheated. Then,
the level of superheat continued to rise as the compressor mass flow rate led the capillary tube mass flow rate, and
the evaporator charge continued to decline. This is verified by Figure 5.9, which shows the mass flow rates into and
out of the condenser. These were obtained from a compressor map and a turbine flow meter respectively. The
turbine flow meter reading was corrected using a steady state offset of +1 lb/hr (Admiraal and Bullard, 1993). Since
the flow rate into the condenser was greater than the flow rate out, the condenser must have been collecting charge.
This is further verified by Figure 5.10 which shows a concurrent increase in condenser subcooling. Near the middle
of the on-cycle, the capillary tube mass flow rate (as measured by the turbine flow meter) rose above the compressor
mass flow rate, and charge began to accumulate back into the evaporator, thus reducing the level of superheat to its
quasi-steady value of about 3 °F. The evaporator was unable to operate with a saturated exit because the additional
charge it required was in the subcooled zone of the condenser.
1400
1200
1000
QEvap (Airside)
800
600
QEvap (air side)
400
200
0
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
38
Evaporator Inlet Air
Evaporator Exit Air
15 Evaporator Exit Refrigerant
Saturation Temperature
10
Temperature (°F)
-5
-10
-15
-20
-5:00 0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
25
20
Mass Flow Rate (lbs/hr)
15
10
5 Condenser Inlet
Condenser Exit
0
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
After the initial transients, the evaporator heat transfer rate again remained nearly constant as a result of the
balance points of the system. The declining level of superheat counteracted the declining temperature difference
between the evaporator and the incoming air. As the superheated area decreased, the two-phase area grew,
corresponding to an increase in the overall evaporator heat transfer coefficient. These opposing effects maintained a
nearly constant evaporator heat transfer rate.
39
100
80
Temperature (°F)
60
40
20 Condenser Exit
Saturation
0
-5:00 0:00 5:00 10:00 15:00 20:0025:00 30:00
Time (m:s)
Figures 5.8 through 5.10 indicate that the charge was redistributing throughout the on-cycle as the system
tried to maintain a balance point. The bulk of the charge redistribution (about 5 ounces) occurred during the first two
minutes of the on-cycle as the evaporator emptied. After this was completed, there was a slower charge
redistribution back into the evaporator (up to 0.5 ounces), which reduced the level of superheat. This secondary
redistribution had a much smaller effect on evaporator performance, as shown in Figure 5.7. It is unlikely that the
addition of the electric heaters used to provide the constant 80W heat load changed the characteristics of the
redistribution. Since the heat was well distributed throughout the cabinet, it was no different than the heat that
leaked through the cabinet walls from the ambient air. Thus, these conclusions should apply equally well for a
refrigerator without heaters.
The loss in evaporator capacity can be quantified as before. In this case, the evaporator reached full
capacity in only two minutes, compared to four minutes for the 90 °F ambient case. The average evaporator heat
transfer rate over the entire on-cycle was 488 Btu/hr, compared with the average quasi-steady value of 491 Btu/hr.
This translates into only a 0.6 % loss in capacity. The COP for this on-cycle is plotted in Figure 5.11. The plot clearly
shows most of the degradation in COP occurred during the first two minutes as the evaporator heat transfer rate was
low and the system power was high. The plot also shows a noticeable degradation in COP in the middle of the on-
cycle when the evaporator had a high level of superheat.
40
1.5
COP
1
COP
0.5
0
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
5.6 Conclusions
In both of the cases analyzed in this chapter, the evaporator operated at a reduced capacity until the charge
could be redistributed to its equilibrium location. Before cooling can begin, the temperature of the two-phase section
of the evaporator must drop below the surrounding air temperature. When the evaporator contains little charge as a
consequence of less off-cycle migration, or rapid emptying upon compressor startup, it can change temperature
quickly. This allows the evaporator to begin cooling very early into the on-cycle. For the 60 °F case, it is believed
that a similar amount of charge migrated to the evaporator during the off-cycle as for the 90 °F case. However, once
the compressor started, the evaporator became immediately starved. This could be a consequence of the distribution
of liquid refrigerant within the evaporator. As the refrigerant boiled during the off-cycle, it may have traveled
through the evaporator tubing towards the exit. Once the compressor started, large slugs of liquid exited thereby
starving the evaporator. This allowed the temperature of the remaining refrigerant to drop rapidly and begin
absorbing heat from the freezer air. In contrast, a flooded evaporator needs more time to boil off some charge before
its pressure and temperature can drop and heat transfer can begin. This is believed to cause the longer period of
reduced evaporator capacity for the 90 °F ambient case. Since no boiling occurred during the off-cycle in that case,
much of the refrigerant may have been trapped near the evaporator inlet, and fewer slugs of liquid were able to exit
the evaporator. Thermodynamically, it is beneficial to remove the excess charge from the evaporator as quickly as
possible, so that the quasi-steady evaporating temperature can be established, and full refrigerating capacity can be
attained. However, this benefit is offset by the risk to compressor reliability.
41
Chapter 6: Capillary Tube/Suction Line Heat Exchanger
6.1 Introduction
Most modern refrigeration equipment uses a capillary tube/suction line heat exchanger to lower the
refrigerant quality at the inlet to the evaporator, thereby increasing its capacity. The efficiency gains for an R-12
system are small (up to 5 percent increase in COP), however there are several other important benefits (Domanski et
al, 1992). By heating refrigerant in the suction line, the heat exchanger ensures that only superheated vapor reaches
the compressor, which helps prolong compressor life. It also heats the suction line above the ambient dew point
temperature to prevent moisture from condensing on it. The price for these benefits is a loss in compressor efficiency
as the suction temperature increases. For some refrigerants, such as R-22, the loss in compressor efficiency is greater
than the gain in evaporator capacity, so suction line/capillary tube heat exchangers are not used. The heat exchanger
also affects the mass flow rate through the capillary tube, which may impede charge redistribution during cyclic
operation. This chapter will outline the transient behavior of the capillary tube, and consider the effects of the
capillary tube/suction line heat exchanger on charge redistribution.
20
Superheat
Subcooling
16
12
°F
0
0:00 5:00 10:00 15:00 20:00 25:00 30:00
Time (m:s)
42
The effect on mass flow rate of varying degrees of condenser subcooling was analyzed using a finite
difference computer simulation of the capillary tube/suction line heat exchanger (Peixoto, 1993). The simulation was
run using inputs in accordance with experimental data. The evaporating pressure was set to 16 psia, and the
condensing pressure to 113 psia. The condenser subcooling was varied from 0 to 7 °F, and the evaporator superheat
was set at 5 °F. The results from this simulation are plotted in Figure 6.2. As expected, an increased level of
subcooling increases the mass flow rate through the capillary tube. The refrigerant entering the capillary tube is
slightly denser, and the flash point moves further up the tube. In this case, the mass flow rate increases by 0.33
lbs/hr for each additional degree of subcooling. By this mechanism, the system is able to allocate more charge to the
evaporator when it needs it, thereby achieving a mass flow balance with the compressor. It is important to note that
this effect ocurrs with or without the presence of a suction line/capillary tube heat exchanger. This is verified by
running the simulation with an adiabatic capillary tube.
13
12.5
Capillary Tube Mass Flow Rate
Mass Flow Rate (lbs/hr)
12
11.5
11
10.5
10
9.5
9
0 2 4 6 8 10
Subcooling (°F)
Figure 6.2 Effect of condenser subcooling on capillary tube mass flow rate
With certain evaporator/accumulator configurations, it is also possible that the compressor is unable to
empty the evaporator at the start of the on-cycle. In this instance, the evaporator operates with a surplus of charge
as the refrigerant slowly boils out. With excess charge in the evaporator, the condenser will have a low level of
subcooling, or perhaps a saturated exit. This reduces the mass flow rate through the capillary tube, thereby aiding
the charge redistribution. Again, this effect is independent of the presence of a capillary tube/suction line heat
exchanger.
43
heat transfer from the capillary tube and moves the flashing point upstream. This reduces the mass flow rate through
the capillary tube into the evaporator, thus it takes longer for the charge to redistribute and reduce the level of
superheat.
The effect on mass flow rate of varying degrees of evaporator superheat was also analyzed using the finite
difference computer simulation of the capillary tube/suction line heat exchanger (Peixoto, 1993). Again the simulation
was run using an evaporating pressure of 16 psia, and a condensing pressure of 113 psia. This time the condenser
subcooling was set at 5 °F, and the evaporator superheat was varied from 0 to 15 °F. The results from this simulation
are plotted in Figure 6.3.
As expected, the maximum mass flow rate through the capillary tube for the specified evaporating and
condensing pressures occurs with the minimum level of evaporator superheat, which corresponds to the coldest
possible suction line. With 15 °F of superheat, the mass flow rate drops from 11.2 to 10.5 lbs/hr. This corresponds to
a 0.05 lb/hr decrease in mass flow rate for each additional degree of superheat.
During the on-cycle of a refrigerator operating in 60 °F ambient room, a 22 minute period was observed
during which the evaporator had on average 10 °F of superheat above its quasi-steady value at the end of the on-
cycle (see Figure 5.8). Assuming that about 0.5 ounces of charge redistributed to reduce the level of superheating,
the average difference in mass flow rates at the evaporator inlet and exit was 0.085 lb/hr. Using the simulation results,
10 °F of superheat reduced the capillary tube mass flow rate by 0.5 lbs/hr. Thus, without the capillary tube/suction
line heat exchanger, the average difference in mass flow rates into and out of the evaporator would have been 0.585
lbs/hr. At this rate, the 0.5 ounces of charge could have been redistributed in as fast as 3.2 minutes. Clearly, the
capillary tube/suction line heat exchanger reduces the flow rate into the evaporator when it needs more charge.
12
11.2
10.8
10.4
10
0 5 10 15
Superheat (°F)
Figure 6.3 Effect of evaporator superheat on capillary tube mass flow rate
Conversely, when the evaporator contains a surplus of charge it operates with a lower quality exit than its
steady state value. The two-phase section of the suction line lengthens, resulting in more heat exchange with the
capillary tube. The additional heat transfer increases the capillary tube mass flow rate, which brings more charge into
the evaporator when it already has too much. Thus, the suction line/capillary tube heat exchanger impedes charge
redistribution in both cases, whether the evaporator is over- or under-charged.
44
6.4 Conclusions
The presence of the capillary tube/suction line heat exchanger slows the redistribution of charge to the
evaporator when it has too much or too little charge. The negative effect of slower charge redistribution is most
likely offset by the steady state efficiency gains that the heat exchanger provides. When the non-thermodynamic
effects of the heat exchanger are considered, such as the protection it gives the compressor, it becomes clear that its
use is warranted in refrigeration systems.
45
Chapter 7: Conclusions and Recommendations
7.1 Introduction
In the introduction of this report, two main goals were stated. The first is to identify and quantify the
individual phenomena which degrade cycling performance. In the preceding six chapters, we have analyzed several
of these factors, and they will be summarized here. The second goal is to determine which transients could be
modeled using simple corrections to a quasi-steady model, and which would require a full transient simulation. This
has not been specifically addressed in the previous chapters, so it will be discussed in this final chapter.
46
condenser geometry, affect how much energy (between 15 and 30 Btu) was brought into the evaporator during the
off-cycle. The liquid migration mass flow rate can be modeled using an adiabatic capillary tube model, and vapor
migration behaves like Fanno flow. The main difficulty in modeling the migration process is in determining how much
liquid migrates before vapor migration takes over.
In some instances, such as when the condensing temperature is high, the migrating refrigerant raises the
temperature of the evaporator above the freezer air temperature. In this case, most of the 15 to 30 Btu are transferred
to the freezer air, which degrades efficiency. However, when the migrating refrigerant is cool, such as when the
condensing temperature is low, the evaporator temperature remains below the freezer air temperature for the duration
of the off-cycle. This results in heat transfer from the air to the refrigerant inside the evaporator. In this case, the 15
to 30 Btu remain in the evaporator and are never transferred to the freezer air, so no loss in efficiency occurs as a
direct result of the migration. The heat transfer from the air for one particular data set was estimated to be only 2 Btu,
hence the refrigerating effect is most likely negligible. However, the heat transfer begins to boil the migrated
refrigerant, which affects the distribution of liquid and vapor within the evaporator. This can affect the redistribution
of charge during the subsequent on-cycle.
At the beginning of the on-cycle, most of the charge that migrated to the evaporator must be removed
before quasi-steady operation can be achieved. Thermodynamically it is best to remove the excess charge as quickly
as possible, so that the evaporating temperature can drop, and heat transfer between the evaporator and the freezer
air can begin. For the data sets used in this analysis, the initial phase of the redistribution process took between 2
and 4 minutes, during which the evaporator heat transfer rate was greatly reduced. This was found to be a major
contributor to cycling losses; the COP during these first minutes rose linearly from 0 nearly up to its maximum value.
The effect on overall COP therefore depends on cycle length and the amount of time it takes to redistribute the excess
charge at the beginning of the cycle.
The speed of the redistribution process depends primarily on whether refrigerant exits the evaporator as
liquid or vapor. If the excess charge vaporizes before leaving the evaporator, an additional 21 Btu could be absorbed
by the refrigerant from the freezer air. However, the redistribution process would be slower, thus prolonging the
period of reduced evaporator capacity. From this standpoint, slugging the liquid out of the evaporator is preferable,
except that it may damage the compressor. Perhaps an accumulator could be added at the compressor inlet, or the
compressor shell itself could be designed to act as an accumulator, to prevent damage from liquid slugs.
If too much charge is pulled out of the evaporator at the beginning of the on-cycle, the system will operate
with a high level of superheat at a reduced capacity. While this affects COP, it appears to be thermodynamically
preferable to the effect of refrigerant maldistribution during the first few minutes of the on-cycle.
For systems with a low-side oil sump, such as those with reciprocating compressors, up to 0.4 ounces of
refrigerant dissolve into the oil during the off-cycle, and boil out during the on-cycle. This effect is reversed and
much more pronounced for high-side oil sump systems, such as those with rotary compressors. In this case, up to 6
ounces of refrigerant may dissolve into the oil during the on-cycle, and boil out during the off-cycle.
By absorbing some of the extra charge from the evaporator during the off-cycle, a compressor with a low-
side sump allows the evaporator to return to its optimum charge level earlier in the on-cycle. However, this might not
47
affect the rate of decrease in low-side pressure and the rate of increase in evaporator capacity. As the refrigerant
boils out of the oil, it tends to increase the low-side pressure, thus prolonging the period of lower evaporator
capacity.
In a high-side sump system, much more charge is available to migrate through the capillary tube into the
evaporator during the off-cycle. In addition to the refrigerant in the condenser, the refrigerant dissolved in the oil (up
to 6 ounces) can migrate as well. With roughly twice the charge in the evaporator by the end of the off-cycle, it is
unlikely that the evaporator could become starved at the beginning of the on-cycle. However, as the compressor
begins depleting the evaporator, some of the refrigerant gets absorbed by the oil, thus the compressor collects
charge. This lowers the mass flow rate into the condenser, hence the mass flow rate into the evaporator is also lower.
In this manner, the high-side oil sump speeds the removal of charge from the evaporator. Future experiments are
being designed to quantify these effects.
The redistribution of refrigerant during the on-cycle substantially affects refrigerator performance.
Unfortunately it is complex to model, primarily because it is difficult to predict how much of the extra charge slugs out
of the evaporator as liquid, and how much of it leaves as vapor. This makes it difficult to predict the redistribution
time, and whether or not the evaporator will become starved. For this reason, the beginning few minutes of an on-
cycle cannot be modeled using a quasi-steady simulation. It would be extremely difficult to make a simple correction
to a steady state model, since the loss in evaporator capacity depends on so many variables.
The addition of a check valve to prevent off-cycle migration may eliminate some cycling losses. However,
the evaporator would be even more likely to become starved upon startup, since it would contain much less charge at
the end of the off-cycle. This would still result in a loss in COP until the evaporator could return to its steady-state
charge level.
48
calibrated using a temperature reference. The calibration of the thermocouple terminal panels in the data acquisition
system should be checked periodically for temperature drift.
This work has shown that microphones may be a valuable tool for analyzing refrigerant movement during
cycling. It is important to use high quality recording equipment to minimize extraneous noise generation. Analog or
digital filters could be developed to remove fan and compressor noise, which dominate the recordings. Every attempt
should be made to reduce the effects on refrigerator performance introduced by the instrumentation. This includes
minimizing thermal mass, added volume, and extra insulation which may occur as a result of mounting instrumentation
devices.
For future projects, it is recommended that cycling data be taken before steady state data. This way, the
cycling data can be examined first, and relevant steady state points can be determined beforehand. It is also
important to take a complete set of both cycling and steady state data before making any changes to the system
which could affect performance. The data should be taken as quickly as possible to minimize the effects of the
inevitable minor leaks that seem to accompany every instrumented refrigerator.
In order to maximize the number of data points for which mass flow meter was operable, much of this
analysis was done using an overcharged refrigerator. This will be less important in the future because the current
mass flow data have been used to adjust and improve the compressor map (Admiraal, 1993). Additional experiments
should be conducted with an optimally charged refrigerator to observe any differences.
49
References
Admiraal, D.M., and Bullard, C.W., Variable Conductance Models of Refrigerator/Freezer Heat Exchangers,
Forthcoming, Air Conditioning and Refrigeration Center, University of Illinois at Urbana-Champaign, 1993.
American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., 1989 ASHRAE Handbook
Fundamentals, ASHRAE, Atlanta, 1989.
Arthur D. Little, Inc., EPA Refrigerator Analysis Program User's Manual Peer Draft, 1992.
Cavallaro, A., University of Illinois at Urbana-Champaign, personal communication, 1993.
Dabiri, A.E., and Rice, C.K., "A Compressor Simulation Model With Corrections For the Level of Suction Gas
Superheat," ASHRAE Transactions, Vol. 87, Part 2, pp. 771-782, 1981.
Domanski, P.A., Didion, D.A., and Doyle, J.P., "Evaluation of Suction Line - Liquid Line Heat Exchange in the
Refrigeration Cycle," Procedings of the IIR-Purdue Refrigeration Conference, pp. 131-139, 1992.
Grebner, J.J., and Crawford, R.R., The Effects of Oil on the Thermodynamic Properties of Dichlorodifluoromethane
(R12) and Tetrafluoroethane (R-134a), ACRC TR-13, Air Conditioning and Refrigeration Center, University
of Illinois at Urbana-Champaign, 1992.
Incropera, F.P., and DeWitt, D.P., Fundamentals of Heat and Mass Transfer, 2nd ed., John Wiley & Sons, New York,
1985.
Janssen, M.J.P., deWit, J.A., and Kuijpers L.J.M., "Cycling Losses in Domestic Appliances: An Experimental and
Theoretical Analysis," International Journal of Refrigeration, Vol. 15, No. 3, pp. 152-158, 1992.
Melo, C., Da Silva Ferreira, R.T., Horn Pereira, R., and Ribeiro Negrao, C.O., "Dynamic Behaviour of a Vapor
Compression Refrigerator: a Theoretical and Experimental Analysis," Status of CFCs - Refrigeration Systems
and Refrigerant Properties, Proceedings of the International Institute of Refrigeration, pp. 141-149, 1988.
Moran, M.J., and Shapiro, H.N., Fundamentals of Engineering Thermodynamics, 1st ed., John Wiley & Sons, New
York, 1988.
Mulroy, W.J., and Didion, D.A., "Refrigerant Migration in a Split-Unit Air Conditioner," ASHRAE Transactions, Vol.
91, Part 1A, pp. 193-206, 1985.
Omega Engineering, Inc., The Temperature Handbook, Vol. 27, 1989.
Pedersen, C.O., Professor of Mechanical Engineering, University of Illinois at Urbana-Champaign, personal
communication, 1992.
Peixoto, R., Modeling of Refrigerant Flow Through Capillary Tubes, Forthcoming, Air Conditioning and
Refrigeration Center, University of Illinois at Urbana-Champaign, 1993.
Porter, K.J., and Bullard, C.W., Modeling and Sensitivity Analysis of a Refrigerator/Freezer System, ACRC TR-31,
Air Conditioning and Refrigeration Center, University of Illinois at Urbana-Champaign, 1992.
Reeves, R.N., Bullard, C.W., and Crawford, R.R., Modeling and Experimental Parameter Estimation of a
Refrigerator/Freezer System, ACRC TR-9, Air Conditioning and Refrigeration Center, University of Illinois at
Urbana-Champaign, 1992.
Sanvordenker, K., Techumseh Products Company, personal communication, 1992.
Staley, D.M., Bullard, C.W., and Crawford, R.R., Steady State Performance of a Domestic Refrigerator Using R12
and R134a, ACRC TR-22, Air Conditioning and Refrigeration Center, University of Illinois at Urbana-
Champaign, 1992.
Stoecker, W.F., and Jones, J.W., Refrigeration and Air Conditioning, 2nd ed., McGraw Hill, New York, 1982.
Stoecker, W.F., Design of Thermal Systems, 3rd ed., McGraw Hill, New York, 1989.
Sugalski, A., Jung, D., and Radermacher, R., "Quasi-Transient Simulation of Domestic Refrigerators," Proceedings of
the XVIII International Congress of Refrigeration, Vol. III, pp. 1244-1249, 1991.
Wang, J., and Wu, Y., "Start-up and Shut-down Operation in a Reciprocating Compressor Refrigeration System with
Capillary Tubes," International Journal of Refrigeration, Vol. 13, pp. 187-190, 1990.
50
Appendix A: Instrumentation
Staley (1992) estimated the accuracy of the original setup experimentally, using Equation A.1. The precision
error was determined to be ±0.53 °F and the bias is supplied by the manufacturer as 0.5°F. In an attempt to further
reduce the uncertainties in temperature measurement, a tare system was implemented on recent data sets. After all of
the thermocouples have soaked in an isothermal chamber for 48 hours, one is chosen as reference, in this case the
chamber air temperature, and the remaining thermocouple readings are adjusted to match the reference. These
adjustments are called the tare values, and they are typically less than 0.5°F. This in agreement with manufacturer's
bias specification. The point of the tare values is to eliminate the bias error, and hence lower the overall uncertainty.
Tare values are taken before each set of tests at each chamber temperature, and they are found to be relatively
constant. By eliminating the bias error, the overall uncertainty becomes the precision error, which is in this case ±
0.53 °F.
51
48 T-type thermocouples
ACM2-12-16 ACM2-12-8A
2 power
transducers
Uncertainty =
( Bias)2 + (Pr ecision error )2 A.1
52
Refrigerant Line
Pressure Tap
Brass Block
An immersion thermocouple is typically placed in an existing copper elbow in the refrigerant line through a
short guide tube. The guide tube is silver soldered to the elbow to provide support for the sheath. After the
soldering is completed, a 0.020 inch diameter hole is drilled through the refrigerant line wall. The thermocouple
sheath can then either be soft soldered or epoxied in place. Placing thermocouples in the evaporator is complicated
because of the difficulty in soldering to aluminum. Fortunately the tube wall is thick enough here to support the
thermocouple. The aluminum is prepared by roughening the surface with sandpaper and then cleaning it with
acetone. The thermocouple sheath is epoxied in place, and then taped along the adjoining evaporator tubing to
provide additional support. No leaks have been detected in these arrangements, and the soft soldering has had no
known ill effects on thermocouple performance.
Copper Elbow
Soft Solder
Stainless or Epoxy
Steel
Sheath
Thermocouple
Connector
Silver Solder Guide Tube
Pressure taps were added directly to the refrigerant lines, and supported with guide tubes as shown in
Figure A.4. The guide tube is first silver soldered in place, and then a 1/16 inch hole is drilled through the refrigerant
line. The pressure tap line is then soft soldered in place. This method minimizes added thermal mass and volume.
Guide Tube
Silver Solder
Copper Tubing
53
A.3 Surface Thermocouples
In order to evaluate the performance and applicability of surface thermocouples, several were mounted
adjacent to immersion thermocouples in the refrigerant lines, as shown in Figure A.5. In order to mount a surface
thermocouple, the copper tube is roughened with sand paper, and then cleaned with acetone. A T-type
thermocouple is wrapped several times around the tubing and secured in place with a thread. This provides strength
and also minimizes heat conduction through the thermocouple wire. It is important to maintain good contact between
the thermocouple junction and the copper tubing. A drop of thermally conducting epoxy (Wakefield Engineering
part number 152-1-A) is placed over the junction. After allowing sufficient drying time, a 3.0 inch wide strip of 0.5
inch thick adhesive-backed foam insulation is wrapped around the assembly to ensure that no fin effect occurs with
the thermocouple wire.
Insulation
Thermocouple wire Thermal
Copper Tubing Epoxy
The goal is to determine if surface thermocouples agree well enough with immersion thermocouples to
eliminate the need for immersion thermocouples on future projects. Surface thermocouples are easier to install
because they do not penetrate the refrigerant line. This prevents pressure drops as well as potential leaks. During
steady state operation the difference between surface and immersion temperatures for 40 data points is summarized in
Figure A.6 and Table A.2.
250
200
Surface Temperature (°F)
150
100
50 Compressor Inlet
Compressor Exit
Condenser Exit
0 Evaporator Inlet
Evaporator Exit
-50
-50 0 50 100 150 200 250
Immersion Temperature (°F)
54
A quick glance at Figure A.6 shows that most of the thermocouples agree quite well, with the exception of
the pair at the compressor exit. Here the surrounding air averages about 100 °F cooler than the refrigerant
temperature. This would tend to cool the surface, and lower the thermocouple's temperature. Secondly, the
refrigerant exiting the compressor is in the vapor phase which does not transfer heat to the tubing as well as a liquid
would. These are probably the major factors contributing to the large temperature difference. The surface
thermocouple itself is not suspected to cause the large error because it agrees perfectly with the immersion
thermocouple when the refrigerator is turned off and at thermal equilibrium with its surroundings. Another possible
explanation is that the thermocouple is mounted improperly, but there are no visual indications of this.
The other four surface thermocouples are in better agreement with their immersion counterparts. The
difference between these surface and immersion temperatures is relatively small, and a correlation can be created
between them. In future projects, immersion thermocouples could be eliminated in favor of surface thermocouples
using these correlations. It is possible that the performance of these surface thermocouples may be improved by
more insulation, although too much insulation could begin to affect system performance.
A surface thermocouple can be crudely modeled as shown in Figure A.7, which contains a simplified
resistive network representation. The thermal resistivity of the metal tubing is very small, thus the tube is assumed to
be isothermal. Under this assumption, the insulation has little effect on the surface temperature. Axial conduction
through the tube would make the wall temperature the same under the insulation as it is adjacent to it. With these
simplifications, the one-dimensional heat transfer is characterized by Equation A.2
55
Tair
(1/hA) ref
Tref
Tref
56
260
220
200
180
160 Predicted
Surface
140
140 160 180 200 220 240 260
Refrigerant Temperature (°F)
40
Predicted and Surface Temperatures (°F)
30
20
10
-10 Predicted
Surface
-20
-20 -10 0 10 20 30 40
Refrigerant Temperature (°F)
The correlations work very well in all locations, with the possible exception of the compressor exit, where the
root mean square error is 2.51 °F. The simple resistive network representation may not work well here because the
difference between refrigerant and air temperatures is nearly 100°F. This large temperature difference amplifies the
importance of an accurate k value. Also, the insulation could be substantially affecting the tube wall temperature
which would invalidate the one-dimensional heat transfer assumption. Finally, the k value is really a function of
refrigerant and air flow rates. While the air flow rate is relatively constant throughout the data set, the refrigerant
mass flow rate may vary by a factor of 2. It is not yet known if the k values would work well on another installation,
however a second refrigerator has been equipped with surface thermocouples and it is recommended to repeat this
analysis on that unit.
The surface thermocouples were also evaluated using a transient data set taken at an ambient temperature of
75°F. Here they do not work nearly as well. Figures A.10 and A.11 compare a subset of the transient data at the
57
compressor exit and evaporator inlet, and Table A.4 summarizes the maximum and RMS errors for the complete data
set for all five sets of thermocouples. The maximum errors occur during sharp transients and are not representative of
typical errors, which are listed in the RMS error column. This indicates that the surface thermocouples may be most
useful when analyzing quasi-steady state operation, such as at the end of on or off cycles.
200
Surface Temperature (°F)
150
100
50
50 100 150 200
Immersion Temperature (°F)
Figure A.10 Transient comparison of surface and immersion thermocouples at compressor exit
20
Surface Temperature (°F)
10
-10
-20
-30
-30 -20 -10 0 10 20
Immersion Temperature (°F)
Figure A.11 Transient comparison of surface and immersion thermocouples at evaporator inlet
58
The surface thermocouples tend to lag behind the immersion ones, as shown in Figure A.12 for the
compressor exit and Figure A.13 for the evaporator inlet. Because the rate of change of temperature in the refrigerant
during cycling is not constant, the lag is not constant either. One explanation for this is that it takes a finite amount
of time for heat to conduct through the tube wall, hence the surface temperature lags behind the refrigerant
temperature. The correlation developed for steady state can be applied to transient data with moderate success as
shown in Figures A.14 and A.15 for the compressor exit and evaporator inlet respectively. Table A.5 summarizes the
maximum and RMS errors using the correlations for all of the thermocouples.
One problem that arises in using the thermocouple correlations for transient data is that phase conditions
might occur that do not exist for steady state. For example, during the initial startup, the condenser exit may be
superheated, and this is clearly unrealistic for steady-state conditions. No k value has been estimated to describe
this condition.
200
Immersion
Surface
180
160
Temperature (°F)
140
120
100
80
60
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
Figure A.12 Surface and immersion temperatures at compressor exit during cycling
80
Immersion
Surface
60
Temperature (°F)
40
20
-20
-0:10 1:08 2:25 3:43 5:00
Time (h:m)
Figure A.13 Surface and immersion temperatures at evaporator inlet during cycling
59
200
160
140
120
100
80
60
60 80 100 120 140 160 180 200
Figure A.14 Comparison of Predicted and measured refrigerant temperatures at compressor exit
20
Predicted Temperature (°F)
10
-10
-20
-30
-30 -20 -10 0 10 20
Figure A.15 Comparison of predicted and measured refrigerant temperatures at evaporator inlet
In general, the thermocouple correlations give large errors during sharp transients such as at the beginning
of the on and off cycles. Near the ends of the on and off cycles the temperatures are close enough to steady state so
that the steady state correlations can be used with some success. Again the compressor exit surface thermocouple is
the farthest from its corresponding immersion thermocouple, although the others are not much better. At this time,
60
surface thermocouples cannot be recommended to reliably measure temperatures in a transient system such as a
cycling refrigerator, especially when analyzing the sharp transients at the endpoints of a cycle.
The simple thermocouple model was developed assuming steady state heat transfer. This could be one
reason that the model does not work as well for transient operation. Perhaps a more complete model could be
developed to include the thermal mass of the tube wall, however this added complexity might make the correlations
impractical to use.
In order to see if the overall accuracy is substantially worse than the manufacturer's specifications, the
PC519E power transducer was compared with a stand-alone power analyzer, a Valhalla Scientific model 2100. Its
accuracy is also listed in Table A.6. Various combinations of fan motors and light bulbs were connected to both
power transducers simultaneously. A comparison of the power measured by the PC519E through the data
acquisition system to power displayed on the Valhalla Scientific 2100 is presented in Figure A.16. Overall the
agreement is very good over the entire range. The largest discrepancies occurred when measuring powers over 230
W, perhaps because the accuracy of the Valhalla Scientific 2100 drops significantly at higher powers. The maximum
absolute error was 2.9 W (1.1%), which occurred when measuring a 260W load. The maximum percentage errors
occurred when measuring powers less than 20 W, in this case 2.7 % which corresponded to a 0.7 W difference.
61
The accuracy of the data acquisition system is bounded by the absolute difference between the two
readings given by the measurement devices, plus the accuracy rating of the Valhalla Scientific 2100, assuming that
the manufacturer's accuracy claims are legitimate. This information is summarized in Table A.6 also. The actual
accuracy is most likely better than these worst case calculations. Since the Valhalla Scientific 2100 is only accurate to
±0.25% of the reading ± 6W when the current is over 2 A, the results of this comparison in the upper power range are
not very meaningful. Additional testing should progress as soon as a more accurate reference power measurement
device becomes available.
350
300
250
Data Acquisition
200
150
100
50
0
0 50 100 150 200 250 300 350
Valhalla Scientific 2100
Figure A.16 Comparison of PC519E and Valhalla Scientific 2100 power transducers
62
Appendix B: Cabinet Heat Loads
Several methods for determining the overall heat transfer conductance of a domestic refrigerator cabinet
have been implemented. All are based on a reverse heat leak test, however different air circulation and heating
methods are used. The least intrusive and most realistic method is preferred, as long as it produces consistent
results. A summary of the results is presented in Table B.1, and an explanation of the methods follows.
The first UA estimates were calculated by D. Staley (1992) using hair dryers to heat and circulate the air
inside the freezer and fresh food compartments. The hair dryer fans were wired to run continuously, while the
heating element was cycled by a proportional controller. The vents connecting the freezer and fresh food
compartments were taped shut to eliminate heat transfer due to migrating air. The cabinets were heated to identical
temperatures above the ambient test chamber temperature in order to transfer heat from inside the cabinet to the
surroundings. By measuring temperatures and heat inputs, Equations B.1 and B.2 were solved independently for
UA frig and UA frez:
63
provided by nichrome wire heating elements strung across the fan inlets. This time, the heating elements were
controlled by a variac. The test and data reduction procedures remained unchanged. The results from this method
are markedly different in the freezer compartment, possibly due to the replacement of interior panels which were
damaged during the summer of 1991. Other possible sources for the differences are the altered air currents, or the un-
filtered, fluctuating power provided to the test equipment.
Reeves' fall UA estimates were verified by D. Admiraal (1993) in the winter of 1992 using a similar procedure.
By this time the power conditioner had been added to the test equipment, and power measurements became more
consistent. Admiraal used the same fans and heating elements as Reeves did for his second test, except he used
proportional controllers to cycle the heating elements. The result for the refrigerator cabinet is similar to Reeves',
however there is a slight difference in the freezer compartment. This may be caused by experimental error or different
fan orientation.
The fan orientation was suspect in affecting the convection inside the compartments, hence a less intrusive
UA test was desired. In order to better simulate the air flow patterns inside the compartments during normal
operation, the evaporator fan was wired to run continuously and the muffin fan was removed from the freezer. The
compartments are no longer isolated from each other; air is allowed to flow between them. The air temperatures from
both compartments were carefully matched to minimize heat transfer from migrating air. New, larger heating elements
were constructed to distribute heat more evenly throughout the cabinets. At first, the test was run with no fan in the
refrigerator compartment; however, this resulted in a large temperature gradient. The 14W, 105cfm muffin fan was
placed in the fresh food compartment to stir the air. It is placed on the wire rack as shown in Figure B.1. The inside
of the refrigerator compartment is isothermal to ± 1.0 °F and in the freezer compartment ± 0.7 °F. A least squares
minimization is implemented to simultaneously calculate UA frig, UA frez, and fz, the evaporator fan air split fraction. The
objective function is given in Equation B.3:
Top View
28"
12"
9"
25"
Side View
heater
glass shelf
36"
heater
64
Sigma = (Pevap + Qfrez - UA frez(Tfrez - Tamb) + UA mullion(Tfrig - Tfrez)
where Pevap is the evaporator fan power, Ttofrig is the air stream temperature into the refrigerator compartment, and
Tfromfrig is the air stream temperature leaving the refrigerator compartment. This objective function is derived from an
energy balance on the compartments as shown in Figures B.2 and B.3.
T T
Qfrez
evaporator
(UAfrez)(²Tfrez)
heater
T
T
T
T
T
T
(1-fz)(m)(cp)(Ttofrig)
(1-fz)(m)(cp)(Tfromfrig)
(UAmullion)(²Tmullion)
(1-fz)(m)(cp)(Ttofrig) (1-fz)(m)(cp)(Tfromfrig)
(UAmullion)(²Tmullion)
T
T
T
heater 1
Qfrig
T T
(UAfrig)(²Tfrig)
heater 2
circulation fan wire shelf
T
T P
T
compressor T ²P
T condenser
65
This parameter estimation results in a fairly steep minimum for the UA's as shown in Figure B.4. Although
the UA values are firmly determined, the value for fz is not. By setting fz to different values and solving for the UA's,
it becomes apparent that sigma is not a strong function of fz. For this reason, fz is set at 85% and only the UA's are
estimated.
UA frez is higher than the values obtained with muffin fans in sealed compartments. When the compartments
were sealed, the space behind the evaporator contained stagnant air. This could act as an insulator and decrease
cabinet conductance. Thus, when air is allowed to flow in this area, another heat transfer surface is introduced,
thereby increasing cabinet conductance. In comparison with Staley's estimate, the evaporator fan may provide less
convection than the hair dryer and hence the overall conductance is lower.
35.00
30.00
25.00
20.00
Sigma
15.00
10.00
5.00 0.5375
0.5225
0.00
0.5075
0.8725
0.88
0.4925 UA frez
0.8875
0.895
0.9025
0.91
0.4775
0.9175
0.925
0.9325
0.94
UA frig
Qfrez is typically between 5 and 20W, and Qfrig is between 35 and 55W. According to manufacturer's
specifications, the uncertainty on these measurements is less than +/- 0.25W. Thermocouple uncertainty is less than
+/-0.9 °F, also according to manufacturer's specifications. The greatest uncertainties in the energy balance are
UA mullion and the air split fraction, fz. UA mullion was estimated by D. Admiraal (1993) to be 1.2 Watts/°F ±20%. This
does not cause a large error since the mullion temperature difference, ∆Tmullion, is very small and the resulting mullion
term is on the order of 1W. The mass flow rate is calculated by multiplying the volumetric flow rate, v, with the air
density at Ttofrig. The volumetric air flow rate was estimated by R. Reeves (1992) to be 45cfm. This is in agreement
with the manufacturer's estimate of 44 to 46 cfm. The air split fraction was estimated to be 85% in the spring of 1991,
and 70% in the fall of 1991. For this analysis, 85% is used. Because Ttofrig and Tfromfrig are nearly equal, the heat
transfer due to migrating air is typically from 2 to 5W, and thus the parameter uncertainties do not bias the results
very much.
Finally, to test the hypothesis that convection is an important contributor to cabinet UA values, the large
muffin fan was replaced by a 4W, 15cfm fan. This smaller fan is sufficient to keep the refrigerator compartment
66
isothermal to +/- 1.4 °F and the freezer to +/- 0.7 °F. The same objective function is used and again fz is set at 85%.
UA frez is essentially unchanged, because nothing is changed in the freezer compartment. UA frig decreases by about
10%. This can be explained by the lower convection coefficients which result from a smaller fan. UA frez is probably
more sensitive to fan selection because the compartment is much smaller.
67
Appendix C: Migration Model Program Listing
This appendix contains a True Basic™ program listing for the off-cycle migration model used in the
Refrigerant Migration chapter. The generalized Newton-Raphson subroutine was provided by C.O. Pedersen (1992).
68
LET Eevapinitial = Eevap(me, Te, Ve) !initial energy in evaporator in BTU
LET Pe = Psat(Te)
LET Pc = Psat(Tc)
LET Qe = UAe*(Te-Tfrez)
LET Eet = DEevapDt(me,Te,Ve)
LET Eem = DEevapDm(me,Te,Ve)
LET DTeDt = -(Qe + mdot*h - Eem*mdot)/Eet
69
LET Te = Te + DTeDt*deltat
LET Tfrez = Tfrez + 0.01
LET ECum = Ecum - mdot*h*deltat
PRINT #1: migtime;CHR$(9);Tc;CHR$(9);Pc;CHR$(9);mc*16;CHR$(9);mdot*16/60;CHR$(9);Qe
PRINT migtime;CHR$(9);Tc;CHR$(9);Pc;CHR$(9);mc*16;CHR$(9);mdot*16/60;CHR$(9);Qe;CHR$(9);Te
NEXT counter
!**********************************************************************
! FUNCTION SECTION
!**********************************************************************
!**********************************************************************
! ENERGY CALCULATION FUNCTIONS
!
FUNCTION Ecv(m, T, V)
!Total energy of the condenser in BTU
DECLARE DEF quality
DECLARE DEF uvap
DECLARE DEF uliq
LET mcp = 0.435 !BTU/R for condenser coils
LET x = quality(m,T,V)
IF x > 1 THEN LET x = 1
LET u = uliq(T) + x*(uvap(T) - uliq(T))
LET Ecv = mcp*(460+T) + m*u
END FUNCTION
FUNCTION Eevap(m, T, V)
!Total energy of the evaporator in BTU
DECLARE DEF quality
DECLARE DEF uvap
DECLARE DEF uliq
LET mcp = 2*0.72025 !BTU/R
LET x = quality(m,T,V)
LET u = uliq(T) + x*(uvap(T) - uliq(T))
LET Eevap = mcp*(460+T) + m*u
END FUNCTION
!**********************************************************************
! DERIVATIVE CALCULATION FUNCTIONS
!
FUNCTION DeDm(mass, Tc, Vol)
DECLARE DEF Ecv
LET deltam = 0.000625 !1/100 ounce
LET DeDm = (Ecv((mass+deltam/2),Tc,Vol) - Ecv((mass-deltam/2),Tc,Vol))/deltam
END FUNCTION
70
FUNCTION DEevapDm(m, T, V)
DECLARE DEF Eevap
LET deltam = 0.000625 !1/100 ounce
LET DEevapDm = (Eevap((m+deltam/2),T,V) - Eevap((m-deltam/2),T,V))/deltam
END FUNCTION
FUNCTION DEevapDt(m, T, V)
DECLARE DEF Eevap
LET deltaT = 0.1 !1/10 degree F
LET DEevapDt = (Eevap(m,(T+deltaT/2),V) - Eevap(m,(T-deltaT/2),V))/deltaT
END FUNCTION
!**********************************************************************
! THERMODYNAMIC FUNCTIONS
!
FUNCTION quality(m, T, V)
DECLARE DEF vliq
DECLARE DEF vvap
LET quality = (V-m*vliq(T)) / ( m*(vvap(T)-vliq(T)) )
END FUNCTION
FUNCTION Psat(T)
!gives saturation pressure in psia, T in F
LET Psat = 23.932968228 + 0.51418662707*T + 0.0037341886015*T^2 + 0.0000190658186*T^3
END FUNCTION
FUNCTION hliq(T)
!gives enthalpy in BTU/lb, T in F, for saturated liquid
LET hliq = 8.536860812 + 0.2177245061*T - 0.000011400220011*T^2 + 0.00000087534587076*T^3
END FUNCTION
FUNCTION hvap(T)
!gives enthalpy in BTU/lb, T in F, for saturated vapor
LET hvap = 77.252285672 + 0.10468129772*T + 0.000022011282416*T^2 - 0.00000087986314726*T^3
END FUNCTION
FUNCTION uliq(T)
!gives internal energy in BTU/lb, T in F, for saturated liquid
LET uliq = 8.48644651 + 0.21641471052*T - 0.000012813919711*T^2 + 0.00000076402277996*T^3
END FUNCTION
FUNCTION uvap(T)
!gives internal energy in BTU/lb, T in F, for saturated vapor
LET uvap = 70.154332782 + 0.095941438918*T + 0.000047077025446*T^2 - 0.00000068031077878*T^3
END FUNCTION
FUNCTION vliq(T)
!gives saturated liquid volume in ft^3/lb, T in F
LET vliq = 0.011038151037 + 0.000013743768887*T - 0.0000000095701304835*T^2 + 0.00000000036043141398*T^3
END FUNCTION
FUNCTION vvap(T)
!gives saturated vapor volume in ft^3/lb, T in F
LET a = 1.6248139964 - 0.032160201866*T + 0.00033008938409*T^2 - 0.0000017596245796*T^3
LET vvap = a + 0.0000000036691583966*T^4
END FUNCTION
71
FUNCTION density(T1, P1)
!returns density of slightly superheated vapor in lbs/ft^3
!used only in vapor mass flow rate calculation
LET R = 1.986/120.93
LET Z = 0.83
LET density = P1/(R*(T1+460)*Z)*144/778.2
END FUNCTION
FUNCTION SVapPres(m,T,V)
!returns the ideal gas pressure for slightly superheated vapor
!uses compressibility factor
!used in calculating the pressure in the consdenser after it's superheated
LET R = 1.986/120.93*778.2/144
LET Z = 0.83
LET rho = m/V
LET SVapPres = Z*rho*R*(T+460)
END FUNCTION
!**********************************************************************
! MASS FLOW RATE FUNCTIONS
!
FUNCTION mflow(P)
!mass flow rate in lbs per hour for liquid
LET mflow = 1.987-3.025*P/25+0.146*(P/25)^2
END FUNCTION
72
IF PC < 30 THEN
LET x(2) = 0.03
LET x(1) = 0.02
END IF
!************************************************************************
! Subroutines from here on
!************************************************************************
SUB nr(x(),r(),k(),nvar,toll,niter,delta)
! Generalized newton-raphson subroutine
! Numerical partial derivative version
! c o pedersen, m&ie dept, u of illinois
! Inputs:
! x = variable array, should contain initial values on entry (nvar long)
! k = constant array
! nvar = number of variables
! toll = convergence criterion (toll*dot(x,x), suggest.001)
! niter = max number of iterations allowed
! delta = increment of x in partial deriv calc (suggest .001)
! Outputs:
! r = residual equation values
! x = final x values
! niter = actual number of iterations
! Required Subroutines:
! rcalc(r(),x(),k()) ! subroutine to evaluate residual
! equations (supplied by user).
! Subroutines used:
! calcfp(r(),ro(),x(),k(),fprime(,),delta)1
! ! evaluates numerical partial derivatives
!
DIM dx(1),ro(1),fprime(1,1),invfprime(1,1)
! initialize, resize and zero arrays
LET maxiter=niter ! pass in max number of iterations, return actual number
MAT r=zer(nvar)
73
MAT dx=zer(nvar)
MAT ro=zer(nvar)
MAT fprime=zer(nvar,nvar)
MAT invfprime=zer(nvar,nvar)
FOR niter = 1 to maxiter
CALL calcfp(r,ro,x,k,fprime,delta)
CALL calcr(r,x,k)
! **
! ** solve for corrections
! ** note +r is used on rhs
! ** corrections will be subtracted from base value
! **
MAT invfprime= inv(fprime)
MAT dx=invfprime*r
MAT x=x-dx
LET err=dot(dx,dx)
LET xnorm = dot(x,x)
IF err < toll*xnorm then ! termination condition
CALL calcr(r,x,k) !reevaluate r
EXIT SUB
END IF
NEXT niter
END SUB
SUB calcfp(r(),ro(),x(),k(),fprime(,),delta)
! **
! Subroutine to fill the partial derivative matrix (Jacobian)
!
! The subroutine calls routine calcr to evaluate all derivatives.
! calcr must be given a set of variable (x) values. It returns
! corresponding values for all the residual equations.
! It initially calculates a set of base values ro, based on the initial
! x vector supplied, and uses that set in all the partial derivative
! evaluations.
!
! Note that it evaluates an entire column of the Jocobian when it has
! incremented one of the x values. This helps to reduce the number of
! calculations for a sparce Jacobian.
!
! The variable excursion for the PD calculation is delta (input) times the
! current value of the variable.
!
! **
LET nvar=size(x)
CALL calcr(ro,x,k)
FOR i=1 to nvar
LET deltax=delta*x(i)
LET x(i)=x(i)+deltax
CALL calcr(r,x,k)
FOR j=1 to nvar
LET fprime(j,i)=(r(j)-ro(j))/(deltax)
NEXT j
LET x(i)=x(i)-deltax
NEXT i
END SUB
74
SUB CALCR(R(),X(),k())
!---------------------------------------------------------
! ** SUBROUTINE TO EVALUATE RESIDUAL EQUATIONS
!---------------------------------------------------------
! ** ASSIGN VALUES TO THE VARIABLES FROM
! ** THE X ARRAY TO SIMPLIFY CALLING PROCEDURE
! ** AND MAKE THE EQUATIONS READABLE
LET f = k(1)
LET L = k(2)
LET D = k(3)
LET g = k(4)
LET Pc = k(5)
LET Pe = k(6)
LET M1 = x(1)
IF M1 < 0.0001 THEN LET M1 = 0.0001
LET M2 = x(2)
LET P1 = x(3)
!---------------------------------------------------------
! ** RESIDUAL EQUATIONS
!---------------------------------------------------------
LET R(1) = -f*L/D + (g+1)/(2*g)*log((1+(g-1)/2*M2^2)/(1+(g-1)/2*M1^2)) - (1/g)*(1/M2^2 - 1/M1^2) -
(g+1)/(2*g)*log(M2^2/M1^2)
LET R(2) = Pc - P1*(1+(g-1)/2*M1^2)^(g/(g-1))
LET R(3) = Pe/P1 - M1/M2*((1+(g-1)/2*M1^2)/(1+(g-1)/2*M2^2))^.5
END SUB
75
Appendix D: Convective Heat Transfer in the Suction Line
One way to determine whether vapor alone could have achieved the rapid temperature drop of the suction
line (during the first 30 seconds of the on-cycle for the 90 °F ambient case) is to compare the observed convection
coefficient to the theoretical coefficient. The observed convection coefficient, h obs, is determined from Equation D.1,
dE
h obsA∆T =
dt D.1
where A is the internal surface area of the suction line (40.3 in ), ∆T is the average temperature difference between the
2
suction line surface and the refrigerant (22 °F), and E represents the energy stored in the interchanger. ∆T is
determined by averaging the measured temperature differences at both ends of the suction line over time. The
amount of energy transferred from the interchanger (for the first 30 seconds of the on-cycle) was estimated to be 1
Btu in Chapter 6, thus the average rate that energy was removed is 1/30 Btu/sec, or 120 Btu/hr. Using these values,
Equation D.1 is solved to find that h obs is 19.5 Btu/hr-ft 2-°F.
The Dittus-Boelter equation, Equation D.2,
h DB D
= 0. 023Re0 .8Pr 0.4
k D.2
can be used to calculate the theoretical heat transfer coefficient for vapor flow through the suction line, h DB. All
refrigerant properties are calculated for superheated vapor at the average refrigerant temperature. Definitions for the
Reynolds number, Re, and the Prandtl number, Pr are provided by Equations D.3 and D.4 respectively.
Ý
4m
Re =
πDµ D.3
µc p
Pr = D.4
k
Descriptions and values for all parameters are found in Table D.1. Using these values, h DB is calculated to
be 19.2 Btu/hr-ft 2-°F.
The theoretical and measured convection coefficients are extremely close, confirming that vapor alone could
have caused the rapid temperature drop of the suction line during the first 30 seconds of the on-cycle.
76
Appendix E: Compressor Power Measurement
Compressor calorimetry data are customarily taken in a 90 °F ambient room with the suction gas maintained
at 90 °F. A curve fit for compressor power is usually made from this data as a function of condensing and
evaporating temperatures alone. Under different ambient conditions, the inlet temperature varies, thus the density of
the suction gas varies as well. This can have a significant impact on the performance of the compressor. This effect
has been recognized by Dabiri and Rice (1981) who have proposed correction factors to account for different levels
of superheat at the compressor inlet.
Figures E.1 and E.2 compare the measured compressor power with the power predicted by the compressor
map for two separate steady state data sets. The compressor map power, Pcomp, is given in Equation E.1:
77
300
100 °F Ambient
250
Compressor Power (W) 75 °F Ambient
200
150 90 °F Ambient
60 °F Ambient
100
Measured
50 Compressor Map
0
0 5 10 15 20 25 30 35 40
Data Point
Figure E.1 Measured vs. predicted compressor power for fall 1992 data
300
75 °F Ambient
250
Compressor Power (W)
200
90 °F Ambient
150
60 °F Ambient
100
Measured
50 Compressor Map
0
0 5 10 15 20 25 30
Data Point
Figure E.2 Measured vs. predicted compressor power for spring 1993 data
78