Sunteți pe pagina 1din 11

Applied Thermal Engineering 25 (2005) 11151125

www.elsevier.com/locate/apthermeng

A simple technique for refrigerant mass measurement


Erik Bjork

Department of Energy Technology, Royal Institute of Technology, SE 100 44 Stockholm, Sweden


Received 12 June 2004; accepted 19 September 2004
Available online 2 November 2004

Abstract
A simple technique for refrigerant mass measurement is described and evaluated. First, quick-closing
valves trap the refrigerant in the section under consideration. Then, the refrigerant is expanded into a tank,
thus reaching a superheated state. Finally the mass is calculated by pvT relationship. The technique was
implemented on a domestic refrigerator and was computer automated (no need for manual intervention).
Preliminary 1 data are reported of the charge distribution during an ono cycle.
2004 Elsevier Ltd. All rights reserved.
Keywords: Charge inventory; Void fraction; Domestic refrigerator

1. Introduction
Knowledge of the quantity of charge has a great importance for heat exchanger design and heat
pump system eciency. Recently, there has been a growing interest in hydrocarbons as refrigerants, which creates a demand for reduced charges due to their ammability. Many studies have
found that an eciency-maximum exists for a certain quantity of charge in a heat pump system.

Corresponding author. Tel.: +46 87908602; fax: +46 8203007.


E-mail address: bjork@energy.kth.se
1
A second article is intended that will focus on the experimental results rather than the experimental technique.

1359-4311/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2004.09.008

1116

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

Nomenclature
R
dR
T
V
X
dX
m
p
v
q

measurement result
uncertainty in R
temperature (C)
volume (m3)
variable
uncertainty in X
mass (kg)
pressure (bar)
specic volume (m3/kg)
density (kg/m3)

Subscript
i
index

The development of correlations for two-phase ow heat transfer and pressure-drop is a research
area where information of the local void fraction is essential.
Hewitt [1] discussed dierent techniques for void measurements; including methods based on
radioactive absorption and scattering, impedance, sound velocity, electro-magnetic ow, microwave absorption, light scattering, and volume of the liquid or vapour phase within the channel.
He noted that the volume measurement or quick-closing valves technique, as being intrusive,
needed repeated re-establishments of the steady-state conditions compared to the dierent on-line
methods. Mulroy and Didion [2] used quick-closing valves to isolate sections of a split-unit air
condition, from which the refrigerant was removed and weighed. Janssen [3] used on-line weighing
of the heat exchangers to nd the mean void fraction at steady-state condition. Whalley [4]
pointed out that the quick-closing valves technique is a simple if rather brutal method of measuring void fraction, but that the method has the great advantage that the results are not open to
interpretation. The technique to measure mass of a superheated gas by the pvT relationship
was for instance used by Philipp et al. [5].
Several of the techniques discussed by Hewitt [1] are today implemented in commercially available products. However, in general all the dierent on-line methods are sophisticated and expensive, and demand special calibration and handling. In contrast, the quick-closing valves technique
is time-consuming and needs hand-operation. Moreover, if the refrigerant is removed from the
system and weighed, there is a risk of errors from refrigerant losses.
It is the purpose of the present paper to present an improvement to the quick-closing
valves technique. By this improvement, the technique can be made autonomous, and without the errors introduced from the remove-and-weigh procedure. The experiments from which
preliminary results are reported were conducted on a domestic refrigerator during ono
cycling conditions. The results, and the proposed technique itself, are evaluated through
an error analysis and through a comparison to the conventional remove-and-weigh technique.

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

1117

2. Methods
2.1. The mass measurement technique
The cooling system is subdivided into control volumes (all with known volumes) by quick-closing valves, hereby trapping the refrigerant. Each control volume is then expanded into a tank (also
with known volume) large enough to ensure a superheated state. As thermodynamic equilibrium
is reached (i.e. at the temperature of the ambient), the temperature and pressure are measured.
Finally, the density is obtained from the equations of state (pvT relationship), and the mass
is calculated according to
mqV:

From the equation it is evident that the combined volume (control volume and tank) must be
known.
2.2. Sizing of the tank
The required tank volume can be calculated (also using Eq. 1) from the maximum quantity of
charge and the refrigerant density (as saturated vapour at the ambient temperature). The experience from the present work is that the tank should be selected as large as possible to reduce the
time to reach thermodynamic equilibrium, to reduce uncertainties, and to expand the possible
parameter space (for instance a lower ambient temperature).
2.3. Experimental arrangement
The technique was implemented on a domestic refrigerator in such way that the experimental
arrangement was made automatic (no need of manual intervention during an experiment). This
was obtained by computer control and by letting the compressor evacuate the tank after a test
(thus bringing the refrigerant back to the cooling system). Four components were investigated
for their charges: the condenser, the lter drier, the evaporator and the compressor. In order to
limit the number of components, three of the control volumes were stepwise expanded into one
common tank. However, the compressor was kept as a separate control volume, as being large
enough to ensure a superheated state. The system is displayed in Fig. 1. Information about the
control volumes is provided in Table 2.
The test object was an Electrolux ER8893C refrigerator with free convection heat exchangers
and a piston compressor (ZEM HQY70AA) with low-pressure oil sump (265 ml mineral oil
charge). The nominal refrigerant charge was 34 g of Isobutane (R600a). The capacity control
was by intermittent run (ono cycling) with self-defrosting in every o-cycle.
The following modications were made to the cooling system: ve valves and a tank were
added. The charge was increased by 1.3 g (total 35.3 g) in order to compensate for the charge that
could not be evacuated from the external tank. An electrical heater with a small fan, positioned in
the cabinet, was used to accelerate the test series. Pressure taps were tted on the tank, on the
suction line, and on the discharge line. The valves were servo actuated ball valves (Swagelok s-series) with internal diameters close to the original system. These valves give a low pressure-drop and

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125


Control Volume 3

cond out

cond mid

cond high

Pcond

discharge valve

condenser

condenser valve

discharge

Control Volume 2

compressor high
filter-drier

filter drier
Pcomp

compressor low

capillary valve

Compressor
Control Volume

1118

suction

cabinet

Control Volume 1

evap in

electrical heater

evaporator

tank

accumulator

capillary tube

suction valve
Ptank

tank valve
evap out

evap mid

acc in

Fig. 1. Experimental apparatus.

are bi-stable, thus generating no heat when passive. The tank was positioned elevated, compared
to the compressor, in such way that any oil would be drawn back to the cooling system by gravity.
Overall, small modications were made to the original system (all original components were at
original positions, connection pipes were kept short and with internal diameters close to the original system, thin tubes were used for external piping). As a result, the behaviour of the modied
system was close to the behaviour of the original system. The experiments were conducted in close
to ISO 7371 conditions. The refrigerator air temperature was controlled to 5 1 C for a test, with
a cut-in of 6 C and a cut-out of 4 C. The climate chamber temperature was 25 1 C with a
relative humidity of 65 10%.
A computer (with software developed in HP-VEE) controlled the compressor, the electrical
heater, and the valves via a relay-box (Utronix SwitchBox44) and a servo-controller (Basic Stamp
Mini SSC II). The signal from the temperature (T-type thermocouple) and pressure transducers
(Druck PDCR 4011) were AD-converted in a logger (HP 34970A) before transferred to the computer where it was recorded for later analysis. Positions for thermocouples and pressure transducers are shown in Fig. 1.
2.4. Experimental procedure
In all, 91 tests were performed, each after re-established steady cycling conditions. By this,
snap-shots of the charge distribution were collected at unique times throughout the ono
cycle, together displaying the charge variation with time.

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

1119

Prior to the test series the cooling system was carefully evacuated and charged so that the exact
quantity of charge in the system was known. The desired points of time to be tested in the ono
cycle were entered into the computer program. After this, the experimental apparatus was
autonomous.
Each test started with pressure equalization to give the compressor an unloaded start. Then,
the compressor was started with the valves positioned in such way that the tank was evacuated
(thus bringing the refrigerant back to the cooling system). After this, the tank was isolated from
the system by the tank valve. Next, four ono cycles were completed, allowing the refrigerator
to approach steady cycling conditions. During the fth cycle, a timer determined the moment
for the valves to close around the control volumes and to open the tank valve so that the evaporator charge (control volume 1) was expanded. Time was then given for the system to thermally equalize with the tank and the ambient, a process that was accelerated by the electrical
heater in the cabinet. Finally, control volumes 2 and 3 were expanded stepwise, each step with
time given for thermal equalisation. After this, a new test started beginning with pressure equalization and evacuation of the tank. During the entire process pressures and temperatures were
recorded for later analysis. Table 1 provides a timeline for the experimental procedure of one
test.
2.5. Calculation procedure
Calculations were made in a spreadsheet program (Excel) coupled to a refrigerant property
add-in program (Refprop). First, the residual charge in the tank was calculated (the pressure in
the tank was about 0.1 bar after the evacuation). Then, control volume 1, 2 and 3 by subtracting
one mass from another. Next, the compressor vapour charge. Finally, the charge dissolved into
the compressor oil by adding all vapour masses and subtracting them from the total charge.
2.6. Experimental comparison with remove-and-weigh technique
The proposed technique was compared to a conventional mass measurement technique by
which the refrigerant is extracted into a tank and weighed. First, the refrigerator compressor
was powered and run for 1 h, thus causing the refrigerant to distribute over the cooling system.
Then, the valves were closed, the compressor switched o, and a tank (which was kept in a bath of
low temperature liquid nitrogen) connected to the cooling system. The refrigerant was then extracted, control volume by control volume, into the low vacuum tank. At each step of the emptying process the tank was carefully weighed. In parallel, the mass was measured by the proposed
technique.
2.7. Inuence of oil
In the examined cooling system inevitably some oil will circulate outside the compressor. Since
Isobutane is soluble in mineral oil, any oil present will reduce the pressure in the control volumes, thus giving an error with the proposed technique. It is not likely, however, that oil accumulates in the condenser (which basically is a 14 m long, 3.5 mm internal diameter steel pipe
with downward return bends), the vertically mounted cylindrical shaped lter drier, the capillary

1120

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

Table 1
Experimental procedure for one test
Experimental step

Time

Compressor Valves
open

(1)
(2)
(3)
(4)

0
300 s
650 s
Four
complete
cycles
017 min

O
On
On
On/o
cycling

1,2,3,4,5
1,2,4,5
3
1,2,3,4
5
1,2,3,4
5

O
O
O
O

1
1
1
1

On or o

1,2,3,4

Last scanf

1,2,3,4 On

2400 s

1,2,3,4 O

(8) Control volume 2c

3000 s

3,5

1,2,4

0.1

(9) Control volume 3d

3600 s

2,3,5

1,4

0.1

15 s after valves 14
closingg
5 last scans before
valve 3 opensh
5 last scans before
valve 2 opensh
5 last scans before
pressure equalizationh

(10) Pressure equalization 5400 s

1,2,3,4,5

Pressure equalizing
Evacuation starts
Evacuation ends
Normal operationa

(5) Timing the moment


to close valvese
(6) Measurement start,
control volume 1b
(7) Heater o

Valves Heater Scan rate Data used


closed
(scans/s) for calculation

Based on a separate experiment, showing that the ono cycle was almost unchanged after the fourth cycle.
Evaporator + expansion tank + part of suction line + capillary tube.
c
Filter drier + note 2.
d
Condenser + note 3.
e
Controlled by computer program by elapsed time from compressor start or stop.
f
Expansion tank pressure and temperature.
g
Compressor pressure and temperature. Point of time based upon a separate experiment, in which the temperature
dierence between the internal parts of the compressor (electrical winding) and the compressor surface temperature was
measured.
h
Average expansion tank pressure and temperature.
b

tube, or the suction line. On the other hand, it is possible that oil accumulates in the evaporator
accumulator. For this reason, a separate experiment was conducted to estimate this quantity (see
Appendix A for details). It was found that only a small quantity of oil was present and a more
detailed analyse, using the Raoults mixing rule (cited by [6]), showed that the error introduced
was small.
2.8. Uncertainty analysis
The uncertainty estimations are based on the computerized method described by Moat [7].
Hence, input values are sequentially perturbed with its estimated uncertainty, and the individual
uncertainty contributions accumulated and combined through root-sum-square. Thus, it is assumed that: (a) each variable is independent, (b) each variable is Gaussian distributed if repeatedly
measured, (c) all uncertainties are expressed with the same odds, and (d) the result is a linear function of the input variables (which is true for small perturbations). The estimated uncertainties are
provided in Table 2.

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

1121

Table 2
Experimental quantities
Quantity

Best estimate/range

Uncertainty

Expansion tank volume


Control volume 1
Control volume 2
Control volume 3
Compressor volume
Temperature expansion tank
Temperature compressora
Pressureb
Density-conversionc
Charge
Oil in evaporatord

2.894 l
3.0194 l
3.0307 l
3.1657 l
1.766 l
2426 C
4150 C
0.13 bar

2%
2%
2%
2%
2%
1 C
2 C
1%
0%
0.2 g
0.2 g

35.3 g
0.2 g

Combined uncertainty from (A) temperature measurement and (B) vapour temperature measured indirectly at
compressor surface.
b
Pressure in expansion tank (0.13 bar) and compressor case (0.81.7 bar).
c
Uncertainty in density-conversion assumed embedded in pressure and temperature uncertainties.
d
Based on a separate experiment. See Appendix A for details.

As a starting point, the experimental result, R, is calculated from several variables, Xi:
R RX 1 ; X 2 ; X 3 ;    ; X N :
The uncertainties are estimated for each variable so that the true value is covered with 201 odds
(95% condence):
X i X i Measured or best estimate  dX i

20 : 1:

Next, the input variables is perturbed with each estimated uncertainty resulting in the individual
uncertainty contributions
dRX i

oR
dX i :
oX i

Finally, the overall uncertainty is calculated by the root-sum-square of the individual


contributions
(
2 )1=2
N 
X
oR
dX i
20 : 1:
dR
oX i
i1

3. Results and discussion


3.1. Charge distribution during one ono cycle
The experimental result is shown in Fig. 2. As can be seen, the largest charge-displacements
occur immediately after a compressor start and stop. At the start, the quantity of charge in the

1122

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125


25

ON

OFF

charge (g)

20

15

evaporator
condenser
10

compressor oil

compressor vapour
filter drier

0
0

200

400

600

800

1000

1200

1400

1600

time (s)

Fig. 2. Cooling system charge distribution during one ono cycle.

evaporator is large due to o-cycle migration. Immediately after the start it decreases rapidly to a
minimum of less than 4 g in half a minute. During this part of the cycle the system is strongly
unbalanced with a high mass ow through the compressoras a consequence of a high evaporation temperature, and a low mass ow through the capillary tubeas a consequence of vapour in
the capillary tube. For the rest of the on-cycle, the evaporator charge increases as the lter drier,
condenser, and compressor charges decreases.
At the compressor stop (after about 600 s), the system undergoes a pressure equalisation. This
can be clearly seen on the evaporator and condenser masses. The refrigerant moves from the condenser to the evaporator in two and a half minutes. It is interesting to note that, for the remaining
o-cycle, the evaporator charge decreases from a peak of 22 g to a low 19 g. Apparently, the
refrigerant is evaporating in the evaporator during the o-cycle, and as can be seen, dissolves into
the compressor oil and increases the mass in the compressor vapour.
Fig. 2 also provides the uncertainties, given at 95% condence. In absolute numbers the uncertainties are low. The largest uncertainties are associated with the quantity of charge dissolved into
the compressor oil. This is a consequence of the calculation procedure in which this charge is calculated as all other charges subtracted from the total charge. Thus, several uncertainties are inuencing this quantity. In total, 17 input variables were perturbed. The largest uncertainty
contributions (averaged over all measurement points) were caused by the uncertainties in the pressure (adding 0.25 g for the condenser charge uncertainty) and the uncertainty in the quantity of
total charge (adding 0.2 g for the uncertainty of the charge dissolved into the compressor oil). The
uncertainty contribution due to the oil in the evaporator was less than 0.06 g.

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

1123

Table 3
Experimental results of comparison between techniques
Volume

Remove-and-weigh technique (g)

New technique (g)

Dierence (g)

Dierence (%)

Evaporator + tank
Filter drier
Condenser
Compressor
Total

12.2
2.1
11.9
8.8
35.0

12.12
2.21
11.67
9.0a
b

0.08
+0.11
0.23
+0.2

0.7
+5.2
1.9
+2.3

a
b

Of which 2.9 g as vapour.


By denition equal to the experimental 35.0 g that was taken as the true mass.

3.2. Comparison between the proposed technique and the remove-and-weigh technique
The quantities of refrigerant, as measured by the remove-and-weigh technique and the new
technique, are shown in Table 3. As can be seen, the dierences are small ranging from less than
1% for the evaporator to 5% for the lter drier. Due to the interaction between the charge in the
compressor vapour and the compressor oil, it was only possible to measure the total compressor
charge (i.e. not explicitly the charge as vapour and the charge dissolved into the oil) with the remove-and-weigh technique.

4. Summary and concluding remarks


The objective of this work was to describe and evaluate a refrigerant mass measurement technique, based on the quick-closing valves technique and the equations of state. The results indicate,
overall, that the proposed technique should make a useful contribution to future charge/void
measurements. Timesavings are possible through autonomous operation. Errors caused by leakages in the remove-and-weigh technique are eliminated. Experiments showed that the technique is
accurate.
The technique was implemented on a domestic refrigerator cooling system. A test series was
carried out from which preliminary results are reported of the charge distribution during one
ono cycle. The results showed o-cycle migration and on-cycle redistribution of the refrigerant.
In general, the error analysis resulted in low uncertainties for the experimental results. It was
found that the amount of oil in the evaporator was small, and that its inuence on the measured
charge was small.
The proposed technique, implemented for autonomous operation, is restricted to compressor
systems in which re-established running conditions, after a measurement, is possible by pump
down of the tank. Another (but general) limitation is if a large and unknown quantity of oil, soluble into the refrigerant, exists within the control volume. In this case, the errors may become
considerable.
It is recommended for future work to use a tank as large as possible. By this, uncertainties are
decreased and measurement conditions are established in a shorter time. In fact, by using a tank
much larger than the control volume, the volume and temperature of the control volume itself

1124

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

becomes less important. Another possible improvement would be to electrically heat the compressor, thus driving the refrigerant out of the oil so that it could be directly measured.
It is also recommended for future work to use the experimental set-up itself for volume determination of the control volumes. First, the tank volume is determined from measurements or
given data. Then, the tank is charged with a suitable gas while the other parts of the system
are evacuated. Finally, the gas is expanded stepwise from the tank into the control volumes while
pressure and temperatures are observed so that the volumes can be calculated.
Acknowledgement
The Swedish National Energy Administration and Electrolux AB provided support for this work.

Appendix A
A separate experiment was conducted in order to estimate the quantity of oil in the evaporator.
A refrigerator of the same type was modied so that the evaporator refrigerant-line (and accumulator) was visualised through an inspection glass.
First, the refrigerator was run for four consecutive ono cycles to stage a base-condition.
Then, the compressor was powered for a certain time (that was varied in the experiment between
0 and 40 h) before switching it o. Since the refrigerant is soluble into the oil, the refrigerant dissolves into the relatively large compressor oil tray. After 20 h the remaining liquid in the evaporator was observed (using a digital camera) and interpreted for its volume. To ease the
observation the refrigerator was slightly tilted, so that liquid in the evaporator accumulated in
the corners. The result is shown in Fig. 3.
3

liquid (ml)

2.5

1.5

0.5

0
0

10

15

20

25

30

continuous compressor-on time (h)

Fig. 3. Observed quantity of liquid in evaporator.

35

40

45

E. Bjork / Applied Thermal Engineering 25 (2005) 11151125

1125

Bearing in mind that the liquid observed also includes some dissolved refrigerant (in fact supported by observations at the compressor start-up), it is seen that the oil accumulates with compressor on time to a quantity of about 2.5 ml. At this level it remains steady even for longer
compressor on times. It is also seen that the oil quantities at shorter compressor on times (representative for the normal ono cycle) is very small. At the compressor start-up, in a regular ono
cycle, it was observed that liquid refrigerant was pumped out from the evaporator as liquid slugs.
It appears that oil is carried back to the compressor at this time in the ono cycle. On the limited
data available and taking into account that a small quantity of oil also could be accumulated as a
thin oil-layer covering the refrigerant-line it was estimated that 0.2 (0.2) g of oil was accumulated
in the evaporator at regular ono cycling mode.

References
[1] G.F. Hewitt, Measurement of Two Phase Flow Parameters, Academic Press Inc., 1978, ISBN 0-12-346260-6.
[2] W.J. Mulroy, D.A. Didion, Refrigerant migration in a split-unit air conditioner, ASHRAE Transactions 91 (Part
1A) (1985) 193206.
[3] M.J.P. Janssen, Cycling losses in cooling circuits, MSc Thesis, WOP-WET 89.002, Eindhoven University of
Technology, 1989.
[4] P.B. Whalley, Boiling, Condensation and GasLiquid Flow, Oxford University Press, 1987, ISBN 0-19-856181-4.
[5] J. Philipp, W.E. Kraus, H. Quack, Numerical simulation of refrigeration cycles with capillary tubes in on/o cycling
mode, in: Proceedings of the 4th IIR-Gustav Lorentzen Conference on Natural Working Fluids, Purdue University,
USA, 2000.
[6] J.J. Grebner, R.R. Crawford, Measurement of temperaturepressureconcentration relations for mixtures of R-12/
mineral oil and R-134a/synthetic oil, ASHRAE Transactions 99 (1) (1993) 387396.
[7] R.J. Moat, Describing the uncertainties in experimental results, Experimental Thermal and Fluid Science 1 (1988)
317.

S-ar putea să vă placă și