REDUCTION OF THE AMOUNT OF FLAMMABLE REFRIGERANT IN

A SMALL-CAPACITY DOMESTIC HEAT PUMPS

O. Samoteeva, B. Palm, P. Lundqvist, P. Fernando

Division of Applied Thermodynamics and Refrigeration
Department of Energy Technology
Royal Institute of Technology
10044 Stockholm, Sweden

ABSTRACT

This paper presents an experimental and theoretical investigation of the charge distribution,
possibilities to reduce the total amount of refrigerant, the void fraction, and pressure drop in a
heat pump system and its components. The research is a part of a project that has as a goal the
development of an efficient domestic heat pump system that would cover a big part of the energy
needs (about 5 kW) of an average single-family house, while safety is carefully considered by
keeping the refrigerant charge under 150 grams.

For this purpose, a test unit was built and propane was chosen as the working fluid. Rectangular
Copper Micro Channel (RCMC) heat exchangers with smaller internal volume were used. A
theoretical model of the charge distribution in the heat exchangers was developed and theoretical
results were compared to the experimental values.

1. INTRODUCTION

During the last years an increasing tendency towards the use of natural refrigerants in all possible
applications is observed. Hydrocarbons have proved to have many advantages in their use as a
refrigerant. Their flammability appears to be the main problem when used in closed or inhabited
spaces. The IEC&ISO standards include a safety standard for household refrigerators (IEC
60335-2-24) that allows the use of flammable refrigerants in quantities of up to 150 grams. This
standard can be applied to household heat pumps as well. The system that contains up to 150
grams of refrigerant requires small internal volumes of heat exchangers, compressors and pipes
in the system.

Several models for prediction of void fraction and pressure drop in two-phase flow have been
proposed. The simplest to use is the homogeneous flow model. Prediction models for two-phase
flows were developed by Lockhart and Martinelli (1949), Hughmark (1962), Bankoff (1960),
Zivi (1964), Smith (1969), Chisholm(1972), Premoli et al. (1971), Graham et al. (1998), Yashar
et al. (1998) and others. Rice (1987) classified void fraction prediction approaches and applied
them to refrigerant charge prediction. Few investigations have been made on the charge amount,
distribution and reduction issue.

For several years already a number of investigations on the topic of heat pump running on
hydrocarbons have been performed at the Royal Institute of Technology, Stockholm. Högström
(1999) [20] and Pelletier (1998) [21] have studied propane as refrigerant in residential heat
pumps. Currently a project on the subject of charge minimisation in residential heat pumps
working on propane is running at the Division of Applied Thermodynamics and Refrigeration. A
test facility representing a domestic heat pump having heating capacity of 5 kW has been
constructed. Experimental and theoretical investigations are made for different types of heat
exchangers, with the aim of minimization of the system’s internal volume [5], [6]. In this paper,
calculated parameters are compared to the measured ones and the most appropriate correlations

1
for this type of heat exchangers are determined. Results are discussed and on their basis
recommendations and planning of future investigations are made.

2. METHODS

The research work consists of two parts: experiments and modelling. Experiments were
performed in order to determine the performance, charge distribution and amount in the system,
as well as to collect data for use in theoretical analysis. Modelling deals with prediction of the
void fraction, pressure drop and calculation of the amount of refrigerant in different system
components.

2.1 Experiments
A test rig resembling a small-capacity domestic heat pump has been built in order to investigate
the goals set by the project. Generally, the system represents a basic heat pump cycle, consisting
of two heat exchangers used as evaporator and condenser, compressor, thermostatic expansion
valve and tubing. System components were chosen using software programs provided by the
companies-producers.

C O MPRESSO R
V2
BRINE INLET
GFM

P
DP

P V3
EVAPO RATO R
W AT ER OUT LET

WFM
C O NDENSER
DP

WAT ER INLET
SIT E GLASS

V4

EXPANSIO N BRINE OUT LET

VALVE
V1

Figure 1. Experimental test rig with RCMC as evaporator and condenser

Experiments were performed with Rectangular Copper Micro Channel heat exchangers used as
evaporator and condenser, scroll compressor designed for R407C and thermostatic expansion
valve. The system is provided with four pneumatically operated valves (V1-V4) placed in order
to separate the refrigerant in different system components when the amount of refrigerant in
them is to be measured. System is also provided with thermocouples, pressure and pressure
difference transducers, connected to the computer through a data logger. The schematic view on
the test rig is given in the figure 1.

Several objectives were followed while performing experiments. One was to determine the
optimal charges for different working conditions and the second main objective was to find the
charge distribution in different system components for optimal charges and to measure pressure
drop in the heat exchangers for each set (each evaporation temperature). Test series were done at
four different evaporation temperatures –16, -8, 0 and +5 0C, at different charges for each
temperature condition. The water outlet temperature from the condenser was fixed at 40 0C and
the superheat was kept about 5 degrees.

2
2.2 Modelling
The same initial conditions as in the experimental part were used for the calculations. At present
modelling of the void fraction and pressure drop is presented only for evaporator. The work on
the condenser is ongoing and will be presented in the following papers. Many void fraction and
pressure drop correlations have been developed for round tubes of different diameters and only
few investigations of the void fraction and pressure drop in rectangular channels were made. The
goal of the present study is to determine which of the existing models developed for round tubes
of different diameters can be used for upward vertical flow in rectangular microchannels.

For the calculations it was assumed that the heat flux is constant and the quality varies linearly
with a 0,01 step from the inlet value until x=1 along the boiling part of the channel. The void
fraction value was integrated along the channel. For void fraction prediction in the current work
several models were used. The homogeneous flow model’s main assumption is that both phases
of the two-phase flow are mixed and have the same velocity so that the flow behaves as one-
phase. The Lockhart-Martinelli model [10] is based on their research on pressure drop in
adiabatic two-phase flow in pipes. They introduced a new parameter now called the Lockhart-
Martinelli parameter (Xtt). Chisholm’s slip-ratio model uses the same equation of the void
fraction as described previously in the homogeneous model, determining the slip ratio in his way
[2]. The CISE model was developed by Premoli et al. (1970) in the work on prediction of the
density of two-phase flow under adiabatic conditions [2]. The void fraction is calculated using
general formula, but the slip ratio is a function of quality, liquid and gas densities, the Reynolds
number (mass flux dependent) and the Weber number [2, 3]. UI ACRC (The University of
Illinois Air Conditioning and Refrigeration Center) has developed a void fraction relation for
evaporation in all tubes. Wilson et al. (2000) [8] have applied this method for flattened tubes of
different sizes and have replaced the diameter term with the channel height. Hughmark’s method
[11] is one of the most complex ways of calculating the void fraction due to its iteratively
determined coefficient. This correlation for void fraction is empirical and based on the use of
Bankoff’s flow coefficient K [12].

The total pressure drop in two-phase flow in a straight tube is a sum of three components,
gravitational (or static) pressure drop, momentum pressure drop and frictional pressure drop. The
component that has most influence on the total pressure drop in tubes is the frictional pressure
drop. Frictional pressure drop in two-phase flow can be predicted in two main ways: using a two-
phase multiplier for the one-phase pressure drop or using friction factors developed for two
phase flow. Friedel’s correlation for the frictional two-phase pressure gradient is written in terms
of the two-phase multiplier, defined in the detailed description of the method [2, 7]. The Pierre
model was developed to calculate frictional pressure drop in horizontal two-phase flow. Values
for the friction factor and the way of calculation of the mean specific volume can be found in the
literature [9]. Granryd (1985) has investigated frictional pressure drop specifically for this type
of heat exchangers. Results in terms of pressure drop were correlated by using a modified Pierre
model with fm=0.083, resulting in the equation: ∆p = f m * G 2 *ν '' * L / Dh
.
4. RESULTS

4.1 Experiments:
The optimal charge for each condition was determined by plotting the coefficient of
performance of the heat pump versus the total amount of refrigerant in the system. According
to the figure 2, the optimal charge is increasing with the increase in evaporation temperature. It
can be seen that the optimal charge varies from about 170 to 250 grams for different
conditions. It implies the necessity to install a receiver, which would help to adjust the
necessary charge amount. In the test system the length of gas and liquid line tubing is bigger
than it can be in the real system due to the presence of the valves. It was estimated that the

3
 amount of refrigerant in the real system can be reduced by about 20 to 30 grams due to the
elimination of the valves.

120
5,8 Evaporator

Amount in the components, g

Condenser
5,3 100
+5°C Liquid Line

4,8 Gas
line&Compressor
0°C 80 Linear (Evaporator)

4,3
COP1

60
3,8 -8°C
3,3 -8°C 40
-16°C
2,8 -16°C 0°C 20
5°C
2,3
0
140 160 180 200 220 240 260 280 170 190 210 230 250 270 290
Total charge in the system, g Total charge, g

Figure 2. Variation of the coefficient of performance Figure 3. Variation of the amount of refrigerant in different
(COP1) with the total charge in the system for different system components with optimum total charge in the system
evaporation temperatures for different evaporation temperatures (-16, -8, 0, +5)

Determination of the amount of refrigerant in different system components was done using
pneumatically operated shut-off valves. The refrigerant was drained from each part of the system
and weighed. Figure 3 indicates the measured amount of refrigerant at the evaporation
temperatures –16, -8, 0 and +50 with the charge giving the highest COP in the system. As can be
seen from the graph, the amount of the refrigerant in the evaporator, condenser and lines doesn’t
vary with the charge due to the fact that the system works at different evaporating temperatures.
The additional charge at increased pressure is accumulating in the compressor. Some of the
refrigerant that accumulates in the compressor is dissolved in the compressor oil and cannot be
easily drained.

4.2 Modelling
Figure 4 presents the calculated mass of refrigerant plotted against the experimental values. As
can be seen from the graph, the use of the Hughmark model for void fraction prediction gives
results that are close to the experimentally determined values (up to 5% max deviation) of the
amount of refrigerant in the evaporator tubes. High precision of the method is achieved by the
iterative calculations of the dependent parameters. The UI ACRC model was also applied to the
present study, first using the values for a and n proposed by Wilson et al. [8] and then adjusted
(curve fit) for the channel height used in the present study - 0.6 mm. As can be seen from the fig.
4, the correlation then predicts the amount of refrigerant in the evaporator with a high precision
(max 5% deviation). The CISE correlation under predicts the amount of refrigerant in the
evaporator tubes by 18-20%. The Lockhart-Martinelli model and Slip ratio-correlated model,
using Chisholm’s (1972) parameter give rather big errors in predicting the amount of refrigerant
in the studied evaporator. The predicted amount of refrigerant given by these two models differs
from the experimentally defined values by up to 70%. The homogeneous flow model shows the
biggest deviation of the predicted values from the experimentally measured data. A model based
on perfect annular flow was developed by Granryd. This model over predicts the refrigerant
charge by about 20 to 30%. These data are not included in the figures.

4
Figure 4. Comparison between predicted and Figure 5. Comparison between predicted and
measured amount of refrigerant in evaporator measured pressure drop

The pressure drop was calculated using three correlations and the obtained results were
compared to the experimentally determined values. Results are presented in figure 5. The Friedel
correlation is recommended by Whalley [2] to be used in the range µl/µg<1000. Our working
conditions for all investigated cases are within this range. As can be seen from the graph, the
deviation from the measured pressure drop is over 50%. It probably can be explained by the fact
that this method does not work very well for the case of rectangular microchannels, depending
on flow development. The Pierre correlation was developed for two-phase flow in horizontal
channels, but it works for vertical flow as well. However, our domain of application is outside
the boundaries defined by Pierre. Nevertheless, the prediction of the pressure drop shows a
deviation of no more than 27 to 50% (see fig 5). The use of Granryd’s empirically defined
friction factor fm=0.083 in Pierre’s correlation and using the arithmetic mean value of the vapour
quality instead of the formula defined by Pierre, gives only deviations of 15% in the low
temperature region and deviates even less for higher evaporation temperatures. These deviations
can be explained by uneven width inside the channels (channel width can vary by 0.01 mm),
which influences calculations and also by some uncertainties in the measurements.

5. CONCLUSIONS

Several objectives have been followed while testing the system with RCMC HE as evaporator
and condenser. Tests at different evaporation temperatures were performed and optimal total
charge in the system was found to increase with increasing evaporation temperature. It was
observed that the amount of refrigerant in all the system components, except in the compressor,
does not change when system works with optimum charge at different evaporation temperatures.
Refrigerant is accumulating in the free volume of the compressor, some being dissolved in the
oil. Theoretical investigations on the void fraction and amount of refrigerant in RCMC heat
exchangers have shown that Hughmark and adjusted IU ACRC model have an accuracy of up to
5% in predicting the amount of refrigerant in the evaporator. The CISE model gives an
underprediction in the range of 20%. Pressure drop prediction shows that the Pierre model
adjusted by Granryd for this specific type of heat exchanger gives the best approximation. The
deviation is only about max 15% and a part of the error can be caused by uneven channel
dimensions and uncertainties in measurements.

5
 6. NOMENCLATURE

Latin symbols v– specific volume [m3/kg]

Dh – hydraulic diameter [m] Xtt – Lockhart-Martinelli parameter [-]
fm – Pierre’s friction factor [-]
G – mass flux [kg/m2s] Greek symbols
L – channel length [m] µ - viscosity [Ns/m2]
p – pressure [Pa]

7. REFERENCES
1. R.W. Lockhart, R.C. Martinelli, 1949, Proposed correlation of data for isothermal two-phase, two-component
flow in pipes, Chemical engineering progress, Vol45, No.1,
2. P.B. Whalley, 1987, Boiling, condensation and gas-liquid flow, Oxford University Press
3. C.K. Rice, 1987, The effect of void fraction correlation and heat flux assumption on refrigerant charge
inventory predictions, ASHRAE Transactions, Vol.93, Part 1.
4. S.A. Klein, and F.L. Alvarado, 2002, Engineering Equation Solver, software, Academic version.
5. P. Fernando, O. Samoteeva, P. Lundqvist, B. Palm, August 2001, Charge distribution in a 5 kW heat pump
using propane as working fluid. Part I: Experimental investigation, Proceedings of 16 Nordiske Kolemode og
9 Nordiske Varmepumpdage.
6. O. Samoteeva, P. Fernando, B. Palm, P. Lundqvist, August 2001, Charge distribution in a 5 kW heat pump
using propane as working fluid. Part II: Modelling of liquid hold-up, Proceedings of 16 Nordiske Kolemode og
9 Nordiske Varmepumpdage.
7. G.F. Hewitt, 1998, Heat Exchanger Design Handbook, Part 2 –4, Begell house, Inc. New York, Wallingford
(UK).
8. M.J. Wilson, J.C. Chato, Ty A. Newell, 2000, A study of refrigerant pressure drop and void fraction in
flattened copper tubes, Proceedings of the 2000 International Refrigeration Conference at Purdue.
9. Bo Pierre, 1949, Pressure drop in boiling refrigerants, Kyltekniksk tidskrift, Dec 1957. In Swedish.
10. R.W. Lockhart, R.C. Martinelli, 1947, Proposed correlation of data for isothermal two-phase, two-component
flow in pipes, Chemical Engineering Progress, Vol. 45, No 1.
11. G.A. Hughmark, 1962, Holdup in gas-liquid flow, Chemical Engineering Progress, Vol. 458, No 4.
12. S.G. Bankoff, November 1960, A variable density single-fluid model for two-phase flow with particular
reference to steam-water flow, Journal of Heat Transfer.
13. G.B. Wallis, 1969, One-dimensional Two-phase Flow, McGraw Hill, New York.
14. E. Björk , 2002, Pressure drop in a plate evaporator for refrigerators, Proceedings of the 1st International
Conference on Heat Transfer, Fluid Mechanics and Thermodynamics.
15. P.L. Spedding, J.J.J. Chen, 1986, Hold-up in multiphase flow, Encyclopaedia of fluid mechanics, Vol.3,
Houston, USA.
16. E.Granryd et all, 1999, Refrigerating engineering, Royal Institute of Technology, Stockholm, Sweden.
17. A.J.Holt and B.J.Azzopardi, 1995, Two phase pressure drop and void fraction relevant to compact two-phase
heat exchangers, 4th UK National Conference on Heat Transfer.
18. R.W.Brown, A.Gomezplata and J.D.Price, 1969, A model to predict void-fraction in two-phase flow Chemical
Engineering Science, Vol.24, pp.1483-1489.
19. J.J.J Chen and P.L.Spedding, 1981, An extension of the Lochart-Martinelli theory of two-phase pressure drop
and holdup, International Journal of Multiphase flow, Vol.7, No.6, 659-675.
20. C. Högström, 1999, Värmepump med propan som köldmedium, Licenciate thesis, Royal Institute of
Technology, Stockholm.
21. O. Pelletier, 1998, Propane as refrigerant in residential heat pumps, Licenciate thesis, Royal Institute of
Technology, Stockholm.