Advance Proof.

Private to members
Copyright © 2013 The Institute of Refrigeration
No publication or reprinting without authority

THE INSTITUTE OF REFRIGERATION

Pioneering Applications of Refrigeration
Technologies
by

Pega Hrnjak
Winner of the IOR J&E Hall Gold Medal

Professor and Co-Director,

Air Conditioning and Refrigeration Center, The University of Illinois, Urbana Champaign
President,
Creative Thermal Solutions
pega@illinois.edu

(Session 2012-2013)

To be presented before the Institute of Refrigeration at

London Chamber of Commerce and Industry, 33 Queen Street, London, EC4R 1AP
On Thursday 7th February 2013 at 5.45pm

Introduction

Refrigeration is a mature technology. Numerous smart and well prepared scientist and engineers have
been building the structure and knowledge in thermodynamics, fluid related areas, heat transfer,
materials, compressors, auxiliaries etc… to be used in refrigeration over a century and thus many
think that in scientific sense refrigeration is a closed book. Needless to say we do not share that
position, neither in scientific nor in engineering aspect of the statement. Numerous are examples to
oppose assumption of exhausted creativity and invention in refrigeration: new compressor types
(geometry and avoidance of lubricants), heat exchangers (microchannel and other new types),
refrigerants (new environmentally friendly man made fluids and resurgence of natural refrigerants like
CO2, ammonia, water,…), new cycles and control options, etc… The truth is that every new concept
in refrigeration requires solid evaluation against well established and carefully optimized existing
solutions which slows down the development process compared to new, emerging technologies.

This paper presents some of authors pioneering concepts with emphasis on microchannel heat
exchangers, condensers and evaporators in particular with flash gas bypass and periodic reverse flow.

Proc. Inst. R. 2012-13. 5-1

Example 1: Microchannel Condensers for Ammonia Chillers

Ammonia is well known to be a very efficient refrigerant, inexpensive and natural. Over a century of
proven service in industrial refrigeration offset its mild flammability, bad or alarming odour and mild
toxicity of this refrigerant. Nevertheless, ammonia was never well accepted as a refrigerant for
applications in urban areas mostly because of the smell that could be even panicking in the case of leak
for unfamiliar people. On the other hand, ammonia is also one of rare refrigerants that have vapour
lighter than air. That characteristic puts ammonia in safety advantage because refrigerant vapour goes
upwards, thus avoiding creation of LFL (lower flammability level) values when unobstructed upward
paths are provided. In addition, ammonia has the lowest demand for charge in microchannel HXs
compared to any currently used refrigerant. All said above leads to an excellent opportunity for
reconsidering ammonia as a refrigerant in urban areas in some applications: very low charged,
hermetic chillers placed on the roof with unobstructed vapour release.

Such small (compared to industrial size), low charged ammonia systems, chillers in particular, used for
refrigeration (commercial – supermarkets, or similar) or air conditioning with a coolant as a part of
secondary loop or cascade system provide excellent potential for new market penetration of
ammonia. In that way the excellent thermodynamic and thermophysical properties of the fluid can be
taken full advantage of and systems are more likely to abide by local codes due to the extremely low
charge.

The author with Andy Litch, a graduate student, pioneered use of microchannel condensers more than
15 years ago when building the first air cooled ammonia chiller with microchannel condenser. At that
time there were no hermetic ammonia compressors so open Bitzer compressor was used, along with
SS Ni brazed plate evaporator Alfa Laval. Since immiscible oil was used, a simple architecture was
implemented to provide oil return. That work resulted in the lowest specific charge air-cooled chiller
for ammonia reported in the literature so far. The chiller is shown in the Fig. 1.

Microchanell
condenser

Sight glass
Receiver

   Expansion     
v   valve
Flowmeter

    Evaporator Compressor

Figure 1. Prototype of an ammonia chiller with air cooled microchannel condenser in the
wind tunnel.

Proc. Inst. R. 2012-13. 5-2

A two cylinder, open reciprocating compressor directly coupled to a 6kW VFD controlled motor was
used. A helical oil separator removed mineral oil and returned it to the compressor crankcase. The
superheated compressor discharge line enters the wind tunnel test section going to the condenser
inlet and subcooled ammonia exits the condenser and wind tunnel going to the receiver, which was
made a high pressure steel sight glass. Liquid ammonia passed through a coriolis type mass flow
meter, which measured mass flow rate and density. Throttling was done by a manually controlled
needle metering valve. The evaporator was a nickel brazed plate HX. The warm glycol from the heat
recovery loop generates a heat load. The ammonia flowed downward as it travelled through the
evaporator (despite distribution concerns) to ensure positive return of the small quantities of mineral
oil back to the compressor. A suction line accumulator was used to prevent liquid slugging of the
compressor in some cases. Two shut-off valves were mostly closed isolating the accumulator from
the rest of the ammonia loop and the accumulator was eliminated in charge determination.

Two condensers were evaluated in the prototype: one with a single serpentine tube and the other
with a parallel tube arrangement between headers having 24 tubes in the first pass and 14 in the
second. Each tube has 19 triangular ports of equal dimensions with a hydraulic diameter less than 1
mm. The fins were multi-louvered. The serpentine condenser had a single tube that passes 16 times
through multi-louvered fins. There are five enhanced square ports in the tube. Additional details of
these condensers may be found in Hrnjak and Litch, 2008 or more detailed in the report Litch &
Hrnjak, 1999.
619

25.45

404

Refrigerant In

28.6 Fin Pitch: 2.03 Refrigerant Out

Tube/Fin Cross Section

19.1 17.1 Louver Cross Section

1.72
Dh = 4.06 27

6.35

28.6

Figure 2: “Macrochannel” serpentine condenser.

Figure 3: “Microchannel” parallel condenser with air exit thermocouple grid.

Proc. Inst. R. 2012-13. 5-3

Overall heat transfer performance and charge measurements were taken for each condenser and the
system as a whole. The microchannel heat exchanger with parallel flow performed better in every
respect. Overall, condenser performance was quantified in terms of U values for different air flow
rates, superheating and subcooling conditions and is presented in Fig. 4.
140
Serpentine
200 Parallel flow
120

160
100

Uair (W/m2-C)
80
Uair (W/m2-C)

120

60
80
< 10 °C Subcooling 7 °C <= Subcooling < 10 °C
>= 10 °C Subcooling 40 10 °C <= Subcooling < 12 °C
12 °C <= Subcooling < 14 °C
40
20

0 0
0 1 2 3 4 5 0 1 2 3 4 5
Vface (m/sec) Vair,face (m/sec)
Figure 4. Overall heat transfer coefficients for two microchannel condensers for
ammonia.

Parallel flow 0.30

Serpentine
0.12 +15%
Newell, 1999
0.25 Butterworth, 1975
0.10 Zivi, 1964

-15%
0.20
0.08
mref,pred (kg)
mref,pred,st (kg)

0.15
0.06

0.10
0.04

0.02 0.05

0.00 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.00 0.05 0.10 0.15 0.20 0.25 0.30
mref,exp (kg) mref,exp (kg)

Figure 5. Predictions and measured values for the charge in two condensers.

Refrigerant inventory measurements of the condenser were taken at different operating conditions.
Refrigerant inventory measurements are compared to model results using different void fraction
model predictions. All void fraction correlations perform similarly in helping to predict total charge.
The use of Newell’s correlation (Newell et al. 1999) for serpentine condenser yields the smallest
average error of 9.3%, with a maximum of 15.7%. With the Butterworth 1975 and Zivi 1965
correlations, the average and maximum errors are 10.1/22.8% and 12.3/24.9%, respectively. The slight
over prediction results in a simulated subcooled region that is larger than the actual region, inflating
predicted charge. Data by Adams, Hrnjak and Newell 2003 fit well in the prediction. These results
are presented in Fig. 5.

Proc. Inst. R. 2012-13. 5-4

Obviously, predictions for a serpentine condenser were much more accurate than for microchannel
when using the same correlation and experimental data. That clearly indicates a significant inaccuracy
in the charge prediction in headers (see Fig. 6). Another insight from Figure 6. (serpentine) is that
liquid subcooling is a large contributor to total charge. The relative predicted charge contributions
from the refrigerant phase zones for the data point with the highest liquid subcooling tested are 0.5%
in superheated zone, 29.2% in the two-phase region, and 70.3% in subcooling region. From the data
point with the lowest liquid subcooling, the contributions are 0.5%, 60.1%, and 39.4% in subcooling.
Even though the subcooling region is only 26% of the total tube length, it comprises 70% of the total
charge. Thus it is advantageous to reduce subcooling not only for increased heat transfer, but to
reduce refrigerant charge.
8
0.008

7
0.007
70
0.0018
Header Section 4 4
Header Section

0.0016
6
0.006
0.0014
Header Section 2
5
0.005 Header Section 2 Header Section
Header Section 3 3

Refrigerant mass [g/m]

0.0012
mref,i (kg)

mref,i (kg)
4
0.004 0.0010
Mass [g]

3
0.003 0.0008

0.0006
2
0.002
Header Section
Header Section 1 1
0.0004
1
0.001
0.0002

0
0.000 0.0000
0.0 00 11 2 33 4 55 66 77 88 99 10
0 0.20.2 0.4
0.4 0.6
0.6 0.8
0.8 1.0
1.0 1.2
1.2 1.4
1.4 2 10
Condenser Circuiting Length (m) Condenser Circuiting Length (m)
Circuiting length [m] Length [m]

Figure 6. Charge distribution in two microchannel condensers.

From the experimental data taken, the microchannel parallel flow condenser appears to outperform
the macrochannel serpentine condenser. The overall heat transfer coefficient for a given surface
velocity was 60-80% higher than for the serpentine condenser (mainly due to higher air side heat
transfer coefficient) and the charge is an average of 53% less. The microchannel condenser has a
smaller volume for approximately the same surface area. Also, it has less charge and better heat
transfer than the serpentine and typical condensers. For details and additional information please see
Hrnjak and Litch, 2008 ore the report Litch and Hrnjak 1999.

A comparison of the system charge to evaporator capacity for commercially available units to that of
the experimental chiller setup used in this study is given in Table 1. The value for the experimental
chiller, 18 g/kW, was chosen from data point 4 (see Litch & Hrnjak, 1999), a nominal high load case
similar to the conditions to which the commercial chillers were rated. Total system charge measured
was 0.24 kg. The numbers for the commercial chillers were selected from the rating tables of the
smallest chiller listed in the literature. It should be noted that the experimental chiller is not really
optimized for charge size. The compressor was oversized even though it is the smallest reliable
ammonia compressor available. The focus of the project was on condensers as the major contributors
to charge and performance. Furthermore, there are instrument fittings and additional tubing for
accessibility and routing that add excess internal volume. Thus, the number for the chiller charge to
capacity ratio could be even lower.

Proc. Inst. R. 2012-13. 5-5

Table 1. Charges of ammonia chillers at the turn of the century

Chiller System Capacity, System

Evaporator specific
[kW] charge
[g/kW]
Air cooled:
Litch & Hrnjak (MC condenser) 13 18
Refcomp VKA16-14 16 125
York YSLC F4F00UW 220 129
N.R. Koeling LK 25 25 159
Water cooled:
ILKA MAFA 100.2-11K45 108 23
ABB (York) BXA 108 157 – 43
Gram (York) LC 38 – 228 228 - 37
Sabroe (York) PAC 57 – 1074 172 - 36

There was a long incubation period for new developments in hermetic ammonia compressors and
microchannel heat exchangers as condensers became a reality in small chillers.

At first the use of microchannel HXs in industrial refrigeration systems seemed to be received
sceptically due to its size, type and perceived capacity issues, along with potential corrosion,
distribution and defrosting issues. Nevertheless, that long incubation period helped professional
circles to accept the concept and start slowly but surely to consider it. Larger sized chillers for
synthetic refrigerants were first to adopt microchannel condensers for units sized for a few hundreds
of kW cooling capacity. It was just a matter of time until ammonia systems would try this approach.

Probably the most important development in this area was the new hermetic compressor for
ammonia that is used for both refrigeration and heat pumping. The wrap is designed for ammonia,
with a nominal capacity in cooling (at -5oC/50oC) of 45 kW and in heat pumping of 47 kW. The motor
is IPM type with aluminium windings. There are two models: for low and high temperature. The
weight of the hermetic version is about 100 kg. This compressor is equipped with an oil pump.

Table 2. Characteristic of the first chiller unit with hermetic scroll compressor
New Mycom unit with hermetic scroll compressor
Shell High pressure chamber
Design pressure 2.7 MPa
Motor IPM type Al windings 15 kW
Operating Design temperature 120 oC
Conditions Condensing temperature 30 to 55 oC
Evaporating temperature -35 to +10 oC
Rotational speed 1800～3600 rpm
Lubrication Oil pump
Oil cooling Liquid injection
Oil type PAG
Weight 100 kg
Refrigerant charge 6 kg
Capacity at -5/50 oC Cooling 45 kW
Heating 47 kW

Proc. Inst. R. 2012-13. 5-6

 270 mm

700 mm

Figure 8. New Mycom chillers with

Figure 7. New Mycom compressor: cross hermetic ammonia scroll and conventional
section and photo. condensers.

Development work on microchannel condensers for ammonia has moved from the Air Conditioning
and Refrigeration Center at the University of Illinois to Creative Thermal Solutions (CTS), a high tech
company that specializes in research and development of novel refrigeration and air conditioning
approaches. Figure 9 presents a photo of a condensing unit with microchannel heat exchanger used in
an experimental facility for evaluating ammonia evaporators that has been in use for 10 years. CTS
works on the evaluation of various systems and refrigerants and has long term experience in this area.

It was great privilege to help Mycom in developing the first air cooled chiller using a hermetic ammonia
compressor with microchannel condenser. That design significantly reduced refrigerant charge,
reduced the size and weight of the condenser and paved the road for expansion of the use of ammonia
in various indirectly cooled refrigeration and air conditioning applications: supermarkets, commercial
air-conditioning, process cooling and similar. Location of the extremely low charged chiller on the
roof dramatically reduces the possibility of even smelling ammonia in the unlikely event of leakage.

A new semihermetic compressor for use with ammonia was announced in 2008 and a complete chiller
last month.

Axial fan

Microchannel condenser

Pressure transducers

Compressor behind

Pressure cut-offs

Figure 9. Condensing unit with microchannel condenser normally used at CTS during
ammonia evaporator studies.

The use of microchannel heat exchangers made of aluminium paves the road for another significant
potential improvement: different material selection. Ammonia always suffered from poor compatibility
with copper. That was perceived as a drawback when compared to systems that were able to use that
lighter weight material, easier to work with and less corrosive all compared to steel. The full
utilization of aluminium is another, yet unrealised potential for greater competitiveness of ammonia.

Proc. Inst. R. 2012-13. 5-7

With the increasing cost of copper even in systems with conventional refrigerants aluminium is
opening new applications. That opportunity should be used for ammonia.

Example 2: Flash Gas Bypass and reversed FGB

Background

There is no doubt that microchannel heat exchangers are more compact than the conventional round
tube plate-fin heat exchangers thanks to much higher heat transfer coefficients both on refrigerant and
air side. However, one of the challenges when using microchannel heat exchangers (MCHX) as
evaporators is distribution of two-phase refrigerant among parallel microchannel tubes, because
vapour and liquid have significantly different densities and thus inertia. Many attempts have been made
to understand this better and improve distribution in the past decades (see Hrnjak, 2004).
Improvement attempts could be divided in two major types: a) mixing distributors and b) flow
arrangements that facilitate unification of refrigerant properties. The first type is almost always
empirical, hoping that a specific design (typically patented) will provide a good distribution over the
wide range of operating conditions. The modifier “over the wide range of operating conditions” is
very important because it is possible to experimentally arrange a design that will work in a specific
state, but it is typically vastly missing the target in off-design conditions. The second type of approach
attempts to either homogenize the fluid (mix vapour and liquid well) or separate liquid from vapour to
feed liquid only to the tubes. Based on the second principle, we have started an FGB (Flash Gas
Bypass) approach for microchannel (actually any parallel flow) heat exchangers, first using CO2 (Beaver
et al. 2000, and Elbe and Hrnjak, 2004) to continue through Milosevic and Hrnjak, 2011 and Tuo and
Hrnjak 2012 and 2013 (focused on additional feature – reverse flow in the tubes). The FGB concept
was conceived at the time when microchannel evaporators had to be used in transcritical CO2 systems
to handle higher pressures than competing refrigerants in compact size. Very little was known at that
time about improving distribution in microchannel evaporators and there was a need for
improvements to offset less than perfect thermodynamic properties of CO2. In the paper Beaver et al.
2000 and even more detailed in technical report Beaver et al 1999 the results show that COP
improvement was measured to be 20% at constant capacity, while in Elbel and Hrnjak, 2004 and
detailed in Elbel and Hrnjak, 2003 the COP and capacity were simultaneously improved 8% and 7%.
The concept of flash gas removal was used in industrial refrigeration but for different purposes and
with different heat exchangers. Later analysis indicated that the flash gas improvements (reduced
pressure drop in evaporator would be even greater in low pressure refrigerants. That coincided with
acceptance of the microchannel evaporator for R134a automotive air-conditioning systems (MAC).
Both gave an impetus to the focus on low pressure fluids. To make the concept viable it was
necessary to solve the separation problem that allows integration of the separator in the header so
that activity resulted in Milosevic and Hrnjak, 2011 and Tuo and Hrnjak, 2012a.

Flash Gas Bypass

The concept of the flash gas bypass is simple and it is shown in the Fig. 10. High pressure liquid
coming from the condenser is expanded in expansion device before reaching flash gas tank, made to be
a part of the header. Liquid is separated from vapour and sent to the inlet header of the evaporator.
Vapour passes through the pressure reducing device that imposes adequate restriction to the flow to
generate the same pressure drop as evaporating two phase flow in the evaporator that will at the exit
have the desired condition (typically low superheat). This evaporator was placed in the facility as
presented in Fig. 11.

Proc. Inst. R. 2012-13. 5-8

Figure 10. The evaporator section and flash gas bypass approach.

Figure 11. Schematic drawing of test facility.

System performance comparison at the same compressor speed

The system in DX and FGB modes were first evaluated with R134a at the same compressor speed, at
compressor inlet superheat in a range of 5 °C to 20 °C. Fig. 12. shows the cooling capacity and COP
for both modes. When in FGB mode the system produced 18% to 13% more capacity and
simultaneously was more efficient (COP increased 7% to 3%) than the baseline DX, in the range
explored.

Figure 12. FGB mode outperforms DX in both cooling capacity Q and efficiency COP at
the same compressor speed.

Proc. Inst. R. 2012-13. 5-9

Fig. 13 compares DX and FGB modes for compressor inlet superheat ∆Tsup,Cp=15 °C. The suction
pressure is higher in the FGB than in the DX mode for 39.4 kPa. Consequently, mass flow rate of the
compressor is increased due to higher density of vapour resulting in higher cooling capacity. Increase
in evaporator exit pressure is attributed to two reasons: 1) the pressure drop across evaporator in
the FGB mode is 6.6 kPa lower than that in the DX mode and 2) better refrigerant distribution (see
Fig. 15.) significantly improved overall heat transfer rate by reducing the superheated area. The total
increase in evaporator exit pressure and reduction of evaporator pressure drop is shown in Fig. 13
and 14. It appears that the improved refrigerant distribution contributes more to performance
enhancement than reduced pressure drop. On the condenser side, condensing pressure is raised
because of increased capacity and despite of more efficient operation condensing temperature
increased to provide larger temperature difference between refrigerant and air side.

Figure 13. Refrigeration cycles in DX and FGB modes under the same compressor speed.

Figure 14. Raise of evaporating pressure mainly contributes to increase of suction

pressure.

Proc. Inst. R. 2012-13. 5-10

Figure 15. More uniform and reduced superheated zone (in red) in FGB mode indicates
improved refrigerant distribution.

Figure 15 shows the evaporator surface temperature obtained by infrared imaging which indicates
refrigerant distribution within microchannel tubes. In the DX mode, a large red area (high surface
temperatures – almost equal to air inlet temperature) on the right side of the evaporator presents a
superheated zone where tubes receive less liquid refrigerant than those on the left in a cold colour
(lower surface temperatures). This is due to uneven quality distribution at the inlet to tubes. Vapour
phase with lighter density has smaller inertia than liquid phase and thus it is more readily turning 90°
than liquid to branch out into vertical tubes. Due to greater inertial forces more liquid refrigerant
reaches the other end of the header decreasing quality at the inlet to the tubes in that region. In the
FGB mode, size of superheated zone is significantly reduced and in a more uniform pattern.
Consequently more evaporator area is in the two phase zone which has larger heat transfer rate.
Thus, larger capacity is obtained. However, it is worth noticing that in the FGB mode tubes on the left
side seems to still get more liquid than those on the right so distribution is not yet perfect. One
reason is that the vapour refrigerant flow has the higher velocity in the outlet header than liquid flow
in the inlet header. Therefore, associated pressure drop in the outlet header is larger. Even being
better than the case when inlet and outlet are at the same side, pressure drop difference in this case
still causes that driving pressure differential experienced by tubes on the left side is larger than tubes
close to the evaporator inlet. Assuming equal air side heat transfer conditions the consequence is
more superheat in the region of lower mass flow rate in the tubes.

Reduction of refrigerant pressure drop, another benefit of FGB method is illustrated in Fig. 16. Total
refrigerant flow rate mtot entering the compressor is higher in the FGB mode, compared to DX mode,
when compressor speed is unchanged. However, the evaporator refrigerant side pressure drop is
lower up to 30%, since only single phase liquid enters the coil. This causes actual refrigerant flow rate
through the evaporator in the FGB mode to be only mtot(1-xin) which is about 11% lower than mtot,DX
through the evaporator in the DX mode. Additionally, refrigerant velocity along the header and tubes
is lower when only liquid phase enters the evaporator. However, since the pressure drop in the
baseline DX mode is relatively low, the absolute reduction of pressure drop in the FGB mode is only
about 5 to 7 kPa. Compared to distribution improvement, performance enhancement due to this
benefit is relatively low, but should be more beneficial for a compact microchannel evaporator with
small size microchannels, in which case pressure drop is relatively high.

Proc. Inst. R. 2012-13. 5-11

Figure 16. FGB mode shows increased mass flow rate but reduced evaporator pressure
drop.

In addition to affecting evaporator performance, maldistribution also causes fluctuation of refrigerant

temperature at the evaporator exit. Liquid droplet mist and superheated vapour from different
microchannel tubes form a mixture (in thermal non-equilibrium) in the outlet header. Fig. 16b shows
the reading of thermocouple located at the evaporator exit for the case when compressor suction
superheat was 15 °C. In the DX mode, the maximum magnitude of temperature fluctuation is about
6 °C, but in the FGB mode that is much more stable with a fluctuation less than 1°C. Due to the long
suction line liquid droplets evaporates in the superheated vapour, reaching equilibrium state before
entering the compressor. It is the same phenomenon described in Hrnjak et al. (2001).

Figure 16b. Thermocouple signal at the evaporator exit in the DX mode shows more
fluctuation than in FGB mode due to presence of droplets in superheated vapour.

COP enhancement at the matched cooling capacity

To have the only one measure of performance improvement (here COP) the cooling capacity (Qe) is
maintained constant by varying the compressor speed in both DX and FGB modes. As shown in Fig.
17, capacities are maintained around 2.92 kW with a deviation less than ±1%. Improvements of COP
are higher than when compressor speed is constant. Compressor speed for the DX mode had to be
increased when superheat was higher to compensate for low evaporator performance. The
compressor speed in the FGB mode is reduced approximately 400 rpm below DX mode to provide
the same capacity. In general, lower speed can result in a higher isentropic efficiency for compressors,
but for current experimental results isentropic efficiency in the FGB mode is only about 1% to 2%
higher than in the DX mode. Thus, system improvement is insignificantly related to the improved
isentropic efficiency of the compressor. Corresponding p-h diagram for ∆Tsup,Cp=15 °C is shown in

Proc. Inst. R. 2012-13. 5-12

Fig. 18. In the FGB mode, evaporator exit pressure is increased 109 kPa primarily due to improved
refrigerant distribution, and discharge pressure is approximately 55 kPa lower because of less amount
of heat rejection in the condenser. Both reduce the compressor work and thus increase COP.

Figure 17. COP improvement of FGB over DX at the same capacity.

Figure 18. COP improvement in FGB mode at the matched capacity by reducing
compressor speed.

Fig. 19 shows refrigerant mass flow rate and evaporator pressure drop. Slightly lower total refrigerant
flow rate is required in the FGB mode to provide the same capacity, compared to DX mode.
Comparing to the fixed compressor speed, reduction of the evaporator pressure drop is much higher
when matching the same cooling capacity. First, the flow rate of refrigerant entering the evaporator is
now much lower in the FGB mode. Additionally, vapour phase density increases at a higher
evaporating temperature, and therefore vapour velocity becomes lower after the same amount of
liquid refrigerant evaporates within the microchannel tube. The pressure drop is further reduced.

Proc. Inst. R. 2012-13. 5-13

Figure 19. Reduction of refrigerant pressure drop in evaporator is much higher when
maintaining the same FGB and DX capacity.

These are good results for improving system performance, but imperfect temperature distribution in
Fig. 15 have inspired us to investigate in more detail and provide for further improvements. We made
heat exchangers with transparent headers to observe the flow with high speed cameras. These
visualizations were extremely revealing.

Flow regime at the inlet header indicated boiling fluctuations even in DX mode

Fig. 20 shows the infrared image of evaporator surface temperatures and two-phase flow regimes at
the inlet header in DX mode at 35°C/35 °C conditions. The entire header is divided into four zones
of flow visualization. The large superheat zone near the evaporator inlet indicates that microchannel
tubes on the left side receive more vapour refrigerant than those on the rear end of the header,
because vapour flows preferentially through the microchannel tubes near the inlet because of its lower
density and inertia force than liquid. However, in zones 1 and 2, the superheated region in each tube
gradually enlarges as flow proceeds downstream, reaching the maximum size around at the tube 10,
and then abruptly decreases. This is mainly because the downward gravitational force intends to
separate vapour on the top and liquid on the bottom. The vapour-liquid flow initially well-mixed at
the evaporator inlet gradually becomes stratified, with lesser liquid droplets entrained by the vapour
flow but thicker bulk liquid flow at the bottom of the header. Thus, flow stratification increases the
difficulty to feed liquid refrigerant up to vertical tubes and the corresponding superheat area. Until
most vapour refrigerant is branched out and elevating surface of the bulk liquid is close enough to the
tube inlets, almost only liquid refrigerant is fed into the microchannel tubes downstream, i.e. zones 3
and 4.

The occasional reverse vapour (and some liquid) flow which is discharged from the microchannels
against the upward bulk flow direction is observed to occur in microchannel tubes, especially those
located downstream of the inlet header (zones 3 and 4) where the liquid pool surface is as high as the
tube inlets. Reverse vapour flow is visible when a bubble enters the liquid pool at the bottom.
Reverse vapour is barely detected in the front part of the header where the liquid surface is relatively
low, probably because reverse vapour is swept away by the fast horizontal flow before being injected
into the liquid pool to generate more visible bubbles.

Proc. Inst. R. 2012-13. 5-14

 1 2 3 4
Reverse
bubble

1 3

Reverse
2 4 bubble

Figure 20. Flow regime at the inlet header in DX cycle at 35oC/35 oC condition.

(1a) Pressure drop oscillation (1b) Pressure drop amplitude spectrum

(2a) Inlet/outlet pressure oscillation (2b) Inlet/outlet pressure amplitude spectrum

Figure 21. Temporal variation of the evaporator pressure drop and inlet/outlet pressures
in DX cycle at 35°C/35 °C condition.

Proc. Inst. R. 2012-13. 5-15

Behaviour of reverse vapour flow could be indicated by oscillations of pressure, flow rate and
temperatures in microchannels. Fig. 21 shows the temporal oscillations of the evaporator inlet/outlet
pressures and pressure drop in DX operation. In order to characterize the oscillation, Fast Fourier
Transform (FFT) analysis is used to convert time-dependent data (Fig.21 (1a) and (2a)) into
corresponding frequency domain data (Fig.21 (1b) and (2b)), in which dominant frequency, if any, and
amplitude could be identified. In Fig. 21 (1a), the pressure drop oscillates considerably but in a
somewhat random fashion with a temporal average of about 26 kPa. Such fluctuation is mainly caused
by: 1) nucleation bubble dynamics and instances of vapour blockage during the confined bubble
growth; 2) the dynamic interaction of reverse vapour flow and the compressible volume provided by
the inlet header. Its random behaviour is postulated to be due to the complicated interaction among
hundreds of parallel channels through the inlet header, since this evaporator consists of 25 parallel
tubes, and in total 250 parallel microchannels. Considering that severe quality maldistribution at the
inlet of the tubes results in the different heat fluxes in each individual channel (see infrared image in
Fig. 20), the pressure drop oscillation is expected to be different among channels. Consequently, the
overall pressure drop measured between inlet and outlet of the evaporator as being the dynamic
superimposition of all pressure drops across individual channel plus appropriate parts of the headers is
unsteady.

From the amplitude spectrum in Fig. 21 (1b), the pressure drop oscillation shows two dominant
frequency bands around 0.5 Hz and 0.83 Hz, indicating two corresponding oscillation periods of about
20 s and 1.2 s. Particularly, the period of that faster oscillation mode is very similar to that of the
vapour backflow observed through the high speed images, which confirms that reverse flow results in
a pressure drop fluctuation. For pressure oscillations, the outlet pressure remains relatively constant
with a small oscillation amplitude of about ±5kPa around the average; the inlet pressure fluctuates
significantly with a peak-to-peak value up to 20 kPa. Pout FFT reveals only a low frequency band around
0.02 Hz, confirming its steady behaviour observed in time domain (Fig. 21 (2a)). It may be inferred
that the operations of the compressor and even condenser are not affected by reverse vapour flow
and boiling fluctuations in the evaporator. Pin has an extremely similar dominant frequency band
peaking at 0.82 Hz as P. Such coincidence demonstrates that P oscillation originates from Pin and
the latter is the consequence of periodic reverse flow.

Flow regime at the inlet header and boiling instability in FGB cycle

The FGB cycle feeds the evaporator with liquid only refrigerant, as flashing vapour generated during
the expansion process is separated and bypassed to the suction line. Fig. 22 shows a typical flow
regime at the inlet header and evaporator surface temperature profile in this case. The superheated
area is reduced and in a more uniform pattern compared to DX cycle. Microchannel tubes at the rear
end of the header appear to receive more liquid refrigerant than those close to the inlet, one of the
reasons is mass flow maldistribution due to the pressure drop of the outlet header (Tuo and Hrnjak,
2012).

Although receiving only liquid refrigerant from upstream the flash gas separator, the entire inlet
header is not filled up with liquid. A clear vapour-liquid interface distinguishes the liquid pool at the
bottom and a vapour pocket at the top, oscillating up and down around the tube inlets.

Proc. Inst. R. 2012-13. 5-16

 1 2 3 4

1 3
Liquid surface

2 4
Figure 22. Flow regime at the inlet header in FGB cycle at 35°C/35 °C condition.

Fig. 23 depicts a sequence of high-speed images for periodic reverse flow at one portion of the inlet
header. During the first 0.2 s no reverse bubble is observed, indicating that liquid refrigerant may
enter the tubes in a forward flow direction. Since t+0.4 s, bubbles are gradually discharged out of the
tubes, rising up in the liquid pool driven by the buoyancy force, and finally breaking up and merging
into the vapour pocket on the top. The vapour reversal continues to about t+1.4 s. Meanwhile, the
level of vapour-liquid interface gradually drops since coalescence of reverse bubbles expands the
volume of the vapour pocket and thus expels out liquid pool retained at the header bottom.
Particularly, during the period of t+1.0 s to t+1.4 s, liquid level in the header is lower than the tube
inlets. Afterwards, liquid surface gradually goes up. It can be inferred that vapour backflow at the tube
inlets ceases and forward upward flow is re-established, with both vapour and liquid refrigerant
temporally retained in the header. The behaviour of reverse vapour flow and induced vapour-liquid
interface oscillation repeats periodically, and the interval was estimated to be about 1.8 s based on the
flow visualization.

Figure 23. Periodic oscillation of vapour-liquid interface and reverse vapour flow (zone 3).

Proc. Inst. R. 2012-13. 5-17

Fig. 24 illustrates temporal variations of evaporator inlet/outlet pressures and overall pressure drop.
Compared to DX cycle, the pressure drop and inlet pressure in FGB cycle fluctuate periodically in a
clearer pattern and with greater oscillation amplitude, since liquid is distributed more evenly among
parallel microchannel tubes. This results in more uniform heat flux and refrigerant mass flux to
synchronize reverse flow oscillation in each tube.

The pressure drop is about 18.65 kPa on average but temporally varies between the instant maximum
of about 39 kPa and the minimum of almost 0 kPa, resulting in the peak-to-peak amplitude of roughly
39 kPa. In frequency domain, P spectrum only has one dominant frequency band cantered at 0.53 Hz.
This peak value corresponds to a period of 1.89 s, which is very similar to that of periodic reverse
flow observed in Fig. 23. Similar to DX cycle, Pin FFT shows exactly the same dominant frequency
band and peak value, but Pout remains almost constant. These results confirm that oscillations of inlet
pressure and therefore pressure drop are caused by periodic reverse vapour flow in microchannels
and dynamic interaction with the upstream compressible volume.

Under the same air inlet temperature and velocity (similar heat flux) the pressure drop oscillation in
FGB cycle differs from that in DX cycle (compare Fig. 24(1b) to Fig. 21(1b)). In FGB case the pressure
drop fluctuates at a lower frequency but with higher oscillation amplitude. This may be because the
flash gas separator in FGB cycle serves as a damper upstream the inlet header and heated
microchannels.

(1a) pressure drop fluctuation (1b) pressure drop amplitude spectrum

(b1) inlet and outlet pressure oscillations (b2) inlet and outlet pressure amplitude spectrum

Figure 24. Variation of evaporator inlet/outlet pressures and pressure drops in FGB cycle
under 35ºC/35 ºC condition.

Proc. Inst. R. 2012-13. 5-18

To further identify the relationship between reverse flow and oscillation of Pin, Fig. 25(1) depicts one
typical oscillation period of inlet and outlet pressures, and Fig. 25(2) shows a consecutive sequence of
synchronized high speed images corresponding to respective data points highlighted on Fig. 25(1). At t
= 5.224 s, the first bubble is vented from the inlet of the second tube on the right, and until t = 5.364
s, vapour reverse is observed at all the tubes and continues until t5. During this period, inlet pressure
gradually increases because of continuous accumulation of reverse vapour within the inlet header. It
reaches the peak value at t4, and then decreases. This drop may be because continuously evaporating
liquid that reside in the microchannel tubes will gradually deplete the residual liquid and increase the
superheated region, without receiving any fresh liquid from the tube inlet. It will slow down bubble
growth and lower the over pressure which drives the reverse flow. Therefore, the inlet pressure
decreases correspondingly. Clearly, vapour reversal appears to be less intensive and even ceased at t5.
When the inlet pressure is sufficient to overcome the over pressure together with the pressure drop
along the tube, forward upward flow is re-instated. However, because the vapour pocket is present
above the liquid and the liquid surface is lower than the tube inlets, most likely only vapour refrigerant
is received by tubes at the beginning. Approximately, at t6, when rising liquid surface reaches the tube
inlets, fresh liquid starts to feed the microchannels. From the above observation and analysis, one
oscillation cycle would be divided into three consecutive steps, as sketched in Fig. 26: 1) t1~t5 –reverse
vapour flow; 2) t5~t6 – re-entraining vapour in a forward flow; 3) t6~t7 – fresh liquid feeding
microchannels in a forward flow. In the third step liquid feeding the tubes only occurs approximately
in the period of t6~t7, about 0.27 s or 16.2% of the entire cycle. It can be inferred that the instant
liquid mass flow entering the microchannels would be about more than 5 times of the average value.

(1) Inlet pressure variation;

(2) Behaviour of reverse bubbles

Figure 25. Evolution of the inlet pressure and synchronized flow regime at inlet header.

Proc. Inst. R. 2012-13. 5-19

Figure 26. Schematic illustration of temporal evolution of periodic reverse flow.

Effect of heat flux on periodic reverse vapour flow and boiling fluctuations in FGB mode

(1) 27 °C/27°C (2) 35 °C/35°C

(3) 43 °C/43°C
Figure 27. Pressure drop amplitude spectrum under different air inlet conditions.

Fig. 27 (1-3) shows the amplitude spectrums of the evaporator pressure drop at three air
temperatures in FGB cycle. As ambient temperature increases, the dominant frequency bands are
shifted up, indicating shorter oscillation periods or more intensive periodic reverse vapour flow. This
is because at the fixed compressor speed, higher air inlet temperatures will increase the evaporator
cooling capacity and the refrigerant flow rate due to an increased refrigerant-air temperature
difference. Average heat flux qvag and refrigerant mass flux Gavg at each channel will both become

Proc. Inst. R. 2012-13. 5-20

higher for the evaporator with fixed heat transfer area and controlled constant exit superheat, defined
as:
Qe
qavg  (1)
Aref , s
m ref (1  xin )
Gavg  (2)
Aref ,c
Where Qe, m  ref , xin are cooling capacity, total refrigerant mass flow rate and quality right after the
expansion valve. Aref,s and Aref,c are heat transfer surface area and microchannel cross sectional area on
the refrigerant-side. Since flash vapour is bypassed, only actual refrigerant flow rate is that for only
 (1  x ) .
liquid refrigerant, i.e. m ref in

Higher average heat flux at a higher air inlet temperature intensifies bubble nucleation and the
following longitudinal expansion both upstream and downstream along the microchannels. On the
other hand, proportionally increased liquid mass flux results in liquid filling the channels at a faster
pace. Both effects intend to speed up the oscillation of the reverse flow. Fig. 28 illustrates that as
average heat flux increases from 8.37 kW/m2 to 10.8 kW/m2 the oscillation frequency rises up from
0.45 Hz to 0.58 Hz. It should be noted that the oscillation frequency just varies slightly because of
relatively small change of average heat flux.

Figure 28. Variation of the dominant oscillation frequency as a function of average wall
heat flux.

Potential impacts of reverse vapour flow on evaporator performances

Periodic reverse flow and triggered boiling fluctuations will cause maldistribution of refrigerant flow
among parallel microchannel tubes. For microchannel evaporators in DX mode, refrigerant
maldistribution results from a combined effect of quality maldistribution due to different vapour and
liquid densities and inertia forces and due to periodic reverse flow. It is very difficult to distinguish
their individual impacts on evaporator performance. The effects of periodic reverse flow are expected
to be less significant than the quality maldistribution for explored evaporator, since as shown in Fig. 3
the large superheated area on the left half of the evaporator is the consequence of vapour refrigerant
predominantly branching out through tubes near to the evaporator inlet.

The impact of reverse flow is a strong disturbing factor in FGB feeding mode which supposedly
eliminates quality maldistribution by feeding liquid only to the evaporator. Even if the trapped vapour
would be initially removed, the reversed vapour that is periodically discharged from the microchannels
definitely creates vapour that is trapped in the top of the header. That creates a situation with
oscillating vapour-liquid interface which obstructs feeding liquid only into each microchannel tube in
spite of the fact that the upstream flash gas separator supplies only liquid to the inlet header. As

Proc. Inst. R. 2012-13. 5-21

shown in Fig. 29(1), a liquid wave propagates from the inlet toward the header end, creating an abrupt
rise in the liquid surface, similar to a hydraulic jump. At a certain moment, as shown in Fig. 29(1) for
example of t0 + 0.10 s, inlets of the first 3 microchannel tubes on the left are submerged in the wave
crest, having the opportunity to receive liquid refrigerant while other tubes are still exposed to the
vapour pocket, since liquid surface ahead the abrupt rise is lower than the tube inlets. These tubes
could only take the vapour refrigerant. This causes mild maldistribution of liquid refrigerant flow
among parallel microchannel tubes. A similar consequence would be expected when the unstable
wave travels backward after hitting the header end as shown in Fig. 29(2).

(a) Forward wave propagation (b) Backward wave propagation

Figure 29. Illustration of possible liquid maldistribution due to oscillating liquid interface
in FGB mode.

In addition, reverse vapour flow may reduce local refrigerant-side heat transfer coefficient. As
discussed in the previous section, backflow vapour trapped in the top header will be re-entrained with
a forward flow as long as the liquid surface does not reach or is not close enough to the tube inlets.
Very likely, the vapour entrainment will form a “dryout” bubble slug without being surrounded by a
thin liquid film between the channel wall and vapour core, as illustrated in Fig. 26. Thus, the local heat
transfer coefficient in this “dryout” vapour slug will be much lower than when tubes only receive
liquid. Finally, the evaporator pressure drop will be increased because trapped vapour will be
recirculated into the evaporator with reinstated forward flow.

Flash Gas Bypass Reversed (FGBR) - a solution to mitigate the impacts of periodic reverse flow

One possible and simple solution is to vent and bypass the reverse vapour trapped in the header.
Particularly, for microchannel evaporators with horizontal headers and vertical upward tubes, this can
be realized by adding any venting mechanism. In one experiment we have added two venting ports, at
the ends of the inlet header and bypassed the reverse vapour through additional venting lines. A

Proc. Inst. R. 2012-13. 5-22

simple layout of revised flash gas bypass method with venting reverse vapour flow is demonstrated in
Fig. 30, as used in initial experiments. Consequences and benefits are: 1) vapour pocket occupying the
top header is removed such that liquid surface is elevated and stabilized; 2) all tube inlets were
submerged in the liquid, therefore taking only liquid refrigerant more uniformly; 3) local heat transfer
coefficient increased by eliminating entrained “dryout” bubble slugs; 4) the evaporator pressure drop
was further reduced because venting the reverse vapour decreases the actual refrigerant flow rate
through the evaporator.

These experiments revealed significant reversed mass flow rate: 5-8% of the total. .

Figure 30. Illustration of venting reversed vapour flow and potential benefits.

Conclusions

The paper presented two novel approaches in refrigeration: microchannel condenser for low charged
ammonia chillers and flash gas bypass system. The first is based on 15 years of application in real
systems while the second is still waiting for full application. Nevertheless, findings in FGB work are
revealing an unexpected reversed flow of refrigerant in the evaporator, mostly vapour. Obviously,
refrigeration is still an open area for innovative approaches - even the path from idea to realization can
take longer than a decade.

Acknowledgment

The author is grateful to the members of ACRC, the industry university cooperative research center
for both financial and technical assistance, as well as Mayekawa, Bitzer, Modine, IIAR, Hydro, Alfa Laval
for their multifaceted support. Without help from CTS, its staff as well as at the time graduate
students Hanfei Tuo, Alen Milosevic, Stefan Elbel, Andy Litch, and Andy Beaver this paper would not
be possible.

References

 Adams, D., P. Hrnjak, and T. Newell, 2003, Pressure Drop and Void Fraction in Microchannels Using Carbon Dioxide,
Ammonia, and R245FA as Refrigerants, ACRC Report TR221
 Butterworth, D. 1975. A Comparison of Some Void-Fraction Relationships for Co-Current Gas-Liquid Flow”
International Journal of Multiphase Flow, 1: 845-850.
 Litch A., P. Hrnjak 1999. "Condensation of Ammonia in Microchannel Heat Exchangers" ACRC Report CR22
 Hrnjak, P.S. and A. D. Litch, “Microchannel Heat Exchangers for Charge Minimization in Air-Cooled Ammonia
Condensers and Chillers,” International Journal of Refrigeration, 31:4, 658-668, 2008

Proc. Inst. R. 2012-13. 5-23

 Newell, T. et al. 1999. An Investigation of Void Fraction in the Stratified/Annular Flow Regions in Smooth, Horizontal
Tubes International Journal of Multiphase Flow.
 Pearson, A. 2003, Low charge ammonia plants: Why bother? Technical paper#5, pp 153-176, IIAR Ammonia
Refrigeration Conference, Albuquerque, NM March 16-19
 Zivi, S.M. 1964. Estimation of Steady-State Steam Void-Fraction by Means of the Principle of Minimum Entropy
Production ASME Journal of Heat Transfer 86: 247-252.
 Litch, A., P. Hrnjak, 2000, Low Charge Air Cooled Ammonia Chiller with Aluminium Microchannel Condenser,
Proceedings of the IIR Gustav Lorenzen Conference on Natural Refrigerants, Purdue, W. Lafayette, IN, 543-551,
 Hrnjak, P. A. Litch, 2001. Charge Reduction in Ammonia Chiller Using Air-Cooled Condensers with Aluminium
Microchannel Tubes, Proceedings of IIAR Refrigeration Conference Long Beach, CA, 235-267
 Hrnjak, P. 2002, “Microchannel Heat Exchangers as a Design Option for Charge Minimization on NH3 and HC
Systems,” Proceedings of IIR Conference Zero Leakage - Minimal Charge, Stockholm, 111-118,
 Hrnjak, P., 2005. Charge Minimization in Ammonia Refrigeration Systems, Proceedings of the IIR Conference on
Ammonia Refrigerating Systems, Renewal and Improvement, Ohrid, Macedonia
 Palm, B. 2009. Summarizing A Decade of Experience on Charge Reduction for Small Hydrocarbon, Ammonia and HFC
Systems, 1st IIR Workshop on Refrigerant Charge Reduction, Paris
 Hwang, Y., Jin, D.H., Radermacher, R., 2007. Refrigerant distribution in minichannel evaporator manifolds.
HVAC&R Research. 13(4), 543-555.
 Fei, P., Cantrak, D., Hrnjak, P.S., 2002. Refrigerant distribution in the inlet header of plate evaporators. 2002
SAE World Congress. Paper # 2002-01-0948.
 Beaver, A.C., Yin, J.M., Bullard, C.W., Hrnjak, P.S., 1999. An experimental investigation of transcritical carbon
dioxide systems for residential air conditioning. Contract Report CR 18, Air Conditioning and Refrigeration Center,
University of Illinois, Urbana-Champaign, IL.
 Beaver, A., P. S. Hrnjak, J. Yin, and C. W. Bullard, “Effects of Distribution in Headers of Micro-Channel
Evaporators on Transcritical CO2 Heat Pump Performance,” Proceedings of the ASME Advanced Energy
Systems Division, Orlando, FL, AES-Vol. 40, 55-64, 2000.
 Tuo, H.F., Hrnjak, P.S., 2012a. Flash gas bypass in mobile air conditioning system with R134a. International
Journal of Refrigeration. 35(7), 1869-1877.
 Brutin, D., Topin, F., and Tadrist, L., 2003. Experimental study of unsteady convective boiling in heated
minichannels. International Journal of Heat and Mass Transfer. 46(16), 2957-2965.
 Bowers, C.D, et al., 2012. Refrigerant distribution effects on the performance of microchannel evaporators.
International Refrigeration and Air Conditioning Conference at Purdue. No.2173.
 Tuo, H.F., Bielskus, A., and Hrnjak, P., 2012b. Experimentally validated model of refrigerant distribution in a
parallel microchannel evaporator. SAE International Journal of Materials and Manufacturing. 5(2), 365-374.
 Bowers, C.D., Wujek, S.S., and Hrnjak, P., 2010. Quantification of refrigerant distribution and effectiveness in
microchannel heat exchangers using infrared thermography. International Refrigeration and Air Conditioning
Conference at Purdue. No.2117.
 Elbel, S. and P. S. Hrnjak “Flash Gas Bypass for Improving the Performance of Transcritical R744 Systems that
Use Microchannel Evaporators,” International Journal of Refrigeration, 27:7, 724-735, 2004.
 Elbel, S. W. and P. S. Hrnjak, “Experimental and Analytical Validation of New Approaches to Improve
Transcritical CO2 Environmental Control Units,” ACRC Report No. CR-52, Dec. 2003.
 Hrnjak, P. S., “Developing Adiabatic Two-phase Flow in Headers – Distribution Issue in Parallel Flow
Microchannel Heat Exchangers,” Heat Transfer Engineering, 25:3, 61-68, 2004.
 Milosevic, A. S. and P. S. Hrnjak, Flash Gas Bypass Concept Utilizing Low Pressure Refrigerants,” ACRC
Report No. TR-283, July 2011
 Tuo, H.F., Hrnjak, P.S., 2012. Flash gas bypass in mobile air conditioning system with R134a. International Journal
of Refrigeration. 35(7), 1869-1877.
 Tuo, H.F., Hrnjak, P.S., 2013. Periodically reversing flow and boiling fluctuations in a microchannel evaporator.
International Journal of Refrigeration. In print.
 Tuo, H. and P. Hrnjak, 2012a, "Experimental Study of Refrigerant Two Phase Separation in a Compact Vertical
T-Junction," ASHRAE Transactions, 118:1.

Proc. Inst. R. 2012-13. 5-24