Applied Thermal Engineering 183 (2021) 116114

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

An experimentally validated model for microchannel condensers with

separation circuitry
Jun Li a, Pega Hrnjak a, b, *
a
ACRC Air Conditioning and Refrigeration Center, Department of Mechanical Science and Engineering, The University of Illinois at Urbana-Champaign, 1206 W Green
St., Urbana, IL 61801, USA
b
CTS Creative Thermal Solutions, 2209 N Willow Rd, Urbana, IL 61802, USA

A R T I C L E I N F O A B S T R A C T

Keywords: The experimental study on a microchannel condenser with the separation circuitry (or separation condenser) was
Microchannel condenser conducted in the past by the same authors. The separation condenser has an inlet in the middle of the height, and
Separation vapor separates from the liquid in the second header after condensation in the 1st pass. This paper presents a
Two-phase flow
condenser model that incorporates a mechanistic model developed to predict the phase separation efficiency in
Second header
the second header. The 1-D finite-volume method is used. The condenser model is validated by condenser testing
results using R134a. The mass flux through the first pass is in the range of 145–330 kg m− 2 s− 1. The difference
between measurement and modeling result is ±5% for capacity and ±15% for pressure drop. The model reveals
that other than the in-header fluid dynamics for the two-phase flow, the downstream flow resistance (a function
of the air velocity) also influences the separation results in the second header, based on the equal pressure drops
in the upper path and lower path. The model could be used for the design optimization for separation condensers,
which will be presented in another study.

transfer area being 37% less than the baseline, the condenser that ex­
1. Introduction tracts liquid performed as good as the baseline condenser. Luo et al. [4]
apply the extraction design in round-tube-plate-fin condensers for
For a condensation process, liquid on the wall is detrimental because organic Rankine cycles (ORCs) and build a mathematic model for the
it decreases the heat transfer coefficient and increases the frictional condenser. Optimization based on the model shows cost reduction up to
factor. Removing condensate is one of the ways to improve condenser 34.5% and exergy efficiency improvement up to 13.7%. Li et al. [5] and
performance. Liu et al. [1] propose the concept of the condenser with Li et al. [6] studied the effect of liquid extraction in shell-and-tube
separation circuity (or separation condenser) and recommend unequal condensers used in ORCs. The design is found to increase the average
numbers of flow passes on the upper side and the lower side in their heat transfer coefficient by 23.8% for the R600/R601a mixture and
design. A bypass line is used to transfer the liquid phase to a specific reduce the heat transfer area by 44.1%. Zhang et al. [7] studied
location in the condenser. Further, the authors propose installing a extraction condenser using a U-shape tube as the low-cost separator in
separator in the header to enhance the phase separation. Compared with the header. The overall heat transfer coefficient of the condenser using
the separation condenser, another configuration, the extraction R134a is found to improve by 10% to 20%. Davies et al. [8] studied the
condenser, has been more widely studied. Wu et al. [2] present a study effect of inclination on heat transfer of water flow in large-tube con­
on a serpentine condenser that extracts liquid in header based on several densers for power plants. The slightly improved heat transfer coefficient
of a published patent by the same group (Peng et al., [3]). The condenser is attributed to the improved water drainage and increased void fraction
is used for a residential air conditioning (AC) system with a refrigerating near the condenser outlet.
capacity of 2300 W. In the vertical headers, liquid–vapor separators are Nowadays, air-cooled microchannel condensers for automobile ap­
included to ensure that pure vapor enters the next flow passage. Both plications adopt designs with multiple passes (one pass is a plural of
capillary length and refrigerant charge were optimized to obtain the parallel tubes brazed as one part through fins). Li and Hrnjak [9,10]
highest COP and refrigerant capacity for condensers. With the heat prove both experimentally and by modeling that a microchannel

* Corresponding author at: ACRC Air Conditioning and Refrigeration Center, Department of Mechanical Science and Engineering, The University of Illinois at
Urbana-Champaign, 1206 W Green St., Urbana, IL 61801, USA.
E-mail address: pega@illinois.edu (P. Hrnjak).

https://doi.org/10.1016/j.applthermaleng.2020.116114
Received 17 June 2020; Received in revised form 19 August 2020; Accepted 25 September 2020
Available online 12 October 2020
1359-4311/© 2020 Elsevier Ltd. All rights reserved.
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Nomenclature c condenser
crit critical
C heat capacity rate, W K− 1 D downward exit plane
D diameter, mm dec deceleration
ΔP pressure drop, kPa down downward
Ef entrainment fraction drop droplet
F force, N eq equivalent
G mass flux, kg m− 2 s− 1 fri frictional
h specific enthalpy, kJ kg− 1 grav gravitational
2 1
HTC heat transfer coefficient, W m− K− h header
loc location i tube number i
ṁ mass flow rate, g s− 1 ihd inlet header
MAC mobile air conditioning IN inlet
MC microchannel L liquid phase
N number max maximum
P pressure, kPa min minimum
Q capacity, kW odn outdoor nozzle
RH relative humidity, – ohd outlet header
RHS right-hand side, – r refrigerant
T temperature, ◦ C ri refrigerant inlet
x vapor quality, – ro refrigerant outlet
t microchannel tube
Subscripts U upward exit plane
1st 1st pass Uai upper-pass air inlet
2L 2nd-lower pass up upward
2U 2nd-upper pass V vapor phase
3L 3rd-lower pass VS vapor superficial
3U 3rd-upper pass
a air Greeks
ai air inlet η separation efficiency, –
ao air outlet ζ split ratio, –
buoy buoyancy ρ density, kg⋅m− 3

condenser with the separation circuitry increases the condensate flow the US.
rate by up to 8.9% for the same inlet and outlet specific enthalpy. A However, for a separation condenser, as shown in Fig. 1, complete
schematic of the pass circuitry and a photo of the real condenser pro­ separation of liquid from vapor in the second header does not usually
totype are shown in Fig. 1. Different from a conventional design of exist for every operating condition. In the limited number of modeling
condensers, the inlet of this condenser prototype is in the middle of the studies of the separation condensers (Won [11]), the predictive model
height. In the second header that is downstream of the 1st pass, the for phase separation inside the header has not been found.
vapor phase is expected to move upward and separate from the denser To first quantify the phase separation, Li and Hrnjak [12,13] defined
liquid phase due to the effect of gravity. Since this design extends the the phase separation efficiencies in the second header in their experi­
relative area of upper passes to lower passes and flow in the upper passes mental study. Following the nomenclature in Fig. 2, Eq. (1) defines the
has a higher heat transfer coefficient due to less liquid film on the wall, liquid separation efficiency ηL as the ratio of liquid mass flow rate going
the overall effectiveness of the condenser can be increased. into the 2nd-lower pass (through the downward exit plane) ṁLD to the
This separation design provides performance improvement without total liquid mass flow rate supplied to the inlet of the second header. Eq.
too much additional cost (only flow pass rearrangement at the design (2) defines the vapor separation efficiency ηV as the ratio of the vapor
phase and a few more baffles for the headers). Based on an estimate by mass flow rate going into the 2nd-upper pass (through the upward exit
the authors, if a 10% improvement in energy efficiency were to be plane) ṁVU to the total vapor mass flow rate supplied to the inlet of the
achieved, that would be translated to a saving of $1.1 billion (2 billion second header.
liters or 550 million gallons) of gasoline annually only for the vehicles in

Fig. 1. Microchannel condenser with the separation circuitry.

2
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Fig. 2. Variables related to the separation efficiencies ηL and ηV.

ṁLD
ηL = (1) ṁVU
ṁLD + ṁLU xU = (3)
ṁVU + ṁLU
ṁVU
ηV = (2) ṁVD
ṁVU + ṁVD xD = (4)
ṁVD + ṁLD
where ṁLU is the liquid mass flow rate at the upward exit, and ṁVD is the Different values of xU represent different separation results in the
vapor mass flow rate at the downward exit. header. When xU > xIN, phase separation happens.
The qualities at the two exits, xU, and xD, are to quantify the flow This paper first briefly demonstrates a preexisting 1-D mechanistic
condition at the upward exit and the downward exit, respectively. xU model ([14]) by the same authors for calculating ηL, ηV, xD, and xU in the
and xD are calculated as Eqs. (3) and (4): second header. Then, this header model is incorporated into the model

Fig. 3. Boundary conditions and three critical locations along the longitudinal direction of the second header: locV,crit, locp0, and locL,crit.

3
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

for the separation condenser to calculate the separation efficiency. The uVS changes the sign, i.e., the vapor critical location locV,crit, as shown in
model is validated by condenser results using R134a. To the authors’ Fig. 3. The criterion for calculating locV,crit can be expressed as
best knowledge, this novel and complete model for separation con­
uVS,locV,crit < 0 < uVS,locV,crit (5)
densers is the first to appear in the literature. This model clearly explains +1

the flow mechanisms of the separation condenser in that the internal The liquid moving upward is considered in two forms: droplet and
two-phase flow dynamics in the second header and the downstream flow film. The first form is the droplet form. Based on how many droplets
resistance (a function of the air velocity) work together to determine the from each inlet tube go upward, another critical location in Fig. 3, the
separation result. This model can be employed in the future to study liquid critical location locL,crit, can be determined by Eq. (6):
various design parameters of the separation condenser with the goal of
optimization. uLS,locL,crit < 0 < uLS,locL,crit +1 (6)

where above locL,crit, the superficial velocity of liquid is positive; below

2. Model description
locL,crit, the superficial velocity of liquid is negative. locL,crit must always
be higher than locV,crit, as shown by Fig. 3, because the positive drag
2.1. Second header model
force for the liquid needs to be provided by upward-moving vapor.
Therefore, above locL,crit, there is upward co-current flow; between locL,
Li and Hrnjak [14] developed a mechanistic model for flow in the
crit and locV,crit, there is counter-current flow with liquid moving down
second header to predict separation efficiency. More details about the
and vapor moving up; below locV,crit, there is downward co-current flow.
second header model can be found in that paper. This paper will focus on
The second form of liquid going upward is the film form. The cor­
the incorporation of that header model into the separation condenser
responding critical location for the liquid film is locfilm,crit. locfilm,crit must
model.
be higher than the segment in which liquid goes up as droplets since it
The computational domain of the second header model is the inlet
requires higher vapor velocity to entrain liquid films than to entrain
region (inlet flow direction is out of the microchannels to the right)
droplets for the same amount of liquid mass flow. Above locfilm,crit all
encompassed by the green dash rectangle in Fig. 3. The model calculates
liquid coming out of the inlet tube(s) is assumed to go upward. In Fig. 3,
inlet flows to the upper region and the lower region of the second
there is another location locp0, which denotes the segment where the net
header. Inlet tube number Nt is shown to be 21 (the same as that of the
mass flux of refrigerant changes sign. After locV,crit, and locL,crit are set,
condenser in Fig. 1) as an example but can vary. The flow is assumed
locp0 will be determined.
steady, adiabatic, and incompressible. One segment is defined as the
The model for droplet entrainment is based on the analogy to the
space in the header between every two adjacent microchannel tubes.
free-fall terminal velocity of a drop made in Turner et al. [15]. Three
Vapor flow in each segment is assumed to have uniform velocity. Up­
forces in the vertical direction are considered for a droplet: the gravi­
ward is the positive direction for velocities and forces.
tational force Fgrav, the buoyancy force Fbuoy, and the drag force Fdrag
As shown in Fig. 3, mass flow and pressure boundary conditions are
exerted by the vapor. The drag force coefficient for Fdrag is chosen as the
applied at the inlets to the computational domain, vapor mass flow
sphere case in GPSA Engineering Databook [16]. The diameter is pre­
boundary condition is applied at each exit of the computational domain,
dicted by the classic liquid flow atomization model by Beale and Reitz
and heat flux (=0) boundary condition is applied at the wall. Therefore,
[17]. This is a selection based on the comparison to several other flow
when incorporated into the condenser model, the inputs of the second
atomization models ([18–20]). The Beale and Reitz model most accu­
header model are ṁIN, xIN, pIN, ṁVU. Outputs of the model are ṁVD, ṁLU,
rately predicts the separation efficiency from the experiments in [14]. A
ṁLD, pU, pD. Based on ṁVU, the vapor split ratio ζV (vapor going up from
Rosin-Rammler distribution is used to describe the droplet size distri­
one tube divided by the vapor supplied to that tube) shown in Fig. 4 is
bution. The droplet entrainment criterion is the sum of Fdrag and Fbuoy
calculated for all the inlet tubes. Then, the local vapor superficial ve­
exceeds Fgrav. Since vapor velocity grows going upwards, once a droplet
locity uVS is calculated for each segment. For example, a few segments
is determined to be entrained at the initial segment, the net force it is
are shown in Fig. 4. The local uVS will be used to determine local liquid
going to experience will always be positive or equal to 0. Therefore, it
entrainment in each segment, as will be shown below. Summing up all
will keep moving upward at either accelerating or constant speed.
the liquid going up from each tube, ṁLU can be calculated, and then ṁLD
For the inlet tubes above locV,crit below locfilm,crit, depending on the
can be calculated based on mass conservation. There is a segment where

Fig. 4. Initialization of the local superficial velocity of vapor flow.

4
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

droplet diameter, the liquid split ratio ζL is used to describe the part of
the liquid going up for those tubes. ζL for one inlet tube is defined as the
portion of the liquid going up in the total liquid coming out of that tube.
ζL for tube i is determined by the largest diameter of the droplets that the
vapor in segment i can entrain.
The model for film entrainment adopts the classic criterion of flow
reversal point. It calculates the minimum vapor velocity at which the
upward flowing annular film starts. A simple expression for this criterion
by Wallis [21] is:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ρV
u*VS = uVS ≈1 (7)
gD(ρL − ρV )

where uVS* is the dimensionless vapor velocity, uVS is the superficial

vapor velocity, g is the acceleration due to gravity, and D is the diameter
of the tube.
Above locfilm,crit, all liquid mass flow rate from each inlet tube goes
upward. Below locfilm,crit and above locV,crit, a portion (ζL) of the liquid
mass flow rate from each inlet tube goes upward. Then, the mathe­
matically equivalent locL,crit in terms of total liquid entrainment can be Fig. 6. Modeling of one single pass in the condenser.
solved by Eq. (8)
loc∑ ( ) and pressure boundary conditions are applied at the inlet, the mass flow
( ) ( ) film,crit

Nt − locL,crit ṁL,t = Nt − locfilm,crit ṁL,t + ζL ṁL,t (8) boundary condition is applied at the outlet, the convection boundary
locV,crit i condition is applied at the microchannel tube wall, and heat flux (=0)
boundary condition is applied at the header wall. In the pass model, each
Summing the liquid from each inlet tube up, ṁLU will be calculated as
microchannel tube is divided into small segments for finite volume
an output. Then, ṁLD will be calculated. From ṁVU and ṁVD, xU and xD
calculation. Each segment has its own set of governing equations for
can also be calculated.
heat transfer and pressure drop. The governing equations are linked
Adding liquid mass flux onto the vapor flux, the mass flux, quality,
through the designation of segment inlet and outlet. Each segment has
and void fraction in each segment can be calculated. Starting from locL,
distinct conditions at the refrigerant inlet and the refrigerant outlet. The
crit going upward, and from locV,crit going downward, the pressure drop
outlet condition of a segment is equal to the inlet condition of the
in each segment in those two directions can be calculated. The pressure
segment downstream. This way, the simulated refrigerant enthalpy and
drop considered includes the accelerational Δp due to the change of the
associated properties will steadily change across the chain of segments.
rate of momentum, the frictional Δp, and the gravitational Δp. Void
While the air face velocity and inlet temperature are the same for each
faction correlation by Zivi [22] is used for calculation of mean density ρ.
segment due to the cross-flow configuration, each segment has its air
The friction Δp is calculated by the correlation of Friedel [23]. The
outlet temperature. In a common simulation case, inputs to one micro­
pressure is deducted to pU at the upward exit plane (exit of segment 21 in
channel tube are the inlet parameters of the very first segment as well as
Fig. 3) and to pD at the downward exit plane (exit of segment 0 in Fig. 3),
parameters that are equal for all segments such as refrigerant mass flux,
respectively.
air velocity, and air inlet temperature, pressure, and relative humidity.
The following assumptions are made to the pass model: (1) both the
2.2. Pass model refrigerant flow and the airflow are at steady-state; (2) uniform liquid
refrigerant flow distribution among the microchannel tubes in one pass
The second header model is combined with the 1-D finite-volume unless otherwise specified; (3) at each microchannel in the same tube,
model for single flow passes from Li and Hrnjak [9] to simulate the the refrigerant mass flow rate is the same; (4) no heat is conducted along
whole separation condenser. Fig. 5 illustrates the model used for each the tube or between the tubes through fins; (5) all headers are adiabatic.
part of the condenser. The green shadowed rectangle denotes the loca­ To model the heat transfer between the air and the refrigerant, the
tion of the second header. Except for the 1st pass, each pass consists of Effectiveness–Number of Transfer Unit (ε-NTU) method is used. As
one inlet header, parallel microchannel tubes, and one outlet header. outlined in Incropera et al. [24], the maximum possible heat transfer in
One pass is one module in the code, marked by one dashed rectangle in one segment is defined as
Fig. 5.
For the refrigerant flow in a single pass, as shown in Fig. 6, mass flow Qmax = Ca (Tri − Tai ) (9)

where Ca is the heat capacity rate of air. The actual heat transfer rate in
one segment is given as
Q = ε Ca (Tri − Tai ) (10)

where ε is the effectiveness of the element, the effectiveness is a function

of the NTU and heat capacity rate ratio Cratio, as defined in Eq. (11).
When a two-phase flow is present, Cratio is zero.
min(Cr , Ca ) Cmin
Cratio = = (11)
max(Cr , Ca ) Cmax

where Cr is the heat capacity rate of refrigerant. Both single-phase flow

and two-phase flow will exist on the refrigerant side in the condenser.
Fig. 5. Second header model incorporated with previous pass model to simu­ Eq. (12) is used for single-phase flow and Eq. (13) for two-phase flow.
late the separation condenser.

5
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

[ ( [ ] )]
ε = 1 − exp Cratio ⋅NTU 0.22 ⋅ exp − Cratio ⋅NTU 0.78 − 1 (12)

ε = 1 − exp( − NTU) (13)

The total thermal conductance, UA, of one segment to calculate NTU
(UA/Cmin) is expressed as:
1 1 δwall 1
= + + (14)
UA ηHTCa Aa kwall Ar HTCr Ar

where HTCa and HTCr are airside and refrigerant-side heat transfer co­
efficients, which are determined by different correlations. The flow
condition used to determine HTCa and HTCr for one segment is chosen to
be the inlet condition of that segment. The heat transfer and pressure
drop correlations used for calculating HTCa and HTCr are listed in
Table 1.
The original single-pass model in Li and Hrnjak [9] has been
improved in the way that it considers the pressure drop in the header
segment by segment (varying with mass flux and quality along the
longitudinal direction of the header), which is more physical. Frictional
pressure drop ΔPfri and gravitational pressure drop ΔPgrav are consid­
ered. Fig. 6 demonstrates the pressure considered in one single pass of
the condenser. The pressure drop for each flow path is expressed by Eq.
(15). It includes the sum of segmental pressure drop in the inlet header,
contraction pressure drop from header to the microchannel tube ΔPi,in,
tube pressure drop ΔPi,tube, expansion pressure drop from tube to outlet
header ΔPi,out, and the sum of segmental pressure drop in the outlet
header. The pass model solves for a solution for mass flow rate and
quality for each microchannel tube to have equal pressure drop for each
flow path, as shown by Eq. (16). The approach of successive under
relaxation is employed in iteration to ensure a stable solution. Due to the
complexity of two-phase flow distribution, for one flow pass having a
two-phase inlet, the liquid flow rate is assumed uniform or estimated
from infrared temperature measurement, and the vapor flow rate for
each tube is solved.
∑
i ∑
Nt
Fig. 7. Procedures for modeling the separation condenser.
ΔPi,flow\_path = ΔPi,ihd + ΔPi,in + ΔPi,tube + ΔPi,out + ΔPi,ohd (15)
1 i
the inner loop using secant method until |p3L,ro - p3U,ro| < 0.05 kPa. The
ΔP1,flow\_path = ... = ΔPNt ,flow path (16) resultant ṁLU,pass will be compared with ṁLU,second header. If they are not
equal, ṁVU will be iterated in the outer loop until |ṁLU,pass - ṁLU,second
Fig. 7 demonstrates the procedures for modeling the whole separa­
header| < 0.1 g/s. With the final values of ṁVU and ṁLU, the final results
tion condenser. From the refrigerant-side and airside inputs, the 1st pass
for heat transfer and pressure drop in the upper passes and the lower
is calculated first. Then ṁVU is initialized, and it will be iterated in the
passes can be calculated. At last, the model will proceed to the 4th pass
outer loop. Using ṁVU and the condition at the 1st pass exit as the inputs,
in which the two streams recombine. The outputs of the whole model
the second header model is run and it outputs the liquid flow rate going
will be the refrigerant and air outlet states, heating capacity, pressure
into the upper pass ṁLU,second header, as well as the two boundary pres­
drop, and subcooling. The code is implemented in MATLAB (R2017a).
sures pU and pD. They are the inlet pressures for the 2nd-upper pass and
Refrigerant properties are got from REFPROP 10.0 [32].
2nd-lower pass. Then, in the inner loop, ṁLU,pass is iterated. With pU, pD,
ṁVU and the initial value of ṁLU,pass as the inputs to the model of
3. Experimentation
downstream passes, the outlet pressure of the 3rd-upper pass p3U,ro and
that of the 3rd-lower pass p3L,ro are calculated. ṁLU,pass will be iterated in
Fig. 8 is a schematic for the mobile air conditioning (MAC) system
used for experiments. The system consists of an open compressor, the
separation condenser under study, a manually-controlled electric
Table 1
expansion valve, a microchannel evaporator, and an accumulator before
Summary of selected heat transfer and pressure drop correlations.
the compressor suction. The compressor is an ACDelco compressor with
Item Correlation
a fixed displacement of 135 cm3 REV− 1. The evaporator is a two-slab
Airside four-pass microchannel evaporator with an overall dimension of W254
Heat transfer coefficient Chang and Wang [25] mm × H225 mm × D39 mm. It has 58 microchannel tubes in total and a
Pressure drop Chang and Wang [26]
fin density of 10 fins per inch. The total heat transfer area on the airside
Refrigerant side – Single-phase region
Heat transfer coefficient Gnielinski [27] is equal to 2.87 m2.
Frictional pressure drop Churchill [28] The airflow rate for a heat exchanger is controlled by a variable
Refrigerant side – Two-phase region speed blower in each environmental chamber. The flow rate is deducted
Heat transfer coefficient Cavallini et al. [29]
from the pressure drop and temperature at the flow nozzle downstream.
Frictional pressure drop Cavallini et al. [29]
Deceleration pressure drop Cavallini et al. [30]
In each chamber, the airside pressure drop across the flow nozzle is
Refrigerant side – Condensing superheated region measured by a differential pressure transducer and the dry-bulb tem­
Heat transfer coefficient Xiao and Hrnjak [31] perature at the nozzle outlet measured by a Type-T thermocouple

6
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Fig. 8. MAC system for testing the separation condenser.

(Omega Engineering) grid. Two thermocouple grids are installed up­

Qa = ṁa (hodn − hai ) (17)
stream and downstream of the evaporator and condenser for measuring
the dry-bulb temperatures. Two chilled-mirror dew-point sensors are
Qr = ṁr (hri − hro ) (18)
also installed at the heat exchanger inlet and nozzle outlet. Immersed
Type-T thermocouples and pressure transducers are placed along the where hodn is the specific enthalpy of well-mixed air at the outdoor
refrigerant circuit for depicting the refrigeration cycle. All thermocouple nozzle downstream of the condenser. ṁa is deducted from the air pres­
tips are ensured to be at the center of the tube. All pressure tubings are sure drop through the nozzle.
installed vertically upward on the tube. An absolute pressure transducer The final capacity is the average of the refrigerant-side capacity and
(Sensotec TJE) is employed to measure the outlet pressure of the sepa­ the airside capacity:
ration condenser, and a differential pressure transducer (Sensotec TJE)
is used to measure the pressure drop. An infrared camera (Mikron Qc =
Qr + Qa
(19)
Thermo Tracer) is used to measure the condenser surface temperature. 2
More information with details for the testing facility and instrumenta­ Using uncertainties for measured variables presented in Table 3, an
tion in the system can be found in Li and Hrnjak [10]. uncertainty analysis of the calculated parameters is carried out in EES
The separation condenser is a parallel-flow, single-slab condenser (2019). The uncertainty for condenser capacity Qc is ±1.5%. When
with louver fins designed for a mid-size sedan. The flow passes of the subcooling ΔTsub is higher than 5 K, the uncertainty for it is ±2%.
separation condenser are made by 54 aluminum microchannel tubes and
D-shaped headers. The tube number for each pass has been shown in 4. Results and discussion
Fig. 1. The condenser has a face area of 0.275 m2 and a total airside heat
transfer area of 5.22 m2. The other geometrical characteristics for the 4.1. Model validation
condenser are listed in Table 2.
Capacities of the condenser are obtained through independently First, the model is validated against experiment data under 67
measured airside capacity and refrigerant-side capacity. On the airside, different operating conditions per SAE Standard J2765 [33]. Air tem­
the condenser capacity is calculated as Eq. (17). The refrigerant-side perature is 35 ◦ C or 45 ◦ C, and air face velocity varies between 1.5 m s− 1
capacity is calculated as Eq. (18). and 3.7 m s− 1. The refrigerant condensing pressure ranges from 1175 to
1674 kPa and the R134a mass flow rate from 15.1 to 34.4 g s− 1, which
Table 2 corresponds to mass flux through the 1st pass in the range of 145–330
Main geometrical dimensions of the microchannel condenser with separation kg m− 2 s− 1. Subcooling at the condenser exit is controlled in the range of
circuitry. 0–20 K. Liquid flow rate distribution is assumed to be uniform for two-
phase inlet at one pass.
Item Value Item Value
Fig. 9(a) shows the comparison of predicted and measured heating
Width w/ headers [mm] 710 Number of MC ports per tube 12
capacities for the separation condenser. 98.5% of the data points are
[–]
Width w/o headers [mm] 687 Fin thickness [mm] 0.1
predicted within ±5% deviation from the experimental results. Fig. 9(b)
Width covered by fins 670 Fin pitch [mm] 1.21 compares the predicted and measured condenser pressure drop. 84% of
[mm]
Height w/ side plates 405 Louver pitch [mm] 0.88
[mm] Table 3
Height w/o side plates 390 Louver length [mm] 4.0 Measurement uncertainties.
[mm]
Measurement Unit Uncertainty
Depth [mm] 12.2 Louver angle [-] 28
MC tube thickness [mm] 1.43 Header type D- Refrigerant pressure kPa ±1
shape Refrigerant pressure drop kPa ±1
MC tube pitch [mm] 7.0 Header equivalent diameter 11.0 Air pressure drop Pa ±1%
[mm] Temperature ◦
C ±0.1
MC port Dh [mm] 0.68 Refrigerant mass flow rate g s− 1 ±0.1

7
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Fig. 9. Comparison of the experiment results and the model results: (a) Condenser capacity; (b) Pressure drop.

the data points are predicted within ±15% deviation from the experi­ diameter is probably one of the most intuitive variables to change for a
mental results. A lot of the underpredicted outliers are within the range parametric study. Fig. 11 simulates another inlet condition from Fig. 9,
of 45–75 kPa, which is at the lower end of ΔPc. The small base values and the diameter of the second header is studied as a variable. Equiva­
contribute the bigger difference. Consider the high sensitivity of pres­ lent diameter Deq of the D-shape header changes from 9.2 mm to 14.7
sure drop to tube diameter (approximately inverse of quartic), especially
in the range for microchannels, ±15% is an impressively small error
band to our best knowledge. Overall, the modeling results show good
agreements with the experimental results.
Fig. 10(a) shows the infrared image of the surface temperature of the
separation condenser at one experimental condition with Tri = 74.0 ◦ C
and Tro = 41.8 ◦ C. Fig. 10(b) shows the modeling results for the same
inlet condition. For this case, the liquid distribution profile for each pass
is estimated based on the length of the subcooled region of each tube
from the infrared image. The two profiles of surface temperature match
well. Tro = 44.2 ◦ C in the model, 2.4 K higher than Tro in the experiment.
Since both outlets are in the subcooled region, the difference in Tro only
causes a difference of 1.6% in Q. xU of the second header, as defined by
Eq. (3), is calculated to be 0.67 by the model. Complete phase separation
does not happen in this case, proving the necessity of the second header
model to calculate the separation efficiency. The calculated values of xU
for all the test conditions from Fig. 9 vary from 0.56 to 0.75, while the
inlet quality to the second header, xIN, is equal to 0.41 to 0.66. xU is
generally higher than xIN by 0.1–0.15.

Fig. 11. Separation efficiencies for different diameters of the second header
4.2. Effect of the diameter of the second header
(Pri = 1426.8 kPa, Tri = 72.9 ◦ C, ṁr = 26.1 g/s, Tai = 35.1 ◦ C, vai = 1.5 m/s,
RHai = 0.22).
Phase separation happens in the second header, so the header

Fig. 10. Comparison of the local wall temperature: (a) Experiment result; (b) Modeling result (Pri = 1333.9 kPa, Tri = 74.0 ◦ C, ṁr = 31.2 g⋅s− 1, Tai = 35.1 ◦ C, vai =
2.0 m⋅s− 1, RHai = 0.17).

8
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

mm. The rest of the condenser geometry is kept the same as in Fig. 1 and The effect of Deq on the condenser capacity will be discussed in a future
Table 2. In addition, Fig. 11 also demonstrates how the ηV and ηL are study.
fixed in a separation condenser from the perspectives of the second
header and the downstream passes. 4.3. Effect of the air velocity distribution
The pass model and the header model each output a curve of ηL a
function of ηV. The intersection of the two curves is the real separation As has been discussed in Section 2.2, in a separation condenser, the
efficiency. A similar trend for the curve of the second header model has upper path and the lower path meet at an integrated receiver. Between
been shown and explained in [13,14]: the more vapor goes up (ηV in­ the two exits of the second header and the integrated receiver, air ve­
creases) in the second header, the more liquid will be entrained upward locity affects the downstream flow resistance of the second header; thus,
(ηL decreases) because the upward interfacial drag force becomes the pressure drop in those passes. That sets the pressure boundary
stronger. The trend of the curve of the pass model is the opposite, which conditions at the downward exit and the upward exit, thus altering the
is based on the pressure drop balance between the upper passes and the separation efficiencies in the header. To demonstrate this, four cases are
lower passes, as discussed in Section 2.2. The more vapor goes up (ηV simulated at the same refrigerant inlet condition (Tri, Pri, ṁr). Airside
increases) into the 2nd-upper pass, the less liquid will go up (ηL in­ face velocity is changed from a uniform distribution of 1.5 m/s all over
creases), because if both ṁVU and ṁLU increase, the ΔP in the upper the condenser to three other non-uniform air distributions for the upper
passes will exceed that in the lower passes. ηV starts from 0.67 because passes and the lower passes, while the total volumetric flow rate of the
when ηV is lower than 0.67 ΔP in the lower passes will be too large for ΔP air is maintained the same. Table 4 shows the four different velocity
in the higher passes to balance. Similarly, ηV ends at 0.89, because when distributions. Deq of the second header is maintained as 11.0 mm, which
ηV is higher than 0.89 ΔP in the higher passes will be too large for ΔP in is the original size.
the lower passes. It is physically impossible for ηV to be out of the range For the same inlet condition as in Fig. 11 except for the air inlet
of 0.67–0.89 for this condenser geometry and this operating condition. velocity, Fig. 13 shows the different values of ηV and ηL calculated by the
When the condenser geometry or the operating condition changes, the model for the four velocity distributions in Table 4, each with the wall
physical range for ηV will change. temperature profile. From (a) to (d), the air velocity for the upper passes,
In Fig. 11, from the inlet condition to the condenser, the refrigerant vUai, increases, with (b) being the case of uniform vai. Comparing all
undergoes de-superheating and condensation in the 1st pass. The inlet cases in Fig. 13, with a higher vUai, a higher ηV, and a lower ηL exist. Also,
condition to the second header is p1st,ro = 1412.8 kPa, ṁ1st,ro = 26.1 g/s, with a higher vUai, the heat transfer coefficient of air becomes higher, so
x1st,ro = 0.553. Then, for this inlet condition and the same ηV, increasing the refrigerant temperature and the wall temperature at the exit of the
Deq reduces the vapor velocity in the vertical direction in the header, 3rd-upper pass become lower. Conversely, the wall temperature at the
thus decreasing the upward interfacial drag. So, less liquid goes up, and exit of the 3rd-lower pass becomes higher due to a lower airside heat
ηL increases. The ηL - ηV curve of the second header is shifted upward. As transfer coefficient for the lower passes.
a result, its intersection with the ηL - ηV curve of the pass model moves Although case (d) has the lowest refrigerant temperature at the exit
toward the upper right direction in Fig. 11. If Deq increases indefinitely, of the 3rd-upper pass, case (b) has the lowest the Tro (38.8 ◦ C) of the
the maximum values of ηL and ηV will be limited by the right end of the condenser. Subcooling of case (b) is 13.1 K. As all the cases in Fig. 13
pass model curve (here ηV = 0.89 and ηL = 0.97). Therefore, increasing have single-phase flows at the condenser outlet, Tro signifies the specific
the diameter of the second header will increase both ηL and ηV in a enthalpy hro at the outlet based on the property of the refrigerant R134a.
separation condenser with constant inlet condition. hro is the lower with a lower Tro. Since all cases have the same refrigerant
Fig. 12 shows the corresponding xU for each header diameter for the inlet condition thus the same ṁr and hri, case (b) has the highest Qc based
same inlet condition, as in Fig. 11. When both ηL and ηV increase, that on Eq. (18) for this particular comparison. The reason for that will be
means ṁVU increases and ṁLU decreases based on Eqs. (1) and (2). xU explained at the end of this section.
will, for sure increase based on Eq. (3), which is shown by the upward- Fig. 14 shows the ηL - ηV curves of the pass model for all the four cases
shifted curves of the second header model in Fig. 12. and explains the reason why ηV becomes higher and ηL becomes lower
The present study focuses on how the model works to calculate the from the case (a) to case (d) in Fig. 13. The upper passes with higher vUai
separation efficiency in the second header for different values of Deq. allow more vapor flow to come in because the higher air velocity lowers
the average quality, and the lower average quality reduces the average
frictional factor. As a result, at the same value of ηL, ηV is higher for
higher vUai, i.e., the curve is translated to the right-hand side (RHS).
Thus, the pass model curve intersects with the header model curve at a
higher ηV but a lower ηL.
Fig. 15 shows the corresponding xU for the same inlet conditions, as
in Fig. 14. The xU - ηV curve is also translated to the RHS as vUai increases.
When ηL decreases and ηV increases, that means both ṁVU and ṁLU in­
crease based on Eqs. (1) and (2), so it is hard to determine the changing
trend of xU directly. But the intersections in Fig. 15 show xU actually
decreases a little bit based on the relative magnitude of ṁVU to ṁLU, but

Table 4
Variable air velocity for the upper passes and the lower passes for studying the
effect of air distribution.
Distribution Distribution Distribution Distribution
1 2 3 4

Upper 1.2 1.5 1.8 2

passes
(m/s)
Lower 2.305 1.5 0.724 0.197
passes
Fig. 12. xU for different diameters of the second header (same inlet condition
(m/s)
as Fig. 11).

9
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Fig. 13. Wall temperature for different air velocity distributions: (a) vUai = 1.2 m/s; (b) Uniform air (vai = 1.5 m/s); (c) vUai = 1.8 m/s; (d) vUai = 2 m/s (Pri = 1426.8
kPa, Tri = 72.9 ◦ C, ṁr = 26.1 g/s, Tai = 35.1 ◦ C, RHai = 0.22).

Fig. 14. Different separation efficiencies for different air velocity distributions
for the upper and lower passes (same refrigerant and air inlet conditions Fig. 15. Different xU for different air velocity distributions for the upper and
as Fig. 13). lower passes (same refrigerant and air inlet conditions as Fig. 13).

the changing amount is quite small (from 0.61 to 0.60) compared to the passes decreases. Whether the total Qc will increase or not depends on
amount caused by reducing Deq in Fig. 12. The change of xU is con­ the trade-off between these two partial capacities. Fig. 16 shows the
strained by the relatively “flat” shape of the xU - ηV curve of the second capacities in the upper passes and in the lower passes for (a) – (d) from
header in the range of the intersections. The slope of the xU - ηV curve of Fig. 13. Here, the uniform-air-velocity case (b) has the highest sum, thus
the second header is a function of Deq and the inlet condition to the the highest Qc. Although (d) has the highest ηV and the highest capacity
second header. If Deq = 14.7 mm as in Fig. 12, the change of xU by the air in the upper passes, the increase in the upper passes is demolished by the
velocity maldistribution will be bigger because the slope of the second decrease in the lower passes. In the end, Qc in (d) is the lowest of all. The
header curve is bigger in magnitude. optimal air distribution to achieve the highest capacity will be further
At last, we explain why case (b) in Fig. 13 has the highest Qc. From discussed in another study.
the analyses above, from (a) to (d), both ṁVU and ṁLU increase, so the
refrigerant mass flux in the upper passes increases, and that in the lower 5. Summary and conclusion
passes decreases. While xU is almost the same, xD also changes little, so
the average HTCr in the upper passes increases due to a higher refrig­ A mechanistic model was built in the past to predict separation ef­
erant mass flux, and the average HTCr in the lower passes decreases due ficiency in the second header of separation condensers. In this study, the
to a lower mass flux. On the airside, the average HTCa in the upper second header model is incorporated into a novel model for separation
passes also becomes higher with a higher vUai. Therefore, from (a) to (d), condensers. The second header model provides inputs to the 2nd-upper
the capacity in the upper passes increases while the capacity in the lower pass and the 2nd-lower pass in a separation condenser. The condenser

10
J. Li and P. Hrnjak Applied Thermal Engineering 183 (2021) 116114

Appl. Therm. Eng. 94 (2016) 435–448, https://doi.org/10.1016/j.

applthermaleng.2015.10.074.
[5] J. Li, Q. Liu, Y. Duan, Z. Yang, Performance analysis of organic Rankine cycles
using R600/R601a mixtures with liquid-separated condensation, Appl. Energy 190
(2017) 376–389, https://doi.org/10.1016/j.apenergy.2016.12.131.
[6] J. Li, S. Hu, F. Yang, Y. Duan, Z. Yang, Thermo-economic performance evaluation
of emerging liquid-separated condensation method in single-pressure and dual-
pressure evaporation organic Rankine cycle systems, Appl. Energy 256 (2019),
113974, https://doi.org/10.1016/j.apenergy.2019.113974.
[7] X. Zhang, J. Li, P. Qi, D. Chao, Experimental study of condensation heat transfer in
a condenser with a liquid-vapor separator, Appl. Therm. Eng. 152 (2019) 196–203,
https://doi.org/10.1016/j.applthermaleng.2019.02.086.
[8] W.A. Davies III, Y. Kang, P. Hrnjak, A.M. Jacobi, Effect of inclination on heat
transfer and flow regimes in large flattened-tube steam condensers, Appl. Therm.
Eng. 148 (2019) 999–1006, https://doi.org/10.1016/j.
applthermaleng.2018.11.078.
[9] J. Li, P. Hrnjak, Separation in condensers as a way to improve efficiency, Int. J.
Refrig. 79 (2017) 1–9, https://doi.org/10.1016/j.ijrefrig.2017.03.017.
[10] J. Li, P. Hrnjak, Improvement of condenser performance by phase separation
confirmed experimentally and by modeling, Int. J. Refrig. 78 (2017) 60–69,
https://doi.org/10.1016/j.ijrefrig.2017.03.018.
[11] S.P. Won, Performance Analysis of Integral Receiver/Dryer Condenser for
Automobile, Proceedings of 2006 International Forum on Strategic Technology,
IEEE, (2006) 84–88. DOI: 10.1109/IFOST.2006.312254.
[12] J. Li, P. Hrnjak, Visualization and quantification of separation of liquid-vapor two-
phase flow in a vertical header at low inlet quality, Int. J. Refrig. 85 (2018)
Fig. 16. Capacities in the upper passes and the lower passes for different air 144–156, https://doi.org/10.1016/j.ijrefrig.2017.09.018.
velocity distributions (same refrigerant and air inlet conditions as Fig. 13). [13] J. Li, P. Hrnjak, Parametric study and improvement of phase separation in
intermediate headers of microchannel condensers, J. Fluids Eng. 142 (1) (2020),
011304, https://doi.org/10.1115/1.4044480.
modeling results are compared to the experimental results of a mid-size [14] J. Li, P. Hrnjak, Phase separation in vertical header of microchannel condensers—a
mechanistic model, J. Fluids Eng. 142 (1) (2020), 011303, https://doi.org/
MAC condenser. The deviation for capacity is ±5%, and that for pressure
10.1115/1.4044285.
drop is ±15%. The model could be used to provide a guide for the design [15] R.G. Turner, M.G. Hubbard, A.E. Dukler, Analysis and prediction of minimum flow
of separation condensers. While separation happens in most cases, the rate for the continuous removal of liquids from gas wells, J. Petroleum Tech. 21
efficiencies (ηV for vapor and ηL for liquid) are not equal to 100% for the (11) (1969) 1475–1482, https://doi.org/10.2118/2198-PA.
[16] Gas Processors Suppliers Association (GPSA), Engineering Databook, 2004.
simulated conditions. The quality going up into the 2nd-upper pass, xU, [17] J.C. Beale, R.D. Reitz, Modeling spray atomization with the Kelvin-Helmholtz/
varies from 0.56 to 0.75, while the inlet quality to the second header xIN Rayleigh-Taylor hybrid model, At. Sprays 9 (6) (1999) 623–650, https://doi.org/
is in the range of 0.41 to 0.66. 10.1615/AtomizSpr.v9.i6.40.
[18] P.J. O’Rourke, A.A. Amsden, The TAB Method for Numerical Calculation of Spray
For a fixed inlet condition, a larger diameter of the second header Droplet Breakup, SAE Technical Paper 872089, SAE, 1987. Doi: 10.4271/872089.
increases ηV and ηL in the condenser. xU also increases as the diameter [19] R.D. Reitz, Modeling atomization processes in high-pressure vaporizing sprays, At.
increases. By changing the downstream air velocities while maintaining Spray Tech. 3 (4) (1987) 309–337.
[20] S.V. Apte, M. Gorokhovski, P. Moin, LES of atomizing spray with stochastic
the total airflow rate a constant, it is shown that the downstream flow modeling of secondary breakup, Int. J. Multiphase Flow 29 (2003) 1503–1522,
resistance has an impact on the separation efficiencies as well. https://doi.org/10.1016/S0301-9322(03)00111-3.
Increasing the air velocity for the upper passes increases ηV but decreases [21] G.B. Wallis, One-Dimensional Two-Phase Flow, McGraw-Hill, New York, 1969.
[22] S.M. Zivi, Estimation of steady-state steam void-fraction by means of the principle
ηL. xU is decreased as the air velocity for the upper passes increases, and
of minimum entropy production, J. Heat Transfer 86 (2) (1964) 247–251, https://
the variation is smaller than that caused by changing the diameter of the doi.org/10.1115/1.3687113.
second header. [23] L. Friedel, Improved friction pressure drop correlations for horizontal and vertical
two phase pipe flow, 3R Int. 18 (7) (1979) 485–491.
The effects of diameter on condenser capacity and the optimal air
[24] F.P. Incropera, D.P. Dewitt, T.L. Bergman, A.S. LaVine, Fundamentals of Heat and
distribution will be discussed in another study. Mass Transfer, 6th ed., John Wiley & Sons, 2007.
[25] Y.J. Chang, C.C. Wang, A generalized heat transfer correlation for louver fin
Declaration of Competing Interest geometry, Int. J. Heat Mass Transf. 40 (3) (1997) 533–544, https://doi.org/
10.1016/0017-9310(96)00116-0.
[26] Y.J. Chang, C.C. Wang, Air side performance of brazed aluminum heat exchangers,
The authors declare that they have no known competing financial J. Enhanced Heat Transf. 3 (1) (1996) 15–28, https://doi.org/10.1615/
interests or personal relationships that could have appeared to influence JEnhHeatTransf.v3.i1.20.
[27] V. Gnielinski, New equations for heat and mass transfer in turbulent pipe and
the work reported in this paper. channel flow, Int. Chem. Eng. 16 (2) (1976) 359–368.
[28] S.W. Churchill, Friction-factor equation spans all fluid-flow regimes, Chemical Eng.
Acknowledgments 84 (24) (1977) 91–92.
[29] A. Cavallini, L. Doretti, M. Matkovic, L. Rossetto, Update on condensation heat
transfer and pressure drop inside minichannels, Heat Transfer Eng. 27 (4) (2006)
The authors thankfully acknowledge the supports from the Air 74–87, https://doi.org/10.1080/01457630500523907.
Conditioning and Refrigeration Center at the University of Illinois, [30] A. Cavallini, D. Del Col, M. Matkovic, L. Rossetto, Frictional pressure drop during
vapour–liquid flow in minichannels: modelling and experimental evaluation, Int. J.
Urbana-Champaign, Creative Thermal Solutions, Inc., and Dr. John
Heat Fluid Flow 30 (1) (2009) 131–139, https://doi.org/10.1016/j.
Meyer of Hanon Systems. ijheatfluidflow.2008.09.003.
[31] J. Xiao, P. Hrnjak, A heat transfer model for condensation accounting for non-
References equilibrium effects, Int. J. Heat Mass Transf. 111 (2017) 201–210, https://doi.org/
10.1016/j.ijheatmasstransfer.2017.03.019.
[32] E.W. Lemmon, I.H. Bell, M.L. Huber, M.O. McLinden, Reference Fluid
[1] Q. Liu, R.M. Gilden, J.P. Luther, High capacity condenser, US Patent No. Thermodynamic and Transport Properties Database – REFPROP, Version 10.0, US
5,752,566, 1998. Secretary of Commerce, 2018.
[2] D. Wu, Z. Wang, G. Lu, X. Peng, High-performance air cooling condenser with [33] SAE International. Procedure for Measuring System COP [Coefficient of
liquid-vapor separation, Heat Trans. Eng. 31 (12) (2010) 973–980, https://doi.org/ Performance] of a Mobile Air Conditioning System on a Test Bench, SAE Surface
10.1080/01457631003638952. Vehicle Standard J2765 OCT2008, 2008.
[3] X. Peng, L. Jia, Z. Cheng, Taiwan, China Patent No. 188060, 2003.
[4] X. Luo, Z. Yi, Z. Chen, Y. Chen, S. Mo, Performance comparison of the liquid–vapor
separation, parallel flow, and serpentine condensers in the organic Rankine cycle,

11