You are on page 1of 32

Accepted Manuscript

Numerical study and visualization on flow characteristics of reflux condensation


in air-cooling condenser

Deng Hui, Liu Jizhen, Yang Tingting, Wu Sai

PII: S1359-4311(18)34730-6
DOI: https://doi.org/10.1016/j.applthermaleng.2018.11.109
Reference: ATE 12990

To appear in: Applied Thermal Engineering

Received Date: 31 July 2018


Revised Date: 16 October 2018
Accepted Date: 26 November 2018

Please cite this article as: D. Hui, L. Jizhen, Y. Tingting, W. Sai, Numerical study and visualization on flow
characteristics of reflux condensation in air-cooling condenser, Applied Thermal Engineering (2018), doi: https://
doi.org/10.1016/j.applthermaleng.2018.11.109

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Original version

Numerical study and visualization on flow characteristics of


reflux condensation in air-cooling condenser

Deng Hui1, Liu Jizhen1, Yang Tingting1, Wu Sai2


1. North China Electrical Power University, Beijing, China
2. China Electric Power Research Institute, Beijing, China

1
Original version

Highlight
Flow of reflux condensation in air-cooling condenser tube is studied numerically.
A model-based approach to determine the condensation length is proposed.
Effect of varying vapor/air operations on reflux condensation is considered.
Dynamic behavior relating to complete and countercurrent condensation is
visualized.

Abstract
A numerical simulation was presented to investigate and visualize the flow
characteristics of reflux condensation in core tube related to air-cooling condenser. The
simulation had been performed in a flat tube under a saturation pressures of 13 kPa. The
sensitivity of condensation length, condensate thickness and interfacial shear was
examined with varying the vapor mass flux and cooling air Reynolds number. In
addition, the condensate-void and vapor-absence corresponding to complete
condensation were visualized in a transient simulation. Interfacial rivulets, film
climbing and liquid entrainment corresponding to flow-through condensation were
identified by contours of liquid void fraction and further explained by the
cross-sectional velocity profiles. Meanwhile, several recommendations to experiment
and operation on upflow finned tubes were proposed based on numerical results.

KEYWORDS: Air-cooling condenser; Flow characteristics; Reflux condensation;


Numerical simulation.

2
Original version

1 Introduction
Air-cooling condensers have been widely appied in fossil fuel power plants
featuring coal-rich but water-scarce, because of their capability to reduce cooling water
consumption in the Rankine cycle. A typical air-cooling condenser consists of hundreds
of uniform finned tubes. A single finned-tube is assembled by one core tube (in one of
flat, oval, or round cross-sectional shapes) with thousands of fins welded. Typically,
finned-tubes condense steam at sub-atmospheric pressures to improve the efficiency of
Rankine cycle. Meanwhile, the low pressure favors non-condensable gas, i.e., the
atmospheric air leaking into tube, which contributes to deteriorate the condensation
perfomance [1]. To reduce the effect of air, current condensers deploy two types of
finned-tubes, as downflow and upflow, according to the orientation of steam in core
tubes (upward or downward). Downflow tubes serve as basic condenser devices, and
the upflow remove air in cooperation with condensers. During a condenser operation,
the excess steam that fails to through-flow condensation in downflow systems is
introduced into core tubes of upflow tubes with the leaked air. Then, the steam
condenses totally in countercurrent, while the air flows upward due to buoyancy and is
eventually removed by an evacuation.
The finned tube functions in an air-cooling condenser as the elemental heat
exchanger, enhancing its condensation performace has been attracting significant study
attention. Because the convective heat transfer coefficient of air convection on fin-side
is much lower than condensation heat transfer coefficient on vapor-side, most study in
past focused on the effects of geometries and structure related to fins (e.g., [2–5]).
There have been several recent studies addressing the thermal performance and flow
characteristics on vapor-side. Owen et al. [6] examined the vapor distribution from the
main duct to downflow flat tubes. O’Donovan et al. [7] conducted an experiment on
pressure losses in a circular downflow bundle. Kekaula et al. [8] modelled the film
condensation in a round core tube with the Nusselt analysis [9] and pool model [10],
respectively. Berrichon et al. [11] carried out a flooding experiment of the
liquid-and-air mixture reflux flow in a vertical round tube without incorporating
condensation into the flow. Moreover, the same authors [1] compared the condensation
heat transfer coefficient between co-current and reflux in a vertical round tube based on
an analytical model. O’Donovan et al. [12] performed an experimental and theoretical
study on a prototype downflow air-cooling condenser to determine the performance of

3
Original version

a 50 MW power plant in terms of steam-side. Kang et al. [13] presented an experiment


to the inclination effect on pressure drop and flow regime of condensation in a
downflow finned-tube with flat core tube. Mahvi et al. [14] developed a model to
quantify effects of vapor-side operation on a downflow air-cooling condenser.
Studies [1, 7, 8, 11−14] suggested that both operations on vapor/air sides should be
considered to improve air-cooling condenser performance, among which most work
focused on the through-flow condensation in downflow tube, the limited has been
conducted on the complete condensation in upflow tube. Two example are the studies
of Berrichon et al. [1, 11], based on analytical and experimental methods, respectively.
Therefore, the flow characteristic in upflow tube is the initial aims of this study to
investigate with a numerical simulation using the volume of fluid (VOF) method [15]
and Lee’s phase change model [16].
The VOF method associated with Lee’s mode has been confirmed as an effective
tool to investigate upflow condensation, especially those regarding flow pattern,
interfacial behavior and pressure drop. For instance, Kharangate et al. [17] explored
the upflow condensation of refrigerant FC-72 in a vertical round tube. The numerical
results captured the 2D flooding and climbing film. Qiu et al. [18] simulated a 3D
simulation of the forced convective upflow condensation. The results showed that the
heat transfer coefficient and frictional pressure drop increased with mass flux and vapor
quality. Qiu et al. [19] performed a 3D simulation of the condensation of wet steam
upflow inside a vertical pipe. The results revealed the bubbly, slug, churn, and annular
flows. Cui et al. [20] developed a 2D simulation of hydrocarbon mixture upflow
condensation in an inclined channel. Parametric studies were performed to investigate
the model under various boundaries. Fiorentino and Starace [21] presented a 2D
simulation to the temporal two-phase flow of falling film against reflux air in an
experimental evaporative condensers [22], the numerical stable and drops film were
observed with varying the mass ratio between water/air. The recent numerical research
are promising, few is related to the flat flow channels and thermos-physical properties
to those of condenser tube in the present study; and the vast majority focused on the
through-flow condensation, instead of the complete condensation reported in the study.
Not limited to these heretofore results, two shortcomings in application of VOF
combined with Lee’s mode is draw out from literature, as that (1) predicted local heat
transfer coefficient exhibited noticeable discrepancy with target at the vapor-inlet
region; and (2) an universal agreement on the evaluation of mass transfer intensity

4
Original version

factor has yet to be reached for Lee’s model. Details of shortcomings would be
discussed in the section related to model validation.
Despite the importance of flow characteristic to understanding the thermodynamic
particulars in reflux systems and to ensure the reliable operation of upflow tubes, the
numerical application addressing this issue is not well-documented, especially on the
condensation length and interfacial behaviors in terms of effects of cooling air flow and
vapor inlet conditions, as well as on the visualization of complex dynamic behaviors
involved in countercurrent flow and complete condensation. As per the literature
review, this is perhaps the first study of its kind to develop a numerical approach to
study the interfacial behaviors and flow visualization of upflow condensation in finned
air-cooling condenser tube.

2 Computational methods
2.1 Governing equations
The VOF method, phase-change model, momentum and energy conservation
equations, species transport equation, and the turbulence model adopted in the
commercial Ansys Fluent ® [23], were used for the formulation of the present
condensation problem and implementation of the numerical simulation.
The VOF method was separately implemented for the computation of the
continuity equations for two phases as follows:

 t ( L  L )  ( L  L uL )  SL

  (  )  (  u )  S . (1)
 t V V V V V V

 L   V  1

Assuming that the mass-related film condensation was transferred at a constant


pressure, accordingly, a quasi-thermal equilibrium state (i.e., Lee phase-change model
[16]) was used for the evaluation of the mass transfer between two phases and
calculation of the source terms in continuity equations, SL and SV, as follows.
For condensation,
T  Ts
S V   S L  r V  V (T  T s ). (2)
Ts

For evaporation,

5
Original version

T  Ts
S V   S L  r L  L (T  T s ). (3)
Ts

In Eqs. 2 and 3, the T and Ts denote the cell temperature and saturate temperature,
respectively.
The momentum and energy equations were calculated for the combined phases,
which are listed as follows.
Momentum equation:

(  u )    (  u  u )  p    [  (u  u T )]   g. (4)
t
Energy equation:

  E     [(  E  p)u ]    ( eff T   h i J i )  Q, (5)
t
the source term in energy equation Q caused by phase-change was computed as
Q  S L   LV , (6)
where λLV is the latent heat of vaporization.
The mixture inside core tube of finned-tube consists of only two species, as water
vapor and air. Therefore, the mass fractions in gas phase cells of the computational
domain satisfies the following equation:
Y V  Y A  1. (7)
Therefore, a species transport equation based on a single species (water vapor) was
adopted to model the mass diffusion between the two species as follows:
( YV )
   (  uYV )  J V  Sd , (8)
t
where Sd denotes the source term in the transport equation, which was assumed to be
caused by the condensation of vapor species and determined as the source term in
continuity equations, SV:
S d  S V. (9)
The mass flux of species diffusion, JV was calculated from the concentration
gradient of vapor species as
t
J V  (  Dm  )YV , (10)
Sct

where μt and Sct are the turbulent viscosity coefficient and turbulent Schmidt number,
respectively; and Dm is the mass diffusion coefficient based on the Chapman–Enskog
equation [24].

6
Original version

The shear stress transport (SST) k–ω turbulence model [25], which was validated
to predict hydrodynamics and heat and mass transfer characteristics of two-phase flow
in several studies [26], was applied in the present study, as follows:

(  k )    (  ku )    ( Γ k k )  G k  Y k  S k ; (11)
t

(  )    ( u )    ( Γ   )  G   Y  D   S  . (12)
t
k and ω on the left-hand-side of Eqs. 11−12 represent the turbulence kinetic energy
and specific dissipation rate, respectively. On the right-hand-side, Gk and Gω represent
the generation of k and ω due to mean velocity gradients; Γk and Γω denote the
effective diffusivity of k and ω; Yk and Yω represent the dissipation of k and ω due to
turbulence; Dω represents the cross-diffusion term; Sk and Sω are user-defined source
terms.

2.2 Computational domain


To reduce CFD mesh quantity, an under-scale 4:1 prototype tube was developed
on the basis of the analogy principle to model a full-sized engineering flat tube, which
is widely deployed in current air-cooling condensers of 600 MW power plants. The
same analogy method and same engineering tube used by Kong et al. [27] were applied
in the design of an experimental condenser module. The geometric dimensions of the
prototype tube are listed in Table 1.
Table 1
Dimensions and geometric parameters of under-scale core tubes.
Geometric Parameter Values of
parameters symbols under-scale tube

Major axis of plate w 50 mm


Long axis of semi-oval a 3.8 mm
Short axis of semi-oval b 2.5 mm
Tube axial length L 2.5 m
Tube inclination angle β 30°
Cross-section area A 280 mm2
Hydraulic diameter DH 10 mm

Fig. 1 demonstrates a symmetrical 3D computational domain that requires


minimal computational effort. The geometric parameters in Table 1 were used during
the simulation, and quadrilateral meshes were adopted for all domains. The meshes
near the cooling wall were non-uniform and carefully refined as boundary cell layers in
a C-grid block [28], in which the length of C-grid, nodes, mesh law, spacing 1, and ratio

7
Original version

1 were adopted to 0.3 mm, 15, Geometric1, 5 μm, and 1.15, respectively. Thus, thin
liquid film formation was captured. The other meshes were uniform.
x-y view x-z view

Pressure z
outlet
x

Cooling-wall
Symmetrical

β = 30°
Gravity
Boundary

w = 50 mm
cells
L = 2.5 m

Two-phase
reflux flow

Cross-section view
(y-z view )

y
Velocity a = 3.8 mm
inlet
x b = 2.5 mm
Fig. 1 Computational domain of two-phase fluid system with local coordinate system.

2.3 Simulation conditions and solution techniques


A velocity inlet and pressure outlet were adopted at the lowest and top ends of flat
tube in Fig. 1, respectively. The vapor-inlet velocity was determined as U = G/ρV. Using
the analogy principle, the range of the mass flux G corresponded to the Reynolds
number of 40−100% operational exhaust at the turbine heat rate acceptance mode with
the assumption that the vapor distribution between downflow/upflow finned-tubes was
7:3. The inlet turbulence intensity Ti was evaluated in the following empirical
correlation [23]:
u
Ti   0.16 R e 1/8 . (13)
u

The saturated temperature Ts, determined by the exhaust pressure Ps at the turbine heat
rate acceptance mode, was adopted at the lowest end of flat tube.

8
Original version

(a)

Temperature (K)
Wall temperature (K)

G = 8 kg/(m2∙s)
Rea = 5,021
Ts = 324 K
x = 25 mm L = 2.5 m

Ts = 324 K
(c)
y (mm)

Two-phase
(b) G = 8 kg/(m2∙s) reflux flow
Wall temperature (K)

Ts = 324 K
y = 1250 mm

Rea Twall
1,340 G = 8 kg/(m2∙s)
2,662 Rea = 5,021
5,021 y Ts = 324 K
7,508
9,978
1,2379 x

x (mm)
Fig. 2 Correlation among Twall, Ts and Rea in upflow tube, as Twall variation (a) along axial
direction, (b) along radial direction; and (c) a filled contour of Twall.
The wall temperature of an operational finned-tube varies at location and depends
on the operations on both vapor/air sides, as the coupled results of crossflow between
the cooling fluid and the two-phase fluid. Refer to the infrared thermographs of
finned-tubes surface in the literature [3, 29], the wall temperature and their location
along the axial and radial orientations in flat tube were extracted using Matlab® Image
Processing Toolbox [30]. With these data, the correlation among the local temperature
Twall, inlet-vapor temperature Ts, cooling air Reynolds number Rea, and the location of
Twall was proposed with a Mathematica® polynomial function [31], which was adopted
on cooling wall of numerical tubes using Fluent’s UDF DEFINED_PROFILE [30], as,
T wall  T x  T x,ref , (14)

T x  k 0  k 1 x *  k 2 x *2  k 3 x *3  k 4 x *4  k 5 x *5  k 6 x *6 , (15)
T x,ref  T y  Ts , (16)

T y  24  23 y *  222 y *2  79 y *3  1191y *4  1546 y *5  611y *6 , (17)

where, x* and y* denote the dimensionless x and y of Twall, determined as x/w and y/L,
respectively. The coefficients in Eq. (15) were determined by the Rea, as listed in Table.
2. Figs. 2(a) and 2(b) present the variation of wall temperature with axial and altitude
locations in this study, respectively. The filled contour of wall temperature illustrated

9
Original version

in Fig. 2(c), visualized the proposed method.


Table 2
Coefficients in Eq. (15).
Rea k0 k1 k2 k3 k4 k5 k6

1,340 −0.058327 −4.38815 146.669 −1287.34 3399.15 −3548.74 1289.67


2,662 −0.805586 11.7927 22.5675 −833.617 2625.54 −2981.41 1153.42
5,021 −1.95535 26.4671 −33.1624 −694.233 2460.83 −2920.11 1160.49
7,508 −1.90839 62.5774 −626.715 1931.52 −2790.95 1983.31 −573.827
9,978 −1.15297 30.3658 −202.775 −67.8927 1535.74 −2334.77 1036.34
12,379 −4.56751 75.7009 −406.243 751.278 −302.825 443.188 323.444

Table 3 provides the thermodynamic conditions of vapor at the core tube inlet and
the cooling flow Reynolds number on fin-side in simulations.
Table 3
Initial conditions of vapor and cooling air in simulations.
Ps Ts ρV G U Rea
(kPa) (K) (kg/m3) (kg/m2∙s) (m/s) (−)

13 324 0.087 8–21 140–220 1,340–12,379

Numerical simulations were conducted using a transient time scheme with a


pressure-based solver. The explicit VOF method [23] was used in the multiphase model
with default values on volume fraction cutoff, and implicit body force was enabled.
Turbulence effects were considered using the SST k–ω model [25] with default model
constants. The semi-implicit method for pressure-linked equation scheme [33] was
utilized to handle pressure–velocity coupling. Green–Gauss node-based formulation
[34], pressure staggering option [35], and compressive interface capturing scheme for
arbitrary meshes [36] were adopted for gradient, pressure, and volume fraction
discretization, respectively. The second-order upwind [37] scheme was used for energy
discretization while first-order upwind scheme [37] was adopted for turbulent kinetic
energy, specific dissipation rate, and momentum discretization.
Specifically to transient computation, the Courant number for solving the VOF
equation was restricted to 0.25. A variable time stepping method with an initial time
step size of 2×10−5 s was adopted to solve transient equations. The number of time steps
for each numerical case was set to 1,000, and the maximum iteration per time step was
set to 60. The absolute convergence criterions for all governing equations were set to
10−3 except the energy equation in the value of 10−6.

2.5 Grid independence


A grid independence analysis of computational domain with vertical state shown

10
Original version

in Fig. 1 was conducted in through-flow condensation mode at a uniform temperature


wall boundary. Domain discretization was sequentially performed at five mesh levels
and with an element size that ranged from 20 μm coarse level to 2 μm refined level.
The average heat transfer coefficient on the cooling wall from 0.3 m to 2.5 m along
axial depth and the dynamic pressure drop ∆P between vapor inlet and two-phase
outlet were monitored for five levels of domain discretization. Based on a mass
transfer intensity factor of 1,200,000 s−1, the deviations on the average heat transfer
coefficient and ∆P between the results for 5 and 2 μm were only 1.0% and 1.1% for
the numerical tube reported, respectively. Therefore, the mesh of 5 μm elements with
a sum of 5,950,800 could be satisfactorily utilized to evaluate the heat-mass-transfer
performance for the study of numerical computations. Fig. 3 illustrates the details of
heat transfer coefficient and ∆P in grid analysis from coarse mesh to refined mesh.

(a) (b)
Averaged HTC (W/m2K)

1.1%
1.0%
ΔP (Pa)

G = 18 kg/(m2∙s) G = 18 kg/(m2∙s)
Ts = 324 K, Ps = 13 kPa Ts = 324 K, Ps = 13 kPa
ΔTcool = 5 K ΔTcool = 5 K

5/5,950,800 15/3,620,000 5/5,950,800 15/3,620,000

2/7,342,400 10/5,044,000 20/2,860,000 2/7,342,400 10/5,044,000 20/2,860,000


Element size/Mesh size (μm/−) Element size/Mesh size (μm/−)
Fig. 3 Analysis of (a) average heat transfer coefficient and (b) dynamic pressure drop to determine
grid independence.

3 Model validation
The mass transfer intensity factor in Lee model r (Eqs. 2 and 3) plays an
important role in simulation of condensation applications. However, a universal
agreement on the evaluation of r has yet to be reached, the differences among adopted
values are a considerably large according to the working fluid, geometry, and operating
parameters [17−20]. In the present study, the suitable value of r was determined by a
trial-and-error procedure. Different values of 400,000; 600,000; 1,000,000; and
1,200,000 s−1, were examined in pursuit of the best agreement between the predicted
heat transfer coefficients and target values estimated by a well-known empirical
correlation, i.e., Shah’s correlation [38]. The data used for the validation in Shah
correlation included water condensation inside various shapes (round, rectangular, and
triangular) of channels in all orientations (horizontal, vertical, and angles in between
upward and downward), with the diameter ranging from 0.1 to 49 mm, reduced

11
Original version

pressure from 0.0008 to 0.946, and flow rate from 1.1 to 1,400 kg/(m2·s), all of which
covered the range of the corresponding parameters in this study.
The predicted heat transfer coefficients were calculated through a series of trial
cases as G = 5−20 kg/(m2∙s), Ts = 324 K, and a uniform temperature wall of ∆Tcool =
3−7 K. On the basis of comparison in predicted heat transfer coefficient with target
value among r = 400,000; 600,000; 1,000,000; and 1,200,000 s−1, the value of r =
1,200,000 s−1 was applied. The predicted average heat transfer coefficients were
compared with Shah’s results in Fig. 4(a). Overall, the predicted average values
underestimated the targets, and the errors were less than −15%. The comparison of
predicted local heat transfer coefficient with Shah’s results was illustrated in Fig. 4(b).
Considering the bias value of 10% in the figure, a good agreement between the
predicted and correlation could be observed in the middle and downstream regions.
However, the predicted values showed noticeable discrepancy from the target in the
vapor-inlet. Such discrepancy might be attributed to the difficulty in initiating the
accurate values of condensate thickness in this region. This assumption explained
several similar numerical discrepancies in simulating film condensations with the VOF
method associated with Lee model [17, 19, 39−41]. Therefore, the local interfacial
behaviors near the vapor-inlet were intentionally excluded for later discussion because
of difficulty modeling this region. Nonetheless, the predicted local heat transfer
coefficient captured the spatial target values well over the region 0.5 m < y < 2.5 m, as
shown in Fig. 4(b).
Predicted average HTC (W/m2K )

+15% (b)
(a)
G = 14 kg/(m2∙s)
Local HTC (W/m2K )

Ts = 324 K 10% bias bar


ΔTcool = 4 K
−15%
Shah correlation [36]
2
G = 5−20 kg/(m ∙s)
Ts = 324 K
Predicted HTC
ΔTcool = 3−7 K

Shah [36] average HTC (W/m2K) y (mm)

Fig. 4 Comparison of predicted heat transfer coefficients with Shah [36] results.

4 Results and discussions


4.1 Data processing
4.1.1 Determination of condensation length

When an upflow finned-tube works in complete condensation mode, two separate

12
Original version

but interconnected sections exist simultaneously in core tube, namely, the phase-change
section and single phase section. Determining the length of phase-change section is the
precondition for analyzing the flow results of complete condensation. A new approach
to estimating the length of phase-change was developed in this study using an
analytical model as follows.
The thickness of condensate film in an upflow flat tube can be modeled as a
partial differential equation [42] considering the interfacial shear between vapor and
condensate in the unit control volume, as shown in Fig. 5.
      (T  T )
( f  v ) g (cos   sin  ) 3  ( I,x   I,y ) 2  f f s wall (18)
x y x y LV f
my
mc |I
z  y |I
mx

g  x |I
x
y m
mx + x dx
m y x
my + dy
y u

w
Fig. 5 Control VOF domain on plate part of flat core tube in upflow finned-tube.
Meanwhile, the outlet condensate flux with an axial length of y mm could be
written as the integration of the rates of unit control volume [42], as:
w y
mc ( y)  2 m( y)  2 m( x) |x  w . (19)
0 0
1 2

The first term in Eq. 19 represents the axial flow flux and the second term is the radial
flow flux; w denotes the length of major axis of the plate (which is 50 mm, as shown in
Table 2). The coefficient “2” of both terms means that the flow flux from the unit
control volume should be doubled because the flat tube is axially symmetrical.
The axial and radial flow rates in Eq. 19 could be calculated using the average
velocity of the condensate film [42], as:
 f ( f   v ) g cos  3  I,x 2
mx  u f  dy =[ 3f
 
2 f
 ]f dy

 . (20)
m  w  dx =[ f ( f   v ) g sin   3   I,y  2 ] dx


y f
3f 2 f
f

13
Original version

By combining Eqs. 18−20, the outlet condensate flow rate was used to determine
the length of phase-change section, i.e., the complete condensation length, in the
upflow tube using the following numerical procedure.
(1) Numerical cases with initial conditions listed in Table 2 were performed
using Ansys Fluent ®.
(2) The values of wall temperature and wall shear with their location coordinates
were conserved using the User-Defined Memory Macros [32] from Step 1.
(3) With the data in Step 2, the correlation of wall temperature and shear versus
their locations were proposed using the polynomial function in the
computation software Wolfram Mathematica® [31].
(4) Eq. 18 was solved using the NSolve function in Wolfram Mathematica® [31]
on the basis of correlations relating to temperature and shear in Step 3.
(5) The initial value of L was set to 2,500 mm, and the decrement was set to 10
mm. From the numerical solution in Step 4, the outlet condensate flow rate
with y = L was determined as
mc ( y)  mc ( y) / A, (21)
where A denotes the cross-sectional area of the flat tube (which is 280 mm2,
seen in Table 2), and mc(y) denotes the outlet condensate flux in Eq. 19 which
was calculated numerically through Eqs. 18 and 20 using the NIntegrate
function in Wolfram Mathematica® [31], until the outlet condensate flow rate
was less than or equal to the inlet vapor flow flux G.
(6) The final value of L in Step 5 was the complete condensation length.

4.1.2 Interfacial parameters

The local liquid film thickness, δ, was defined as a function of axial length y mm
in the following equation:
N
 i    ij ( y ) / N , (22)
j 1

where the superscript i denotes the values of upper semi-oval (USO), plate (PL), and
down semi-oval (DSO) of flat core tube; δj(y) is the distance from the wall to the center
of cell at the two-phase interface; y is the tube axial length, and N is the sum of these
cells. In the study, the cell at two-phase interface was determined by whether its value
of liquid void fraction ranged within a target scale, defined as follows:
0.3  a L  0.9. (23)

14
Original version

The interfacial shear between the vapor and liquid condensate, τ, was calculated
using the following function of interfacial velocity gradient [43]:
u
   . (24)
z

The shear in Eq. 24 was simplified by the axial shear on cooling wall. The
simplification has been proven reasonable in several related works (e.g., Li [44]),
according to the negligible vapor radial velocity compared with axial velocity and the
insignificant condensate thickness compared with the two-phase flow channel
dimensions.

4.2 Flow characteristics


The upflow condensation in computational domain of Fig. 1 was simulated
considering effects of vapor-inlet mass flux G and cooling air flow Reynolds number
Rea. When the simulations were conducted to verify effects of G, the flat tube was
assumed operation at a fixed Rea on air-side. Similarly, when the simulations were
conducted to verify effects of Rea, a fixed G on vapor-side was assumed.

G = 16.3 kg/(m2∙s)
Ts = 324 K, Ps = 13 kPa
Wall z+ (−)

Rea = 2,662

y (mm)
Fig. 6 Variation of wall z+ along condensation length representative of flow characteristics cases.
Before discussing the numerical results, the variation of dimensionless wall
distance z+ is illustrated in Fig. 6 with G = 16.3 kg/(m2∙s), Ts = 324 K, and Rea =
2,662. Corresponding to first cells nearest to cooling wall in computational domain of
Fig. 1, z+ is defined as:
z    u  z / . (25)

The friction velocity uτ is evaluated as:


u   y / , (26)

15
Original version

where τy denotes the wall shear in y orientation. These low values (z+ < 2.3) in Fig. 6
indicate that the fluid-flow and heat-transfer physics near the cooling wall were well
resolved. The values shown were representative of z+ used to achieve convergence for
all the operating conditions given in Section 4.2.

4.2.1 Length of condensation

The variations of length of complete condensation (LCC) with increases of G and


Rea are illustrated in Figs. 7(a) and 7(b), respectively, predicted by the analytical mode.
As shown in Fig. 7(a), LCC increased monotonously from 1,650 to 2,100 mm with G
increase from 8.1 to 16.3 kg/(m2∙s). Fig. 7(b) indicates an overall decrease of LCC from
2,250 to 1,750 mm with Rea raising from 1,340 to 12,379. Thus, it was indicates that a
large Rea effectively enhanced the mass-transfer level in upflow condensation.
However, the mass performance could not be deemed to be worsened by G increase
because of the overall mass of vapor entering tube, which depended on G, also
increased with LCC.

Rea = 2,662
Ts = 324 K, Ps = 13 kPa (b)
y (mm)

y (mm)

(a) G = 17.2 kg/(m2∙s)


Ts = 324 K, Ps = 13 kPa
Lcc predicted by analytical model
Lcc predicted by analytical model

8.1 9.8 11.4 13.0 14.6 16.3 1,340 2,662 5,021 7,508 9,978 12,379
G (kg/m2∙s) Rea (−)
Fig. 7 Length of complete condensation varying with (a) increasing vapor-inlet mass flux from 8.1
to 16.3 kg/(m2∙s) and (b) increasing cooling air flow Reynolds number from 1,340 to 12,379.
4.2.2 Interfacial behaviors

The local δ and τ on three parts of flat tube varying with axial length y as G
increased from 8.1 to 16.3 kg/(m2∙s) are illustrated in Figs. 8(a) and 8(b), respectively.
As indicated in Fig. 8(a), δUSO and δDSO lied in the range of 10–30 μm, as nearly twice
large as δPL in 5–15 μm, which was contributed to the relatively larger values of τPL
comparing with τUSO and τDSO, as illustrated in Fig. 8(b). Moreover, δUSO, δPL, and δDSO
exhibited a reasonable fluctuation in the whole condensation section. On the contrary,
τUSO, τPL, and τDSO displayed a monotonous decline with the condensation approaching
LCC.

16
Original version

Fig. 8 Predicted local film thickness (left) and interfacial shear (right) with increasing vapor-inlet
mass flux from 8.1 to 16.3 kg/(m2∙s).

δ G (kg/[m2∙s]) 8.1
9.8
(a) 8.1 (b)
Cross-sectional δ (μm)

Cross-sectional τ (N/m)

9.8 11.4
11.4 13.0
14.6
16.3

Rea = 2,662 13.0


Ts = 324 K, Ps = 13 kPa 14.6
16.3

y (mm) y (mm)
Fig. 9 Predicted cross-sectional values of (a) film thickness and (b) interfacial shear stress, with
increasing vapor-inlet mass flux from 8.1 to 16.3 kg/(m2∙s).
To better illustrate the effects of G on interfacial behaviors, the values of δUSO,
δPL, and δDSO in Fig. 8 were averaged to one cross-sectional value and replotted in Fig.
9. The similar average and plot were carried out to τUSO, τPL, and τDSO. As illustrated in
Fig. 9(a), the δ lied in the range of 10−20 μm, meanwhile displayed an insensitivity of
G, since the effect of G on δ was unfounded in the figure. On the contrary, based on Fig.
9(b), the τ was relatively dependent to G because of a high G resulting in a large τ; the τ
raised by approximately 4 times as G increased from 8.1 to 16.3 kg/(m2∙s).
The local δ and τ on the three parts of the flat tube varying with y as Rea increased
from 1,340 to 12,379 are illustrated in Figs. 10(a) and 10(b), respectively. According to
the results of Fig. 10(a), the gap among δUSO, δPL, and δDSO decreased with increase of
Rea, which was contributed to the decreasing gap among τUSO, τPL, and τDSO with
increase of Rea, as illustrated in Fig. 10(b). Similar to the results in Fig. 8(a), δUSO, δPL,

17
Original version

and δDSO in Fig. 10(a) exhibited a fluctuation in the whole condensation section;
whereas, τUSO, τPL, and τDSO displayed a monotonous decline with the condensation
approaching.

Fig. 10 Predicted local film thickness (left) and interfacial shear (right) with increasing cooling air
flow Reynolds number from 1,340 to 12,379.

τ Rea (−)
δ Rea (−) G = 17.2 kg(m2∙s)
1,340
1,340 Ts = 324 K
Cross-sectional δ (μm)

2,662
Cross-sectional τ (N/m)

2,662 5,021
5,021 7,508
9,978
12,379

7,508
9,978 (b)
12,379 (a)

y (mm) y (mm)
Fig. 11 Predicted cross-sectional values of (a) film thickness and (b) interfacial shear stress, with
increasing cooling air flow Reynolds number from 1,340 to 12,379.
Fig. 11 is plotted using the cross-sectional values of data in Fig. 10. Based on the
Fig. 11(a), the δ lied in the range of 5−25 μm. A non-negligible difference of δ with
change of Rea was observed, nevertheless, determining whether δ was upsized or
downsized by Rea was very difficult from the figure. On the contrary, the τ exhibits an
obvious sensitivity of Rea in Fig. 11(b) since the τ was reduced effectively by the
increased Re, especially when Re was larger than 5,021.
Several early literature (e.g., Berrichon et al. [1]) related to reflux condenser tubes
with large hydraulic diameter and relatively low inlet-vapor flow rates, indicated that

18
Original version

the δ varied in a relatively gentle trend and the value was much greater (in the range of
50−150 μm) than the results in Figs. 8−11. However, this study suggested that the
thickness of condensate in upflow finned-tubes fluctuated frequently in the whole
condensation section and its value lied in the range of 5−30 μm. Thus, the precision of
measuring instrument should to achieve the magnitude of 1 μm for the basic accuracy
regarding to the in-tube experiments.

4.3 Visualization and analysis


The reflux condensation in upflow finned-tube is distinguished through two
characteristics, namely, complete condensation and countercurrent flow. The former
leads to a vapor-absence in the upper section of tube and subsequent a
low-temperature zone, and the latter contributes to liquid entrainment, film climbing,
and flooding. To capture these instantaneous behaviors, a transient simulation is
essential although in an overwhelming time cost. In the study, to balance the
numerical accuracy and computational efficiency, the transient simulation was
conducted in an upflow tube with a same cross-section but a reduced length (L = 1,000
mm) relative to those in Fig. 3.
Following the approach presented in Section. 2, Fig. 12 shows the variation of
wall z+ along the condensing length with G = 3.2 kg/(m2∙s), Ts = 324 K, and Rea =
5,021. These low values (z+ < 0.7) indicate that near-wall fluid-flow and heat-transfer
physics were well resolved. The values shown were representative of wall z+ values
used to achieve convergence for all the operating conditions given in Section 4.3.

G = 3.2 kg/(m2∙s)
Ts = 324 K, Ps = 13 kPa
Wall z+ (−)

Rea = 5,021

y (mm)
Fig. 12 Variation of wall z+ along condensation length for visualization case.

19
Original version

4.3.1 Complete condensation

The complete condensation was simulated as G varying from an initial value of


3.2 kg/(m2∙s), increased to 5.4 kg/(m2∙s), decreased to 3.2 kg/(m2∙s), and then finally
reduced to 1.6 kg/(m2∙s) in a duration of t = 0.018–0.042 s. Since Fluent® prohibits
any void computational domain in any simulation case with the VOF method [21], a
vapor–air mixture in an air fraction value of 1% was introduced to tube regardless of
whether the air affected the in-tube thermal performance. In the simulation, the Rea
and Ts remained constant as 5,021 and 324 K, respectively.
The instantaneous contours of liquid void fraction on tube wall are plotted in Fig.
13, where individual images in the sequence are 0.0020 s apart. From these figures, the
lower half section of tube wall was fully covered by condensate film, whereas the
upper half section was film-void because vapor had been totally condensed in the
lower section. Moreover, it was founded that the increased G considerably extended
the range of liquid-phase (LCC), whereas the decreased G shortened it, guided by the
red auxiliary line to link the upward peak of the condensate among the 11 contours.
The animation of Fig. 13 was presented as an attachment to the paper (Animation
01.avi) for a more clear visualization to dynamic formation of condensate with change
of G.

Fig. 13 Computed sequential images of liquid void fraction on plate of flat tube during t =
0.0237–0.0402 s.
The instantaneous mass fraction of vapor species on symmetry plane of flat tube
is illustrated in Fig. 14, in which individual images in the sequence are 0.0020 s apart.
According to the figure, the vapor was completely condensed in the lower half tube,
subsequently, an obvious vapor-absence zone existed in the upper section, where the
latent heat of vaporization was no longer available. This might bring a low-temperature

20
Original version

zone to upflow finned-tubes in a cold weather, possibly taking tubes a high risk of
freezing because the temperature of zone depended on the atmosphere temperature and
cooling air flow rate instead of the coupled heat. Furthermore, Fig. 14 demonstrates that
the range of vapor-absence was more sensitive to change of G than that of LCC
illustrated in Fig. 13 since a high G considerably narrowed the vapor-absence zone
whereas a low G led to expansion. Similarly, the animation of Fig. 14 was presented as
an attachment to the paper (Animation 02.avi) for a more clear visualization for the
dynamic vapor-absence zone with change of G.

Fig. 14 Computed sequential images of air mass fraction on symmetry plane of flat tube during t =
0.0237–0.0402 s.
Combined with the results from Fig. 7(b), which illustrated the decrease of Rea
effectively raised the value of LCC, the results of Figs. 13 and 14 suggested that the
axial fan of upflow air-cooling condenser should run at a very low rotation speed, for
the purpose to shorten the low-temperature zone on cooling wall and to
prevent finned-tubes freezing when a turbine operated in a low exhaust flux at a cold
winter.

4.3.2 Countercurrent flow

Interfacial waviness, film climbing, and liquid entrainment are inherent features
of countercurrent condensation [17]. Moreover, the experiments by Mouza et al. [45]
and Zapke et al. [46] suggested that the flooding velocity was strongly dependent on the
fluid physical properties, and geometries and inclination angle of the flow channel.
Focusing on visualizing the two-phase dynamic behavior involved these features
rather than flooding velocity, two series of transient simulations to countercurrent
through-flow condensation were conducted under the following conditions. (1) Case
A: G increasing from 5.4 kg/(m2∙s) to 6.7 kg/(m2∙s); fixed Ts = 333 K, Xa = 0.1 %, and

21
Original version

a uniform temperature wall of ∆Tcool =10 K in the duration of t = 1.7720–3.7873 s. (2)


Case B: G increasing from 18.3 kg/(m2∙s) to 29.4 kg/(m2∙s); fixed Ts = 340 K, Xa =
0 %, and a varying temperature wall at Rea = 5,021 in the duration of t =
0.0100–0.0436 s. The former case was conducted to investigate the dynamic process
of liquid flooding and associated interfacial waviness, while the latter featured film
climbing and liquid entrainment.
Fig. 15 shows the instantaneous formation of condensate on tube wall during t =
1.7720–2.852 s of Case A, where individual images in the sequence are 0.0848 s apart.
Two interfacial behaviors caused by the countercurrent can be summarized. First, the
reflux interfacial shear overwhelmed the gravity and surface tension in the liquid,
which damaged the original integrity of liquid and dragged the film moving
longitudinally and transversely simultaneously. As a result, numerous voids were
formed in the interface, which finally generated evident waves similar to a rivulet in
the film’s interface. Second, the reflux shear elongated the liquid ligaments from a
narrow area in the middle stream. Then, the ligaments simultaneously flooded toward
inlet and outlet ends of tube, this trend was marked by two black auxiliary lines
linking the upper and lower summits of the condensate film among the 14 contours in
Fig. 15.

Fig. 15 Computed sequential images of liquid void fraction on plate of flat tube during t = 1.7720
to 2.8152 s in Case A.
The interfacial waves are discussed quantitatively in Fig. 16 by comparing the
three-time-enlargement features in rectangular area within the 1st, 4th, 9th, and 13th
in Fig. 15 and the experimental photos reprinted from Figs. 5(b)–5(d) of Berrichon et
al. [1]. Although the experimental device was a round vertical tube with a varying G,

22
Original version

comparing to the numerical flat tube inclined a 30° angle with a varying time but a fixed
G, the interfacial details on two tube walls were in good agreement. Fig. 16(a) shows
that the liquid droplet rose at the bottom of experimental and numerical tubes,
detaching from the condensate under reflux shear. In Fig. 16(b), evident waves
appeared on the film surface primarily visible at the bottom of experimental tube
compared with the upper numerical tube. Meanwhile, the film thickness grew under
reflux shear, which corresponded to the post-flooding accompanied with droplet at
tubes’ upper side. Fig. 16(c) illustrates the occurrence of flooding in the major parts of
both tubes, with the liquid moving upward with the increasing experimental air
velocity and the passing of transient simulation time, respectively; meanwhile, wave
amplitude evidently increased due to the interfacial shear. Fig. 16(d) exhibits a fully
developed flooding in both tubes, in which, the cyclic wave and rivulet formation
were present in the entire range of experimental and numerical tubes. In this state, the
liquid film spilled over the upmost end of tube due to the extremely small gravity
effect compared with the reflux shear.

Fig. 16 Contours of computed liquid void fraction in Case A compared with experimental photos
by Berrichon [1] at (a) t = 1.7720 s and Ug = 4.3 m/s, (b) t = 2.0282 s and Ug = 7.6 m/s, (c) t =
2.2726 s and Ug = 10.3 m/s, and (d) t = 2.5167 s and Ug = 13.1 m/s.
For a more clear visualization of the film behaviors, an animation of the
full-scale liquid void fraction contour during t = 1.7720–3.7873 s in Case A is attached
to the paper as an AVI file (Animation 03.avi).
23
Original version

Fig. 17 shows the instantaneous film climbing on the upper USO in


eight-time-enlargement sequential contour plots during t = 0.0158–0.0318 s in Case B,
in which individual images in the sequence are 0.0020 s apart. As highlighted within
the yellow rectangles and ovals, small liquid masses first detached from the
condensate by reflux vapor. Then, the film climbed upward along USO, accompanied
with an evident mass increase due to the climb grabbing the liquid from the
condensate surface until the tube outlet, where nearly all climbing film spilled over,
only a few droplets managed to escape and fell back into the condensation tube.

Fig. 17 Computed sequential void fraction contour on upper section of symmetry plane of USO of
flat tube during t = 0.01581–0.3181 s.
The climbing film is quantitatively explained in Figs. 18(a) and 18(b) depicting
the liquid void fraction and cross-sectional velocity in rectangles within the 7th in Fig.
17. From the figures, the Y-velocity of liquid near the wall remained positive in the
range of 1–10 m/s, corresponding to the climbing film. The Y-velocity of liquid was
larger than the X-velocity, but both velocities of liquid were lesser by two orders of
magnitude than those of vapor in core area, resulting in a steep boundary layer
between the vapor and condensate. Both of Z-velocity of the vapor and liquid were
low at approximately 0.1 m/s, thus, the dynamic behavior of two-phase was nearly
negligible in this direction.

24
Original version

Fig. 18 Features of semi-transparent green rectangle in Fig. 15 of (a) liquid void fraction contour
and (b) variation of vapor–liquid velocity across the section of y = 988 mm in Fig. 16(a).
Fig. 19 shows the instantaneous liquid entrainment on upper tube in
eight-time-enlargement sequential contour plots during t = 0.0354–0.0402 s in Case B,
in which individual images in the sequence are 0.0006 s apart. As highlighted within
the white rectangles, firstly, a relatively large liquid mass attempted to change shape
from its integral formation from the 1st and 2nd images; afterward, the formation of
entire film was torn and the droplet formed, as show in the 3rd and 4th images; then,
the droplet was separated from the entire formation and jumped into the vapor region
in the 5th–7th images; finally, the droplet overflowed to the void zone outside the
computational domain, as illustrated in the 8th and 9th images.

Fig. 19 Computed sequential void fraction contour on upper section of DSO symmetry plane of
flat tube during t = 0.0354–0.0402 s.

25
Original version

Fig. 20 Features of semi-transparent green rectangle in Fig. 17 of (a) liquid void fraction contour
and (b) variation of vapor–liquid velocity across the section of y = 988 mm in Fig. 18(a).
Similarly, the liquid entrainment is quantitatively explained in the combination of
Figs. 20(a) and 20(b). From the figures, the X- and Y-velocity of the droplet lied in the
range of positive 10–100 m/s near the core vapor, indicating that the droplet did not
reach the summit and still was in an upward motion. The velocity of vapor between
the droplet and climbing film on DSO lied in the range of 1–10 m/s, which much less
than that of vapor in the core region, which was due to flow block by the liquid
subsequently weakening the heat-mass-transfer level in the device. The Z-velocity of
vapor and liquid near the droplet raised to approximately 1 m/s, which indicated that
the dynamic behavior of the two-phase was slightly unstable in this direction.
In summary, Figs. 15−20 illustrated the through-flow condensation mode in
upflow finned-tube leaded a spillover of condensate because of the interfacial rivulet,
climbing film and liquid entrainment, which brought up the quality loss of working
fluid in Rankine cycle and contributed to deteriorate the heat-mass-transfer
perfomance. Thus, the through-flow condensation mode should be prevented in
operating an upflow air-cooling condenser as far as possible.
For a more clear visualization of climbing film and liquid entrainment, an
animation of four-time-enlargement contour in upper flat tube during t =
0.0158–0.0318 s in Case B is attached to the paper as an AVI file (Animation 04.avi).

5 Conclusion
A numerical simulation was conducted to study the length of complete

26
Original version

condensation LCC, condensate thickness δ and interfacial shear τ during vapor


condensation in an upflow flat tube of air-cooling condenser, considering effects of
vapor mass flux G and cooling air Reynolds number Rea. In addition, a visualization of
two-phase flow involving complete and countercurrent condensation was investigated.
The key findings and recommendations were as follows.

(1) The LCC was extended by almost 400 mm as G increases from 8.1 to 16.3
kg/(m2∙s), and shortened by almost 500 mm as Rea increases from 1,340 to
12,379. The δ exhibited an obvious fluctuation in the range of 5−25 μm from
the vapor inlet to LCC, meanwhile presented a strong insensitivity to both
changes of G and Rea. The τ hugely depended on Rea and G because of a high
G and a low Rea resulting in a large τ.

(2) The distribution of film on tube wall corresponding to complete condensation


was captured as a full cover on the lower tube and void on the upper tube;
meanwhile, the vapor-absence was evident in the upper tube. The ranges of
condensate film and vapor-absence zones were hugely sensitive with
fluctuating G.

(3) The interfacial waviness was predicted to a rivulet corresponding to


through-flow condensation mode, which was in a good agreement with
experimental images. The instantaneous film climbing and liquid entrainment
were demonstrated and further explained by the velocity profile of two-phase
flow.

(4) Recommendations regarding on experiment and operation. The precision to


measure δ should achieve 1 μm for the basic experimental accuracy. The axial
fan of upflow air-cooling condenser should run at a very low speed to prevent
finned-tubes freezing at cold winters. Upflow condensers should avoid
running in through-flow condensation mode.

Acknowledgment
We acknowledge the financial support of the Special Science and Technology
Program of Shanxi Province in China (No. MD2016-02) and the National Key
Research and Development Program of China (No. 2017YFB0902100).

27
Original version

Nomenclature
Nomenclature X air inlet mass fraction
A area Y species mass fraction
DH hydraulic diameter Greek symbols
E energy α volume fraction
gravitational force β inclination angle
G mass flow rate κ thermal conductivity
h heat-transfer coefficient λ latent heat of vaporization
I turbulence intensity μ dynamic viscosity
diffusion mass flux ν kinematic viscosity
k turbulent kinetic energy ρ density
L length interfacical shear
Nu Nusselt number τ interfacical shear magnitude
P pressure ω energy dissipation rate
q heat-transfer flux Subscripts
Q phase-change heat c condensation
Q source term in energy equation d diffusion
r phase change coefficient eff effective
Re Reynolds number f liquid film
S source term in continuity equations k turbulent kinetic energy
T temperature L liquid
velocity vector O overall
u,v,w velocity magnitude s saturated
U vapor inlet velocity magnitude TP two-phase
x,y,z local descartes coordinate V vapor

Reference
[1] Berrichon J D, Louahlia-Gualous H, Bandelier P, et al. Local heat transfer during
reflux condensation at subatmospheric pressure and with and without
non-condensable gases for power plant application[J]. International
Communications in Heat and Mass Transfer, 2016, 76: 117-126.
[2] Kumar A, Joshi J B, Nayak A K, et al. 3D CFD simulations of air cooled
condenr-III: Thermal–hydraulic characteristics and design optimization under
forced convection conditions[J]. International Journal of Heat and Mass Transfer,
2016, 93: 1227-1247.
[3] Xu C, Yang L, Li L, et al. Experimental study on heat transfer performance
improvement of wavy finned flat tube[J]. Applied Thermal Engineering, 2015, 85:
80-88.
[4] Du X, Feng L, Li L, et al. Heat transfer enhancement of wavy finned flat tube by
punched longitudinal vortex generators[J]. International Journal of Heat and Mass
Transfer, 2014, 75: 368-380.
[5] Li L, Du X, Yang L, et al. Numerical simulation on flow and heat transfer of fin
structure in air-cooled heat exchanger[J]. Applied Thermal Engineering, 2013,

28
Original version

59(1): 77-86.
[6] Owen M, Kröger D G. A numerical investigation of vapor flow in large air-cooled
condensers[J]. Applied Thermal Engineering, 2017, 127: 157-164.
[7] O'Donovan A, Grimes R. Pressure drop analysis of steam condensation in
air-cooled circular tube bundles[J]. Applied Thermal Engineering, 2015, 87:
106-116.
[8] Kekaula K, Chen Y, Ma T, et al. Numerical investigation of condensation in
inclined tube air-cooled condensers[J]. Applied Thermal Engineering, 2017, 118:
418-429.
[9] W. Nusselt, Die Oberflachenkondesation des Wasserdamffes, Zetrschr. Ver.
Deutch. Ing. 60 (1916) 541–546.
[10] Fiedler S, Auracher H. Experimental and theoretical investigation of reflux
condensation in an inclined small diameter tube[J]. International Journal of Heat
and Mass Transfer, 2004, 47(19): 4031-4043.
[11] Berrichon J D, Louahlia-Gualous H, Bandelier P, et al. Experimental study of
flooding phenomenon in a power plant reflux air-cooled condenser[J]. Applied
Thermal Engineering, 2015, 79: 214-224.
[12] O'Donovan A, Grimes R. A theoretical and experimental investigation into the
thermodynamic performance of a 50 MW power plant with a novel modular
air-cooled condenser[J]. Applied Thermal Engineering, 2014, 71(1): 119-129.
[13] Kang Y, Davies III W A, Hrnjak P, et al. Effect of inclination on pressure drop and
flow regimes in large flattened-tube steam condensers[J]. Applied Thermal
Engineering, 2017, 123: 498-513.
[14] Mahvi A J, Rattner A S, Lin J, et al. Challenges in Predicting Steam-Side Pressure
Drop and Heat Transfer in Air-Cooled Power Plant Condensers[J]. Applied
Thermal Engineering, 2018, 133: 396-406.
[15] Hirt C W, Nichols B D. Volume of fluid (VOF) method for the dynamics of free
boundaries[J]. Journal of Computational Physics, 1981, 39(1): 201-225.
[16] W.H. Lee, A pressure iteration scheme for two-phase flow modeling, in: T.N.
Veziroglu (Ed.), Multiphase Transport Fundamentals, Reactor Safety,
Applications, vol. 1, Hemisphere Publishing, Washington, DC, 1980.
[17] Kharangate C R, Lee H, Park I, et al. Experimental and computational
investigation of vertical upflow condensation in a circular tube[J]. International
Journal of Heat and Mass Transfer, 2016, 95: 249-263.
[18] Qiu G D, Cai W H, Wu Z Y, et al. Numerical simulation of forced convective
condensation of propane in a spiral tube[J]. Journal of Heat Transfer, 2015, 137(4):
041502.
[19] Qiu G, Wu Z, Jiang Y, et al. Numerical Simulation of Condensation of Upward
Flow in a Vertical Pipe[C]//ASME 2014 4th Joint US-European Fluids
Engineering Division Summer Meeting collocated with the ASME 2014 12th

29
Original version

International Conference on Nanochannels, Microchannels, and Minichannels.


American Society of Mechanical Engineers, 2014:
V01DT32A003-V01DT32A003.
[20] Cui X, Li X, Sui H, et al. Computational fluid dynamics simulations of direct
contact heat and mass transfer of a multicomponent two-phase film flow in an
inclined channel at sub-atmospheric pressure[J]. International Journal of Heat and
Mass Transfer, 2012, 55(21): 5808-5818.
[21] Fiorentino M, Starace G. Numerical investigations on two-phase flow modes in
evaporative condensers[J]. Applied Thermal Engineering, 2016, 94: 777-785.
[22] Fiorentino M, Starace G. Experimental investigations on air side heat and mass
transfer phenomena in evaporative condensers[J]. International Journal of Heat
and Technology, 2017, 35: S399-S404.
[23] ANSYS Inc.. ANSYS FLUENT 14.0 Theory Guide[M]. Canonsburg, USA,
2011.
[24] Chapman S, Cowling T G. The mathematical theory of non-uniform gases: an
account of the kinetic theory of viscosity, thermal conduction, and diffusion in
gases[M]. Cambridge: Cambridge University Press, 1990
[25] Menter F R, Kuntz M, Langtry R. Ten years of industrial experience with the SST
turbulence model[J]. Turbulence, Heat and Mass Transfer, 2003, 4(1): 625-632.
[26] Nikou M R K, Ehsani M R. Turbulence models application on CFD simulation of
hydrodynamics, heat and mass transfer in a structured packing[J]. International
Communications in Heat and Mass Transfer, 2008, 35(9): 1211-1219.
[27] Kong Y, Wang W, Huang X, et al. Circularly arranged air-cooled condensers to
restrain adverse wind effects[J]. Applied Thermal Engineering, 2017, 124:
202-223.
[28] ANSYS Inc.. ANSYS ICEM CFD 12.1 Help Manual[M]. Canonsburg, USA,
2009.
[29] Ge Z, Du X, Yang L, et al. Performance monitoring of direct air-cooled power
generating unit with infrared thermography[J]. Applied Thermal Engineering,
2011, 31(4): 418-424.
[30] Solomon C, Breckon T. Fundamentals of Digital Image Processing: A practical
approach with examples in Matlab[M]. John Wiley & Sons, 2011.
[31] Wolfram Inc.. Wolfram Mathematica 9 Manual [M]. Champaign, USA, 2009.
[32] ANSYS Inc.. ANSYS FLUENT UDF 14.0 Manual[M]. Canonsburg, USA, 2011.
[33] Patankar S V, Spalding D B. A calculation procedure for heat, mass and
momentum transfer in three-dimensional parabolic flows[J]. International Journal
of Heat and Mass Transfer, 1972, 15(10): 1787-1806.
[34] Ham F, Mattsson K, Iaccarino G. Accurate and stable finite volume operators for
unstructured flow solvers[J]. Annual Research Briefs, Center for Turbulence
Research, Stanford University/NASA Ames, 2006: 243-261.

30
Original version

[35] Leonard B P. A stable and accurate convective modelling procedure based on


quadratic upstream interpolation[J]. Computer Methods in Applied Mechanics
and Engineering, 1979, 19(1): 59-98.
[36] Ubbink O. Numerical prediction of two fluid systems with sharp interfaces[D].
University of London, 1997.
[37] Patankar S V. Numerical Heat Transfer and Fluid Flow[M]. Hemisphere
Publishing, New York: 1980.
[38] Shah M M. Comprehensive correlations for heat transfer during condensation in
conventional and mini/micro channels in all orientations[J]. International journal
of refrigeration, 2016, 67: 22-41.
[39] Lee H, Kharangate C R, Mascarenhas N, et al. Experimental and computational
investigation of vertical downflow condensation[J]. International Journal of Heat
and Mass Transfer, 2015, 85: 865-879.
[40] Wen J, Gu X, Wang S, et al. Numerical investigation on condensation heat transfer
and pressure drop characteristics of R134a in horizontal flattened tubes[J].
International Journal of Refrigeration, 2018, 85: 441-461.
[41] Da Riva E, Del Col D, Garimella S V, et al. The importance of turbulence during
condensation in a horizontal circular minichannel[J]. International Journal of Heat
and Mass Transfer, 2012, 55(13): 3470-3481.
[42] Deng H, Modelling and numerical calculation for heat and mass transfer of
single-row finned tube, Ph.D. Thesis, North China Electrical Power University,
2016 (in Chinese).
[43] Munson B R, Young D F, Okiishi T H, et al. Fundamentals of Fluid Mechanics[M],
6th Edition SI Version, 2013.
[44] Li J D. CFD simulation of water vapour condensation in the presence of
non-condensable gas in vertical cylindrical condensers[J]. International journal of
heat and mass transfer, 2013, 57(2): 708-721.
[45] Mouza A A, Paras S V, Karabelas A J. Incipient flooding in inclined tubes of small
diameter[J]. International Journal of Multiphase Flow, 2003, 29(9):1395-1412.
[46] Zapke A, Kröger D G. The influence of fluid properties and inlet geometry on
flooding in vertical and inclined tubes[J]. International Journal of Multiphase
Flow, 1996, 22(3):461-472.

31

You might also like