You are on page 1of 9

FLOW AND HEAT TRANSFER IN NARROW

CHANNELS WITH CORRUGATED WALLS


A CFD Code Application
A. G. KANARIS, A. A. MOUZA and S. V. PARAS

Department of Chemical Engineering, Aristotle University of Thessaloniki, Thessaloniki, Greece


I
n an effort to obtain information on the local ow structure inside compact heat
exchangers made of corrugated plates, a commercial CFD code (CFX
w
) is employed
to simulate the ow through an element of this type of equipment. For simplicity, the
channel used for the simulation is formed by only one corrugated plate, while the other
plate is at. A two-equation turbulence model (SST) is used for the calculations and, in
addition to isothermal ow, heat transfer simulations are conducted for a Reynolds number
range (4001400), for the case of hot water (608C) in contact with a constant-temperature
wall (208C). The results, presented in terms of friction factor, wall shear stress, wall heat
ux and local Nusselt numbers, are consistent with the description of the uid motion
inside similar conduits by other investigators. The calculated mean heat transfer coefcients
and friction factors are found to be in reasonable agreement with the limited published
experimental data.
Keywords: compact heat exchanger; narrow channel; corrugation; CFD; Nusselt number;
pressure drop.
INTRODUCTION
The development of compact heat exchangers has been
mainly driven by the need for economical, high performance,
yet small in size and light weight equipment. Novel compact
heat exchangers made of corrugated plates hold signicant
advantages over conventional equipment. Plate exchangers
with corrugated walls provide a large surface area to
volume ratio and enhanced heat transfer coefcients, while
allowing ease of inspection and cleaning (Kays and
London, 1998; Shah and Wanniarachchi, 1991). Such types
of exchangers, like the herringbone or the chevron type, are
being rapidly adapted by food, chemical and refrigeration
process industries replacing shell-and-tube exchangers.
Unfortunately, unlike the conventional heat exchangers,
there is lack of a generalized design method for plate heat
exchangers. Variations in design of the basic geometrical
features (e.g., aspect ratio, inclination angle of the corruga-
tions) make it almost impossible to generate an adequate
heat transfer database covering all possible congurations.
The heat transfer augmentation in conduits with corru-
gated walls is accompanied by a substantial increase in
pressure drop. An optimum design must involve a balance
between friction losses and heat transfer rates and thus the
designer must decide how to trade off between these two
factors. Nevertheless, the requirement for detailed and
accurate measurement of the design parameters (e.g.,
temperature, pressure and velocity elds) is very difcult
to be achieved, because the ow passages in compact
heat exchangers are complex in geometry and of relatively
small dimensions. The rapid development of computational
tools permits the prediction of ow characteristics using
CFD code simulation which is considered an effective
tool to estimate momentum and heat transfer rates in this
type of process equipment. Consequently, as CFD is
more widely used in engineering design, it is becoming
of essential importance to know how reliably the ow
features and the hydrothermal behaviour can be reproduced
in such conduits.
Kays and London (1998) state that the applicable range
of Reynolds numbers for compact heat exchangers is
between 500 and 15 000. In addition, when these equip-
ments are used as reux condensers the limit imposed by
the onset of ooding reduces the maximum Reynolds
number to a value less than 2000 (Paras et al., 2001).
The type of ow prevailing in such narrow passages is
still an open issue. Shah and Wanniarachchi (1991) declare
that, for the Reynolds number range 1001500, there is
evidence that the ow is turbulent. Heggs et al. (1997)
suggest that pure laminar ow does not exist in the
Reynolds number range they tested (15011,500) and sup-
port their conclusion by studying local transfer coefcients.
Ciofalo et al. (1998), in a comprehensive review article

Correspondence to: Professor S. V. Paras, Department of Chemical


Engineering, Aristotle University of Thessaloniki, Univ. Box 455, GR
54124 Thessaloniki, Greece.
E-mail: paras@cheng.auth.gr
460
02638762/05/$30.00+0.00
# 2005 Institution of Chemical Engineers
www.icheme.org/journals Trans IChemE, Part A, May 2005
doi: 10.1205/cherd.04162 Chemical Engineering Research and Design, 83(A5): 460468
concerning modelling heat transfer in narrow ow
passages, state that, in the Reynolds number range of
15003000, transitional ow is expected, a kind of ow
among the most difcult to simulate using conventional
turbulence models. Recently, Vlasogiannis et al. (2002),
who experimentally tested a plate heat exchanger under
single and two-phase ow conditions, verify that the ow
is turbulent for Re . 650. Lioumbas et al. (2002), who
studied experimentally the ow in narrow passages
during counter-current gasliquid ow, suggest that the
ow exhibits the basic features of turbulent ow even
for the relatively low gas Reynolds numbers tested
(500 , Re , 1200). Focke and Knibbe (1986) performed
ow visualization in narrow passages with corrugated
walls using an electrode-activated pH reaction. They con-
cluded that the ow patterns in such geometries are very
complex and suggested that the local ow structure controls
the heat transfer process in the narrow passages. The salient
feature of the ow is the existence of secondary swirling
motions along the furrows of their test section.
The choice of the most appropriate turbulence model
for CFD simulation is another open issue in the literature.
The most common two-equation model, based on the
equations for the turbulence energy k and its dissipation
1, is the k1 model (Davidson, 2001). Ciofalo et al.
(1998) state that the standard k1 model using wall
functions overpredicts both wall shear stress and wall
heat ux, especially for the lower range of the Reynolds
number encountered in this kind of equipment. Menter
and Esch (2001) note that the overprediction of heat
transfer is caused by the overprediction of turbulent
length scale in the region of ow reattachment, which is
a characteristic phenomenon appearing on the corrugated
surfaces in these geometries.
An alternative to the k1 model is the kv model devel-
oped by Wilcox (1988). The kv model, which uses the
turbulence frequency v in place of turbulence dissipation
1, appears to be more robust, even for complex appli-
cations, and does not require very ne grid near the wall
(Davidson, 2001). The main disadvantage of kv model
is its sensitivity to the free stream values of turbulence fre-
quency v outside the boundary layer, which affects the
solution and, in order to avoid this, a combination of the
two models, k1 and kv, i.e., the SST (Shear-Stress
Transport) model is proposed (Menter and Esch, 2001).
The SST model can switch automatically between the
two aforementioned turbulence models using specic
blending functions that activate the kv model near the
wall and the k1 model for the rest of the ow. Although
the SST model combines the most widely used
two-equation turbulence-models, other models, like LES
(Large-Eddy Simulation) is considered more appropriate
in turbulent ow simulation. However, the LES model is
considered less robust and requires high-computational
power (Ciofalo et al., 1998).
Due to the modular nature of a compact heat exchanger,
a common practice for computational expense reasons is to
think of it as composed of a large number of unit cells
(RES, Representative Elementary Unit). The results are
obtained using a single cell as the computational domain
and imposing periodicity conditions across its boundaries
(e.g., Ciofalo et al., 1998; Mehrabian and Poulter, 2000;
Blomerius and Mitra, 2000). However, since the validity
of this assumption is not generally accepted in the literature
(Ciofalo et al., 1998), an alternative method is to consider
the complete corrugated plate as the computational domain,
but this results in an increase of both computational space
and time.
In a previous work in this Laboratory (Paras et al., 2001)
the ow in a vertical channel of a model plate heat
exchanger was studied. This model, manufactured by
VICARB-Alfalaval, comprises of two plates having
corrugations machined at a 458 angle and two side-channels
[Figure 1(a)]. The experiments revealed that, during
counter-current gasliquid ow (which resembles the ow
when this equipment is used as a condenser), the side
channels play a signicant role in the liquid ow through
the furrows of the corrugations, promoting even distri-
bution. The lateral drainage into the side-channels tends to
increase with increasing gas ow rate, leading to a progress-
ive elimination of the liquid lm. This situation, referred to
as maldistribution, is considered favourable for the oper-
ation of such a device as a condenser, because of the
exposure of nearly fresh wall to the condensing vapours.
In this paper, CFD modelling is employed to investigate
the ow and thermal characteristics within the complicated
passages of a plate heat exchanger as described above.
The approach used here is based on a more complex geo-
metry consisting of a whole channel instead of a single
cell. More specically, the solution domain employed,
due to computational power limitations, is a simplication
of the real conduit and comprises of one corrugated and one
at plate, adjacent to each other. Nevertheless, the results
from this simplied geometry can be used to study the
basic features of the ow inside narrow channels with cor-
rugated walls and to validate the CFD code. Experimental
results on overall pressure drop, obtained from a Plexiglas
w
test section of the same geometry, are compared to the
calculated values. Information on local heat transfer coef-
cients is also obtained and validated with data from the
literature, in order to quantitatively evaluate the thermal
performance of a corrugated-plate compact heat exchanger
with side-channels. In addition, the ow pattern prevailing
inside the furrows and the side-channels of the conduit,
which affects the local momentum and heat transfer rates
of this type of equipment, is predicted.
MODEL PARAMETERS AND SOLUTION
PROCEDURE
The geometry studied in the present simulations is
consistent with an existing compact heat exchanger
described in detail elsewhere (Paras et al., 2001). The
Plexiglas
w
text section is formed by two plates, 70 cm
high and 15 cm wide, which simulates a vertical channel
of a corrugated plate heat exchanger. On each plate, corru-
gations are machined at a 458 angle, as well as side-chan-
nels [Figure 1(a)]. The two plates were superposed so that
the opposite corrugations formed a cross-type pattern with
the crests of the corrugations nearly in contact. However,
in order to keep the computational demands at acceptable
levels, a simpler channel is studied. This channel is formed
by only one of the corrugated plates [Figure 1(b)], which is
comprised of fourteen equal sized and uniformly spaced
corrugations and two side-channels (Figure 2), while the
second plate is at. Details of the corrugated plate
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
FLOW AND HEAT TRANSFER IN NARROW CHANNELS 461
geometry are presented in Figure 2 and Table 1. The
simpler case of single-phase ow of water is investigated
here.
The Reynolds number is dened at the conduit
entrance as:
Re
u d r
m
(1)
where u is the uid velocity dened as Q/A, Q is the volu-
metric ow rate, A is the ow cross section at the entrance,
d is the distance between the plates at the conduit entrance
(d 10 mm), while r and m are the density and the
viscosity of the uid respectively, at entrance conditions,
calculated by the CFD code. The Reynolds numbers exam-
ined (400, 900, 1000, 1150, 1250, 1400) lay at the lower
part of the compact heat exchanger operability range and
correspond to the working conditions of reux condensers.
It must be noted that the denition of Reynolds number is
an open issue in literature, as the geometry of these devices
do not allow a unique calculation method.
In this study, in addition to isothermal ow, heat transfer
simulations are carried out for the same Reynolds numbers,
where hot water (608C) is cooled in contact with a constant-
temperature wall (208C). The latter case is realized in
condensers and evaporators. Additionally, it is assumed
Figure 1. Model plate heat exchanger: (a) view of two superposed corrugated plates; (b) detail of a single plate with side channels.
Figure 2. Geometry of the CFD model and sectional view of a corrugation.
Table 1. Plate geometric characteristics.
Plate length 0.200 m
Plate width 0.110 m
Maximum spacing between plates, d 0.010 m
Number of corrugations 14
Corrugation angle 458
Corrugation pitch, h 0.005 m
Corrugation width, w 0.014 m
Plate length before and after corrugations 0.050 m
Channel (groove) width 0.005 m
Heat transfer area 2.7 10
22
m
2
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
462 KANARIS et al.
that heat is transferred only through the corrugated plate,
while the rest of the walls are considered adiabatic.
A commercial CFD code, namely CFX
w
version 5.6,
developed by AEA Technology, was employed to explore
its potential for computing detailed characteristics of this
kind of ow. The code uses a nite volume approach,
and the simulation is performed in steady state. The grid
size used is chosen by performing a grid dependence
study, since the accuracy of the solution depends on
the number and the size of the cells (Versteeg and
Malalasekera, 1995). The grid dependence study was
conducted by altering the cell size inside the channel, and
by rening the grid size on the corrugated wall.
To overcome computer power limitation, a compromise
was made and the SST model was preferred to LES
model in the calculations, as mentioned previously. The
mean velocity of the liquid phase was applied as boundary
condition at the channel entrance (i.e., Dirichlet boundary
condition on the inlet velocity) and no-slip conditions on
the channel walls. A constant temperature boundary
condition was applied only at the wall of the plate with
the corrugations (including the at entry and exit sections),
whereas the rest of the plate walls are considered adiabatic.
The choice of constant temperature on the wall was
selected for simplicity, by assuming that on the other side
of the wall, the ow rate of the cool uid is high enough
to ensure this condition.
Calculations were performed on a SGI O
2
R10000 work-
station with a 195 MHz processor and 448 Mb RAM. The
CFX
w
5.6 code uses a nite volume method on a non-
orthogonal body-tted multi-block grid. Unstructured
tetrahedral mesh was used, modied near the walls by
applying prism layers, in order to simulate the wall bound-
ary layer accurately. The use of prism layers on the walls is
advised for conned geometries (CFX
w
Manual, 2003).
The mesh was also checked for inappropriate generated
cells (e.g., tetrahedral cells with sharp angles) and xed
and the nal number of cell elements was 870,000. In the
present calculations, the CFD code uses a method similar
to that used by Rhie and Chow (CFX
w
Manual, 2003)
with the SIMPLEC algorithm for pressure-velocity decou-
pling and the QUICK scheme for discretisation of the
momentum equations (Versteeg and Malalasekera, 1995;
CFX
w
Manual, 2003). The grid was constructed using
CFX
w
-Build 5.6 and ICEM CFD 4.CFX, while CFX
w
-
Post was used for post-processing. The normalized mass
residual, i.e., the measure of the local imbalance of each
conservative volume equation (CFX
w
Manual, 2003) is
used by the CFD code as the convergence criterion and
its value was set to be less than 10
28
.
RESULTS AND DISCUSSION
The results of the present study conrm the dominant
role of the side-channels in ow distribution and suggest
that uid ow is mainly directed to the right side-channel
of this model plate (Figure 3). Part of the uid ows over
the corrugation crests and after being reected on the
right side wall, follows the furrows and reaches the oppo-
site side-channel. It appears that if two corrugated plates
with angles 458 and 2458 were superposed (as in a real
heat exchanger) part of the uid phase would also be
directed to the right channel, creating a symmetrical overall
ow distribution. Experiments performed in this Labora-
tory (Paras et al., 2001) suggest that the above ow distri-
bution promotes the drainage of the liquid phase through
the side-channels in counter-current two-phase ow.
This type of ow behaviour is also described by Focke
et al. (1985), who made visual observations of the ow
between two superposed corrugated plates without side
channels. They conrm that the uid, after entering a
furrow, mostly follows it until it reaches the side wall,
where it is reected and enters the anti-symmetrical
furrow of the plate above, a behaviour similar to the one
predicted by the CFD simulation. More specically, the
velocity inside the left side-channel progressively increases
[Figure 4(a)], while that in the right side-channel decreases
[Figure 4(b)]. It seems that most of the ow passes through
the furrows, where enhanced heat transfer characteristics
are expected, a fact that is also reported by Heggs et al.
(1997).
Figure 5 shows details of the ow inside a furrow, where
secondary, swirling ow is identied. It is suggested
(Heggs et al., 1997) that this kind of secondary ow is
the result of interaction between the ow inside the
narrow channel and the highly accelerated ow over the
crest. This ow is considered capable of bringing new
uid from the main stream close to the walls, augmenting
heat transfer rates. Focke and Knibbe (1986) describe
also this kind of swirling ow. Blomerius and Mitra
(2000) have also observed longitudinal vortices in narrow
channels, while Won et al. (2003) consider this secondary
ow responsible for the increase in turbulence shear
stress and turbulence production.
A typical distribution of the z-component of shear
stress is presented in Figure 6 for Re 1400, since the
Figure 3. Typical ow pattern (streamlines) inside the channel, predicted
by CFD; Re 900.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
FLOW AND HEAT TRANSFER IN NARROW CHANNELS 463
distribution is similar for all Re numbers. Shear stress
increases with Reynolds number, as expected, and it attains
its maximum value at the crests of the corrugations. It may
be argued here that, during gasliquid counter-current ow
in such geometries, this shear stress distribution tends to
prevent the liquid layer from falling over the crest of the
corrugations and to keep it inside the furrows. The visual
observations of Paras et al. (2001) seem to conrm the
above behaviour.
Wall heat ux through the corrugated plate was predicted
by the CFD code [Figure 7(a)]. The local Nusselt number,
Nu
x
, was then calculated (by means of a Fortran subroutine)
using the expression:
Nu
x

q d
(T
b
T
w
)k
(2)
where q is the local wall heat ux, d the distance between
the plates at entrance, T
w
the wall temperature, T
b
Figure 4. Velocity vectors inside channels: (a) left-side channel; (b) right-side channel.
Figure 5. Swirling ow inside a furrow; Re 900.
Figure 6. Wall y-shear stress distribution on the corrugated plate;
Re 1400.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
464 KANARIS et al.
the local uid temperature and k the thermal
conductivity of the uid. The local uid temperature
over a point of the plate wall, T
b
, is the average uid
temperature calculated by numerical integration of the
uid temperature across a line vertical to the corrugated
wall at this point. In addition to the local Nusselt
number various modied Nusselt numbers were evalu-
ated as follows:
. a mean Nusselt number, Nu
c
, calculated by numerically
integrating the local Nu over the corrugated area only;
. an overall average Nusselt number, Nu
ave
, calculated by
numerically integrating the local Nu over the whole
plate;
and their values are presented in Table 2. These Nusselt
numbers were calculated in CFX
w
-Post using a function
that computes the average value of a variable by taking
into account the mesh element sizes. It must be noted
that in an effort to validate the CFD code predictions,
values of Nu
ave
for the smooth plate were computed by
CFX
w
and are found to be in accordance with analytically
obtained results. These non-dimensional quantities are also
calculated in order to study the effect of non-corrugated
area to the whole heat transfer augmentation.
Figure 7 shows typical wall heat ux and local Nusselt
number distributions over the corrugated wall for
Re 1400. The distributions of heat ux and Nu are
similar for all Reynolds numbers studied. It is noticeable
that on the top of the corrugations the local Nusselt
number attains its maximum value. Heggs et al. (1997)
also notice that the mass and heat transfer performance exhi-
bits peaks on the crests of the corrugations. This conrms the
strong effect of the corrugations, not only on the ow distri-
bution, but also on the heat transfer results. Ligrani and Oli-
veira (2003) also denote that vortices and secondary ows in
general contributes in heat transfer augmentation, as they
increase secondary advection of uid between the central
parts of the channel and the near-wall region, and provide
regions of high turbulence production. Secondary ows
also decrease the probability of appearance of stagnation
areas that promote heat and mass transfer.
To the best of the authors knowledge, experimental
measurements of heat transfer and pressure drop for the
corrugated plate geometry under consideration are not
available in the open literature. In an effort to validate
the simulation results, data by Heavner et al. (1993),
Vlasogiannis et al. (2002) and Ligrani and Oliveira
(2003) are used. The data by Vlasogiannis et al. (2002)
concern measurements of the heat transfer coefcients
both for single (Re , 1200) and two-phase ow in a
plate heat exchanger with two corrugated walls and a
corrugation inclination angle of 608. It should be also
noted that heat exchangers used in Vlasogiannis et al.
(2002) and Heavner et al. (1993) experiments lack the
side channels employed in the present simulation.
The results presented by Ligrani and Oliveira (2003) con-
cern geometries with rib turbulators in 458 continuous
arrangements.
In heat exchanger analysis j-Colburn factor, commonly
used to express the heat coefcients, is dened as (Blomerius
and Mitra, 2000):
j
Nu
Re Pr
1=3
(3)
Figure 7. Heat transfer results for Re 1400: (a) wall heat ux distri-
bution on the corrugated plate; (b) local Nu distribution on the corrugated
plate.
Table 2. Nusselt number: calculated and experimental data (Vlasogiannis
et al., 2002).
Re Nu
ave
Nu
c
Nu
vlasog
65% Nu
vlasog
Nu
sm
Nu/Nu
sm
400 20.7 20.5 13.2 8.6
900 27.5 27.3 38.0 24.7 9.4 2.9
1000 28.8 28.6 41.2 26.8 10.2 2.8
1150 30.0 28.8 44.2 28.7 11.0 2.7
1250 31.1 30.9 46.8 30.4 11.7 2.7
1400 32.2 32.0 49.5 32.2 12.5 2.6
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
FLOW AND HEAT TRANSFER IN NARROW CHANNELS 465
In Figure 8 the j-Colburn factor values, calculated using
Nu
c
, are compared with experimental values for various
Reynolds numbers. Focke et al. (1985), who measured
heat transfer coefcients in a corrugated plate heat
exchanger by placing a partition of celluloid sheet
between the two plates, report that the overall heat trans-
fer rate is reduced to 65% of the value for the plates with-
out the partition. This statement was taken into account, in
an effort to compare the CFD results of this study with the
experimental data by Vlasogiannis et al. (2002), which are
acquired in similar geometries but with two corrugated
plates. Thus, the results are found to be in good agreement
with the 65% of the corresponding experimental values.
The exponent of Re in the resulting correlation has a
value 0.42 which agrees with that proposed in the litera-
ture. The above remark holds true for all Reynolds num-
bers except for the smallest one (Re 400). In the latter
case the Nusselt number, and consequently the j-factor, is
greatly overpredicted by the CFD code. This is not unex-
pected, since a two-equation turbulence model is not
capable of correctly predicting the heat transfer character-
istics for such low Reynolds numbers.
Since the Nusselt numbers for the corrugated area (Nu
c
)
and the overall average Nu
ave
(Table 2) have practically
the same value, it can be concluded that the presence
of the smooth part of the plate does not signicantly inu-
ence the heat transfer coefcient. Consequently, the pre-
sence of the side-channels, whose area is only a small
percentage of the total plate area, apart from inhibiting
ooding (Paras et al., 2001), seems to have practically no
effect on the thermal behaviour of the plate in single
phase ow.
In an effort to validate the CFX
w
code, the CFD predic-
tions were checked against the corresponding values for the
smooth wall plate calculated by an analytical method
(White, 1991) and found to be in excellent agreement, as
expected. The pressure drop for the corrugated plate was
also predicted by the CFD code, converted into terms of
friction factor and compared with experimental data
collected in this Laboratory (Table 3). Figure 9 presents
the friction factor experimental data and CFD predictions
for the corrugated plate as a function of the Reynolds
number. It appears that the experimental values follow a
power law of the form:
f mRe
n
(4)
where m and n are constants with values 0.27 and 0.14,
respectively. Heavner et al. (1993) propose a similar
empirical correlation based on their experimental results
on a plate heat exchanger with 458 corrugation angle, but
with two corrugated plates and without side channels. It
must be noted that, in spite of the differences in geometry,
the slope of the correlation derived from the present data
has the same value with the one proposed by Heavner
et al. (1993).
Figure 10 presents the normalized values of Nu
ave
and f
obtained both from the present study (Tables 2 and 3) and
the work by Taslim and Wadsworth, as referred by Ligrani
and Oliveira (2003), concerning a 458 continuous rib
arrangement. Values of Nu
ave
and f are normalized by the
corresponding values from a smooth plate heat exchanger,
in order to evaluate the heat transfer augmentation versus
the friction losses increase due to the existence of corruga-
tions. The CFD results are once more in agreement with the
experimental results taking under consideration that the rib
arrangement setup, justies an increase in heat transfer,
which is attributed to the existence of ribs on both sides
of the plate heat exchanger.
Table 3. Friction factor: calculated and corresponding experimental data.
Re f
exp
f
cfx
f
sm
f/f
sm
900 0.1044 0.0915 0.0074 12.4
1000 0.1025 0.0886 0.0069 12.8
1150 0.1020 0.0866 0.0064 13.5
1250 0.0984 0.0850 0.0061 13.9
1400 0.0981 0.0839 0.0058 14.5
Figure 9. Friction factor vs. Reynolds number.
Figure 8. j-Colburn factor vs. Reynolds number.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
466 KANARIS et al.
CONCLUDING REMARKS
The present work examines the ability of a commercial
CFD code to predict the ow and heat transfer character-
istics in a narrow channel with corrugated wall, with a
certain corrugation angle, width and height. The use of a
CFD code allows computation for various geometrical
congurations, in order to evaluate their effects and to
study them closely. In this way the engineer is able to opti-
mize the efciency (i.e., maximize the ratio of heat transfer
to friction losses) for a given geometry.
The simulation results reveal that, compared to a
smooth-wall plate heat exchanger, corrugations improve
both ow distribution and heat transfer. The comparison
of Nusselt number values for the simplied model shows
that the side-channels, besides improving the operability
of the heat exchanger by shifting the ooding limit to
higher gas velocities when used as reux condensers
(Paras et al., 2001), do not affect the overall heat transfer
augmentation negatively. Pressure drop, a variable that
could also trigger interest on economically optimizing
a plate heat exchanger, has, as expected, higher values
compared to a smooth plate channel.
Additional experimental work is needed, and indeed is in
progress in this laboratory, to obtain more data in order to
validate the results of the present work. More specically,
local velocity prole determination will clarify the type of
ow prevailing in such geometries, while local temperature
measurements will allowthe prediction of heat transfer rates.
NOMENCLATURE
A cross section area at the entrance of the channel
d distance between the plates at the entrance of the channel,
equation (1)
f friction factor
k uid thermal conductivity
m constant
n constant
Nu
x
local Nusselt number
Nu
c
mean Nu calculated by numerical integration over the
corrugated area
Nu
ave
average Nu calculated using the total wall heat ux through the
whole corrugated plate
Nu
sm
average Nu calculated using the total wall heat ux through the
whole smooth plate
Pr Prandtl number
Q volumetric ow rate
q local wall heat ux
Re Reynolds number
T
b
local uid temperature
T
w
wall temperature
u mean velocity at channel entrance
Greek letters
1 turbulence energy dissipation
m viscosity
r density
f corrugation angle
REFERENCES
Blomerius, H. and Mitra, N.K., 2000, Numerical investigation of convec-
tive heat transfer and pressure drop in wavy ducts, Numerical Heat
Transfer, Part A, 37: 3754.
CFX
w
Release 5.6 User Guide, 2003, AEA Technology, CFX Inter-
national, Harwell, Didcot, UK.
Ciofalo, M., Collins, M.W. and Stasiek, J.A., 1998, Flow and heat transfer
predictions in ow passages of air preheaters: assessment of alternative
modeling approaches, in Sunden, B. and Faghri, M. (eds). Computer
Simulations in Compact Heat Exchangers (Computational Mechanics
Publ., UK).
Davidson, L., 2001, An Introduction to Turbulence Models, Department of
Thermo and Fluid Dynamics, Chalmers University of Technology,
Publication 97/2 (Goteborg, Sweden).
Focke, W.W. and Knibbe, P.G., 1986, Flow visualization in parallel-plate
ducts with corrugated walls, J Fluid Mech, 165: 7377.
Focke, W.W., Zachariades, J. and Olivier, I., 1985, The effect of the
corrugation inclination angle on the thermohydraulic performance of
plate heat exchangers, Int J Heat Mass Transfer, 28: 14691497.
Heavner, R.L., Kumar, H. and Wanniarachchi, A.S., 1993, Performance
of an industrial plate heat exchanger: effect of chevron angle, AIChE
Symp Ser No 295, Vol 89, Heat Transfer, Am Inst Chem Eng, Atlanta,
GA, 262267.
Heggs, P.J., Sandham, P., Hallam, R.A., Walton, C., 1997, Local transfer
coefcients in corrugated plate heat exchangers channels, Trans
IChemE, Part A, Chem Eng Res Des, 75(A7): 641645.
Kays, W.M. and London, A.L., 1998, Compact Heat Exchangers, 3rd
edition (Krieger Publ. Co., Florida, USA).
Ligrani, P.M. and Oliveira, M.M., 2003, Comparison of heat transfer
augmentation techniques, AIAA Journal, 41(3): 337362.
Lioumbas, I.S., Mouza, A.A. and Paras, S.V., 2002, Local velocities inside
the gas phase in counter current two-phase ow in a narrow
vertical channel, Trans IChemE, Part A, Chem Eng Res Des, 80(6):
667673.
Mehrabian, M.A. and Poulter, R., 2000, Hydrodynamics and thermal
characteristics of corrugated channels: computational approach, Applied
Mathematical Modeling, 24: 343364.
Menter, F. and Esch, T., 2001, Elements of industrial heat transfer predic-
tions, 16th Brazilian Congress of Mechanical Engineering (COBEM),
2630 Nov. 2001, Uberlandia. Brazil.
Paras, S.V., Drosos, E.I.P., Karabelas, A.J. and Chopard, F., 2001,
Counter-current gas/liquid ow through channels with corrugated
wallsvisual observations of liquid distribution and ooding, World
Conference on Experimental Heat Transfer, Fluid Mechanics &
Thermodynamics, September 2428, Thessaloniki, Greece.
Shah, R.K. and Wanniarachchi, A.S., 1991, Plate heat exchanger design
theory, in Buchlin, J.M. (ed.). Industrial Heat Exchangers, von
Karman Institute Lecture Series 19912004.
Versteeg, H.K. and Malalasekera, W., 1995, An Introduction to Compu-
tational Fluid Dynamics (Longman Press, London, UK).
Vlasogiannis, P., Karagiannis, G., Argyropoulos, P. and Bontozoglou, V.,
2002, Airwater two-phase ow and heat transfer in a plate heat
exchanger, Int J Multiphase Flow, 28(5): 757772.
Figure 10. Globally averaged Nusselt number ratio vs. friction factor ratio.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
FLOW AND HEAT TRANSFER IN NARROW CHANNELS 467
White, F.M., 1991, Viscous Fluid Flow, 2nd edition (McGraw-Hill Inc.,
New York, USA).
Wilcox, D., 1988, Reassessment of the scale-determining equation, AIAA
Journal, 26(11): 12991310.
Won, S.Y., Mahmood, G.I. and Ligrani, P.M., 2003, Flow structure and
local Nusselt number variations in a channel with angled crossed-rib
turbulators, Int J Heat Mass Transfer, 46: 31533166.
ACKNOWLEDGEMENTS
The authors wish to thank Professor A.J. Karabelas for his helpful
comments and suggestions and Mr A. Lekkas for the technical support.
The manuscript was received 2 June 2004 and accepted for publication
after revision 17 February 2005.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A5): 460468
468 KANARIS et al.

You might also like