International Journal of Thermal Sciences 67 (2013) 152e166

Contents lists available at SciVerse ScienceDirect

International Journal of Thermal Sciences

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

Numerical study of fully developed flow and heat transfer in a wavy

passage
Abhishek G. Ramgadia, Arun K. Saha*
Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur, UP 208016, India

a r t i c l e i n f o a b s t r a c t

Article history: A two-dimensional numerical study of fully developed fluid flow and heat transfer through a horizontal
Received 6 January 2012 wavy surface is presented. Time dependent NaviereStokes and energy equation have been solved using
Received in revised form finite volume method. Numerical calculations are performed for a wavy surface described by sine
28 August 2012
function y ¼ 2a  sin2 (px/L) with ratios Hmin/Hmax and L/a fixed to 0.4 and 8 respectively. Attempts have
Accepted 6 December 2012
been made to foresee the effect of streamwise domain length, an integer multiple of periodic domain
Available online 1 February 2013
length, on the flow and heat transfer characteristics. It is established that the flow and heat transfer
characteristics do not show any dependence on the length of the periodic domain thereby confirming
Keywords:
Wavy channel
that geometric and flow periodicity are identical. The effect of Reynolds numbers, for Re in the range 25
Heat transfer e1000, on the flow field and heat transfer has been presented. For steady flow the heat transfer rates are
Fluid flow found to be very low and for unsteady flow with increased mixing between core and near wall fluids
Finite volume method enhanced heat transfer rates are obtained.
Collocated grid Ó 2012 Elsevier Masson SAS. All rights reserved.

1. Introduction that forms on the heat exchange surface. As mentioned earlier this
is generally achieved passively by the modification of the geometry
Enhancement of heat transfer is the process of improving the to generate large scale vortical structures. These structures carrying
thermal performance of heat exchanging devices. In the current era, heat are convected out of the heated surface by the main flow once
heat transfer enhancement has gained serious momentum due to they are away from the surface. Consequently, the fresh cold fluid
increased demands by industry for efficient heat exchange equip- replaces this large scale vortical region, thereby escalating the heat
ments as well as stringent environmental norms. As a result many transfer. Relevant examples are off-set fins, louvers, vortex gener-
efforts in the form of research have been made to produce heat ators, communicating channels, acoustic excitation of the flow,
exchangers that satisfy the ecological norms. One of the frequently oscillatory inflow, wavy fins, etc. The heat transfer enhancement by
used methods to attain heat transfer enhancement is by geometric such a method is also accompanied by an undesirable increase in
modification prompting earlier transition to turbulence, creating the pressure drop resulting in higher friction factor.
vortices that increase mixing or restarting the thermal boundary A study of the relevant literature reveals a fairly large body of
layer to decrease its thickness. For this purpose, various heat literature dealing with fully developed flows in periodic geome-
transfer augmentation devices have been studied, both numerically tries. Heat and mass transfer for laminar, transitional and low
and experimentally, and are available in the open literature. A Reynolds number turbulent flows in the case of a wavy walled
special attention has also been focused to save materials and en- channel was first studied experimentally by Goldstein and Sparrow
ergy used in many industrial applications such as heat exchangers [1]. They showed that in the Reynolds number range of 1000e1200,
for air conditioning and refrigeration. It is imperative that they are the heat transfer was comparatively larger than in the case of
made as compact and light as possible. The same constraint is a parallel-plate channel. They also found that in the case of low
applicable for the cooling systems in automobiles and spacecrafts, Reynolds number turbulent flow range from 6000 to 8000, the heat
where volume and weight constraints are crucial. transfer was three times greater than for a parallel-plate channel.
The main objective of increasing heat transfer in the aforesaid For a channel with symmetric wavy walls, Nishimura et al. [2]
devices is achieved by interrupting efficiently the boundary layers experimentally studied mass transfer and flow patterns for range
of Reynolds number varied from 20 to 300. Two different wall ge-
ometries were considered: the sine shaped and the arc shaped.
* Corresponding author. Mass transfer rates, for both the geometries studied, for steady flow
E-mail address: [email protected] (A.K. Saha). was found to be negligible whereas it was reported that as the flow

1290-0729/$ e see front matter Ó 2012 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.ijthermalsci.2012.12.005
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 153

Nomenclature Pr Prandtl number, mCp/k

Re Reynolds number, ruavgHavg/m
Cf skin friction factor, sw =0:5ru2avg St Strouhal number, fr$Havg =uavg
Cp specific heat T dimensional temperature
f Darcy friction factor, ðDp=0:5ru2avg ÞðHavg =LÞ u and v dimensionless Cartesian velocity components in x and
fo friction factor for fully developed channel flow (based y directions respectively
on channel height), 24/Re x and y dimensionless distances along x-axis and y-axis.
fr frequency of vortex shedding
Hmin minimum height between two wavy walls Greek symbols
Hmax maximum height between two wavy walls b(t) mean periodic pressure gradient in the x-direction
Havg average height between two wavy walls, q non-dimensional temperature, ðT  Tw Þ=ðTb1  Tw Þ
(Hmin þ Hmax)/2 m dynamic viscosity
hx local heat transfer coefficient r mass density
L characteristic length of single period s cycle time
n wall normal direction uz z-component of vorticity, vu/vy  vv/vx
np number of modules in the computational domain
Nux local Nusselt number, List of subscripts
hx Havg =k ¼ ððvq=vnÞjn¼0 Þ=ðqw ðx; yÞ  qb ðxÞÞ min minimum value
Nu wall averaged Nusselt number, max maximum value
Z L avg average value
ð1=LÞ Nux ð1 þ y02 Þ1=2 dx w value at the wall
0
P periodic component of pressure b bulk value
p pressure 1 quantities evaluated at domain inlet

becomes unsteady, the mass transfer rates are high. The local heat and wavelength on the flow and heat transfer characteristics.
transfer and flow behaviour for a sinusoidal wavy channel was Decreasing the channel height and increasing the amplitude caused
investigated experimentally by Rush et al. [3]. It was found that at the flow to become more unstable thereby increasing both friction
low Reynolds number (Re w 200) instabilities were present near factor as well as heat transfer, with the variation of wavelength
the exit of the channel. But as the Reynolds number was increased, having the minimal effect. A numerical study of a sinusoidal wavy
the instabilities moved towards the entrance. At Re w 800, the plate channel with a constant wall temperature was done by Zhang
entire channel flow was unstable. They observed through experi- et al. [8]. With increasing values of Reynolds number, the lateral
ments that with increasing instability, there was an increase in the vortices formed in the trough regions, were found to increase in
heat transfer rate in the wavy channels. Nishimura et al. [4] magnitude. Also the separation between the plates was found to
experimentally investigated the flow and mass transfer character- affect the flow. For large separations, vortices were formed in the
istics in an axisymmetric sinusoidal wavy-walled tube in the Rey- valley region of the plates. Wen and Ho [9] experimentally inves-
nolds number range from 50 to 1000. It was observed that steady tigated three different fin configurations including wavy fins and
flow changed into unsteady one when the Reynolds number found the heat transfer enhancement for the wavy channel was
exceeded above 160. Combined instantaneous wall shear stress higher compared to other configurations.
measurements and flow visualizations showed that in the transi- Haitham et al. [10] conducted a numerical study on the heat
tional flow regime, the laminar-like motion and turbulent-like transfer along periodic wavy walls in compact heat exchangers.
motion alternatively appeared at different time intervals, signify- Both sine and arc-shaped channels were studied. With an increase
ing intermittent flow behaviour. in either the length ratio or the height ratio, the recirculation length
Tanda and Vittori [5] numerically studied heat transfer for and the pressure drop were found to decrease. Additionally, sine-
a laminar flow in the case of a 2D wavy wall. The per-cycle pressure shaped channels had a lower normalized pressure drop in com-
drop for a wavy wall was found to be larger than that for a straight parison with the arc-shaped one. It was pointed out that in the case
channel whereas the heat transfer coefficient was found to be of non-staggered wall, the recirculation size depended on the
larger than that for a straight channel. It was also reported that the Reynolds number. At higher Reynolds number the recirculation size
heat transfer coefficient had its maximum value near the crests and was found to be larger. But in the case of staggered walls, the dis-
the minimum value near the troughs of the wavy channel. Wang placement factor also played a role in the recirculation size. In the
and Chen [6] numerically analyzed the rates of heat transfer for numerical report of Mahmud et al. [11], they found that higher the
flow through a sinusoidally curved convergingediverging channel surface waviness the earlier the flow separation, larger the pressure
using a simple coordinate transformation applied to the spline drop and higher the heat transfer rate.
alternating-direction implicit method. The effects of wavy geome- Most of the aforesaid numerical studies used multiple sine
try, Reynolds number and Prandtl number on the skin-friction co- waves to attain the fully developed state and reported the heat and
efficient and Nusselt number were considered. The obtained results mass transfer results attained in the fully developed regime.
illustrated that the amplitudes of the Nusselt number and the skin- Patankar et al. [12] first proposed a numerical method where the
friction coefficient increases with increase in Reynolds number and geometric periodicity can be utilized to reduce the length of the
amplitudeewavelength ratio. computational domain thereby reducing the computational cost.
Hossain and Islam [7] in their numerical report found that the They studied the heat transfer and flow characteristics for the case
flow was steady up to a critical Reynolds number which in turn of a heat exchanger having successive segments of isothermal
depended on the geometric configuration. Beyond a critical Rey- plates placed transverse to the direction of the fluid flow and the
nolds number, a self-sustained oscillatory flow was observed. They dependence of the Nusselt number on Reynolds number was
also reported the effect of variation of minimum height, amplitude reported. Wang and Vanka [13] numerically simulated fully
154 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

developed flow through periodic wavy convergingediverging simulate the flow physics. To the best of the authors’ knowledge,
channels to obtain the heat transfer rates. They observed that the there has not been any information reported whereby the effect of
flow was steady and laminar up to a Reynolds number of 180 and number of periodic modules considered in the domain on fluid flow
the heat transfer rates were very low. As Reynolds number was and heat transfer has been discussed.
increased beyond this value, the flow becomes more complex and
the heat transfer enhancement was showing a large increase. The 2. Governing equations
heat transfer coefficient was found to be enhanced by a factor of 2
compared to the case of a parallel-plate channel. But this was In the present study, the incompressible NaviereStokes equa-
accompanied by a higher friction factor. Niceno and Nobile [14] tions along with the energy equation in the primitive variables
numerically performed the analysis of fully developed fluid flow form are used to simulate the flow field and heat transfer. As
and heat transfer through periodic wavy and arc shaped channels. the domain is periodic in the streamwise direction, it necessitates
They found that the flow remained steady up to a critical Reynolds the mean pressure gradient to be isolated from the fluctuating
number and beyond a critical value at which transition to an un- component, which can be expressed as follows:
steady regime was observed, the heat transfer rate increased as
! !
a result of self-sustained oscillations. Recently, Xie et al. [15] Pð x ; tÞ ¼ pð x ; tÞ  bðtÞx (1)
numerically studied the fully developed laminar flow and heat
transfer characteristics inside a two-dimensional wavy channel where, b(t) is the linear component of the non-dimensional pres-
used in compact heat exchangers. Their results confirmed that the sure and is adjusted at each time step to achieve the desired mass
overall Nusselt number and friction factor increased with Reynolds flow rate. Unlike the studies by Guzmán and Amon [21], Wang and
numbers. Recently Sui et al. [16], Sui et al. [17] and Gong et al. [18] Vanka [13], and Niceno and Nobile [14] where they have used
reported numerical studies of fluid flow and heat transfer in wavy a constant pressure drop, the present study uses an unsteady
walled microchannels of rectangular cross-section. pressure drop. For a constant pressure drop, the term b(t) used in
One of the objectives of the present study is to evaluate the equation (3) does no longer remains a function of time. This causes
effect of Reynolds numbers, for Re in the range from 25 to 1000. The the velocity scale at inlet or any streamwise locations to fluctuate in
other objective is to confirm the number of periodic modules that time, which may not be justified to use to define the Reynolds
gives simultaneous flow and geometric periodicity. In the open number based on average velocity. Hence, the alternate velocity
literature, almost all the studies except by Wang and Vanka [13] scale (generally the frictional velocity), which results in an un-
and Niceno and Nobile [14] deal with developing flow and heat known average velocity is used. Hence, the Reynolds number based
transfer in wavy channels. However, the treatment of flow perio- on average velocity for the simulation is not known in advance and
dicity in Wang and Vanka [13] and Niceno and Nobile [14] has been can be obtained only during the simulation. Moreover, for any de-
done by considering the mean pressure gradient to be constant as vice, the inlet mass flow rate is generally specified which results in
opposed to the present study where it is considered as time a constant average velocity at inlet. Depending on the nature
dependent. There is a dearth of literature available on developed (steady or unsteady) of flow, the shear stress either remains con-
flow and heat transfer for such geometries. stant or varies in time. But the constant pressure drop method
In the present paper the geometry, as shown in Fig. 1, is gen- assumes a constant shear stress which is also not correct for un-
erated by using sine waves with a phase shift of 180 . This geometry steady simulations. The algorithm to calculate the pressure drop is
is easy to generate and can provide significant heat transfer aug- such that the mean pressure gradient (b(t)) is iterated to a value
mentation if operated within an appropriate Reynolds number that ensures the inlet velocity to be equal to the constant
range. It is a well-established fact that the geometric periodicity average velocity corresponding to the considered Reynolds number
does not guarantee flow periodicity, as shown by Saha and Acharya (Saha and Acharya, [22]). As a consequence, the mean pressure
[19] and Saha [20]. This is because of the fact that some of in- gradient is expected to vary if the flow is unsteady and results in
stabilities such as TollmieneSchlichting (TeS) waves in the flow unequal frictional forces from one time instant to another.
largely depend on the domain length. Most of the fluid flow and On substituting into the NaviereStokes and energy equations,
heat transfer studies were done using multiple numbers of waves the governing equations in dimensionless form are expressed as:
to generate the channel geometry whereas for fully developed flow
vu vv
cases, only one periodic module was taken into account. This pro- þ ¼ 0 (2)
vides the prime motivation to perform a study in order to predict vx vy
the effect of number of periodic modules taken into account in the !
computational domain on heat and mass transfer estimation. This vu vðuuÞ vðuvÞ vp 1 v2 u v2 u
may lead to a reduction in computational cost for high Reynolds þ þ ¼  þ þ þ bðtÞ (3)
vt vx vy vx Re vx2 vy2
number computations as a smaller domain size may be used to
!
vv vðuvÞ vðvvÞ vp 1 v2 v v2 v
þ þ ¼  þ þ (4)
vt vx vy vy Re vx2 vy2

!
vq vðuqÞ vðvqÞ 1 v2 q v2 q
þ Lq þ þ ¼ þ (5)
vt vx vy RePr vx2 vy2

The unknown function ‘L’ in the energy equation is given by

1 v
L ¼ ðT ðtÞ  Tw Þ (6)
Tb1  Tw vt b1
The coupling between q and L can be solved iteratively as
Fig. 1. Geometry of sinusoidal wavy channel. described by Wang and Vanka [13]. The governing equations are
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 155

non-dimensionalized by using the length scale Havg i.e. the average 5. Code validation
height of wavy channel, the velocity scale uavg, the pressure by ru2avg
and the time by Havg/uavg. Temperature is non-dimensionalized by The present code is validated by comparing the results obtained
using the scale q ¼ ðT  Tw Þ=ðTb1 ðtÞ  Tw Þ, where Tw is the wall at Reynolds numbers of 25, 100 and 400 with the numerical results
temperature and Tb1 is the time dependent bulk temperature of the of Haitham et al. [10]. The geometry studied by Haitham et al. [10]
fluid evaluated at the inlet of periodic domain. For the present study, had the following ratios Hmin/Hmax ¼ 0.3 and L/a ¼ 8 and a similar
the working fluid is chosen to be air for which the Prandtl number geometry was simulated for purpose of validating the code. The
is 0.7. comparison between the two sets of results, shown in Fig. 2, reveals
Finite volume method (FVM) on collocated grid is used to solve good agreement. The velocity profiles in Fig. 2 are plotted at centre
equations (2)e(5). To prevent the pressure velocity decoupling the of the computational domain where H ¼ Hmax. The present code has
Momentum Interpolation Method (MIM) originally proposed by Rhie also been tested for a fully developed laminar channel flow having
and Chow [23] has been used. Both convective and diffusion terms are constant wall temperature. The computed friction factor and Nus-
discretized using a central difference scheme. Time advancement is selt number are found to have excellent match with analytical so-
achieved semi-implicitly using first order backward-time method. lution at various Reynolds numbers. The comparison of friction
factor and wall averaged Nusselt number with the numerical data
of Haitham et al. [10], Wang and Vanka [13], Niceno and Nobile [14]
3. Boundary conditions and Nishimura et al. [24] listed in Table 1, shows good agreement.
The reported friction factor values by all the researchers seems to
No-slip boundary condition (u ¼ v ¼ 0) is imposed on all the be that of Darcy friction factor rather than Fanning friction factor
wavy walls. The periodic boundary condition generalized by though the definition of friction factor in their papers corresponds
equation (7) is used in the streamwise direction to Fanning friction factor. Wang and Vanka [13] and Niceno and
  Nobile [14] defined the Reynolds number as Re ¼ uav Havg =n, where
f x þ np Lx ; y; t ¼ fðx; y; tÞ (7) uav is the time-averaged mean velocity in the channel. The differ-
ence in results between the present study and those of Wang and
where, ɸ ¼ u, v or p, ‘L’ is the length of domain in the streamwise Vanka [13] and Niceno and Nobile [14] are due to the difference in
direction and ‘np’ represents the number of modules included in the the Reynolds numbers between the two data sets. The difference
computational domain. The corresponding periodic condition for may also be due to the developing flow and heat transfer in certain
the energy equation is studies and solution technique (constant pressure drop versus the
present method) in others. For the purpose of validation, Reynolds
qðx þ nLx ; y; tÞ qðx; y; tÞ
¼ (8) number values are converted as per definition used in the present
qb2 ðtÞ qb1 ðtÞ study and the obtained non-integer values are written in brackets.
The same code has also been validated thoroughly with the pub-
where, qb1(t) and qb2(t) are the bulk temperatures calculated at the
lished results of Casarsa and Arts [25] (flow field) and Wagner et al.
inlet and outlet of the computational domain, respectively. Constant
[26] (heat transfer) for rib roughened channel flow, the details of
wall temperature (qw ¼ 0) is imposed on all the confining surfaces.
which can be found in Ramgadia and Saha [27].

4. Flow configuration 6. Grid independence test

The geometry under consideration is shown in Fig. 1 where one To ensure the accuracy and validity of the numerical results,
periodic module, which is also varied to see the effect of number of a careful check for the grid independence of the numerical solu-
modules on the flow and heat transfer characteristics, is consid- tions has been made using five different grids for the geometry
ered. A single period of the sinusoidal wavy channel has been with ratios Hmin/Hmax ¼ 0.4 and L/a ¼ 8. A series of computations, at
depicted here. In this study, numerical calculations are performed Reynolds number of 800, are performed to ensure mesh-
for a wavy surface described by y ¼ 2a$sin2 ðpx=LÞ where, 2a is the
amplitude of the curve, x is the distance along the x-direction and L
is the characteristic length of the channel. The flow domain consists
of two wavy walls with amplitude of 2a, having a minimum and
maximum distance of Hmin and Hmax respectively between them. In
the present study, the ratios Hmin/Hmax and L=a are fixed to 0.4 and 8
respectively to generate the flow geometry. For steady state solu-
tions, the convergence criterion applied was 1  105.
The local Nusselt number and the wall averaged Nusselt number
are defined as:

hx Havg ðvq=vnÞn¼0
Nux ¼ ¼  (9)
k ðqw ðx; yÞ  qb ðxÞÞ

ZL  1=2
1
Nu ¼ Nux 1 þ y02 dx (10)
L
0

where,

4pa px px Fig. 2. Comparison of transverse variation streamwise velocity at an axial location
dy
y0 ¼ ¼ sin cos (11) where H ¼ Hmax with Haitham et al. [10] at Re ¼ 25, 100 and 400 (Hmin/Hmax ¼ 0.3 and
dx L L L L/a ¼ 8).
156 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

Table 1
Comparison of friction factor and wall averaged Nusselt number with reference data (Hmin/Hmax ¼ 0.3 and L/a ¼ 8).

Re Friction factor, f

Present Haitham et al. [10] Niceno and Nobile [14] (Re) Nishimura et al. [24] Wang and Vanka [13] (Re)
25 1.483 1.294 1.820 (28.3) e e
100 0.458 0.415 0.583 (109.3) 0.275 (125) e
200 0.264 e 0.368 (208.2) 0.157 (225) e
400 0.150 0.165 0.202 (435.4) 0.160 (415) e

Nusselt number, Nu
25 4.39 4.39 3.94 (28.3) e 4.07 (21.8)
100 4.27 4.59 4.04 (109.3) e 4.04 (109.3)
200 4.76 e 4.26 (208.2) e 4.04 (172.6)
400 5.31 5.77 4.99 (435.4) e 5.47 (478.0)

independent solutions. The details of the grid sizes studied are

given in Table 2. Fig. 3 shows the comparison of time-averaged
streamwise velocity component plotted along the mid-height of
the wavy channel. It is found that except for the coarsest grid size
with 150  63 (6615) nodes, no noticeable change in the stream-
wise velocity component is observed among other four grid sizes.
Comparing the values of wall averaged friction factor and Nusselt
number from Table 2, the variation is found to be 0% and 2.5%
respectively between intermediate grid with 13,500 nodes and the
finest grid with 22,815 nodes. Therefore all the simulations in the
present study have been done using intermediate grid with
150  90 (13,500) nodes.
In addition to the grid independence test, a test for independ-
ency of time-step (Dt) was also performed a priori. The test was
performed at Reynolds number of 800 and the grid size used was
150  90 (13,500) nodes. The details of the study are given in
Table 3. The friction factor remains constant at all the four values of
the time-steps. The variations in the Nusselt number values are Fig. 3. Comparison of time-averaged streamwise velocity component (u) for five
different grid sizes plotted along the mid-height of the wavy wall at Re ¼ 800.
4.3% and 1.6% respectively between the time-step values of 1  103
and 8  104 compared to Dt ¼ 6  104 . Based on the study the
time-step interval chosen for all the unsteady simulations in the
along the streamwise direction. As mentioned earlier, it has been
present study is 8  104.
observed that for such computational domains, sometimes the
geometrical periodicity does not guarantee flow periodicity and
7. Results and discussion multiple number of periodic length are required to achieve flow
periodicity. Guzmán and Amon [21] reported that the frequency of
The results obtained for the geometry studied having the ratios decay mode in a sinusoidal channel is close to the plane channel
Hmin/Hmax and L/a fixed respectively at 0.4 and 8 are reported in the mode having the same wave spatial structure. Saha [20] demon-
present section. First the results obtained to study the dependence strated in his three-dimensional simulation of turbulent flow
of selected length of the periodic domain are presented. The across an array of pin fins that the instantaneous flow and tem-
obtained fluid flow and heat transfer characteristics for the Rey- perature fields revealed differences in flow and geometric perio-
nolds number spanning the range from 25 to 1000 are discussed in dicity while time-averaged fields show both flow and geometric
the subsequent section. periodicity to be same. They further verified by them that the in-
stabilities caused thereby were related to the TeS waves or trav-
7.1. Domain independency test elling waves. The low frequency unsteadiness of the main stream
flow is responsible to drive the instabilities as well as the flow
Since the present investigation involves the study of fully unsteadiness within the wavy channel. A single periodic module
developed flow and heat transfer in a wavy channel, the flow in the with a domain length smaller than the wavelength of the aforesaid
long periodic geometry is modelled using a domain of smaller low frequency oscillations (Nyquist criterion), damps out these
streamwise length with periodic boundary conditions imposed oscillations and is thereby unable to appropriately reproduce the

Table 2
Grid independence study performed at Re ¼ 800. Table 3
Time-step independence study performed at Re ¼ 800.
Grid size (nodes) Friction factor f Nusselt number Nu
105  63 (6615) 0.080 10.497 Time-step Dt Friction factor f Nusselt number Nu
128  77 (9856) 0.080 10.543 4  103 0.080 10.179
150  90 (13,500) 0.080 10.933 1  103 0.080 10.643
173  104 (17,992) 0.080 11.172 8  104 0.080 10.933
195  117 (22,815) 0.080 11.211 6  104 0.080 11.105
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 157

flow field. In the present study, attempts have been made to

identify the effect of the streamwise domain length, which is an
integer multiple of geometric domain length, on the instantaneous
flow and heat transfer characteristics. Four different streamwise
lengths namely one, two, three, and four periodic lengths, also
known as pitch, are considered and the Reynolds number value is
fixed to 800.
Fig. 4 illustrates the instantaneous streamlines for four different
periodic modules plotted at the same phase. The streamlines are
visibly identical in all modules in the crest and the trough regions of
the channel for the instantaneous field. Comparing the isotherms,
not shown here, the instantaneous temperature field is also found
to be irrespective of the length per number of periodic geometries
considered. The time variation of wall averaged Nusselt number,
not shown here, for the four periodic modules also matched exactly
with each other revealing module independency on heat transfer
predictions.
Time-averaged streamline plots for four different periodic
modules, not shown here, showed formation of a recirculation
bubble in both the crest and the trough regions of the channel. The
formation of these bubbles facilitates mixing of the main stream
fluid with the near wall fluid thereby enhancing the heat transfer.
The streamlines patterns in the crest and the trough regions of the
channel for all four modules are found to be identical and the
pattern is similar to that obtained for single module geometry
shown in Fig. 14(c). It is also found from the time-averaged iso-
therms in Fig. 5 that the flow is also thermally independent of the Fig. 5. Time-averaged isotherms for four periodic modules at Re ¼ 800 plotted at the
same phase (a) one periodic module (b) two periodic module (c) three periodic module
number of periodic modules chosen. The values of maximum and
(d) four periodic module.
minimum magnitude of temperature being 1.2 in the bulk flow and
0.05 near the wall respectively, are similar for all the four periodic
modules. This confirms that particularly for this geometry at the of the velocity fields at selected locations is undertaken. Fig. 6(a)
studied Reynolds number range, the flow periodicity and geometric compares the variation of the time-averaged streamwise velocity
periodicity for both instantaneous and time-averaged flow field plotted along the streamwise and transverse directions respec-
match. tively. For the single module geometry the velocity variation is
So far the comparisons of the spatial distributions were made. plotted along the geometrical centreline of the domain whereas for
To have better confidence in our claim, the quantitative comparison multiple module geometries the results are plotted along the
geometrical centreline of the last module for the respective ge-
ometries. Here again the velocity profiles match immaculately with
each other thereby verifying that the single module geometry is
enough for Reynolds number range of 50e1000 to resolve the flow
field without suppressing the low frequency oscillations caused by
the TeS waves.
Table 4 depicts the comparison of the wall averaged friction
factor, Nusselt number and Strouhal number for the geometry
considered with four different periodic modules. From this table it
is evident that the wall averaged friction factor, Nusselt number and
Strouhal number per module remain fairly constant, thus signifying
that the existence of temperature fields are independent from the
domain length. The non-dimensional frequency also called Strouhal
number has been calculated by computing FFT of the velocity var-
iables. The FFT of time dependent Nusselt number also reveals
identical non-dimensional frequency. The variation of the time-
averaged Nusselt number and skin friction factor over the arc
length is plotted in Fig. 6(b), which are self-repetitive in each pe-
riodic length for multiple periodic modules. From the obtained
thermal patterns it is concluded that a single module of the channel
is appropriate to fully capture heat transfer for the chosen Reynolds
number range.

7.2. Effect of Reynolds number

Since it has been found that a single module of the wavy channel
Fig. 4. Instantaneous streamline for four periodic modules at Re ¼ 800 plotted at the
is sufficient to fully capture the fluid flow and heat transfer phe-
same phase (a) one periodic module (b) two periodic module (c) three periodic module nomena without suppressing the low frequency oscillations caused
(d) four periodic module. by the TeS waves, only one module of the wavy channel is used to
158 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

Fig. 6. Variation of time-averaged streamwise velocity variation, Nusselt number and skin friction factor along streamwise direction and transverse direction of the channel at
Re ¼ 800. (Lines without symbols in (b) represent Nusselt number and lines with symbol indicate skin friction factor variation).

investigate the effect of Reynolds number on the fluid flow and heat the flow remains steady and the streamlines are found to be
transfer. The grid size used is 150  90 (13,500) and the effect of Re symmetric even for Re ¼ 200. At Re ¼ 100, the centre of the recir-
is investigated in the range from 25 to 1000. It is quite obvious for culation region is located close to the wall of the geometry but with
any flow configuration to have both steady and unsteady flow fields the increase in Reynolds number to 200, it shifts towards the
depending on the Reynolds number. Long-time integration of the centreline. The streamline pattern shows a single trapped pair of
governing equations reveals that the flow becomes unsteady vortices in the cavities and almost perfectly straight flow in
beyond a specific Reynolds number. Fig. 7 presents the time var- the core. As a result, the warm core fluid passes straight through the
iation of the streamwise velocity near the top wall of the channel at centre of the passage, without interaction and any momentum
x ¼ 0.85 and y ¼ 0.5. The velocity is seen to be time-independent at transport with the cooler fluid near the walls.
Re ¼ 350, however, it shows undulation at Re ¼ 400. The signals Because of the flow is unsteady at Re ¼ 400 (see Fig. 7), the
stored at various locations within the domain also reveal identical symmetry is completely lost. The centre of the recirculation un-
behaviour. Therefore, the critical Reynolds number for the studied dulates continuously from the left to the right end of the concave
geometry is found to lie between Reynolds numbers of 350 and 400 area due to higher inertia of fluid trapped in this region. As shown
and it agrees well with the values reported by other researchers. in Fig. 8(e), at Re ¼ 600, the vortices grow alternatively in the upper
The corresponding values obtained in Wang and Vanka [13], Niceno and lower region of the wavy wall as it is shown later (see Fig. 12).
and Nobile [14], and Guzmán and Amon [21] are 391.3, 380.4 and The viscous forces are not strong enough to keep the vortices
375.6 respectively. confined in the furrows and the vortices are alternatively ejected
into the core fluid from each furrow. With further increase in
7.2.1. Instantaneous field Reynolds number to 800, the flow becomes highly complex. This
The instantaneous streamlines plots for the geometry at various causes higher momentum transport between the core fluid and the
Reynolds numbers are depicted in Fig. 8. At Re ¼ 25, the streamlines fluid in the furrows.
follow the wavy wall and are symmetric about the channel cen-
treline and no separation is observed. The flow is first observed to
bifurcate from the wall at Re ¼ 100 forming two separation bubbles,
one each in the crest and in the trough region respectively of the
channel. The two vortices formed are however steady due to the
low inertia of the fluid in the region and the symmetric nature of
flow is still retained about the centreline of the channel. As the
Reynolds number is increased further, the size of the recirculation
region increases and it shifts downstream towards the outlet of the
domain. These vortices become stronger and increase in size but

Table 4
Friction factor, time-averaged Nusselt number and Strouhal number for different
periodic modules at Re ¼ 800.

Re ¼ 800 Friction factor f Nusselt number Nu Strouhal number St

1-Periodic 0.080 10.933 0.305
2-Periodic 0.080 10.933 0.305
3-Periodic 0.080 10.824 0.305
4-Periodic 0.081 10.828 0.305
Fig. 7. Time variation of streamwise velocity at x ¼ 0.85 and y ¼ 0.5.
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 159

Fig. 8. Instantaneous streamlines at different Reynolds numbers (a) Re ¼ 25, (b) Re ¼ 100, (c) Re ¼ 200, (d) Re ¼ 400, (e) Re ¼ 600, (f) Re ¼ 800.

Instantaneous vorticity contours at different Reynolds numbers recirculating fluid in the furrows of the wavy passage. Because of
are shown in Fig. 9(aef). The dashed lines indicate negative vor- the breakage of the thermal boundary layer by the interaction be-
ticity whereas the continuous lines signify positive vorticity values. tween the core flow and the shear layer within the furrows, the
The upper and lower shear layers within the top and bottom fur- heat transfer increases. When the flow is steady, a minimal increase
rows are nearly symmetrical at low Reynolds numbers of 25, 100 in heat transfer rate is expected due to the isolation of cooler fluid
and 200. At Re ¼ 400, the instability causes the vortices to become in the two furrows from the main streamwise flow stream. For the
unstable and to roll up in the crest and trough of the wavy wall. steady flow at lower Reynolds number (Re < 400), the thermal
Intensity of the vortices increases with the Reynolds number. With boundary layer along the curved wall is thicker and consequently
an increase in Reynolds number the flow pattern becomes complex the temperature gradient is higher compared to its counterpart in
and vortices move up and down in the crest and trough of the wavy the unsteady regime. It is evident from Fig. 10 that the thermal
passage. Because of this intense interaction between the core flow boundary layer becomes thinner with increasing Reynolds number.
and the flow within the furrows, the momentum transport be- Also the waviness in the temperature field appears once the flow
tween the cold fluids near the wall with the hot fluid from the core becomes unsteady. Fig. 11 shows the temporal variation and the
occurs. corresponding power spectra of the surface averaged Nusselt
Fig. 10 illustrates instantaneous isotherms at different Reynolds number for Re ¼ 400, 600 and 800. The temporal variation of
numbers. As mentioned earlier, heat transfer enhancement in the Nusselt number shows asymmetric behaviour as it involves addi-
wavy passage is due to the mixing of the core fluid with tional frequencies due to secondary structures in the flow field. The
160 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

Fig. 9. Instantaneous vorticity contours at different Reynolds numbers (a) Re ¼ 25, (b) Re ¼ 100, (c) Re ¼ 200, (d) Re ¼ 400, (e) Re ¼ 600, (f) Re ¼ 800.

heat transfer augmentation obtained at Reynolds number of 800 is lower furrow, A1 and A2. The size of the bubbles in the upper furrow
significantly higher than that obtained at the other two values. The is found to be smaller in comparison to the lower one. The centre of
corresponding values of the computed root mean square (rms) primary recirculation zone ‘A0’ in the lower furrow shifts to the
variation about the mean value at Reynolds numbers of 400, 600 right of the domain with elapsed time and its size also reduces,
and 800 respectively are 1.15, 1.66 and 1.77. The rms values are also Fig. 12(b) at time ‘t þ ðs=4Þ’. Meanwhile, the size of the small
seen to increase with Reynolds number. Although the rms variation counter-clockwise zone ‘A1’ correspondingly increases and it also
between Reynolds number values of 600 and 800 is very small, still starts shifting to the right of the domain. The size and location of
a significant increase in mean Nusselt number is noticeable when the clockwise rotating secondary zone ‘A2’ formed by flow sepa-
the Reynolds number increases from 600 to 800. With a further ration near the domain inlet, during this phase remains fairly
increase in Reynolds number to 1000 the obtained Nusselt number stagnant. At a later instant, not shown here, the size of clockwise
and rms values are 12.09 and 1.44 respectively, confirming the rotating secondary zone also starts to increase and it shifts down-
increase in heat transfer with increasing Reynolds number. stream along the wall.
Fig. 12 depicts temporal evolution of the flow field at Re ¼ 600 A stage is reached, Fig. 12(c) at time ‘t þ ð2s=4Þ’, when all the
through four instantaneous snapshots of streamlines within a time- three recirculation zones become eventually of equal size and at
cycle (s). Initially at time ‘t’ (Fig. 12(a)) one large recirculation zone this moment the centre of intermediate counter-clockwise zone
(A0) is present in the lower furrow region. A pair of tiny recircula- falls on the domain centreline. At this time the development cycle
tion zone is also seen to develop adjacent to the large bubble in the beings on the top wall, and Fig. 12(c) resembles Fig. 12(a) with the
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 161

Fig. 10. Instantaneous isotherms at different Reynolds numbers (a) Re ¼ 25, (b) Re ¼ 100, (c) Re ¼ 200, (d) Re ¼ 400, (e) Re ¼ 600, (f) Re ¼ 800.

recirculation regions swapped between the crest and the trough forming the clockwise rotating small recirculation. The following
regions of the channel. As time elapses, at time ‘t þ ð3s=4Þ’, all the second point where Cf again becomes positive corresponds to the
three recirculation zones merge with one another forming a single region where the developing two small developing counter-
primary recirculation region ‘B0’, Fig. 12(d). As the three recircula- rotating recirculation regions interact with each other. The value
tion zones merge with one another, a complex interaction occurs of Cf becomes negative again at the point where the counter-
through which fresh warm fluid from main stream flow is sucked clockwise rotating small recirculation region separates from the
into and the cold fluid coming from the wall is ejected from the wall and the value remains negative up to the point where the
resultant newly formed primary recirculation region. The whole primary recirculation region reattachment takes place. The Nusselt
cycle then repeats itself in time, after ‘t þ s’, thereby taking cold number in the region of two newly formed counter-rotating recir-
fluid from the wall and mixing it with hot fluid in the main stream culation zones is low due to the fact that they remain trapped
bringing about augmentation in both heat and mass transfer rates. within the main stream fluid and the primary recirculation region
Fig. 13(aed) shows the instantaneous variation of Nusselt thereby suppressing heat and mass transfer. The Nusselt number
number and skin friction factors plotted on the bottom wall of the peaks at a value of 19.2 in the region where the reattachment of
channel with the instants corresponding to the phase at which primary vortex to the channel wall, bringing fresh fluid from the
streamlines were plotted in Fig. 12. The location of first point, at core, takes place. The Nusselt number gradually decreases along the
which Cf becomes negative in Fig. 13(a) corresponds to the position wall as this clockwise rotating fluid moves upstream of the channel
where the flow separation from the wall takes place in Fig. 12(a) exchanging heat with the wall. In Fig. 13(b) also, each point at
162 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

Fig. 11. Temporal Nusselt number variation (left) and its power spectra (right) at Re ¼ 400, 600 and 800.

which Cf changes its sign corresponds either to a separation or to transfer begins to decrease in the region where the flow first sep-
a reattachment of the fluid from the wall pertinent to Fig. 13(b). The arates from the wall forming the first recirculation zone. It becomes
highest Nusselt number with a magnitude of 21.1 again is found to the lowest in the region where the second recirculation region
occur in the region where the primary recirculation region reat- develops and begins to increase in the region of interaction be-
tachment takes place. tween the primary recirculation region and the developing sec-
The lowest Nusselt number regions with a value lower than 2.0 ondary zone. The lowest heat transfer rates, Fig. 13(d), are found at
occurs where the initially formed small counter-rotating region is the instant at which there exists only one vortex region where the
developing and is unable to exchange heat with the wall. The value Nusselt number remains very low along the channel wall and the
of Cf in this region also remains fairly constant. In Fig. 13(c) the magnitude of Cf also remains very low and almost constant. The
highest heat transfer also occurs in the primary recirculation region highest Nusselt number value observed at this instant is about 10.0.
and the following region where the boundary layer is thin. The heat It is generally observed that the locations of high and low Nusselt

Fig. 12. Streamlines plots at four different time instants at Re ¼ 600.

 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 163

Fig. 13. Nusselt number and skin friction factors variation as a function of arc-length at four time instants at Re ¼ 600.

number values are closely correlated to the existence of corre- Reynolds numbers. The streamline pattern shows the formation of
sponding large and small magnitudes of skin friction factor. The a pair of recirculation bubbles one within each furrow, with their
Nusselt number is also perceived to decrease or increase sig- centre located towards the downstream direction of the wavy
nificantly respectively in the regions where the skin friction factor geometry. However, at Re ¼ 1000, two tiny recirculation bubbles
either becomes negative or positive. are observed to be trapped between the primary recirculation
zone and the channel wall. The size of this secondary bubble is
7.2.2. Time-averaged field very small at Re ¼ 1000. The size of this bubble may grow in size
The spatial contours of the time-averaged streamlines and with increasing Reynolds number which cannot be predicted
isotherms for unsteady flows at Re ¼ 400, 600, 800 and 1000 are without doing computations at higher Reynolds numbers. From
shown in Fig. 14(aed). It is to be noted that the time-averaging of Table 5, it can be easily accredited that with increase in Reynolds
the flow and the temperature fields have been done over 25 number, both vorticity magnitude and temperature at the centre
flow-through periods. The averaging has also been carried out of primary recirculation zone increase indicating higher vortex
using 30 flow-through periods but no noticeable changes in the strength and better mixing respectively. It is also clearly identified
results are observed. For the aforesaid values of Reynolds number that the location of the primary recirculation bubble remains fairly
the locations and magnitudes of properties such as peak vorticity constant for Reynolds number ranging from 400 to 1000. It is
magnitude and temperature at the centre of the recirculation observed from the isotherms that as the Reynolds number in-
bubble are shown in Table 5. No significant differences in the creases the warm core fluid penetrates the furrow region thus
mean flow characteristics are observed among the values of increasing the heat transfer.
164 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

Fig. 14. Time-averaged streamlines and isotherms at different Reynolds numbers (a) Re ¼ 400, (b) Re ¼ 600, (c) Re ¼ 800, (d) Re ¼ 1000.
 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166 165

Table 5 the respective normalized values. The drop in Nusselt number

Location and magnitude of properties at centre of recirculation bubble for various value at a Reynolds number of 100 compared to Re ¼ 25 is due to
Reynolds numbers.
the formation of a recirculation region, characterized by small
Re x y Peak vorticity ðjuz jÞ Temperature (q) magnitudes of the cross-stream velocity component, which de-
100 0.823 0.506 1.022 0.243 creases the heat transfer between the wall and the core fluid. At the
400 1.005 0.435 2.235 0.385 Reynolds number of 25, with no separation, the obtained thermal
500 1.022 0.426 3.122 0.548
gradients are relatively higher (not shown here) than those
600 1.022 0.426 3.514 0.613
800 1.017 0.425 3.839 0.813 obtained at Re ¼ 100 which explains the higher Nusselt number at
1000 1.010 0.426 4.045 0.870 the former Reynolds number value. As the Reynolds number is
0.880 0.700 0.207 0.036 increased to 200, the strength of the recirculation increases.
Because of higher inertia (at Re ¼ 200), the thermal gradients on the
The heat transfer augmentation with increase in Reynolds wall become higher resulting in increase of heat transfer rates.
number is characterized by comparing the variation of the mean It is evident from the normalized Nusselt number values that
streamwise velocity component and the temperature profiles along within the steady regime the heat transfer enhancement obtained
the channel centreline (Fig. 15(a, b)). The average velocity obtained using wavy wall geometry is comparable to that obtained using the
along the streamwise direction is evidently higher for the Reynolds parallel plate channel. Hence the wavy walled geometry does not
number of 1000 depicting higher mass transfer rates and better provide any consummate performance over a parallel plate channel
mixing with the fluid in the channel furrows. In addition, the within the steady regime. It is apparent that heat transfer
temperature profile variation along the streamwise direction de- enhancement at Re ¼ 600 with Nu value of 7.76 is superior as
marcates more uniform temperature with increase in Reynolds compared to Re ¼ 400 and 500 having Nu values of 5.69 and 6.58
number. Further the variation of the temperature profile along the respectively. As the Reynolds number is increased further to 800
cross-stream direction also becomes more uniform at higher Rey- there is a sharp rise in the Nusselt number value by 40.1% to a value
nolds number. This is explained by flatter profiles at high Reynolds of 10.93 from the value of 7.76 at Re ¼ 600. The value of Nu is fostered
number values, indicating better diffusion rates. This supports the further to 12.10 as Re is increased to 1000. As discussed earlier this is
fact that heat transfer performance of the studied geometry in- due to the high intensity of mixing between the core and the near
creases with Reynolds number. wall fluid at higher Reynolds number values. As the Reynolds
The effectiveness of the wavy channel can also be quantified by number increases higher unsteady complex momentum transport
evaluating the Thermal Performance Factor (TPF) defined as the due to higher non-linearity causes the rupture of the thermal
ratio of heat transfer enhancement to friction factor (Gee and Webb Table 6
[28]) and given by, Friction factor and Nusselt number and thermal performance for the range of
Reynolds number studied.

TPF ¼ Nu=Nuo =ðf =fo Þ1=3 (12) Re Friction Nusselt TPF Strouhal
factor f ðf =fo Þ number NuðNu=Nuo Þ number St

where, Nuo ¼ 3:77 is the Nusselt number for fully developed flow 25 1.287 (1.340) 4.539 (1.204) 1.092 e
100 0.382 (1.593) 4.444 (1.179) 1.009 e
in a parallel-plate channel subjected to a constant wall tempera-
200 0.212 (1.763) 4.551 (1.207) 0.999 e
ture, Incropera and DeWitt [29] and fo ¼ 24/Re is the friction factor 400 0.124 (2.065) 5.687 (1.508) 1.185 0.293
for parallel flow in a straight channel (Muralidhar and Biswas [30]). 500 0.108 (2.252) 6.580 (1.745) 1.331 0.293
The friction factor, Nusselt number, thermal performance factor and 600 0.096 (2.400) 7.760 (2.058) 1.537 0.305
Strouhal number for the wavy geometry for the range of Reynolds 800 0.080 (2.653) 10.933 (2.900) 2.095 0.305
1000 0.069 (2.888) 12.099 (3.209) 2.254 0.305
numbers studied are compared in Table 6. The quantities written
within brackets for friction factor and Nusselt number represent

Fig. 15. Streamwise and spanwise variation of time-averaged streamwise velocity and temperature at different Reynolds number.
166 A.G. Ramgadia, A.K. Saha / International Journal of Thermal Sciences 67 (2013) 152e166

boundary layer and replaces it with a fresh boundary layer on the moderate Reynolds number for steady flow, Int. J. Heat Mass Transfer 33 (5)
(1990) 835e845.
walls of the wavy geometry. This is the root cause of heat transfer
[3] T.A. Rush, T.A. Newell, A.M. Jacobi, An experimental study of flow and heat
amplification for the wavy channel. It is also pointed out that as transfer in sinusoidal wavy passages, Int. J. Heat Mass Transfer 42 (1999)
opposed to a constant Nusselt number value of 3.77, independent 1541e1553.
from the Reynolds number for the case of fully developed channel [4] T. Nishimura, Y.N. Bian, Y. Matsumoto, K. Kunitsugu, Fluid flow and mass
transfer characteristics in a sinusoidal wavy-walled tube at moderate Rey-
flow, the Nusselt number shows strong dependency on Reynolds nolds numbers for steady flow, Heat Mass Transfer 39 (2003) 239e248.
number for fully developed flow through a wavy walled channel. [5] G. Tanda, G. Vittori, Fluid flow and heat transfer in a two-dimensional wavy
The friction factor is also found to decrease with Reynolds number channel, Heat Mass Transfer 31 (1996) 411e418.
[6] C.C. Wang, C.K. Chen, Forced convection in a wavy-wall channel, Int. J. Heat
within the studied range. However, the normalized friction factor Mass Transfer 45 (2002) 2587e2595.
increases with Reynolds number and it indicates, as f/fo > 1.0, that the [7] M.Z. Hossain, A.K.M.S. Islam, Fully developed flow structures and heat transfer
resistance to the flow field is always higher for a wavy walled channel in sine-Shaped wavy channels, Int. Commun. Heat Mass Transfer 31 (6) (2004)
887e896.
at a given Reynolds number than that offered by a parallel plate [8] J. Zhang, J.J. Kundu, R.M. Manglik, Effect of fin waviness and spacing on the
channel. The highest TPF value of 2.254 with peak Nusselt number lateral vortex structure and laminar heat transfer in wavy-plate-fin cores, Int.
and lowest friction factor is obtained at the highest Reynolds number J. Heat Mass Transfer 47 (2004) 1719e1730.
[9] M.Y. Wen, C.Y. Ho, Heat-transfer enhancement in fin-and-tube heat exchanger
of 1000. Similarly to the Nusselt number value, the TPF value is also with improved fin design, Appl. Therm. Eng. 29 (2009) 1050e1057.
found to increase steeply by 36.3% from the value of 1.537 at Re ¼ 600 [10] M.S.B. Haitham, N.K. Anand, H.C. Chen, Numerical study of heat and mo-
to the value of 2.095 at Re ¼ 800. It is also pointed out that the TPF mentum transfer in channels with wavy walls, Numer. Heat Transfer Part A 47
(2005) 417e439.
value is considerably greater than unity for Reynolds number greater
[11] S. Mahmud, A.K.M.S. Islam, C.M. Feroz, Flow and heat transfer characteristics
than 400, when the flow becomes unsteady. This further supports the inside a wavy tube, Heat Mass Transfer 39 (2003) 387e393.
earlier point that to obtain the best thermal performance the wavy [12] S.V. Patankar, C.H. Liu, E.M. Sparrow, Fully developed flow and heat transfer in
wall geometry should always be operated under unsteady flow con- ducts having streamwise e periodic variations of cross-sectional area, J. Heat
Transfer 99 (1977) 180e186.
ditions. From the friction factor data obtained numerically, the cor- [13] G. Wang, S.P. Vanka, Convective heat transfer in periodic wavy passages, Int. J.
relation given by equation (13) is obtained for the range of Reynolds Heat Mass Transfer 38 (17) (1995) 3219e3230.
number studied in the present study. The correlation gives a max- [14] B. Niceno, E. Nobile, Numerical analysis of fluid flow and heat transfer in
periodic wavy channels, Int. J. Heat Fluid Flow 22 (2001) 156e167.
imum error of 8% at Reynolds number of 1000. [15] G.N. Xie, Q.W. Wang, M. Zeng, L.Q. Luo, Numerical investigation of heat
transfer and fluid flow characteristics inside a wavy channel, Heat Mass
f ¼ 15:4Re0:797 (13) Transfer 43 (2007) 603e611.
[16] Y. Sui, C.J. Teo, P.S. Lee, Y.T. Chew, C. Shu, Fluid flow and heat transfer in
microchannels, Int. J. Heat Mass Transfer 53 (2010) 2730e2772.
[17] Y. Sui, C.J. Teo, P.S. Lee, Direct numerical simulation of fluid flow and heat
8. Conclusion transfer in periodic wavy channels with rectangular cross-sections, Int. J. Heat
Mass Transfer 55 (2012) 73e88.
[18] L. Gong, K. Kota, W. Tao, Y. Joshi, Parametric numerical study of flow and heat
The numerical simulation of fluid flow and heat transfer through
transfer in microchannels with wavy walls, J. Heat Transfer 133 (2011)
a sinusoidal wavy channel whose surface is described by 051702.1e051702.10.
y ¼ 2a$sin2 ðpx=LÞ has been performed to elucidate the effect of [19] A.K. Saha, S. Acharya, Unsteady simulation of turbulent flow and heat transfer
number of streamwise periodic modules considered and Reynolds in a channel with periodic array of cubic pin-fins, Numer. Heat Transfer Part A
46 (2004) 731e763.
number on the fluid flow and heat transfer. It is found that within the [20] A.K. Saha, Effect of number of periodic module on flow and heat transfer in
range of Reynolds number studied, the channel geometry formed by a periodic array of cubic pin-fins inside a channel, J. Enhanced Heat Transfer
single sine wave is sufficient to predict the fluid flow and heat transfer 15 (3) (2008) 243e260.
[21] A.M. Guzmán, C.H. Amon, Transition to chaos in converging-diverging
without suppressing the low frequency oscillations caused by the Te channel flows: RuelleeTakenseNewhouse scenario, Phys. Fluids 6 (6)
S waves. The critical Reynolds number for the geometry is found to be (1994) 1994e2002.
in the range of 350e400. It is also established that to get preeminent [22] A.K. Saha, S. Acharya, Parametric study of unsteady flow and heat transfer in
a pin-fin heat exchanger, Int. J. Heat Mass Transfer 46 (2003) 3815e3830.
heat transfer the flow through the channel should be unsteady and at [23] C.M. Rhie, W.L. Chow, Numerical study of turbulent flow past an aerofoil with
high Reynolds number. Then only the enhanced unsteady mo- trailing edge separation, AIAA J. 21 (11) (1983) 1525e1532.
mentum transport caused by the high inertia of the fluid will lead to [24] T. Nishimura, Y. Ohori, Y. Kawamura, Flow characteristics in a channel
with symmetric wavy wall for steady flow, J. Chem. Eng. Jpn 17 (5) (1984)
rupture of the thermal boundary layer near the wall eventually 466e471.
leading to high thermal performance ratios. For the range of Reynolds [25] L. Casarsa, T. Arts, Experimental investigation of the aero thermal perfor-
number studied, the highest heat transfer rates along with the best mance of high blockage rib-roughened cooling channel, ASME, J. Turbo Ma-
chinery 127 (2005) 580e588.
thermal performance are obtained at the highest studied Reynolds
[26] J.H. Wagner, B.V. Johnson, R.A. Graziani, F.C. Yeh, Heat transfer in rotating
number value of 1000. A correlation for the friction factor for the serpentine passages with trips normal to flow, ASME, J. Turbo Machinery 114
geometry within the range of Reynolds number studied in the pre- (1992) 847e857.
sent study has also been found, and gives a maximum error of 8%. [27] A.G. Ramgadia, A.K. Saha, Large eddy simulation of turbulent flow and heat
transfer in a ribbed coolant passage, J. Appl. Math. 21 (2012). Article ID
246313.
References [28] D.L. Gee, R.L. Webb, Force convection heat transfer in helically rib-roughened
tubes, Int. J. Heat Mass Transfer 23 (1980) 1127e1136.
[1] J.L. Goldstein, E.M. Sparrow, Heat and mass transfer characteristics for flow in [29] F.P. Incropera, D.P. DeWitt, Introduction to Heat Transfer, fourth ed., John
a corrugated wall channel, J. Heat Transfer 99 (1977) 187e195. Wiley & Sons, New York, 2002.
[2] T. Nishimura, S. Murakami, S. Arakawa, Y. Kawamura, Flow observations and [30] K. Muralidhar, G. Biswas, Advanced Engineering Fluid Mechanics, second ed.,
mass transfer characteristics in symmetrical wavy-walled channels at Narosa Publishing House, New Delhi, 2005.