You are on page 1of 14

Heat transfer enhancement by using pulsating flows

Efrn Moreno Benavides


Citation: Journal of Applied Physics 105, 094907 (2009); doi: 10.1063/1.3116732
View online: http://dx.doi.org/10.1063/1.3116732
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/105/9?ver=pdfcov
Published by the AIP Publishing

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

JOURNAL OF APPLIED PHYSICS 105, 094907 2009

Heat transfer enhancement by using pulsating flows


Efrn Moreno Benavidesa
Dpto. Motopropulsin y Termofluidodinmica, ETS Ingenieros Aeronuticos, Universidad Politcnica
de Madrid, Pza. Cardenal Cisneros, 3, 28040 Madrid, Spain

Received 30 April 2008; accepted 15 March 2009; published online 8 May 2009
The paper presents a theoretical model of convective heat exchangers working with internal
pulsating flows. It aims for a better physical understanding of the processes leading to a heat transfer
enhancement inside these devices. When the frequency of the pulsation is increased, some
geometries exhibit a maximum response, measured by its temperature rise, similar to those obtained
in some dynamical resonant systems. The work explains the nature of this characteristic behavior
and produces a simplified theoretical model that isolates the main physical features of the fluid
dynamics involved. Two characteristic frequencies, measured by its Strouhal numbers, are
theoretically found. The first one is associated with the spatial-averaged thermal response of the
fluid near the wall and the second with the response of the velocity field. It is found, for a general
device, that both Strouhal numbers and the maximum enhancement are mainly defined by the
geometry of the device. Finally, the heat transfer enhancement of a straight channel, a backward
facing step channel, and a two heated blocks inside an adiabatic channel are used to validate the
model. Enhancements calculated with the present model are compared with the results reported in
the scientific literature showing a good agreement for the tested cases. 2009 American Institute
of Physics. DOI: 10.1063/1.3116732
I. INTRODUCTION

The specific power of thermoacoustic devices is increased by using in the heat exchangers solid walls separated
only by a few thermal penetration depths of gas and by increasing the surface to volume ratio.14 This feature needs to
be implemented in the design for practical applications and
hence several actual designs include flat plates, wires,
grooves, sharp bends, stepped walls, and stepped or largeaspect ratio channels, among others. Besides, nowadays,
there are many other engineering systems that present similar
configurations such as microelectromechanical systems, microcombustors, or microcoolers. The state of the art in the
theoretical characterization of these heat transfer processes
appears when the pulsating flow is considered, either because
it appears in a natural way such in some kind of thermoacoustic devices15 or well because it is forced by pulsating
the flow with vibrating or moving parts placed far enough in
the upstream or downstream path.6,7 Since the refrigeration
power depends on the heat fluxes, an interesting way to
modify these fluxes could be to change the mean velocity of
the fluid just as it has been experimentally corroborated in
baffled pipes.8 Other interesting applications of the oscillating baffled tubes are the oscillatory flow reactors9,10 where
the transport phenomena, both mass and heat transfer, are of
the highest interest.
Due to the interest, the theme is being discussed in the
current scientific literature where it remains controversial. A
brief review11 of this situation is best summarized by classifying previous work into four categories according to the
conclusion being reached: a pulsation enhances heat
transfer,8,12 b pulsation deteriorates heat transfer,13 c pula

Electronic mail: efren.moreno@upm.es.

0021-8979/2009/1059/094907/13/$25.00

sation does not affect heat transfer,11 and d heat transfer


enhancement or deterioration may occur depending on the
flow parameters.14
After the review,11 where the main conclusion was that
pulsation neither enhances nor deteriorates heat flow, it has
been reported that pulsation has no effect on the timeaveraged heat transfer along straight channels15 and that
forced flow pulsation enhances convective mixing and affects Nusselt number:16 it reaches its maximum for a specific
pulsating frequency and decreases for both higher and lower
values of frequency.
Since this maximum looks like the one obtained in the
case of many damped resonating dynamical systems, it was
suggested16 that this behavior was due to the existence of a
nonlinear coupling of resonant nature between thermal effects and fluid dynamics parameters, i.e., it was concluded
that the Nusselt number enhancement appeared to be of a
resonant nature. However, since the energy equation is decoupled from the momentum one for an incompressible flow,
the velocity field is not determined by the temperature field
except for the change in properties due to the temperature
variations. In addition, the decoupled energy equation does
not have any temporal derivative of order greater than 1, so
that there is no resonant frequency associated with the temperature field. A resonant frequency could appear if variable
properties were considered. However, it was verified, by
means of numerical calculations with a two-dimensional
2D backward facing step at low Reynolds numbers, that the
maximum is obtained even when the properties are considered constant16 and hence the explanation based on the resonant frequency is not satisfactory.
Therefore, this study is motivated by the need to find a
comprehensive explanation that fixes the basic mechanisms
leading to a heat transfer enhancement when a pulsating flow

105, 094907-1

2009 American Institute of Physics

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-2

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

is present. Although the theory is highly simplified, it proposes an alternative method to deal with this complex phenomenon. Since the formulation of the presented model is
supported by integral equations, the main results are general
and, hence, useful for a broad class of geometries. Designers
could use these ideas to devise means to enhance heat transfer processes or to improve the current devices.

By defining T pcT pT p as the spatial-averaged specific


internal energy over the volume V p, the energy balance equation leads to
d
dQ
= V P T PtcT PtT Pt
dt
dt
+ GcTHTH cTLTL.

II. PROBLEM FORMULATION

As usual, the heat flow amounts to

In order to establish which ones are the main physical


phenomena that explain the above exposed disparity of results, a general geometrical configuration will be studied and
the physics involved will be simplified to retain only the
relevant terms, so that, we focus on incompressible flow
without any source of mechanical work. This approach will
retain only the most important aspects for the dynamical response.
Consider the unsteady temperature field Tx , t, where x
is the three-dimensional vector and t represents the time, and
the three-dimensional unsteady velocity field, ux , t, defined
in every point of a volume V limited by the surface SV. The
Reynolds transport theorem applied to the energy balance
leads to

dQ
= hSWTW T P.
dt

dQ
=
dt

u2 3
cT +
dx
2
V t

SV

cT +

P u2
u dA.
+
2

The term on the left represents the heat flux, is the


density, c the specific heat capacity of the fluid, P the pressure, and dA a surface element vector. Since the fluid is
liquid, P / cT 1 and u2 / 2cT 1 hold, and the energy balance can be reduced to
dQ
=
dt

cTd3x + GcTHTH GcTLTL .

t
V

Here, G is the mass flow rate and TLcTLTL and


THcTHTH are the spatial-averaged specific internal energies over the inlet and exit ports, respectively.
We assume a stationary inlet temperature profile. However, since the problem is not stationary, the temperature
profile at the exit port must exhibit a periodical response.
This temporal oscillation is the result of a spatial wave traveling toward the exit port. Consequently, the spatial-averaged
internal energy over a region with a characteristic length
much greater than the wavelength must be stationary. Let us
define V-V p as the volume where transient effects are negligible over the spatial-averaged properties and V p as the volume where the spatial-averaged internal energy is time dependent. By this definition, the internal geometry and the
unsteady fluid field have been divided into two characteristic
regions with physical relevance. Remembering that the temporal oscillation of the internal energy is negligible outside
the volume V p, the following equation holds:
dQ
=
dt

cTd3x + GcTHTH GcTLTL .

Vp t

Here h is the spatial-averaged heat transfer coefficient


and SW is the wall area at the temperature TW.

III. EFFECT OF THE PULSATION ON THE


TEMPERATURE FIELD

For stationary flows, it is well known that the heat transfer coefficient is obtained, with enough precision, by relating
Nusselt, Reynolds, and Prandtl numbers, Nu= fRe, Pr. We
assume that this relationship preserves all the important
physics for our purpose its validity, for the pulsating flows
considered here, is shown in Sec. VII, so that it is used as
the instantaneous heat transfer model. Here, the Nusselt
number is defined as usual, Nu= hL / kT P, where L is a characteristic length associated with the device geometry and
kT P is the fluid thermal conductivity at the transient representative temperature in that region; the Reynolds number is
chosen in this context as Re= GL / T PA with A the inlet
cross-sectional area and T P the viscosity at temperature
T P; and the Prandtl number as Pr= T PcT P / kT P.
In order to obtain an analytical solution by means of a
perturbation method around the stationary solution, which is
labeled with the subindex 0, the nonstationary flow is modeled as G / G0 = 1 + at with the initial prescription 1
note that does not have any physical meaning, it is used
only with the purpose of obtaining an analytical solution and
must be set to 1 once the solution is found. Therefore, the
solution is expressed as T P / T0 = 1 + bt + 02 so hence, the
heat transfer coefficient is reduced to h / h0 = 1 + a + b
+ 02 with = - + + -, = Re dLnf / d ReG0,T0,
= Pr dLnf / d PrG0,T0, = T / kdk / dTT0, = T / d
/ dTT0, = T / cdc / dTT0, and = T / d / dTT0 defining the dependency of the fluid properties and the heat
transfer coefficient with the temperature and the velocity.
The order-0 energy equation is
h0SWTW T0 = G0cTHTH cTLTL.

By using this information, the next order in the energy


equation leads to
db
= 1za + z 1b,
d

V PT0cT01 + +
t
= tc =
,

h 0S W

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-3

z=

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

TW
1.
T0

The pulsating flow is obtained when a is set as a


= a+ei + aei / 2 subindexes + and indicate that a+
and a are conjugate complex constants, and = tc relates
nondimensional and dimensional pulsations. Then b is calculated from Eq. 7 to be b = b+ei + bei / 2 with
b+ = a+

1z
.
1 z + i

10

The modulus of this amplitude, see Eq. 11 below,


shows that the effects of the mass flow pulsation on the temperature field are damped if the period of the pulsation is
much greater than the characteristic time tc,
b+2 = b+b =

12z2
a+2 .
1 z2 + 2

11

Note that the denominator does not vanish in any frequency, so that there is no resonant frequency in the temperature field. Since the maximum perturbation in the temperature field is obtained for zero frequency, the relative
importance of , , and z determines the maximum response.
Coefficients , , , and depend on the properties of the
fluid, whereas and depend on the geometrical configuration of the device and on the topology of the fluid field
inside it. Coefficient mixes fluid properties and flow topology and, finally, coefficient z measures the thermal rise inside the fluid.
Typical values of these coefficients for liquid water17
which implies z 0.15 are = 0.60, = 6.37, 0.17,
and = 0.14. The specific heat for water has a minimum
near 313 K, so that the coefficient can be positive or negative depending on the temperature. With respect to the topology, if the dominant solid structure inside the channel is
composed of wires in crossflow surrounded by a liquid at
Reynolds numbers lower than 1000 and greater than 40,
a good approach18 could be to use fRe, Pr = 0.52 Re0.50
Pr0.62 / PrTW0.25, so that = 0.50 and = 0.62. However, for
a turbulent movement inside an empty pipe Re 104 typical values could be = 0.80 and = 0.33. Another interesting
case is the incompressible laminar flow in circular tubes that
assumes all the properties to be constant and both hydrodynamic and thermally fully developed, where it is known18
that the Nusselt number in this case does not depend on the
Reynolds and Prandtl numbers, i.e., = = 0. As a result,
coefficient spans from 0.7 to 3.4 the lower value holds
for the low-Reynolds-number crossflow case. By considering other fluids and other topologies, the coefficients could
be very different from the ones presented here.
Since z moves from 0.1 to 0.5, the expected error due
to neglecting the variation of the properties with the temperature can become significant. For example, a system characterized by a low Reynolds number over a flat plate
= 0.50, = 0.33, = 1.5, and z = 0.12 would have an error of
the order of 18% z = 0.18. This result compares well with
previous numerical calculations,16 which in similar conditions of low Reynolds numbers reported a maximum difference of 15%. However, although just as we have demon-

strated, exact calculations oblige to retain the dependency


with the temperature, this dependency is not necessary for
explaining the nature of the heat transfer enhancement.
Therefore, the properties are fixed as constant and hence
the temperature field response is given by
b+ = a+

1z
.
1 + i

12

Additionally, the heat transfer coefficient by


h
= 1 + a + 02.
h0

13

This result shows that there is no variation in the instantaneous heat transfer coefficient if = 0 just as some
researches15 pointed out for a range of pulsating frequencies
covering two orders of magnitude. They have shown that a
circular tube in the laminar regime under pulsating flow conditions does not have any oscillation of the local Nusselt
number if the flow is both thermally and hydrodynamically
developed. This can be explained with the use of the stationary solution for a constant wall temperature,18 Nu= 3.66,
which shows that in this case is zero and hence, avoids any
fluctuation of the heat transfer coefficient.
However, they also found15 that the Nusselt number varies in time in the near-entry region of the pipe. The explanation comes again from Eq. 13 by taking into account that
the thermal entry flow with a fully developed velocity profile
has a behavior given by the Lveque solution18 Nu Re0.33,
i.e., = 0.33, and hence the expansion given by Eq. 13
retains the pulsation effects.
On the other hand, when = 1, the oscillation of the
convection coefficient would be maximum while the oscillation of the temperature would be minimum. This is nearly
the situation in turbulent flows where = 0.8.

IV. EFFECT OF THE TEMPERATURE DEFINITION

Since the heat transfer coefficient replaces the unknown


heat flux by the introduction of a temperature difference, its
behavior is conditioned by the definition of this
difference.12,18,19 This problem is illustrated in this section
and solved in the next one.
A quantity related with the heat flow appears if the heat
transfer coefficient, Eq. 13, is multiplied by the temperature
fluctuation, Eq. 12. This quantity is an initial approach to
the heat flow,
dQ/dt
= 1 + a z1b 2z1ab + 02.
h0SWTW T0
14
Since the time-averaged heat flux is obtained by integrating the above expression along a complete period, the terms
of order vanish, and the nonlinear terms produce a heat
transfer given by
2a+2 1
dQ/dtaveraged
=1+
+ 02.
2
h0SWTW T0
1 + v2

15

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-4

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

This heat flux does not exhibit any resonant behavior


since all the terms in the denominator are always positive
quantities. Indeed, it always decreases with the frequency
and hence does not have any maximum it is always positive,
i.e., it predicts a heat transfer enhancement. This kind of
solutions can be found in the literature,8,11 when a periodical
perturbation without any constant component is added to the
stationary temperature field to give the periodical response.
For example note that this solution is maximum for
= 1 / 2, laminar flow, and negligible frequency, and that the
maximum pulsation without reversing flow is given by a+
= 1, the maximum heat transfer enhancement predicted by
this approach is 12.5% 6% for a+ = 1 / 2. Calculations for a
small-frequency periodic motion imposed on a fully developed steady laminar flow inside a cylindrical pipe show the
same result.11 It is significant that they conclude that the
interaction between the velocity and temperature oscillations
introduces an extra term in the energy equation which reflects the effect of the pulsations in producing higher heat
transfer rates, which increases approximately 6% compared
to its nonpulsating value when the amplitude is half of the
mean velocity. Besides, they conclude that the solution
would depend on the square of the frequency if terms of
higher order were kept in their solution just as the above
expression shows.
However, the solution given by Eq. 15 is not always
correct since the average of the temperature perturbation
over one period is implicitly assumed to be zero. Indeed, the
greater the heat transfer established by Eq. 15, the greater
the averaged local temperature over the wall, so that the
expansion of the temperature as a series of powers of
should include a constant term of order 2. The new terms
modify the solution and, hence, depending on the model used
for this feature, the calculated Nusselt number could change
significantly. As a consequence, it is possible to define different Nusselt numbers. It has been shown19 that, for the
same spatial and temporal temperature distribution, different
definitions of the average Nusselt number for pulsating flow
lead to contradictory results. The difference depends on how
the time-averaged temperature is constructed.
Although a definition of the Nusselt number for pulsating flow, based on the local bulk temperature, can be defined
in a rational way,12 the next sections will use the timeaveraged heat flux because it is the entity with physical
meaning. Besides, this quantity can be obtained from the
energy equation as the time-averaged increment of enthalpy
along the device. As it will be shown, only the steady Nusselt
number, which is well defined in the literature, will be
needed in this article. Of course, since the steady Nusselt
number depends on the Reynolds number, it must follow the
mass flux variations. This approach to the problem is developed in Sec. V by using the minimum number of terms in the
Taylor expansions that leads to a consistent solution.

wall temperature TW. Then, it is convenient to define a dimensionless measure of the heating efficiency as

Consider a general device with the limitations described


in Sec. II where the exit temperature TH is limited by the

16

Of course, the fluid temperature between the initial and


the final part of the hot wall changes significantly. This is
considered, in the calculation of the heat transferred from the
wall, by using the logarithmic mean temperature,18
T P = TW

TH TL
.
TW TL
ln
TW TH

17

The energy balance comes from Eqs. 4 and 5,


hSWTW T P = V Pc

dT P
+ GcTH TL.
dt

18

The definitions of two dimensionless ratios, given by


Eqs. 19 and 20, show the energy balance as

P =

TW T P
,
TW TL

19

0 =

TW T0
,
TW TL

20

d P G 0c G
h
.
P +
=
h0
d h 0S W G 0

21

By using 0 as the stationary heating efficiency, the energy balance is reduced to

h 0S W 0
= ,
G 0c 0

h P d P/ 0 G
+
=
.
h0 0
d
G0 0

22

23

In addition, the dimensionless form of the equation defining the logarithmic temperature Eq. 17 is
ln1 + / P = 0.

24

This equation relates the ratio P / 0 and the exit temperature measured by / 0. Its zeroth and first order terms
lead to

0 = 1 e ,

25

1
0
= Z,
P
1
0

26

Z=
V. HEAT TRANSFER FOR AN ARBITRARY GEOMETRY

TH TL
.
TW TL

1 0
.
0 + 0 1

27

Coefficient Z measures the temperature amplification inside the device. Normally, actual devices have small values
of , typically 0.1, and hence the following holds: 0

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-5

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

1, 0 1, and Z 21 1. Since Z is much greater


than 1, this states that the outlet fluctuations are driven by
very small fluctuations in the spatial-averaged specific internal energy over the volume V P. Although in this case the
terms of order 01 / Z could be set aside, it is convenient to
retain them for further discussions.
Therefore, when the fluctuation is taken into account,
these ratios are expressed for any value of as

1 = Zmt + s2 + 03,
0

28

P
1 = mt Ymt22 s2/Z + 03,
0

29

where Y is calculated from the logarithmic temperature definition by retaining the second order terms,
Y = Z0

Z/2 + 10 1
.
1 02

30

In order to solve the energy balance equation for mt


and s, it is necessary to know h / h0. This heat transfer coefficient is of a convective nature, and hence it must depend on
a characteristic velocity over the wall, which, in general, is
different from the spatial-averaged one obtained from the
mass flow rate. The quasisteady behavior suggests to use a
Taylor expansion h / h0 = 1 + d + d22 + 03, with and
constants and d a function that depends on the wall velocity. With this in mind, Eq. 23 evolves to
1 + Zm +

s=

dm
= a d,
d

31

dm
Z
d2 + m Za + d Y m + 2
Z+1
d

32

This system of differential equations determines m and s


by knowing a and d. Since dt represents the oscillation of
the dimensionless velocity over the wall, a pure-harmonic
approach imposes d = d+ei + dei / 2 with d+ and d two
conjugate complex constants. A simplified estimation of d+
is given in Sec. IX. Then, the solution is obtained from Eqs.
31 and 32 as
m = m+ei + mei/2,
m+ =

s=

33

a+ d+
,
1 + Z + i

Z
2Z + 1

34

+ 2
Z+1
Z+1
2
1+
Z+1

0 + 1

35

Z2 1 2Y
a+d+cosa+ d+,
Z + 12

37

2 = d+2 .

38

Note that the time-averaged outlet temperature is the one


that a thermocouple would measure in the exit port of an
experimental arrangement and hence the constant part of s is
a measure of the heat transfer enhancement. Note also that
the effect of fluid oscillation amplitude on heat transfer enhancement appears in Eqs. 3538. In Eq. 35 the heat
transfer enhancement s depends on the parameters 0, 1,
and 2, given by Eqs. 3638. In this scenario the parameters a+ and a, which are the complex numbers related to
the phase and to the amplitude of the mass flow rate fluctuation, and d+ and d, which are related to the phase and amplitude of the velocity over the wall, explicitly appear. Thus,
this heat transfer modeling takes into account the effects
present in oscillatory baffled tubes where the pulsating flow
creates a periodic mass flow rate that contributes to an increase in the cycle-average velocity near the walls. Furthermore, the model predicts that the dependence is quadratic in
the parameters describing the mass flow rate and the velocity
near the wall. Besides the peak superficial fluid velocity is
measured by the modulus of d+ that appears in 1 and 2 as
linear and quadratic, respectively. Therefore, both the peak
superficial fluid velocity and the averaged flux are used to
characterize the flow, just as it is normally done for describing the mixing in oscillatory baffled tubes.20 However, Eq.
35 also depends on the frequency and on the relative phase.
In fact, this is a proof that shows that more than two parameters must be considered for a full description.
For very low frequencies, this solution Eq. 35 can
produce the enhancement or deterioration of the heat transfer
process depending on the sign of the coefficient 0 which
changes, for example, with the relative phase between the
velocity over the wall and the mass flow rate. The same can
happen at very high frequencies depending on the sign of 2
which changes with the sign of , which normally is not
positive and hence it will not give an enhancement of the
heat transfer process. Besides, the solution given by Eq. 35
presents an extreme at the frequency,

2 0
2 0 2

+ 1.
39
Z+1
1
1
This frequency gives the maximum response and hence establishes a critical Strouhal number. Since this extreme disappears for those devices where 1 vanishes, the above expression is able to reproduce all the results found in the
current scientific literature. In the next sections, this solution
is validated for several practical devices.
VI. HEAT TRANSFER IN STRAIGHT CHANNELS

As shown before, for very high frequencies, the solution


given by Eq. 35 can be approached by

ZZ + 1 Y
Z+1+Y
a+2 + 2
d+2
0 =
Z + 12
Z + 12

1 = a+d+sina+ d+,

36

s=

Z
d+2 .
2Z + 1

40

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-6

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

This states that the final response depends on the parameter Z


which depends on the efficiency of the device, on the
square of the amplitude of the pulsation, and on the parameter which depends on the behavior of the stationary
convective coefficient. Assuming that the stationary convective coefficient follows a potential law of the form h Re,
its Taylor expansion gives = -1 / 2 which is negative
for those applications with 1. For example, a turbulent
flow = 0.8 suffers a discouragement of the heat transfer
process. From the expression, we obtain that the maximum
discouragement is produced for = 0.5 laminar flow and Z
tending to infinity very low efficiency. Without reverse
flow on the wall, the maximum value for d+ is 1, and hence
the maximum attainable discouragement is near 6%.
Another interesting limit is the one obtained for very
long pipes, where the ratio length/radius l / R tends to infinity. For a pipe, = h0SW / G0c can be expressed as
= 2l / RNu0 / Pr Re0, with the Nusselt and Reynolds numbers based on the radius of the pipe. This means that can
be much greater than 1 if the length increases sufficiently:
l / R Pr Re0 / Nu0. In this limit, 1, 0 1, Z e
1, and Y / 2 1 hold, and hence, the solution given by
Eq. 35 tends to zero exponentially independently of the
flow parameters. This means that the difference between pulsating and nonpulsating flows decreases along the tube
length just as it has been found in a previous work.21
In addition, for fully developed stationary flows in circular pipes, the value of is 1 Ref. 18 and thus
= -1 / 2 = 0 holds: the discouragement is near zero for
high frequencies independently of the length of the pipe and
of the amplitude of the oscillation. It has been theoretically
demonstrated12 that the difference between the pulsating and
nonpulsating flows is very small less than 1% of the steady
one for very low frequencies and zero for high frequencies.
For turbulent flows 0.8, 0.08 or pipe entries
0.33, 0.11, the behavior depends on the heat
flow imposed by the stationary mass flow rate, on the operational temperatures, and on the propagation of the pulsation
toward the wall against the viscosity. This last fact obliges us
to seek a model, as general as possible, that relates the averaged velocity over the wall and the heat transfer coefficient.
Section VII aims to attain this, while Sec. IX describes a
model that relates the velocity over the wall with the mass
flow rate pulsation.
VII. ESTIMATION OF THE HEAT TRANSFER
COEFFICIENT FOR AN ARBITRARY GEOMETRY

This estimation lies in the dimensionless equations governing the heat transfer for an incompressible flow under the
assumption that the properties of the fluid are temperature
independent18 Einsteins notation is applied with i = 1 , 2 , 3

ii = 0,

41

St0 0 j + ii j = j p +
St0 0 + ii =

1
i i j ,
Re0

1
Ec0
i i
jii j + ji.
Pe0
Re0

42

43

Dimensional and nondimensional variables are related to


each other by means of

= 0,
t

= L1i,
xi

P=p

G20
,
A2

St0 =

AL
,
G0

Ec0 =

G20
.
2A2cTW TL

x i = L i,

ui = i

G0
,
A

T = TW TW TL,

Re0 =

LG0
,
A

Pe0 =

LG0c
,
kA
44

Note that the dimensionless numbers St0, Re0, Pe0, and


Ec0 are evaluated at the mean mass flow rate, and hence do
not change with the pulsation. This fact is indicated by the
subindex 0. Eckert number Ec0 only affects the temperature
field and only has to be taken into account when friction
gives rise to a noticeable warming of the fluid which is not
the case because the fluid velocity is considered to be negligible when compared with the speed of sound and there is no
any large velocity gradient fluid is being modeled as incompressible. In this situation, the equation governing the maximum attainable heat transfer enhancement is
Pe0 St0 0 + Pe0 ii = ii .

45

The Strouhal number St0 typically is near to one in these


applications, and hence the complete device cannot be
treated as stationary. However, the Pclet number Pe0 is normally much greater than 1, and hence the dominant effects in
the heat transfer forced by the wall temperature can be restricted to those that are important in a region near the wall.
Indeed, it fixes the thickness of the thermal convective
boundary layer. The thickness of this thermal boundary layer
can be obtained by calculating the characteristic length that
leads to a Pclet number of order of 1, i.e., kA / G0c
L. The energy equation can be formally rewritten for this
layer by changing the characteristic length L for L, so
that, near the wall, St0 0 / L 1 and Pe0 1 hold.
Here, the nomenclature Pe0 and St0 indicate that those
numbers have been calculated by using instead of L. This
fact allows the removal of the temporal derivative from the
energy equation near the wall. Therefore, although the complete device is not stationary, the thermal boundary layer can
be modeled as quasisteady, and Eq. 45 can be reduced to
Pe0 ii = ii .

46

Note that we have retained the convective term in the energy


equation near the wall because it has to be as important as
the dominant one. Since it is proportional to the velocity, the
maximum heat transfer enhancement is obtained when the
velocity near the wall is maximum. This allows us to find an
upper bound to the heat transfer enhancement by considering
an inviscid flow, i.e., by taking the Reynolds number to infinity. Thus, the vanishing viscosity allows to approach the
velocity near the wall by a parallel flux i = 1 which does not
depend on the transversal coordinate i = 2. In this situation,

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-7

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

the thermal conduction in the direction of the flow can be


neglected as well as all the derivatives in the third component i = 3, and hence the equation governing the maximum
attainable heat transfer enhancement results
Pe0 11 = 22 .

47

In this equation 1 is assumed to be maximum and independent of the coordinates i = 1 , 2 , 3. Although we have presumed an infinite Reynolds number, the Pclet number Pe0
has been maintained finite because in a convective heat
transfer problem, it is always as important as the conductive
heat transfer. Since liquids have Prandtl numbers Pr
= Pe/ Re above 1, and then a thermal boundary layer thinner
than the velocity boundary layer removed in the present
case because of the large Reynolds number assumed, the
last assumption could seem so drastic. However, it is useful
because it emphasizes the effects derived from the velocity
fluctuation while retaining a finite heat transfer rate.
A solution which satisfies the boundary conditions
1 , 0 = 0 and 0 , / 2 = P is

= Pe

2/ Pe
1
0 1

sin2.

48

In this expression, 0 is a constant related to the upper


limit of the convective boundary layer, note that the flux
of heat is zero at 2 = / 2 = / L. The heat transferred
over the entire wall amounts to


SW

T
x2

dSW =
x2=0

kSWTW TL
L

2/ Pe
1
0

2=0

2k Pe0
1 .
L

= 1,

49

50

51

This solution is formally similar to that calculated for a


thermal entry in a laminar tubular flow where the velocity
profile is described by the HagenPoiseuilles law, Nu
= 3.66+ 0.05/ X, with X = L / 2R Pe and R the pipe radius,18
if the heat transfer at zero velocity is discounted. It is interesting to note that this solution assumes a finite low Reynolds number, whereas the one derived here assumes the
Reynolds number to be infinite with the purpose of obtaining
an upper bound to the actual heat in a given configuration.

= 0.

52

Z
+
1
Z
,
s=
2Z + 1
2
1+
Z+1
0 + 1

53

ZZ + 1 Y
Z+1+Y
Z2 1 2Y
2
2
a+
2 a+
2 d+
Z + 1
Z + 1
Z + 12
d+cosa+ d+,

1 = a+d+sina+ d+.

Remembering that Pe0 1 and that / L 1, the


above expression can be approached by
h=

2
Re0 Pr 1,
L

The advantage of this approach is that it explicitly retains the effect of the local velocity near the wall. This expression states that the mean Nusselt number is proportional
to the dimensionless velocity over the wall. Since this velocity is induced by the pulsating flow, it has to exhibit an
induced oscillation which also makes the instantaneous mean
Nusselt number to oscillate. The effect of the pulsation on
the velocity near the wall can be described by 1 / 10 = 1
+ dt, and hence the instantaneous spatial-averaged heat
transfer coefficient as h / h0 = 1 + dt + 2dt2 = 1 + dt.
It does not have any term of order greater than because of
the linear model obtained.
In this situation, the energy balance Eqs. 35 and 38
provides the following equations for the heat transfer:

d1

This heat has to be equal to the one calculated from the


convective coefficient by means of the expression hSWTW
T P and hence, by using the averaged thickness of the thermal boundary layer, , the convective coefficient is converted to
2k
2
1 Pe01 e L//41 Pe0.
h=
L

Nu =

0 =

kSWTW T P 1 Pe0

L
1 e

Since the Pclet number is the product of the Reynolds


and Prandtl numbers, the model obtained for the heat transfer
process is

54
55

The time-averaged energy transfer is measured by the


time-averaged exit temperature and hence by s. In addition
the Strouhal number and the dimensionless pulsation are related to each other by means of a characteristic Strouhal
number Stc defined by
St0 =

LA LA
=
= Stc .
G0
G 0t c

56

Since the mean thickness of the thermal boundary layer


is given by , the volume V P can be approached by V P
= SW, and hence the characteristic time by tc = c / h0 so that
the characteristic Strouhal number leads to
Stc =

L A
.
SW

57

This solution shows that the heat transfer enhancement is


dominated by the relative phase between the pulsating mass
flow rate and the displacement induced over the wall. In
particular, the expression has an extreme at the frequency
given by Eq. 39 with 2 = 0.
If the solution given for small frequencies coincides with
the one obtained for nonpulsating flows, 0 should be zero
and, hence, the averaged phase between the velocity near the
inlet port and the wall should accomplish

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-8

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

cosa+ d+ =

Z + 1 + Y d+
ZZ + 1 Y a+
2
.
2
Z 1 + 2Y d+ Z 1 + 2Y a+
58

It is convenient to note that the condition 0 = 0 is not


attainable when 0 approaches to 1. In fact, near this point of
maximum efficiency, 1, 0 1, Z e 1, and Y
/ 2 1 hold and hence, the last expression does not produce any real solution for the cosine except when a+ = d+.
This establishes that the heat transfer enhancement that will
be obtained in Sec. VIII does not exist for high efficiency
devices.
For those cases where the pulsation on the wall velocity
coincides with the mass flow rate pulsation, i.e., 1 = 0 and
a+ = d+, there is no any effect in the heat transfer process at
any frequency, just as occurs for straight channels with completely developed flows. This result reproduces the one obtained in Sec. VI.
It is also interesting to note that the result can be an
enhancement of the heat transfer flow if 1 0 and a discouragement if 1 0. The first case is obliged by the causality if the hot wall is placed near the inlet port the phase
difference is positive and less than and the second one if
the distance transversal to the flow from the hot wall to the
inlet port is adequately increased. The maximum enhancement will be produced if the pulsation over the wall had a
phase delay near / 2.
Note that the dominant parameters retained in this model
for the heat transfer process near the wall are the velocity
over the wall and the thickness of the thermal boundary
layer. Since the nonslip condition imposed by the viscosity
produces also a viscous boundary layer, the exact value of
the averaged velocity near the wall is not adequately determined. This crucial problem has been solved in this section
by substituting the exact solution of the problem by an upper
bound of it. Since the maximum enhancement will come
from the maximum heat transfer coefficient, the partial derivative equations of the fluid mechanics have been simplified in order to obtain the maximum attainable heat transfer
coefficient near the wall. As shown, this solution comes from
retaining the thermal boundary layer and from removing the
nonslip condition. Hence, the velocity over the wall is just
related with the velocity in the nonviscid core. This approach
is feasible because the thermal and momentum equations are
decoupled in the incompressible problem presented.
VIII. DISCUSSION AND VERIFICATION FOR LOW
VALUES OF

When 1, the heating efficiency of the device is very


low, just as occurs in some devices of reduced size. This
application can be modeled by the following practical limit:
0 1, 0 1, and Z 2 / Y 1. In this limit, the
above expressions lead to
1
s=
2

St0
Z Stc
,
St0 2
1+
Z Stc

0 + 1

59

FIG. 1. Dimensionless definition of the 2D backward facing step. Not at


scale.

0 = a+2 a+d+cosa+ d+,

60

1 = a+d+sina+ d+.

61

In the case of having no effect at zero frequency Eq.


58 holds, Eq. 59 can be simplified to
s=

a+d+2 a+2 St0/Z Stc


.
2
1 + St0/Z Stc2

62

From Eq. 57, the Strouhal number for maximum enhancement can be expressed as
Z Stc = 2

L A
.
SW

63

This is the ratio of two volumes: the first one is related to


the size of the device, and the second one, with the volume
of fluid affected by the convective heating near the hot walls.
This formula states that the Strouhal number for maximum enhancement is given by St0 = Z Stc, which is of order
of 1 in those cases where LA / SW is of order of 1 and leads
to a maximum enhancement given by
smax =

a+
d+2 a+2 .
4

64

In addition, it states that this kind of enhancement requires that the relative importance of the pulsation over the
mean velocity must be greater in the wall than in the inlet
port, i.e., a+ d+, something that can be implemented by
suddenly increasing the cross-sectional area of the channel.
For example, a backward facing step like the one presented in Fig. 1 exhibits this feature.16 Since the flow channel
duplicates its area, the mean velocity is divided by a factor of
2 this will be derived in Sec. IX, and hence, the amplitude
of the pulsation can be estimated by considering that it
doubles the mean velocity over the wall: a good approach to
the maximum enhancement can be obtained by considering
a+ = 1 and d+ = 2. This produces a maximum value for the
heat transfer enhancement of 43% which compares well with
the previous literature16 which states that the maximum Nusselt number is 42% higher than in the steady case for ideal
water and 44% greater if the properties of the fluid are considered to depend on the temperature.
The device establishing the characteristic length L as
the outlet height is described16 by L = 450 m and SW / A
= 10 and works with ideal water k = 0.598 W K1 m1,
= 103 Pa s, c = 4180 J kg1 K1, and = 998 kg m3 in an
operational point identified by Nu0 = 5.18 and Re0 = 100.
These values give the following nondimensional character-

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-9

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

Dimensionless heat transfer


enhanccement

0.5

Another is the indirect effect given by the displacement of


some fluid characteristic structures such as stagnation points
or contracting or expanding recirculation regions. These phenomena appear only if the geometry of the device is adequate, and add an extra velocity in the wall that, in general,
will have a phase different from the forcing pulsation:
G / G0 = 1 + at with a = a+ei + aei / 2. In Sec. IX A,
we will characterize the longitudinal problem in the core
where the effects due to the viscosity can be neglected, and
afterwards, the viscous problem near the wall.

0.4
0.3
0.2
0.1
0
0

0.5

1
1.5
2
2.5
Strouhal number (St0)

FIG. 2. Comparison between the theoretical dimensionless heat transfer


enhancement solid line and data obtained from direct numerical simulation
Ref. 16 diamond points for different values of the frequency at Re0
= 100.

ization of the stationary heat flow: = 0.074, 0 = 0.071, 0


= 0.96, Z = 26, and Y = 0.34. By looking at the configuration
of the device, it is plausible to assume that the thermal
boundary layer extends over the 5% of the channel height
/ L 0.05. This leads to consider Stc = 0.037, and hence,
the maximum enhancement is obtained for a Strouhal number equal to Z Stc = 0.96. These values have been employed
to feed Eq. 62 whose results are represented in Fig. 2 together with data proceeding from previous numerical
calculations16 for different frequencies.
Discrepancies observed in Fig. 2 between the numerical
calculations and the approach given in this section can be
due to the neglected nonlinear terms, which can introduce
harmonics of order greater than 1 and to the fact of having
neglected the influence of the Strouhal number on the amplitude and phase of the wall velocity. Additionally, the lack of
experimental data produces an extra error whose amplitude
depends on the assumptions of the 2D numerical model,
mainly on the boundary conditions. However, presented results show that the main physical phenomena governing the
pulsating heat transfer process, which have been taken into
account by the present work, are i the existence of an internal characteristic transient temperature, given by the
model that is summarized by Eq. 24, ii the transient response of an internal volume of fluid, given by the model
leading to Eq. 35, and iii the convective heat transfer
driving the transient response of that internal volume, given
by the characteristic limit that leads to Eq. 52.
Besides, it can be anticipated that the results obtained by
considering d+ constant in Eq. 62 will be incorrect if the
device geometry exhibits a hot wall that is not flat in the
direction of the flow. These kinds of geometries need to predict d+ as a function of the Strouhal number, in the most
general way possible.
IX. ESTIMATION OF THE VELOCITY OVER THE WALL

In a general device there are different phenomena conducting to a pulsation on the wall velocity. One of them is
the direct effect that modifies the intensity of the entire velocity field with a delay depending on the device geometry.

A. Longitudinal distribution of pressure imposed by


the nonviscid core

Providing that the Reynolds number is much greater than


1, the mass flow rate along the channel is related with the
pressure by means of the momentum equation applied to the
nonviscid core. The integral equation of momentum in the
longitudinal direction is then expressed as

u 1 3
d x+
t
V

u1uidAi = F1 .

65

SV

Since the fluid has been assumed to be incompressible,


the longitudinal velocity can be obtained from the mass flux
as
u1 =

G
.
Ax1

66

With this change, the integral of volume leads to the


length of the channel, l, and the surface integral, to the dynamic pressures in the inlet and exit ports whose areas are,
respectively, A and AE,
l

dG G2 A
+
1 = F1 .
dt A AE

67

The force over the volume, which includes the one due
to the pressure in the inlet and outlet ports and the one over
the walls, can be modeled by a force linked to the pressure
drop and a drag force. The drag is modeled as a friction
coefficient c f multiplied by the inlet dynamic pressure and
the dimensionless length of the channel,
F1 = PI PEA

c f G2 l
.
2 A L

68

After some algebra, the momentum equation can be


written in a nondimensional form,

G
G
L
pI pE = St00 +
l
G0
G0

cf L A
+
1
2 l AE

. 69

Normally, these devices have an aspect ratio much


greater than 1, i.e., l / L 1, while A AE and c f 1 due to
the pulsation,22 and hence the second term in the square
bracket can be neglected. When the pulsation is taken into
account, the mass flow rate is given by G / G0 = 1 + a, and the
last equation can be expanded up to terms of first order in .
During this expansion, the friction coefficient in the channel
can be considered as constant. In addition, the first term is

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-10

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

related to the longitudinal gradient of pressure and, hence,


the pressure in the nonviscid core is governed by the following equation:
cf
+ St00a + c f a + 02.
2

70
FIG. 3. Dimensionless definition of the 2D two-heated-block channel.

B. Longitudinal distribution of velocity

The solution over the walls that are parallel to the flow
comes from considering the external gradient of pressure
given by the last equation and by retaining the viscosity, such
as the following momentum equation derived from Eq. 42
expresses:
St0 01 + 111 = 1 p +

1
2 2 1 .
Re0

71

By considering the velocity over the wall as

1/10 = 1 + d+ei0 + dei0/2,

72

and the pressure gradient imposed by the inviscid core Eq.


70, the order- contribution of Eq. 71 results
2
10St0 0d+ + 10
1d +

10
22d+ = St0 0a+ + c f a+ .
Re0
73

The spatial derivatives can be removed by using the


dominant component of its spatial Fourier expansions,

1d+ = i1d+ ,

74

2d+ = i2d+ .

75

With this in mind, the pulsating solution evolves to


d+ =

a+
10

St0 i + c f
St0 101i +

22
Re0

76

We are interested in the modulus, which is


d+ =

a+
10

St20 + c2f

22
St0 101 +
Re0
2

2.

77

This expression shows that the velocity over the wall can
be amplified if the Strouhal number coincides with the eigenvalue describing the dominant spatial distribution of the
geometry.
C. Comparison with previous 2D numerical
simulations

This expression takes into account the behavior of the


velocity near the wall and can be used to explain the results
for the device described in the Ref. 22. This is a 2D channel
with two heated blocks working with a pulsating flow of air
see Fig. 3.
Numerical calculations22 of the heat transfer enhancement are plotted in Fig. 4 for two different Reynolds numbers and a fixed amplitude of the pulsation as a function of

the Strouhal number. The global Nusselt number that is represented in Fig. 4 for the full device can be obtained by
taking into account that the Nusselt numbers for the steady
nonpulsating flow are 10.3 and 11.2, at respective Reynolds
numbers of 500 and 700, for the first block and 7.50 and 8.96
for the second block.22
From these results, we can calculate that the Strouhal
number for maximum enhancement is near 5. However, the
isothermal lines22 indicate that the dimensionless thickness
of the region affected by the transient field of temperature is
near 0.25, and therefore / L 0.25, in addition, the hot surface is Sw / A = 3. This allows to calculate Z Stc 2.7 which is
roughly the half of the one observed in Fig. 4. The discrepancies are solved by using Eq. 77.
The value of 1 in Eq. 77 arises from the longitudinal
distribution of the blocks, so that the main longitudinal
wavelength introduced by this configuration is obtained by
considering that one block induces a complete period
2 rad over its dimensionless length 1, i.e, by taking
1 2.
The coefficient 2 is related to the reciprocal of the
thickness of the viscous boundary layer, so that it can be
approached by a nth-power law of the Reynolds number:
2 = k Ren0. For stationary, laminar, and flat boundary layers,
it is known18 that the value n = 0.5 holds. Although, the problem considered here is not flat and neither stationary, the
boundary layer can be modeled as quasisteady since its
thickness is expected to be very low and hence n is fixed to
be 0.5. Considering that the dimensionless boundary layer
thickness is near 5% of the channel height, k can be obtained
from 5001/20.05k / 2 which leads to k 1.4.
The averaged velocity over the wall inside the thermal
layer, 10, is governed by the mean flow over the wall of the
0.25
Dimensionless heat transfer
enhanccement

1 p =

0.20
0.15

Re0=700

0.10

Re0=500

0.05
0.00
0

4
6
8
10
Strouhal number (St0)

12

14

FIG. 4. The effect of the Strouhal number on the heat transfer enhancement
with an amplitude of the pulsation of 0.2 for the device described in Fig. 3.
Points are numerical calculations carried out with a direct simulation at two
different Reynolds numbers Ref. 22 and solid lines are the theoretical
calculation carried out by using Eqs. 62 and 77.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-11

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides


6
5

|dd+|

Re0=700

3
2

Re0=500
1
0
0

4
6
8
10
Strouhal number (St0)

12

14

FIG. 5. Dimensionless amplitude of the pulsation of the velocity over the


wall estimated as a function of the Strouhal number for the device described
in Fig. 3.

channel and by the defect of momentum due to the viscous


layer. Since the thermal boundary layer and the obstacles
have a similar dimensionless thickness 0.25, which is much
greater than the viscous layer 0.05, it is plausible to expect
that the dominant effect was due to the continuity equation,
which takes into account the averaged area of the channel,
E10 = 1,

78

1 l Ax1
dx1 .
79
l 0 A
Note that E is 2 this result was used in Sec. VII to
establish d+ = 2 for the device of Fig. 1 and 0.75 over the
blocks in the device of Fig. 3. However, the second device
has four vertical hot walls that are not present in the first
device. Since these vertical walls are not contemplated by
the averaged velocity given by Eq. 79, it must be corrected
in order to consider their contribution to the averaged velocity over the wall. Every vertical wall generates a stagnation
point in its lower corner that reduces the velocity to zero in
its surrounding. Since the velocity field near a stagnation
point changes linearly with the coordinate, the averaged velocity over the vertical walls is roughly half of the one at the
upper corner. In addition the vertical component of the velocity on the upper corner is reduced considerably by the
deflection imposed by the main stream and the viscous layer
reduces even more this velocity. Hence, we neglect the velocity on the vertical walls: this gives a dimensionless
spatial-averaged velocity of 10 0.9.
Since the pulsation modifies the velocity field, the force
over the blocks is typically amplified by the pulsating flow.
For example, the amplitude of the instantaneous friction factor increases dramatically, easily reaching 50 times the stationary value:22 for a+ = 0.2, Re0 = 500, and St0 6.5, the
friction factor under pulsating conditions is near 35 times the
stationary one, which is typically of order one of 1. We have
used values of c f = 29 for Re0 = 500 and c f = 48 for Re0 = 700.
The plots in Fig. 4 compare the numerical results22 with the
theoretical ones obtained by using Eq. 62 with d+ given by
Eq. 77.
Figure 5 shows the effect of the Strouhal number on the
velocity over the wall. As St0 increases, the gain in the pulE=

sating wall velocity gradually increases to a maximum


around St0 5.6 and decreases afterward. This peak is due to
the dominant spatial harmonic induced by the blocks.
Therefore, in this case, the heat transfer enhancement is
given by the effect of the pulsating mass flow rate on the
velocity field. The effect on the temperature field does not
produces a significant enhancement, although it is slightly
appreciated in the change in the slope in the region near
St0 2.7 for the lower Reynolds number see Fig. 4.
The physics implemented in this scheme is the following: the mean flow pulsation induces a velocity pulsation
over the wall, which depends on the changes in the crosssectional area available to the flow, in the augmentation of
the pressure forces, in the longitudinal velocity distribution
along the channel, and in the viscous boundary layer thickness. The gas dynamic, described mainly by these parameters, substantially increases the pulsating wall velocity for a
given Strouhal number measured by Eq. 77 and results in
the augmentation of convective thermal transport measured
by Eq. 62.
X. CONCLUSION

A theoretical study was carried out for determining the


effect of pulsating flows in channels with different solid
structures on the inside. The work shows that, although the
temperature field is substantially affected by such pulsation,
the heat transfer is not changed for straight channels with
fully development profiles of velocity. The study also shows
that there are changes in the heat transfer where the velocity
profile is not developed, for example, near the pipe entry or
where there are abrupt changes in the internal geometry of
the channels.
Two different geometries of this last case have been
studied in representation of the typical topologies that appears in these types of devices. The first one is a 2D backward facing step with the hot surface parallel to the flow in
an adiabatic channel. In this case, a maximum enhancement
appears for a characteristic Strouhal number defined by the
ratio of two volumes: one related with the size of the device
and the other to the volume of the fluid region supporting
transient effects on the spatial-averaged temperature near the
hot wall. This number results to be quite stable for a given
device. The second one is a 2D channel with two heated
blocks at uniform temperature in an adiabatic channel. In this
case, the maximum enhancement is mainly controlled by the
response of the wall velocity to the pulsating excitation. This
new effect introduces another characteristic Strouhal number
for the device which is mainly defined by the periodical longitudinal distribution of the blocks along the channel. Results
show that this number is also quite stable.
The relative importance of these two physical phenomena establishes the final response for a given device. In particular, the maximum attainable enhancement should be expected when both Strouhal numbers coincide. Besides, it has
been proven that, due to the first effect, the enhancement
exists even when there is no resonant frequency in the system.
Another interesting conclusion is that, while the steady

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-12

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

heat transfer depends on the variations in fluid properties


with the temperature, the heat transfer enhancement does not
depend significantly on them. As a result, the performance of
these devices is summarized in an equation which shows that
the enhancement is easily obtained in those devices whose
efficiency is low if the pulsation over the wall has the correct
amplitude and phase.
Although the validation has been made in straight and
constricted 2D channels operating with a pulsating flow superimposed on a continuous and constant mass flow rate, the
conclusions of the theory are quite general because of the
integral formulation of the physical laws. This allows the use
of the main results for complex three-dimensional geometries.
Additionally, by using an upper bound of the convective
heat transfer in the boundary layer, it has been concluded that
the main information relating to the details of the device is
the closure model for the characteristic distribution of the
velocity over the wall, as a function of the inlet flow pulsation, and the averaged thickness of the thermal boundary
layer. This result reveals that this model is very sensitive to
the geometry and to the flow patterns in the channel, especially when stagnant or recirculation regions are involved. A
comparison with numerical calculations obtained from the
opened literature shows that the results agree well, at least
for the cases presented, if these models are correctly implemented. It also shows that, as expected, the results are very
dependent on the pressure over the blocks under pulsating
conditions because the pressure field determines the spatial
distribution of the velocity. Therefore, the use of the model
with complex geometries requires the correct estimation of
the velocity near the wall, that requires a good approach for
the drag force and a good knowledge of the spatial distribution of the velocity; both could be the object of future studies
for particular configurations.

NOMENCLATURE
a dimensionless measure of the flow rate
oscillation
A inlet cross-sectional area of V
AE cross-sectional area of the exit port
b dimensionless measure of the temperature
oscillation
c specific heat capacity
c f friction coefficient
d dimensionless measure of the oscillation of the
velocity near the wall
Ec Eckert number
F force exerted over the fluid
G dimensional mass flow rate
h convective heat transfer coefficient
k thermal conductivity
L characteristic length of the device
l longitudinal length
m dimensionless measure of the first-order term of
the heat efficiency oscillation
Nu Nusselt number

P
p
Pe
Pr
Q
R
Re
S
s
St
Stc
SW
T
t
T0
tc
TL
TH
Tp
u
V
Vp
x
Y
z
Z

0
p

dimensional pressure field


dimensionless pressure field
Pclet number
Prandtl number
heat
radius
Reynolds number
boundary of the control volume
dimensionless measurement of the second-order
heat efficiency oscillation
Strouhal number
characteristic Strouhal number
area of the solid surface on the boundary of V p
dimensional temperature field
time
spatial-averaged stationary temperature of V p
characteristic time that measures the delay of
the device response
inlet low temperature
outlet high temperature
spatial-averaged transient temperature of V p
dimensional velocity field
control volume
volume inside V where the spatial-averaged internal energy is time dependent
spatial vector
dimensionless coefficient affecting the secondorder term in the transient temperature
dimensionless coefficient measuring the stationary jump of temperature imposed by the wall
dimensionless coefficient comparing pulsation
effect on the heating efficiency with the pulsation effect on the wall temperature jump.
dimensionless coefficient that measures the influence of Re in Nu
dimensionless coefficient measuring the influence of the velocity oscillation in the secondorder term of h
dimensionless coefficient that measures the influence of Pr in Nu
dimensionless coefficient that measures the influence of T in h
characteristic thickness of the thermal boundary
layer
positive dimensionless number used for the
mathematical treatment
dimensionless coefficient that measures the influence of T in c
dimensionless measurement of the heating
efficiency
dimensionless coefficient related to the temperature field
dimensionless measure of the temperature jump
near the wall at stationary conditions
dimensionless measure of the temperature jump
near the wall at oscillating conditions
dynamic viscosity
dimensionless pulsation

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

094907-13

dimensionless coefficient that measures the influence of T in k; also is the dimensionless spatial coordinate
density
dimensionless coefficient that measures the influence of T in
dimensionless time
dimensionless velocity field
phase of a complex number the number is indicated by a subindex
ratio of the stationary efficiency to the stationary dimensionless jump of the temperature near
the wall
dimensionless coefficient that measures the influence of T in
dimensional angular pulsation
Subindexes
+, indicate complex conjugated numbers
i , j indicate the spatial coordinate
0 indicates stationary conditions; also indicates
temporal coordinate
S. Backhaus and G. W. Swift, J. Acoust. Soc. Am. 107, 3148 2000.
D. L. Gardner and G. W. Swift, J. Acoust. Soc. Am. 114, 1905 2003.
3
Y. Ueda, T. Biwa, U. Mizutani, and T. Yazaki, J. Acoust. Soc. Am. 115,
1134 2004.
1
2

J. Appl. Phys. 105, 094907 2009

Efrn Moreno Benavides

E. M. Benavides, J. Appl. Phys. 1019, 094906 2007.


E. M. Benavides, J. Appl. Phys. 9911, 114905 2006.
6
R. S. Wakeland and R. M. Keolian, J. Acoust. Soc. Am. 116, 294 2004.
7
R. S. Wakeland and R. M. Keollian, J. Acoust. Soc. Am. 115, 2873 2004.
8
M. R. Mackley and P. Stonestreet, Chem. Eng. Sci. 5014, 2211 1995.
9
C. T. Lee, M. R. Mackley, P. Stonestreet, and A. P. J. Middelberg, Biotechnol. Lett. 23, 1899 2001.
10
A. P. Harvey, M. R. Mackley, and T. Seliger, J. Chem. Technol. Biotechnol. 78, 338 2003.
11
J. C. Yu, Z. X. Li, and T. S. Zhao, Int. J. Heat Mass Transfer 47, 5297
2004.
12
M. Faghri, K. Javadani, and A. Faghri, Lett. Heat Mass Transfer 6, 259
1979.
13
H. N. Hemida, M. N. Sabry, A. Abdel-Rahim, and H. Mansour, Int. J.
Heat Mass Transfer 45, 1767 2002.
14
H. W. Cho and J. M. Hyun, Int. J. Heat Fluid Flow 114, 321 1990.
15
H. Chattopadhyay, F. Durst, and S. Ray, Int. Commun. Heat Mass Transfer
33, 475 2006.
16
A. Velazquez, J. R. Arias, and B. Mendez, Heat Mass Transfer 51, 2075
2007.
17
D. R. Lide, CRC Handbook of Chemistry and Physics: A Ready-Reference
Book of Chemical and Physical Data: 20042005, 85th ed. CRC, Boca
Raton, FL, 2004.
18
H. D. Baehr and K. Stephan, Heat and Mass Transfer Springer-Verlag,
Berlin Heidelberg, 2006.
19
Z. Guo and H. J. Sung, Int. J. Heat Mass Transfer 40, 2486 1997.
20
X. Ni and N. E. Pereira, AIChE J. 46, 37 2000.
21
G. Hafez and O. Montasser, 11th International Mechanical Power Engineering Conference, Cairo, 57 February 2000 unpublished, pp. H128
H137.
22
S. Y. Kim and B. H. Kang, Int. J. Heat Mass Transfer 41, 625 1998.
4
5

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
193.52.108.46 On: Wed, 15 Jan 2014 12:24:30

You might also like