Você está na página 1de 14

Pergamon Chemical Engineering Science, Vol. 50, No. 14, pp.

2211 2224, 1995


Elsevier Science Ltd
Printed in Great Britain
0009-2509/95 $9.50 + 0.00
0009-2509(95)00088-7
HEAT TRANSFER AND ASSOCIATED ENERGY DISSIPATION
FOR OSCILLATORY FLOW IN BAFFLED TUBES
M. R. MACKLEY and P. STONESTREET
Department of Chemical Engineering, University of Cambridge, Pembroke Street, Cambridge CB2 3RA,
U.K.
(Received 15 Augus t 1994; accepted in revised f o r m 13 February 1995)
Abstract--We report experimental data on the heat transfer performance of a periodically baffled tube
subject to both steady (net) flow and oscillatory flow. The data show that, in particular, at a low net flow
Reynolds number, significant heat transfer enhancement can be achieved with the superposition of fluid
oscillations. A general correlation is derived for the measured Nusselt number as a function of both net flow
and oscillatory Reynolds number. Dynamic pressure drop data for oscillatory flow are also reported, and
estimates of energy efficiency for obtaining heat transfer enhancement made from these measurements are
compared with smooth wall turbulent flow equations. For large amplitudes of oscillation (equivalent to half
the tube diameter) the overall power dissipation follows the quasi-steady theory. At smaller amplitudes of
oscillation the power dissipation was larger than predicted by the quasi-steady theory, indicating an
increased eddy interaction.
INTRODUCTION
Heat transfer by forced convection in tubular systems
is dependent on the flow conditions in the tube and
the heat transfer characteristics for turbulent and
laminar flow in smooth walled tubes are well estab-
lished and predicted by the Dittus Boelter and Sieder
Tate correlations respectively [see for example Kay
and Nedderman (1980), Holman (1976)]. When the
flow is fully turbulent, rates of heat transfer are rela-
tively high due to the presence of radial mixing. At
intermediate and especially, lower tube velocities
where the flow is laminar, radial flow is not present
and rates of heat transfer are correspondingly re-
duced.
A method for improving tube heat transfer is to use
static mixer inserts which can modify the flow, and
promote radial mixing (an improvement in conduc-
tive heat transfer probably occurs as well). Designs
vary, ranging from a simple helical coil insert, to
a more complex design such as proprietary "wire
matrix" inserts (Oliver and Soji, 1993). While this
approach is often successful, the flow conditions ob-
tained with a particular static mixer is ultimately
a function of the bulk (net) flow velocity, and control
of heat transfer is dictated by this parameter.
An alternative method for efficient mixing in tubes
is to apply an oscillatory flow to tubes which are
periodically baffled. The oscillatory flow promotes
chaotic mixing in the tube, of which radial velocity
components are significant (Brunold e t al . , 1989;
Howes e t al., 1991). An advantage of this system is
that mixing effects are decoupled from the mean flow
velocity, since it is determined by the superposition of
the fluid oscillations which can be precisely controlled
by varying frequency or amplitude of oscillation.
The use of oscillatory flow for heat transfer en-
hancement has already been investigated by Baird e t
al . (1966), and by Kiel and Baird (1971), who exam-
ined the effect on the overall heat transfer coefficient
of a shell-and-tube heat exchanger by flow oscillations
on the tube-side. The tubes were unbaffled and a small
enhancement effect was observed, and correlated to
oscillation frequency. Examples of similar experi-
ments can be found in the literature, see for example,
West and Taylor (1952), Mizushina e t al . (1972), and
Gupta e t al . (1982). More recently, heat transfer en-
hancement was investigated by Mackley e t al . (1990).
Fluid oscillations were superimposed on a steady net
flow on the tube side of a shell-and-tube heat ex-
changer, in which periodically spaced orifice-type
baffles were inserted. A sevenfold increase in tube-side
Nusselt number was achieved (relative to steady, un-
baffled flow) in a tube subjected to a low Reynolds
number, R e , bulk flow ( R e < 200).
The aim of this present study is to extend the
preliminary findings of Mackley e t al . (1990). This
paper reports the results of experiments to determine
the heat transfer enhancement that could be achieved
under both steady and oscillatory flow conditions for
a single phase fluid in a conventional shell-and-tube
heat exchanger containing periodic baffles. A full
range of frequencies and amplitudes are explored and
a heat transfer correlation derived. The power dissipa-
tion required to achieve the oscillatory mixing is
evaluated and related to the heat transfer enhance-
ment. The power dissipation of oscillatory flow is
compared to quasi-steady theory (Jealous and
Johnson, 1955). Power dissipation is also determined
in relation to the heat transfer enhancement, and
compared to the effect of turbulent flow mixing in
smooth walled tubes.
2211
2212
BACKGROUND AND THEORY
Oscillatory flows can be characterised by three
parameters. The first is the oscillatory Reynolds num-
ber or peak Reynolds number given as
Reo = XocoDv i (1)
where Xo is the centre-to-peak amplitude, co the fre-
quency of oscillation, D is the tube diameter, and v the
kinematic viscosity. If there is an additional (external)
net flow in the tube, a net flow Reynolds number is
defined as:
R e , = U D v - 1 (2)
where U is the mean net flow velocity. In steady flow
the Reynolds number is the indicator of the relative
importance of inertial and viscous forces. However,
when there is unsteady flow, the relative importance
of these two forces is not constant, and an additional
parameter is used to describe the flow, called the
Strouhal number
D
Sr = (3)
4~rxo "
I n this form the Strouhal number represents an
amplitude ratio, as used by Sobey (1980), rather than
the alternative Xoco/U for smooth tubes [see Schlicht-
ing (1979)]. We choose an orifice type baffle geometry
that has been shown in previous work to be effective
for fluid mixing and heat transfer (Brunoid e t al., 1989;
Mackley e t al., 1989). The baffle spacing employed is
1.5 tube diameters, and the constriction ratio of each
baffle is S = 0.6, where S = (Do~D) 2, D is the tube
internal diameter, and Do is the orifice diameter.
HEAT TRANSFER EX PERI MENTS
The heat transfer efficiency can be determined by
evaluating the heat transfer coefficient h, expressed as
a dimensionless number, the Nusselt number, Nu:
ht D
Nu , - (4)
k
where D is the tube i.d., k is the thermal conductivity
of the fluid, and h, is the (tube side) heat transfer
coefficient. This is determined from the measured heat
transfer coefficient, for the shell-and-tube heat ex-
changer, given by the following equation:
1 1 D O l n ( Of f D)
Uo - h, 1- ~ + 2 k ~ (5)
Uo is the overall heat transfer coefficient (ignoring
"fouling factors"), k~ is the thermal conductivity for
the tube wall material (stainless steel), D~ is the outside
tube diameter, and h~ is the shell side heat transfer
coefficient. Uo is related to heat flux as follows:
Q = A U o A T l m : Qoi l = m' Cl , ATo i l (6)
where ATu. is the log mean temperature difference
(LMTD) of the heat exchanger, A To, the temperature
difference of tube side oil over the tube length, m' the
M. R. MACKLEY and P. STONESTREET
time averaged mass flow rate of the tube-side oil, Cp
the specific heat capacity of the oil, and A the total
area for heat transfer based on the tube i.d. For the
entire length of the tube, the LMTD is given by
AT2 - ATx (Tout - T1) - (Tin - T2)
ATI m-
I n [ A T 2 / A T 1 ] l n[ ( Tout - T O / ( T i . - T2)]
where Tout and Ti. are the oil temperatures leaving
and entering the heat exchanger, respectively, and T~
and T2 are the water temperatures entering and leav-
ing the shell side, respectively. Note: it is assumed that
all the temperatures are time averaged bulk temper-
atures. Substituting into eq. (6), this gives the follow-
ing equation for Uo:
Uo __ m ' C p A T o i l l n[ ( To~t - Tx ) / ( Ti . - 7"2)] (7)
A (ATo, + ATwater)
Uo is then determined experimentally and the tube
side Nusselt number is calculated using eq. (5) as fol-
lows:
1 k _ o . [ _ 1 O O l n ( O , / O ! ]
N u t - D LUo O~ , 2ks, l ' (8)
To calculate the wall resistance for 1 mm stainless
steel, it was assumed that k~s = 16. The shell-side heat
transfer coefficient, h,, is assumed constant and is
estimated by plotting 1/Uo vs 1~Re m, where m is deter-
mined iteratively, h~ was estimated at 1800Wm -2
C- 1 at the intercept 1~Re '7 = 0 which corresponds
to infinitely small tube side resistance. This was in
close agreement with a value obtained by Mackley et
al. (1990) for earlier experiments.
Comparison of the experimentally determined
Nusselt number, Nu t will be made with existing cor-
relations for smooth walled tubes.
Re,, < 2100
Nu , = 1.86 [ R e P r ( D / Z ) ] ~/3(#/#w)'a4
(Kay and Nedderman, 1985)
2100 < R e < 10,000
N u , = 0.116IRe - 1 2 5 ] ( Pr ) :/3
x [1 + ( O/ Z) Z/ 3]( l a/ l aw) ~"
(Levenspiel, 1984)
R e > 10,000
Nu t = O. 023( Re) ' a( Pr ) l / 3( #/ #w) ' 14
(Holman, 1976).
EQ UI PMENT
A l aborat ory scale shell-and-tube heat exchanger
was used for both the heat transfer experiments and
the power dissipation experiments. A schematic dia-
gram of the experimental apparatus is shown in Fig. 1.
The heat exchanger was installed in a horizontal ori-
entation to be used in single pass, counter-current
mode. The tube used was a 12 mm i.d. stainless steel
pipe with a wall thickness of I mm. The active length
o il return
water in
Heat transfer and associated energy dissipation
[ i i i i i i i i i i i l
12mm i. d . b a ffle d tu b e
2213
Isometric of baffles
tube insert
18 rnm i ~ 1 1 ~ 7 r n m
11.5 rnm
O scilla tin g
P istons
t f
Oscillation d ire ctio n
W ater o u t
" " [
flowmeter
, TI
D r iv e h u b
D isplacem ent
transducer
KEY
Ti~ermoco,Jple
q- - D Pressure b'ansducer
non re tu rn valve
~ - - e le c tro n ic lin k
T ube in te rn a l d ia m e te r: 1 2 .0 rnrn
CHURCHILL o il heater/pum p
Fig. 1. Schematic diagram of the apparatus used for heat transfer and pressure drop experiments.
was 1 m. The shell diameter was 80 mm. Tap water at
mains pressure was used in the shell as the cooling
medium. A constant flow rate of approximately
101min-~ was maintained on the shell side.
A mineral oil, Shell Tellus R10, was used as the
active fluid in the tube. The oil is a Newtonian fluid
and has the properties of p = 840 kgm -3, k = 0.137
Wm- ~ C -1, and / ~=0. 005Pas at 50C. A Con-
air-Churchill oil heater-circulator was used for the
oil heating and bulk flow transport in the heat ex-
changer tube. Fluid oscillations within the tube were
provided by an electric motor driving two horizon-
tally opposed pistons, operated by a connecting rod
and yoke arrangement. Hydraulic hosing of equal
lengths connected each piston to either end of the heat
exchanger tube. As the motor turned, the crank and
yoke caused one side to apply a positive pressure to
the fluid while the other moved exactly 180 out of
phase with the first piston to apply a negative pres-
sure. The speed (frequency) was adjusted electroni-
cally, and the amplitude was set by adjusting the
connecting rod pivot point on the motor hub. The
resultant time-dependant displacement was measured
by a calibrated displacement transducer attached to
the piston yoke.
Four K-type, 2 mm diameter thermocouples were
used to measure temperature. The specification of
each thermocouple-instrumentation was 0.IC res-
olution, 0.2% accuracy in the range 0-100C, and
a response time of approximately 0.4 s to approach
65% of a 100C step change. The calibration of each
thermocouple was checked before installation for lin-
earity, and once per week for drift. Each ther-
mocouple was inserted together with a short section
of dense wire mesh in the tube to ensure that an
average bulk temperature was measured. The temper-
atures measured were the oil inlet-outlet to the tube,
and the water inlet-outlet to the shell. The volumetric
flow rate was determined using a "Litremeter" helical
gear, positive displacement flow meter, which is able
to measure the absolute value of flow rate, indepen-
dent of viscosity effects. The resolution of the flow
meter was 0.01 I mi n- 1, with an accuracy of0.1% and
a repeatability of 0.01% for the range 0-10 l mi n-~.
The analogue signals corresponding to temperature
and flow rate were connected to an analogue-to-
digital converter (ADC), which, in turn, was con-
nected to an IBM-compatible 486 computer. The
digital resolution was 0.1% of the full scale of any
measured value. This recorded the data from each
channel on to a disk ASCII file. All heat transfer
results were calculated by a Pascal program and the
output data files were subsequently manipulated by
graphics software.
EXPERIMENTAL METHODS
Before a set of experiments, the oil heater and pump
were activated, and the oil was allowed to re-circulate
until the desired temperature of 70C was attained.
A constant net flow rate through the heat exchanger
was set using the flow meter and adjusting inlet and
outlet valves to maintain a severe flow constriction at
each valve. These constrictions, coupled with a non-
return valve at both the inlet and the outlet, ensured
(as far as possible) that the propagation of the oscilla-
tion occurred only in the heat exchanger. This was
confirmed by a steady-reading observed on the up-
stream flow meter. Note: it is possible that, while the
2214
time averaged net flow was constant, a slight pulsing
of the net flow occurred because of the effect of the
main tube oscillations on the non-return valves (unde-
tectable by the flow meter). However, this would
probably not affect the temperature measurements,
which were made in a section of insulated pipe where
no heat transfer occurred.
For all the experiments, a constant shell-side water
flow rate was maintained, and it was assumed that the
shell side heat transfer coefficient was constant for
every run. For each change of flow conditions (oscilla-
tory or steady net flow), the average Nusselt number
for the tube was calculated from the recorded data
corresponding to steady-state operation, i.e. where the
thermocouple and flow meter values had become
steady. Furthermore, values of p, la, Cp, Re, , Reo, and
Pr were calculated on the basis of the length averaged
tube side temperature at the particular steady net flow
rate of the oil within the active length of the heat
exchanger tube.
RESULTS AND DISCUSSION
The first set of experiments were performed to de-
termine the heat transfer characteristics of the stan-
dard case of smooth tube, and to determine the effect
of the baffles relative to the smooth tube where the
flow of oil in the tube was steady (non-oscillatory).
Figure 2 shows the relationship of the (tube-side) Nus-
selt number, Nu, , to Re . for the non-oscillatory mode
of operation. It can be seen that in both cases Nu,
increased as the steady flow, Re . , increased but the
effect was much greater for the baffled tube. The
M. R. MACKLEY and P. STONESTREET
maximum Nusselt number obtained was Nu, = 95 at
Re. ~ 800, which is a fivefold improvement over the
standard case at the same value of Re. . With steady
flow only, the addition of baffles to a smooth tube
results in a significant improvement in heat transfer.
These results are in general agreement with those
given by Mackley et al. (1990). The similar effect has
been also observed with the use of proprietary tube
inserts, such as wire matrix inserts investigated by
Oliver and Soji (1993).
The second set of experiments was performed to
investigate the effect of oscillatory flow on the heat
transfer characteristics of the baffled tube relative to
a smooth tube. In all the experiments, a small, fixed
steady-flow component was maintained, correspond-
ing to Re, = 130. For the first oscillatory flow experi-
ment, the amplitude was fixed at 6 mm ( St = 0.16) and
the frequency varied from 2-10 Hz to give a range of
Reo. Figure 3 shows the results of these experiments
and demonstrates that significant heat transfer en-
hancement occurs when fluid oscillation is combined
with the presence of baffles. At Reo = 760 it can be
seen that the Nusselt number for baffled-oscillatory
flow is of order 95, whereas for the smooth tube it is of
order 20. Compared to earlier experiments (Mackley
et al., 1990), the Nu values for the baffled-oscillatory
flow are of order 20% greater for the same oscillatory
Reynolds number (Reo), although the same general
behaviour is observed.
Further experiments were performed by selecting
different oscillation frequencies at various fixed ampli-
tudes. Figure4 shows the results as a plot of the
100
=-
Z
o
o~
Z
80
60
40
20
O
O
o
o
oo
o
o O
, o 0 0 0 0 0
~ , o o
. . , ~ 8 d ~ oo~O o o o
o o O t 1 ~ ~ "
, , , I h , , I , I I I
200 400 600 800 1000
Net flow Reynolds number, Re n (tube)
o smooth tube
o baffled tube
0 i
0 1200 1400
Fig. 2. Comparison of the heat transfer obtained for a smooth tube and a baffled tube for steady
(non-oscillatory) flow.
=-
Z
.g
E
e-
o~
Z
100
80
60
40
20
Heat transfer and associated energy dissipation
I I I
o
o
o
B
o
[] Baffled tube
smooth tube
J ~ , I t ~ J I , , t [ J , , I J ,
0 2 0 0 4 0 0 6 0 0 8 0 0 1000
Oscillatory Reynolds number, Re
Fig. 3. Comparison of the heat transfer obtained in a smooth tube and a baffled tube where an oscillatory
flow is superimposed on a low Re bulk flow: Ren = 130, Sr = 0.16.
2215
100
Z
o
e~
Z
10
O
X
O
+
X
o . ?
O X A
6
R e n = 1 3 0
O St r = 0. 27 ( Xo= 3. 5 r a m)
X St r = 0. 21 ( x0- - 4. 5 mm)
+ St r - - 0. 18 ( Xo= 5. 2 r a m)
A Str=0.16 (Xo--6.0 ram)
Str=0.14 (x0=6.5 ram)
0. 1 1
O)X
o
Fig. 4. Tube side heat transfer as a function of maximum oscillatory velocity for oscillatory baffled flow
with a low Re bulk flow, where Re, = 130.
Nussel t number , Nut, vs t he maxi mum osci l l at ory
velocity, xoto, obt ai ned for five different ampl i t udes.
On t he di agr am, t he St r ouhal number is i ndi cat ed as
well as t he ampl i t ude. I t can be seen t hat var yi ng t he
ampl i t ude has onl y a smal l effect; al t hough t here is
some scatter, t he dat a poi nt s all lie fairly close t o-
gether. The maxi mum Nussel t number obt ai ned is of
or der Nut = 100 at xoco = 0.41, whi ch is vi rt ual l y
a 10-fold i ncrease over t he non- osci l l at or y resul t (at
t he same Ren). Var yi ng t he frequency has a st r ong
2216 M. R. MACKLEY and P. STONESTREET
effect on the heat transfer. For example, where
Xo = 4.5 mm at a frequency of oscillation of 2 Hz
(COXo = 0.056), the Nusselt number obt ai ned was of
order 17, whereas at f = 11 Hz (ogx o = 0.31), the value
is approximately 93. It is obvi ous that the heat trans-
fer is dependent on oscillatory velocity ffox,), and that
a desired Nu, can be obt ai ned by choosing a high
enough frequency for any given ampl i t ude (obviously
within practical limits).
I n order to determine the effect of the bul k flow rate
on the oscillatory flow heat transfer, experiments were
performed in which the flow rate was varied for each
of a number of different oscillation frequencies at
a fixed ampl i t ude of 6.4 mm. Re, was varied in the
range 100-1200 which was the practical upper limit of
the oil circulator. The results are shown in Fig. 5, in
which dat a are plotted as Nusselt number vs Re, for
a number of different oscillatory condi t i ons in the
baffled tube. The points for Reo values depicted cor-
respond to f = 0, 4, 6, 8, and 10 Hz for Reo = O, 300,
450, 680 and 800, respectively. It can be seen that for
a given Reo, the heat transfer rate increases with
increasing Re,, and that for a given Re,, the heat
transfer is greater for a higher Reo. It can be seen that
at large values of Re., the oscillatory curves tend
asymptotically towards the steady-flow curve, be-
cause the net flow component becomes much larger
relative to the oscillatory flow component for higher
net flow rates I n addition, at low values of net flow,
high rates of heat transfer are still obt ai nabl e
(Nu > 100) provided the oscillatory flow component
is present. For Reo = 800 the first dat a poi nt corres-
ponds to where Re, = 130, and it can be seen that the
Nusselt number is of order 120. For the steady flow,
smoot h t ube case Nu ~ 4 for the same value of Re,,
and thus oscillatory baffled flow resulted in a 30-fold
increase in heat transfer. This is a significant enhance-
ment. I n the extreme case it is expected that if the Reo
curves were extrapolated back to where Re, = 0, good
heat transfer woul d still take place, as the mixing
effect of oscillatory flow is still present. This would
correspond to a batch heat transfer situation.
Figure 5 also shows a correlation which was fitted
to the experimental data. The correlation is as follows:
Nu, = 0.0035 Re~'3 prl/3 + 0.3 [(Re, Re22 1
( 9 )
This fits the dat a well, although it is acknowledged
that it is a purely phenomenol ogi cal model. The first
term of the correlation corresponds to the steady-flow
cont ri but i on to heat transfer, and is deliberately
chosen to be similar to the well-known Di t t us Boelter
t urbul ent flow equat i on, but the exponent of Re, is
greater to account for the presence of baffles in the
tube. Note: the Pr number was calculated for each
experiment, but the dependence of Nu on Pr was not
specifically determined. An expression for the wall
viscosity effect has not been included because the
significance of this term could not be determined
within the scope of the experiments. The second term
accounts for the additional oscillatory behaviour, by
assumi ng that when Reo >> Re,, the effect of oscillation
is superimposed on the steady behavi our by addi ng
the oscillatory term to the steady term. I n the case for
=-
Z
, J ~
Z
100
10
r r i _ ~ - _ _ _ _ ~ , I . . . . . . . . . . . . . . . . . . .
. . + - . -
R%=S00 " -*-+-+- ~ -~- - ~ " ' * + 7 +
- - +- ' - - - +- i t - +- ~ . - - C- ~ / ~ ' O"
+ @ # _ - ~ - - ' ~ - - / . / . ~ 1~
. - - ~ / . /
Re =680~ ~ + + ' - ~ - + - + - ' +
~ 9 . . . . . . - _ Cg- +- ~ " . ~ "
Re =450 , 4-
o * _ - ~ -
. - -
Reo=300 ~- -
/
+
~ ~ - - ~ , / + ~ +"
. " 4:
/
+ ~,
# . /
Steady baffled flow
/ 4-
+/
+ Experimental data
- - - - correlation (Eq. 9)
. . . . . . . i z a . . . . . . . . . . . . t ~ t Z . L .
100 11300
Net flow Reynolds number, Re n
Fig. 5. Experimental and predicted tube side heat transfer as a function of both Reo and bulk flow Re, for
a bamed tube.
Heat transfer and associated energy dissipation
steady flow ( Reo =0) , the equation is simply
N u t = 0.0035 p r l / 3 Re ~. "3, and this is plotted on the
bot t om curve giving good agreement with the experi-
mental points. Similarly, when Re . >>Re o, the oscilla-
tory term is small in relation to the steady term, and
the curve collapses onto the steady flow behaviour.
While obviously the correlation was only fitted for
measured N u values within the experimental range of
R e . and Re o, it displays the correct behaviour that
could be extended somewhat outside the experimental
range of R e . and Reo.
Figure 6 shows a plot of eq. (9) plotted against
standard heat transfer correlations for the flow of the
oil in a smooth tube heat exchanger of the same
dimensions, which gives an overview of the heat trans-
fer regimes corresponding to different types of flow
behaviour. As can be seen from Fig. 6, oscillatory flow
can provide similar heat transfer rates to those pos-
sible with smooth wall turbulent flow, but in the range
R e , = 100-1200. The results in Fig. 6 show one of
the significant process advantages of using oscillatory
flow in a baffled tube. Whereas turbulent flow is
usually required in smooth tube devices to obtain
acceptable heat transfer, Fig. 6 shows that the re-
quired level of heat transfer can be selected by choos-
ing an appropriate value of oscillation amplitude and
frequency. This allows the freedom to choose the
appropriate net flow to give the required residence
time or throughput within a given tube. I n particular,
high heat transfer (i.e. N u > 100) and long residence
times (z ~ rain h) can be achieved in tubes of modest
length (Z ~ meters).
2217
POWER DISSIPATION
M ode l l i n9
The oscillator acts on the fluid to give the following
variations of displacement, x, velocity, u, and acceler-
ation, a:
x = - Xo cos (cot) (10a)
u = xo~o sin (tot) (10b)
a = xoo~ 2 cos (rot). (10c)
The time periodic flow will result in a correspond-
ing unsteady pressure gradient. For oscillatory flows
in smooth tubes, the relationship between velocity
and pressure gradient is evaluated using unsteady
boundary-layer theory, see for example Edwards and
Wilkinson (1971) or Schlichting (1979). For oscilla-
tory flow in a baffled tube, however, we assume that
the frictional component in the flow results mainly
from the cumulative effect of flow through each ori-
rice, rather than from events at the wall. At any instant
in time the pressure along the tube (relative to the
dat um pressure) is assumed to be a linear function of
length. This is illustrated in Fig. 7. For a fluid direc-
tion from left to right, the left side is under high
pressure, while the right is under negative pressure
(the pistons operate 180 out of phase). Under these
conditions the pressure at point 1 shows a positive
peak relative to the datum, and at point 2 a negative
peak (at the same time instant). Thus, for normalised
distance along the tube:
p = p= sin ~b(l - 2f) + p'
( i t )
1ol)o
Z 100
o
e~
E
Z
Z
"~ IO
.8
[...,
' ' ' ' ' ' ' 1 ' ' ' ' ' ' ' 1
Oscillations + ba f f l e s
, ~ Re--3(X)-800
- + . + V + ~ . + ~ f ~ + + . + + - + + * + +
+ ++ ++
~ r i f i c e ba f f l e s onl y

t - e e e
+
Laminar regime
( Si ~ l e r Ta t e )
x
O
Transitional
, , , , , , ,
x
x
x
x
x Turbulent
(Dittus Boelter)
l i i i i L J t i [ i i = J L i L i i i i i J J L i
100 1000 10000 100000
Bulk flow Reynolds number, Re n
Fig. 6. Comparison of oscillatory flow heat transfer correlation to standard heat transfer correlation for
smooth tubes.
2218
I
O
Pl
Pi = psin~
4
~- x g=x/Z
M. R. MACKLEV and P. STONESTREET
dt~
p+dp
Pressure gradient--dp/dg

P2
p2 = -PmSin~
Pm
P
-Pm
~/ p=pm(1-2g), sin~=l
~ ~ _ ~ p=PmSin~(1-2g)
~ p=pm(1-2e), sin~--1
Pm
-Pm
Fig. 7. lllustration of pressure-time behaviour along the length of a baffled tube subjected to oscillatory
flow.
where p' is the mean pressure, f is the normalised
length (0 < d < 1), and sin ~b =f(t ), where the form of
is still to be specified. Thus, for d = 0, p =Pm sin q~,
and ~ = 1, p = - p, sin q~, and we define a time-de-
pendent pressure difference between positions 1 and 2:
Ap(t) = 2pro sin 4-
More generally,
Ap(t) = Apo sin ~b (12)
where Apo is the overall pressure difference between
positions 1 and 2.
The power at any instant is a function of the pres-
sure gradient and the instantaneous superficial velo-
city which is assumed constant over the length of the
tube:
P ( t ) = a c u ( t ) A p o sin ~b (13)
where ac is the cross-sectional tube area. The instan-
taneous power given by eq. (13) is more usefully ex-
pressed by the time-averaged power:
ac u ( t ) A p o sin q~dt
o
P.ve = (14)
T
where T is the period of oscillation (T = 2 n / t o ) . The
power density is obtained by dividing eq. (14) by the
system volume, and substituting eq. (10b) for the fluid
superficial velocity:
acxoto I ~T
sin (tot)Apo sin 4) dt (W m- 3).
e v = V T .Jo
05)
To obtain an expression for ~b, it is convenient to
define a phase shift between pressure and velocity,
defined as 6:
Ap(t) = Apo sin ~b = Apo sin (tot + 3) (16)
where Apo is the maximum pressure fluctuation
(centre-to-peak). Applying a trigonometric expansion
to eq. (16), we obtain:
Apo sin (tot + 6) = Apo sin (3)cos (tot)
+ Apo cos (6)sin (tot). (17)
Substituting eq.(17) into eq. (15), we obtain the
following:
A p o x o t o cos(6) f ~ sin2 (tot) dt
e v- Z
where Z is the tube length (m). After integration, this
reduces to:
1 Apo c o s ( 6 ) x o t o
e,~ = 2 Z (18)
Heat transfer and associated energy dissipation
From which the power density may be determined.
For a particular experiment the phase angle, t$, is
obtained by measuring the time difference between
the peak of the pressure trace and the velocity trace on
a plot of Ap, u, x vs time, and determining the fraction
of a complete cycle (in radians) which this time differ-
ence represents.
Quasi-steady theory f or power density
For oscillatory flows it has been a common practice
to assume "quasi-steady" behaviour (Jealous and
Johnson, 1955). This model assumes that the frictional
pressure drop in a time-periodic flow at a certain
instantaneous velocity is assumed to be identical to
the pressure drop that would be obtained at a steady
velocity of the same magnitude of the instantaneous
velocity. The maximum frictional pressure drop is
obtained according to the standard pressure drop
relation across an orifice (at high Reynolds numbers),
i.e.:
np(tOXo)2(1/S 2 -- 1)
Apl 0 = 2Co2 (19)
where n is the number of baffles and Co orifice coeffi-
cient (usually assumed to be 0.6). The power density
for quasi-steady flow is given by:
2np(toxo)3(l/S 2 - 1)
(20)
ev = 37rC 2 Z
Graphical determination of power dissipation
For the area traced out on a pressure-displacement
diagram, we can define energy dissipation per cycle, Ec
where:
= ac~Ap( t ) dx. (21) Ec
By measuring the area from actual Ap vs x plots, one
is able to calculate the power density
a , acf a , f
e ~ - - - ( 2 2 )
V Z
where a= is the measured area andf i s the frequency of
oscillation (Hz). The value of ev so obtained can be
compared to the value obtained by eq. (18). The two
values should be in close agreement.
EXPERIMENTAL ME T H O D S
Power density (W m-3) was determined for oscilla-
tory flow by measuring the time-dependent pressure
fluctuations in the same heat exchanger used in the
heat transfer experiments. The pressure fluctuations
were measured with two Druck model 810 pressure
transducers. These transducers contain a solid state
silicon strain gauge and thin diaphragm in direct
contact with the fluid. With the associated instrumen-
tation these devices are accurate to within 0.1% of the
full scale capacity (3 bar), and the expected 15 Hz
maximum frequency of the pressure fluctuations are
well within the 2000 Hz bandwidth. The outputs from
2219
the pressure transducer instrumentation and the dis-
placement transducer were connected to an ADC and
a 486 computer was used to record the pressure-time
and displacement-time data to a disk file. The corres-
ponding power density calculations were performed
using a spreadsheet program.
Before each experiment was made, the system pres-
sure was equilibrated by opening the bleed valve to
the system and the transducer conditioners individ-
ually zeroed. The oil was then allowed to circulate
through the tube to force air to escape through the
bleed valve. The bleed valve was then closed. For
oscillatory measurements, it was found necessary to
pressurise the tube to a constant mean pressure of
50 kPa to avoid cavitation.
In order to obtain the pressure-time behaviour
uniquely for oscillatory flow in the active length (i.e.
where heat transfer took place), and also to avoid
effects from various components in the external flow
circuit, all oscillatory flow experiments were con-
ducted for a closed system (no net flow). The pressure
drop at different (steady) net flow rates (i.e. where no
oscillatory flow was imposed) was determined in
a separate experiment. The protocol was to select
a particular amplitude within a range 1-7 mm and
then carry out a frequency sweep from 3 Hz to as high
a value as was practical. Usually, it was possible to
reach up to 15 Hz at small amplitudes (1-2 mm), but
at the larger amplitudes it was not possible to go
beyond 10 Hz, as the pressure fluctuations exceeded
the range of the pressure transducers. In total, fre-
quency sweeps for 6 amplitudes were performed, at
approximately 6, 5, 4, 3, 2 and 1 mm. This corres-
ponded to a Strouhal number range of 0.15-0.9.
RESULTS AND DISCUSSION
Figure 8 shows typical directly measured data of
the pressure and displacement both plotted against
time. The data shown are for a 4 mm amplitude at
10 Hz oscillation frequency. The pressure traces p~
and P2 are the direct output from the transducers and
are shifted upwards on the diagram due to the system
pressure of 50 kPa gauge. Because of the "push-pull"
action of the oscillator, the pressure traces from the
transducers are inverted with respect to each other.
The output from the transducer closest to the com-
pressing piston shows an increase in pressure, whereas
the downstream transducer (closest to the expanding
piston) registers a decrease in pressure relative to the
datum. The thick solid line on the diagram is the net
pressure fluctuation, Ap, obtained by the difference of
the two pressure transducers. Data were found to be
highly reproducible, provided all air bubbles were
excluded from both the oil and the pressure trans-
ducer ports.
From Ap vs t and x vs t data, it is possible to
calculate the power density by eq. (18). It is necessary
to measure the phase angle, 6, and this can be ob-
tained directly from Fig. 8, where 6 is the phase lead of
pressure with respect to velocity, or where x = 0. For
2220 M.R. MACKLEY and P. STONESTREET
100
50
-50
I &P= P 1 - P2
oe
F ;
I I I ~ i!
I, ". "
mi e I
%
f :
%. :
. a
o P2 (gauge) -0- P1 (gauge) . . . . . . . x (ram)
-2
-4
-6
- 1 0 0
- 8
0 0.05 0.1 0.15 0.2 0.25
Time (s)
Fig. 8. Pressure-time, displacement-time and velocity-time data for an oscillatory flow pressure drop
experiment.
>
the dat a presented in Fig. 8, the time difference be-
tween the pressure peak and x = 0 is 0.007 s, and thus
= 0.007 x (2~f) = 0.53 radians. The peak pressure is
approximately 75 kPa, Omo = 0. 24ms -1, and the
power density can then be calculated [eq. (18)] as
7807 W m -3. These calculations are performed for
each experiment.
Figure 9 shows typical pressure vs displacement
curves for oscillatory flow. For the purposes of illus-
tration only three plots are shown, corresponding to
frequencies of 6, 10 and 12 Hz at a fixed amplitude of
3.2 mm, although obviously such figures were ob-
tained for every experiment. I t can be seen that the
shape of the plots approximate a tilted ellipse. An
interesting phenomenon is observed for the data at
6 Hz. The measured plot of Ap vs x collapses to cusps
at each extreme of x. This suggests that no net energy
dissipation occurs over the outer parts of the oscilla-
tion stroke. This may be due to the unwinding of
a vortex around the baffle edge as the flow changes
direction, resulting in some recovery of energy. This
effect is not observed at higher frequencies, at which
the flow is presumably more chaotic.
From the area of each pressure-displacement plot,
the power can be determined. Taking the 10 Hz plot
as an example, the area corresponds to the work done
per cycle, and the power density is calculated accord-
ing to eq. (22) to be approximately 7750 W m- 3. This
agrees well with the calculated value of 7807 W m- 3
calculated from eq. (18). A number of such calcu-
lations were performed, all with good agreement be-
tween the two methods, and consequently the power
density for the rest of the experiments was calculated
according to eq. (23), which is a simpler method than
measuring the area for each Lissajou diagram for each
experiment.
Dat a for all the power density experiments, corres-
ponding to frequency sweeps at six different ampli-
tudes are shown in Fig. 10 (note: this is for Re , = 0).
Also shown is the power density that would be ob-
tained from the quasi-steady state time-averaged
power model [eq. (20)]. It can be seen that the results
for the highest amplitude (6.4 mm) are only slightly
greater than the quasi-steady model prediction; how-
ever as the amplitude is reduced the frictional pressure
drops generally exceed prediction. This trend was also
observed by Hafez and Baird (1978) in their measure-
ments of the power consumption in a reciprocating
plate column. I n that case the operating frequencies
did not exceed 3.5 Hz and the lowest amplitude was
3.5 mm, so the effect was less pronounced than in the
present work.
It is interesting to relate the observed power density
behaviour to the Sr number, where the flow is non-
quasi-steady, the amplitude is small and therefore the
Sr large (0.5-0.9). At larger amplitudes the quasi-
steady behaviour is approached, and the Sr numbers
are correspondingly small. This agrees with the work
of Sobey (1983) who investigated flow separation with
oscillatory flow in furrowed channels. Although the Sr
employed by Sobey is defined slightly differently, it
was nevertheless found that at sufficiently low Sr
numbers the flow was quasi-steady, while at larger
Strouhal val,ms, the flow was described as an "un-
Heat transfer and associated energy dissipation 2221
70
50 ~r , ' j ~ - ' ~ ~ ~ ' ~
.lO
-70
-4 -3 -2 -1 0 1 2 3 4
Displacement, x (ram)
Fig. 9. Pressure-displacement plots for three different oscillatory flow pressure drop experiments, where
xo = 3.2 mm.
10000
1000
~ ~oo
10
Quasi-steady
, , , , i [ , , , , ,
0.1
COx o
Fig. 10. Power density as a function of maximum oscillatory velocity for experimental data and predicted
by the quasi-steady-model.
steady separated flow". I t is possible that similar re-
gimes exist for oscillatory flow i n baffled tubes. Where
Sr = 0.16, the power density indicates the flow follows
more or less the quasi-steady prediction, whereas for
Sr > 0.2, the quasi-steady prediction is not valid.
I n order to relate the heat transfer obt ai ned in the
previous experiments to the energy requirements i n
oscillatory flow, the tube-side Nusselt number can be
plotted against a normalised power density. This was
done as follows: At a part i cul ar value of toxo, the
2222
corresponding Nu value is obtained from Fig. 4, and
the oscillatory flow power density from Fig. 10. Be-
cause the data in Fig. 4 had a steady-flow component,
Ren = 130, we have to add the steady-flow power
density component to each value of power density
from Fig. 10. The steady-flow power density for the
net flow rate corresponding to Ren = 130 was meas-
ured as 35 W m-3, but when coupled with the oscilla-
tory flow, a correction has to be made for non-linear
interactions in the tube. This can be approximated by
calculating an enhancement factor for the steady-flow
power density given by (Baird, 1994):
{ 1 + [4 ReofiRe. x)] 3 } 1/3. (23)
Using this formula, the enhanced net flow power
density was calculated and added to the oscillatory
flow power dissipation for each value of cox.
It is also useful to rank the heat transfer and power
density performance for oscillatory flow alongside the
performance for a conventional laminar or turbulent
flow. To determine the power density relationship for
laminar and turbulent flow regimes in smooth pipes,
the power density is calculated for a specific Re. using
the friction factor equations (Levenspiel, 1984)
For laminar flow:
Re2,, i a3
ev = 32
For turbulent flow:
(Re./~)3
e~ = 2/ s Dap2
(130 ~ Ren ~ 2100).
(2100 ~ Re. <~ 20000).
M. R. MACKLEY and P. STONESTREET
)e I is the fanning friction factor obtained from a fric-
tion factor chart for a smooth walled tube. Using
these equations, and the standard heat transfer cor-
relations for Nusselt number, we plot Nu vs power
density for a given net flow Reynolds number, Re.. All
the experimental and calculated ev values were nor-
malised by dividing by ev = 6.5 W m-3, which corre-
sponded to the power required to maintain a flow rate
of Re, = 130 in a smooth tube. This value of Re. was
used for the oscillatory flow experiments (see Fig. 4),
and thus it was considered appropriate to use the
steady-power value at Re. = 130 as the reference
value in all cases. Note: We were only interested in the
power density values for the flow regimes in the active
length of the baffled heat exchanger tube. Any addi-
tional frictional losses in the external flow circuit are
characteristic of the specific design of an external flow
circuit. Also, we did not attempt to evaluate the elec-
trical or mechanical efficiency of the oscillator unit,
which is a feature inherent to the mechanical design of
the unit.
Figure 11 shows the Nu vs normalised e,v data for
oscillatory flow at two different amplitudes for
Ren = 130, and data for turbulent and laminar flow
regimes obtained as described previously. Additional
data is shown for steady baffled flow. It can be seen
from Fig. 11 that oscillatory baffled flow, where
(24) xo = 6,4 mm, gives the best heat transfer response (the
highest Nu) for a given power number for all the flow
regimes shown. Oscillatory flow at this amplitude has
been observed to approach quasi-steady behaviour
(25) (see Fig. 10). In contrast, where Xo = 3.5 mm, the Nu
achievable for a given power input was less than for
Z
Z
O3
100
10
' ' ' ' ' ' " 1
/
x
al l ~
P
aW
0
c' ,
, , , , i , i i I f , , i f , , , ] J i , r r l ~ E[ r ~ T
w
/
f
i x" ~ Xo= 6.4 mm
-x steady flow
s mo o t h t ube
- Turbulent flow
. . . . La mi na r f l OW
i i J , , ~ l l i , i I J , l , I i _~ i i 1, 1, 1 i i J t t J , , I , i i i l l
1 10 100 1000 10000 100000
Relative power density(~v/g ' Re=130 )
Fig. 11. Comparison of the oscillatory bamed flow heat transfer as a function of power density with
predicted behaviour for smooth tube steady-flow heat transfer.
Heat transfer and associated energy dissipation
either oscillatory flow at Xo = 6.4 mm, or for turbulent
flow in a smooth tube. For this amplitude the flow
was observed to be non-quasi-steady.
For the steady flow with baffles the data seem to
fall in a transition region, eventually coinciding with
the quasi-steady oscillatory flow curve (xo = 6.4 mm)
at high power values. This indicates that for non-
oscillatory flows in tubes, there is a clear advantage
for the use of baffles in favour to a smooth tube, as the
heat transfer performance obtained would be higher
at a given power input than for either laminar or
turbulent flow in a smooth pipe.
In general terms, it may be observed from Fig. 11
that oscillatory flow in a baffled tube gives a superior
heat transfer performance than a turbulent flow for
a given power input. Furthermore, for power values
below 800, oscillatory flow can produce a Nusselt
number performance that is not accessible to conven-
tional steady pipe flow.
CONCLUSI ONS
It has been shown that oscillatory flow leads to
a substantial enhancement in tube-side heat transfer
in a shell-and-tube heat exchanger. While the presence
of baffles in a tube results in heat transfer enhance-
ment, the greatest enhancement is obtained where
both baffles and oscillations are present. For a broad
range of oscillatory conditions tested, it was found
that the heat transfer rate was strongly dependent on
the product of frequency and amplitude of oscillation
and, by choosing a particular frequency and ampli-
tude, precise control of heat transfer enhancement can
be obtained.
The greatest advantage of oscillatory flow appears
to be found at low net flow Reynolds numbers (low
tube flow rates). It was seen that a 30-fold improve-
ment in Nusselt number could be obtained under
certain conditions. A correlation was fitted to the
experimental oscillatory flow data, and this could be
used with confidence to predict the Nu within the
range 100 < Re, < 1200, and 0 < Reo < 800. The cor-
relation for heat transfer presented in this paper
should enable design calculations to be carried out for
heat exchange. This, coupled with residence time data
(Dickens et al., 1989) and data on the particle suspen-
sion characteristics of the type of flow (Mackley et al.,
1993) should now provide sufficient information to
enable pragmatic reactor design for single and some
mixed phase reactions.
The heat transfer and power density data presented
here illustrates the similarities and differences of oscil-
latory flow in baffled tubes to other types of flow. At
high net flow, as seen from the heat transfer data given
in Fig. 6, the presence of oscillations has a diminishing
effect, and the system approaches the non-oscillatory
flow behaviour. However, at low net flow rates the
presence of flow oscillations significantly alters the
(convective) heat transfer behaviour, which suggests
a profound effect on the fluid mechanics.
It is apparent from the power density data given in
Fig. 10, that at small amplitudes of oscillation and
2223
higher frequencies the power density is greater than
predicted by the quasi-steady model. We believe the
device is operating in a non-quasi-steady regime,
where particularly interesting effects have been re-
ported in other papers. It has been observed from flow
visualisation studies (Brunold et al., 1989; Mackley
and Ni, 1994) that complex flow structures and signifi-
cant Eddy interaction exist. Effective control over
residence time behaviour can also be obtained in this
regime (Dickens, 1989), and efficient heat transfer has
been demonstrated. I n our opinion this non-quasi-
steady operating regime offers a different type of fluid
mechanics to either quasi steady or turbulent flow,
and we believe that significant process advantage can
be found by operating in this regime.
Acknowledoements--The authors are indebted to Prof. M.
H. I. Baird for valuable discussions regarding the power
dissipation work. Financial support from Shell KSLA (Am-
sterdam) and the Science and Engineering Research Council
(now EPSRC) is gratefully acknowledged.
NOTATI ON
a acceleration, m s- 2
as measured area of pressure-displacement plot
ac cross-sectional area of tube, m 2
Cp specific heat capacity, J kg- ~ C- 1
Co orifice coefficient
D tube diameter, m
Do orifice diameter, m
D~ tube outside diameter, m
Ec work done per cycle, J
f frequency, Hz
L spacing between baffles, m
f dimensionless distance
m' mass flow rate, kg s-1
n number of baffles
P Power, W
Pr Prandlt number ( = Cpll/k)
Reo oscillatory Reynolds number
Ren steady-flow Reynolds number
Sr Strouhai number
S fractional open area of baffle
t time, s
T Period of oscillation, s
u velocity, m s- t
U bulk average velocity, m s-
x displacement, m
xo amplitude of oscillation (centre-to-peak, m)
Z tube length, m
Greek
c5
Ap
Apo
Apy
Aplo
P
(0
letters
phase angle between pressure and velocity
power density, W m- 3
pressure drop, Pa
maximum pressure drop, Pa
pressure drop due to friction, Pa
maximum pressure drop due to friction, Pa
kinematic viscosity, m 2 s- 1
density, kg m- 3
viscosity, Pa s
angular frequency, radian s- 1
2224 M. R. MACKLEY and P. STONESTREET
REFERENCES
Baird, M. H. I., Duncan, G. J., Smith, J. I. and Taylor, J.,
1966, Heat transfer in pulsed turbulent flow. Chem. Engn#
Sci. 21, 197-199.
Baird, M. H. I., 1994, Personal communication.
Brunold, C. R., Hunns, J. C. B., Mackley, M. R. and
Thompson, J. W., 1989, Experimental observations on
flow patterns and energy losses for oscillatory flows in
ducts with sharp edges. Chem. Engng Sci. 44, 1227-1244.
Dickens, A. W., Mackley, M. R. and Williams, H. R., 1989,
Experimental residence time distribution measurements
for unsteady flow in baffled tubes. Chem. Engng Sci. 44,
1471-1479.
Edwards, M. F. and Wilkinson, M. A., 1971, Review of
potential applications of pulsating flow in pipes. Trans. I.
Chem. E. 49, 85-94.
Gupta, S. K., Patel, R. D. and Ackerberg, R. C., 1982, Wall
heat/mass transfer in pulsatile flow. Chem. Engn# Sci. 37,
1727-1739.
Hafez, M. M. and Baird, M. H. I., 1978, Power consumption
in a reciprocating plate extraction column. Trans. I. Chem.
E. 56, 229-238.
Holman, J. P., 1976, Heat Transfer, Fourth Edition.
McGraw-Hill, New York.
Howes, T., Mackley, M. R. and Roberts, E. P. L., 1991, The
simulation of chaotic mixing and dispersion for periodic
flows in baffled channels. Chem. Enong Sci. 46, 1669-1677.
Jealous, A. C. and Johnson, H. F., 1955, Power requirements
for pulse generation in pulse columns. Ind. Engng Chem.
47, 1159-1166.
Kay, J. M. and Nedderman, R. M., 1985, Fluid Mechanics
and Transfer Processes. Cambridge University Press,
Cambridge.
Keil, R. H. and Baird, M. H. I., 1971, Enhancement of heat
transfer by flow pulsation. Ind. Engng Chem. Process. Des.
Develop. 10, 473-478.
Levenspiel, O., 1984, Engineering Flow and Heat Exchange.
Plenum, New York.
Mackay, M. E., Mackley, M. R. and Wang, Y., 1991, Oscilla-
tory flow within tubes containing wall or central baffles.
Trans. I. Chem. E. 69A, 506-513.
Mackley, M. R. and Ni, X., 1994, Experimental fluid disper-
sion measurements in periodic baffled tube arrays. Chem.
Engng Sci. 48, 171-178.
Mackley, M. R., Smith, K. and Wise, N. P., 1993, The
suspension of particles using oscillatory flow in baffled
tubes. Trans. I. Chem. E. 71A, 649-656.
Mackley, M. R., Tweddle, G. M. and Wyatt, 1. D., 1990,
Experimental heat transfer measurements for pulsatile
flow in baffled tubes. Chem. Enang Sci. 45, 1237-1242.
Muzushina, T., Maruyama, T., Ide, S. and Mizukami, Y.,
1973, Dynamic behaviour of transfer coefficient in pulsat-
ing laminar tube flow. J. Chem. Engng Japan 6(2),
152 159.
Oliver, D. R. and Soji, Y., 1992, Heat transfer enhancement
in round tubes using three different tube inserts: non-
newtonian liquids. Trans. I. Chem. E. 70A, 558-564.
Schlichting, H., 1979, Bounday-Layer Theory, Seventh Edi-
tion, pp. 436-438. McGraw-Hill, New York.
Sobey, I. J., 1983, The occurrence of separation in oscillatory
flow. J. Fluid Mech. 134, 247-257.
West, F. B. and Taylor, A. T., 1952, The effect of pulsations
on heat transfer. Chem. Engng Prog. 48(1), 39-50.

Você também pode gostar