Chemical Engineering Science, 197 1, Vol. 26, pp. 1987-2001. Pergamon Press.

Printed in Great Britain

A mechanism for heat transfer in a Votator $-type scraped-surface

heat exchanger
A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKENS
Unilever Research Laboratories, Vlaardingen/Duiven, The Netherlands

(Receioed 22 June 1970)

Abstract-The mechanism for heat transfer in a SSHE consists of three steps, viz.:
Penetration of heat by conduction in a thin layer close to the heat transfer surface in the time
between two scrapings.
Partial temperature equalization in the boundary layer that builds up on the scraper blade.
Convective radial transport from the cooled layer to the bulk of the liquid. Taylor vortices contri-
bute to this radial transport.
Heat transfer measurements were conducted under various conditions.
The heat transfer coefficient can be described by means of an equation resulting from penetration
theory, modified by an empirically determined correction factor ‘p. In ‘pare included:
The incompleteness of the temperature equalization in the boundary layer.
The effect of radial dispersion.
The decrease in driving force for heat transfer due to axial dispersion.
When the influence of the axial dispersion on heat transfer is absent (high mass flow rates), cp is a
function of Pr. If Taylor vortices are present, the value of cpis independent of ReA. Below the critical
Reynolds number ‘p decreases due to insufficient radial mixing.
At low mass flow rates the influence of the axial dispersion on the driving force for heat transfer is
important. The apparent decrease in heat transfer coefficient due to this effect can be calculated. To
do so, the plug flow with dispersion model is applied in which the axial dispersion coefficient is cal-
culated from the standard deviation of the residence time distribution curve.

INTRODUCTION transfer. However, no measurements of the

IN PREVIOUS articles[ 1,2] the flow phenomena dispersion coefficient were available at that
and the power consumption in a Votator-type time.
scraped-surface heat exchanger (further referred Heat transfer measurements are reported
to as SSHE) have been discussed. This paper, by various authors. The most extensive in-
which is the last one of the series, deals with heat vestigations using SSHEs are those by
transfer. Skelland, Oliver and Tooke[6] and by Trom-
The literature on heat transfer is far more ex- melen [7].
tensive than on flow phenomena and power con- Penetration theory as a mechanism for heat
sumption. A review of the papers published transfer has been discussed by various authors
before 1967 is given by Penney and Bell [3]. [7-121. The measured values of the heat
Important aspects are: transfer coefficient when viscous liquids are
The mean temperature difference in the used as working fluid are always lower than
SSHE. A mathematical description has been those calculated from a theory of penetration
given by Bott, Azoory and Porter[4] and by of heat followed by complete mixing.
Penney and Be11[5] for the influence of axial Trommelen[7] added a correction factor gcto
dispersion on the mean temperature difference the equation resulting from penetration theory
and therefore on the effectiveness of heat followed by complete mixing:

tTrade name ofthe Chemetron Corporation, Louisville, Kentucky, U.S.A.

$Authors’ address: Olivier van Noortlaan 120, Vlaardingen, The Netherlands.

1987
 A. M. TROMMELEN. W. J. BEEK and H. C. VAN DE WESTELAKEN

was the same as described earlier[7]. Five

variables were investigated: shaft speed, shaft
diameter, number of rows of scraper blades, mass
His measurements could be correlated by: flow rate and fluid viscosity. The experiments
were carried out with glycerol/water mixtures of
cp= 1 - 3.28 pe-o= (for Pe < I&). (2) three different concentrations.
In the previous paper[7] the heat transfer
For Pe > 2500, (c was a function of Pr. coefficient was calculated from the amount of
For the experimental data from Skelland, heat transported, the area of scraped surface and
Oliver and Tooke[6] it followed that: the logarithmic mean temperature difference
between the cooling water and the liquid in the
cp= 1 - 2.78 (Pe + 200)-““s. (3) annular space; by using a correlation for the
heat transfer coefficient between cooling water
Equations (2) and (3) are very similar. and wall [ 151, the heat transfer coefficient between
The work described in this article was under- liquid and scraped wall could also be calculated.
taken for three reasons, viz.: In the above calculation it is assumed that the
(1) Measurements on the axial dispersion in a power consumed by the scraping of the blades is
SSHE have been reported[ 11. The in- a part of the heat flux from the liquid in the
fluence of axial dispersion on the effective annular space to the cylinder wall.
temperature difference can therefore be However, as is stated above, this heat is gene-
investigated as suggested by Azoory and rated very close to the scraped wall and the heat
Bott[l3]. transfer resistance to this wall can therefore be
(2) In the recent work on power consumption neglected. It can easily be shown that the real
in SSHEs[2] a distinction was made heat transfer coefficient at the scraped side is
between the power that is dissipated in the equal to that calculated in the above manner
bulk of the liquid in the annular space and multiplied by ql/q2, in which q1 is the heat flux
the power needed to overcome the fric- from the scraped wall (excluding the power that
tion between the blade edges and the wall. is dissipated in the clearance between the edge of
It appeared that the latter part is greater the blade and the tube wall) and q2 is the total
than the former. The clearance between heat flux.
the edge of the blades and tube wall is of All heat transfer coefficients were calculated
the order of 1 pm[2] and from the tem- in this way. The correction factor cp, defined in
perature profile of the liquid in this clear- Eq. (l), was also calculated.
ance [ 141 it can be calculated that this heat
RESULTS
flows directly into the cylinder wall, with-
out participating in the heat transfer Influence ofaxial dispersion
mechanism from the bulk of the liquid in In the paper on flow phenomena [ l] it is argued
the annular space to the scraped wall. This that the assumption of plug flow in the SSHE
effect should be accounted for in calculat- may lead to too high values for the driving force
ing oc,, the heat transfer coefficient on the for heat transfer. Axial dispersion decreases the
scraped side. effective temperature difference between work-
(3) As the flow pattern in a SSHE is known ing fluid and heat transfer medium. For the plug
[ 11, an attempt can be made to find the heat flow with axial dispersion model the decrease in
transfer mechanism. heat transfer rate is dependent on two dimension-
less parameters, NT = +4/&c, (Stanton-
EXPERIMENTAL number), and PeD = vL/D, (see Fig. 1). The mass
The apparatus and the experimental procedure flow rate has a great influence on the decrease
1988
 A mechanism for heat transfer

m viscosities was derived from measurements of

IO0 the heat transfer coefficient, using the 56 mm
60 shaft with two rows of scraper blades. The
0.8 40
shaft speed was constant at 10 rev. set-‘. Three
20
glycerol/water concentrations were used. The
-: 0"
n 0.6 '0 I, viscosities were approximately O-045, O-15 and
4 0,
E 5.0 a
0.45 N . sec. mm2.The results are plotted in Fig. 3.
a" 3.0 In Figs. 2 and 3 the factor cpis plotted against
0.4 the mass flow rate. For each series of measure-
I.0
0.5 ments with constant N and 7, cp is proportional
o.eL I I1111111 , I 11111~ to the effective heat transfer coefficient. At high
0.1 0.2 0.4 0.6 I.0 2.0 4.0 6.0 IO" mass flow rates (& > cu. 0.1 kg, set-I), the
a, A
N, = - value of the effective heat transfer coefficient is
@ mCP
independent of & and this value is assumed to
Fig. 1. Effect of P6clCt number Pe and NT on the effective be the actual heat transfer coefficient cz,. The
heat transfer coefficient. lower heat transfer coefficients for c$,,,< ca.
O-1 kg. see-’ are caused by assuming plug flow
in heat transfer rate due to axial dispersion. A in cases where axial dispersion is important.
decrease in C& results in increasing NT and de-
creasing Pe, (v is proportional to 4,); Fig. 1 I.0
shows that this gives a decrease in heat transfer 0.6

rate. 0.6E

Measurements of the heat transfer coefficient

were carried out at shaft speeds of 4, 10 and 25
rev. set-’ using the 56 mm shaft with two rows of
scraper blades. The viscosity of the liquid varied
between O-1 and O-2 N . sec. m+ (these figures
refer to the viscosity of the liquid at the arithmetic 0.1 I
0.01 2 4 6 60.1 2 4 6 6 I.0
mean of inlet and outlet temperature). The mass Mass flow rate , kg set-1
flow rate variedbetween O-013and 0.58 kg . set-I.
Fig. 3. Influence of the mass flow rate on the heat transfer
The results of the measurements are plotted in coefficient at various viscosities. n(N . sec. m-*): O-CCL
Fig. 2. Detailed results are given elsewhere[ 161. 0.45, O-CL 0.15, A-cu. OM5. N = lOrev. set-I, d, =
The influence of the mass flow rate at various 56mm.n=2.

Our results will now be compared to the model

I.0
04 by Bott, Azoory and Porter[4] using Fig. 1.
0.6 Previously[l] it has been calculated that the
A axial dispersion coefficient in the SSHE with the

!
@4
9
a+.." Lv 56 mm shaft rotating at 8 rev. set-’ with n = 2
A b
OAO 0
. 0 is approximately 2.3 X 10e4 m2 . see-l for a liquid
il.2 O . o 0 0
0 viscosity of O-2 N . sec. rne2. This value for the
A l
dispersion coefficient was used in the calculation
0.11
0.01
l ’
2 4
I

660.1
I I

2
I I

4 6
I I,

61.0
of the decrease in effective heat transfer co-
Mass flowrate , kg. set-1 efficient according to Fig. 1. In Fig. 4 the
Fig. 2. Influence of the mass flow rate on the heat transfer
measured and calculated ratio of the effective
coefficient at various shaft speeds. N(rev. set-I): a-25; and actual heat transfer coefficient are plotted
A-10;O-4.n=0~1-0~2N.sec.m-2,d,=56mm,n=2. against the mass flow rate. The measurements
1989
 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

kg. set-‘. The results of the measurements are

plotted in Figs. 5-8.
The influence of the number of rows of blades
on the heat transfer coefficient was investigated
by carrying out experiments in which 2, 3, 4 or
6 rows of blades could be fitted to the 56 mm
shaft. The mass flow rate was about 0.17 kg. set-1
and the shaft speed 10 rev . set-l. The viscosities
of the glycerol/water mixtures were O-04, 0.14
and 0.44 N . sec. m-*. The results are plotted in
Fig. 9.
The last parameter studied was the width of
the annular space. The shaft speed was 10 rev.

1 I I I ,I I I II,

IO-2 2 4 660' 2 4 6 8 1.0

Mass flow rate , kg. set-1
Fig. 4. Ratio of the effective and actual heat transfer coeffi- 0
cients as a function of the mass flow rate. Measured: O-at
r) = 0.45 N . sec. m-2
indi.a~,;&~.=$;$~;. set . M-Z. Lines
e 0.4

.
.
=
A.
0

shown in Fig. 3 for the viscosities of approxi- l/. l

/ l --
l

/
mately 0.15 and 0.45 N . sec. m-* were used for
this purpose, as the ratio of shaft speed and vis-
cosity is close to that applying during the 0.1’ I , I 1 I I

measurements of the dispersion coefficient [ 11. 4 6 6 IO 20 40

Shoft speed, rev set-1
From Fig. 4 it can be seen that the theoretical
model predicts the decrease in heat transfer due Fig. 5. Influence of the shaft speed on the heat transfer coeffi-
to backmixing fairly well. cient at various viscosities. q(N . sec. m-*): l - cu. 0.45;
A-cu. 0.15; 0-cu. O+IO. d,=46mm, n=2, &=0.17
Industrial votators are usually operated at such kg. set-I.
high mass flow rates, and hence low values of NT
and high values of PeD, that the influence of the I.0
axial dispersion on heat transfer may be neg- 0.6
lected. The measurements reported in the follow-
0.6
ing sections were made at $I,,,= 0.17 kg. set-l F
and the influence of axial dispersion was there- 0.4_ O-VV-CT
fore negligeable. % rAAA A LJ A

Heat transfer without influence of axial dis-

l .
persion
Measurements of the heat transfer coefficient
were carried out using all four shafts with two
rows of scraper blades. The shaft speed varied O,I4 Shoft speed, rev see-1
40

between 4 and 33.3 rev , set-l. Three glycerol/

water concentrations, having viscosities of Fig. 6. Influence of the shaft speed on the heat transfer coefi-
cient at various viscosities. ?(N . sec. m-*): 0-w 0.45;
approximately O-04, 0.13 and 0.44 N . set . me2 A-co. O-15; 0-w. 0.040. d,=56mm, n=2, &=0+17
were used. The mass flow rate was about 0.17 kg. set-I.

1990
 A mechanism for heat transfer

I.0 0.5 )-
0.6
0.4 -
0.6
oe
0.3 -

9 ‘\ A
0.2 - .

\ .

1 I II I I
4 6 0 IO 20 40
0.1 L I I I I
Shaft speed, rev see-1 2 3 4 6
Number of rows of blades
Fig. 7. Influence of the shaft speed on the heat transfer coeffi-
cient at various viscosities. q(N . sec. m-*): O-U. 0.45; Fig. 9. Influence of the number of rows of blades on the heat
A-W. 0.15; O-W. O&IO. d,=62mm, n=2, &=0.17 transfer coefficient for vti.ous viscosities. n(N . sec. m-*):
kg. set+. @-cu. 044; A-W. 0.14; O-W. OWO. d, = 56mm, N =
10 rev. set-I, c#J,,,= 0.17 kg. see-*.

I.0
0.5
0.6
F
0.6 - 0.4

F
0
%0.4 - 8 6 4f8 s 0.3t

l . A :

0.2 - .

1 I I I 1
0.1 1 I
4 6 6 IO 20 40
Shaft speed, rev set-1 I 1 I I
4 I5
dri
Fig. 8. Influence of the shaft speed on the heat transfer 2 (mm)”
coefficient at various viscosities. q(N . sec. m-*): O-co.
0.45; A-cu. 0.15; O-W. 0340. d, = 68mm, n = 2, 4,. = Fig 10. Influence of the annular space on the heat transfer
O-17 kg. set-*. coefficient for various viscosities. q(N sec. m-z): l -ea.
0.44; A-cn. 0.14; O-W. O@tO. N= lOrev. set-I, n=2,
4, = 0.17 kg. set-I.
set-’ during these experiments. Two rows of
scraper blades were applied. The viscosities of temperature of the liquid as a function of
the glycerol/water mixtures were again O-040, time and position is given by penetration
O-14 and 044 N . set . m-*. The results are given theory.
in Fig. 10. (2) The thin and cooled layer is scraped off the
wall and mixed with the bulk of the liquid
THE MECHANISM OF HEAT TRANSFER in the annular space. The mixing is
Up to now it has been assumed that the assumed to be imperfect and this qualita-
mechanism of heat transfer in a SSHE consists tively explains why the measured heat
of two steps: transfer coefficients are lower than is cal-
(1) In the time between two scrapings a thin culated from penetration theory [ 121.
layer of liquid close to the heat transfer From the study of the flow pattern [ 11, it has
surface is cooled by conduction. The appeared that the second step does not take place
1991
 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

in the Couette flow regime. Neither is this parallel to this stationary wall somewhere in the
mechanism valid for the Taylor vortices, as can liquid. At t = 0 the temperature is zero for x > 0.
be seen as follows: The depth of the penetration For t > 0 the wall temperature is T,,.
of heat in the time between two scrapings is If the oscillating wall is absent and heat transfer
cu. 0.1 mm. The vortices in the Taylor instability takes place by conduction only, the temperature
regime will not cause a mixing of such a thin in the liquid can be described with the penetra-
layer close to the wall with the liquid in the tion theory:
annular space.
It would seem, therefore, that the mechanism T= TOerfcL
for heat transfer described above is not valid. In 2vz
the search for alternatives, two possible mechan-
isms were considered: The heat flux is in this case:
(1) The contraction and subsequent expansion
of the streamlines caused by the passage (5)
of the blade increase the heat flux to the
heat transfer wall: The photographs of the
flow pattern [ 11 show that contraction and It has been shown [16] that the heat flux, as
subsequent expansion of the streamlines given by Eq. (5), is not much influenced by a
take place in the blade area. It is possible flow induced in the liquid by a plane oscillating
that this effect influences the penetration in the x-direction.
of heat in the liquid close to the heat One of the two possible mechanisms for heat
transfer surface. transfer is now rejected. We may conclude, there-
(2) The penetration of heat in the time fore, that high heat transfer coefficients in a
between two scrapings is followed by a SSHE, and hence high temperature gradients
temperature equalization: In the hydro- near the wall, are caused by a temperature
dynamic boundary layer that builds up on equalization in the hydrodynamic boundary
the scraper blade the velocity of the liquid, layer on the scraper blade.
relative to the blade velocity, is low.
Therefore, after the liquid has been scraped Temperature equalization in a boundary layer
off the wall, the time it remains in the A second possible mechanism for heat transfer
boundary layer before contacting the wall that will be considered is the penetration of heat
again is relatively long. In this boundary followed by a temperature equalization in the
layer a temperature equalization may there- boundary layer that builds up on the blade. A
fore take place, resulting in a high tempera- schematic view of the geometry is given in Fig.
ture gradient at the wall, and consequently 11. The heat transfer surface is at x = 0. In the
in a high heat transfer coefficient. model the scraper blade is taken to be parallel
In the following sections we will consider the to the heat transfer surface. The heat transfer
merits of these possible mechanisms. tube moves and the blades are stationary.
In the boundary layer on the scraper blade the
Znjuence of an oscillating wall on heat penetra-
tion - “0 X
The contraction and subsequent expansion of 2
the streamlines caused by the passage of the
blade can be compared with the flow between
two plane walls, one of which is oscillating. Let Adiabatic
liquid fill the space x > 0 and be bounded by a Fig. 11. Geometry of model for temperature equalization in
stationary wall at x = 0 and an oscillating wall a boundary layer.

1992
 A mechanism for heat transfer

velocity of the liquid is relatively low com- in which:

pared to v,,. A temperature equalization may
therefore occur in the boundary layer, resulting
in a high temperature gradient at the wall after
A(z) =2 X-O
_-
passage of the blade. It may not be expected that
the temperature equalization is complete, and From the boundary layer equations [ 171 it
therefore the equation resulting from penetra- follows:
tion theory, Eq. (l), will yield too high values for
the heat transfer coefficients. A correction factor 3 1
A(z) =v0’?‘4+j4-
cpwill allow for this, as well as for other deficien-
cies of the model.
For z < 0 both wall (x = 0) and liquid (x > 0) and hence:
have a velocity u0 in positive z-direction. The
blade (between z = 0 and z, = 1) is supposed to
be adiabatic. Hence for z < 0 is valid: (8)

21,= v0 for x L 0 A further assumption is:

v,=O for x20. v,= 0. (9)

For z > 0:
Using the velocity profile given in Eqs. (8) and
vZ=O at x=0. (9) the equation of energy must now be solved to
find the temperature profile in the boundary
The wall temperature for z < 0 is T,, and the layer.
temperature of the liquid for z < 0 and x = ~0is The equation of energy reads for this case [ 171:
equal to zero. Let the temperature profile in the
liquid at z = 0 be given by:
(10)
T = To for x 6 aT
(frictional heat is not taken into consideration).
The temperature gradients in the x-direction
and
are much greater than in the z-direction and
T= Oforx > S,. therefore a*T/az* is neglected with respect to
a2T/ax2. After elimination of v, with Eq. (8) the
(The error function, which describes the tempera- equation of energy now reads:
ture profile during heat penetration, is approxi-
mated by a linear function.) x aT a2T
Equations describing the velocity profile that “qdz=s (11)
builds up on the scraper blade are known from with
the literature [ 171. As the thickness of the thermal 0.324 vo3’2
boundary layer that builds up on the blade is E= a& *
usually smaller than that of the hydrodynamic
boundary layer, the velocity profile is approxi- The boundary conditions are:
mated by a linear function in x in the following
way:
at
v, = A (z)x (6) for x > S,
1993
 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

at x=O:aT=O.
ax

(11) is solved by substituting:

8 = 23’*;q = $zl/* dz

5= x31*; dt = qx1i2 h_

(11) now reads:

- 2 exp {-x3(y)-‘}]. (14)
aT
-=-_ 3 a*T
2~ a? (12)
at
The temperature profiles at z = 1 will now be cal-
with boundary conditions: culated with (14) for the following conditions:

at‘ 5=O[T= T,(l-f) for 5 < aT3/* u. = 2.5 m .set-I

[T=O for 5 > aT3/* I= 0.02 m

at tg= 0: g = 0. &= 1-37X 10e4m

a = 0.94 X 10W7m* . set-1

The solution to (12) is known [ 181:
O-8 x 10m5m* . set-’ (q= 0.01 N . sec. m-*)
Y = O-8 x 10m4m* . se6 (1) = O-1 N . sec. m-*)
1O-8 X 10m3m* . set-’ (7) = 1 N.sec.m-*).

(5-d2 The results are given in Fig. 12. They show that a
x eXP-4(3t/2e) +exp-4(3~/2r) good temperature equalization takes place in the
(13) boundary layer on the blade. At viscosities of O-1
and 1 N . sec. m-* the temperature difference
This is a rather awkward expression. The prob- between bulk and liquid that reaches the wall
lem can be simplified by linearizing (1 - (z*‘~/S,)). after leaving the scraper blade is reduced to 19
If this expression is plotted against z . &-1.5 it is and 11 per cent.
found that it may be approximated by 0.8-0.8 In this boundary layer approximation the flow
z . &-l+. in the x-direction has been neglected. If this flow
After substituting z and x for 4 and 4, the solu- is taken into account the temperature equaliza-
tionis[18]: tion will be even better than calculated. On the
other hand it has been assumed that the scraper
6 = @4&-I”[ (f&1.5_x1.5) blade is adiabatic. Often the scraper blade is
made of stainless steel. Heat conduction will
take place in the z-direction through the blade
erf
1 (6,1.5-x1.5) (!90’5}+ (8T1.5+xl.5)
and this results in less temperature equalization.
This effect may be more important for stainless
x erf[ (&I.” +x’.5) (7O~5)
steel blades than for plastic blades with low
thermal conductivity.
1994
 A mechanism for heat transfer

x.(mm)

Fig. 12. Temperature equalization in a boundary layer. 7 (N . set . m-2): curve I, 0.01; curve 2,O. 1; curve 3, I.

The preceding calculations have shown that

important aspects of the heat transfer mechanism
are:
Penetration of heat by conduction in a thin
layer close to the heat transfer surface in the
time between two scrapings;
Partial temperature equalization in the
boundary layer that builds up on the blade.

Penetration of heat after successive scrapings

0
The temperature equalization that takes place Dislonce from heot tronsfer wall
in the boundary layer on the blade is not com-
Fig. 13. Incomplete temperature equalization in the boun-
plete, as can be seen in Fig. 12. Therefore the dary layer on the scraper blade. The curve indicates the
driving force during the penetration of heat after temperature of the liquid contacting the wall after the first
the first scraping is smaller than in the period blade passage.
before this scraping. This is shown in Fig. 13:
The driving force, TO- Tb, is smaller than To - 0, During the second scraping a temperature
that is assumed in the derivation of the heat equalization in the boundary layer will again
transfer coefficient from penetration theory: take place, but this time the temperature of the
liquid on contacting the wall after the scraping
cY8 = 2 (hpC,Nn)“.5. (15)
will be higher than Tb in Fig. 13. This results in
an even lower heat transfer coefficient in the
period between the second and third scraping
In the period between the first and second than that given by Eq. (16). As the mean resi-
scraping the average heat transfer coefficient will dence time in a SSHE is some orders of magni-
therefore be approximately: tude higher than the time between two scrapings
this process is repeated many times.
CY~
= ~-+-(hp~~Nn)~.‘. (16) During the first periods of the process the
7r depth of the penetration of heat is sasmall that
1995

C.E.S. Vol. 26 No. 12-C

 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

the radial convective heat transport due to dis- It is obvious that the value of (To - Ta)/To is of
persion does not influence the process of heat paramount importance. If the influence of axial
penetration. However, after a few scrapings, disp ‘rsion is absent, this factor is equal to the
followed by partial temperature equalizations in co R ection factor cpin Eq. (1).
the boundary layer, the layer of liquid influenced $‘here are three competing effects influencing
by the penetration of heat becomes thicker and the value of (To - Ta)/To and hence of cp:
the radial dispersion begins to play a role, The depth of the penetration of heat, &, in
especially if Taylor vortices are present. the time between two scrapings. For constant
The radial dispersion can be represented by a thermal diffusivity, S, is proportional to the
stagnant film of thickness 8, (influenced by the square root of time.
operating conditions), assuming good mixing in The temperature equalization in the bound-
the liquid outside this film. This implies that the ary layer on the scraper blade. It has been
temperature of the liquid at x = &, is equal to shown that the equalization is more complete
the bulk liquid temperature during the penetra- when the shaft speed is low, the viscosity is
tion of heat. After a number of scrapings, followed high, and aT is low.
by partial temperature equalizations, the pene- The radial dispersion and hence 8,. The
tration depth of the heat conduction will be equal greater the rotational Reynolds number, the
to S,. A pseudo-equilibrium, which is schemati- smaller will be 8,.
cally sketched in Fig. 14, will then be established. Equation (14) shows that increasing the shaft
speed does not influence gc when radial disper-
sion is absent (where E - Vet’*- ST3). Equation
\ (14) proves that (Tj,,,)IT, and hence p is inde-
\ pendent of shaft speed. If in the case of radial
\
dispersion the value of 8, is proportional to N-1’2,
\
I-
\
cp is not influenced by the shaft speed either. In
that case, the temperature profiles in Fig. 14 are

;l--b
similar when the shaft speed is varied.

COMPARISON OF EXPERIMENTAL RESULTS

WITH THE PROPOSED MECHANISM
1 I
The influence of shaft speed and viscosity on
80
x (distance from heat transfer wall 1 the heat transfer coefficient as shown in Figs.
5-7 will first be compared with the proposed
Fig. 14. Penetration and temperature equalization after
various scrapings. Dashed line: temperature profile of the mechanism. In the experiments with the 46, 56
liquid contacting the wall after the blade passage. and 62 mm shaft diameter, it appears that for a
liquid having a viscosity of ca. O&t0 N . set . m-*,
During the penetration of heat the temperature (p is not influenced by the shaft speed. When
is T,, at x = 0 and zero at x = 8,. During the liquids with a higher viscosity are used (ca.
temperature equalization in the boundary layer, 044 N . sec. m-*) it appears that p decreases
the temperature at x = 0 drops from T, to Tb. The below a certain value of the shaft speed.
temperature at x = 8, increases, because the In Fig. 15 cp is plotted as a function of Re,.
radial dispersion is smaller in the blade area. The For ReR > 280 Taylor vortices will occur[l]. In
heat transfer coefficient for this situation is this area cp is not influenced by the shaft speed.
again: Apparently in this flow regime 8, is influenced by
the shaft speed in the same way as &. This
as = ~+(A~c,N~)‘~. (16) means that 8, is inversely proportional to the
?r square root of the shaft speed. The Couette
1996
 A mechanism for heat transfer

IQ-
06 -
0.6 -
n ” 0 “0

9 0.4 -

0.2- ./+== AA A

Fig. 15. The influence of the Reynolds number on the heat

transfer coefficient at various viscosities. n(N . sec. mm2): Fig. 17. pPr0,Z5 as a function of Re,. v(N . sec. m-‘): l -cu.
O-M. 0.45; A-cu. 0.15; q -ea. OXMO.d, = 56mm. n = 0.45; V-ca. 0.15; Cl-cu. 0.040. d,=56mm, n=2, c#+,=
= 0.17 kg. set-I.
2, c#J,,, 0.17 kg. set-I.

flow regime prevails for Re, < 280. Apparently In Fig. 18 cp . PFz5 is plotted vs. ReRIRecr for
SD is more dependent on the shaft speed in the the three different shaft diameters.
Couette flow regime than in the Taylor instability For the values of ReJRe,, > 1, the measure-
regime. Experiments with the 46 and 62mm ments can be correlated by
shafts show the same trend.
From Figs. 5-7 it appears that cp is also in- (0 = 2.0 Pr-“‘2s_ (18)
fluenced by the viscosity of the working fluid.
cp decreases with increasing viscosity as a result For values of ReJRe,,. < 1, q PPz5 is smaller
of a higher value of 8, in the case of high- than 2.0.
viscosity liquids. In Fig. 16 the value of cp for The experiments with the 68 mm shaft (Fig.
ReR > 280 is plotted vs. the Prandtl number. It 18) show, at low shaft speeds and high viscosities,
a higher value of the heat transfer coefficient than
with the other shafts. Under these conditions the
boundary layer that builds up on the scraper
blade is thicker than the gap between the blade
and the shaft. The velocity in this layer will be
lower than in an undisturbed boundary layer,
causing a better heat equalization. The mechan-
ism for heat transfer is apparently different in
this case, as can be seen from Figs. 8 and 19.
We will now consider the experiments by
Skelland, Oliver and Tooke[6]. As the mass
Fig. 16. The influence of the Prandtl number on the final flow rate during their experiments was rather
value of the heat transfer coefficient. n = 2, & = 0.17 low, axial dispersion was important. The form
kg.sec-1,d,(mm):U-46;O-56;A-62. of the correction factor for the experiments was:

appears that the relationship between cp and Pr cp= 1-2*78(Pe+200)-0’18 (400 < Pe < 6000).
is given by: (3)

cp= constant . Pr-“.25. (17) The influence of the axial dispersion (the mass
flow rate is proportional to Pe) is included in this
In Fig. 17 cp. PF5 is plotted vs. Re,. For the function. The experimental results of Skelland
46 and 62 mm shaft the same relationship et al. and those reported in this chapter are in
between cpand Pr is found. agreement with each other.
1997
 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

IO
6
6

“_

0.1 I IO
Re,/ Ret,

Fig. 18. (PP~O’~~

as a function of ReR/Recr for various shaft diameters d, (mm): O-46; O- 56; v-62. n = 2,
9., = 0.17 kg. set-I.

caused by a measuring error, or an incorrect

mounting of the scraper blade.

CONCLUSIONS AND DESIGN

The mechanism for heat transfer in a SSHE
consists of three steps:
0.11 I I I Penetration of heat by conduction in a thin
14 103 It+
R’R layer close to the heat transfer surface in the
Fig. 19. The influence of the Reynolds number on the heat time between two scrapings.
transfer coefficient at various viscosities. q(N . sec. me2): Partial temperature equalization in the
l -cu. 0.45; A-cu. 0.15; q -ea. O@+O.d, = 68mm, n = 2,
$J* = 0.17 kg. set-*. boundary layer that builds up on the blade.
A convective radial transport from the
From Fig. 9 it can be seen that p decreases cooled layer to the bulk of the liquid. Taylor
slightly with an increasing number of rows of vortices contribute to this radial transport.
blades. This is caused by the fact that more rows The heat transfer coefficient can be described
of blades decrease the radial dispersion. The by an equation resulting from penetration
transition from Couette flow to Taylor vortices theory, modified by an empirically determined
will take place at a higher Reynolds number, if correction factor:
many rows of blades are present [ 193. Further-
more the contact time of the liquid and the wall Nu = 1.13 (Re, . Pr . PZ)O.~.cp.
between two scrapings is less than l/Nn. The
scraped liquid does not flow back from the cphas been found to be a function of Pr:
scraper blade to the edge along the back of the
blade, but contacts the wall again a short distance p = 2.0 Pr-O‘25forReR > Re,.
from the blade edge. This effect increases with
an increasing number of rows of scraper blades. At low mass flow rates the influence of the
From Fig. 10 it appears that the width of the axial dispersion on the driving force for heat
annular space has no influence on heat transfer transfer is important. The apparent decrease in
if liquids with viscosities of O&IO and 0.14 the heat transfer coefficient due to this effect can
N . sec. m-* are used. For a viscosity of 044 be calculated. To do so, the plug flow with dis-
N . sec. m-* the heat transfer coefficient is higher persion model is applied, in which the axial
at an annular space of 4 mm. This has been ex- dispersion coefficient is calculated from the
plained previously. The low heat transfer co- standard deviation of the residence time dis-
efficient at an annular space of 7 mm is probably tribution curve.
1998
 A mechanism for heat transfer

As has been stated in a previous article [2], the the length of the votator is increased proportion-
power consumption in a SSHE consists of two ally, the power consumption per unit mass of
parts: the power that is dissipated in the bulk of working fluid retains its value. If the heat transfer
the liquid in the annular space and the power area is increased by increasing the diameter of
needed to overcome the friction between blade the SSHE, the shaft speed should be reduced. If
edges and tube wall. For the calculation of cu, the blade velocity and the annular space are the
only the power dissipated in the bulk of the liquid same for the small and large unit, the power
must be taken into consideration. consumption is higher for the large unit. By in-
This power consumption is given by: creasing the number of rows of scraper blades
the heat transfer coefficient can be kept constant.
PO = R . ReRml for Re, < Re, (19) A check should be made to ascertain whether the
flow regime is the same for both cases.
and
For calculation of the total power consumption
PO = R . ReR-o’5. Re,;P5 for ReR > Re, (20) in a SSHE, an emperical correlation [2] has been
derived:
with
p = 251 (Nd,)‘.79 . $.% . lP* . L
(dt - d8)@31 *
High heat transfer coefficients are obtained
when Taylor vortices are present. A wide annular (for m .k . s. units). (23)
space promotes instability. The power dissipa-
tion in the annular space is less when the value The total amount of heat that has to be removed
of (dt - d,) /dt is high. For the same value of the at the cooling side of a SSHE is the sum of the
blade velocity the flow is more unstable in a net enthalpy decrease of the product and the
SSHE with small tube diameter than in one with power dissipation given by Eq. (23).
a large tube diameter. If a certain heat transfer
area is required, a long, narrow SSHE is prefer- Acknowledgement-The authors thank Mr N. A. M.
able to a short, large diameter apparatus. The Lansbergen for carrying out the experiments.
increase in pressure drop can be reduced by
increasing the annular space. If a SSHE is NOTATION
operating in the Taylor vortices regime, the heat a MPCP = thermal diffusivity, L2r1
transfer coefficient is proportional to the square CP specific heat of fluid, L2r2 T-l
root of the number of scrapings per unit time. DE axial dispersion coefficient, L*t-’
However, the power consumption in the annular 4 shaft diameter, L
space is proportional to W5 and independent of 4 diameter of heat transfer tube, L
n. The heat removal is therefore more efficient L length of scraped surface, L
with a large number of scraper blades and a low N shall speed, t-l
shaft speed than with a smaller number of blades NT +Gl.~p = dimensionless group in Fig. 1
and a higher shaft speed, given the same number Nu cu,d,/A= Nusselt number
of scrapings per unit of time in both cases. The number of rows of scraper blades
same is true in the Couette regime. Theoretically, ft power consumed by rotating shaft plus
then, the best way of scaling up is by increasing blades, ML* r3
the length of a SSHE. Increase in pressure drop Pe v(dt-d,)/a = PC&t number
can be offset by widening the annular space. The PeD vL/D, = P&let number for axial dis-
heat transfer coefficient does not change and the persion
power consumption increases linearly with the PO PIpN3dPL = power number
length. When for an increase in mass flow rate Pr r)c,/h = Prandtl number
1999
 A. M. TROMMELEN, W. J. BEEK and H. C. VAN DE WESTELAKEN

R parameter in power consumption model heat transfer coefficient at scraped side,

R% &No/r] = Reynolds number for rotation
Rect critical rotational Reynolds number 6 clearance between blade edge and tube
T temperature, T wall, L
TO temperature of heat transfer tube wall, T thickness of stagnant layer in radial dis-
AT increase in temperature of liquid between persion model, L
scraper blade edge and wall, T depth of penetration of heat, L
Tb temperature of the liquid contacting the parameter in Eq. (1 l), Lm3j2
wall after blade passage, T transformed coordinate, L3’2
t time, t fluid viscosity, ML-‘t-l
V mean axial velocity of fluid, Lt-’ thermal conductivity of fluids, MLt-3T-1
Vo tangential velocity of wall of model SSHE, kinematic viscosity, L2t-’
Lt-’ transformed coordinate, L3j2
fluid density, A4Lp3
Greek symbols (Pm mass flow rate of liquid, Mt-’
aeff effective heat transfer coefficient, Mt-3T-1 cp correction term in Eq. (1)

REFERENCES
[I] TROMMELEN A. M. and BEEK W. J., Chem. Engng Sci. 197 I 26 1933.
[2] TROMMELEN A. M. and BEEK W. J., Chem. Engng Sci. 197 I 26 1977.
[3] PENNEY W. R. and BELL K. J., Ind. Engng Chem. 1967 59 (apr) 47.
[4] BOTH T. R., AZOORY S. and PORTER K. E., Trans. Inst. Chem. Engrs. (London) 1968 46 T 37.
151 PENNEY W. R. and BELL K. J.. Chem. Enan~ Prop. Svmo. Ser. 1969 65 1.
[6] SKELLAND A. H. P., OLIVER D. R. and TGOKE S.:B;. Chem. Engng 1962 7 346.
[7] TROMMELEN A. M.. Trans. Inst. Chem. Engrs. (London) 1967 45 T 176.
[8] KOOL J.. Trans. Inst. Chem. Engrs. (London) 1958 36 T 253.
[9] HARRIOT P., Chem. Engrs. Prog. Symp. Ser. 1959 55 137.
[IO] BRAGINSKII L. N., BEGACHEV V. I. and PABLUSHENKO I. S., Zh. Prikl. Khim. 1964 37 1984.
[ll] DIERENDONCK L. L. VAN, Paper read at the Symposium Hanteren oan uiskeuze uloeistoflen II, Utrecht, 1965.
1121 LATINEN G. A., Chem. Engng Sci. 1958-59 9 263.
[ 131 AZOORY S. and BOTT T. R., Trans. Inst. Chem. Engrs. (London) 1968 46 T 42.
[14] TROMMELEN A. M., Paper presented at the 3rd congress CHISA 1969, 15-20 September 1969, Marianske L&z&,
Czechoslovakia.
[I51 DITIUS F. W. and BOELTER L. M. K., Univ. Cal$ (Berkeley) Pub/s. 1930 2 443.
1161 TROMMELEN A. M., Physical aspects of scraped-surface heat exchangers, Thesis, Delft 1970.
[ 171 BIRD R. B.. STEWART W. E. and LIGHTFOOT E. N., Transport Phenomena, Wiley. New York 1960.
[ 181 CARSLAW H. S. andJAEGER J. C., Conduction ofHeat in Solids. Clarendon Press, Oxford 19592.
[19] SNIJDER H. A., Physics Fluids 1968 11 1606.

R&.un16- Le mecanisme du transfert de chaleur dans un Cchangeur de temperature a surface balayee

se fait en 3 stades:
Penetration de la chaleur par conduction dans une couche fine pres de la surface du transfert de
chaleur dans le temps compris entre deux passages de lames.
Egalisation pattielle de la temperature dans la couche limite qui se forme sur la lame.
Transport radial de convection de la couche refroidie a l’ensemble du liquide. Les tourbillons de
Taylor contribuent ace transport radial.
Des mesures du transfert de chaleur furent effectuees dans differentes conditions. Le coefficient du
transfert de chaleur peut etre decrit a l’aide d’une equation resultant de la theorie de la penetration
et mod&e par un facteur p de correction, determine empiriquement.
cpcomprend:
L’inachsvement de l’egalisation de la temperature dans la couche limite.
L’effet de la dispersion radiale.
La diminution de la force d’entrainement du transfert de chaleur, due a la dispersion axiale.
Quand l’influence de la dispersion axiale sur le transfert de chaleur est absente (passage de grandes
quantites de liquide) cp est fonction de Pr. Si les tourbillons de Taylor sont presents, la valeur de cp
est independante de Re,. En dessous du nombre critique de Reynolds rpdiminue a cause d’un melange

2000
 A mechanism for heat transfer

axial insuffisant. Avec un Ccoulement restreint, I’influence de la dispersion axiale sur la force d’entrain-
ement du transfert de chaleur est importante. La diminution apparente du coefficient de transfert de
chaleur du a cet effet peut &re calculee. Pour se faire il faut appliquer le modble d’ecoulement sans
melange et avec dispersion, dans lequel le coefficient de dispersion axiale est calculC d’apres la
deviation standard de la courbe de distribution du temps de passage.

Zusanunenfassung-Der Mechanismus der Wkmetibertragung in einem SSHE setzt sich aus drei
Stufen zusammen, namlich:
Eindringen von WLme durch Warmeleitung in einer diinnen Schicht nahe der WLmetibertragungs-
obertlache in dem Zeitraum zwischen zwei Schabvotgangen.
Teilweiser Temperaturausgleich in der sich an der Schaufelkante bildenden Grenzschicht.
Konvektiver Radialtransport aus der gekiihhen Schicht zur Masse der Fliissigkeit.
Taylorsche Wirbelstromungen tragen zu diesem Radialtransport bei.
Die Warmetibertragungsmessungen wurden unter verschiedenen Bedingungen durchgefiihrt.
Der Warmeiibertragungskoeffizient kann mittels einer sich aus der Penetrationstheorie ergebenden
Gleichung, abgeiindert durch einen empirisch ermittehen Korrektionsfaktor cp beschrieben werden.
Der Faktor ‘pumfasst:
Die IJnvollstlndigkeit des Temperaturausgleichs in der Grenzschicht.
Die Wirkung der Radialdispersion.
Die Abnahme in Triebkraft zum Wimeaustausch infolge axialer Dispersion.
Wenn eine Wirkung der axialen Dispersion nicht vorhanden ist (hohe Stoffstriimungsgeschwindig-
keiten) wird cp eine Funktion von Pr. Wenn Taylorsche Wirbelstriimungen vorhanden sind, wird der
Wert von 9 unabhangig von Re,. Unterhalb der kritischen Reynoldsschen Zahl nimmt cp ab infolge
ungeniigender Radialmischung.
Bei niedrigen Stoffstrijmungsgeschwindigkeiten wird die Wirkung der axialen Dispersion auf die
Triebkraft zum Wiirmeaustausch bedeutend. Die scheinbare Abnahme des Warmetibertragungs-
koeffizienten infolge dieser Wirkung kann errechnet werden. Zu diesem Zweck wird Pfropfstromung
bei dem Dispersionsmodell angenommen, wobei dann der axiale Dispersionskoeffizient aus der mitt-
leren quadratischen Abweichung der Verweilzeitverteilungskurve berechnet wird.

2001