You are on page 1of 15

Chemical Engineering Science-, Vol. 44, No IO, pp. 2207.2221. L989. ooo9-2509/89 $3.oO+O.

W
Printed in Great Britain 8 1989 Pergaman Press pk

LASER-DOPPLER MEASUREMENTS OF TURBULENT-FLOW


PARAMETERS IN A STIRRED MIXER

H. WU and G. K. PATTERSON
Department of Chemical Engineering, University of Arizona, Tucson, AZ 85721, U.S.A.

(Received 24 April 1987; accepted 29 Nouember 1988)

Abstract--It is recognized that a detailed knowledge of turbulence parameters, as well as velocities, can aid
in understanding and modelling mixing rate dominated phenomena in stirred vessels. Turbulent-Row
parameters were measured in a baffled, Rushton turbine agitated vessel with a laser-Doppler velocimeter.
The necessary corrections for the periodic, nondissipative velocity fluctuations in the near impeller region
were made by an autocorrelalion method. Two components of periodic fluctuation, one corresponding to
impeller blade frequency, the other corresponding to twice that frequency, were found to be significant. With
the periodicity removed, meaningful turbulence data including turbulence intensities, autocorrelation
functions, turbulence scales, energy spectra, and turbulence energy dissipation rates were obtained. Integral
scales and turbulence energy dissipation rates were a particular objective in this work because of their
usefulness in modelling local mixing rates in turbulent flows. From an energy balance around the impeller
and impeller stream, it was found that 60% of the energy transmitted into the tank via impeller was
dissipated in that region, and 40% was dissipated in the bulk of the tank. An equation for calculating local
energy dissipation rates from resultant fluctuation velocities and resultant turbulence macroscales,
312
E = A”, appeared adequate. Constant A was found to be 0.85.

INTRODUCTION spectra(Kim and Manning, 1964, Cutter, 1966; Sato et


Mechanically stirred tanks are widely used in in- al., 1967, 1970; Mujumdar et QI., 1970; Rao and
dustries for a variety of mixing processes such as Brodkey, 1972; Komasawa et u1., 1974; Fort et al.,
blending miscible or immiscible liquids, dispersing 1974; Gunkel and Weber, 1975; Van? Riet et al., 1976,
gases or solids in liquids, mixing reactants, homo- Van Der Molen and Van Maanen, 1978) have also
geneous or heterogeneous, to effect chemical reac- been measured, but with more inconsistent results
tions, etc. The design of these mixers and the associ- than the mean velocity measurements because of the
ated principles have been reported and discussed (Uhl inherent measuring difficulties of turbulent flows and
and Gray, 1966; Brodkey, 1975; Nagata, 1975; the different experimental techniques with different
Ulbrecht and Patterson, 1985). The traditional design limitations appiied. The performance and reliability of
methods, based on correlations of overall vessel-aver- the experimental methods such as flow displacement,
age parameters, have been satisfactory in some appli- pitot tubes, photographic techniques, and hot-wire,
cations, but they cannot provide information about hot-film and laser-Doppler vclocimeters were reviewd
the “local” phenomena of mixing and their inter- by Mujumdar et al. (1970), Rao and Brodkey (1972)
actions with other transport and chemical reactions. and Patterson (1983).
These local phenomena are especially important when The nonrandom velocity fluctuation caused by the
a good quality of mixing is desired. To reach some periodic passage of impeller blades makes more com-
theoretical understanding of turbulent mixing, par- plicated the turbulence structure in the region close to
ticularly its local effects, a detailed knowledge of flow the impeller. Van? Riet and Smith (1973, 1975) detec-
and turbulence characteristics within the mixer is first ted a pair of strong trailing vortices originating from
required. behind each of the blades. The periodic sweeping of
Most studies on the stirred-tank flow have been these vortices over the stationary measuring point was
empirical rather than theoretical due to the extreme the source of the nonrandom, periodic fluctuation.
complexities of the flow. Flow patterns (Sachs and This periodic fluctuation, pseudo-turbulence, as called
Rushton, 1954; Metzner and Taylor, 1960; Kim and by Van? Riet et al. (1976), is embedded in the total
Manning, 1964; Cutter, 1966; Schwartzberg and fluctuation resulting in correlation functions showing
Treybal, 1968; Cooper and Wolf, 1968; Mujumdar et permanent periodic oscillation (Mujumdar et al., 1970;
al., 1970; Rao and Brodkey, 1972; Gunkel and Weber, Rao and Brodkey, 1972), and energy spectra exhi-
1975), impeller pumping capacities (Sachs and biting either a peak at the impeller blade frequency
Rushton, 1954; Cooper and Wolf, 1968; Gunkel and (Kim and Manning, 1964, Rao and Brodkey, 1972,
Weber, 1975), and power consumption rates Sato et al., 1967, 1970; Komasawa et al., 1974) or
(Calderbank, 1958; Bates et al., 1963; Schwartzberg multiple peaks at that frequency and its harmonics
and Treybal, 1968; Bertrand et al., 1980) in various (Mujumdar et al., 1970;Fort et al., 1974; Gunkel and
mixers have been measured and correlated. Turbu- Weber, 1975; Van’t Riet et al., 1976; Van Der Molen
lence intensities, correlation functions, and energy and Van Maanen, 1978). This turbulence periodicity

2207
2208 H. Wu and G C. PATTERSON

was also found in the concentration field (Manning plication of suitable convection velocities, and esti-
and Wilhelm, 1963; Reith, 1965). It must be excluded mation of turbulence scales and energy dissipation
in any calculation of turbulence parameters since they rates.
are mainly, if not entirely, determined by the irregu-
larity and randomness of the real turbulent motion APPARATUS AND DATA HANDLING
[see Hinze (1975)]. To account for this pseudo-turbu- The apparatus used in this work consisted of a
lence, Rao and Brodkey (1972) considered the sum of baffled stirred tank with a six-blade disk turbine, and
the positive and negative areas under the oscillating an LDV with computer data acquisition. A schematic
correlation function curve as the correct integral scale; diagram of the tank and impeller, and the coordinates
Mujumdar et al. (1970) and Gunkel and Weber (1975) used for the presentation of the results are shown in
obtained the contribution of the periodic fluctuation Fig. 1. The tank was made of plexiglass with 27 cm
from the narrow portion of the energy spectrum at the diameter and height. The four baffles were each one-
impeller blade frequency; Van Der Molen and Van tenth of the tank diameter in width. The turbine,
Maanen (1978) measured this periodic component which was of standard design with a diameter about
directly by recovering the instantaneous velocity from equal to one-third of the tank diameter, was placed
the correlator which was triggered at impeller blade one-third of the way up from the tank bottom. The
frequency; Laufhutte and Mersmann (1985) assumed tank, filled with water as the working fluid, was open
the periodic fluctuation velocity to be proportional to at the top, and was placed in a square tank also filled
the impeller speed, and obtained it by extrapolating with water to minimize optical distortion.
the value in the laminar range into the turbulent The LDV used was a standard one-direction model
range. by DISA with coaxial optics operating in a dual-beam
The turbulence parameters of most direct relation differential mode (fringe mode), and using a tracker for
to the mixer capacity and performance are the turbu- signal processing. The laser was a 15-mW He-Ne
lence scale and turbulence energy dissipation rate. The Hughes 3327H-PC. A Tektronix 60-MHz oscilloscope
former is a measure of eddy size, i.e. a measure of was used to monitor the signal quality. The measuring
intermaterial surface area for molecular mixing; the volume, i.e. the intersection point of the two focused
latter, according to Kolmogoroff, is an important laser beams out of the optical unit, was an ellipsoid
characteristic of turbulence when the Reynolds num- with 0.5 mm diameter and 1 mm length. This size was
ber is sufficiently high. Some mixing models that stress about the same order of magnitude as the turbulence
molecular transport, such as Corrsin’s (1964) isotropic microscales. making direct measurement of these
mixer model, Ottino et al’s (1979) lamellar model, and
Angst et al’s (1982) reaction-diffusion eddy model, are
primarily based on a known field of energy dissipation
rates and length scales. The estimation of these two
parameters in stirred tanks has been made by Cutter T/IO-
(1966), Sato et al. (1967, 1970), Mujumdar et al. (197Q
Rao and Brodkey (1972), Komasawa et al. (1974),
Gunkel and Weber (1975), Okamoto et al. (1981),
o/4

J-El
Barthole et al.(1983) and Laufhutte and Mersmann z
(1985). They all concluded that energy was mostly Id +
dissipated in the impeller stream, except Gunkel and D/5
f
Weber (1975) who found most energy dissipated in the +0=9.3cm-i f
bulk. The results available, however, are very in- T/3
consistent
sible.
and no reconciliation of them seems pos-
I
AT=27cm-
The inconsistency of the previous sets of data may
be caused by the various measuring methods used,
some of which were not ideal. For a stirred-tank flow
with high intensity of turbulence and in some lo-
cations poorly defined flow directionality, an ideal
measuring device should have a high-speed, wide-
range, linear and direction-sensitive transducer re-
sponse, and, more importantly, should not interfere
with the flow. The laser-Doppler velocimeter (LDV)
meets all these requirements despite its ambiguity
noise problem (George, 1975; Durst et al., 1976;
Durrani and Greated, 1977). The objective of this
study was to use the LDV to measure the flow and
turbulence parameters in a standard stirred mixer.
Particular emphasis was placed on accuracy of
measurements, correction of pseudo-turbulence, ap- Fig. I. Tank and impeller with dimensions and coordinates.
Laser-Doppler measurements of turbulent-flow parameters 2209

scales impossible. However, with an extrapolation which correspond to impeller Reynolds numbers
method described in the next section a rough esti- (ND2/v) of 1.4, 2.9 and 4.3 x 104, respectively. Most
mation of these scales can still be made. The light- results reported here are the data for the 45” plane and
scattering particles were M-pm styrene spheres with 200 rpm impeller speed.
almost the same density as water. The general princi-
ples and techniques for the LDV can be found in Durst
et al. (1976) or Durrani and Greated (1977). RESULTS AND DISCUSS&ON
A special feature of the tracker used in this study Mean velocities
was its ease of computer interfacing. Through its built- The profiles of the mean radial, tangential and axial
in analog-to-digital conversion circuit, the tracker velocities normalized with the impeller tip velocity are
transmitted data at a constant speed to the interface shown in Figs 2, 3 and 4, respectively. The jet-like
with the computer. The constancy of the data speed, pattern of the impeller stream, the center-line vel-
which is important to the computation of time- ocities decreasing and the flow entrainment expanding
dependent turbulence quantities, was always main- with the increase in radial distance from impeller, is in
tained even if the tracker was not “in lock”. When the agreement with that of earlier works (Sachs and
lock detector indicated a signal “drop out”, which Rushton, 1954; Kim and Manning, 1964; Cutter, 1966;
might be caused by weak signals, discontinuous par- Cooper and Wolf, 1968; Mujumdar et aE., 1970; Rao
ticles through the measuring volume, or any type of and Brodkey, 1972; Gunkel and Weber, 1975). It
noise, an especially small digit was supplied to the
computer as a warning flag. These false data acquired
I
by the computer were then detected and excluded in
200 r pm
any computation. The data transmission speed, how- rkm) -
ever, was limited to a value in accordance with the 0 5.0

frequency range selected on the tracker within which


the Doppler signal was processed. The highest frc-
quency range used in this work was from 0.3 to
3.3 MHz, which gave approximately a data speed of
7000 Hz. This maximum data speed was about the
same as or even lower than the rate of the fastest
change of velocity fluctuations, adding one more
difficulty for the measurement of turbulence micro-
scales. Nevertheless, these resolution problems, the
size of the measuring volume and the finite speed of the
data response, should place little limitation on the
macroscale measurement because their effects are
profound only in the high-frequency range of turbu- -2.
lence, while most turbulence energy, to which the Bo d I (312 013 ial ds 016 0; 08
uru,,
macroscale is a corresponding characteristic, is con-
tained in the lower-frequency range. Fig. 2. Mean radial velocity profiles at various radial
positions.
A Commodore microcomputer with 32-k memory
was used for data acquisition and computation. In
each measurement, 10,240 instantaneous velocity val-
ues were collected for computation. The program was 200 rpm
written in machine code to ensure high-speed data rkml -
0 5.0
handling. It included a dummy subroutina which can 0 6.0 _
h 7.0
slow down the speed of data read-outs if needed. Thus, l 7.7
while the maximum data speed is required for the 0 9.0 -
p 10.5
measurement of high turbulence, a slower data speed
can be used for lower turbulence to assure sufficient
averaging time for the calculation of turbulence quan-
tities from those fixed number of instantaneous data.
The calculation steps in order were mean velocity,
root-mean-square fluctuation velocity, Eulerian
autocorrelation function, then turbulence energy spec-
trum. The periodic velocity fluctuation was corrected.
The autocorrelation function and energy spectrum
before and after the correction were plotted on the
computer for visualization. Measurements were made -25
0.0 0 I 0.2 03 04 0.5 0.6 0.7 0.6
in the impeller stream and in the near region around U&Jt,p

the impeller, at planes 45 and 0” with the baffle planes. Fig. 3. Mean tangential velocity profiles at various radial
The impeller speeds used were 100, 200 and 300 rpm, posltions.
2210 H. Wu and G. K. PATTERSON

_ 2.5, , sistency of measurements. The only work in the litera-


ture that checked such a flow balance was by Gunkel
and Weber (1975). They also found a good data
consistency for average velocities.

Impeller pumpiny capacities


The radial pumping capacities of the impeller were
calculated as the integral of the mean radial velocities
oo- over the entire impeller stream, as in eq. (l), with z1
and z2 being the axial positions where the mean radial
-05 -
velocities reached zero. These pumping capacities,
-10 - normalized by ND3, along with other data are plotted
in Fig. 5. To compare the results of different studies, as
-1.5 -
pointed out by Rao and Brodkey (1972), one shouId be
-zo- aware of the effects of the differences in vessel and
impelIer design, experimental methods, and operating
-251 I
-010 -005 000 005 0 IO 0 15 0.20 conditions. Our data show that the normalized
Wh,p pumping capacities are independent of the impeller
Fig. 4. Mean axial velocity profiles at various radial posi- speed, and increase with radial distance from the
tions impeller tip due to the fluid entrainment. This result
generally agrees with others. Sachs and Rushton’s
(1954) data with a four-blade disk turbine using a
should be noticed that these mean velocity distri-
photographic technique and Rao and Brodkey’s
butions are not symmetrical about the z = 0 plane,
(1972) data with a six-blade open turbine using hot-
but shift upwards as the wall is approached. This is
film velocimetry are both below our data. Gunkel and
because the impeller is not symmetrically located and
Weber (1975) and Cooper and Wolf (1968) both used
the top of the tank is a free surface. The axial mean
hot-wire anemometry and the same type of turbines as
velocities, as shown in Fig. 4, are much smaller than
used in this work, but their data are somewhat differ-
the other two components throughout the impeller
ent from ours. Cooper and Wolf (1968), with different
stream. At r = 10.5 cm, they almost vanish. This may
sizes of turbines, found that at r/R = 1, where there is
be the place where the impeIler stream is about to split
no fluid entrainment, the values of Q,/ND3 varied
into two streams, one flowing upward, the other
between 0.73 and 0.89. Gunkel and Weber’s (1975)
flowing downward into the bulk of the tank.
velocity data appeared to be the highest in the litera-
The reliability of the data can be tested by checking
ture.
the volume flow rate balances around some control
volumes containing the impeller. Table 1 shows the
Turbulence correlation functions
results of such balances around two control volumes,
As described in the Introduction, the periodic vel-
one is set as the region from r = 0 to 5 (cm), the other
ocity fluctuation resulting from the disturbance of
from r = 5 to 9 (cm), both ranging from z = -2 to 2
(cm). The radial and axial flow rates through the
controt surfaces were calculated as follows:
PZ
Q, = 27cr U,dz (1)
s =I

Q,=2rr i ‘*r(U,,-UU,,)dr (2)


J.,
wherez, =,--2cm,z, =2cm,r, =Oor5(cm),andr,
= 5 or 9 (cm). In eqs (1) and (2) the average values of
mean velocities at the 45” and the 0” plane were used
so that axisymmetry could be applied. The differences
between the inflow and outflow as shown in Table 1
are within 4%, providing strong support to the con-

Table I. Flow rate balance around the impeller at 200 rpm, 001 I
-2GZ<2 (cm) IO I2 14 16 18 2.0 22 24

Flow in Flow out Difference


Region Fig. 5. Profiles of radial pumping capacity in the impeller
(Us) (I/s) (“/)
stream: (A) this study (0 = 1OOrpm; II = 200 rpm, x
OGr<5(cm) 2.24 2.32 3.6 = 300 rpm); (B) Gunkel and Weber (1975); (C) Sachs and
5Sr<9(cm) 5.29 5.22 1.3 Rushton (1954); (D) Cooper and Wolf (19h8); (E) Rao and
Brodkey (1972).
Laser-Doppler measurements of turbulent;flow parameters 2211

impeller vortices must be subtracted from the total


turbulence. This correction can be made by an
autocorrelation method. By assuming some particular
function to approximate the periodic fluctuation vel-
ocity, its contribution to the autocorrelation function
can be calculated explicitly. Then other turbulence
data such as’ root-mean-square ffuctuation velocities,
turbulence scales, and energy spectra.can be obtained
from the corrected autocorrelation functions. The best
function to represent such a periodic fluctuation veloc-
ity would be a complete Fourier series as a function of
time or frequency, but only the first two terms, one
corresponding to the impeller blade frequency, the
other corresponding to twice that frequency, were
used. The reasons are: (1) more terms are not practical
in use; (2) in most cases only two major peaks were
observed in the energy spectra; and (3) if too many
terms are used, the higher-frequency terms will be
T

essentially correlated with the real random turbulence, Fig. 6. Example of the periodic components of Eulerian auto-
making the periodic part inseparable from the random correlation function: (- - -) first component [ = A, cos
(Znr/T)], (- . -) second component [ = A, cm (47x/T)],
part. Therefore ) sum of first and second components.
(
u,,,(t) = &,“d(f) + u,,,(t) (3)
Uper(t) = A cos 2mnt + B cos 4mxt (4) i._z
%ot- %.nd
+p (7)
txr
where u is the fluctuation velocity, a function of time, 2 psr = A2/2 + B2/2. (8)
subscripts tot, rand and per denote the total, random
and periodic component, respectively, m is the impel- Equations (7) and (8) were used to correct the
lcr blade frequency, and A and B are constants, the turbulence intensities; the results are presented in the
amplitudes of the two periodic components. Assuming next section.
there is no correlation between periodic and random It should be noticed that the original correlation
turbulence, the autocorrelation functions can be ob- curve in Fig. 8 is not as symmetrical as that in Fig. 7
tained as because the proportion of the second periodic compo-
nent, B2/2, near the edge of the impeller blades is
larger than that at the center of the impeller tip. This
uCr)u(t - z),,, = u(t)u(t - ZLd + u(t)u(t - ?Ipsr
behavior can be explained by the vortex structure
(51
detected by Van? Riet et al. (1976). The periodic
A2 vortex disturbance is not a perfect sine-wave function
u(tM-z),,, = -cos2mxr +$cos4mnr. (6) of time as the trailing vortices continuously sweep
2
over the fixed measuring point. In most of the
measurements, the second periodic component was
When T is large the random part of the correlation
found to be smaller than the first one, yet in some
function approaches zero, the only thing left is the
locations it was significant; using only the first compo-
pure periodic correlation function. An arbitrary part
nent was not enough to remove all the periodicity.
of such a periodic function is shown in Fig. 6 where T
= l/m, i.e. the time period for the impeller to rotate a
distance interval between two blades, and A, = A2/2, t?$)
A, = BL/2. The oscillation amplitudes, A, and A,, can
be determined from the correlation function values at
points a-e. Since the correlation function at large z
oscillates nicely with the impeller blade frequency, its
amplitudes and even the impeHer speed can be easily
found and checked. The whole correlation function
was then corrected with eqs (5) and (6). The success of -02-

this correction can be seen from Figs 7 and 8. The -04 -


original correlation coefficients oscillate at the blade
-06-
frequency; the corrected ones decrease with 5 toward 0 5 IO 15 20 25 30 35 40 45 50 55
zero with essentially no period&y. Notice that these r (msec)
correlation coefficients are the correlation functions
normalized with the total mean-square fluctuation Fig. 7. Original and corrected Eulerian autocorrelation
functions near the center of the impeller tip. Measured in r-
velocities, I&. These & also contain a random and a direction at 200 rpm, r=5 cm, I =0 cm: (-) original,
periodic part. At T = 0, eqs (5) and (6) become (- - -) corrected.
2212 H. Wu and G. K. PATTERSON

04

02

00

-0 2

Fig. 8. Original and corrected Eulerian autocorrelation


function near the corner of the impeller tip. Measured in r-
direction at 200 rpm, r=S cm, z= 1 cm: (---) original,
(- - -) corrected. 2.5 1 I
0.00 0.05 010 0 IS 020 025 030 035
4’utlp

Turbulence intensities Fig. 10. Profile of tangential turbulence intensity near the
With eqs (7) and (8) the random part of the turbu- impeller tip.
lence intensity can be separated from the total. The
distributions of these turbulence intensities near the
impeller tip are presented in Figs 9-l 1. The periodic as
r=5cm
well as the random turbulence intensities, like the 200 rpm
20 -
mean velocities, were proportional to the impeller A tomi
V random
speed. This can be seen in Fig. 9. For the radial and 15-
tangential components, the largest correction, i.e. the
to -
largest contribution of the periodic fluctuation, oc-
curred at z=O, but for the axial component it oc- 0.5 -
curred at 22/w = 0.5 and - 0.5. This is again consistent
with the trailing vortices described by Van? Riet and 00 -

Smith (1973, 1975) and Van’t Riet et al. (1976). They -0.5 -
found that a pair of vortices, one above, the other
below, the disk originated from behind each blade and -10 -

rolled out into the impeller stream with strong but


-1.5 -
opposite directions of rotation around their axes,
which, at a short distance out from the impeller, were -2.0 -

nearly perpendicular to the blade. Thus at the impeller I


- 2.5 ’
center plane, which is approximately at the middle 0.00 005 0 IO 0 I5 020 0.25 0.30, 035
WJl,P

Fig. 11. Profile of axial turbulence intensity near the impeller

Y-----
tip.

15-

IO -
between the two roiling vortices, the circumferential
velocities of both vortices lie in the horizontal (r-0)
05 - plane, making maximum disturbance in the radial and
tangential directions, but having little influence on the
00 -
axial direction. By contrast, if the measuring point is at
-05 - the center plane of either of the two vortices when the
vortices are sweeping over the point, the rotation of
-10 -
vortices will have little impact on the radial or tangen-
-15 - tial direction but maximum disturbance on the axial
direction. Therefore, from the correction distributions
-2O-
in Figs 9-11 one can expect that z = 0 is approxi-
-25L I mately at the edges of the two vortices, 22/w = 0.5 and
000 005 0 IO 0.15 020 025 030 035
-0.5 are approximately at the centers of the upper
u:AJ,,c
and the lower vortices, respectively, and consequently
Fig. 9. Profile of radial turbulence intensity near the impeller the diameters of the vortices are each about half the
tip. width of the blade.
2214 H. WV and G. K.PATTERSON

E,(n) 1 ' . """I 0 ' "'*"I 1 and Brodkey (1972) and Komasawa et al. (1974), a
_ slope of * may not necessarily indicate the existence
4ufTE. ,o~_ ,fY of an isotropic field. Such a slope region in fact has
t often been found in shear flows with moderate
Reynolds numbers. At higher frequencies, say beyond
IDo* -4--Q& 200 Hz, no further steeper slope was observed, prob-
,” ably because the spectrum has been obscured by the
! L\ i’‘11 presence of ambiguity noise caused by the limited
frequency response of the tracker. The limitation of
10-I t T;--
this sort in LDV applications has been reported by
: \ George (1975), Van Maanen et al. (1975) and Durst et
[ \ ‘r i
% al. (1976).
10-Z- ‘. The original spectra and their corrections measured
t
200 rpm \ in the other two directions were all similar to those in
r=5cm, z=Ocm t \I
- -A- orlglnol ! & the radial direction. Beyond r = 1.9R, the periodic
- --.?-- COrrCCtCO fluctuation disappeared, and no peak showed up on

m-3. ' ' '.**'a' ' 1 "'#'a' 0 ' ' ..8,1 the original spectra. In view of the success of correc-
100 IO' 102 103
n [seci) tions to the correlation functions, then to the energy
spectra, the approximations used for the periodic
Fig. 14. One-dimensional (radial) energy spectrum near the fluctuation velocity and periodic correlation function,
center of the impeller tip.
eqs (4) and (6), seem to be fairly good. In the literature,
only Mujumdar et al. (1970), Gunkel and Weber
(1975) and Laufhutte and Mersmann (1985) made the
other, a smaller one, at about 40 Hz, appear on the correction for turbulence intensities, no correction for
original spectrum. It also exhibits local minima at the correlation functions and energy spectra has been
n cz 5, 35 and 55 (Hz). These maxima and minima are reported before.
all attributed to the periodic fluctuation embedded in
the spectrum function. To analyze whether this behav-
ior is consistent with the initial assumption made for
the periodic velocity fluctuation, one can check the Turbulence macroscales
The Eulerian time macroscales were calculated
characteristics of the periodic energy spectrum. Sub-
stitution of eq. (6) into eq. (10) yields from the random autocorrelation coefficients

A% sin (2nnjm) E%sin(2~n/m) O”u;(t)u,(t - ~),.“d


E,(n),,, = + uE, = dr (13)
7c(?l2 - mZ) 7Z(n2 -4m2) u:.,d
(12) where i means the J-, 0 or z component. The time scales
where a blade time,
l/m, instead of infinity has been were then converted to the length scales with suitable
used as the upper integration limit in the Fourier convection velocities, UCi:
transformation. It can be seen that E,( n)Pcr is a
I+ = UC,UEi. (14)
periodic function oscillating around zero. In the low-
er-frequency range, or when 1n2 - m21or 1rt’ - 4mZ 1is These convection velocities can be approximated with
small, IE,(n),..I is relatively large compared with the the local mean velocities if Taylor’s frozen field hy-
random spectrum. Thus the original spectrum, pothesis can be applied. Sato et al. (1967, 1970),
E,(n),,,, should exhibit the same IocaI maxima and Mujumdar et al. (1970), Rao and Brodkey (1972),
minima as E, ( PI&_ has, namely, maxima at n = m and Komasawa et al. (1974) and Barthole et al. (1983) all
2m; minima at n = m/4, 7m/4 and llm/4. This predic- used this hypothesis to transform temporal results
tion is evidenced in the original spectrum shown in into spatial results. However, for high-intensity turbu-
Fig. 14. In a higher-frequency range E,(n),,, is smaller lent shear flows, such as the flows in stirred tanks, the
but may still be important because E1(n),,,d is also applicability of this hypothesis is questionable. In this
small at high frequencies. This means that El(n),,, study a method proposed by Heskestad (1965) was
may also show local extrema at high frequencies. The adopted. He derived from equations of motion the
original spectrum in Fig. 14, however, does not show convection velocity in rectangular coordinates for the
this oscillating behavior at frequencies beyond 60 Hz, flow where only one component of mean velocity
but it would have done if denser data had been exists:
calculated. Notice that the sine functions in eq. (12) are
vigorously oscillating when n is large.
The corrected spectrum seems to have retained the
typical characteristics of a random turbulence energy
spectrum. Most of the energy is contained in the
lower-frequency eddies; in the higher-frequency range This equation has been extended by Van Doorn (1981)
a % slope emerges. However, as mentioned by Rao to suit three-dimensional flows:
Laser-Doppler measurements of turbulent-flow parameters 2215

of the blade or the diameter of the trailing vortices.


For this region of fluid, the impeller blades should be
the only turbulence sources. At 2z/w = 1 and - 1, i.e.
near the blade corners, the turbulence is mainly gener-
where i, j and k denote the three principal components, ated by the blade corners rather than by the central
and the fluctuation velocities are the random parts part of the blade or by the vortices. The macroscales
only. This convection velocity was used in this work there are probably a measure of the sharp blade
although its form in cylindrical coordinates should corners. Outside the core of the impeller stream, say
include two more terms, the centrifugal and Coriolis )22/w I> 2, the macroscales are much larger than those
force term. These two terms may be significant when in the stream. In this region, part of the fluid is
r/R is small; their neglect here only serves as an entrained from the bulk and, in its traveling history,
approximation. may have interacted with the impeller, baffles, tank
Qualitatively, eq. (16) agrees with some reports in wall, or other parts of the fluid. Thus the macroscales
the literature. It indicates that the convection velocity there should be an “average” size of these turbulence
is larger than the local mean velocity. Goldschmidt et sources. Clearly, the macroscales are a function of
al. (1981) found a similar result in a plane jet from the location. Due to the complexity of the flow, the
space-time correlation measurements. If turbulence is impeller stream becomes the only part of the tank
isotropic and the mean flow is in direction 1, eq. (16) is where the macroscales can be identified, i.e. half the
reduced to width of the blades.
Figure 16 presents the distributions of macroscales
2

Lumley (1965)
u,:

found
= u:
(
by
1+5u'.
u:

spectral
) analysis
(17)

that
at the center of the impeller stream. The data of this
work show that the three components of the scales
have about the same order of magnitudes, ranging
from 0.25~ to OSw, throughout the impeller stream.
(1 + 5uz/LJf) was the factor needed to correct Taylor’s Minimum scales are located at a short distance from
hypothesis in the calculation of energy dissipation the tip, which is also the place where the random
rates. This correction method was later used by turbulence intensities are the highest. Away from the
Okamoto et al. (1981) in the study on stirred-tank proximity of the blades, the radial and axial compo-
flows. nent slightly increase with I while the tangential
Figure 15 presents the calculated macro length component remains about constant. In the literature a
scales near the impeller tip at 200 t-pm. These length range from 0.1~ to 0.5~ has been reported. Cutter
scales were found to be independent of impeller speed (1966) obtained radial and tangential rrtacroscales
because, while the time scales decreased with the from space correlation measurements. Mujumdar et
impeller speed, the convection velocities increased al. (1970) and Komasawa et al. (1974) calculated radial
with that speed., It is believed that turbulence macro- macroscales from the zero frequency intercept of one-
scales, which are a measure of energy-containing dimensional energy spectra. Rao and Brodkey (1972)
eddies, should be about the same order of magnitude
as turbulence-generating sources. This can be seen in
Fig. 15. In the region near z = 0, the macroscales are
about equal to 0.35w, which is close to half the width

Fig. 16. Radial distribution of turbulence length microscales


at the center of the impeller stream: (A) this study (200 rpm,
-25l
0.0
"
0.2

0.4

0.6

06

IO
radial); (B) this study (200 rpm, tangential); (C) this study
(200 rpm, axial); (D) Cutter (1966) (radial); (E) Cutter (1966)
(tangential);(F) Mujumdar et al. (1970) (radial. 6-in. impeller);
Fig. 15. Axial distribution of turbulence length macroscales (G) Mujumdar et al. (1970) (radial, 5-in. impelier); (H)
near the impeller tip. Komasawa et al. (1974) (radial).
2216 H. Wu and G. K. PATTERSON

used correlation functions, and Sato et al. (1967) used change in tE with At was linear and that even using the
energy spectra; both obtained macroscales about smallest At, i.e. the fastest data speed, the 7E was still
equal to 0.3~. not converging to some constant value. This is again
an evidence of the LDV resolution limitation. Never-
Turbulence microscales theless, from the variation of tE with At as shown in
The turbulence microscales, A,, cannot be deter- Fig. 17, it can be suspected that a meaningful ~~ may
mined from the spectra obtained in this study with an be located somewhere between the value obtained at
LDV because they were significantly distorted at high the smallest At and that at the At = 0 intercept._ With
frequencies due to the LDV ambiguity noise. Thus TV estimated this way, Ar then A, c A,/,,/2 and the
these scales were calculated from the time derivaiives turbulence Reynolds number, Re, = u’i,/v, were ob-
of fluctuation velocities with the use of convection tained. It was found that at the center of the impeller
velocities derived in eq. (16), both being measured in stream 1, was within 0.3-1.0 (mm) and Re, was within
the radial direction: %X190. These values and the values available in the
literature were listed in Table 2.
1 1 au, z Since the microscales are so difficult to estimate, the
2 (18)
TE =-(-->
2u; fJt t=0 values in Table 2 ranging from 0.3 to 3 mm can be
considered to be in reasonably good agreement.
1, = U&7, Mujumdar et aZ. (1970) calculated IE, from the frc-
where zE is the Eulerian time microscale. It is under- quency where the dissipation function, n2E,(n), was a
stood that only the random parts of turbulence quan- maximum. Sato et al. (1967) obtained it from the
tities can be used in eqs (18) and (19). Similar to the integral of the dissipation function over the entire
corrections of T, the random parts of mean-square frequency domain. For these methods to succeed, the
derivatives of fluctuation velocities were obtained as
spectra must be free of any unwanted turbulence such
as the periodic turbulence in the near impeller region,
and must have a good accuracy, particularly in the
higher-frequency range, to ensure the convergence of
dissipation functions. Rao and Brodkey (1972) used
-(2A* + 8B2)m2z2. (20) the same approach as this study used, but did not
Unfortunately, the calculation of the time derivative of report the variation in the derivative of fluctuation
the fluctuation velocity was difficult. This derivative velocity with respect to the data speed. All of these
was found to be sensitive to the speed of the data read- literature results were obtained in the impeller stream.
out. It increased with the data speed, that is, increased Gunkel and Weber’s (1975) data, however, were ob-
with the decrease in the time interval between individ- tained in the bulk. They calculated them from the total
ual read-outs, At. However, the mean-square fluctu- power consumption rates, based on the isotropy as-
ation velocity remained about constant with the sumption and on their finding that most of the energy
change of At. Therefore the Eulerian time scale calcu- was dissipated in the bulk.
lated with eq. (18) decreased with the decrease in At.
Figure 17 shows this result. It can be, seen that the Turbulence energy dissipation rates
The energy dissipation rate in a particular region
can be determined by analyzing the energy balance
I I I r , 1 I I around the region. In the impeller region or impeller
stream, where the turbulence intensity is high, the only
energies concerned are the power input through the
25
impeller, the fluid flow kinetic energy, and the turbu-
lence dissipation energy; other forms of energy such as
20 the potential energy, pressure work, and viscous dissi-
pation energy are all assumed negligible. Therefore the
turbulence energy dissipation rate in a region is simply
equal to the difference between the inflow and the

Table 2. Turbulence microscale and turbulence Reynolds


number
i d,

000 00 01 0.2 0.3 04


I
05
At (msecl I Mujumdar
Reference

Sat0 et al. (1967)


et al. (1970)
(mm)

2.0-3.0
I .cL2.0
Re,

120-270
4CkI20
Rao and Brodkey (1972) 0.7-1.0 17C-245
Fig. 17. Variation of turbulence time microscale, calculated Gunkel and Weber (1975) 1.5-3.0 72-193
with eq. (18), with the time interval ofdata acquisition at the This study 0.3-1.0 50-190
center of impeller stream.
Laser-Doppler measurements of turbulent-flow parameters 2217

outflow of kinetic energy through the region boundary the differences in the measuring techniques and calcu-
plus, if this region contains the impeller, the power lation methods, his result agrees well with this work.
input via the impeller. By further neglecting the contribution of the axial
To be specific, the impeller region was defined as the kinetic energy flow rate, Cutter concluded that 20% of
cylindrical region containing the impeller with r from the total energy was dissipated in the impeller region,
zero to 1.08R and z from z, = - 2 cm to z2 = 2 cm; the 50% in the impeller stream, and 30% in the bulk. A
impeller stream was defined as the annular region very different result was reported by Gunkel and
from r = 1.08R to 2.26R with the same range of z as Weber (1975) who, using a similar approach to this
the impeller region. The radial kinetic energy flow rate work, found that most of the energy was dissipated in
through the surface at r is the bulk of the tank, not in the impeller region or

[ ( u, + u,)~ + ( u, + I+)* + ( U, + u,)‘]( U, + u,)r d0 dz


L J,, Jo

= p-w U,(U,Z+U,2+UZ+3u,2+utlgl+u,2)dZ (21)

where the cross and the higher-order correlation terms impeller stream. Besides the accuracy of velocity
have been neglected. Similarly. the axial kinetic energy measurements that may significantly affect the result
flow rate through the surface at z (either z, or zZ) with r of cncrgy dissipation rates, a possible reason for the
from r, to r2 is inconsistency between the result of Gunkel and Weber
and those of Cutter and this work is that the impeller-
to-tank diameter ratio used by Gunkel and Weber,
l/2, is larger than what was used by Cutter and in this
+ 3$)r dr. (22) work, l/3. Sato et al. (1970) has found that using a
larger impeller causes the turbulence characteristics to
In the calculation of these kinetic energy flow rates the be more uniformly distributed in the vessel. In other
fluctuation velocities should include not only the words, a larger impeller can spread turbulence not
random parts but also the periodic parts because the only into the impeller stream but also throughout
latter are also a form of kinetic energy. The power other parts of the tank, resulting in the energy dissi-
input through the impeller was calculated from P pation rate in the bulk being as significant as that in
= N,pN3D5, where N, = 5 was used. This value of the impeller stream. Gunkel and Weber (1975) at-
power number for open tanks with six-blade disk tributed their result of most of the energy being
turbines and Re > lo4 has been confirmed by Bates et dissipated in the bulk to the finding that the turbu-
a1. (1963) and Bertrand et al. (1980). lence in the bulk was more random, had a smaller
With these calculations it was found that about periodic component, and a higher relative intensity
30% of the total energy was dissipated in the impeller than the turbulence in the impeller stream. The valid-
region (the volume swept by the impeller blades ity of this reasoning is questionable. It should be the
projected back to r = O), about the same amount was
dissipated in the impeller stream, and the rest, about
40% was dissipated in the bulk of the tank. Clearly, IO 1 1 I 1
the dissipation in the impeller region and impeller
09 -
stream is significant. The fluid in this area represents
only 9% of the total. The energy dissipation rate per
unit mass there is about 15 times as much as that in
other parts of the tank. It should be mentioned that
since the error in the mean velocity data as discussed
before was within 4% and the error in the turbulence
velocity data was only slightly higher than that, the
error in the kinetic energy flow rates, which are
proportional to the cube of the velocities, was prob-
ably within 15%. Considering the difficulty in the
estimation of energy dissipation rates, this error
should be acceptable.
For comparison, the radial kinetic energy flow 0.1 -

rates, KE,, normalized with the total power con-


I I I I I I
00
sumption rate were plotted along with Cutter’s (1966) 10 12 14 16 16 20 22 24
r/R
data in Fig. 18. Applying Laufer’s (1954) approach, he
obtained the change in kinetic energy with respect to I Fig. 18. Radial flow rate of energy in the impeller stream:
from the transport equation of kinetic energy. Despite (0) this study (200 rpm), (~ ~ -) Cutter (1966).
2218 H. Wu and G. K. PATTERSON

absolute intensity of turbulence rather than the rela- 2.5


22
tive intensity that determines the energy dissipation 200 rpm
r(cm) -
rate if other turbulence characteristics are kept con- 0 50
stant.
In addition to the average distribution of energy
dissipation rates, an attempt was also made to seek an
expression to predict the local dissipation rates. Ac-
cording to Batchelor’s (1953) energy cascade theory,
the energy dissipation rate, which is a characteristic of
small eddies, can be calculated from the turbulence
velocity and the macroscale, these being the character-
istics of large eddies. He proposed on dimensional
z312
grounds that E = $A---- where A was an empirical
L ’
constant. In a nearly isotropic, grid-generated turbu-
lent Aow, Batchelor found that A z 1.0. This equation -2 5 t I
0 5 IO I5 20 25
and its results have been confirmed by experimental S/i
data in mariy flows that are close to be isotropic [see
Fig. 19. Distribution of local turbulence energy dissipation
Hinze (1975) and Tennekes and Lumley (197211. How- rate normalized with overall average energy dissipation rate.
ever, its validity in anisotropic shear flows is less clear.
Townsend (1976) and Tennekes (1977) suggested that
E oc q3f2/L, where q = +(q + z + 2). Townsend 35
(1976) further stated that, in self-preserving shear $
flows, the length scale used for the calculation of E 30
should be about 3 times the turbulence macroscale. It
is felt, however, that in three-dimensional flows the 25
most representative of the scale that is associated
with the resultant turbulence kinetic energy, y, must
20
be a “resultant” macroscale, L,,, = ,/L: + L$ + ,!,z.
Therefore
15

IO

Again, 4 should not contain any contribution of 5


periodic Auctuations. Each component of the resultant
macroscale can be calculated from the corresponding I I I I I I --1
4 12 14 1.6 I.8 20 22
time scale and convection velocity. By integrating the r/l R2
local values of q3”/Lrer over each section of the
impeller stream, and comparing the integral with the Fig. 20. Radial distribution oflocal energy dissipation rate at
energy dissipation rate in the section obtained from the center ut”the impeller stream: (A) this study (200 rpm), (B)
Cutter (1966), (C) Lauf-hutte and Mersmann (1985). (D) Rao
the calculation of kinetic energy flow rates, the con- and Brodkey (1972), (E) Komasawa et al. (1974), Okamoto er
stant A in each section of the stream was found. ill. (1981)
Throughout the stream an average value of 0.85 with a
deviation of &-10% was obtained. This value is re-
markably close to Batchelor’s. Antonia et al. (1980) E decays with the radial distance from the impeller
measured E in jets, and found that A should be about except in the near impeller region where most of the
1.0 for round jets and 0.5 for plane jets. turbulence is periodic, and the random turbulence is
Figure 19 shows the distributions of local energy yet developing. Cutter (1966), Rao and Brodkey (1972)
dissipation rates calculated with eq. (23), normalized and Laufhutte and Mersmann’s (1985) results are in
with the tank-average power consumption rate good agreement with ours. Komasawa et al. (1974)
through impeller. It can be seen that maximum dissi- and Okamoto et aZ’s (1981) data are somewhat lower.
pation rates occur at the central part of impeller Without correcting the non-dissipative periodic tur-
stream where turbulence intensities are high and bulence, Cutter and Rao and Brodkey overestimated E
macroscales are small. Similar to the turbulence inten- in the region close to the impeller. Cutter’s approach
sity distributions, double peaks appear on the curve at has been described. Rao and Brodkey used E
r=5cm. = 15v(u2/1,2). This method suffers from the uncer-
To compare with some literature values the energy tainty of the isotropy assumption and the extreme
dissipation rates at the center of impeller stream were difficulty in microscale measurements. Laufhutte and
replotted in Fig. 20. The result of this work shows that Mersmann applied Batchelor’s equation but assumed
Laser-Doppler measurements of turbulent-fow parameters 2219

a constant scale proportional to the impeller diameter. spectra after correction seemed to have retained the
However, the macroscale, in fact, varies with location typical spectra1 characteristics, but cannot be used to
in the tank; thus their c values will be too low in the calculate turbulence microscales or energy dissipation
region where the scale is small, and will be too high in rates because of their distortion at high frequencies
the region where the scale is large. Komasawa et al. due to the response limitation of the LDV.
and Okamoto et al. used the same equation as Rao By checking the energy balance around the impeller
and Brodkey did, but with the microscale being calcu- and the impeller stream, it was found that about 60%
lated from the dissipation function. As discussed in the of the total energy input was dissipated in that region,
previous sections, this method suffers from the strong and about 40% was dissipated in other parts of the
bias of periodic turbulence and the poor convergence tank. If E is a parameter for turbulent mixing to
of the dissipation function at high frequencies due to molecular scale, the impeller region and the impeller
the resolution limitation of the measuring instrument. stream would be the most important parts of the tank
Sato et al. (1967) and Barthole et al. (1983) also used for this process, although the role of the bulk is also
this approach and reported values close to what significant, An equation for calculating local energy
3/z
Komasawa et al. and Okamoto et al. obtained.
dissipation rates, E = A 5, was found to be adequate
res
with A z 0.85.
CONCLUSIONS

The three components of mean and fluctuation Acknowledgments-Partial support for this research as ob-
velocity have been measured in a fully baffled, turbine- tained from the National Science Foundation and Imperial
agitated vessel with an LDV. The accuracy of Chemical Industries Ltd.
measurements was confirmed by checking the volume
flow rate balance around the impeller. The stream
ejected from the impeller blades flowing toward the NOTATION

tank wall behaved like a jet. The mean velocities, A, A,, constants
pumping capacities, and turbulence intensities in the A,, B
impeller stream were approximately proportional to D impeller diameter
the impeller speed. Their values normalized with the E,(n) one-dimensional energy spectrum in fre-
impeller tip velocity became a function of position in quency domain
the tank only, not a function of impeller speed. These kinetic energy flow rate
results were in good agreement with many earlier turbulence macroscale
works. resultant turbulence macroscale,
Due to the periodic disturbance of the vortices ZZJL: + L: + L:
generated by the impeller blades, the measured fluctu- impeller blade frequency
ation velocities in the vicinity of the impeller tip impeller speed
contained a large portion of periodic components. power number
This periodic fluctuation was evidenced in the oscil- frequency
lating autocorrelation functions and in the peaked power consumption rate through impel-
energy spectra. By approximating this periodic fluctu- ler
ation with a two-component periodic function, the Q pumping capacity
random parts of the autocorrelation functions, turbu- turbulence kinetic energy, E ,(a,l-r + iI+
4
lence intensities, and energy spectra were recovered
from their originally biased quantities. The correc- +u:)
tions appeared successfuh the periodic characteristics R impeller radius
in the correlation functions and energy spectra were R, Eulerian autocorrelation function
totally removed. The results showed that the periodic Re impeller Reynolds number, z ND*/v
fluctuation dominated the turbulence field close to the ReA turbulence Reynolds number, I u’i,/v
impeller, but rapidly diminished away from the tip. At r radial coordinate
r = 1.5R it was reduced to about 20% of the total T = l/m; also tank height and diameter
fluctuation. This was also the place where the random t time
turbulence became fully developed and had the high- u mean velocity
est intensity in the impeller stream. uti* impeller tip velocity
The macroscales were obtained from the time in- I_4 fluctuation velocity
tegral of autocorrelation functions multiplied by pro- W impeller blade width
per convection velocities taking into account the flow z axial coordinate
three-dimensionality and anisotropy. At the center of
the impeller stream these scales were slightly smaller Greek letters
than the half width of impeller blades. Outside the turbulence energy dissipation rate
stream they were much larger. The microscales were 1 longitudinal microscale
also estimated in the impeller stream, and were found A: lateral microscale
to be about one-tenth of the macroscales. The energy V fluid kinematic viscosity
2220 H. Wu and G. K. PATTERSON

fluid density Gunkel, A. A. and Weber, M. E., 1975, Flow phenomena in


Eulerian time macroscale stirred tanks, Part I & II. A.I.Ch.E. J. 21, 931-949.
time delay in autocorrelation function Heskestad, G., 1965, A generalized Taylor hypothesis with
application for high Reynolds number turbulent shear
Eulerian time microscale
flows. Trans. ASME, J. appl. Mech. 32, 735.
Hinze, J. O., 1975, Turbulence, 2nd Edition. McGraw-Hill,
Subscripts New York.
Kim, W. J. and Manning, F. S., 1964, Turbulence energy and
1, 2, 3 three principal directions intensity spectra in a baffled, stirred vessel. A.I.Ch.E. J. 10,
r, 0, 2 radial, tangential, axial 747-752.
i any principal direction Komasawa, I., Kuhoi, R. and Otake, T., 1974, Fluid and
periodic particle motion in turbulent dispersion. Chem. Engng Sci.
per
29, 64-650.
rand random
Laufer, J., 1954, NACA Report 1174.
tot total Laufhutte, H. D. and Mersmann, A. B., 1985, Dissipation of
power in stirred vessels. In 5th Europ. Cant on Mixing, pp.
331-340, Wurzburg, F.R.G.
Superscripts
Lumley, J. L., 1965, Interpretation of time spectra measured
root-mean-square value in high-intensity shear flows. Phys. Fluids 8, 1056-1062.
_ average quality Manning, F. S. and Wilhelm, R. H., 1963, Concentration
fluctuations in a stirred baffled vessel. A.1.Ch.E. J. 9, 12-19.
Metzner, A. B. and Taylor, J. S., t960, Flow patterns in
agitated vessels. A.1.Ch.E. J. 6, 109-l 14.
Mujumdar, A. S., Huang, B., Wolf, D., Weber, M. E. and
REFERENCES
Douglas, W. J. M., 1970, Turbulence parameters in a
Angst, W., Bourne, J. R. and Sharma, R.N., 1982, Mixing and stirred tank. Can. J. them. Engng 48, 475483.
fast chemical reaction-IV. The dimensions of the reaction Nagata, S., 1975, Mixing: Principles and Applicatuzms.
zone. Chem. Engng Sci. 37, 585-590. Halsted Press, New York.
Antonia, R. A., Satyaprakash, B. R. and Hussain, A. K. M. F., Okamoto, Y., Nishikawa, M. and Hashimoto, K., 1981,
1980, Measurements of dissipation rate and some other Energy dissipation rate distribution in mixing vessels and
characteristics of turbulent plane and circular jets. Phys. its effects on liquid-liquid dispersion and solid-liquid mass
Fluids 23, 695-700. transfer. Znt. them. Engng 21, 88-94.
Barthole, J. P., Maisonneuve, J., Gence, J. N., David, R., Ottino, J. M., Ranz, W. E. and Macosko, C. W., 1979, A
Mathieu, J. and Villermaux, J., 1983, Measurement ofmass lamellar model for analysis of liquid-liquid mixing. Chem.
transfer rates, velocity and concentration fluctuations in Engng Sci. 34, 877-890.
an industrial stirred tank. Chem. Engng Fundam. 1, 17-26. Patferson, G. K., 1983, Turbulent mixing and its measure-
Batchelor, G. K., 1983, The Theory of Homogeneous Turbu- ments. In Handbook ofFluids in Motion (Edited by N. P.
lence. Cambridge University Press, Cambridge. Cheremisinoffand R. Gupta), Chap. 5. Ann Arbor Science,
Rates, R L., Fondy, P. L. and Corpstein, R R., 1963, An Ann Arbor, MI.
examination of some geometric parameters of impeller Rao, M. A. and Brodkey, R. S., 1972, Continuous flow stirred
power. Ind. Engng Chem. Process Des. Dev. 2, 31&314. tank turbulence parameters in the impeller stream. Chem.
Bertrand, J., Couderc, J. P. and Angelino, H., 1980, Power Engng Sci. 27, 137-156.
consumption, pumping capacity and turbulence intensity Reith, Ir. T., 1965, Generation and decay of concentration
in baffled stlrred tanks: comparison between several tur- fluctuations in a stirred baffled vessel. A.I.Ch.E.-I.Ch.E.
bines. Chem. Engng Sci. 35, 2157-2163. Symp. Ser. No. 10, p. 14.
Brodkey, R. S. (Ed.), 1975, Turbulence in Mixing Operations. Sachs, J. P. and Rushton, J. H., 1954, Discharge flow from
Academic Press, New York. turbine-type mixing impellers. Chem. Engng Prog. 50,
Calderbank, P. H., 1958, Physical rate process in industrial 597-603.
fermentation, Part I: the interfacial area in gas-liquid &to, Y., Ishii, K., Horie, Y., Kamiwano, M. and Yamamoto,
contacting with mechanical agitation. Trcrns. Znsln them. K., 1967, Turbulent flow in a stirred vessel. Xagaku
Engrs 36, 443463. Kogaku 31, 271-281.
Cooper, R. G. and Wolf, D., 1968, Velocity profiles and Sato, Y., Kamiwano, M. and Yamamoto, K., 1970, Turbulent
pumping capacities for turbine type impellers. Can. J. flow in a stirred vessel--effects of impeller types. Kagaku
them. Engng 46, 9&100. Kogaku 34, lw11 I.
Corrsin, S., 1964, The isotropic turbulent mixer: Part II. Schwartzberg, H. G. and Treyhal, R. E., 1968, Fluid and
Arbitrary Schmidt number. A.1.Ch.E. J. 10, 870-877. particle motion in turbulent stirred tanks. Znd Engng
Cutter, L. A., 1966, Flow and turbulence in a stirred tank. Chem. Fundam. 7, l-11.
A.I.Ch.E. J. 12, 35-45. Tennekes, El., 1977, Turbulence: diffusion, statistics, special
Durrani, T. S. and Greated, C. A., 1977, Laser Systems in dynamics, In Handbook ofTurhu[ence (Edited by W. Frost
Flow Measurements. Plenum Press, New York. and T. M. Moulden), Vol. 1, Chap. 5. Plenum Press, New
Durst, F., Melling, A. and Whitelaw, J. H., 1976, Principles York.
and Practice of LDA. Academic Press, New York. Tennekes, H. and Lumley, J. L., 1972, A First Course in
Fort, I., Placek, J., Kratky, J., Durdil, J. and Drbohlav, J., Turbulence. MIT Press, Cambridge, MA.
1974, Tukbulent characteristics of the velocity field in a Townsend, A. A., 1976, The Structure of Turbulent Shear
system with turbine impeller and radial baWes. Cohn Flow, 2nd Edition. Cambridge University Press,
Czech. them. Commtm. 39, 181&1822. Cambridge.
George, W. K., Jr, 1975, Limitations to measuring accuracy Ubl, V. M. and Gray, J. B. (Eds), 1966, Mixing-Theory and
inherent in the laser Doppler signal. In Proc. ofrhe LDA- Practice, Vols I and 2. Academic Press, New York.
Symp., Copenhagen. Ulbrecht, J. J. and Patterson, G. K. (Eds), 1985. Mixing of
Goldschmidt, V. W., Young, M. F. and Ott, E. S., 1981, Liquids by Mechnnical Agitntion. Gordon & Breach, New
Turbulent convective velocities (broad-band and wave- York.
number dependent) in a plane Jet. J. Fluid Mech. 105, Van Der Molen, K. and Van Maanen H. R. E., 1978, Laser-
327-345. Doppler measurements of the turbulent flow in stirred
Laser-Doppler measurements of turbulent-flow parameters 222 1

vessels to establish scaling rules. Chem. Engng Sci. 33, Van? Riet, K., Bruiin, W. and Smith, J. M.. 1976, Real and
1161-1168. pseudo-turbulen& in the discharge stream from a Rushton
Van Doorn, M., 1981, On Taylor’s hypothesis in turbulent turbine. Chem. Engng Sci. 3L407-412.
shear flows. Internal note 811123, University of Van’t Riet, K. and-Smith, J. M., 1973, The behavior of
Missouri-Rolla. gas-liquid mixtures near Rushton turbine blades. Chem.
Van Maanen, H. R. E., Van Der Molen, K. and Blom, J., Engng Sci. 28, 1031-1037.
1975, Reduction of ambiguity n&c in laser-Doppler velo- Van’t Riet, K. and Smith, J. M., 1975, The trailing vortex
cimetry by a crosscorrelation technique. In Proc. of the system produced by Rushton turbine agitators. Chem.
LDA-Symp., Copenhagen. Engng Sci. 30, 1093-l 105.

You might also like