Sie sind auf Seite 1von 10

chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

Suspension of solid particles in vessels agitated


by axial flow impellers

R.K. Grenville a,∗ , A.T.C. Mak b,c , D.A.R. Brown d


a Philadelphia Mixing Solutions Ltd., Palmyra, PA, USA
b Air Products & Chemicals Inc., North Point, Hong Kong
c Department of Chemical and Biomolecular Engineering, Hong Kong University of Science and Technology,

Hong Kong
d Fluid Mixing Processes, BHR Group, Cranfield, UK

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

Article history: In the design of agitators for solid–liquid mixing processes there are two important criteria.
Received 3 March 2015 The first is “just suspended” where all the particles are in motion on the base of the vessel
Received in revised form 29 April but not necessarily uniformly distributed. The second is “uniform distribution” where the
2015 particles are distributed throughout the liquid phase so that, at each axial location within
Accepted 21 May 2015 the slurry, the solids concentration is very close to the calculated average.
Available online 29 May 2015 The correlation proposed by Zwietering (1958. Chem. Eng. Sci. 8, 244) has been widely used
for estimating the “just suspension” speed, NJS , and is recommended for agitator design
Keywords: in the Handbook of Industrial Mixing (Atiemo-Obeng et al., 2004. Solid-liquid mixing. In:
Mixing Paul, Atiemo-Obeng, Kresta (Eds.), The Handbook of Industrial Mixing. John Wiley & Sons,
Agitation Hoboken, NJ, p. 558). However, there is significant evidence that the correlation does not
Stirred tank properly account for the effect of key variables, including liquid viscosity, density difference,
Solids suspension and scale. This is likely related to the correlation’s lack of physical basis and development
Turbulence using simple dimensional analysis, along with the fact that it was based on limited data
Eddies taken over a narrow range of physical properties and scales. Another problem with the
correlation is that in order to use it the value of a geometric specific constant must be
known.
This paper reports data taken at three vessel scales and over a wide range of liquid phase
viscosities using axial and mixed flow impellers in standard baffled vessels with torispher-
ical or dished bases. The model used to correlate the data is based on the work of Davies
(1986. Chem. Eng. Proc. 20, 175). Analysis shows that there are two regimes; one where the
large and medium sized particles interact with turbulent eddies in the inertial sub-range
and a second where smaller particles interact with eddies approaching the viscous sub-
range. The correlation developed for the inertial sub-range mechanism can be rearranged
to show that the appropriate scale-up rule for the just suspension speed is constant power
input per unit mass.
© 2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction of the vessel but not necessarily uniformly distributed. The


agitator operating speed at this condition is defined as the
In the design of agitators for solid–liquid mixing processes “just suspension speed”, NJS . The second is “uniform distri-
there are two important design criteria. The first is “just sus- bution” where the particles are distributed throughout the
pended” where all the particles are in motion on the base liquid phase so that, at each axial location in the slurry, the


Corresponding author. Tel.: +1 717 202 7976; fax: +1 717 832 1740.
E-mail address: rkgrenville@philamixers.com (R.K. Grenville).
http://dx.doi.org/10.1016/j.cherd.2015.05.026
0263-8762/© 2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291 283

solids concentration is close to the calculated average (see for


Nomenclature example Angst and Kraume, 2006).
The just suspended condition is the optimum for mass
Ar Archimedes number (–) transfer between particles and the liquid phase since the par-
a exponent in Eq. (2) (–) ticles are all in motion and in contact with the liquid for the
b exponent in Eq. (2) (–) minimum power input. If any particles are stationary on the
C impeller clearance above vessel base (m) vessel base, the mass transfer rate is limited by the renewal of
c exponent in Eq. (2) (–) interfacial area and hence, diffusion through the settled bed
D impeller diameter (m) of particles. If the agitator operates at a speed above NJS the
d particle size (m) mass transfer coefficient will increase slightly as the velocity
D diffusivity (m2 /s) of the liquid passing the surface of the particle increases (see
e exponent in Eq. (2) (–) for example Nienow, 1985).
Fl impeller Flow number (–) The most widely used method for estimating the just sus-
f exponent in Eq. (3) (–) pended speed was proposed by Zwietering (1958) and it is
g acceleration due to gravity (m/s2 ) recommended for use in agitator design by Atiemo-Obeng
Kx a constant (x = 1, 2, 3. . .) (–) et al. (2004). Many workers in this area have fitted their data
kL mass transfer coefficient (m/s) to the Zwietering correlation with some success but very few
lE turbulent length scale in inertial sub-range (m) have studied a range of vessel scales and varied the liquid
N impeller speed (s−1 ) viscosity as discussed below.
NJS impeller speed when particles are “just sus-
pended” (s−1 ) 2. Literature
n exponent in Eq. (17) (–)
P power (W) Kneule (1956); Jones (2010) measured the particle dissolution
Po impeller Power number (–) rate as a method for identifying the optimum mixing condi-
p constant in Eq. (43) (–) tions for mass transfer. At power inputs below the optimum,
q exponent in Eq. (43) (–) the mass transfer coefficient was found to be approximately
Re impeller Reynolds number (–) proportional to the agitator power input. At speeds above the
Re* Reynolds number in Eq. (14) (=Re D/T) (–) optimum, the mass transfer coefficient increased by only ∼5%
ReE Reynolds number of turbulent eddies (–) for a 10 fold increase in power input. There is very little ben-
Sc Schmidt number (–) efit, in terms of dissolving time, to be gained by operating the
Sh Sherwood number (–) agitator at a speed above this optimum value.
s constant in Zwietering’s correlation, Eq. (1) (–) Zwietering (1958) defined the “just suspended” condition;
T vessel diameter (m) “when no deposits remained stationary on the bottom (of the
u fluctuating velocity of turbulent eddies (m/s) vessel) for more than one second the suspension was consid-
uP  fluctuating velocity of turbulent eddies in slurry ered complete”. The correlation that Zwietering produced for
(m/s) estimation of NJS , is:
X Zwietering’s concentration (mass of parti-
cles/mass of liquid) (%) 0.45
0.10 d0.20 X0.13 (g/L )
x exponent in Eq. (3) (–) NJS = s (1)
D0.85
XV volume fraction (volume of particles/volume of
slurry) (%) It is widely used and recommended for agitator design in
Y Baldi’s turbulence parameter in Eq. (13) (–) the Handbook of Industrial Mixing (Atiemo-Obeng et al., 2004).
y exponent in Eq. (3) (–) The constant, s, must be measured for each impeller and
Z liquid depth m geometry tested and it is assumed to be scale independent.
z constant in Grenville, Mak and Brown correla- In order to develop the correlation, prior to experimenta-
tion, Eq. (33) (–) tion, Zwietering carried out dimensional analysis to develop:
˛ constant in Eq. (31) (–)
˛ constant in Eq. (32) (–)
ˇ exponent in Eq. (31) (–)   1/3  a  b  c
T D d
 exponent in Eq. (31) (–) NJS = K1 Xe
g2 (g/2 )
1/3
(g/2 )
1/3
(g/2 )
1/3
 exponent in Eq. (32) (–) (2)
ı exponent in Eq. (31) (–)
ı exponent in Eq. (32) (–)
ε̄ vessel averaged power input per mass (W/kg)
Re-arranging Eq. (2):
ε turbulent kinetic energy dissipation rate (W/kg)
K Kolmogorov length scale (m)  T a
 dynamic viscosity (N s/m2 ) NJS = K1 Df dc Xe x gy (3)
D
L liquid density (kg/m3 )
S particle density (kg/m3 ) where f = a + b, x = −(1 + 2f + 2c)/3 and y = (2 + f + c)/3
 density difference (=S − L ) (kg/m3 ) In order to determine the values of x and y, c and f must be
 kinematic viscosity (m2 /s) found experimentally.
constant in Eq. (18) (–) The variables chosen for the dimensional analysis are
incomplete since the densities of the solid, liquid and/or the
density difference are not included.
284 chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291

Five impellers were tested; four radial flow turbines and


one axial flow marine propeller at five vessel scales ranging
from 0.15 to 0.60 m in diameter. Initially NJS was measured in
water using sand as the solid particle to be suspended. Then,
“from over 1000 experiments with sand in water” the expo-
nents c and e were found to be 0.20 and 0.13, respectively. The
exponents a and b “varied appreciably” but “f was fairly con-
stant . . . between −0.78 and −0.94”. Zwietering then adopted
the value of −0.85 for all the impellers tested so, x = 0.1 and
y = 0.45.
Zwietering then carried out experiments in which the vis-
cosity and density difference were changed and he found that:

  −  0.45
S L
NJS ∝ 0.1 (4)
L Fig. 1 – Plot of power input per unit mass of slurry versus
volume fraction of particles.
Fortuitously, the exponent on viscosity was 0.1, agreeing
with the value obtained from experiments in water alone. Also
the exponent on the density term was 0.45 agreeing with the Nienow (1968) measured NJS in a flat-based vessel 0.14 m
exponent on g obtained from the experiments in which the in diameter agitated by Rushton turbines and found that the
density difference had not been changed. This meant that exponents on , d, D and X were similar to those reported
these two terms could be combined into (g/L ). by Zwietering (1958). Although the liquid viscosity and vessel
Zwietering’s correlation predicts that the power input by scale were not varied, he assumed that their effect on NJS , as
the impeller per mass of slurry will decrease on scale-up. If reported by Zwietering, were correct and subsequent analysis
the scale-up is geometrically similar with Po and D/T constant, was carried out accordingly.
from Eq. (1): Raghava Rao et al. (1988) used three vessel scales ranging
from 0.30 to 1.50 m in diameter measuring NJS for Rushton and
NJS,L
 D 0.85 pitched-blade turbines but only used water as the test liquid.
S
= (5) The data were found to fit Zwietering’s correlation with small
NJS,S DL
changes in the exponents on d (0.11 versus 0.20) and X (0.10
Scaling-up with geometrical similarity in the turbulent versus 0.13).
regime: Myers et al. (1994) measured NJS in a flat-based vessel
0.29 m in diameter agitated by pitched blade turbines and
ε̄L N3 D2 hydrofoil impellers. Liquid density was varied by adding
= L3 L2 (6) sodium chloride and this increased the kinematic viscosity
ε̄S NS DS
by about 40%. The data were fitted to Zwietering’s correlation
but it was stated that “there are indications that the depend-
Substituting Eq. (5) into (6):
ence of the just-suspended speed on particle size may have

ε̄JS,L
 D 3×0.85  D 2  D 0.55 two regimes”.
S L S
= = (7) Armenante et al. (1998) measured NJS for Rusthon and flat
ε̄JS,S DL DS DL
blade turbines in vessels of 0.19, 0.24 and 0.29 m in diameter.
The particles studied were glass beads and resin particles at
Even though re-arranging the correlation predicts that the
0.5% (wt/wt) concentration and the liquid was distilled water.
power input by the impeller per unit mass of slurry decreases
Since the solids concentration and liquid viscosity were not
on scale-up, which is very attractive from an industrial point of
varied it was assumed, for data analysis, that their effect on
view, Zwietering recommended scaling-up using a criterion of
NJS as reported by Zwietering (1958) was correct.
constant power input per mass because it is “on the safe side”.
Uby (2006) has shown that there are three regimes in
The use of Zwietering’s correlation to estimate NJS , using
which the mechanisms by which the particles are suspended
a value of s measured at small scale, implies that the power
and distributed change depending on the particle size and
input per unit mass at the plant scale will be lower than mea-
fluid properties and their effect on the tank flow and sett-
sured at the lab scale (Atiemo-Obeng et al., 2004). The only
ling regimes. The three regimes are laminar tank flow with
way to apply Zwietering’s recommendation of scaling up using
laminar settling, turbulent flow with laminar settling and tur-
constant power per mass would be to carry out experiments,
bulent flow and settling. In reality the tank flow regimes are
measure the actual power input per mass and use this result
not laminar (Re < 50) but transitional and turbulent.
as a basis from which to scale-up.
Baldi et al. (1978) proposed a model that could account for
Fig. 1 shows a plot of power input per unit mass of slurry for
the values of the exponents that had been found empirically by
a pitched blade turbine (D = T/2 and C = T/4) operating in vessels
Zwietering. They wrote an energy balance where the turbulent
that were 0.305 and 0.610 in diameter versus volume fraction
kinetic energy (TKE) imparted to a particle is proportional to
of particles (Mak, 1992). It is clear that power per unit mass
its potential energy gain as it is lifted from the vessel base:
does not decrease on scale-up. If the data are forced to fit Zwi-
etering’s correlation, s for the 0.30 m vessel is 5.08 ± 5.48% and
L (u ) ∝ gd
2
for the 0.61 m vessel, 5.44 ± 1.26%. The probability that these (8)
two s-values are from the same population is 1.11 × 10−3 so
there is some risk associated with using s-values and Zwieter- If the particles are larger than the Kolmogorov length scale
ing’s correlation for design of agitators for solids suspension. they will be picked up by eddies in the inertial sub-range and
chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291 285

the turbulent fluctuating velocity can be related to the energy 1


 
1/2  Z 1/3  T 
n/2
dissipation rate and eddy length scale by: NJS ∝ g d1/6 (1 + XV )(1 − XV )
Po1/3 D2/3 L T D
(20)
u ∝ (εlE )
1/3
(9)

Particles will be suspended from the vessel base by eddies Davies’ model predicts that NJS is independent of the liq-
of a similar size so setting lE = d and substituting Eq. (9) into uid’s viscosity and scale-up should be carried out using a rule
Eq. (8): of constant power input per unit mass.
Poirier et al. (1999) measured NJS for Zeolite particles in
2/3 vessels of 1.5, 6 and 18.75 feet in diameter using submerged,
L (εd) ∝ gd (10)
shrouded axial impellers and also concluded that the scaling
  1/2 was best described by the constant power per volume method.
ε1/3 ∝ g d1/6 (11) Ibrahim and Nienow (1998) measured NJS for a variety
L
of impellers including Rushton and pitched blade turbines,
Relating the power input per mass to the Power number, hydrofoils and the Ekato Inter-MIG in a vessel that was 0.29 m
speed and diameter of an impeller for a vessel where Z = T: in diameter. They varied the impeller clearance above the
vessel base from T/6 to T/8 and four fluid viscosities were
PoN3 D5 studied; 0.001, 0.01, 0.10 and 1.00 Pa s. The data were forced
ε̄ ∝ (12) to fit Zwieterings’s correlation and the resulting s-values
T3
compared with those measured in water. They found that
Substituting Eq. (12) into (11): the discrepancy between the s-value measured in water and
those measured in the viscous fluids increased with increas-
T
  1/2
ing viscosity. No attempt to explain this discrepancy was
NJS = g d1/6 (13)
YPo1/3 D5/3 L made.
Ayranci et al. (2012) measured the just suspension speed
If the local turbulent kinetic energy dissipation rate for hydrofoil impellers of T/3 and T/2 diameter. They also mea-
(TKEDR) at the base of the vessel is proportional to the vessel sured the mean and fluctuating velocities near the vessel base
average TKEDR, then Y will be constant. If the value is lower using Particle-Image Velocimetry and predicted their values
at the base, due to decay of turbulence, then Y will not be a using a Large–Eddy Simulation Computational Fluid Dynam-
constant. Baldi et al. (1978) found that: ics model. They concluded that the mechanism by which the
particles are picked up changes from turbulent eddy domi-
Y = Re∗0.2 X−0.15 (14) nated for the T/3 hydrofoil to a combination of the mean flow
and turbulence for larger one.
where Ayranci and Kresta (2014) have reviewed the limitations of
Zwietering’s correlation for the design of industrial scale mix-
D
Re∗ = Re (15) ing equipment and proposed a modification, using Baldi et al.
T
(1978)’s model, which is more accurate in predicting NJS . This
Substituting Eq. (15) into (13): modification allows s-values, that have been determined by
forcing data to fit Zwietering’s correlation, to be used for esti-
0.42 mation of the agitator’s operating speed irrespective of scale,
0.17 d0.14 X0.125 (g/L )
NJS ∝ (16) provided that the scale at which the experiments were per-
D0.89
formed is known.
The form of this correlation is very similar to Zwietering’s, Many of the papers found in the literature report results of
and the constant of proportionality is a function of impeller suspension experiments carried out using Rushton turbines
type, impeller diameter and clearance above the vessel base yet these impellers are rarely used in industry for solids sus-
in the same manner as Zwietering’s s-value. pension applications (Barnak, 2013; Janz, 2013; Kehn, 2013;
Davies (1986) wrote a force balance equating the force act- Simpson, 2013).
ing on a particle to the sedimentation force and included a Thus, the goals of the study reported here are:
term to account for the volume fraction of the particles:

L (u ) d2 ∝ gd3 (1 − XV )
2 n
(17) • To study and correctly account for the effect of scale on NJS .
• To study and correctly account for the effect of liquid vis-
Also he assumed that the presence of the particles reduces cosity on NJS .
the turbulent fluctuating velocity such that: • To study impellers commonly used in industry to suspend
particles in a liquid phase.
u
uP = (18)
1 + XV
3. Model and dimensionless numbers
Re-writing Eq. (9) using this corrected fluctuating velocity
and the more general definition of the average power input Re-arranging Eq. (11):
per mass

  1/2
Po1/3 ND5/3
ε̄1/3 ∝ (19) ε1/3 ∝ Po1/3 NJS D2/3 ∝ g d1/6 (21)
T 2/3 Z1/3 L
286 chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291

Multiplying both sides of Eq. (21) by d4/3 / and the left-hand


side by (D/D)4/3 :

NJS D 2 d
 4/3   1/2 d3/2
Po1/3 ∝ g (22)
 D L 

The term on the left-hand side of Eq. (22) is the Reynolds


number of the turbulent eddies at the scale of the particles
when the impeller operates at NJS . The term on the right-hand
side is the square root of Archimedes number. Eq. (22) can be
re-written:

ReE = K2 Ar1/2 (23)

If the particles are approaching the Kolmogorov length


scale in size, they will be picked up by eddies in the viscous
sub-range and the turbulent fluctuating velocity can be related
to the energy dissipation rate and Kolmogorov length scale by:

 ε 1/2
u = K (24)

Fig. 2 – Vessel and impeller geometry.
Setting K = d and substituting into Eq. (24):
The solid particles were sand and urea formaldehyde.
ε The range of physical properties investigated was:
L d2 ∝ gd (25)

d: 100, 180, 200, 325, 360 and 605 ␮m
g
ε∝ (26) S : 1500 and 2650 kg/m3
L d
L : 1000 kg/m3
Substituting Eq. (12) into (26): : 0.52–2000 cP
XV : 0.04–20.22%
 g 1/3 Re: 3.9 × 102 –6.5 × 105
Po1/3 NJS D2/3 ∝ (27)
L d
The viscosity of the liquid phase was adjusted using solu-
Or tions of low molecular weight carboxymethyl cellulose which

1
 g 1/3 are Newtonian at the concentrations used.
NJS ∝ (28)
Po1/3 D2/3 L d
5. Results
Multiplying both sides of Eq. (27) by d4/3 / and the left-hand
side by (D/D)4/3 :
5.1. Regimes

NJS D2
 d 4/3  g 1/3 d Fig. 3 shows a plot of Eddy Reynolds number versus
Po1/3 ∝ (29) Archimedes number for pitched blade turbines and hydrofoil
 D L 2/3
impellers of T/3 and T/2 diameter at a clearance of T/4. The
Or plot shows that there are indeed two regimes; one in which
Eddy Reynolds number is proportional to the square root
ReE = K3 Ar1/3 (30) of Archimedes number and a second, where Eddy Reynolds

4. Experimental

NJS was measured visually using Zwietering’s criterion: “when


no deposits remained stationary on the bottom (of the ves-
sel) for more than one second the suspension was considered
complete”.
Experiments were carried out in geometrically identical
dish based vessels of 0.30, 0.61 and 1.00 m diameter with liq-
uid depth equal to the vessel diameter. The vessels had four
vertical baffles that extended the length of the straight side
and these were T/12 wide and mounted T/60 from the vessel
wall.
Pitched blade turbine and hydrofoil impellers of T/2 and T/3
diameter were located at three clearances, C = T/3, T/4 and T/6
above the vessel base. The vessel and impeller configuration Fig. 3 – Plot of Eddy Reynolds number versus Archimedes
is shown in Fig. 2. number (XV = 1.9%, d = 180 ␮m).
chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291 287

Table 1 – Regression results for pitched blade turbine.


Coefficient Upper 95% Lower 95%
Conf. limit Conf. limit

log ˛ 0.181 0.200 0.162


ˇ 0.511 0.517 0.506
 0.131 0.154 0.108
ı 0.103 0.167 0.038
log ˛ 0.181 0.202 0.161
 0.146 0.170 0.122
ı 0.085 0.154 0.017

number is proportional to the cube root of Archimedes num-


ber. The boundary between the two regimes occurs at an eddy
Reynolds number of approximately 1 × 10−1 .

5.2. Correlation for predicting NJS Fig. 4 – Plot of correlation for pitched blade and hydrofoil
impellers.
The data were divided into two groups; one with Ar > 2 × 10−2
where the particles interact with inertial sub-range eddies and correlate both the pitched blade and hydrofoil performance. In
the other with Ar < 2 × 10−2 as shown in Fig. 3. fact, the probability that the two data sets are from the same
For the data sets taken with Ar > 2 × 10−2 , a regression was population is 2.58 × 10−28 and the individual constants should
carried out for each impeller type of the form: be used for calculation of NJS and agitator design.
Eq. (33) can be re-arranged for NJS :
 C ı
ReE = ˛Arˇ XV

(31) z
 g 0.500  C 0.100
D NJS = d0.167 XV
0.154
(34)
Po0.333 D0.667 L D
The results for the pitched blade turbine are shown
in Table 1 and for the hydrofoil, Table 2. The correlation Only the constant, z, and the exponents on XV and C/D
coefficients for both impellers were over 99% and the P-values have been determined from the regression. The other four are
for all the coefficients were less than 0.05. predicted by Davies’ (1986) model.
Since the turbulence based model predicts that the expo- Also, for a given impeller type, geometry and system:
nent, ˇ, should have a value of 0.50, the exponent was
rounded-off and the regression performed on: Po0.333 NJS = K4 (35)

ReE
 C ı Brown (2010) has shown that the Flow number is propor-

= ˛ XV (32) tional to the cube root of the Power number. For axial and
Ar1/2 D
mixed flow impellers operating in the turbulent regime:
The results of the second regression are also shown in
Tables 1 and 2. Fl = 0.75Po0.333 (36)
Finally the effect on the data fit of taking the average values
of   and ı and applying them to both the pitched blade turbine So Eq. (35) can be re-written as:
and hydrofoil impellers was tested. The constant, ˛ , changed
in the third decimal place so the correlation can be written: FlNJS = K5 (37)

 C 0.100
Wu et al. (2001) measured velocities using laser-Doppler
ReE = zAr0.500 XV
0.154
(33)
D anemometry and NJS , for glass Ballotini beads, using a variety
of impellers in a vessel 0.39 m in diameter. They plotted their
where z = 1.528 (with relative standard deviation of ±11.0%) for
data according to Eq. (37) but also concluded that:
the pitched blade turbines and z = 1.213 (with relative standard
deviation of ±14.2%) for the hydrofoils.
Fls = K6 (38)
The correlation is plotted in Fig. 4.
Visually it appears that the two data sets are the same and
It has been shown in this paper that the Zwietering s-value
that the average value of the constant, z, could be used to
is scale dependent so Eq. (38) can only be valid at the scale
where the data were collected.
Table 2 – Regression results for hydrofoil impeller.
Coefficient Upper 95% Lower 95% 6. Discussion
Conf. limit Conf. limit
6.1. Impeller efficiency
log ˛ 0.085 0.119 0.052
ˇ 0.510 0.519 0.502
 0.128 0.212 0.043 Eq. (34) can be further re-arranged and multiplying both sides
ı 0.103 0.180 0.026 by D/T:
log ˛ 0.080 0.114 0.046
 0.163 0.244 0.081 Po0.333 NJS D1.667 g
 0.500  C 0.100  D 
ı 0.115 0.193 0.036 =z d0.167 XV
0.154
(39)
T L D T
288 chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291

Or eddy Reynolds number (see for example Calderbank, 1967;


 g 0.500  C 0.100  D  Levins and Glastonbury, 1972; Armenante and Kirwan, 1989),
ε̄0.333 =z d0.167 XV
0.154
(40) as defined in Eq. (22):
JS
L D T
 4/3  1/3
For a given system (particle size, concentration, densities NJS D 2 d εJS d 4
ReE = Po1/3 = (42)
etc.) Eq. (40) shows that:  D 3

 D 3
ε̄JS = K7 z3 (41) The Frössling equation relates the Sherwood number to
T Reynolds and Schmidt numbers.
The power input by a T/2 impeller operating at NJS will be
3.38 times higher than the power input by a T/3 impeller of kL d q
Sh = = 2 + pReE Sc1/3 (43)
the same type. Similarly, a pitched blade turbine will require D
2.00 times the power of a hydrofoil of the same diameter when
For particles that are much larger than the Kolmogorov
operating at NJS .
length scale, the exponent q is 3/4. So combining Eqs. (42) and
Ayranci et al. (2012) also concluded that the mechanism by
(43), setting q = 3/4 and rearranging gives:
which particles are suspended by a T/3 hydrofoil is dominated
by turbulence with convection playing a secondary role. For
D
+ p(εJS ) Sc−2/3
1/4
a T/2 hydrofoil both turbulence and convection contribute to kL = 2 (44)
d
the suspension mechanism. Consequently the level of turbu-
lence required by the T/3 impeller was approximately 33% of
So, even though the power input by the T/2 pitched blade
the level required by the T/2 turbine when operating at NJS .
turbine is approximately seven times higher than the T/3
Fig. 14 of their paper shows that PJS for the T/2 turbine is
hydrofoil when both are operating at NJS , the mass transfer
approximately double PJS for the T/3 turbine.
coefficient achieved will only be 63% higher.
Ayranci et al. (2012) studied suspension in a flat based ves-
sel while the data reported here were measured in a dish
based vessel. The region where the last particles are sus- 6.3. Comparison with Zwietering’s correlation
pended changes from the corner, where the wall meets the
base, in a flat based vessel to the lowest point in the centre of Fig. 1 shows the power input per mass of slurry versus the
the base, for a dish based vessel. It is likely that this difference volume fraction of sand in water for a pitched blade turbine
may contribute to the difference in the suspension mecha- T/2 in diameter with a clearance of T/4 above the vessel base.
nisms observed and the relative power inputs by the T/3 and The open squares represent data measured in a vessel 0.31 m
T/2 hydrofoils operating at NJS . in diameter and the open diamonds represent data measured
Brown (2010) presented velocity profiles generated, mea- in a 0.61 m vessel. From this plot it is clear that power per mass
sured by laser-Doppler anemometry, by pitched blade turbines does not decrease on scale-up.
of T/4, T/3 and T/2 in diameter in a dish based vessel. The pro- Zwietering’s correlation implies that the power input per
file for the T/4 turbine is predominantly axial while, for the mass at NJS in the 0.61 m vessel should be ∼70% of the power
T/2 turbine, the profile has both axial and radial components. input per mass in the 0.31 m vessel and the filled diamonds
In a dish based vessel the last particles are suspended from show the power input per mass predicted by the correlation
the lowest point in the dish and the smallest turbine directs for a vessel 0.61 m in diameter. The filled triangles show the
its flow to this location. For the T/2 turbine the particles are power per mass predicted by Zwietering’s correlation using the
indirectly picked up in the flow that travels down the dish and s value measured at 0.31 m scale to calculate NJS for a 3.10 m
returns to the suction of the blades. This is less efficient than vessel.
the direct pick up achieved by the smaller turbine. Also, there is no theoretical reason to support the value
A T/2 pitched blade turbine will require approximately of 0.1 for the exponent on kinematic viscosity in the corre-
seven times the power of a T/3 hydrofoil when operating at lation. If the particles are larger than the Kolmogorov Length
NJS in a dish based vessel. The effect on the mass transfer Scale and interacting with turbulent eddies in the inertial sub-
coefficient between the particles and the liquid must also be range, the exponent on viscosity should be zero. Zwietering’s
considered. correlation predicts that NJS in a fluid of 10 cP would be approx-
imately 26% higher (i.e. 100.1 ) than required which means that
6.2. Mass transfer coefficient the power drawn by the impeller would be twice as high.

Operation at NJS is the optimum for mass transfer between the 7. Conclusions
particles and the liquid phase. At this condition the resistance
to mass transfer is due to diffusion through the film surround- A correlation for estimation of the just suspension speed, NJS ,
ing the particles. When the agitator operates at a lower speed for agitated slurries has been developed from a physical model
and some of the particles are stationary on the vessel base the for particle—fluid interactions proposed by Davies (1986). The
limiting mass transfer rate is controlled by diffusion through model can be rearranged into two dimensionless numbers,
the bed of settled particles (Nienow, 1985). the eddy Reynolds and Archimedes numbers. In this form
Once all the particles are in motion and in contact with the model predicts that for large and medium sized particles,
the liquid phase, the mass transfer coefficient, kL , can be esti- interacting with turbulent eddies in the inertial sub-range, the
mated from the Frössling equation (see for example Davies, exponent on Archimedes number should be 1/2. Regression
1972). The Reynolds number commonly used to account for analysis of the data confirms that the predicted exponent is
the interactions between the fluid and the particles is the correct with correlation coefficients greater than 99%.
chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291 289

The new correlation is applicable to downward pumping University of Birmingham, UK, Dr. Arthur W. Etchells III of
axial flow pitched blade turbines and hydrofoils with geomet- Rowan University, Glassboro NJ, USA and Dr. Lars Uby of Xylem
ric ratios limited to: Inc., Stockholm, Sweden for invaluable discussions. Also, we
are particularly grateful to Dr. Steven P. Bailey of DuPont Engi-
0.33 ≤ D/T ≤ 0.50. neering for the guidance he provided on the statistical analysis
0.17 ≤ C/T ≤ 0.33. that we carried out.
0.30 ≤ C/D ≤ 1.00.
Appendix A. Comparison of methods for
Data were measured in three dished base vessels of 0.30, calculating NJS
0.61 and 1.00 m diameter with four standard baffles installed.
In all cases the liquid depth was equal to the vessel diameter. Atiemo-Obeng et al. (2004) in the Handbook of Industrial Mixing
One of the main conclusions of this work is that, in the iner- give the following example (Example 10-2). Aluminium chlo-
tial sub-range, NJS is independent of the fluid’s viscosity. This ride particles are to be suspended in methylene chloride.
is at odds with much of the work published since Zwietering The concentration expressed as Zwietering’s X, is 0.4% so a
(1958)’s paper appeared. It is clear that the reason for this is slurry containing 100 g of liquid will contain 0.4 g of particles.
that most of these studies used only water as the test fluid The mass concentration of the slurry is 0.398%.
and assumed that the exponent of 0.1 on kinematic viscosity NJS is calculated for a pitched blade turbine with D = T/3 and
in Zwietering’s correlation was correct. C = T/8. Zwietering’s s-value is given as 3.7 for this impeller.
Zwietering’s correlation predicts that, on scale-up with Since the correlation presented here cannot be applied to an
geometrical similarity, the power input by the impeller per impeller located this close to the vessel base, the clearance
mass of slurry will decrease by the ratio of the vessel diame- has been increased to T/4 and the s-value for this geometry is
ters raised to an exponent of 0.55. Davies (1986) model predicts 4.4 (Atiemo-Obeng et al., 2004, Table 10-3).
that the scale-up rule of maintaining just suspension condi- Other data required are:
tion should be constant power input per unit mass and the
data reported here support this conclusion. Particle size 5 mm or 5 × 10−3 m
There are two outstanding questions regarding this work
Particle density, S 2440 kg/m3
that need to be addressed. The first is the effect of the shape of
the vessel base on the pick-up mechanism. The data reported Liquid density, L 1326 kg/m3
here were taken in vessels with dished or torispherical bases Kinematic viscosity,  7.541 × 10−7 m2 /s
and the correlation developed from Davies’ (1986) turbulence
model accounts for the effect of impeller to vessel diameter Atiemo-Obeng et al. (2004) give a vessel diameter of 48 in.
ratio. Ayranci et al. (2012) concluded that, in a flat-based ves- but use an impeller diameter of 28.5 in. in their calculation.
sel, turbulence is the dominant pick-up mechanism for the For an impeller diameter of T/3, the vessel diameter would be
smaller T/3 impeller while turbulence and flow contribute for 85.5 in. and this value is used in the following calculations:
the larger T/2 impeller. Even though the mechanisms might From Eq. (1):
be different in both vessels, the smaller impellers achieve NJS
with the lowest power. 0.10 d0.20 X0.13 (g/)
0.45
NJS = s (1)
D0.85
0.10 0.20
(7.541 × 10−7 ) × (5 × 10−3 )
0.45
× 0.40.13 × (9.81 × (2440 − 1326)/1326)
NJS = 4.4 × 0.85
0.724
NJS = 1.12 RPS or67.4 RPM

This is 18% higher than the original solution and it is higher


The second outstanding question is the effect of the by the ratio of the turbine’s s-values at the two clearances.
impeller geometry and the direction of its primary discharge. 68 RPM is a standard AGMA gearbox output speed so this
The correlation presented here was developed for axial and operating speed would be chosen.
mixed flow pitched blade turbines and hydrofoils, while much Then the power drawn by the turbine would be calculated
of the work reported in the literature has used radial flow from:
Rushton turbines and it is not clear that the new correlation
will be applicable to Rushton turbines. These are rarely used in P = PoSL N3 D5 (40)
industry for processes requiring solids suspension, but it may
be of academic interest to determine if their performance can The Power number for the T/3 diameter turbine at T/4 clear-
be described by the Davies (1986) model. Should this be done ance is 1.73 (Mak, 1992) and, since the slurry is so dilute it is
it is likely that given the change in flow patterns, the effects safe to assume that its density is equal to the liquid phase
of the geometric variables C/D and D/T would be different. alone.
 68 3
Acknowledgments P = 1.73 × 1326 × × 0.7245
60
We gratefully acknowledge the members and staff of the P = 665 W or 0.89 HP
Fluid Mixing Processes consortium for their permission to
publish the data used in this paper and Dr. Pip N. Jones of The volume of slurry in the vessel will be 8.044 m3 and the
Axalta Coating Systems, Wuppertal, Germany for translating mass, 10666.77 kg. The power input by the impeller per mass
Kneule’s 1956 paper. We thank Prof. Alvin W. Nienow of the of slurry will be 0.062 W/kg.
290 chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291

The agitator specified will have a 1 HP motor and operate obtained. This strongly supports the argument, made above,
at 68 RPM. that Zwietering’s correlation for predicting the “just suspen-
Taking the same conditions and using Eq. (34): sion” speed in an agitated vessel, using s-values measured at
small scale, cannot be used reliably for design of full scale
z
 g 0.500  C 0.100 plant equipment.
NJS = d0.167 XV
0.154
(34)
Po0.333 D0.667 L D

References
The clearance above the vessel base, C, is 21.37 in. and the
z value for a Pitched blade turbine is 1.528.
Angst, R., Kraume, M., 2006. Experimental investigations of
100 kg of slurry will contain 0.398 kg of particles stirred solid/liquid systems in three different scales: particle
and 99.602 kg of liquid. The volume of particles will distribution and power consumption. Chem. Eng. Sci. 61,
be 1.631 × 10−4 m3 and the volume of liquid will be 2864–2870.
7.511 × 10−2 m3 . The volume fraction of solids is 0.217%. Armenante, P.M., Kirwan, D.J., 1989. Mass transfer to
microparticles in agitated systems. Chem. Eng. Sci. 44,
2781–2796.
 0.500 Armenante, P.M., Uehara Nagamine, E., Sustano, J., 1998.
1.528 9.81 × (2440 − 1326) Determination of correlations to predict the minimum
NJS = ×
1.730.333 × 0.7240.667 1326 agitation speed for complete solid suspension in agitated
 21.37 0.100 vessels. Can. J. Chem. Eng. 76, 413–419.
0.167
× (5 × 10−3 ) × 0.2170.154 × Atiemo-Obeng, V.A., Penney, W.R., Armenante, P.M., 2004.
28.5 Solid–liquid mixing. In: Paul, E.L., Atiemo-Obeng, V.A., Kresta,
S.M. (Eds.), The Handbook of Industrial Mixing. John Wiley &
NJS = 1.44 RPS or 86.4 RPM
Sons, Hoboken, NJ, pp. 558–562.
Ayranci, I., Machado, M.B., Madej, A.M., Derksen, J.J., Nobes, D.S.,
84 RPM is a standard AGMA gearbox output speed so this Kresta, S.M., 2012. Effect of geometry on the mechanisms for
operating speed would be chosen and Eq. (32) would be solved off-bottom solids suspension in a stirred tank. Chem. Eng. Sci.
for a new impeller diameter fixing NJS as 84 RPM (or 1.40 RPS). 79, 163–176.
Ayranci, I., Kresta, S.M., 2014. Critical analysis of Zwietering
In this case it is 29.5 in. or 0.750 m.
correlation for solids suspension in stirred tanks. ChERD 92,
Eq. (40) would then be used to calculate the power drawn
413–422.
by the impeller: Baldi, G., Conti, R., Alaria, E., 1978. Complete suspension of
particles in mechanically agitated vessels. Chem. Eng. Sci. 33,
 84 3 21–25.
P = 1.73 × 1326 × × 0.7505 Barnak, S., 2013. Private Communication.
60
Brown, D.A.R., 2010. Mixer performance characteristics: impeller
P = 1500 W or 2.01 HP and process efficiency. In: MIXING XXII, Victoria, BC, Canada.
Calderbank, P.H., 1967. Mass transfer. In: Uhl, V.W., Gray, J.B.
(Eds.), Mixing: Theory and Practice, 2. Academic Press, New
In this case, with a slurry mass of 10666.77 kg, the power York, NY, pp. 76–81.
input by the impeller per mass of slurry will be 0.141 W/kg. Davies, J.T., 1972. Eddy transfer near solid surfaces. In: Turbulence
The agitator specified will have a 3 HP motor and operate Phenomena. Academic Press, New York, NY, pp. 143–151.
at 84 RPM. Davies, J.T., 1986. Particle suspension and mass transfer rates in
agitated vessels. Chem. Eng. Proc. 20, 175–181.
The Zwietering s-values reported in Table 10-5 of Atiemo-
Ibrahim, S., Nienow, A.W., 1998. Comparing impeller performance
Obeng et al. (2004) were measured by Mak (1992) in a vessel
for solid–liquid mixing in the transitional flow regime with
which was 0.61 m (24 in.) in diameter. Newtonian fluids. In: Proc. Int. Conf. on Mixing and
From Eq. (5), NJS in the 24 in. vessel can be calculated: Crystallization, Tioman Island, Malaysia.
Janz, E.E., 2013. Private Communication.

NJS,L
 D 0.85 Jones, P.N., 2010. Private Communication.
S
= (5) Kehn, R.O., 2013. Private Communication.
NJS,S DL Kneule, F., 1956. Die prüfung von rührern durch
löslichkeitsbestimmung. Chem. Ing. Tech. 28, 221–225.
Or Levins, D.M., Glastonbury, J.R., 1972. Application of Kolmogoroff’s
theory to particle-liquid mass transfer in agitated vessels.
 D 0.85 Chem. Eng. Sci. 27, 537–543.
L
NJS,S = NJS,L Mak, A.T.C., 1992. Solid–liquid mixing in a mechanically agitated
DS vessel. University College London (Ph.D. Thesis).
 85.5 0.85 Myers, K.J., Fasano, J.B., Corpstein, R.R., 1994. The influence of
NJS,24 = 67.4 × = 198.45 RPM solids properties on the just-suspended agitation
24 requirements of pitched-blade and high-efficiency impellers.
Can. J. Chem. Eng. 72, 745–748.
The power input, calculated from Eq. (40) from this speed, Nienow, A.W., 1968. Suspension of solid particles in turbine
is 28.85 W agitated vessels. Chem. Eng. Sci. 23, 1453–1459.
The volume of slurry in the vessel will be 0.165 m3 and the Nienow, A.W., 1985. The mixer as a reactor: liquid/solid systems.
In: Narnby, N., Edwards, M.F., Nienow, A.W. (Eds.), Mixing in
mass, 226.76 kg. The power input by the impeller per mass of
the Process Industries. Butterworth & Co. (Publishers) Ltd.,
slurry will be 0.132 W/kg.
London, UK, pp. 362–363.
The power input per mass calculated from the Grenville, Poirier, M.R., Powell, M.R., Gladki, H., Rodwell, P.O., 1999.
Mak and Brown correlation, Eq. (32), for the plant scale vessel Suspending zeolite particles in tanks. In: Seventh Int. Symp.
is 0.141 W/kg which is within 7% of the value calculated for the on Liquid–Solid Flows, Joint ASME-JSME FED Summer Meeting,
24 in. diameter vessel from which the Zwietering s-value was http://sti.srs.gov/fulltext/ms9800893/ms9800893.pdf.
chemical engineering research and design 1 0 0 ( 2 0 1 5 ) 282–291 291

Raghava Rao, K.S.M.S., Rewatkar, V.B., Joshi, J.B., 1988. Critical Wu, J., Zhu, Y., Pullum, L., 2001. Impeller geometry effect on
impeller speed for solid suspension in mechanically agitated velocity and solids suspension. In: Proc. ISMIP 4, Toulouse,
contactors. AIChE J. 34, 1332–1340. France.
Simpson, T.A., 2013. Private Communication. Zwietering, T.N., 1958. Suspending of solid particles in liquid by
Uby, L., 2006. New regimes and correlations for just suspended agitators. Chem. Eng. Sci. 8, 244–253.
speed. In: Proc. 12th Euro. Conf. on Mixing, Bologna, Italy,
pp. 249–255.

Das könnte Ihnen auch gefallen