44 www.cepmagazine.

org May 2003 CEP

Solids/Liquids Handling
IQUID AGITATION IS A COMMON UNIT
operation in the chemical engineering and bio-
logical process industries, and its practical use
is widely published (1, 2). However, most of the litera-
ture on liquid agitation addresses Newtonian uids,
while only limited design information is available on
the agitation of non-Newtonian power-law uids.
These uids are either shear-thinning or shear-thicken-
ing i.e., their apparent viscosity,
a
, decreases
(shear-thinning) or increases (shear-thickening) with
the shear rate, .
This article describes the relationships that must be
understood in order to scale-up uid mixing processes
in non-Newtonian power-law uids. The principles de-
scribed apply to both close-clearance impellers, such as
anchors and helixes, and open-turbine impellers.
Non-Newtonian uids
It would be ideal to develop a design procedure for
general mixer design that would work reliably for all
non-Newtonian uids. However, the prospect of such
logic is highly unlikely, due to the complexity and vari-
ability of the materials encountered in the process in-
dustries. Any design procedure should be veried by
actual testing at a scale that is as close to the produc-
tion scale as possible. The nal design of mixers for
non-Newtonian power-law uids should be based on
scaleup of successful operating experience.
Power-law uids Fluids with non-Newtonian
characteristics exhibit a non-linear relationship be-
tween the shear stress, , and the shear rate, . The rela-
tionship can be measured with an appropriate viscome-
ter (3). This non-linear behavior can provide some ap-
pealing process characteristics. For instance, paints and
inks that are highly shear-thinning and have a yield
stress can be easily applied using a high shear rate.
These materials will not drip or run once they have
been applied.
Fluid rheology strongly impacts the design of an ag-
itation system. The basics of this subject can be charac-
terized using the steady shear ow between two paral-
lel plates, as illustrated in Figure 1. In this simple ow
eld, is the force per area required to maintain Plate 2
in motion at velocity, u
x
, while Plate 1 is held station-
ary. The shear rate (), which is the velocity gradient,
du
x
/dy, is proportional to the lower plate velocity when
the gap between Plates 1 and 2 is xed. Newtonian u-
Apply these principles of power-law fluid
behavior to scale up mixing processes from
the laboratory to the production plant.
How to Scale-Up
Mixing Processes in
Non-Newtonian Fluids
L
Robert J. Wilkens
Univ. of Dayton
Christopher Henry
Northwestern Univ.
and Lewis E. Gates
ReyNo Inc.
y
x u
x
Plate 1
Plate 2
I Figure 1. Fluid velocity distribution between parallel plates.
CEP May 2003 www.cepmagazine.org 45
ids are those for which and have a linear relationship:
= (1)
The proportionality in this relationship, , is the New-
tonian viscosity. It is material-dependent and typically,
temperature-dependent. Many fluids exhibit Newtonian
behavior, including all gases, liquid metals, low-molecu-
lar-weight organic liquids and aqueous solutions of low-
molecular-weight solutes (e.g., CaCO
3
dissolved in
water).
There are many types of non-Newtonian behavior, all of
which complicate the analysis of mixing phenomena. Only
one of the most common behaviors, that of a power-law
uid, will be analyzed in this article. A power-law uid ex-
hibits the following relationship between and , where K
and n are the material-dependent ow consistency coeffi-
cient and ow behavior index, respectively.
= K
n
(2)
When n = 1, the uid is Newtonian. his relationship can
be used to dene the apparent viscosity,
a
, of a power-law
uid.
Under shear conditions, a power-law uids
a
is:

a
= / = K
n1
(3)
When n < 1,
a
decreases with increasing , and the uid
is termed shear-thinning (or pseudoplastic). For a shear-
thinning uid, increases with increasing . However, as
shown in Figure 2, the rate of increase is less than linear.
This is the most common non-Newtoni-
an behavior, often occurring in polymer
melts and biological uids:
When n > 1,
a
increases with in-
creasing , and the uid is termed shear-
thickening (or dilatant). For a shear-
thickening uid in Figure 1, increases
faster than linearly with increasing .
This behavior is also depicted in Figure
2. Shear-thickening behavior is most
often observed with concentrated slur-
ries, such as titanium dioxide and starch
pastes. One explanation is that at rest
and low shear rates, there is enough liq-
uid present in the void space between
solid particles to prevent the particle-
particle interactions that lead to greater
friction and an increase in
a
. However,
at higher shear rates, the suspension ex-
pands (dilates), such that the amount of
liquid present cannot overcome the fric-
tional forces between the particles and
the consequent increase in
a
.
The non-Newtonian uids discussed in this article are
the simplest of many complex behaviors, and do not in-
clude yield stresses, time dependence or elastic phenome-
na. Chhabra and Richardson (4) provide a thorough review
of non-Newtonian rheology.
Although the power-law model is an attempt to curve-t
real data with a simple mathematical relationship, it is
nonetheless useful in the engineering sense. For instance,
most shear-thinning uids behave in a Newtonian manner
(i.e., their viscosity is independent of ) at very high and
very low shear rates (4). Yet, little data spanning the com-
plete range of shear rates is available and typically cannot
be obtained using a single rheological device.
Thus, power-law model parameters may only be appli-
cable over a few orders of magnitude of . Care must be
taken to ensure that data should correspond to encoun-
tered during processing.
How does non-Newtonian behavior
affect agitation?
Relating the simple, steady shear ow of Figure 1 to the
ow eld in an agitated vessel is no simple task. The pri-
mary complication is that the ow eld in an agitated ves-
sel, including , is a function of position, and in many
cases, time. Therefore, for non-Newtonian liquids,
a
will
also depend on time and position within the mixed vessel.
The highest values of in an agitated vessel occur near
the impeller at points of highest local velocity, v. For open-
turbine impellers, v and are highest at the impeller tip;
decreases with increasing distance from the impeller and is
lowest at the vessel wall. Therefore, shear-thinning liquids
will have their lowest
a
in the impeller region of an open
Newtonian Fluid
Pseudoplastic Fluid
Dilatant Fluid
Shear Rate (g), 1/s
S
h
e
a
r

S
t
r
e
s
s

(

)
,

P
a
I Figure 2. Power-law uid response to shear.
turbine. With close-wall-clearance impellers, such as heli-
cal ribbons, shear-thinning uids will have the lowest
a
near the vessel wall. Common industrial open-turbine and
close-clearance impellers are shown in Ref. 2.
For open-turbine impellers, cavern formation may
occur in power-law fluids (5, 6). In these caverns, liquid
near the impeller will have motion. Pilot studies will in-
dicate whether this motion is sufficient to achieve pro-
cess objectives. The liquid farther from the impeller and
out of the cavern may have very low, or even negligible
velocities (7).
Because of the many complications that arise from
non-Newtonian fluids, this article assumes that mixing or
agitation must have been conducted successfully at a lab-
oratory scale (i.e., Scale 1). The requirement, then, is to
predict the intensity of mixing that is required in produc-
tion (i.e., Scale 2) to duplicate the process result that was
observed in Scale 1. Figures 3 and 4 illustrate Scales 1
and 2, respectively.
First the assumption of geometric similarity will be
made. Geometric similarity is achieved in Scale 2 by mak-
ing the ratio of all linear dimensions in Scale 2 identical to
those in Scale 1. Figures 3 and 4 are geometrically similar.
In mixing scaleup, the objective is to predict the rotational
shaft speed in Scale 2 (or n
2
), that will duplicate the perfor-
mance measured or observed in Scale 1 as a result of rota-
tional shaft speed in Scale 1 (or n).
Effective impeller shear rate
The uids varies throughout the volume of the tank.
Because of this,
a
also varies. For power-law uids, Met-
zner and Otto (8) suggest using an effective shear rate,
eff
,
to obtain an effective viscosity for the agitated tank:

eff
= k
s
n (4)
Bakker, et al. (7) suggest the use of impeller constants
in Table 1 to evaluate
eff
for a wide variety of impellers.
Substituting
eff
in Eq. 3 for the particular uid gives
eff
,
which can then be used to calculate the impeller Reynolds
number for power-law uids, Re
n
:

eff
= /
eff
= K(k
s
n)
n1
(5)
Impeller performance
characteristics database
There are several important dimensionless
numbers that are required to design mix-
ers. All must be determined experimental-
ly for a given impeller conguration.
These numbers can be used to quantify the
performance characteristics of an impeller.
Dimensionless numbers are affected by
geometric factors, such as the ratio of im-
Solids/Liquids Handling
46 www.cepmagazine.org May 2003 CEP
Nomenclature
b = cavern model parameter, m
C = height of the impeller (or shaft) above the vessel oor, m
D = diameter of the impeller, m
D
c
= cavern diameter, m
D/T = relative impeller diameter, dimensionless
D
c
/T = relative cavern diameter, dimensionless
du
x
/dy= velocity gradient
F = total impeller force, N
F
ax
= impeller axial force, N
F

= impeller tangential force, N

Fr = Froude number
g = gravity, m/s
2
K = ow consistency coefficient of a power-law uid, (N/m
2
)s
n'
k
s
= effective shear rate constant
N
B
= blend number, dimensionless
N
F
= force number, dimensionless
N
P
= power number, dimensionless
N
Q
= pumping number, dimensionless
n = rotational shaft speed, rev/s
n
2
= rotational shaft speed in Scale 2, rev/s
n = ow behavior index of a power-law uid
P = impeller power, W
P
i
= pitch of a helical ribbon impeller, m
P
M
= minimum motor power, W
q = theoretical impeller pumping rate, m
3
/s
Re
n
= Reynolds number for a power-law uid, dimensionless
s.g. = specic gravity, dimensionless
S
t
= impeller tip speed, m/s
t = impeller torque, N-m
T = tank diameter, m
u
x
= velocity of lower plate at a certain shear rate, m/s
V = volume, m
3
v = local uid velocity, m/s
v
bf
= bulk uid velocity, m/s
v
o
= cavern boundary velocity, m/s
Z = depth of the liquid in a vessel, m
Greek Letters
= shear rate, 1/s

eff
= effective shear rate,1/s
= absolute or dynamic viscosity, Pa-s

a
= apparent viscosity for a power-law uid, Pa-s (Eq. 3)

eff
= effective dynamic viscosity for a power-law uid, Pa-s (Eq. 5)
= blend time, s
= density, kg/m
3
= shear stress, N/m
2
(Eqs. 1 and 2)
T
2
T
D
2
n
2
n
Z
2
D
Z
I Figure 3. Scale 1 (left) or laboratory-scale mixer is the design
basis for scaleup. Process conditions are assumed to be ideal.
I Figure 4. Scale 2 (right) or production-scale mixer needs
sufficient agitation to replicate Scale 1 process conditions.
peller to tank diameter, D/T, and the ratio
of clearance from the tank bottom to tank
diameter, C/T. All these dimensionless pa-
rameters are normally correlated against
Re
n
for a baffled tank.
The impeller blend number, N
B
, is used
to predict the blend time, , in a mixed
system. N
B
attempts to predict the effect of
impeller D/T on the results:
N
B
= n(D/T)
2.3
(6)
The impeller power number, N
P
, is
used to predict impeller power, P, directly
and torque, t, indirectly:
N
P
= P/n
3
D
5
(7)
The impeller pumping number, N
Q
, is
used to predict the impeller pumping
rate, q, directly and the bulk fluid veloci-
ty, v
bf
, indirectly:
N
Q
= q/nD
3
(8)
The impeller force number, N
F
, corre-
lates the axial force, F
ax
, or the thrust
generated by an impeller. F
ax
is used in
the correlation to predict cavern dimen-
sions and is also important to mechanical
design considerations:
N
F
= F
ax
/n
2
D
4
(9)
Finally, Re
n
for power-law uids mea-
sures the ratio of inertial to viscous forces
within the mixing environment:
Re
n
= nD
2
/(k
s
n)
n1
(10)
As stated previously, all of the dimension-
less numbers just discussed are correlated
with Re
n
.
Examples of the correlations for N
P
,
N
B
, N
Q
and N
F
are given in Figures 58.
These correlations depict the trends ob-
served for axial-ow impellers, such as
pitched-blade or high-efficiency turbines,
for different values of D/T. Note the sud-
den drop in N
F
(Figure 8) when turbulent
ow is reached
While there is considerable informa-
tion in the literature about these dimen-
sionless numbers, no single reference pro-
vides a complete picture. The authors
CEP May 2003 www.cepmagazine.org 47
D/T = 0.25
D/T = 0.35
D/T = 0.45
Impeller Reynolds Number, Re
n
P
o
w
e
r

N
u
m
b
e
r
,

N
P
I Figure 5. N
P
is inversely proportional to Re
n
in the laminar and transitional ow regimes and
becomes constant in the turbulent regime.
I Figure 6. N
Q
generally increases with Re
n
in the laminar and transitional ow regimes and
becomes constant in the turbulent regime.
D/T = 0.25
D/T = 0.35
D/T = 0.45
Impeller Reynolds Number, Re
n
P
u
m
p
i
n
g

N
u
m
b
e
r
,

N
O
have assembled an extensive database
of impeller performance parameters for
scaleup. This database was stored as in-
dividual data points in a text le. In
order to nd values of the dimension-
less numbers between data points of im-
peller performance parameters, an inter-
polation algorithm is used. With this li-
brary, N
P
, N
B
, N
Q
and N
F
are available
for a variety of common industrial im-
peller types.
Scaleup criteria
The scaleup criteria discussed in this
article are based on the assumption that
Scale 1 and Scale 2 are geometrically
similar, and that Scale 2 is larger than
Scale 1. Scaleup criteria will be dis-
cussed in the order of increasing conser-
vatism in mixer design (i.e., mixer size
will be increasing as Scale 1 is convert-
ed to Scale 2 for the subsequent crite-
ria). Asample scaleup is performed later
in this article.
Equal Reynolds number Holding
Re
n
xed with scaleup results in the
smallest production mixer of any of the
criteria discussed in this paper. For this
reason, scaleup based on equal Re
n
is
not usually recommended.
Equal relative cavern diameter
Wichterle, et al. (9), who studied
pseudoplastic and plastic suspensions
of finely divided particulate solids,
used the term cavern to describe
the well-mixed, turbulent region
around the impeller. Amanullah, et
al. (6) proposed an axial-force model
for predicting the cavern diameters,
D
c
, in highly shear-thinning liquids
using axial-flow impellers:
D
c
= [4v
o
(2/n 1)(4
2
K/F)
1/n
+ b
(n 2)/n
]
n/(n 2)
(11)
where b = T/4 (12)
F = (F
ax
2
+ F

2
)
1/2
(13)
F

= 8t/3D (14)
F
ax
is derived from Eq. 9 and
t = N
P
n
2
D
5
/2 (15)
Solids/Liquids Handling
48 www.cepmagazine.org May 2003 CEP
D/T = 0.25
D/T = 0.35
D/T = 0.45
Impeller Reynolds Number, Re
n
B
l
e
n
d

N
u
m
b
e
r
,

N
B
Figure 7. N
B
is inversely proportional to Re
n
in the laminar and transitional ow regimes and
becomes constant in the turbulent regime.
D/T = 0.25
D/T = 0.35
D/T = 0.45
Impeller Reynolds Number, Re
n
T
h
r
u
s
t

N
u
m
b
e
r
,

N
F
Figure 8. N
F
decreases with Re
n
in the laminar and transitional ow regimes, and suddenly
increases before fully turbulent ow is reached.
Figure 9.
Shown is the ow
pattern created by a
high-efficiency |im-
peller rotating in a
power-law uid. Red
shows areas of high
velocity. Blue shows
areas that are stagnant
or that have very low
velocities. Courtesy of
Denise Minch, graphic
artist (DeniseMinch@
hotmail.com).
In principle, Eq. 11 can also be used to predict the size of
caverns generated by radial-ow impellers. It shows that
scaleup based on equal relative cavern diameter, D
c
/T, re-
sults in production mixers for Scale 2 that are slightly larg-
er than those suggested by scaleup based on equal Re
n
.
Although uid in the cavern is highly mobile, further
away from the impeller, where shear stresses are below the
yield stress, the uid may be stagnant. It is possible to in-
crease the size of the cavern by increasing the impeller di-
ameter or speed (Figure 9). The authors recommend that
D
c
of mixed regimes be evaluated, but they do not recom-
mend the use of D
c
/T as a scaleup mechanism.
Equal tip speed The impeller tip speed, S
t
, is a com-
mon scaleup criterion for industrial mixers, often associat-
ed with shear-sensitive mixing phenomena, such as parti-
cle- or droplet-size control:
S
t
= nD (16)
The authors recommend that equal S
t
be used as a scale-
up criterion only when small-scale studies have clearly
shown that desirable process performance correlates with
S
t
. One example might be when S
t
has proven to be an im-
portant predictor of desirable product turbidity for a liquid
product, or desirable particle-size distribution in a slurry.
Equal torque per volume Torque per volume, t/V, is
a measure of torque invested by a uid mixer per unit of
mixed volume. It is calculated as the ratio of t (from Eq.
15) to V, dened as:
V = T
2
Z/4 (17)
t/V is a practical and common scaleup criterion for uid
mixers because it relates directly to the size and torque ca-
pability of the mixer.
Equal bulk fluid velocity Bulk fluid velocity,
v
bf
, is often used as a scaleup parameter in low-viscos-
ity applications:
v
bf
= 4q/T
2
(18)
In Eq. 18, q is the volumetric flowrate of the fluid leav-
ing the impeller blades and
2
/4 is the cross-sectional
area of the tank. In addition, it can be shown that equal
S
t
, equal t/V and equal v
bf
are identical for turbulent-flow
conditions in Newtonian fluids when geometric similari-
ty is maintained.
Equal power per volume Scaleup based on equal
power per volume, P/V, is probably the most commonly
used criterion in mixing because it is easily understand-
able and practical. To calculate P/V, rearrange Eq. 7 to
solve for P:
P = N
P
n
3
D
5
(19)
and calculate V using Eq. 17. Other advantages of using
P/V as a scaleup criterion are that it correlates well with
mass-transfer characteristics in the mixer, and it is conser-
vative enough to provide adequate performance in produc-
tion-scale equipment particularly when no other strong
correlating parameter has been determined from small-scale
testing. The authors recommend P/V as a scaleup criterion
when no others are obvious.
Equal Froude number The Froude number, Fr, is
the ratio of inertial to gravitational force:
Fr = n
2
D/g (20)
Rotational speed tends to decrease with most scaleup crite-
ria. As shown in Eq. 20, n will have a powerful, non-linear
effect on Fr, causing Fr to decrease with increasing vol-
ume, even though D is increasing. Thus, scaleup based on
equal Fr is rarely used as a scaleup criterion because it re-
sults in a relatively large and expensive industrial mixer.
Equal blend time Because blend time, , increases
with increasing tank volume, maintaining equal with
scaleup is difficult and expensive so much that the se-
lection of equal as a scaleup criterion is often consid-
ered impractical.
= N
B
/(n(D/T)
2.3
) (21)
However, a moderate increase in in the large vessel
may reduce the power and torque requirements to a more
acceptable level from a cost perspective. Such tradeoffs are
often made in order to purchase agitation equipment of rea-
sonable size (2). If equal is an absolute design criterion,
economics will dictate how large the production-scale
mixer shall be.
From process design to production
Amixer design procedure must enable the user to speci-
fy the critical components of mixer construction. The criti-
cal components of a uid mixer are the motor, drive (speed
reducer or torque increaser), shaft and impeller.
Mixer motor The motor provides the power to drive
the impellers in the process uid at speeds required to de-
liver the desired process result. The motor must also pro-
CEP May 2003 www.cepmagazine.org 49
Table 1. k
s
values for effective shear rate (
eff
) model
(Eq. 5, Ref. 7).
Impeller Type k
s
High-efficiency 10
Pitched-blade 11
Straight-blade 11
Disc-turbine 11.5
Anchor (D/T = 0.98) 24.5
Helical-ribbon (D/T = 0.96, P
i
/D = 1) 29.4
The values for the anchor and helical ribbon
impellers are for the specied standard geometries only.
For other geometries, use:

eff
= 25 (D/T)
1/2
[P
i
(p
2
D
2
+ P
i
2
)1/2]
0.152
n and
eff
= 25 (D/T)
1/2
n
for anchor and helical ribbon impellers, respectively.
vide additional power for losses that occur in mixer drives,
seals and couplings. One must also plan for process upsets
and process-uid variability. After accounting for all power
requirements, one should select the next-largest available
motor from a database of mass-produced motors so that
costs are controlled.
Mixer drive The mixer drive must be designed for
the maximum torque required by the impeller system,
whether the mixer is in start-up or at steady-state opera-
tion. At startup, the low shear rates produced by the im-
peller will create high
eff
in a shear-thinning uid. At
steady-state conditions, the mixer drive transmits power,
reduces motor speed and increases torque. In addition, the
drive bearings normally support the mixer shaft and im-
peller system.
To specify a drive that will have sufficient torque for the
application, the user should select the American Gear Man-
ufacturers Assn. (AGMA) speed that is just below the cal-
culated operating speed. Selecting the next-lower AGMA
speed for drive increases the torque of the drive, and is
therefore conservative. The combination of higher power
and lower motor speed will require an impeller that is larg-
er than the geometrically similar impeller used as the basis
for calculations. This larger impeller will effectively im-
prove the nal mixer performance.
Mixer shaft The mixer shaft transmits power to the
impeller and must be designed properly for torque, bend-
ing moments and natural vibration frequencies. The calcu-
lations required for shaft design are complex and beyond
the scope of this paper. Detailed coverage is provided in
Ref. 10. Users should seek the expertise of consultants or
mixer manufacturers when performing calculations for
shaft design.
Mixer impeller After selecting the most appropri-
ate impeller configuration (e.g., axial, radial, pitch-blade,
high-efficiency, helical-ribbon, etc.) for Scale 1, a geo-
metrically similar impeller should be chosen for Scale
2. The mixer impeller must be properly sized to transmit
the power and torque from the motor to the process fluid
for proper function. Many of the mechanical design
specifications required to completely design a fluid
mixer are beyond the scope of this paper. In this article,
the emphasis is on the specification of the motor horse-
power, the mixer shaft speed, impeller configuration and
impeller size for proper process function. The complete
mechanical design of the fluid mixer should be left to
manufacturers experienced in that design technology.
Algorithm for scale-up of power-law uids
The necessary calculation sequence for scale-up of mix-
ing phenomena in power-law uids must consider the fol-
lowing physical conditions: Any impeller rotating in a uid
will shear the uid; the
eff
of the impeller must be deter-
mined. The induced shear alters the
a
of the uid as dic-
tated by power-law parameters. Thus,
eff
is used to calcu-
Solids/Liquids Handling
50 www.cepmagazine.org May 2003 CEP
Scaleup Algorithm
Step I. Dene successful small scale (Scale 1) conditions
Measure the power-law characteristics of the uid viscosity
measure the uid properties s.g., K and n
Measure the system geometry of Scale 1
measure C, D, T and Z
calculate V (Eq. 17), D/T and Z/T
Dene the impeller style used in Scale 1
nd k
s
for the specied impeller (e.g., disc, pitched-
blade, high-efficiency, helical-ribbon, etc.)
Find the optimum operating conditions
observe n
Calculate the Re
n
of successful conditions in Scale 1 (Eq. 10)
Look up N
B
, N
F
, N
Q
and N
P
for that calculated Re
n
and for
the geometry (C/T, D/T, etc.) in Scale 1
Calculate
eff
(Eq. 5)
Calculate F
ax
, F

, and F (Eqs. 9, 13 and 14)

select v
o
calculate D
c
(Eq. 11)
Calculate D
c
/T
Calculate P (Eq. 19)
Calculate P
M
(P
M
= P/0.85)
Calculate t (Eq. 15)
Calculate P/V and t/V
Calculate Fr (Eq. 20)
Calculate S
t
(Eq. 16)
Calculate v
bf
(Eq. 18)
Calculate (Eq. 21)
Output results to a spreadsheet for later comparison.
Step II. Dene Scale 2
Calculate the system geometry of Scale 2 from Scale 1 and
geometric similarity.
Calculate n
2
required to produce Re
n
of Scale 1
Calculate
eff
Calculate D
c
/T
Calculate P
Calculate P
M
Calculate t
Calculate P/V and t/V
Calculate Fr
Calculate S
t
Calculate v
bf
Calculate
Output the results to a spreadsheet or database
Compare the calculated characteristics in Scale 2 to Scale 1
Repeat the calculations in Step II, incrementing Re
n
, (and
thus, n
2
) and updating N
B
, N
F
, N
Q
and N
P
(Eqs. 69).
Step III. Compare the results until the desired
characteristics of Scale 2 match those of Scale 1.
Step IV. Specify a motor and a mixer drive.
late Re
n
in order to determine N
B
, N
F
, N
P
and N
Q
of the im-
peller system at those conditions. N
B
, N
F
, N
P
and N
Q
are
variable and are correlated against Re
n
and the system ge-
ometry (Figures 58). For a given impeller n, solve for P, t,
q and . This design procedure requires iterative calcula-
tions described in the sidebar algorithm.
Scaleup example
Scale 1 is a 20-gal pilot reactor with a six-bladed
Rushton disc impeller with a 5.96-in. dia. It is important
to note that the axial force produced by a Rushton im-
peller, or by any other radial-flow impeller, is effectively
zero, so that N
F
= 0 for all conditions studied in Scale 1.
The process fluid is a shear-thinning power-law fluid with
K = 40 N-m
2
s
n
and n = 0.224. After numerous runs, it
is determined that the optimum process performance is
achieved at n = 244 rev/min. The performance conditions
in the 20-gal reactor are summarized in Column 2 of
Table 2.
The planned production size of Scale 2 is 1,000 gal.
The objective is to select conditions in Scale 2 that will
provide equivalent performance to that observed in Scale
1. There is no one parameter in mixed systems that will
assure success in scaleup. The authors recommend look-
ing at a variety of parameters expected to influence per-
formance (e.g., S
t
, P/V, t/V, etc.) and then to use the cal-
CEP May 2003 www.cepmagazine.org 51
Table 2. Summary of sample scale-up analysis.
Column 1 Column 2 Column 3 Column 4 Column 5 Column 6 Column 7
Scale 1 Scale 2 Scale 2 Scale 2 Scale 2 Scale 2
Scaleup Basis Base Equal Equal Equal Equal Equal
Condition Re
n
T/V P/V Fr Relative
V, gal 20 1,000 1,000 1,000 1,000 1,000
D/T 0.33 0.33 0.33 0.33 0.33 0.33
s.g. 1 1 1 1 1 1
n 0.224 0.224 0.224 0.224 0.224 0.224
K, N-m
2
s
n
40 40 40 40 40 40
n, rev/min 244 57 71 109 127 244
T, in. 18.1 66.5 66.5 66.5 66.5 66.5
Z, in. 18.1 66.5 66.5 66.5 66.5 66.5
D, in. 5.96 21.9 21.9 21.9 21.9 21.9
S
t
, ft/min 380 327 408 626 730 1402
P, hp 0.0329 0.283 0.486 1.65 2.63 20.3
P
M
, hp 0.0387 0.333 0.571 1.95 3.1 23.9
t, in-lb
f
10 368 507 1,125 1,537 6,182
t/V, in-lb
f
/1,000 gal 500 368 507 1,125 1,537 6,182
P/V, hp/1,000 gal 1.93 0.333 0.571 1.95 3.1 23.9
N
P
4.61 4.58 4.07 3.83 3.85 4.2
N
F
0 0 0 0 0 0
Fr 0.256 0.0516 0.08 0.189 0.256 0.945
Re
n
48 49 72 154 202 645

eff
, cP 1,959 6,052 5,104 3,659 3,250 1,958
D
c
, in 17.7 65.1 66.1 66.5 66.5 66.5
D
c
/T 0.980 0.979 0.995 1.00 1.00 1.00
Relative
*
1 (assigned) 4.3 3.4 2.2 1.9 1.0
Recommended 190 56 68 100 125 190
shaft speed, rev/min
Recommended 0.25 0.5 0.75 2 5 25
motor power, hp
* Blend times in Columns 27 are relative to Scale 1.
culated values of those parameters from Scale 1 as a
database from which to select the final mixing conditions
in Scale 2.
The authors developed a computer program based on
the algorithm discussed in this article to facilitate the
analysis of mixer scaleup. The program starts the analysis
at a shaft speed specified by the user and increments to a
maximum shaft speed, also provided by the user. The cal-
culated information for each speed is exported to a file
that can be read by Excel or by another equivalent soft-
ware. The information is then summarized in a spread-
sheet where it can be studied in order to select the appro-
priate scaleup parameters.
Table 2 contains the scaleup calculation summary for
the 1,000-gal Scale 2 in Columns 37. Notice that each
successive scaleup criterion results in a n that is higher
than the previous criterion. The result is a bigger mixer
with higher torque, higher rst cost and higher running
costs (reected by power requirement), as expected.
While the authors generally recommend a P/V scaleup
criterion when no other criterion is obvious, Table 2 illus-
trates that relative for the P/V column is 2.2 times
greater than the equal relative scaleup shown in the last
column. This compromise must be considered versus the
higher costs of the mixer designed on the basis of equal .
The equations dened in this article are presented in SI
units instead of English units, because the latter would require
correction factors to make certain formulas dimensionless.
However, the example problem uses English units (e.g., P/V is
dened in hp/1,000 gal) with the understanding that most of
readers would prefer to analyze the problems in those terms.
Conclusions
Scaleup of mixing in power-law uids is complicated
because
a
varies with . Changes in the
a
can shift the
Re
n
into the transitional or laminar ow regimes, which, as
illustrated in Figures 58, causes a change in N
B
, N
P
, N
Q
and N
F
. The solution to the equations involved for power-
law uids becomes a trial-and-error process.
The design logic described in this article depends on hav-
ing reliable values of N
B
, N
P
, N
Q
and N
F
over the laminar,
transitional and turbulent ow regimes for the impeller sys-
tem being analyzed. Often, data are unavailable from a single
source. But, once obtained, these data allow selection of the
appropriate scaleup criterion, ultimately leading to successful
scaleup of power-law uid agitation.
CEP
Solids/Liquids Handling
52 www.cepmagazine.org May 2003 CEP
ROBERT J. WILKENS is an assistant professor of chemical engineering at
the Univ. of Dayton (300 College Park, Dayton, OH 45469-0246; Phone:
(937) 229-2627; Fax: (937) 229-3433; E-mail: [email protected]).
His research interests lie in fluid mechanics (agitation and pipe flow),
namely, impeller performance and multiphase drag reduction. Wilkens is
also the author of several papers in these areas. He holds bachelor and
master degrees from the Univ. of Dayton and a PhD from Ohio Univ., all
in chemical engineering, and has held a post-doctorate research
engineering position at Shell Westhollow Technology Center (Houston,
TX). He is a registered professional engineer in Ohio.
CHRISTOPHER HENRY is currently pursuing his doctorate at
Northwestern Univ. (1311 Chicago Ave, Apt. 204; Evanston, IL 60201;
Phone: (847) 467-6720; E-mail: c-henry@northwestern. edu), having
earned a BSChE from Univ. of Dayton (OH) in May 2002. During his
time as an undergraduate student, Henry worked for International
Papers, where he conducted development work for ink-jet coatings.
His current research is in bioinformatics.
LEWIS E. GATES is president of ReyNo Inc. (921 Howard Lane, Vandalia,
OH 45377; Tel: (937) 264-4112; E-mail [email protected]),
where he develops products, software and application procedures for
manufacturers and users of process equipment, while focusing
primarily on fluids mixing. After receiving both BS and MS ChE degrees
from the Ohio State Univ., Gates worked for DuPont and Chemineer
before launching ReyNo, Inc. He is a member of AIChE and a registered
professional engineer in the state of Ohio. ReyNo funded the work
upon which this paper is based.
Literature Cited
1. Brodkey, R. S. and H. Hershey, Transport Phenomena: A Unied
Approach, McGraw-Hill, New York, NY, pp. 359399 and pp.
752790 (1988).
2. McCabe, W. L., et al., Unit Operations of Chemical Engineering,
Sixth Ed., McGraw-Hill, New York, NY, pp. 238285 (2001).
3. ASTM D4040-99, Standard Test Method for Viscosity of Printing
Inks and Vehicles by the Falling-Rod Viscometer, (1999).
4. Chhabra, R. P. and J. F. Richardson, Non-Newtonian Flow in the
Process Industries, Butterworth Heinemann, Boston, MA (1999).
5. Nienow, A. W. and T. P. Elson, Aspects of Mixing in Rheological-
ly Complex Fluids, Chemical Engineering Research & Design,
Dept. of Chem. Eng., The Univ. of Birmingham, U.K., 66 (1988).
6. Amanullah, A., et al., A New Mathematical Model to Predict Cav-
ern Diameters in Highly Shear Thinning Power Law Liquids Using
Axial Flow Impellers, Centre for Bioprocess Engineering, School
of Chem. Eng., The Univ. of Birmingham, Birmingham, U.K.;
Chem. Eng. Science, 53 (3), pp. 455469 (1998).
7. Bakker, A. and L. E. Gates, Properly Choose Mechanical Agita-
tors for Viscous Liquids, Chem. Eng. Prog., 91 (12), pp. 2534
(Dec. 1995).
8. Metzner, A. B. and R. E. Otto, Agitation of Non-Newtonian Flu-
ids, AIChE J., 3 (1), pp. 310 (1957).
9. Wichterle, K. and O. Wein, Paper B4.6, International Congress of
Chemical and Process Engineering, CHISA 75, Prague, Czech Re-
public (1975).
10. Fasano, J. B., et al., Consider Mechanical Design of Agitators,
Chem. Eng. Prog., 91 (8), pp. 6071 (Aug. 1995).
Related Course Offered by AIChE
Course #090: Industrial Fluid Mixing
Member fee: $995, Non-member fee: $1,155
Houston, TX; October 2728, 2003
For more information or to register, visit www.aiche.org/education
or call (800) 242-4363. Onsite courses are also available.