Chemical Engineering Science, 1974, Vol. 29, pp. 2229-2234. Pergamon Press.

Printed in Great Britain

POWER CONSUMPTION SCALE-UP IN AGITATING

NON-NEWTONIAN FLUIDS

F. RIEGER and V. NOVAK

Department of Chemical Engineering, Fakulta strojni CVUT, 16607 Praha 6, Czechoslovakia

(Received 22 February 1974; accepted 9 May 1974)

Abstract-A method for the determination of power consumption of industrial agitators in

non-Newtonian fluids on the basis of measurements on a model device is presented. The applicability
of this method for purely viscous, time dependent and viscoelastic fluids is discussed. The method was
tested experimentally using agitators of two different sizes and results of measurements confirmed the
applicability of the present method for engineering calculations.

1. INTRODUCTION paper [6]. Neglecting the body forces, the equation

For calculation of agitator power consumption in of motion in terms of stress has the form
mixing non-Newtonian fluid, the following two
methods are usually used. The most frequent Lh
px= -vp +v. 7
method for calculation of power consumption was
suggested by Metzner and Otto[l]. The other one,
based on criterial equations is used for example in We suppose that rheological behaviour of the
papers of Mitsuishi and Hirai[2], Kelkar et al [3,4] mixed fluid can be described by the constitutive
and Rieger and Nov&k[5]. The majority of pro- equation of a simple fluid (see e.g.[7])
posed correlations is limited to the creeping flow
regime and was tested with pseudo-plastic fluids 7 = X$::F&) (2)
only. However, in mixing equipment of industrial
size, the condition of creeping flow regime very in which the dynamic stress tensor T is a functional
often does not apply. The method for scale-up of history of relative deformation gradient F*(t,).
suggested for purely viscous fluids in our previous In the next step we introduce the following
paper [6] is also applicable for creeping flow regime dimensionless variables
only.
The subject of this paper is to suggest a new u* = u/Nd, p * = p /pN2dZ, V* = dV, t * = Nt,
scale-up method which can be used above the
creeping flow regime and we generalize its applica- T* = 7/+jN, S* = S/d’, x* =x/d (3a + g)
bility also to time dependent and viscoelastic fluids.
The proposed method is suitable especially for such where 1 is a characteristic viscosity function, the
types of agitators and such rheological behaviour of form of which is dependent on the specific form of
mixed fluid, for which neither criteria1 equation nor rheological equation used. The characteristic vis-
value of the k constant, necessary for application cosity function q can always be expressed through
of Metzner-Otto method[l], can be found in the constants of the rheological equation and,
literature. An important advantage of this method is eventually, speed of agitator. e.g. for a power-law
a fact, that the rheometric measurement of mixed fluid we used the definition fl = KN”-’ (see
fluid behaviour is not necessary. ref. [S]), Kelkar et a[.[31 used fj = p. for an Ellis
fluid.
Inserting the dimensionless variables (3) to the
2. THEORY Eqs. (1) and (2), we obtain
The method of dimensional analysis of the
dynamic equation and the rheological constitutive pNd2 Du* pNd2
ii*=- -v*p*+v*~r*
q (4)
equation is applied, similarly as in our previous
2230 F. RIECER and V. NOVAK

and Regarding to the Eqs. (6), (7) and (8), it follows that

(5) &=f(t*, Re, Rhl, Rhz, . . . Rh) (12)

We should add the equation of continuity and

For Newtonian fluids Eq. (12) reduces to the
boundary conditions (see e.g.[5]) to the complete
dependence among the power and Reynolds num-
description of flow in a vessel, but no additional
bers
informations would be obtained. The continuity
equation and boundary conditions influence only
the form of resulting dimensionless equations. -&f&=f”(Jy) (13)
The inspection of Eqs. (4) and (5) leads to the
following dimensionless general equations for
Defining the inverse function, we can express the
velocity u*, pressure p * and dynamic stress tensor
Reynolds number
r* in the mixed fluid

U* = u*(x*, t*, Re, Rhl, Rhz, . . . Rhi) (6) ReN =%=gN(&) (14)

p* = p*(x*, t*, Re, Rhl, Rhz, . . . Z&l (7)

Expressing the Reynolds number from general Eq.
(12), we receive
T* = 7*(x*, t*, Re, Rh,, Rhz,. . . Rhi) 03)

=Nd’p=g P
All the quantities above depend on the dimension- Re
- t*, Rh,, Rht, . . Rhi
rl pN’d” )
less coordinate x*, the dimensionless time, t*,
(15)
Reynolds number Nd’plq and on additional
dimensionless criteria Rhi, containing the constants
The principle of the suggested method is an
of rheological constitutive equation and eventually
assumption that the form of the Reynolds number
the speed of impeller N. The specific form of
dependence on power number is independent of
dimensionless criteria Rhi depends on the form of
rheological behaviour of mixed fluid and is deter-
functional X in rheological Eq. (2). e.g. for a
mined by the type and geometry of agitator only.
power-law fluid i = 1 and Rh, = n (see ref. [5]), for
This assumption has the following physical in-
Ellis fluid i = 2, Rh, = Q, Rhz = Nt, as shown by
terpretation: to the same value of the power
Kelkar et al.[3].
number corresponds the same relative influence of
Agitator power consumption P can be deter-
inertial and surface forces represented by the form
mined as an integral of the scalar product of
of the dependence of the power number on the
velocity u and elementary force dF acting on the
Reynolds number. Expressing the above statement
differential area of agitator surface, dS
mathematically, we can rewrite the function g in
Eq. (15) using the known form of Newtonian
P= u.dF= u.a.dS (9) Reynolds-power number relation (14) and we
f xv f .YM
obtain
in which force dF is expressed as a product of the
total stress tensor o and elementary area dS. The ~=g’+&)g’(t*,Rh,.Rhl,...Rh,) (16)
total stress tensor m can be expressed in the form
Let us define the effective viscosity of non-
a=7--pPS (10) Newtonian fluid pef as a viscosity which would
have a Newtonian fluid of the same density at the
After inserting Eq. (10) in Eq. (9) and using
same power consumption and speed of agitator.
dimensionless variables (3) we get the following
According to this definition, the agitator power
expression for power number
consumption in a non-Newtonian fluid can be
expressed from Eq. (13) replacing I_Lby pCf
 Agitating non-Newtonian fluids
and for inverse function
Dividing Eq. (16) by Eq. (18), we obtain the basic
equation for scale-up
/he, = qg’(t*, Rh,, Rhz, . . . Rhi)
From Eq. (19) it follows that the effective viscosity
IL6 for a given agitator depends on rheological
behaviour of mixed fluid, speed of impeller and on
the time of mixing. For a given fluid and agitator
geometry, pef is a function of the agitator speed N
and the time of mixing t only and is independent of
the size of mixing equipment. Therefore the
relation pCLd(f, N) obtained experimentally on a
model device can be used for the calculation of
power consumption of industrial agitators used for
mixing the same fluid. Power consumption of
industrial agitator can then be calculated from the
Newtonian power curve on the basis of Eq. (17).
Furthermore, let us focus our attention to the
specific types of rheological behaviour of mixed
fluids.
(i) Purely viscous fluids
The dynamic stress tensor in purely viscous
fluids depends on the present value of the relative
deformation gradient F,(t, = t) and is independent
of its history. The present value of the relative
deformation gradient can be expressed uniquely by
the rate of deformation tensor A = f [Vu + (Vu)‘].
For this reason, the constitutive rheological equa-
tion will not contain time and consequently the time
variable t disappears from all the above equations.
(The time dependency in Eqs. (6)-(8) caused by
impeller rotation can be eliminated by the choice of
coordinate system rotating with impeller.) Then for
purely viscous fluids Eq. (12) takes the form
& =f(~e, Rh,, Rh2,. . . Rh)
and the Eq. (19) becomes
I*ef= +jg’(Rh,, Rhz, . . . Rhi)
From the last Eq. (21) it follows that the effective
viscosity pef for a given fluid is a function of the
impeller speed only
cLef = weAN)
CES Vol. 29No. 1I-G
2231
(ii) Viscoelastic fluids
Scale-up for viscoelastic fluids can be based on
Eq. (19) from which we have for the effective
(18) viscosity
cLer= F&N, t) (23)
In most of the viscoelastic fluids however, the
(19) natural fluid time is small and in many cases is
shorter than the agitator starting-up. So that the
time dependence of effective viscosity is hot
measurable. For this reason, the scale-up in mixing
of the most viscoelastic macromolecular solutions
can be based on equation for purely viscous fluids
(22). On the other hand Bulena[8] reported that in
kneading dough-like materials the time dependence
cannot be neglected.
(iii) Thixotropic fluids
Scale-up for thixotropic fluids can be based on
the same equation as for viscoelastic fluids (23). It
means that the effective viscosity for a given
thixotropic fluid depends on the impeller speed N
and time of mixing t.
3. EXPERIMENT
The aim of the experiments was the confirmation
of the assumption that the effective viscosity for a
given fluid and agitator geometry depends only on
agitator speed N by Eq. (22) or on agitator speed N
and time t by Eq. (23) and is independent of the
agitator size.
Power measurements were carried out on an
adapted viscometer Rheotest description of which
can be found in[9] or on the turntable described
in [ lo]. In the case of thixotropic fluid agitation, the
force acting on the plate of turntable was transmit-
ted from strain-gauges to the input terminals of a
fast recorder. Rheological properties of the fluids
used in measurements were tested on the rotational
viscometer Rheotest and on a rheogoniometer with
plate-plate and cone-plate arrangements.
The plot of effective viscosity pIf vs impeller
(20) speed N measured with two sizes of anchor
agitator (see Fig. la) in the vessels of diameters 100
and 150 mm is shown in Figs. 2 and 3. Results in
Fig. 2 were obtained with inelastic solution of
(21) carboxymethylcelulose. In Fig. 3 are results of
experiments with slightly elastic solution of polyac-
rylamide. From both figures, it is obvious that the
experimental values of pFLcf for both sizes of anchor
agitators lie on the same curve.
In Fig. 4 are shown results of measurements with
(22) solution of polyisobutylene in decaline carried out
2232 F. RIEGER and V. NOVAK

(b)

Fig. 1. Agitators used in measurements.

Fig. 2. p,r vs N for anchor agitators in CMC solution.

Fig. 4. y,, vs N for anchor agitators in solution of

polyisobutylene in decaline. OD = 150 mm; AD =
100 mm.

20
I I
I
Fig. 3. jar vs N for anchor agitators in PAA solution.

with the anchor agitators mentioned above. In spite

of the fact that the solution was characterized by
considerable normal stress phenomena, its natural
time was not long enough, to obtain measurable oD=l50mm
*II=IOOmm
time dependence of the power consumption. I
Therefore the results were plotted in the form 05
% -.-.-_.-.
r*_&N) which is independent on the size of agitator .-

(see Fig. 4). 1 1

In measurement with thixotropic plastic plaster, 1 I
the screw agitator with draught tube shown in Fig. 0 500 1000

t (51
lb was used. The course of torque at constant
speed of agitator N = 60 l/min was recorded and is Fig. 5. Time dependence of torque of screw agitators in
shown in Fig. 5. Using values of torque from Fig. 5 plastic plaster mixing (constant speed N = 1 sect ‘J.
 Agitating non-Newtonian fluids 2233

and Eq. (18), the values of pIf for various times t P -

were calculated and the plot CL+vs t is shown in
Fig. 6. Experimental results are in a relatively good
agreement with the theoretical assumption rep-
resented by Eq. (23).
m = f(N) (27)
is for a given fluid and agitator geometry indepen-
dent of equipment size. This method is suitable for
purely viscous fluids and for viscoelastic fluids with
short natural time. The Eq. (17) in the creeping flow

‘O’2 regime takes the form

P
PN 3d (%
3=A 0 >-’ (28)

Expressing pef from the latter equation, we obtain

----P 1 P
&f = AN?& - AN Nd’ (29)

Inserting for P/Nd3 from Eq. (27), we get

1 -
5 IO 102 IO'
cLef= AN f(N) = p&N)
t rs1
which is in agreement with relation (22). On the
Fig. 6. pd vs t for screw agitators at constant speed
N = 1 set-’ in plastic plaster. other hand, substituting pcL6from Eq. (23) to Eq.
(29), we obtain the relation

4. DISCUSSION

As mentioned above, the Metzner and Otto &=fW,t) (31)

method[l] is usually used for the calculation of
power consumption of agitators in non-Newtonian Equation (31) is convenient for scale-up of power
fluid mixing. According to this method, the consumption in mixing of time-dependent fluids
effective viscosity equals to the apparent viscosity and viscoelastic fluids with longer natural time in
determined from the rheogram of a given fluid at the creeping flow regime.
the shear rate jam,calculated from equation From the preceding discussion it is clear, that
Eq. (22) resp. (23) reduce to Eqs. (27) resp. (31) in
yef = kN (24) the creeping flow regime and both methods of
scale-up are equivalent. Nevertheless scale-up by
where the coefficient k is a constant for a given Eqs. (27) resp. (31) is more simple in the creeping
type and geometry of agitator. The determination flow regime, because the Newtonian power
of effective viscosity from rheogram can be number-Reynolds number curve is not necessary.
described by equation
5. CONCLUSION

On the basis of theoretical considerations, the

Inserting yd from Eq. (24), we arrive for a given
agitator geometry to the relation
per= /4N) (26)
which is in agreement with our equation for purely
viscous fluids (22).
For scale-up in the creeping flow regime, in
ref. [6], we proposed the method based on the
statement that the function
method for scale-up of power consumption of
agitators in mixing non-Newtonian fluids was
suggested. The following procedure for estimation
of power consumption is proposed:
(1) Carry out power measurements with a given
fluid at a given temperature on a geometrically
similar model
(2) Obtain the dependence of power consumption
on agitator speed or, if necessary, on agitator
speed and time.
2234 F. RIEGER and V. NOVAK

(3) For each value of the power consumption N speed of agitator

calculate the power number and, from the P pressure
corresponding Newtonian power curve for the P power
same agitator, determine the Reynolds number. s area
(4) Calculate the effective viscosity by the relation t particulate time
pef = Nd’plReN and plot pd vs N or, if b time variable
necessary, pef vs N and t. U velocity
(5) The power consumption of an industrial agitator Re Reynolds number Nd*p/tj
can be calculated using the Newtonian power ReN Reynolds number for Newtonian fluids
curve inserting pef for p into Reynolds number. Nd*pIp
The value of effective viscosity pef can be Rh dimensionless criterium containing the
determined from the plot per(N) respective rheological constants and eventually the
pd(N, t) obtained (by the procedure suggested speed of agitator N
in steps l-4) from measurements on a geometri- shear rate
cally similar model. fi characteristic viscosity function
CL dynamic viscosity
This procedure is quite exact in the creeping flow
PLd effective viscosity
regime, above the creeping flow regime it is based
P density
on the validity of the assumption defined by Eq.
u total stress tensor
(16). Even if this assumption is only approximate, it
7 dynamic stress tensor
was confirmed by our experiments with close-
clearance agitators. Also the application of this REFERENCES
method to the calculation of power consumption of
[l] Metzner A. B. and Otto R. E., A.Z.Ch.E. JI. 1957 3 3.
industrial equipment for plastic plaster mixing was [2] Mitsuishi N. and Hirai N., Chem. Engng. Japan 1969
successful. On the other hand, the experimental 2 217.
results with turbine agitator above the creeping [3] Kelkar J. V., Mashelkar R. A. and Ulbrecht J., Chem.
flow regime reported in literature (see e.g. refs. [1 l] Engng Sci. 1973 28 664.
[4J Kelkar J. V., Mashelkar R. A. and Ulbrecht J., Trans.
or [12]) showed that the Metzner-Otto method
Inst. Chem. Engr. 1972 50 343.
(which is the special case of the presented method [5] Rieger F. and Nov&k V., Trans. Inst. Chem. Engr.
as shown in discussion) leads to overestimated 1973 51 105.
values of power consumption. Since for most [6] Rieger F. and Novak V., Chem. Engng Sci. 1972 27
engineering purposes, slightly conservative values 39.
[7] Truesdell C., The Non-Linear Field Theories of
are desirable, we believe that eventual overestima- Mechanics, Encyklopedia of Physics, Vol. III,
tion will keep the design on the safe side. Springer-Verlag. Berlin 1965.
[S] Bulena V., Scale-up for Kneading Systems (in
Czech), 20th National Conference CHISA, Brno
NOTATION
1973.
d agitator diameter [9] Novak V. and Rieger F., Trans. Inst. Chem. Engr.
D vessel diameter 1969 47 335.
[IO] Novak V. and Rieger F., Colln. Czech. Chem.
F force Commun. 1973 38 350.
F, relative deformation gradient [I I] Metzner A. B., Feeks R. H., Ramos L. H., Otto R. E.
k coefficient in Eq. (24) and Tuthil J. B., A.Z.Ch.E. JL 1961 7 3.
K consistency index [12] Rieger F. and NovBk V., Technika u Chemii, 1972 48
n flow behaviour index
1.