Professional Documents
Culture Documents
Short communication
a r t i c l e i n f o a b s t r a c t
Article history: Rotor-stator devices are increasingly important for process intensification applications but the under-
Received 17 August 2007 standing of their operation and scale-up is poor. Consequently a new expression is presented for the
Received in revised form 3 March 2008 calculation of the delivered power, which consists of three terms. Firstly a term reflecting the power
Accepted 4 April 2008
required to rotate the shaft in response to the resistance of the fluid. Secondly a term reflecting the con-
Available online 12 April 2008
vection of power away from the mixing chamber due to the flow of product. Thirdly, a term reflecting
losses. A constant is required for each term and the expression describes the power consumption over the
Keywords:
range 5–20 kW with a RMSE of 0.5 kW.
Power number
Power draw © 2008 Elsevier B.V. All rights reserved.
Power consumption
Process intensification
Rotor-stator device
High shear mixer
2. Experimental methods
∗ Tel.: +44 151 641 3668. The model is validated with data from a Siefer Trigonal mill
E-mail address: Adam.Kowalski@Unilever.com. (http://www.siefer-trigonal.de/) as a function of rotor speed, flow
0255-2701/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2008.04.002
582 A.J. Kowalski / Chemical Engineering and Processing 48 (2009) 581–585
mill in air and PM is the motor power. The motor power can be
determined directly from the current measurement by
√
PM = 3 VI cos (3)
where V is the line voltage (V), I the current (A), cos ϕ is the power
factor and the 30.5 comes from using a three-phase supply. Normally
cos ϕ varies with load but we will assume that the load is relatively
high and so cos ϕ will be fairly constant [12] with a value of about
0.8. The delivered power (P) and current (I) are therefore related
(Eqs. (2) and (3)) by the constant 460 W/A.
The first term on right hand side of Eq. (1) (PT ) is due to the
torque on the motor shaft in the absence of any flow. Following
the traditional approach [3–6,13,14] the Reynolds number (Re) and
power consumption for a batch process are
PT = P0 N 3 D5 (4)
(ND)h
Re = (5)
Fig. 1. Schematic of (a) Siefer Trigonal mill and (b) experimental arrangement.
Fig. 2. Milling data for soda ash slurry showing the effect of flow rate on current Fig. 4. Comparison of measured and predicted power consumption for soda ash
draw for various rotor speeds and milling gaps (see legend). slurry.
4. Results 5. Discussion
Fig. 2 presents a sample of the data for the soda ash slurry The similar values for k1 and PL for the two slurries are reassur-
showing current draw as a function of flow rate for different rotor ing as they indicate that they are characteristic of the rotor-stator
speeds for a fixed milling gap of 100 m. Fig. 3 presents the same device rather than the fluids. However, the difference in the values
data in terms of MD2 N2 from the second term in Eq. (9) which for k0 is of concern and moreover their magnitude is smaller than
describes the effect of flow rate. A comparison of the two figures we expected. Fig. 6 includes 95% confidence intervals for all three
shows that the second term of Eq. (9), introduced to account for constants which show that even for k0 that there is a significant
the effects of flow rate, provided an adequate description of flow overlap in the confidence intervals indicating that the differences
rate. are not statistically significant. This is in part due to the relatively
To complete the full model fit we need to estimate the vis- few data points available for the Durcal 130 resulting in broad con-
cosity of the product as part of the PT term in Eq. (7). The fidence intervals. Nevertheless, the value of k0 is still unexpectedly
radial flow field due to the product flow through the mill results low when typical values are of order 10 or 100 [4] or even up to
in a range of shear rates, which are a minimum at the centre 1000 [5]. The Siefer mill has some geometric similarities to a cone
and a maximum at the periphery. The shear rate will vary with and plate rheometer and purely for the sake of interest the value
rotor speed and gap but for a rotation speed of 4000 rpm and of k0 was estimated for the latter. The data used was obtained dur-
a gap of 100 m a shear rate of order 106 s−1 is characteristic. ing calibration of a TA Instruments AR2000 rheometer with a 4-cm
Unfortunately conventional rheometry is limited to a shear rate diameter, 2◦ cone using 1000 cSt calibration grade silicone oil. The
of approximately 1000 s−1 . Although it must be recognised that value obtained for k0 of about 190 contrasts strongly with values
this is a fairly crude approximation, we will use the viscosities obtained for the Siefer mill which are of order 0.1 (Fig. 6) and sug-
measured at the limit of rheometric technique as a representa- gests that the value for the Siefer mill is probably too low. This is
tive viscosity. Thus for the 50% soda ash slurry we use a value thought to be a consequence of the viscosity values used for the two
of 0.16 Pa s and for the 60% calcite slurry a value of 0.094 Pa s. slurries. Two main concerns have been identified which are set out
We shall discuss shortly the consequences of these approxima- below:
tions.
The soda ash and calcite data are fitted separately by the • Viscosity can be sensitive to the size and shape of particles both
model with the constants k0 , k1 and PL obtained as part of the of which change as the slurry is milled. In addition for soda ash
fit. Figs. 4 and 5 present a comparison between the measured and there is the added complication that it is a porous particle and
predicted powers and shows that the expression provides a good so the phase volume reduces as the particles are milled. These
description of power draw(see Fig.6 for fit parameters). considerations suggest that using a single value for the unmilled
Fig. 3. Data of Fig. 1 as a function of the flow rate term of Eq. (8). Fig. 5. Comparison of measured and predicted power consumption for calcite slurry.
584 A.J. Kowalski / Chemical Engineering and Processing 48 (2009) 581–585
slurry is inappropriate and it should instead be determined for The comparison of measured with the calculated values of
each data point. power obtained from Eq. (9) and using the constants shown in Fig. 6
• The viscosities used in the model correspond to a shear rate, should result in a correlation of gradient 1 and an intercept pass-
which is not representative of the mill operation. Rheological ing through the origin. However, for both slurries the gradient of
techniques typically only measure viscosities up to a shear rate the line is less than 1 with a non-zero intercept (Figs. 4 and 5).
of 1000 s−1 [17]. Typically as shear rate increases viscosities drop Fig. 6 presents the standard error in the gradient and intercept
(shear thinning) and so the values used are too big. Furthermore for the linear fit together with the resulting 95% confidence inter-
the temperature rise as a consequence of milling will further vals. The confidence intervals for the Durcal 130 are again large
reduce the viscosity. So, for example, for the calcite slurry in water due to the relatively few data points but nevertheless suggest
if we were to assume that in the extreme the slurry took on the that the anticipated values of gradient and intercept are plausi-
viscosity of water (i.e. 0.001 Pa s) we would expect the value of k0 ble. However, for the soda ash slurry the 95% confidence intervals
to be proportionally increased and take a value of approximately only just include a gradient of unity and a zero intercept. A pos-
15 which is now consistent with values reported from stirred tank sible cause for this is again due to the single viscosity value used
studies. for the slurry. The first term in Eq. (9) is proportional to viscos-
ity and a poor description of viscosity will be reflected in the
The other source of error is in the definition of the motor char- value of k0 and in the calculation of the power draw. Future work
acteristics. In particular it is known that the power factor, efficiency should start with simple Newtonian fluids where viscosity is well
and losses vary with load but the form of these relationships is not known.
known for this mill. An indication of some of the contributions to Finally we note that the power dependence on rotor speed N
the power losses is obtained from the power drawn when running varies from a squared term, where the flow term dominates, to a
in air of about 2–3 kW whereas the value of PL from the fit is higher cubic term in N, where the power number term dominates. Thus
at about 6–7 W. the experimentally determined relationship mentioned earlier
A.J. Kowalski / Chemical Engineering and Processing 48 (2009) 581–585 585
showing a dependence of N2.5 [9] can now be understood as PT power required to rotate rotor and overcome fluid resis-
resulting from contributions from the two terms. tance (W)
PM the motor power (W)
6. Conclusions P0 power number
V voltage (V)
An expression for the delivered power for a rotor-stator device
including the effects of flow rate has been successfully developed. Greek letters
This was based on the traditional power number approach used the efficiency of the motor (approximately 0.8)
for batch vessels but with the introduction of an additional term to viscosity (approximately 0.16 Pa s for soda ash and
account for the effects of flow rate and required three experimen- 0.94 Pa s for Durcal 130)
tally determined constants. product density (kg/m3 )
Two of the constants (k1 and PL ) are similar for the two slurries
indicating that they are, as expected, characteristic of the equip- References
ment rather than the materials. The other constant (k0 ) is different
for the two slurries although the difference does not appear to be [1] V.A. Atiemo-Obeng, R.V. Calabrese, Rotor-stator mixing devices, in: E.L. Paul,
statistically significant. It is however much smaller than expected V.A. Atiemo-Obeng, S.M. Kresta (Eds.), Handbook of Industrial Mixing, John
Wiley & Sons, Inc., 2004 (chapter 8).
and it is suggested that this is a consequence of a poor description [2] D. Cohen, How to select rotor-stator mixers, Chem. Eng. 105 (1998) 76–79.
of the slurry viscosity. [3] L. Xie, C.D. Rielly, W. Eagles, G. Ozcan-Taskin, Dispersion of nano-particle clus-
The correlation between predicted and measured power draw ters using mixed flow and high shear impellers in stirred tanks, Chem. Eng. Res.
Des. 85 (2007) 676–684.
is not exactly one to one. This thought to be due to assumptions [4] L. Douchet, G. Ascanio, P.A. Tanguy, Hydrodynamics characterisation of rotor-
about the motor performance and again the poor description of the stator mixer with viscous fluids, Trans IChemE A, Chem. Eng. Res. Des. 83 (2005)
slurry viscosity. Future work should start with Newtonian fluids. 1186–1195.
[5] K.J. Myers, M.F. Reeder, D. Ryan, Power draw of a high-shear homogeniser, Can.
Nevertheless the model accounts for the effects of flow rate in J. Chem. Eng. 79 (2001) 94–99.
an appropriate way and predicts the power over a range from 5 to [6] M.F. Edwards, M.R. Baker, J.C. Godfrey, Mixing of liquids in stirred tanks, in: N.
20 kW with a root mean square residual of about 0.5 kW. Harnby, M.F. Edwards, A.W. Niewnow (Eds.), Mixing in the Process Industries,
2nd ed., Butterworth Heinemann Ltd., 1992 (chapter 8).
[7] J.R. Bourne, J. Garcia-Rosas, Rotor-stator mixers for rapid micromixing, Chem.
Acknowledgement Eng. Res. Des. 64 (1986) 11–17.
[8] G.-W. Chu, Y.-H. Song, H.-J. Yang, J.-M. Chen, H. Chen, J.-F. Chen, Micromix-
ing efficiency of a novel rotor-stator reactor, Chem. Eng. J. 128 (2007)
The author would like to acknowledge Mr Geraint Roberts of 191–196.
Unilever R&D for providing the raw data from the TA Instruments [9] J.R. Bourne, M. Studer, Fast reactions in rotor-stator mixers of different size,
AR2000 rheometer, which allowed for the extraction of k0 . Chem. Eng. Process. 31 (1992) 285–296.
[10] T.G. Sparks, D.E. Brown, A.J. Green, Assessing rotor/stator mixers for rapid chem-
ical reactions using overall power characteristics, in: Proceedings of the First
Appendix A. Nomenclature International Conference on Process Intensification for the Chemical Industry,
Paper 3.1.4, 1995.
[11] J. Bałdyga, A.J. Kowalski, M. Cooke, M. Jasińska, Investigations of micromixing
in the rotor-stator mixer, in: XIX Polish Conference of Chemical and Process
cos ϕ power factor (0.8 approximately) Engineering, September, Rzeszów, Poland, 2007.
D rotor diameter (0.22 m approximately) [12] D.E. Brown, Impeller power input measurement in a production-scale fer-
menter, in: Paper N IChemE Symposium Series No. 64, 1981.
h rotor-stator gap (m) [13] D.A.R. Brown, P.N. Jones, J.C. Middleton, G. Papadopouleos, E.B. Arik, Experi-
I current (A) mental methods, in: E.L. Paul, V.A. Atiemo-Obeng, S.M. Kresta (Eds.), Handbook
k0 , k1 constants obtained from fitting procedure of Industrial Mixing, John Wiley & Sons, Inc., 2004 (chapter 4).
[14] M.R. Baker, Mixed flow hydrodynamics, in: N.P. Cheremisinoff (Ed.), Advances
M mass flow rate (kg/s) in Engineering Fluid Mechanics Series, Gulf Publishing Company, 1996 (chapter
N rotor rotation rate (rps) 16).
PF power convected away from the milling chamber by the [15] G. Ascanio, B. Castro, E. Galindo, Measurement of power consumption in stirred
vessels—a review, Chem. Eng. Res. Des. 82 (2004) 1282–1290.
flow
[16] S. Triton, Physical Fluid Mechanics, CUP, 1982.
PG gross power dissipated in the product (W) [17] H.A. Barnes, Viscometry, in: A Handbook of Elementary Rheology, Institute of
PL power losses obtained from fitting procedure (W) Non-Newtonian Fluid Mechanics, Wales, UK, 2000 (chapter 7).