R Grenville Thesis 1992 PDF

-
CRANFIELD INSTITUTE OF TECHNOLOGY
SCHOOL OF MECHANICAL ENGINEERING
DEPARTMENT OF FLUID ENGINEERING AND INSTRUMENTATION
PhD Thesis
Academic Year 1991 - 1992
R. K. GRENVILLE
BLENDING OF VISCOUS NEWTONIAN AND

PSEUDO-PLASTIC FLUIDS
Supervisor: Prof. M. Sanderson
April 1992
is
This thesis submitted in partial submission for the degree of Doctor of Philosophy
ABSTRACT
Blend times and power consumption have been for Newtonian and pseudo-
measured
plastic fluids in agitated, baffled vessels with diameters of 0.30,0.61 and 1.83 m. The
viscosities of the fluids were selected to give a range of Reynolds numbers between 200
and 105. Four impellers were tested with diameters between T/3 and T/2 positioned at
T/3 clearance above the lowest point of the vessel base. The impellers were typical of
those found in the chemical industry.
The blend times were measured using a conductivity technique. A critical Reynolds
number, Rec,;l, was identified for each impeller. For both the Newtonian and pseudo-
plastic results the data could be divided into two groups:
Re > Rec,; NO = constant

l
Re < Recrij NO a Re-'
The data taken with the four impellers could be combined and expressed in terms of
three dimensionless groups; power, Reynolds and Fourier numbers.
Analysis of the Newtonian blend time data shows that blending in the vessel as a whole
is controlled by blending near the vessel wall. Analysis of the pseudo-plastic blend
time data is based on estimation of the fluid viscosity in this region of the vessel,
calculated from a torque balance between the agitator shaft and the vessel surfaces.
Both Reynolds and Fourier numbers are calculated using this value of viscosity. The
correlation of the pseudo-plastic blend time data, using this method, is the same as the
Newtonian correlation and the two are not statistically different.
A cost-benefit analysis has been carried out to compare the running and capital costs
.
of agitators for blending. Running costs are based on the agitator's power consumption
and maintenance costs. The capital cost is based on the shaft torque since this will
determine the mechanical strength of the agitator drive and shaft.
The cheapest blending of a Newtonian fluid is achieved by an agitator with a low power
number impeller while the cheapest blending of a highly pseudo-plastic fluid is achieved
by an agitator with a high power number impeller.
ACKNOWLEDGEMENTS
Firstly, I would like to thank the members of the Fluid Mixing Processes consortium
for funding the research reported in this thesis and for the interest they have taken in
my work.
Secondly, I would like to thank the members of my Total Technology PhD Study Panel:
namely Prof. Mike Sanderson of Cranfield Institute of Technology, Prof. John Smith
of the University of Surrey, Mr. Mike Sweeney of Cranfield School of Management and
Prof. Roger King and Mr. Colin Armstrong of the British Hydromechanics Research
Group Ltd., Cranfield, for their support and advice over the last four years.
Next, I would like to thank my collegues at BHRG Ltd. for their interest and
constructive comments on my work and, in particular, I would like to thank Dr. Nick
Fawcett and Mr. Joe Hannon for helping to bounce ideas around and Mr. Mike Whitton
for proof reading the text of my thesis. Also, I am very grateful for the help the BHRG
technicians and students gave me in gathering experimental data - especially Mr. Bob
Burnapp and Mr. Kevin Lee (who also drew some of the figures in this thesis).
Finally, I would like to thank my wife, Rita, for her patience as she became a "PhD
widow" for a large part of 1991 I
as tried to write up the thesis and the rest of my
family for their interest in my work and support over the years.
LIST OF CONTENTS
Page No.
ABSTRACT
ACKNOWLEDGEMENTS
LIST OF CONTENTS
LIST OF FIGURES
LIST OF TABLES
NOMENCLATURE
CHAPTER 1 GENERAL INTRODUCTION 1
1.1 INTRODUCTION 1
1.2 NEWTONIAN AND NON-NEWTONIAN FLUIDS 1
1.3 LAY OUT OF THE THESIS 2
CHAPTER 2 LITERATURE REVIEW 4
2.1 INTRODUCTION 4
2.2 BLENDING IN THE TURBULENT REGIME 5
2.3 BLENDING IN THE TRANSITIONAL AND LAMINAR
REGIMES 14
2.4 NON-NEWTONIAN FLUIDS 20
2.4.1 Determination of the Characteristic Shear Rate 20
2.4.2 Blending of Non-Newtonian Fluids 28
2.4.3 Reviews 32
2.5 DESIGN METHODS AND ECONOMIC EVALUATION 33
2.6 VISCO-ELASTICITY 36
2.7 CONCLUSIONS 38
CHAPTER 3 EQUIPMENT 40
3.1 INTRODUCTION 40
3.2 VESSELS 40
3.3 IMPELLERS 40
3.4 CONDUCTIVITY PROBES 41
3.5 TEST FLUIDS 42
3.5.1 Newtonian Test Fluid 42
3.5.2 Non-Newtonian Fluid 44
3.5.3 Tracer 45
3.6 RHEOLOGY MEASUREMENT 45
3.6.1 Newtonian Fluids 45
3.6.2 Non-Newtonian Fluids 46
CHAPTER 4 BLENDING OF NEWTONIAN FLUIDS 48
4.1 INTRODUCTION 48
4.2 METHOD 48
4.2.1 Instrumentation Chain 48

4.2.2 Blend Time Determination 49
4.3 DATA ANALYSIS 50
4.4 TURBULENT DATA RE-ANALYSIS 52
4.5 ANALYSIS OF VISCOUS BLEND TIME DATA 55
4.5.1 Individual Probe Blend Times 55
4.5.2 Root Mean Square Blend Times 57
4.5.3 Correlation of Data 58
4.6 DISCUSSION 60
4.6.1 Comparison using Power Input 60
4.6.2 Use of the Correlation for Agitator Design 63
4.7 CONCLUSIONS 65
CHAPTER 5 BLENDING OF PSEUDO-PLASTIC FLUIDS 68
5.1 INTRODUCTION 68
5.2 PSEUDO-PLASTIC FLUIDS 68
5.2.1 The Power Law Model 68
5.2.2 Shear Rates 70
5.2.3 Wall Shear Rate Estimation 70
5.2.3.1 The Vessel Wall 72
5.2.3.2 The Vessel Base 73
5.2.3.3 The Baffles 73
5.2.3.4 Combining 74
5.3 ANALYSIS OF PSEUDO-PLASTIC BLEND TIME DATA 76
5.3.1 Individual Probe 'Blend Times 76
5.3.2 Root Mean Square Blend Times 77
5.3.3 Correlation of Data 77
5.3.3.1 Turbulent Data 77
5.3.3.2 Viscous Data 78
5.4 DISCUSSION 79
5.4.1 Controlling Viscosity 79
5.4.2 Comparison using Power Input 80
5.4.3 Measurement of Rheology 81
5.4.4 Data Correlation 83
5.5 CONCLUSIONS 84
CHAPTER 6 POWER MEASUREMENT 88
6.1 INTRODUCTION 88
6.2 METHOD 88
6.3 POWER NUMBER 90
6.4 RESULTS 91
6.5 DISCUSSION 92
6.5.1 Shear Rate and Reynolds Number 92
6.5.2 Comparison with Other Workers 92
6.6 CONCLUSIONS 94
CHAPTER 7 ECONOMIC ANALYSIS 95
7.1 INTRODUCTION 95
7.2 METHOD 95
7.2.1 Capital Cost 96
7.2.2 Running Cost 96
7.2.3 Analysis 98
7.3 RESULTS 100
7.4 DISCUSSION 101
7.4.1 Comparisson of Experimental and Economic Results 101
7.4.2 Power Number 102
7.5 CONCLUSIONS 102
CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS FOR

FUTURE WORK
104
8.1 INTRODUCTION 104
8.2 CONCLUSIONS 104
8.2.1 Newtonian Blend Times 104
8.2.2 Pseudo-plastic Blend Times 106
8.2.3 Economic Analysis 108
8.3 RECOMMENDATIONS FOR FUTURE WORK 108
8.3.1 Vessel Geometry 108
8.3.2 Fluid Physical Properties 109
8.3.2.1 Rheological Properties 109
8.3.2.2 Physical Property Differences 110
8.3.3 Measurement of Flow
REFERENCES 111
TABLES
FIGURES
APPENDIX
LIST OF FIGURES
Figure No. Title
3.1 Vessel Arrangement

3.2 Pitched blade turbine
3.3 Flat blade turbine
3.4 Hydrofoil
3.5 Conductivity Probe
3.6 Conductivity Probe Positions
4.1 Typical Conductivity Response

4.2 Typical Plot of 2 versus Time
4.3 Plot of Individual Probe NO versus Reynolds number
Newtonian data - T/3 Pitched blade turbine
Newtonian data - T/3 Flat blade turbine
Newtonian data - T/2 Lightnin A310 Hydrofoil
4.7 Plot of Root Mean Square NO versus Reynolds number
4.11 Plot of Blend Time Correlation
Newtonian data -. T13 Pitched blade turbine
Newtonian data - All impellers
5.1 Plot of Shear Stress versus Shear Rate for a Pseudo-plastic Fluid
5.2 Vessel Geometry assumed for estimation of the Wall Shear Rate
Pseudo-plastic data - T/3 Pitched blade turbine
Pseudo-plastic data - T/2 Pitched blade turbine
Pseudo-plastic data - T/3 Flat blade turbine
Pseudo-plastic data T/2 Lightnin A310 & Chemineer HE3 Hydrofoils
-
Pseudo-plastic data T/3 Pitched blade turbine
-
Pseudo-plastic data T/2 Pitched blade turbine
-
Pseudo-plastic data T/3 Flat blade turbine
-
Pseudo-plastic data - T/2 Lightnin A310 & Chemineer HE3 Hydrofoils
All data - T/3 Pitched blade turbine
All data - T/3 Flat blade turbine
All data - T/2 Lightnin A310 & Chemineer HE3 Hydrofoils
Newtonian & Pseudo-plastic data - All impellers
5.16 "Max-Blend" Impeller by Sumitomo Heavy Industries
6.1 Plot of power number versus Reynolds number

All data - T/3 Flat blade turbine
All data - T/2 Lightnin A310 & Chemineer HE3 Hydrofoils
7.1 Plot of Costs versus power number -K=2.00, n=1.0

7.7 Plot of Costs versus power number -K= 11.22, n=0.4
LIST OF TABLES
Table No. Title
3.1 Impeller Dimensions

3.2 Conductivity Probe Positions
3.3 Power Law Constant and Index of Non-Newtonian Fluids
4.1 Newtonian Blend Time Results - T/3 Pitched blade turbine

4.2 Newtonian Blend Time Results - T/2 Pitched blade turbine
4.3 Newtonian Blend Time Results - T/3 Flat blade turbine
4.4 Newtonian Blend Time Results - T/2 Lightnin A310 Hydrofoil
4.5 NO versus Reynolds number - Regression Results
4.6 Correlation Regression Results
4.7 Comparison of Constants
5.1 Pseudo-plastic Blend Time Results - T/3 Pitched blade turbine

5.2 Pseudo-plastic Blend Time Results - T/2 Pitched blade turbine
5.3 Pseudo-plastic Blend Time Results - T/3 Flat blade turbine
5.4 Pseudo-plastic Blend Time Results - T/2 Lightnin A310 & Chemineer
HE3 Hydrofoils
5.5 NO versus Reynolds number - Regression Results
5.6 Correlation Regression Results
5.7 Comparison of Constants
6.1 Power Numbers - T/3 Pitched blade turbine

6.2 Power Numbers T/2 Pitched blade turbine
-
6.3 Power Numbers T/3 Flat blade turbine
-
6.4 Power Numbers - T/2 Lightnin A310 & Chemineer HE3 Hydrofoils
7.1 Selling Price (Capital Cost) versus Shaft Torque

(from Rigden (1991))
7.2 Power Law Constants and Indices of Fluids in Economic Analysis
NOMENCLATURE
Symbol Meaning Units
a Constant or Exponent
-
b Number of blades
-
C Concentration kmol m-'
CN Natrosol concentration %wt. /wt.
c Constant
-
c Clearance (= (T - D)/2) m
D Impeller diameter m
F1 Pumping number (= Q/N D3)
-
Fo Fourier number (= p0/ .t T2)
-
Fr Froude number (= N2 D/ g)
-
g Acceleration due to gravity ms
H Fluid depth m
K Power law constant Ns" m"2
k; Constant relating to quantity i -
KP Constant (= Po Re) in the laminar regime
-
M Mixing index -
N Impeller rotational speed s"'
n Power law index -
P Power W
Po Power number (= P/p N3 D5) -
Q Flow rate m3 s"'
Re Reynolds number (= pN D2 / p) -
T Vessel diameter m
t Time s
V Volume m3
v Velocity m s"'
w Blade width m
x Distance m
y Exponent -
Greek Symbol Meaning Units
cx Kinematic fluid property m2 s"'

F Time interval s
Shear rate s-'
Mean power input per unit mass W kg"'
8 Blend time s
K Constant relating shear rate to impeller speed -
A Torque Nm
Dynamic viscosity Ns M2

Density kg m3
P
Standard deviation of quantity i Units of i
6j
Shear stress N m"2
It
Subscript Meaning
a Apparent
c Cavern
crit Critical
FBT Flat blade turbine
Hyd Hydrofoil
large Large scale
RMS Root mean square
small Small scale
y Yield
1,2,3 Probe 1,2,3
Other symbols used in this thesis are defined in the text

CHAPTER 1
GENERAL INTRODUCTION
1.1 INTRODUCTION
Blending is a widely used operation in the chemical and process industries. Its
objective is to take two or more miscible fluids and blend them to a pre-determined
degree of homogeneity. The time taken to reach this degree of homogeneity is called
the "blend time".
The most common processes where blending is used as the design basis are simple
mixing duties where the product is simply the mixture, reactions where the chemicals
are blended in order to promote the reaction and storage vessels where "minimum
motion" is required to prevent stratification or a skin forming on the surface of the
fluid.
There is a great deal of useful information on the blending of low viscosity fluids and
this is rarely considered a difficult duty since blend times are usually short relative to
other times associated with the process, such as the time taken to fill or empty the
vessel. As the fluid viscosity increases the process becomes increasingly difficult and
blend times longer until they limit the rate at which the process can proceed. In these
cases knowledge of the blend time is much more important.
Most of the work done in the past on the blending of viscous fluids has concentrated
on those exhibiting a Newtonian rheology yet the vast majority of viscous fluids
processed in industry are non-Newtonian. Thus, there is very little relevant information
available to engineers who have the responsibility to select agitation equipment for non-
Newtonian blending duties.
1.2 NEWTONIAN AND NON-NEWTONIAN FLUIDS
The rheology of a fluid is determined by measuring the shear stress at different shear
A Newtonian fluid is characterised by a linear relationship
rates using a rheometer.
between shear stress and shear rate and, by definition, a constant viscosity. The
blend time data taken in Newtonian fluids is fairly simple because the
analysis of
is the same in all regions of the vessel, ie. those experiencing high shear, near
viscosity
impeller, and those experiencing low shear, near the fluid surface and vessel wall.
the
2
An agitator rotating in a moderately viscous fluid is usually operating in the transitional

regime (100 < Re < 10). Here, the product of the impeller rotational speed and blend
time (N8) is inversely proportional to Reynolds number which is a property of the fluid
viscosity.
Non-Newtonian fluids are characterised by a non-linear shear stress shear rate

-
relationship and their viscosity is dependent on the shear rate which they experience.
So the viscosity of the fluid will vary according to its position in the vessel and the
type of rheology it exhibits.
This research project is concerned with the blending of shear-thinning, or pseudo-

plastic, fluids. These have a low viscosity at high shear rates and high viscosity at low
shear rates. Therefore, in an agitated vessel, the viscosity will be lowest near the
impeller and highest near the walls and fluid surface. An important question which
must be answered is can the fluid viscosity be characterised in such a way that a
Newtonian type relationship between the blend time and Reynolds number be obtained?
If not, can the effect of non-Newtonian behaviour on blending performance be
quantified in another way?
Other questions which need to be answered for engineers selecting and designing
agitators are: is a certain size and type of impeller best suited for the blending of
pseudo-plastic fluids if
and so, over what range of conditions can it be applied? Also,
the penalties for using non-optimum impeller design need to be investigated.
1.3 LAY OUT OF THE THESIS
The thesis contains seven main chapters and they are briefly introduced below:
Chapter 2 is a literature review, the conclusions of which have identified the "gaps" in
the current knowledge of non-Newtonian blending. This has been used to determine
the experimental objectives of this research.
Chapter 3 describes the equipment, instrumentation and test fluids in

used the project.
Chapter 4 covers the experimental technique, data analysis method and the results of
Newtonian blend time measurements. A new blend time correlation is presented
the
dimensionless groups: power number, Po, Reynolds number, Re,
which contains three
Fourier number, Fo.
and
3
Chapter 5 covers the measurements made in pseudo-plastic fluids. A new method of
correlating the pseudo-plastic blending data is proposed and comparison between
Newtonian and pseudo-plastic data is made.
Chapter 6 covers power measurements and presents plots of power number versus
Reynolds number for each of the impellers tested. Comparison with literature data is
made.
Chapter 7 describes a cost-benefit analysis of the research presented in this thesis.
Chapter 8 is a final general discussion of the results and presentation of the of the
conclusions of the research. Finally, suggestions for future work are presented.
4
CHAPTER 2
LITERATURE REVIEW
2.1 INTRODUCTION
The performance of various impeller types is discussed in the literature review and
detailed information on their geometries can be found in text books such as "Mixing
in the Process Industries" edited by Harnby, Nienow & Edwards (Butterworths, 1985)
or "Fluid Mixing Technology" by Oldshue (M`Graw-Hill, 1983).
The objectives of this chapter are to compare the various experimental techniques used
to carry out blend time measurements in agitated vessels and the results reported in the
open literature. The advantages and disadvantages of the various measurement
techniques are discussed and the results compared. The literature has been divided into
three sections: the first reviews the measurement of blend time in turbulent systems, the
second discusses blend time in viscous, Newtonian systems and the third, viscous, non-
Newtonian systems.
The subject of this thesis is blending of viscous fluids but it is important to consider
turbulent blending because this gives a base-line against which the performance of
agitators in viscous blending can be compared.
Methods of estimating shear rates are compared because the viscosity of a non-
Newtonian fluid is determined by the shear rate it is experiencing. The viscosity of the
fluid will vary according to its position in an agitated vessel and it is important to be
able to identify the viscosity which controls the performance of an agitator in any non-
Newtonian mixing duty.
A brief catalogue of non-Newtonian mixing reviews is followed by a review of agitator

design methods available and a brief review of literature on the subject of visco-
elasticity.
Finally, the work covered in this thesis is put into context with the conclusions drawn
from the information available in the open literature and the identification of "gaps" in
current knowledge.
5
2.2 BLENDING IN THE TURBULENT REGIME
The most widely used blend time measurement technique involves the addition of a
small volume of tracer to an agitated vessel and recording the change which takes place
in the vessel as the tracer is blended into the bulk fluid. The tracer will have a property
which is different from the bulk fluid but does not affect the way in which the two
fluids mix. The property must be identifiable and measurable as it is diluted in the bulk
fluid. This property can be electrolyte concentration, temperature or acidity.
i) Concentration differences are measured using conductivity probes. The

concentration is measured at one or more points in the vessel and the blend time
is defined as the time taken for the tracer to be dispersed to a specified degree
of homogeneity.
ii) Temperature differences are measured using a thermocouple and the same
principle applies to this technique as to the conductivity technique, ie. a slug of
hot fluid is added to the vessel as tracer and the blend time is the time taken for
the temperature differences to be homogenised.
iii) An acid, or alkaline, tracer is added to the bulk fluid in order to cause a colour
change in a chemical indicator. Blend time is determined from the time taken
for the last wisp of colour to disappear as mixing progresses. This technique
also allows poorly mixed regions of the vessel to be identified, since this is
where the last wisp of colour will disappear. One drawback with this technique
is the effect of stoichiometry on the apparent blend time and this has been
discussed by Norwood & Metzner (1960).
The first two methods have the advantage of being objective measurement techniques.
The fluctuations recorded by the conductivity probe or thermocouple can be recorded
using a chart recorder or digitally, on a computer disk, and analyzed after the
experiment has been completed. The colour change technique is subjective and relies
on the experimenter's judgement on when blending is complete.
Care must be taken when using conductivity or temperature differences to ensure that
density of the tracer is not significantly different from the bulk fluid in the vessel
the
because it is possible that the density difference between the tracer and the bulk fluid
the way in which the blending process takes place.

will affect
6
Landau & Prochzka (1961) compared these three blend time measurement techniques
and concluded that the results obtained were mutually consistent. So it is possible to
compare the blend time measurements reported in the literature, provided one of these
three techniques was used and the results are compared with consideration given to the
drawbacks associatedwith each technique.
Blend time experimental data can generally be divided into two groups: in the first, the
dimensionless group, NO, is a constant, dependent on impeller diameter and type, and
independent of Reynolds number and, in the second, the value of NO is dependent on
impeller diameter, type and Reynolds number. This is the main contradiction in the
correlation of the data and it has important implications for scale-up.
Kramers et al. (1953), Prochazka & Landau (1961), Biggs (1963), Khang (1975), Rielly
& Britter (1985), Sano & Usui (1985,1987), Mackinnon (1987), Komori & Murakami
(1988) and Raghav Rao & Joshi (1988) all used the conductivity technique to measure
blend times. The majority of the work reported in these references was carried out in
vessels of up to 0.61 m diameter. Only Mackinnon (1987) worked at large scale with
vessels of 0.61,1.83 and 2.67 m diameter. The methods employed and results are
described below:
Kramers et al. (1953) measured blend times at two scales, 0.31 and 0.64 m with H=
T, using two conductivity probes positioned T/8 above the vessel base and below the
surface of the liquid. A marine propeller, of T/4 diameter, was studied at both scales
and flat and curved blade Rushton turbines, of the same diameter, were studied at the
large scale only. A variety of baffle types were examined. They concluded that NO
is constant, within 10% and for scale-up, with geometrical similarity and constant
impeller tip speed, constant power per unit mass is required for equal blend time. They
also concluded that vortex formation increases blend time and that Froude number
should be included in a mixing time correlation to take this into account.
Prochazka & Landau (1961) applied the conductivity technique to measuring blend
times with a propeller, pitched blade and Rushton turbines. The ratio DIT was varied
between 0.10 and 0.33, which is smaller than industrial standards (where D/T is
0.2 0.7). They their results by:
typically between and correlated
T 2`) Egn(2.1)
N6 =k ...
-D
7
ie, NO is a constant, and its value is dependent on the type and geometry of the
impeller.
The values of k were 3.48,2.02 and 0.91 for the propeller, pitched blade turbine and
Rushton turbine respectively. Similarly, the values of a were 0.05, 0.20 and 0.57.
They also concluded that the correlation of the data was unaffected by the size of the
tracer charge or the position of its addition.
Biggs (1963) used a conductivity technique to measure the time taken for homogeneity
to be obtained in an agitated vessel with continuous flow in and out (known as a
continuous stirred tank reactor or CSTR). The vessel was 0.24 m diameter and
propellers, pitched blade, flat blade and Rushton turbines with DIT varying between
0.26 and 0.42, were studied. Three times were identified: the first was the time for the
conductivity trace to reach 10% of the fully mixed value; the second was the time for
the trace to reach its maximum value and the third, the "terminal mixing time" or blend
time. The Rushton turbine blend times were correlated by:
e=1.55 Egn(2.2)
N'-' D 1.93 ...
Biggs compared the terminal mixing time in the CSTR to batch mixing times measured
by other workers and justified this comparison because the ratio of residence time to
blend time was of the order of 100: 1. The conductivity fluctuations were unaffected
.
by the flow of the fluid through the vessel.
Khang (1975) and Khang & Levenspiel (1976a, 1976b) used residence time distribution
theory and probability to predict the mixing rate of a slug of, tracer in an agitated vessel.
A conductivity technique was used in 0.56 and 1.22 m diameter vessels using two sizes
Rushton turbine and marine propeller. The data were correlated by:
of
(N (D a- Egn(2.3)
k ...
lA 1 /
2.3 for the Rushton turbine and 2.0 for the marine propeller. A is
The value of a was
decay constant, found from analysis of the peaks and troughs in the
an amplitude rate
trace for each experiment. A design method is proposed, showing how the
conductivity
8
data correlation can be used to size agitators for blending duties.
Rielly & Britter (1985) measured mixing times in a vessel 0.29 m diameter, with H=
T, using a Rushton turbine. They investigated the effect of the probe and tracer volume
on the measured blend time. They concluded that a large probe volume measures a
shorter blend time than a probe of small volume but the volume of tracer does not
affect the measurements. They also concluded that NO is not constant in the turbulent
regime and this may be an effect of the probe volume and its resolution. A plot of NO
versus Reynolds number was presented but the mean value of NO and its standard
deviation were not, so it is not possible to determine how their results compare with
other workers.
Sano & Usui (1985,1987) measured blend times, based on the conductivity technique,
and impeller discharge flow, using a flow follower technique, in vessels of 0.2 and 0.4
m diameter. The impellers studied were a Rushton turbine and two-bladed paddle.
They concluded that NO and Fl are constants, independent of Reynolds number and
correlated them with the impeller dimensions. They found that the blend time was
proportional to the circulation time, with the number of circulations required to achieve
homogeneity given by:
D .5bw...
Eqn(2.4)
m=3.8
TT
where m is the number of circulations.
Mackinnon (1987) reviewed the research which had been carried out by the FMP
FMP measured blend time using the same technique at three scales; 0.61,
consortium.
1.83 and 2.87 m. Eleven radial, mixed and axial flow impellers were studied and the
data taken for all impellers at all scales were correlated by:
3/6
D Eqn(2.5)
Pol NO =k...
T
ie. NO is a constant for a given impeller.
Blend times were measured in three regions of differing agitation intensity; close to the
9
impeller, behind a baffle and halfway between the agitator shaft and vessel wall. The
blend time was not found to vary significantly around the vessel.
Komori & Murakami (1988) measured power consumption and blend and circulation
times in a baffled vessel 0.29 m diameter with H=T and H= 2T. The test fluid was
water and the measurements were carried out at Reynolds numbers of 2.5 x 104,4.8 x
104 and 6.2 x 104. The blend times were measured using a conductivity technique and
the flow patterns using a laser-Doppler anemometer. The impellers were single and
dual T/2 flat blade turbines and the spacing between the dual impellers was varied.
They concluded that turbulent blending is strongly dominated by large recirculating
flows in the axial and radial directions. Greater energy efficiency is achieved by
creating larger scale flows and reducing the number of small scale recirculating flows
in the vessel.
Raghav Rao & Joshi (1988) have measured blend times at four scales: 0.3,0.57, 1.0
and 1.5 m diameter and compared their results with a theoretical approach to the
blending process. Rushton turbines and upward and downward pumping pitched blade
turbines were compared and the effect of impeller clearance examined. They concluded
that NO is constant and dependent on the vessel and impeller diameters, fluid depth and
impeller off-bottom clearance. Of all the impellers tested, the most efficient was the
downward pumping pitched blade turbine of T/3 diameter, a conclusion which differs
from every other worker in this area.
The thermal method was used by Havas et al. (1978b) and Shiue & Wong (1984).
Havas et al. (1978b) measured blend times at two vessel scales, 0.40 and 0.53 m., using
standard propellers, proprietary S. E. M. and VEGYTERV propellers and a pitched
blade turbine. They correlated their data, by:
Egn(2.6)
k= Po (N8)' ...
lT
k was plotted against DIT for the VEGYTERV propeller and it was found that the
k
lowest value of occurs when DIT = 0.55. This shows that this diameter will achieve
desired blend time for the lowest power input and is, therefore, the most energy
a
efficient.
10
Shiue & Wong (1984) measured blend times in a 0.4 m diameter vessel using a wide
variety of impeller types and sizes, including the VEGYTERV propeller and using
draught tubes for the two and four blade pitched turbines studied. They compared the
efficiency of the various impellers by plotting the relationship proposed by
Hoogendoorn & Den Hartog (1967) and concluded that the pitched blade turbines in
draught tubes were the most efficient impellers tested and Rushton turbines at H/2
clearance, the least efficient.
The colour change method was used by Fox & Gex (1956), Norwood & Metzner (1960)
and Brennan & Lehrer (1976)
Fox & Gex (1956) measured blend times at three scales, 0.3,1.5 and 4.2 m diameter,
using propellers and jet mixers in unbaffled vessels. The technique used at the small
scale was an acid-base colour change with blend time measured with the agitator
running. At the large scale, measurement of iodine value (a measure of the degree of
saturation) of batches of vegetable oil was used, with the agitator stationary at the start
of an experiment. They correlated their low viscosity propeller data by:
Fr'6H112T Egn(2.7)
NO ...
Re D 3/2
and jet mixer data by an identical relationship, based on the Reynolds and Froude
numbers at the jet nozzle.
Fox & Gex concluded that the difference in measured blend times, due to the different
techniques used at small and large scales, was the order of 25%. The fact that Froude
in the correlation of the data may be due to the fact that blend times
number appears
in unbaffled vessels, as suggested by Kramers et al. (1953), causing a
were measured
vortex to form on the liquid surface.
Norwood & Metzner (1960) measured blend times, with an acid-base colour change and
fluid velocities, using an electric velocity probe, the resistance of which was a function
fluid flowing past it. Three scales were studied ranging from 0.14
of the velocity of the
0.39 The impeller type studied was a Rushton turbine with DIT ranging
to m. only
0.18 0.74 blade widths. Blend times were correlated by:
between and with various
11
Fr1/6 H/2 T3rz
NO Eqn(2.8)
D2 ...
Norwood & Metzner concluded that, although their correlation was of a similar form
to Fox and Gex's, turbine impellers were more efficient for blending than propellers or
jet mixers.
Brennan & Lehrer (1976) measured blend times in a vessel of 0.42 m diameter for
Rushton and flat blade turbines of DIT between 0.15 and 0.24. A colour change
technique was used to determine blend times and the main conclusion of the work was
that NO # constant. They did not give a correlation of the data.
Finally, Van de Vusse (1955a, 1955b) used the Schlieren method to measure blend
times in vessels ranging between 2 and 45 litres volume. The Schlieren method
involves creating two layers of fluid having significant differences in refractive index.
In practise this was achieved by superimposing two layers of fluid with different
densities. Differences in refractive indices of the two fluids were detected by shining
a beam of light through the fluid, forming Schlieren lines on a screen on the opposite
side of the vessel. As mixing proceeds the intensity of the Schlieren lines decreases as
the density difference and refractive indices are homogenised. The results obtained
using this experimental method can be compared with those obtained using other
techniques but there are two main differences which must be noted: the first is that
large density differences are required to achieve clear Schlieren lines and second, the
agitator must be stationary at the..start of an experiment, with blend time measured from
the moment the agitator starts to rotate. Four types of impeller were studied giving
predominantly radial, tangential or axial flow patterns. The data were correlated by:
Tp N2 D20.3 Egn(2.9)
NO a ...
L\p gH
This correlation holds for Re > 103- 104and shows that for a given set of experimental
conditions, NO is a constant.
Van de Vusse (1955b) includes a section on optimum agitator design and the key rules
if
for low viscosity mixing given are to use small vessels possible, or several small
in large vessel; the optimum DIT = 0.4 and HIT = 0.8 and the impeller blade
agitators a
be as large as possible. The blade angle was found to have little effect
width, w, should
12
on blending efficiency.
Van de Vusse found that blend time is proportional to circulation time, calculated from
the pumping capacity of the impeller.
Hiraoka & Ito (1977) re-analyzed the data of Kamiwano et al. (1962) and produced the
following relationship:
0.3s T1.4 Ho.s

NO = 7.2 Egn(2.10)
Dw ...
D-'
This shows that, for a given impeller and vessel geometry, NO is constant.
Several of the blend time studies have included measurements of circulation time (eg.
Komori & Murakami (1988)) and blend time has been found to be proportional to
circulation time. Bryant & Sadeghzadeh (1982) and Middleton (1982) both measured
circulation time using a "radio pill" technique.
Bryant & Sadeghzadeh (1982) measured circulation times in a vessel of 1 m3 volume
using a neutrally buoyant flow-follower with an internal radio transmitter. An aerial

was positioned around the impeller and each time the flow-follower passed through the
impeller its passage was recorded. The mean circulation time was correlated by:
kH Egn(2.11)
t= ...
ND3
and the standard deviation of t by:
H'13 Egn(2.12)
6=k ...
aN D3
Equations (2.11) and (2.12) were combined to give a relationship for blend time:
H " Eqn(2.13)
e=k 0N ...
D3
13
Middleton (1982) has measured circulation times in three scales of vessel between 0.18
and 4.4 m3 volume using the technique proposed by Bryant & Sadeghzadeh (1982).
Rushton turbines of diameters between T/2 and T13 were studied and the effect
of
gassing was also investigated. The mean circulation time was correlated by:
1T...
t=0.5 V0-3 Egn(2.14)
N5
and the standard deviation of the mean by:
1.23 V'. 45
Eqn(2.15)
DN2S ...
Mean circulation times in plant scale vessels were found to be greater than would be
predicted by the ratio of the vessel volume to impeller pumping capacity.
The majority of references in the literature report that NO is constant for a given
impeller geometry. This list of references includes the one collection of data taken at
large scale which used the same experimental technique at all scales (Mackinnon
(1987)). Those that report the data correlated with NO a weak function of Reynolds
number were:
i) taken at one, small scale (Rielly & Britter (1985)).
ii) used the subjective colour change technique (Brennan & Lehrer (1976)).
used the colour change technique at small scale and a different technique
at large scale (Fox & Gex (1956)).
iv) drew conclusions regarding the performance of batch mixing agitators

from measurements made in a CSTR (Biggs (1963)).
When an agitator is to be designed for turbulent blending duty, and blend time is to be
design criterion, the value of NO should be taken as a constant, scaling-up
used as a
from literature or pilot-plant data. This has the advantage of a great deal of supporting
being more conservative than the alternative correlation where NO is
evidence and
14
proportional to Reynolds number raised to the -1/6 power.
2.3 BLENDING IN THE TRANSITIONAL AND LAMINAR REGIMES
The type of impeller used for the blending of viscous fluids usually depends on the
viscosity of the fluid. Low and medium viscosity fluids are blended using "open"
impellers, such as turbines and propellers with D <_T/2, while high viscosity fluids are
blended with "close-clearance", or positive displacement, impellers, such as helical or
anchor impellers.
One of the problems encountered in the measurement of blend times using a

conductivity probe or thermocouple in viscous fluids, is the way in which the boundary
layer around the probe affects the measurement being made (Metzner & Astarita
(1967)). This has not been found to affect blend time measurements in medium
viscosity fluids but measurement of blend time in high viscosity fluids has almost
exclusively been made using acid/alkali colour change or dye addition.
Several of the references discussed in the previous section include measurements of

blend times with viscous fluids. In these cases one of the objectives of the work was
to examine the way in which fluid viscosity affected blending performance and to
identify the Reynolds number at which the turbulent blending regime ends. The results
of these are discussed below:
Van de Vusse (1955a, 1955b) correlated data taken at Re < 250 by:
T
(p_N2 D2 -03
3 Eqn(2.16)
N6 Re ...
D ApgH
There is in Van de Vusse's correlation where no data appear to have been taken,
a gap
for 250 < Re < 104. Other workers have found that NO Re` in this Reynolds number
The design rules suggested for Re < 250 are the same as those for the turbulent
range.
that the maximum impeller diameter should be selected.
regime, except
Fox & Gex (1956) correlated data taken for 100 < Re < 104 by:
Fr '16 H"' T Eqn(2.17)

NO a ...
Re D 3R
15
They also suggested that a different relationship between NO and Reynolds number may
exist for more viscous fluids.
Norwood & Metzner (1960) measured blend times at Reynolds numbers down to 1, but
only presented a correlation for the turbulent data, equation (2.8). Instead they plotted
the turbulent correlation versus Reynolds number and identified five regimes: turbulent;
late, middle and early transitional; and laminar. The boundaries of these regimes
coincide with the changes in gradient of the power number versus Reynolds number
curve for a Rushton turbine. The fluid viscosity started to affect blend time at a
Reynolds number of about 2000 and the laminar regime started at a Reynolds number
of about 5.
Landau et al. (1963) proposed that Reynolds number and Archimedes number should
be considered in the blending of viscous fluids. Archimedes number was proposed to
take account of density differences between the bulk fluid and the added tracer. They
examined the rate of blending at various positions within the vessel but did not present
a correlation for the data. They found that the fluid viscosity starts to affect blend time
at Reynolds numbers of about 10, depending upon the impeller type. They also found
that the blend time was dependent upon the measuring position within the vessel.
Figures are presented showing the agitation intensity in different regions of the vessel
for different impeller types. The conclusion drawn from the work is that impellers
should be fitted in draught tubes to improve flow around the in
vessel viscous blending
processes.
Khang (1975) measured blend times at a minimum Reynolds number of 500, for
Rushton turbines and marine propellers. Like Norwood & Metzner (1960), Khang
plotted the turbulent correlation, equation (2.3), versus Reynolds number for both
impeller types and the boundary between the transitional, and turbulent regimes was
found to be at 2x 103 for the Rushton turbines and 2x 104 for the marine propellers.
No method of taking the fluid viscosity into account in agitator design is given so the
correlations presented can only be applied to blending in the turbulent regime.
Havas et al. (1978a) measured blend times for screw, helical ribbon, gate, anchor and
the same technique as employed for their studies of low
multi-paddle agitators using
blending (Havas et al. (1978b)). The range of Reynolds numbers studied were
viscosity
2- 500. Plots of NO versus Reynolds number showed that, in the laminar region, NO
is a constant for positive displacement impellers, such as helical ribbons and screws.
impellers were compared by plotting the dimensionless
The blending efficiencies of the
16
groups proposed by Hoogendoorn & Den Hartog (1967). The most efficient impeller
was found to be the screw agitator in a draught tube while the least efficient were the
multi-paddle and gate agitators.
Shiue & Wong (1984) measured blend times in viscous fluids for the same range of
impellers as studied at low viscosity. Plots of NO versus Reynolds number were given
for all impellers tested and the fluid viscosity was found to start to influence the blend
time at Reynolds numbers in the region 103 - 104 depending upon the impeller type.
Pitched blade turbines in draught tubes were found to be least affected by viscosity and
were also found to be the most energy efficient agitators when compared with the others
tested using the Hoogendoorn & Den Hartog (1967) technique.
Other workers who have studied the blending performance of open impellers in viscous
fluids are discussed below:
Hoogendoorn & Den Hartog (1967) measured blend times in vessels of 0.24 and 1.8
m diameter using Rushton turbines, multiple pitched blade turbines, helical screws and
ribbons, marine propellers and anchors. The effect of baffling and draught tubes on
blending efficiency was studied. Several important results were found:
For Rushton turbines:
i) Re >4x 103 NO = constant
ii) 170 < Re <4x 103 NO Re-'
iii) Re < 170 NO Re-a a>1
The increase in blend time at low Reynolds numbers was found to be caused by "dead
behind the baffles so blend times were measured in unbaffled vessels. Blend
zones"
but at Re < 150 stable toroidal vortices were formed above
times were generally shorter
below the impeller. Similar results were found for the other open impeller types
and
Installing the propeller and multiple pitched blade turbine in a draught
studied. marine
found to improve blending at low Reynolds numbers.
tube was
The helical screw was found to behave in a similar way to the Rushton turbines, with
decreasing Reynolds number. Installing the impeller in a draught
NO increasing with
in the performance such that, for 0.1 < Re < 100, NO was
tube caused a change
17
constant. Similarly, over the same Reynolds number range, NO was constant for the
helical ribbon mixer.
The anchor agitator was studied at 100 < Re

< 105 and was found to be inefficient
because the fluid rotated at the same speed
as the impeller.
The blending performance of the impellers

was compared by plotting two dimensionless
groups: the first is based on the power required to achieve a certain blend time and is:
P 82
Eqn(2.18)
T3 ..
This number is a constant for positive displacement impellers,

such as helical ribbons
and those in draught tubes, and varies with Reynolds number for open impellers in the
laminar regime.
The second dimensionless group was defined as a Reynolds number from

which the
impeller rotational speed has been eliminated and is in fact a reciprocal Fourier
number,
1/Fo:
p (T/6) T Egn(2.19)
...

Equation (2.18) was plotted versus (2.19) and the most efficient impellers were found
to be those in draught tubes followed by the helical ribbon in the regime where equation
(2.19) has a value less than one. The general conclusion of, the work was that agitation
of viscous fluids requires good top-to-bottom motion with all streamlines passing
through the impeller and sufficient motion to disturb the laminar wall layers.
Zlokarnik (1967) measured blend times with a variety of impellers and produced a set
of design rules regarding which impeller type to use under different conditions of
viscosity and scale. The determining factor was the Fourier number given in equation
(2.19) and the rules are summarised below:
18
i) 1/Fo>104 Marine propeller in a baffled vessel
ii) 2x10'<1/Fo<10' Gate impeller in a baffled vessel
iii) 12.5 < 1/Fo <2x 103 Paddle impeller in a baffled vessel
iv) 0.25<1/Fo<12.5 Paddle impeller in an unbaffled

vessel
v) 1/Fo < 0.25 Close clearance helical ribbon
Zlokarnik's list of impellers shows that the laminar regime exists at 1/Fo < 0.25 since
this is the value below which use of a helical ribbon is recommended. This is in good
agreement with the conclusions of Hoogendoorn & Den Hartog (1967) who found the
laminar regime exist when 1/Fo <1 for all impellers they tested.
Johnson (1967) measured blend times at two scales, 0.10 and 0.15 m, with H=T, using
two sizes of Rushton turbine, a marine propeller and a helical ribbon. Blend times were
determined by adding dye to the fluid in the range 10 < Re < 100. The blend time data
taken for the three open impellers were correlated by NO a Re"13. NO was found to be
almost constant for the helical ribbon over the range of Reynolds numbers studied, with
a slight decrease with decreasing Reynolds number. This decrease was attributed to the
increased drag causing an increase in bulk movement of the fluid.
Other workers have studied the blending performance of close clearance impellers in
high viscosity blending duties and these are described below:
Gray (1963) measured blend times in a vessel 0.23 m diameter by 0.30 m deep with a
helical ribbon and screw, four-bladed flat and curved turbines and a rod and paddle
agitator in the laminar regime. A dye dispersion technique was used, placing a small
quantity of coloured test fluid on the surface of the vessel contents. Blend time was
determined from the time taken to fully disperse the dye after starting the agitator. The
helical ribbon and off-centre helical screw were found to be the most efficient impellers
for blending in this regime. Blending was found to be hindered by the formation of a
"doughnut shaped" stagnant region which formed below some of the impellers and
blending was found to be improved by incorporating methods of reducing its size, such
as placing the screw agitator off-centre.

19
Novak & Rieger (1969,1975,1977 see also Rieger & Novak (1973,1974)) carried out
an extensive study of the power consumption and blending efficiency for several
geometries of helical screw agitator, anchors and turbines using a conductivity technique
in vessels ranging between 0.1 and 1.0 m diameter. For the screw agitators, they found
that NO was constant in the laminar regime (Re < 100), decreased in the transitional
regime and had a lower constant value in the turbulent regime. The most energy
efficient positions for installing screws were centrally in a draught tube or off-centre
with a small gap between the screw and the vessel wall. The optimum diameter was
T/2.
Coyle et al. (1970) measured blend and circulation times for various geometries of
helical ribbon impeller in a 0.46 m diameter vessel, using a dye addition method. NO
was found to be constant and the blend time about three times longer than the
circulation time.
Hall and Godfrey (1971) have also studied blending with helical ribbon impellers and
found that NO is independent of Reynolds number in the laminar regime and that a twin
flight helix is more efficient than a single flight.
Vonken et al. (1964) measured blend and circulation times for an axial flow impeller
in a draught tube and found that NO was constant in the laminar regime and decreased
with increasing Reynolds number. The blend time was proportional to the circulation
time and showed the same dependence on Reynolds number. A stable toroidal vortex
formed at the lower end of the draught tube at 100 < Re < 300 which could be
prevented by increasing or decreasing the impeller speed.
The effect of viscosity on the pumping capacity, Fl, of impellers has been investigated
by Wong & Huang (1988). They measured pumping capacities of radial flow turbines
three hot-wire anemometers in Newtonian fluids up to 50 cP viscosity.

at scales using
They concluded that the pumping capacity is constant for Re >2x 104 and decreases
Reynolds decreases below this value. Tables of Fl and Re were included in

as number
data have been correlated to show that, when Re <2x 104, the
this paper and these
pumping capcity is proportional to Re0'2.
The conclusions that can be drawn from the work reported in the literature on blending
fluids are that there are two blending regimes at Reynolds numbers outside
of viscous
Once Reynolds number decreases below a value of around 103
the turbulent regime.
being dependent on impeller type, NO is inversely proportional
to 104, the exact value
20
to Reynolds number. At Reynolds numbers below about 200 the exponent on Reynolds
number decreases from -1 to around -10 for open impellers. In practise, if high
viscosity blending is required, a close-clearance impeller, such as a helix, should be
used. Alternatively, a pitched blade turbine or propeller can be installed in a draught
tube. The use of a helix or draught tube produces a constant value for NO in the
laminar regime.
2.4 NON-NEWTONIAN FLUIDS
The papers reviewed in the previous two sections have studied the blending of low,
medium and high viscosity Newtonian fluids. The effect of non-Newtonian behaviour
in blending performance has been studied by several workers and these references are
reviewed below.
2.4.1 Determination of the Characteristic Shear Rate
When studying the blending of Newtonian fluids the viscosity of the fluid does not vary
with position in the vessel as it is independent of the shear rate which it is experiencing.
The apparent viscosity of non-Newtonian fluids is dependent on the shear rate and,
therefore, the shear rate must be taken into account in determination of the viscosity of
the fluid.
The work of Metzner & Otto (1957) is the most well-known in this area, although they
were using a procedure first reported by Magnusson (1952), to determine the shear rate
near the impeller in order to predict its power consumption. Many workers have used
this method to determine shear rates produced by other impellers but it has been found
that it only works well for prediction of shear rates near the impeller. It is poor for
in
predicting the characteristic shear rate processes which are limited by flow in regions
of the vessel away from the impeller, eg. heat transfer from coils or the vessel wall into
the fluid. The various methods are described in chronological order below.
Metzner & Otto's (1957) method involves comparison of the power number versus
Reynolds number curve, measured in the laminar regime where power number is
inversely proportional to Reynolds number, for Newtonian and non-Newtonian, pseudo-

fluids. The the constant of proportionality, KP, is measured in the
plastic value of
Newtonian fluid and the relationship between power number and Reynolds number
found. Power numbers are calculated from the measured power consumption in the
fluids. The apparent Reynolds number can then be calculated from:
non-Newtonian
21
K
Re = Egn(2.20)
a Po ...
The apparent viscosity of the non-Newtonian fluid can be found by re-arranging the
definition of Reynolds number:
N D2
Egn(2.21)
Rea ...
The shear rate can then be calculated by assuming that the power law, or Ostwald
model describes the rheology of the fluid:
Egn(2.22)
7= ...
The shear rates calculated in this way are proportional to the impeller speed. This
it
relationship could, was claimed, be applied to any flow conditions. When the power
numbers were plotted against apparent Reynolds number, the Newtonian and non-
Newtonian curves coincided at Re < 10 but the non-Newtonian power numbers were
less than the Newtonian ones for 10 < Re < 200. The two curves coincided at Reynolds
numbers above 200. It was suggested that the laminar regime may extend to higher
Reynolds numbers in pseudo-plastic fluids than in Newtonian fluids.
Metzner & Otto (1957) worked at vessel diameters between 0.15 and 0.56 m with D/T
varying between 0.27 and 0.77 in the laminar regime and 0.18 and 0.5 in the
transitional regime. Reynolds number was varied between. 2 and 270.
Metzner et al. (1961) extended the work reported by Metzner & Otto (1957) to include
impeller types and found that the proportionality between shear rate and impeller
other
held for the other impellers tested. The shear rate was related to impeller speed
speed
by:
Egn(2.23)
y= KN ...
The values of K were 11.5 ( 12%) for the Rushton turbine, 13 ( 15%) for the fan
(a blade turbine) and 10 ( 9%) for single and double pitch marine
turbine pitched
22
propellers. The experimental conditions were essentially the same as those used by
Metzner & Otto (1957), but the Reynolds number range
was extended to a minimum
of 0.1 and maximum of 620.
Good agreement between the Newtonian and non-Newtonian power

number - Reynolds
number curves was obtained for all the impellers tested. All showed the same trend as
the Rushton turbine, reported previously by Metzner & Otto (1957), that the laminar
regime exists at a higher Reynolds number in non-Newtonian fluids than in Newtonian.
Preliminary work with dilatant fluids indicated that a much more complex relationship
between shear rate and impeller speed exists than the direct proportionality found with
pseudo-plastic fluids.
Calderbank & Moo-Young (1959,1961) proposed that the apparent viscosity of the
fluid at the impeller in non-Newtonian fluids can be calculated from a similar
relationship for the apparent viscosity of a fluid in pipe flow, from which the apparent
viscosity of a fluid is:

3n +1" Egn(2.24)
(O1/)(1-n) 4n ...
\Oc"1
This relationship assumes that the value of K is 8, which is lower than the value found
by Metzner et al. (1961) for three different types of impeller.
Rushton turbines, with 0.33 < D/T < 0.67, two-bladed paddles, with 0.33 < DIT < 0.5,
and three and four-bladed marine propellers, with D/T = 0.6 for the three-bladed
propeller and 0.47 for the four-bladed propeller, were studied in a baffled vessel of 0.25
m diameter. Reynolds number was varied between 0.2, and 1000 and fluids with
pseudo-plastic, Bingham plastic and dilatant rheologies were examined. Equation (2.23)
was found to hold for fluids with pseudo- and Bingham plastic rheologies with K= 10,
while the shear rate in dilatant fluids was correlated by:
(D . Egn(2.25)
y= 38 N ...
lT
Pseudo- and Bingham plastic fluids showed the same trend as observed by Metzner et
(1961), for 10 < Re < 250, that the power number in a non-Newtonian fluid is less
al.
23
than the power number in a Newtonian fluid.
The "pipe flow" Reynolds number, calculated using the viscosity given in equation
(2.24), was plotted versus apparent Reynolds number and the resulting relationship, for
pseudo- and Bingham plastic fluids, can be re-arranged to give:
K [3n1 "
U 0.8 Egn(2.26)
4n ...
(8N)c-n
No relationship could be found for the dilatant fluid data using this method of analysis.
It appears that the Reynolds number determined in this way, calculated using a viscosity
based on a shear rate equal to 8N, over-estimates the measured apparent viscosity by
20%, leading to a constant of proportionality of 0.8.
Beckner & Smith (1966) measured power consumption of various anchor impellers in
a 0.23 m diameter vessel with Reynolds number varied between 0.2 and 100. The
effects of fluid depth, clearance (distance between impeller blade and vessel wall) and
breaker bars, connecting the anchor blade to the shaft, were examined and the data were
correlated by:
-0.93
T '4 p N2-n D2 Eqn(2.27)
Po = 82 ...
cK [a(. 1 - n)]rl-l
where c is the gap between the impeller blade and vessel wall and a is a constant
dependent on the geometry of the impeller. It follows that the shear rate is proportional
to the impeller speed, dependent on the impeller geometry, and given by:
Egn(2.28)
y =a(1 -n)N ...
The constant, K, is dependent on the impeller geometry, which was not the case for
by Metzner & Otto (1957) and Calderbank & Moo-
turbines and propellers as reported
Young (1959,1961) and the degree of non-Newtonian behaviour exhibited by the fluid,
law flow index, n, is included in the equation. The shear rate is
since the power
directly proportional to the impeller speed, for a given impeller geometry and fluid.
Equation (2.27) shows that Po Re-0-93,

yet it is generally accepted that, in the laminar
24
regime, Po a Re"'. It is not clear why the lower exponent was obtained.
Pollard & Kantyka (1969) measured jacket and coil heat transfer coefficients and power
consumption in anchor agitated vessels of 0.30.0.60 and 1.14 m diameter for
Newtonian and non-Newtonian fluids. Reynolds number was varied between 5 and 106
for power measurements and 500 and 106 for the heat transfer measurements. They
concluded that, for Re < 50, the average shear rate is proportional to the impeller speed
but at higher Reynolds numbers:
aN3.75 Egn(2.29)
av ...
Nagata et al. (1971) measured power consumption for a helical ribbon, anchor, flat
blade and Rushton turbine in pseudo- and Bingham plastic fluids over a Reynolds
number range between 0.2 and 1000. They concluded that, in the laminar regime, the
shear rate was proportional to the impeller speed with the constant, K, dependent on the
impeller type for all the impellers tested. The values of K were in good agreement with
those reported by other workers. Good correlation between power number and
Reynolds number, for Newtonian and non-Newtonian fluids, was obtained for ribbon
and anchor impellers in the transitional regime using this method of shear rate
estimation but it did not work for the turbine impellers tested. No alternative was
proposed.
Bourne et al. (1981) measured heat transfer coefficients and confirmed the findings
reported previously by Pollard. & Kantyka (1969), that the apparent shear rate is
proportional to the impeller speed only in the laminar regime and derived an equation
for the shear rate as a function of the impeller speed and diameter, the flow regime and
the flow index of the fluid:
r 2(1- ),
3-2mnan1
Egn(2.30)
r
...
is the log power number versus log Reynolds number plot for
where -m the gradient of
the impeller.
Hcker et al. (1981) the "drag reducing" effects found when dilute solutions
examined
Power consumption was measured in flat-based,
of polymers are pumped or agitated.
0.40 diameter using Newtonian and non-Newtonian fluids. Single
baffled vessels of m
25
and dual Rushton turbines and dual Ekato Mig impellers were studied with DIT ranging
between 0.33 and 0.7 at Reynolds numbers between 0.1 and 106.
In order to calculate the apparent Reynolds numbers of the impellers in non-Newtonian

fluids a shear rate was derived from:
1
7
derived I
N2 D3 KK
P
...
Egn(2.31)
The derived shear rate was plotted against impeller speed and a linear relationship found
with a gradient of about 11 for all the impellers tested. This is in good agreement with
the value found by Metzner & Otto (1957).
Wichterle et al. (1984) measured shear rates on the blades of a Rushton turbine using
an electrochemical technique in a baffled vessel 0.30 m diameter with one impeller of
DIT = 0.33. Newtonian and non-Newtonian fluids were studied with Reynolds number
varied between 1 and 104. The data were correlated by:
(R) ' Egn(2.32)

y (1 + 5.3n) Remy' N ...
,_
where:
N2_" D2 Egn(2.33)
Re =p ...
mK
It was proposed that this relationship could be used to estimate the shear rate on the
blades of other impeller types.
Wichterle et al. (1985) used the same technique to study the distribution of shear
the walls of agitated vessels. A vessel of 0.30 m diameter was studied with
stresses on
Rushton turbines of 0.15 < DIT < 0.5 with further measurements made in 0.14 and 1.0
diameter The measured shear stress was converted to a dimensionless

m vessels.
friction factor, f, and the maximum occurring value was correlated by:
26
T 2(1n))
f -1 Egn(2.34)
max
(1 +15n) Rem ...
D
The shear rate in the fluid at the wall of the vessel was found to be approximately 30%
of the shear rate, measured previously, at the impeller blades and was five times higher
than the shear rate estimated from the method reported by Bourne et al. (1981).
Tebel & Zehner (1985) studied the mixing of pseudo-plastic fluids in loop reactors and
defined a shear rate representative of the mixing process, as opposed to a Metzner &
Otto (1957) shear rate which is used in power correlations. They applied this analysis
to blend time data from the literature and obtained a very good correlation.
Shamlou & Edwards (1985) measured power consumption of helical impellers in

Newtonian and non-Newtonian fluids at 0.15 and 0.40 m scale over a Reynolds number
range between 0.1 and 105. They found that the shear rate constant, K, could be
correlated by:
34- 144 Egn(2.35)

...
D
This is similar to the relationship reported by Beckner & Smith (1966). They also
concluded that as the clearance between the impeller and the vessel wall increases the
value of the constant, K, would decrease until it reaches a limiting value of 12, close
to that reported by Metzner & Otto (1957) and others.
Wang & Yu (1989) measured heat transfer coefficients achieved with a variety of
impellers over a Reynolds number range between 7 and 1.84 x 104. They showed, from
be estimated from:
a theoretical approach, that the shear rate can
f(2-^)) Egn(2.36)
'Y =
kt (1) (2
N1 ...
where:
1-m D2 N(2-") p Eqn(2.37)

f= exp ...
K
27
The values of k and m were found to be 0.4 and 0.00705 respectively.
Elson (1988,1990) studied the growth of "caverns" in fluids possessing a yield stress
as they are agitated and found that the cavern diameter could be correlated by:
Dc Po Rey
DH1 Egn(2.38)
...
+ T2
D3C
where Dc is the cavern diameter, He is the cavern height and Rey is the yield stress
Reynolds number given by:
N2 D2...
Rey =p Egn(2.39)
T
y
where by is the yield stress of the fluid.
Elson suggests that, in pseudo-plastic fluids, a "pseudo-cavern" exists with the impeller
effectively surrounded by slow moving fluid. Substituting the shear stress for the yield
stress in equations (2.38) and (2.39) and re-arranging, the shear rate on the boundary
of the well-mixed and slow moving regions can be found.
It is clear from the information available in the literature that the method of shear rate
determination proposed by Metzner & Otto (1957) works well for predicting the shear
rate and apparent viscosity of a non-Newtonian fluid in the region of an agitated vessel
near the impeller, provided it is operating in the laminar regime. Nagata et al. (1971)
have found that it does not work for open impellers in the transitional regime and
Pollard & Kantyka (1969), Bourne et al. (1981) and Wang & Yu (1989) have found
that it does not work for determining a characteristic shear rate in an agitated vessel in
which heat transfer rates are to be predicted.
Wang & Yu have shown that it is important to be able to estimate the fluid viscosity
in the region of the vessel which controls the rate at which the process under
investigation takes place. In their work the region in is
question the wall of the vessel.
28
The same approach can be applied to blending provided the region in which the
blending rate is controlled can be identified and the viscosity of the fluid estimated.
2.4.2 Blending of Non-Newtonian Fluids
Several references report the results of blending experiments with non-Newtonian fluids.
Some do not produce a correlation for prediction of blend time but a method for
determining the minimum Reynolds number at which an impeller will operate
efficiently. The references are described below, in chronological order.
Metzner et al. (1961) compared the blending of viscous Newtonian and non-Newtonian
fluids using single and dual Rushton turbines, fan turbines (pitched blade turbines) and
marine propellers. The vessel diameter was varied between 0.14 and 0.36 m with DIT
varied between 0.28 and 0.98. Reynolds number was varied between 0.1 and 620.
The minimum Reynolds number, required for movement on the fluid surface, was
identified for each impeller type at the different DIT ratios studied. Larger impellers
could operate satisfactorily at lower Reynolds numbers than small ones. Of the single
impellers, the fan turbine was found to be most effective but the dual Rushton turbines
were most efficient of all when compared in this way. When the performance of the
impellers was compared in terms of their power input there was little advantage to be
gained by using multiple impellers of D/T > 0.5 but there was a distinct advantage in
using dual impellers when D/T < 0.5. The use of dual impellers would position one
closer to the fluid surface than. a single impeller, so it is not surprising that the dual
impeller agitator was found to be the most efficient when judged in this way.
A design method was proposed allowing the impeller size and speed to be determined
in order to achieve a desired "mixing rate".
Godleski & Smith (1962) measured blend times, using a colour change technique, in
baffled vessels of 0.14,0.29 and 0.44 m diameter with T/3 Rushton turbines at each
Blend times were found to be 10 to 50 times longer than predicted by the
scale.
Norwood & Metzner (1957). Calculation of the apparent viscosity
correlation of
that the data points had been taken under laminar conditions, Re < 10, although
revealed
the flow indicated that the flow appeared to be middle transitional, as
observation of
defined by Norwood & Metzner. The blend time data appear to have been taken at
Reynolds numbers below the minimum values required for motion on the fluid surface,
defined by Metzner et at. (1961). This may explain why Godleski & Smith found
as
29
that blend times were an order of magnitude greater than predicted by Norwood &
Metzner's correlation. Also, the Norwood & Metzner approach to blend time data
analysis does not take account of any pseudo-plastic behaviour a fluid may exhibit.
Other workers (eg. Moo-Young et al. (1972) discussed below) have found that pseudo-
plastic behaviour increases blend time when compared with blend times measured in
Newtonian fluids.
Moo-Young et al. (1972) measured blend times and power consumption in a vessel of
0.29 m diameter using a T/3 Rushton turbine, 2T/3 helical ribbon and several novel
tubular agitators. Reynolds number was varied between 5 and i0 calculated from the
,
definition of Calderbank & Moo-Young (1959,1961).
The Newtonian and non-Newtonian power data were found to correlate well with this
Reynolds number, as reported previously, but the non-Newtonian blend times were
longer than those measured in Newtonian fluids at the same Reynolds number. This
was said to be due to eddies decaying more rapidly in pseudo-plastic fluids causing
blend times to increase above the value expected in a Newtonian fluid, at the same
Reynolds number. The fact that the Reynolds number used to correlate the data was
originally developed for correlation of power data, and is not suitable for correlation of
blend time data, is a more likely explanation.
Demange et al. (1977) measured blend times in unbaffled vessels of 0.15,0.20 and 0.28
m diameter with Rushton turbines of 0.064,0.076 and 0.178 m diameter. Seven
pseudo-plastic fluids were tested, selected for their rheological properties and industrial
relevance.
The data were correlated by:
(T/D) (H/D)` Eqn(2.40)

NO =k ...
Fr1Re-"
in
The values of the constant and exponents equation (2.40) were found to be functions
the law index, n, of each fluid. The exact form of this function was a fourth
of power
a table of the constants was given. In the case when n=1 the
order polynomial and
found to be identical to those in the correlation of Fox & Gex (1956) for
values were
Newtonian fluids. The correlation predicted mixing times to within 11 %.
viscous
30
Wichterle & Wein (1981) examined the effect of non-Newtonian behaviour on the
minimum degree of agitation required to create motion throughout the whole vessel by
studying the size of the "nucleus" of fluid in motion around the impeller in relation to
the size of the vessel. This is a similar approach to that taken by Elson (1988) in the
study of the motion of fluids exhibiting a yield stress.
The dimensions of the nucleus created by a Rushton and pitched blade turbine were
determined by moving the impeller towards the fluid surface or the base of the vessel
and recording the conditions under which the fluid moved.
Two values of Reynolds number were defined: the first is the Reynolds number at
which distinctive fluid motion appears and the second, the Reynolds number at which
all dead spaces disappear in the fluid, Re,,,,,,. The best design criterion was found to be
the second definition and this was given by:
1.8 T Eqn(2.41)
Re = ...
ww aD
where a=0.6 for a propeller and = 0.3 for a Rushton turbine.
Having calculated Re,,,,,,the impeller speed is found by equating this value with the
definition of Reynolds number as defined by equation (2.32). It was suggested that this
design method could be applied to other impellers with the value of the constant, a,
calculated from:
3 Egn(2.42)
0.375 PV
o ...
a
where Po is the impeller power number under turbulent conditions.
(1982) studies to develop scale-up rules for

Hemrajani et al. carried out pilot-plant
heat in in which a solid material is added to a high
blending and transfer remelters
fluid. The are melted in these vessels prior to further
viscosity, non-Newtonian solids
The design based on good surface motion, to incorporate the
processing. agitator was
into liquid, and adequate heat transfer for remelting.
solids the
in 0.3 m diameter vessel with single and dual T/3 pitched

Studies were carried out a
in the vessel with four baffles with widths of
blade turbines at a variety of positions
A fluid was used for the tests. The best impeller
0.083T and 0.02T. model
31
combination, on the basis of power input, was equally spaced dual impellers with the
smaller baffles.
Nouri & Whitelaw (1990) studied the flow characteristics, using laser-Doppler
anemometry, of a Rushton and 60 pitched blade turbine, both of T/3 diameter, in a
vessel of 0.14 m diameter, in water, a viscous Newtonian and a viscous non-Newtonian
fluid. The discharge from the Rushton turbine was predominantly radial at all Reynolds
numbers studied while the discharge from the pitched blade turbine had an axial and
radial component, with the radial component increasing with decreasing Reynolds
number. This trend was observed for the Newtonian and non-Newtonian fluids. The
axial components, of the discharge from the pitched blade turbines, were found to
decrease dramatically at a Reynolds number of 650 in the Newtonian fluid and 4.8 x
103 in the non-Newtonian fluid. The difference in Reynolds numbers was attributed to
the drag-reducing effect of the non-Newtonian fluid and the definition of shear rate used
to calculate the apparent viscosity of the fluid and, hence, Reynolds number.
Koutsakos et al. (1990) also used laser-Doppler anemometry to study the effect of non-
Newtonian properties on the discharge from a T/3 Rushton turbine in a vessel of 0.15
m diameter. Three regions were identified:
i) Re > 104 Fl = 0.78
ii) 60 <Re < 104 Fl a Re0.2
iii) Re < 60 Fl a Re
The discharge mean velocity profiles were weakly dependent on Reynolds number when
Re > 60. The fluctuating velocities increase with increasing Reynolds number until the
flow becomes fully turbulent, leading to the conclusion that viscosity and pseudo-
Overall motion in the impeller region of the vessel was
plasticity reduce turbulence.
found to be proportional to the mean centre-line velocity of the discharge flow.
The exponent on Reynolds number in the region 60 < Re < 104 is in good agreement
by Wong & Huang (1988) who used hot-wire anemometry to
with that reported
of several radial flow turbines in viscous Newtonian
measure the purring capacity
fluids.
blend times to be predicted for Newtonian and non-

The only reference which allows
32
Newtonian data is that developed by Demange et al. (1977). Moo-Young et al. (1972)
measured blend times but could not correlate their data because they used a definition
of Reynolds number which was originally developed for correlation of power data.
Metzner et al. (1961) and Wichterle & Wein (1981) have found the minimum Reynolds
number at which open impellers can operate efficiently in viscous fluids and, generally,
this minimum occurs at a Reynolds number of about 200. This agrees well with the
findings of Hoogendoorn & Den Hartog (1967) who studied Newtonian fluids. It also
agrees, indirectly, with the findings of Godleski & Smith (1962) who found that the
blend times they had measured were between 10 and 50 times longer than would be
predicted by the correlation of Norwood & Metzner (1960). It appears that Godleski
& Smith (1962) measured blend times at Reynolds numbers below 200.
2.4.3 Reviews
Many reviews have been written on the problems of mixing non-Newtonian fluids and
a few of these are briefly mentioned below.
Bourne (1964) wrote a survey of industrial mixing problems, with the co-operation of
about fifty companies producing a wide range of chemicals. Several recommendations
were made regarding the knowledge required to improve agitator design.
Su & Holland (1968a, 1968b) wrote two papers: the first was concerned with defining
the types of non-Newtonian fluids which could be encountered in industry and a
description of the factors which should be taken into account in agitator selection and
design. The second gave a comprehensive review of the data that had been published
in the literature on non-Newtonian mixing.
Ford et al. (1972) wrote a review of the experimental techniques used in blend time
and summarised many correlations from the literature.

measurements
Blasinski & Rzyski (1976,1978) wrote two papers: the first reviewed power
impellers in non-Newtonian fluids and the second, power
measurements made with open
impellers.
measurements with close-clearance
Short & Etchells (1982) reviewed unsolved problems they had experienced as consultant
large company. They highlighted the fact that, often blend
engineers within a chemical
33
time is not an important design criterion. Maintaining a minimum degree of agitation

to avoid stagnation is often more desirable. Rheological properties of interest were
pseudo-plasticity, fluids exhibiting a yield stress and visco-elasticity. Although blend
time is not an important design criterion, it can be used, as it has in this research
project, to compare the blending efficiency of various impeller types under different
operating conditions.
Silvester (1984a, 1984b) reviewed mixing of non-Newtonian fluids giving industrial

examples. Topics covered included power measurement, blend times, heat transfer,
flow patterns and aerated systems.
Nienow & Elson (1988) reviewed their work, with other co-workers, in the mixing of
non-Newtonian fluids and covered many topics including power consumption, flow
patterns, scale-up and blend times.
2.5 DESIGN METHODS AND ECONOMIC EVALUATION
Reavell (1951) wrote a review of the types of mixing problems experienced in industry
and discussed the advantages of using various types of impellers as solutions to these
problems.
The most well-known agitator design method, available in the literature, are the "Liquid
Agitation" (Chemineer (1976)) series of twelve articles, originally published in
"Chemical Engineering" between December 1975 and December 1976, and later as a
in booklet, also known as the "Chemineer Articles". Design methods were
collection a
for blending, solids suspension and gas dispersion with advice on mechanical
presented
design, shaft sealing, cost estimation and scale-up.
The design method for fluid blending is based on a "bulk fluid velocity" which
determines the agitation intensity. The agitation intensity is given a "ChemScale"
between 1 10, one point on the scale equivalent to a bulk fluid

number and with
of 0.03 m s"' (6 ft min-'). Having decided on the scale of agitation required
velocity
knowing of the vessel, the pumping capacity of the impeller
and the cross-sectional area
that can be calculated and the impeller size and speed
required to achieve velocity,
If flow is turbulent an iterative process is required to do the design.
calculated. the not
written for a Texas Instruments TI-59
Dickey (1984) published a program,
that performed this calculation. No way of taking non-
programmable calculator,
into in the agitator design was included in the procedures.
Newtonian behaviour account
34
Mersmann et al. (1976) reviewed the data on blending, solids-suspension, gas-dispersion

and heat transfer available in the literature to produce design guidelines for each
process. They concluded that, in the turbulent regime, the product NO was constant and
could be estimated from:
6.7 Tr
NO Egn(2.43)
Po ' D ...
In the laminar regime, use of a helical ribbon is recommended with NO having a

constant value of 55. No advice is given on blending in the transitional regime or on
the effect of non-Newtonian behaviour on blending.
Flagg et al. (1977) compared the blending and heat transfer characteristics of open and
close-clearance impellers including an examination of the economics of agitator
selection and design. They concluded that, at the extremes of Reynolds number
encountered, ie. turbulent or laminar flow, the selection of impeller type is
straightforward. In transitional flow the choice of open or close-clearance impeller is
not so easy and other factors such as initial capital cost, available power supply or the
need for a "self-cleaning" impeller need to be taken into account.
Cheng & Tookey (1977) described a series of pilot-plant experiments carried out to
study the effect of agitator geometry and speed on the yield of a herbicide. The yield
5
obtained at plant scale was to.. 10% less than the yield achieved in the laboratory and
the difference was attributed to lack of agitation at the plant scale. The fluid rheology
changed during the reaction from low viscosity Newtonian at the start, to highly viscous
non-Newtonian with a "porridge-like" consistency at the end. It was found that this
property effected the flow of the fluid when agitated by an.anchor impeller, which had
been used in the plant. A replacement impeller was tested and installed at the plant
improvements in product yield. They concluded that detailed
with subsequent
knowledge of the process chemistry and thermodynamics is not required to determine
design. This is true in some cases, but there are chemical reactions
optimum agitator
knowledge of these parameters combined with knowledge of the mixing is
where
design.
required to achieve an efficient reactor
Corpstein et al. (1979) reviewed the factors which must be considered in pilot-plant
Advice is given on the type of impeller to use at given scale and fluid viscosity
work.
on the interpretation of results. The relationship between the small
combination and
35
and large scale impeller speeds, with geometrical similarity, is:
aN_D
small Egn(2.44)
large small D ...
large
The value of the exponent, a, establishes the

resulting scale-up rule as follows:
a=1 equal fluid velocity
a=0.75 equal solids suspension uniformity
a=0.67 equal overall mass transfer coefficient
a=0 equal blend time
Since scaling-up can sometimes lead to impractically large agitator specification

Corpstein et al. recommended the use of ChemScale for interpretation of the
pilot-plat
data. Curves of energy per unit mass versus Reynolds number are given for
various
ChemScale levels of agitation. The energy per unit mass term can be rearranged to
show that it is equivalent to torque per unit volume divided by the fluid density.
Advice is given on the selection and control of pilot-plant equipment, measurement and
analysis of data and the economic considerations which must be given to the scale of
operation.
Garrison (1983) states that the design of agitators for processing Newtonian fluids
where the scale-up rules are known can be done using the Chemineer Articles
(Chemineer (1976)). A review is given of the factors which must be considered when
pilot-plant work is needed. The rules given are applicable to Newtonian fluids but their
application to non-Newtonian fluids is discussed.
Bowen (1985) produced a design method for the scaling-up of systems where the fluid
viscosity is the controlling variable. This method is based heavily on the Chernineer
Articles (Chemineer (1976)) but includes refinements, such as taking the fluid depth
into account in the calculations. It is shown that scale-up, on the basis of equal bulk
fluid velocity, can be performed by maintaining constant torque per unit volume
indicating that any size of impeller can be to
used achieve a desired bulk fluid velocity.
36
In a second paper Bowen (1986) produced a design method for shear-sensitive systems.
The shear rate is calculated from the mean velocity profile of a Rushton turbine:
)). 3
`
kND Egn(2.45)
l ...

with k=4.2 for calculation of the average shear rate and 9.7 for the maximum.
Equation (2.45) is based on the velocity profile measurements of other workers.

Fawcett (1990) has shown that the Reynolds, or turbulent, stressesin water are several
orders of magnitude greater than the viscous stresses, which are created by the mean
flow from the impeller. Using Bowen's technique to estimate the shear rates in a low
viscosity fluid will give a value several orders of magnitude lower than the true value
because the turbulent stresses are not considered.
2.6 VISCO-ELASTICITY
Visco-elasticity is an important rheological effect and is recognised by phenomena such

as "shaft climbing", fluid recoil or relaxation and flow reversal. It is found in
polymerisation processes where small molecules react to form a long chain.
This thesis is concerned with the effect of pseudo-plasticity on the blending

performance of impellers but a brief review of some of the literature available on visco-
elasticity is given below.
Ulbrecht (1974) has reviewed the literature on mixing of visco-elastic fluids and
reviewed (1975) more practical aspects-of mixing these fluids and the influence of
rheology on product quality.
Chavan et al. (1975) extended the work of Coyle et al. (1970), studying the influence
helical ribbon geometry on blending performance, to visco-elastic fluids. This was
of
done by following the path of a small, neutrally buoyant particle through the vessel.
They found that the axial component of velocity was suppressed in favour of the
with recirculating vortices into which the particle was drawn. They
angular velocity
the consequence of this flow pattern could have a detrimental effect on
concluded that
the molecular weight distribution in a polymerisation reaction.
37
Kelkar et al. (1972) measured the power required to rotate spheres, discs and impellers
in fluids exhibiting pure visco-elasticity, pure pseudo-plasticity and a combination of
the two. They concluded that, when the two rheological properties were present, the
viscous effects were much greater than the elastic effects and, as a result, the power
consumption of the impellers could be calculated by considering the viscous drag effects
alone.
Ide & White (1974) measured the rheological properties of styrene-polystyrene solutions
at concentrations between 15 and 57% and observed the flow patterns in a vessel
agitated by a sphere, disc, turbine and propeller. They found that at low concentrations
the flow patterns were as expected, flow reversal taking place at high concentrations
and segregated flow taking place at intermediate concentrations. They suggested that
the presence of segregated regions within the reactor vessel could lead to distorted
molecular weight distributions of the polymer product.
Ford & Ulbrecht (1976) measured the influence of rheological properties of polymer
solutions on blend and circulation times in a vessel agitated by a helical screw impeller.
They found that blend time was strongly dependent on the visco-elasticity of the fluid
while circulation times were dependent on both the visco-elasticity and pseudo-
plasticity. They concluded that, when scaling-up, velocity gradients do not increase by
the scale factor and in certain circumstances may reduce. This reduces the normal
in
stresses the fluid which, in turn, reduces the secondary flow effects caused by visco-
elasticity.
Carreau et al. (1976) measured blend and circulation times in vessels agitated by helical
ribbons. They found that blend times in Newtonian fluids were 3 to 7 times shorter
than those measured in visco-elastic fluids and concluded that blend time is highly
dependent on the fluid's degree of visco-elasticity. They found that the axial velocity
by
component was suppressed visco-elasticity and the angular velocity increased. Blend
time data were correlated by plotting dimensionless blend time versus a dimensionless
factor, which is a function of the geometrical properties of the impeller.
pumping
Boger (1980) has reviewed the development of non pseudo-plastic, elastic fluids and
their use in isolating elastic effects in rheological experiments.
Greene et al. (1982) used streak photography to investigate the change in flow
pitched blade turbine due to increasing visco-elasticity. They found

characteristics of a
degree of visco-elasticity increased, the axial component of flow decreased
that, as the
38
and the radial component increased, until the flow pattern resembled that of a radial
flow impeller.
Prtner et al. (1988) used an optical technique to measure blend times in pseudo-plastic
and visco-elastic fluids. Data were correlated by:
Re
Ne Egn(2.46)
1+kWi ...
where Wi is the Weissenberg number, which is a dimensionless group relating the

inertial and elastic forces in the fluid. Visco-elasticity was found to increase blend
time.
Brito-De la Fuente et al. (1990) measured power consumption, blend and circulation
times in Newtonian and visco-elastic fluids. They found that visco-elasticity affected
both power and blend time and fitted their data to a correlation which included
Wiessenberg number.
2.7 CONCLUSIONS
The objective of this section of the literature review is to summarise the work reported
in the preceding sections and to put the work reported in this thesis in context.
A great deal of work has been done on the blending of low viscosity fluids in the
turbulent regime (Re > 104). The majority of workers agree that NO is constant and its
value is dependent on the impeller type and diameter.
As Reynolds number decreases a value is reached where the impeller enters the
transitional regime. This occurs at 103 < Re < 104 with the exact value dependent on
impeller type. Only Hoogendoorn & Den Hartog (1967) have measured blend times at
large scale using consistent experimental techniques and they found that, in the
NO is inversely proportional to Reynolds number. Fox & Gex
transitional regime,
(1956) also measured blend times at large scale but used a different technique to their
work at small scale.
As Reynolds number decreases further the impeller enters the laminar regime where,
for open impellers, NO is inversely proportional to Reynolds number raised to the power
39
-10. All workers who have measured blend times in this regime agree that close
clearance or axial flow impellers in draught tubes should be used to achieve efficient
blending.
No large scale work has been carried out to measure non-Newtonian blend times and
only one correlation exists which can be used to predict blend times in non-Newtonian
fluids (Demange et al. (1977)). Also, no consistent method has been proposed for
determination of shear rate and viscosities in agitated vessels in which the blend time
must be predicted. For these reasons the objectives of the work reported in this thesis
are:
i) to measure blend times in Newtonian fluids at three scales using open impellers
which are common in industry, to correlate the data and to give rules for their
design. The range of Reynolds numbers will be between 200 and 5x 10', since
there is a great deal of evidence that such impellers cannot operate efficiently
at lower Reynolds numbers.
ii) to measure blend times in non-Newtonian fluids at three scales using the same
impellers as tested in Newtonian fluids, to correlate the data and to give rules
for their design. The same range of Reynolds numbers will be used.
iii) to determine in in
the characteristic shear rate an agitated vessel which blending
takes place in order to estimate a viscosity which can be used to correlate the
blend time data.
iv) to use the correlation to explain the effect of non-Newtonian behaviour on the
blending process in agitated vessels.
impeller best suited to the blending of fluids in the

v) to determine which types are
transitional regime. The criteria used will be capital, running and total cost of
the agitation equipment.

40
CHAPTER 3
EQUIPMENT
3.1 INTRODUCTION
The equipment used for the experimental programme reported in this thesis was
developed prior to the start of the project and had been used for the measurement of
power consumption and blend times in water (Mackinnnon (1987)). New equipment
was acquired and tested for the measurement of rheology but the basic experimental
equipment and technique was the same. The equipment and test fluids are described
in this chapter.
3.2 VESSELS
Three geometrically similar vessels were used for the tests with diameters of 0.305,
0.609 and 1.830 m. The vessels had a dished base with a radius of curvature equal to
the vessel diameter and were fitted with four baffles, equally spaced around the
circumference of the vessel. The baffles were T/12 wide and mounted vertically at a
spacing of T/60 off the vessel wall and extended from above the fluid surface to the
tan-line of the vessel.
The 0.305 and 0.609 m vessels were constructed of perspex and allowed observation
of the blending process to be made. The 1.830 m vessel was constructed of steel and
was fitted with four viewing windows. In all cases the vessels were filled so that the
fluid depth was equal to the vessel diameter.
The vessel arrangement is shown in figure 3.1.
3.3 IMPELLERS
Much of the research carried out into blending of miscible fluids has concentrated on
Rushton turbines and marine propellers. These two impellers are
the performance of
in industry today for blending duties. Current industrial practices were
not widely used
discussed with members of the FMP consortium, prior to the start of the experimental
was made to ensure that "industrially relevant" equipment
programme and every effort
was tested.
41
Pitched and flat blade turbines are used because they are relatively cheap to
manufacture, consisting of rectangular blades bolted or welded onto a hub. The other
impeller type which is used is the hydrofoil. These have been designed to
produce the
same axial flow patterns as marine propellers but are claimed to have advantages in
terms of manufacture and installation over the propeller (Oldshue (1983)). These tend
to be more expensive to manufacture because they have profiled blades and more care
needs to be taken in balancing them because they have three blades, compared with the
four blades normally fitted to pitched and flat blade turbines.
The three impeller types used in the tests were:
i) pitched blade turbines of T/3 and T/2 diameter.
ii) a flat blade turbine (ie. vertical blades) of T/3 diameter.
iii) Lightnin A310 and Chemineer HE3 hydrofoils of T/2 diameter.
The geometrical ratios were kept constant in scale-up and full dimensions of each
impeller used are given in table 3.1 and they are shown in figures 3.2 to 3.4.
All impellers were positioned such that their centre-lines were T/3 above the lowest
point on the base of the vessel. They rotated clock-wise when viewed from above and
the pitched blade turbines and hydrofoils pumped downwards.
3.4 CONDUCTIVITY PROBES
The method used for measurement of blend times relied on the detection of changes in
conductivity in different parts of the vessel after an electrolytic tracer had been added
on to the surface of the fluid. The conductivity changes were detected using
placed in regions of the vessel with differing agitation intensities.
conductivity probes
The conductivity probes used were made following the design originally described by
Khang & Fitzgerald (1975). They were constructed by embedding pieces of platinum
a screened signal lead, in a "bullet" of epoxy resin 6 mm in
wire, soldered onto
diameter and 30 mm long (Figure 3.5).
The outer electrode was coiled around the probe tip in the shape of a cone and this
for the measuring electrode and as a screen to prevent interference
acted as an earth
42
from other electrodes or earthed objects in the vessel. The probes were coated in
platinum black in order to make their surfaces inert.
The probes' measuring volume was made as small as practically possible to minimize
averaging effects when measuring conductivity. Their presence in the vessel would
affect local mixing but this was also minimised by making the probes as small as
possible and by arranging the probe supports in such a way that they did not interfere
with the flow of fluid around the vessel. This was achieved by attaching the probe
supports to the baffles.
As mentioned above, the three conductivity probes were positioned in regions of

differing agitation intensity:
i) Probe 1 was positioned beneath the impeller and near the blade tips in the
region of highest agitation intensity.
ii) Probe 2 was positioned halfway between the agitator shaft and vessel wall.
iii) Probe 3 was positioned behind a baffle in the region of lowest agitation
intensity.
The probes were placed at different heights and on different sides of the vessel and, on
in the same positions. These positions are given in table 3.2 and
scale-up, were placed
shown in figure 3.6.
The instrumentation chain used for blend time measurements is described in chapter 5.
3.5 TEST FLUIDS
Two types of test fluid were used: the first exhibited a Newtonian rheology and the
second, a non-Newtonian, pseudo-plastic rheology.
3.5.1 Newtonian Test Fluid
in the choice of the Newtonian fluid used for the power

Several factors were considered
Muskett (1984) described criteria which were set down
and blend time measurements.
for choice of fluid. These were:
43
i) Cost: Measurements were made in the 1.830 m diameter vessel,
which has an operating volume of 4.455 m3, so it was
important to find a relatively cheap material that would
not put too great a strain on the project budget.
ii)
Transparency: Flow visualisation was an important facet of the
experimental programme so it was important to find a
clear fluid which allowed all regions of the vessel to be
observed during experiments.
iii) Foaming: If the fluid was a foam stabiliser air, drawn into the fluid
from the surface, would form small, stable bubbles
reducing the mean density of the fluid and effecting the
power consumption of the agitator. The presence of
bubbles would also affect the performance of the
conductivity probes, causing large fluctuations in the
signal as bubbles passed between the probe electrodes.
iv) Handling/ The laboratory in which the experiments were performed

Storage: was originally built for Civil Engineering research and did
not have the facilities to handle hazardous chemicals. For

this reason the choice of fluid was limited to those which
did not require any special handling procedures. The
fluid also had to be stable over a period of days, once it
had been exposed to air.
v) Salt Solubility: Blend times were measured using a conductivity

technique and involved the addition of a small volume of
tracer with a high conductivity, to the vessel and
measurement of the conductivity fluctuations. For this
reason the fluid selected had to be capable of dissolving
an electrolyte without altering its rheology.
vi) Disposal: As mentioned above, the laboratory did not have facilities
to handle hazardous chemicals and disposal of the used
materials had to be considered carefully.
44
Several fluids were considered: Glycerol was rejected

on the grounds of cost (over
1000 per tonne) and its affinity for atmospheric moisture; Glucose
syrup was rejected
because its high dissolved sugar content reduced its ability to dissolve
electrolytes; and
Breox 75 W 30000, supplied by BP Chemicals Ltd., was found to be
a good foam
stabiliser.
The fluid selected was a grade of Versicol. These are polyacrylic acids,
polyacrylamides and polymethacrylic acids and are supplied by Allied Colloids of
Bradford, England. Several grades of Versicol were examined and Versicol E-11
was
finally selected. It dissolved electrolytes readily and, at a cost of around 400
per
tonne, was considered to be financially acceptable.
The visco-elasticity of Versicol E-11 solutions was checked by sending samples to the
Warren Springs Laboratory in Stevenage, England. Solutions of desired
viscosity were
made up and sent to the laboratory and were not found to exhibit visco-elasticity as
supplied but were found to evaporate readily and high concentration solutions were
visco-elastic.
3.5.2 Non-Newtonian Test Fluid
Many literature sources report the use of Carboxy-methyl Cellulose (CMC) solutions
in the study of the mixing of non-Newtonian, pseudo-plastic fluids. The initial stage
of this part of the project was to approach suppliers of CMC to discuss the
requirements. Aqualon (UK) Ltd. supply a wide range of Cellulose-based polymers and
recommended the use of a grade of Hydroxy-ethyl Cellulose, Natrosol 250MR.
This met all the criteria set out for the Newtonian test fluid and cost around 500 per
tonne. The fact that test solutions. were, - at most, 3/4% wt. /wt. meant that the cost of
materials used for this part of the project was very low.
The visco-elasticity of the Natrosol 250MR solutions was checked by sending samples
to the Department of Chemical Engineering at the University of Birmingham, England.
The solutions made-up did have a normal stress component and, strictly speaking were
visco-elastic, but under the operating conditions the ratio between the shear and
slightly
indicated that the effect of the fluid's visco-elasticity on the flow in the
normal stresses
be negligible. This was confirmed by the fact that none of the classic
vessel would
behaviour, such as shaft-climbing or flow-reversal were observed.
signs of visco-elastic
45
Later tests were made with solutions of Natrosol 250HHR, which has
a heavier
molecular weight and exhibits more severe non-Newtonian behaviour. 3/4% wt. /wt.
solutions made up with this material did shaft-climb so the most concentrated solution
of this grade used was 1/2% wt. /wt.
3.5.3 Tracers
The tracer used for Newtonian blend time measurements was made by adding
concentrated nitric acid to a Versicol E-11 solution ensuring that the viscosity of the
tracer was the same as the viscosity of the fluid in the vessel being used for the
experiment. The concentration of nitric acid in the tracer was 4.8 mols per litre.
The viscosity was always within 5 cP of the viscosity of the fluid in the vessel. The
viscosity was adjusted by adding more Versicol or water to the batch of tracer being
made and was checked during the experiment to ensure that it stayed within the limits
required.
The tracer used for the non-Newtonian blend time measurements was made by taking
a small volume of fluid from the vessel and adding sufficient salt (sodium chloride) to
it to make a saturated solution. After mixing, the undissolved salt settled at the bottom
of the container and clear fluid was taken from the surface for addition to the vessel.
3.6 RHEOLOGY MEASUREMENT
3.6.1 Newtonian Fluids
The viscosity of the Newtonian test fluids was measured using a Contraves Rheomat
108 portable viscometer. Its operation is based on the coaxial cylinder principle (see
Solomon (1980), Hickman (1985)). The measuring system conforms to the DIN 53019
standard. The rotational speed of the bob could be raised in eight steps from 50 to
1000 rpm giving a range of shear rates for the measuring system used of 64 to 1290 s-'.
A direct read-out of the fluid's viscosity was given at each shear rate, in Newton
seconds per square metre (Ns m-2).
The viscosity was measured at the start and end of each experiment and it did not
Also, the viscosity was measured at three of four shear rates to
change significantly.
of Versicol supplied was not "out-of-spec" and exhibiting a non-
ensure that the sample
Newtonian rheology. None of the samples ever failed this test.
46
3.6.2 Non-Newtonian Fluids
Solomon (1980) and Hickman (1985) have covered a great deal of the background on
non-Newtonian fluid rheology and measurement and will not be repeated in this thesis.
The technique used in this project to determine the fluid rheology is described below:
The rheology of the pseudo-plastic fluids was measured using a Contraves Rheomat 115
fully automated viscometer controlled by software run on a Tulip personal computer.
This also works on the coaxial cylinder principle and complies with the relevant DIN
standard.
The interface between the Rheomat 115 viscometer and the computer was a
"Rheoanalyzer" which converted the viscometer's output into digital form.
The computer was programed with the measurement routine and the following
information was input:
i) low shear rate.
ii) holding time at low shear rate.
iii) high shear rate.
iv) holding time at high shear rate
time taken for shear rate to rise from low to high value (the ramp time).
v)
time taken for shear rate to fall from high to low value.
vi)
hard disk, it could be fitted

Once the data had been taken and stored, on the computer
by the program. The power law model was selected
to different rheological models
K, index, n, were evaluated by performing a linear
(see chapter 6) and the constant, and
regression of:
Egn(3.1)
in ,r=n ln? + 1nK ...
47
The low and high shear rate values were taken as 10 and 500 s_' respectively because
the viscometer system could not measure accurately at lower shear rates. The holding
times were 30 seconds and the ramp time was 250 s. Once the data had been recorded
it could be fitted to the power law over different shear rate ranges in order to
investigate the effect of the range on the values of K and n. Very little effect was
observed when the data were fitted over any range within the overall sample range
provided that the lower shear rate was not altered. Increasing the lower shear rate
reduced the degree of pseudo-plasticity the fluid exhibited, increasing n.
The power law constants and indices of each of the Natrosol solutions used in the
measurements of blend times are given in table 3.3.
48
CHAPTER 4
BLENDING OF NEWTONIAN FLUIDS
4.1 INTRODUCTION
This chapter of the thesis is concerned

with the measurement and analysis of blend
times in viscous, Newtonian fluids. The blend time measurement method and
instrumentation chain are described in section 4.2
and the data reduction method is
described in section 4.3.
The data taken by FMP, reported by Mackinnon (1987), have been

re-analyzed to
produce a correlation of three dimensionless groups: impeller power number, Po;
Reynolds number, Re and Fourier number, Fo. The reason for
re-analyzing the data
and the analysis are described in section 4.4.
The analysis of the viscous Newtonian blend time data is described in section 4.5
and
discussion of the results and comparison with literature is given at
relevant points in
each section and a general discussion is given in section 4.6. The conclusions of this
section of the project are given in section 4.7.
4.2 METHOD
4.2.1 Instrumentation Chain

.
The instrumentation used for the measurement of blend times in viscous Newtonian
fluids was identical to that used to measure low viscosity blend times, reported by
Mackinnon (1987) and described by Ruszkowski (1985a)., A brief description of the
equipment is given below.
The conductivity probes, described in section 3.5, were energised by a conductivity

meter manufactured by the Instrumentation Section of BHR Group Ltd. for FMP and
was designed especially for blend time measurement.
The output from the meter was a0- 200 mV analogue signal, directly proportional to
the conductivity measured. The signal was amplified to 5V, passed through an anti-
filter and then input to the analogue-digital converter (ADC) board of a PDP
aliasing
11/23 computer. The low-pass cut-off frequency to which the filters were set was
49
dependent on the sampling frequency of the ADC.
The computer collected 4096 data points for each probe signal, irrespective of the
sample time. The sampling frequency was adjusted so that the sampling time was
always between three and four times longer than the blend time. For example, if a
sampling frequency of 50 Hz was selected the sample time was 81.92 seconds and the
expected blend time would be between 20 and 30 seconds.
The data were stored in the computer memory during an experiment and written to the
hard disk once the sampling routine was completed.
4.2.2 Blend Time Determination
The blend time measurement method used in this work involved the addition of a small
volume electrolyte to the bulk fluid in the vessel and recording the conductivity at three
points in the vessel, shown in figure 3.6. The tracer was added from a solenoid valve
positioned above the fluid surface and its exact position is also shown in figure 3.6.
A typical conductivity trace is shown in figure 4.1. The solenoid valve was activated
by a Schmidt trigger which also signalled the computer to start recording the output
from the conductivity meter.
The blend time was measured 8 times for each set of conditions and the data are
presented in tables 4.1 to 4.4. The blend time reported is the mean of the eight
repetitions and the relative standard deviation is
associated with each mean value given
where the relative standard deviation is given by:
RSD= 0x100% Egn(4.1)

...
0
4.3 DATA ANALYSIS
The technique used for the analysis of the viscous blend time data taken for this project
identical to the method used by the workers who measured turbulent blend times
was
for FMP. The method is described by Ruszkowski (1985a) and is briefly described
below:
50
Root mean square concentration has been chosen as the quantity used to characterise
the blending in agitated vessels. The time varying concentration in a vessel at any point
may be represented by the mean concentration in the vessel and a time-varying
component:
C(t) =C+ C'(t) Egn(4.2)

...
where: C(t) is the local concentration at time t.

C is the mean concentration.
C'(t) is the local fluctuating component of concentration at time t.
The root mean square concentration fluctuation over a time F is given by:
rr
f Egn(4.3)
6 (C(t) - C)2 dt =1f (C'(t))2dt ...
c2 ro ro
This approach is commonly used for dealing with turbulent fluctuating quantities such
as fluid velocities measured by laser-Doppler anemometry. The situation in an agitated
vessel is very complicated since the concentration fluctuations are transient, whereas the
magnitude of the root mean square velocity is constant with time. C and are
functions of I', the time over which the integration takes place. This problem has been
overcome by calculating the root mean square concentration for small sections along
the conductivity-time history. 4 typical plot of 6c2 versus time is shown in figure 4.2.
The concentration within the vessel was measured at three points, 4096 times during
an experiment, and the analysis program normalised the responses so that the magnitude
in concentration was represented by 2048 computer digits,
of the total change
irrespective of the actual change in conductivity in the vessel.
The root mean square concentration fluctuation is then calculated for a small section of
the trace. The position of the "calculation section" is then moved along the root mean
fluctuation and calculated again. There were 32 points in each
square concentration
moved forward by 4 points to carry out the next calculation.
section and the section was
A series of estimates of the root mean square concentration fluctuation are obtained:
4i+31
i=0,1,2 1016 Egn(4.4)

c2(1) _1 (C _ C(o))2 ...... ...
32 4i
51
where:
6C2(i) is the i`h estimate of the root mean square concentration fluctuation.
CCis the concentration of the jh data point.
C(oo) is the concentration in the vessel when blending is complete.
The number of points over which the root mean square is calculated is 32 since this
covers a significant portion of the circulation time in the vessel.
Having converted the data from a raw conductivity trace to root mean square
concentration fluctuation versus time, the blend time was defined as the time taken for
the degree of homogeneity to reach a pre-set value. The degree of homogeneity was
calculated by defining a mixing index, M, where:
AC
-(
M= c2)1/2 Eqn(4.5)
...
AC
where: AC = C(oo) - C(O)
M expresses the reduction in the root mean square concentration fluctuation as a

fraction of the total change in concentration in the vessel due to the injection of tracer.
The value of M selected was 0.95, ie. the blend time is defined as the time taken for
the blending process to be 95% complete.
The individual probe blend times, 6,, 02 and 03, were all calculated in this way. A
fourth blend time, the root mean square blend time 6RMS,was calculated to give an
estimate of the blend time for the vessel as a whole. This is calculated from the data
taken at each individual probe by defining a root mean square concentration in the
vessel:
R
Cj. 2 C!. 2 Egn(4.6)
C,
RMS = 3 CZ
J. 1
+ 2 + 3 ...
Cj, is the concentration of the nthdata point measured at probe 1, etc.

where ,
52
C;, is substituted for Cj in equation (4.4) to calculate
RMS C,RMS2which in turn is
substituted into equation (4.5) to calculate MRMS.
All data correlation and comparison with other workers'

results has been made using
BRMSsince it is defined as the blend time for the in
whole vessel and results the
literature are presented on the same basis.
4.4 TURBULENT DATA RE-ANALYSIS
Mackinnon (1987) reviewed the results of the blend time measurements made in
water
by the FMP consortium. The first data were fitted to:
3/6
D
Po' NO Egn(2.5)
T ...
This was developed by taking the dimensionless power per unit mass, as reported by
Mersmann et al. (1976) and Havas et al. (1978b), ie:
D r30 Egn(4.7)
Po 1/3 NO ...
-T T2 =k'
and introducing a scale factor:
DD a= Egn(4.8)
Po" NO k ...
TT
The exponent a was determined to be 1/2 by analyzing the data of Khang (1975)
because, when the correlation was first presented by Ruszkowski (1985b), of the six
impellers tested, five were of T/3 diameter and one T/2. As more data were taken they
fitted to equation (2.5) without checking to see if a better fit could be obtained
were
with a different value of a.
Before analysis of the viscous blend time data was started the data taken in water were
by carrying out a multiplicative regression of NO on Po and DIT using the
re-analyzed
Statgraphics computer package, ie:
53
NO aPob_ Egn(4.9)
T ...
The values of b and c were

-0.319 and -2.062 respectively with a correlation coefficient,
r2, of 91.71% and F-ratio of 139, for 26 data points.
After rounding-up these values, this showed that the

value of a is 1/3, not 1/2, ie:
Po "NO = k" Egn(4.10)

T ...
Having carried out this regression, equation (4.10)

was re-arranged and both sides
multiplied by p/ to give:
PND 2= PT29...
Po 1/3 k11 Egn(4.11)

where: k" = 5.20 10.0%
Equation (4.11) can be re-written:
5.20
Pow Re Eqn(4.12)
_ Fo ...
The terms on the left hand side of equation (4.12) are the impeller power number and
Reynolds number which are commonly used and well-known dimensionless groups.
The power number can be regarded as a drag coefficient and relates the power input by
an impeller to its speed and diameter. Reynolds number is a ratio between the inertial
and viscous forces acting on the fluid.
The term on the right hand side of equation (4.12) is Fourier number and has been used
by other workers in this area to correlate blend time data (eg. Hoogendoorn & Den
Hartog (1967), Shiue & Wong (1984)).
54
Fourier number is a dimensionless group used in unsteady-state heat
mass, or
momentum transfer problems (Parkin (1984)). It can be written in a general form as:
Fo= a6
Egn(4.13)
T2 ...
T is the characteristic length and 8 is the characteristic time of the process in question.
a is a kinematic property of the fluid and is:
i) the diffusion coefficient in mass transfer processes.
ii) the thermal conductivity in heat transfer processes.
iii) the kinematic viscosity in momentum transfer processes.
The blend time in an agitated vessel is controlled by the viscosity of the fluid so the
third version of Fourier number is the appropriate one to use in analysis of the data.
An important point to note regarding this analysis is that it implies that different
impellers of equal diameter will achieve the same blend time at the same power input.
The correlation can be re-arranged to show:
I Egn(4.14)
Ne=k'(T ...
Other workers have found that the exponent on T/D is dependent on impeller type. For
example, Prochzka & Landau (1961) found the exponent was 2.05,2.20 and 2.57 for
marine propellers, pitched blade turbines and Rushton turbines respectively. Khang
(1975) found the exponent was 2.0 for marine propellers and 2.3 for Rushton turbines.
The associated constants of proportionality were different for each impeller type
in this thesis shows that one constant can be applied to
whereas the analysis presented
the impeller types tested if the exponent is the same. The relative standard deviation
all
is 10% which is the same as the deviation reported by Kramers et al.
of the constant
(1953) for their NO values. Since power number and D/T are constant in the turbulent
for impeller the standard deviation of the correlation constant is in
regime a particular
fact the standard deviation of the NO values.
55
Although previous workers have reported a difference in the
exponents on D/T the fact
that the scatter of the data about the correlation is as good as the scatter of the raw data,
and that the data correlated were taken at three scales (0.61,1.83 and 2.67 m diameter
vessels) gives confidence in the use of this correlation for agitator design and impeller
companson.
4.5 ANALYSIS OF VISCOUS BLEND TIME DATA
The blend time data have been analyzed and are reported in three ways:
i) individual probe blend times which indicate how the blending process
proceeds in regions of different agitation intensity.
ii) root mean square blend times which indicate how blending proceeds for
the vessel as a whole.
iii) fitting the data to a correlation of the same form as the one developed
for the turbulent data, equation (4.12)
4.5.1 Individual Probe Blend Times
Individual probe blend times were measured in order to investigate the way in which
the blending process proceeds in regions of high and low agitation intensity. Three
probes were used: the first was-positioned beneath the impeller, the second half way
between the shaft and vessel wall and the third, behind a baffle. Mackinnon (1987)
found that the individual probe blend times measured in water were almost equal at a
particular impeller speed.
The results of the viscous blending experiments are given in tables 4.1 to 4.4. The
tables show vessel scale, nominal fluid viscosity, impeller speed, three individual probe
blend times and root mean square blend time for each set of experimental conditions.
The blend time reported is the mean of eight repetitions of the measurement and the
number given in brackets beneath is the relative standard deviation of the mean, given
as a percentage. The results of the individual probe measurements are plotted as N6;
Reynolds number in figures 4.3 to 4.6. The results reported by Mackinnon
versus
(1987), taken in water, are included in the plots for comparison.
The plots show that a critical Reynolds number, ReC1i,,

exists for each impeller. When
56
the viscosity of the fluid is low, and Re > Rec, the blend times are approximately
't,
equal. As viscosity is increased, and Re < Recj, the blend ti mes diverge such that:
N81 <NO2 <NO3 Egn(4.15)

...
The degree of divergence increased with decreasing Reynolds number for the two
pitched blade and flat blade turbines while the degree of divergence is fairly constant
for the hydrofoil. This appears to indicate that the way in which the tracer is blended
into the bulk fluid differs for the various impellers.
Landau et al. (1963) carried out a similar exercise with five axial and three radial
combinations of probe position (ie. fifteen possible positions) and found that the value
of critical Reynolds number could not be strictly identified since it was determined by
a number of factors which they did not correlate. They presented plots showing the
distributionof NO with position for a number of impellers in baffled and unbaffled
vessels at Reynolds numbers of 2.7 x 103 and 5.0 x 104. The results are generally in
good agreement with the results reported here. Once Re < Re,,;,, the blend times
throughout the vessel will be longer than those measured under turbulent conditions but
the blend times measured near the impeller will be shorter than those measured in
regions of low agitation intensity, near the vessel wall, near the fluid surface or behind
a baffle.
Several workers have modelled agitated vessels as a well-mixed region near the impeller
in which the agitation intensity is high and regions in which the agitation intensity is
lower with transfer of fluid into and out of these regions as the blending process
proceeds (eg. Khang (1975), M`Manamey (1980)). In the turbulent regime, the
pumping number of the Fl,

impeller, is constant and flow between the well-mixed
region and other parts of the vessel will be determined by the impeller speed since:
Eqn(4.16)
Q= Fl ND3...
As Reynolds number is decreased a value will be reached where F1 is a function of

Reynolds number and decreases as Reynolds number decreases. This has been reported
by Wong & Huang (1988) who found that the boundary between these two regimes
Re W. The most likely explanation for the increase in blend times
occurred at =
the wall and surface of the vessel is the reduction in pumping capacity
measured near
hence, in transfer of well-mixed fluid to these regions and unmixed fluid
and reduction
Reynolds determined from the NO versus Re
into the impeller. The critical number
57
plots (figures 4.3 to 4.6) occurs at about the same value supporting this explanation of
the process.
A point to note from these figures is that the degree of divergence between NO1 and
NO3 varies with impeller type. The most marked difference is shown by the T/2
Lightnin A310 hydrofoil where the blend times do not diverge but have a constant
difference as Reynolds number decreases with NO3being 2-3 times greater than NO1.
This indicates that the proportion of the energy input to the vessel devoted to flow and
mixing may vary according to impeller type but, the correlation of the data (see section
4.6.1) shows that there is very little difference between the root mean square blend
times when impellers of the same diameter are compared at equal power input.
Most workers have reported their data in the form of a blend time representative of the
blending in the vessel as a whole. The root mean square blend times have been
calculated in order to obtain a blend time for the whole vessel and are more suitable for
comparison with the bulk of data in the literature.
4.5.2 Root Mean Square Blend Times
The root mean square blend times are given in tables 4.1 to 4.4 and are plotted in
figures 4.7 to 4.10. Again the data of Mackinnon (1987) are included for comparison.
The critical Reynolds number can be identified clearly on these plots and it can be seen
that the Lightnin A310 hydrofoil is affected by viscosity at a higher Reynolds number
than the other three impellers. This result has been reported by Zlokarnik (1967) and
Khang (1975). Both measured blend times for marine propellers and Rushton turbines
found that, for the marine propeller Reri, = 2.0 x 104 and for the Rushton turbine RecR,
2.0 x 103. The results reported here show the same trend- with the highest Re, 1.0
n, =
x 104 for the Lightnin hydrofoil and the lowest Rec,; = 3.0 x 103 for the flat blade
1
turbine. The Rec,;,for the two pitched blade turbines occurs at about 5.0 x 103.
value of
The impeller with the highest power number (the flat blade turbine) has the lowest Rec,;,
This is because the power input to the fluid by the four impellers at a
and vice versa.
Reynolds number will be different. For example, if a T/3 flat blade turbine
particular
Po 3.0 is compared with a T/2 hydrofoil with Po = 0.3 at equal Reynolds
with =
flat blade turbine will input approximately 15 times the power of the
number the
hydrofoil.
be made between the data reported here and in the

Another comparison which can
58
literature is the relationship between NO and Reynolds
number. It is generally accepted
that, in the viscous mixing regime:
NO a Re -' Eqn(4.17)
...
Most workers agree that x=I in the transitional regime.
The Statgraphics computer package was used to perform a

regression of NO on
Reynolds number for the data taken in the viscous regime (ie. Re < Rec,
i) and the
results are shown in table 4.5. Both pitched blade turbines fit this relationship given
by equation (4.17) but the flat blade turbine and hydrofoil do
not.
Zlokarnik (1967) found that the exponent, x, on Reynolds number for

a marine propeller
was -0.7 which agrees quite well with the value of -0.84. The value of x for the flat
blade turbine, of -1.26, is probably due to the number of points in the data set.
4.5.3 Correlation of Data
Initially the data were plotted as Po'l' Re versus 1/Fo, ie. the turbulent data correlation,
for each impeller, in figures 4.11 to 4.14, and all the data sets together. This is
presented in figure 4.15 and shows that two mixing regimes exist. The boundary
between the two regimes occurs at:
1/Fo = 103 Egn(4.18)

...
So the data groups were split at this value using the facility in the Statgraphics package.
The data in the groups defined by equation (4.18) have been correlated by performing
a regression of:
Eqn(4.19)
Po 1/3 Re Fo ...
and the results are given in the table 4.6.
A very good fit of the data to the relationship was obtained but the exponent, y, is
different for each impeller. The correlation would be to
easier use, for agitator design
and impeller comparison, if the exponents are equal so the value of y was rounded-up
for each impeller and the constants and their standard deviations compared, ie.
to -0.5
59
the constants in the following two equations were compared:
Po" Re
= ky Egn(4.20)
Fo'' ...
Po 1/3Re
= ko. Eqn(4.21)
Fo -0.s s ...
The comparison of the constants and their deviations is given in table 4.7.
The table shows that rounding-up the exponent has very little effect on the scatter of
the data. Thus all the data tken at Re < Rec, can be fitted to:
i,
183
Po If' Re = Egn(4.22)
...
Fo
The critical values of the two correlating variables were identified by solving equations
(4.12) and (4.22). This shows that:
Po"' Rec,;,= 6370 Egn(4.23)

...
1/Fo,; = 1225 Egn(4.24)

t ...
1/Focri: is close to the value taken to split the data (1/Fo = 1000) as expected.
The critical value of Reynolds number for a particular impeller can be found by re-
arranging equation (4.23) to show:
6370 Egn(4.25)
Rec; - ...
1 Po l
This shows that an impeller with a high power number will have a lower critical
Reynolds number than an impeller with a low power number which can be seen in plots
4.7 to 4.10 and has been reported by Zlokarnik (1967) and Khang (1975).
60
4.6 DISCUSSION
4.6.1 Comparison using Power Input
The information that will be of most interest to agitator designers is

the type of impeller
which gives the best performance in blending fluids for the lowest power input. The
data taken in the turbulent regime, equation (4.12), be
can re-arranged to obtain a
relationship between blend time, impeller diameter and vessel scale:
5.20 T2
Egn(4.26)
Po" ND 2 ...
which can be re-arranged to show:

T3T /3 Eqn(4.27)
8=5.20 T2 ...
PoN3D5
which can in turn be re-arranged to:
8e1DT T2 Eqn(4.28)
...
Equation (4.28) shows that if impellers are compared at equal power input, a large
impeller (smaller TID) will give a shorter blend time. This rule has been confirmed by
many other workers (eg. Prochzka & Landau (1961), Khang (1975)) but Havas et al.
(1978b) foundthat the optimum diameter of a VEGYTERV propeller was
approximately 55% of the vessel diameter and Van de Vusse (1955b) found that the
optimum impeller diameter was approximately 40% of the vessel diameter. The
impeller sizes studied by Mackinnon (1987), whose data have been re-analyzed,
Prochzka & Landau (1961) and Khang (1975) were less than or equal to 50% of the
vessel diameter so it is likely that insufficient data were taken to identify a similar
minimum in power requirement.
A similar re-analysis can be performed on the data taken in the viscous regime,
equation (4.22), which shows:

61
f3 (T
1 Egn(4.29)
e T -2
FD ...
Equation (4.29) also shows that large impellers give a shorter blend time when
compared at the same power input and scale. Van de Vusse (1955b) found that
impellers should be as large as possible when Re < 250, but no information was
presented on impellers greater than 50% of the vessel diameter in the regime studied
here, ie. 200 < Re < 10.
There are no great surprises in the results of the effect of impeller diameter on blend
time in the turbulent and viscous regimes, but the effect of scale at first may appear
unusual. In the turbulent regime, at constant power input per unit mass and
impeller/vessel geometry, the blend time increases with the ratio between the blend
times at large and small scale given by:

p
v large L large
Egn(4.30)
small
...
v
small
This relationship also shows that if a process is to be scaled-up at constant blend time
and geometry, the ratio in power inputs is given by:
T
Plarge large Egn(4.31)
T
...
Psmall
small
Equation (4.29) can be re-arranged into the same form as equation (4.30) to show that,
at constant power input per unit mass and impeller/vessel geometry:

0large
small Egn(4.32)
...
_T T
large
small
is
ie. blend time reduces as the process scaled-up!
If a process is scaled-up at constant power input per unit mass and geometry the ratio
62
between the impeller speeds at large
and small scale will be:
Marge Dsmail
= Egn(4.33)
Nsmall D ...
large
The ratio between Reynolds numbers be:

will
Relarge MargeD2
large Egn(4.34)
Re NsmallD 2 ...
small marl
Substituting equation (4.33) into (4.34):

Re Dlarge
large Egn(4.35)
Re ` ...
Dsmall
small
It has been shown that, in the viscous regime:
NO Re-' Eqn(4.17)
...
Combining equations (4.35) and (4.17) shows the relationship between NO at different
scales:
`NO) large IDmaiiJ' Eqn(4.36)

D ...
`Ne) small large
Cross-multiplying the impeller speeds at the two scales and substituting equation (4.33):
Dsmall
large Eqn(4.37)
D
...
large
small
blend time at constant geometry is being considered the ratio between

and since
impeller and vessel diameter is constant so equation (4.37) can be re-written:
63

e T
large small Egn(4.38)
0 T ...
small large
In practise, it is more likely that an agitator will be designed at constant blend time, in
which case the power per unit mass required to achieve that blend time will be reduced.
The ratio between the power requirements at the two scales will be given by:
Plage Tlarge Egn(4.39)

...
4.6.2
L
Use of the Correlation for Agitator Design
Blend times have been analyzed for 1< 1/Fo < 106 and two regimes have been
identified with the boundary occurring at:
1/FO = 103 Egn(4.18)

...
If 1/Fo < 103, blend time reduces with increasing scale and, if 1/Fo > 103, blend time
increases with increasing scale provided power input per unit mass and impeller/vessel
geometry are constant. This result indicates that the optimum scale for operation of an
agitator, required to achieve a desired blend time, is the one which operates at the
boundary between the two regimes. This is the point at which the blend time will be
achieved for minimum power input per unit mass The vessel diameter will be
.
calculated from:
10' 8 Egn(4.40)
T_ ...
The form of data correlation presented here is particularly useful because it allows
impeller types to be compared with one equation and it splits the variables
various
determined by conditions at the impeller from those determined by the scale of
operation and the desired process result (blend time).
If blend time is design criterion for a particular agitation duty, the

an appropriate
determined by the scale of operation and process result will be known, ie.
conditions
(7), fluid physical properties (p, t) and desired blend time (0) are
the vessel scale
64
specified at the start of the design and Fourier number (Fo) can be calculated. The
mixing regime can immediately be identified by comparing the calculated value with
Foc,; and the value of Poly Re can be found. This will allow the impeller type, speed
1
and diameter to be chosen in order to satisfy the blend time requirements.
Similarly, if an existing agitator is to be rated or vendors' quotations compared, for a

blending duty, the impeller type (Po), speed (N), diameter (D), vessel scale (T) and fluid
properties (p, t) will be known. Po"3 Re can be calculated, the mixing regime
identified, Fourier number calculated and the blend time found by re-arranging the
Fourier number.
The correlation shows that different impeller types will achieve the same blend time at
a given scale, provided the impellers are the same diameter and operating at the same
power input per unit mass. This is true in both the turbulent and viscous mixing
regimes. Thus there is no saving to be made in agitator running costs by choosing one
impeller type over another.
The capital cost of an agitator can be related to the shaft torque since the a high torque,
requires a greater mechanical strength of gearbox, shaft and impeller. So, savings in
capital cost can be achieved by selecting an impeller which operates at a low torque.
The torque is related to the impeller power by:
P= 2n AN Egn(4.41)
...
When two impellers are compared at equal power input the impeller with the higher
rotational speed will have the lower torque.
If a flat blade turbine, with Po = 3.0, and hydrofoil, with. Po = 0.3 are compared at
equal impeller diameter and power input, the difference in rotational speeds will be:
3_
NHyd rPOFBT 2.154 Egn(4.42)
= ...
NFBT '
Hyd
Therefore, a large diameter hydrofoil will give the most cost effective operation.
Further analysis of the economic implications of the results of the research are given
in chapter 7 of this thesis.
65
4.7 CONCLUSIONS
The conclusions which can be drawn from the re-analysis of blend time
measurements
made in water, reported by Mackinnon (1987) are:
i) blend time data can be correlated by a relationship

containing three
dimensionless groups; power number, Reynolds number
and Fourier number:
5.20
Po" Re = Eqn(4.12)
Fo ...
The relative standard deviation of the constant of proportionality is 10.0%.
ii) This correlation is based on data taken with two pitched blade turbines, of T/3
and T/2 diameter, a flat blade turbine of T/3 diameter and a Lightnin A310
hydrofoil of T/2 diameter in standard, baffled vessels of 0.61,1.83 and 2.67 m
diameter. In all experiments the impeller was positioned T/3 above the base of
the vessel.
iii) This relationship can be re-arranged to show that the dimensionless group, NO,
is constant for each impeller type.
iv) Individual probe blend times; measured beneath the impeller, halfway between
the agitator shaft and behind a baffle are approximately equal.
The conclusions which can be drawn from the data taken in viscous Newtonian fluids
for this research project are:
i) two mixing regimes have been identified with the boundary between them
occurring at 1/Fo = 103.
ii) for 1/Fo > 103 NO constant
for 1/Fo < 103 NO a Re"'
iii) 1/Fo < 103 blend time can be correlated by:

when
66
Po 1/3Re = 183
Egn(4.22)
Fo ...
The relative standard deviation of the constant

of proportionality is 17.4%.
iv) The same four impellers were used as those for

used the water tests, reported
by Mackinnon (1987) in 0.30,0.61 and 1.83
m diameter vessels of the same
geometry.
v) This relationship can be re-arranged to show that NO is inversely

proportional
to Reynolds number, as mentioned previously.
vi) When Re < Rec,; individual probe blend times diverge the longest
n with
measured behind the baffle and the shortest beneath the impeller. The degree
of divergence increases with decreasing Reynolds number.
The two correlations can be re-arranged to show the dependence of blend time on
power per unit mass, impeller/vessel geometry and scale:
i) for 1/Fo > 103 1e '3

Egn(4.28)
0 T2fl ...
DT
1,3 /TT
ii) for 1/Fo < 103 1 Egn(4.29)
oc T-213 ...
eD
iii) in both the turbulent and viscous mixing regimes a large diameter impeller will
achieve a shorter blend time than a small one at equal power input per unit
mass.
iv) on scale-up at constant D/T in the turbulent regime:
13
input large Trage Egn(4.30)

at equal power per unit mass: ...
8
1
small small
ie. blend time increases with increasing scale of operation.

67
for equal blend time: Plarg e jl
arge Egn(4.31)
' T ...
sma ll mall
ie. the power required to achieve equal blend time on scale-up increases
dramatically.
v) on scale-up at constant D/T in the viscous regime:
e T
at equal power input per unit mass: large Eqn(4.38)
0 T mau ...
small large
ie. blend time decreases with increasing scale of operation.
for equal blend time: Plarge Tlarge Egn(4.39)

...
Psmall T
small
ie. the power required to achieve equal blend time increases with increasing
scale but not as quickly as in the turbulent regime.
There are no savings to be made in agitator running costs because the blend time is
dependent on the power input to the fluid in both mixing regimes. Choosing a large
impeller diameter will reduce the running cost required to achieve a desired blend time.
There are potential savings to be made in capital cost if an impeller with a low power
is
number chosen. This will run at a high speed to deliver the required power input and
will therefore operate at a lower torque. Less mechanical strength will be required in
the gearbox, shaft and impeller in
giving potential savings .the agitator's capital cost.
68
CHAPTER 5
BLENDING OF PSEUDO-PLASTIC FLUIDS
5.1 INTRODUCTION
The vast majority of viscous fluids processed in the chemical industry are non-
Newtonian and a large proportion of these are pseudo-plastic. Short & Etchells (1982)
have highlighted the importance of researching the effects of non-Newtonian fluid
properties on agitation.
The previous chapter dealt with the results of blend time measurements in viscous,
Newtonian fluids, the correlation of data and the implications for agitator selection.
This chapter covers similar ground, but deals with the measurements made in viscous,
pseudo-plastic fluids.
The experimental and data reduction methods used were identical to those used for the
Newtonian measurements, described in sections 4.2 and 4.3.
The viscosity of the non-Newtonian fluids varies with position in an agitated vessel
because the shear rate differs. The viscosity of a pseudo-plastic fluid will be low near
the impeller where the shear rates are high and high near the fluid surface or the wall
of the vessel where the shear rates are low. Section 5.2 discusses pseudo-plasticity, its
effect on blending and proposes. a method of account for it in the correlation of blend
time data.
The application of this correlation method to the data is discussed in section 5.3 and a
general discussion is given in section 5.4. The conclusions are given in section 5.5.
5.2 PSEUDO-PLASTIC FLUIDS
5.2.1 The Power Law Model
The rheology of a fluid is determined by measuring the shear stress over a range of
When the data are plotted, a Newtonian fluid is characterised by a straight
shear rates.
line relationship between the two parameters. The gradient of the line is the dynamic
of the fluid, ie:

viscosity
69
ti = ' Egn(5.1)
...
A pseudo-plastic fluid is characterised by a progressively decreasing gradient on a plot

of shear stress versus shear rate. The apparent viscosity of the fluid at a particular
shear rate is the ratio between the corresponding shear stress and the shear rate, ie:
Egn(5.2)
...
I
At very high and low shear rates the apparent viscosity becomes constant (ie. the fluid
is Newtonian) and equal to - and o. These relationships are shown in figure 5.1.
There are several empirical correlations which describe the relationship between shear
stress and shear rate in a pseudo-plastic fluid and of these, the simplest is the power law
model. This is written as:
i =KY" where n< 1 Eqn(5.3)

...
K is a measure of the fluid's viscosity, or "thickness", and n is a measure of the fluid's

deviation from Newtonian behaviour. For a Newtonian fluid, n=I and K=.
If equations (5.2) and (5.3) are combined:
Egn(5.4)
a =K yn-1 ...
As the shear rate increases the apparent viscosity of the fluid decreases, hence shear-
thinning behaviour.
There are several objections that have been raised against the use of the power law
model. Among these are:
i) the shear rate is zero, the predicted viscosity of the fluid is infinite.
when
ii) for real fluids, n is not constant over a wide range of shear rates.
iii) the dimensions of K are dependent on n.

70
The second objection to the use of the power law can be overcome by specifying the
shear rate range over which the rheological measurements have been made and the
values of K and n can be applied.
Other, more complex models have been proposed that overcome these objections but
the power law is widely used because of its simplicity and because it has been found
to adequately describe pseudo-plastic fluid rheology in many applications.
The apparent viscosity at a particular point in a non-Newtonian fluid can be estimated

if the shear rate in the fluid at that point is known. One of the most important aspects
of this project is to identify a method of characterising the shear rates in agitated,
pseudo-plastic fluids so that the blend time data can be correlated.
5.2.2 Shear Rates
Examination of the Newtonian data shows that the blending in the vessel as a whole is
controlled by the rate of mixing at probe 3, which was positioned behind a baffle (see
figure 3.6). Thus, the viscosity in this region of the vessel controls the overall blending
process. A method for estimating the characteristic shear rate is therefore required in
order to calculate this viscosity. Ideally, the viscosity at probe 3 should be estimated
but this is very difficult. The viscosity at the wall of the vessel can be found more
easily and should be significantly different from the baffle viscosity. This viscosity has
been used to correlate the blend time data and its estimation is discussed in the
following section.
5.2.3 Wall Shear Rate Estimation
The estimation of the shear rate is based on a torque balance which equates the shear
the wall of the vessel with the torque per unit volume. This is a simplification
stress at
of the derivation of this equation which should actually be written:
Torque = Shear stress at surface
x Area of surface
x Radial distance from vessel axis to surface.

71
Bird et al. (1960) give an equation relating the
shaft torque to the fluid viscosity, the
velocity gradient (or shear rate) in the fluid at the wall and base of the vessel and the
pressure drop due to the fluid velocity changing as it impinges on the baffles:
JJR+ JJR
Egn(5.5)
A ar pb...
S wA
where: ve is the tangential component of the fluid

velocity.
S is the total surface area of the vessel in contact fluid.
with
A is the total surface area of the baffles.
R is the radial distance to any surface area element dS dA from
or
the impeller axis of rotation.
r is the distance measured into the fluid from any
element of tank
surface, dS or dA.
p is the pressure exerted on the baffles by the fluid.
In order to solve equation (5.5), it is necessary to have an estimate

of the tangential
velocity gradient at the vessel surfaces. If an average shear rate at the surfaces of the
vessel is defined, this term can be re-written:
ffR av Eqn(5.6)
e as
ar
J'R
dS
J'JR
d
...
s w S S
No assumptions need to be made about the velocity gradient since this has been
combined with the viscosity term to give the shear stress in the fluid at the vessel
surfaces. In order to define an average shear rate, in equation (5.6), it has been
assumed that the shear stress is constant at all points on the vessel surface. Wichterle
et a1. (1985) have measured the shear stress on the wall of an agitated vessel using an
electrochemical technique and shown that it does vary with position. It is also likely
to be a function of impeller type with an axial flow impeller giving higher shear
stresses on the base of the vessel than a radial flow impeller which will give high shear
stresses on the wall. The assumption of constant wall shear stress has been made in the
following analysis even though this over simplifies the true relationship between the
shear stress and position on the vessel wall because there are no relevant date which can
be applied to the four impellers tested and it has been used previously by Elson (1988,
1990) in the analysis of cavern formation in agitated yield stress fluids.
72
Equation (5.5) can be written in terms of three componentswhich are balancedwith the
shaft torque. These are:
i) the contribution of the cylindrical vessel wall.
ii) the contribution of the vessel base.
iii) the contribution of the pressure drop on the baffles.
Therefore the vessel will be considered as three sperate elements, each balancing a
proportion of the shaft torque. If a term G is defined as the product of the area of a
small element of the vessel surface and its radial distance from the vessel axis, equation
(5.5) and (5.6) can be re-arranged to:
ffdGcyl JJPbI Egn(5.7)

A ti +ff dGb. + dGbff ...
w Ste, Sb-"
A
In order to carry out this integration, the vessel geometry has been simplified and this
is shown in figure 5.2. The limits of each integral are also shown.
5.2.3.1 The Vessel Wall
The area of a small element of the vessel wall is dScy,. Therefore, the product of the
.
surface area and radial distance for each element of the vessel wall can be written:
R dScy, = dGcy, =R ((R do) dz) Eqn(5.8)

...
The limits of the integration are the arc swept by the vessel wall, 0, which ranges from
0 to 2it radians and the axial height of the vessel wall, z, which ranges from 0 to
')3 T/2. The radial distance between the vessel axis and the surface under consideration
is
is constant for all elements and equal to vessel radius, T/2. The integral equation can
be written:
2n 3-TR
Jdf Egn(5.9)
Gcy! dZ ...
200
73
Integrating:
3T=1.360 Egn(5.10)
Gcy` _T 27t T3 ...
42
5.2.3.2 The Vessel Base
The vessel in which the blend time measurements were made has a dished base (see
figure 3.1) but, in order to simplify the integration, it has been assumed that the vessel
has a flat base. The area of a small element of the vessel base is dSb,., and the radial
distance between the vessel axis and each element varies and is equal to r (see figure
5.2). Therefore, the product of the area and the radial distance for each element of the
vessel base is given by:
r dSbase= dGbase=r ((r do) dr) Egn(5.11)

..
Again the arc swept by the vessel base must be considered but the axial height of the
base is constant, since it has been assumed that it is flat. The radial position, r, varies
from 0 to T/2. The integral can be written:
2n rrz
Jdo Jr2 Egn(5.12)
Gbase = dr ...
00
Integrating:
3
T=0.262 Eqn(5.13)
Gbare=1 2n T3 ...
38
5.2.3.3 The Baffles
Two quantities contribute to the effect of the baffles on the torque balance: the first is
the baffles and the second, the change in velocity of the fluid as it impinges
the area of
on the baffles.
74
The pressure drop due to the baffles is:
pb01f p (Av)2 Egn(5.14)

2 ...
The area of a small element of the baffle is dA and the radial position of the baffle
varies and this is defined as r. The product of the area and the radial distance can be
written as:
r dA dGbaff (dz dr) Egn(5.15)

= =r ...
The height of the baffles, 0 43

z, ranges from to T/2 and they are T/12 wide and
mounted T/60 off the vessel wall, giving a total distance between the inner edge of the
baffle and the vessel wall of T/10. Considering one baffle, the integral can be written:
3-T2 (T/2 - T/60)

fr Egn(5.16)
Gbq(f =f dz dr ...
0 (T2 - T/10)
Integrating:
FT 29 TT=0.03187 Eqn(5.17)
G_1 T3 ...
baff 22 60 5
5.2.3.4 Combining
The three components of the equation can be combined to give:
(G1 +Gbase)+4x pbafI GbI Eqn(5.18)

A =2 w ...
(w)2
A- ti (1.360 + 0.262) T3 +4xP 0.03187 T3 """Egn(5.19)
w2
Re-arranging equation (5.19):

75
'Z =n-0.0638 p (Av)2 Eqn(5.20)

1.622 T3 ...
The only quantity in equation (5.20)

which is not known is the velocity of the fluid
impinging on the baffles. This has been
estimated from the data of Schwartzberg &
Treybal (1968) who used streak-photography to
measure the fluid velocities in vessels
agitated by Rushton turbines. They correlated their data by:
N D2
vST = kST Egn(5.21)
(T2 H)" ...
where ksT is 1.586 for the average total velocity and 1.387 for the average vertical-
plane fluid velocity. If it is assumed that the velocity of the fluid after impact on the
baffles is zero, equation (5.20) can be written as:
T=1n-0.0638 p vsT Egn(5.22)

1.622 T3 ...
The values of ksr have been taken as: Flat blade turbines k=1.3
Pitched blade turbines k- = 0.75
Hydrofoils kST=0.35
The reason for varying the constant with impeller type is that Fawcett (1989) found that
the tangential component of velocity is largest for impellers which produce a
predominantly radial flow pattern and smallest for impellers which produce a
predominantly axial flow pattern. It must be stressed that these constants are assumed
and further work on the measurement of fluid in
velocities - agitated vessels will be
required to provide better confidence in this analysis.
The shear stress at the wall of the vessel has been calculated for each data point using
equation (5.22). The shear rates and viscosities of the fluids have then been calculated
using the power law, ie. from equation (5.3):
'IwK Yew-i) Egn(5.23)

...
The wall viscosity is to

then used calculate Reynolds and Fourier numbers.
76
5.3 ANALYSIS OF PSEUDO-PLASTIC BLEND TIME DATA
The pseudo-plastic blend time data have been analyzed in the

same way as the
Newtonian data:
i) individual probe NO values plotted versus Reynolds number, based on

the wall viscosity derived in the previous section, to show how the
blending process proceeds in regions of differing agitation intensity.
ii) root mean square NO values plotted versus the same Reynolds number
to show how blending proceeds in the vessel as a whole.
iii) fitting the data to the correlation developed for Newtonian data.
5.3.1 Individual Probe Blend Times
As mentioned in the last chapter, individual probe blend times were measured to
investigate the way in which blending proceeds in regions of differing agitation
intensity.
The results of the pseudo-plastic blending experiments are given in tables 5.1 5.4.
-
The tables show vessel scale, the concentration and grade of Natrosol used, the
individual probe blend times and the root mean square blend time for each set
experimental conditions. The blend time reported is the mean of eight repetitions and
the standard deviation of the mean is given as a percentage in brackets below the blend
time result. This is identical to the way in which the Newtonian data are reported. The
individual probe results are plotted as N6; versus Reynolds number in figures 5.3 - 5.6.
The data reported by Mackinnon (1987), taken in water are included in the plots for
comparison.
The plots are almost identical to figures 4.3 - 4.6 which show NO,.versus Reynolds
number for the Newtonian fluids, ie:
i) Re > Rec,;,: NO1= NO2 = NO3
ii) Re < Recn,: NO1< NO2 < NO3

77
111 the degree of divergence of the mixing times increases with decreasing
Reynolds number.
For this reason, the explanation given in section 4.5.1, for the behaviour of Newtonian
fluids can be applied to the behaviour on pseudo-plastic fluids.
5.3.2 Root Mean Square Blend Times
The root mean square blend times are given in tables 5.1 5.4 and plotted in figures
-
5.7 - 5.10 with the data taken by Mackinnon (1987) in water included for comparison.
The trends observed in the non-Newtonian root mean square blend times are identical
to those observed in Newtonian fluids, ie:
i) Re > Rec,; NO = constant

t:
ii) Re < Rec,;,: NO Re`
The value of x for each impeller is given in table 5.5 with the relevant statistical
information. The results are similar to those found for Newtonian fluids and, as with
the individual probe mixing times, the explanation presented for the blending of
Newtonian fluids can be applied to pseudo-plastic fluids. The fit of data is worse than
that obtained for the Newtonian data and this is probably due to errors in the estimation
of the fluid viscosity.
5.3.3 Correlation of Data
The values of the two dimensionless groups, PoW Re and i/Fo were calculated for all
the Newtonian and pseudo-plastic data points and plotted. The boundary between the
turbulent and viscous mixing regimes occurred at:
1/Fo = 103
as for the Newtonian fluids alone, so the data sets were split at this value.
5.3.3.1 Turbulent Data
The turbulent data have been analyzed by performing a multiplicative regression of NO

78
on Po and DIT in the same way as the Newtonian data. The result of the regression is:
D -1.933
NO po -0.357 Egn(5.24)
T ...
The correlation coefficient, r2, was 82.74% and F-ratio is 121 for 51 data
points. Again,
these statistics are worse than the corresponding values obtained for the Newtonian data.
After rounding-up the data can be described by equation (4.12) but

with the constant
of proportionality changed from 5.20 to 5.47, ie:
5.47
Po l Re = Egn(5.25)
Fo ...
The relative standard deviation of the constant is 13.6%.
5.3.3.2 Viscous Data
The data have been correlated in the same way as the Newtonian data by performing
a regression of:
Po" Re a Fo Egn(4.19)
...
and the results are given in table 5.6.
A very good fit of the data was obtained and, again, the exponents were rounded-up to
-0.5. A comparison of the of the exact and rounded-up correlations are given in table
5.7. Rounding-up does not significantly worsen the standard deviation of the correlation
so the data are presented in this form. All the viscous data (Newtonian and pseudo-
plastic) are correlated by:
193
Po" Re = Eqn(5.26)
VFo ...
The relative standard deviation of the constant is 32.5%.

79
Also the Newtonian and pseudo-plastic data sets for each impeller have been compared
using the "Student's t-test". They are not statistically different at the 95% confidence
level.
The Newtonian and pseudo-plastic data, substituted into the correlation, are plotted for
each impeller in figures 5.11 - 5.14 and for all impellers and fluids in figure 5.15.
5.4 DISCUSSION
5.4.1 Controlling Viscosity
Moo-Young et al. (1972) measured blend times in Newtonian and pseudo-plastic fluids
and plotted NO versus Reynolds number, with the fluid viscosity defined by equation
(2.26). They found that the Newtonian data behaved as expected with:
NO a Re' Eqn(4.17)
...
but the pseudo-plastic data did not fit the same relationship. The exponent on Reynolds
number was more negative than -1 indicating that, when compared on the basis of equal
Reynolds number, blending of pseudo-plastic fluids is more difficult than Newtonian
blending.
Analysis of the Newtonian data, discussed in the previous chapter, showed that the
blending of the whole vessel was controlled by the longest blend time, measured behind
the baffle. Blend times measured in the bulk of the vessel and near the impeller were
always much shorter than the limiting blend time. For this reason, the analysis of the
pseudo-plastic blend time data started with finding a method of estimating the fluid
viscosity in this region of the vessel.
The method used to estimate the wall shear rate and viscosity is based on a torque
balance between the agitator shaft and the baffles and vessel surfaces. This has been
used in the past by Elson (1990) to correlate mixing data taken in yield stress fluids.
The inclusion of the baffles in the analysis has not been done before. If the baffles are
ignored, a worse correlation between the Newtonian and pseudo-plastic data is obtained.
If the data are analyzed ignoring the baffles, the two sets fail the t-test and are
statistically different.
The definition of Fourier number, on the basis of wall viscosity, is sensible because it
80
contains the controlling blend time and the viscosity in the region of the vessel where
this blend time occurs.
Reynolds number is determined by conditions at the impeller and should, therefore,

include the viscosity of the fluid in the impeller region. Defining Reynolds
number
with the wall viscosity gives a ratio of the inertial forces produced by the impeller and
the viscous forces in the fluid in this region of the vessel. This is sensible because,
again, the region of the vessel where the controlling blend time occurs is being
considered.
The fact that a much better correlation between the Newtonian and pseudo-plastic blend
time data has been produced than has appeared in the literature previously indicates that
this is a good model of the process but, as mentioned previously, further work will be
required for its validation.
5.4.2 Comparison using Power Input
As mentioned in the last chapter, the information of most interest to agitator designers
is the type of impeller which gives the best performance in blending fluids for the
lowest power input. The basic conclusions drawn from the Newtonian blend time data
were:
i) for a given process operated at constant geometry (D/7) and scale, there is no
saving to be made in running cost because blend time is dependent on power
input. A large diameter impeller will achieve a desired blend time for less
power than a small impeller.
ii) for a given process, there are potential savings to be made if an impeller with
low is chosen because it must run at a higher speed to input the
a power number
same power as an impeller with a high power number.
These conclusions are true in both the turbulent and viscous regimes.
Equation (5.26) can be re-arranged to show:
W Eqn(5.27)
ealT T-2 ...
eDp
81
This equation is identical to equation (4.29) for Newtonian fluids. The fluid physical
properties do not appear in equation (4.29) because they are constant for a Newtonian
fluid. They must be included when a pseudo-plastic fluid is examined because they
will
vary depending on the impeller type used to blend the fluid. The derivation of the wall
viscosity shows that its value is a function of the torque acting on the agitator shaft -
the higher the torque, the higher the wall shear rate and the lower the wall viscosity.
If a blending duty is considered at constant power input per unit mass and geometry (ie.
D/T is constant) and at one scale the blend time will be determined by the wall
viscosity of the fluid. An impeller which minimizes the wall viscosity at a given power
input will achieve the shortest blend time. When the four impellers tested are compared
at equal power input per unit mass, the highest torque is produced by the T/2 pitched
blade turbine and the results of the blend time measurements show that it is the best
impeller to use for pseudo-plastic mixing duties.
In fact, the results imply that an impeller with a higher power number would be even
more efficient for blending these fluids. Zlokarnik (1967) concluded that, for 0.25 <
1/Fo < 104, the most energy efficient impellers to use are gates and large diameter
paddles. This approach to blending viscous fluids has been pursued by two Japanese
agitator manufacturers who offer wide blade paddle impellers for these duties.
Sumitomo Heavy Industries have the "Max-Blend" impeller and Shinko-Pantec have the
"Full-Zone" impeller which operate on the same principles as the gates and paddles
tested by Zlokarnik (see figure 5.16).
5.4.3 Measurement of Rheology
The results of the blend time experiments show how important the measurement of fluid
rheology is to agitator selection and design.
Equation (5.26) shows how the blend time varies with the viscosity of the fluid at
constant power input per unit mass and impeller/vessel geometry:
Eqn(5.28)
e cc K
cc 'y ' ...
Equation (5.22) and (5.23) can be combined to show the relationship between the fluid's
power law constant and index and the agitator shaft torque at constant scale:
82
Y oc
n-w Eqn(5.29)
..
If equations (5.28) and (5.29) are combined the effect that errors in the
measurement
of the power law constant and index will have on blend time can be examined:
nnl
eaKn Egn(5.30)
aKn
n ...
Equation (5.30) shows that, if the fluid is Newtonian, ie. n=1, the mixing time is
proportional to the fluid viscosity, as expected.
The mixing time in two fluids, 1 and 2, can be compared at constant torque. The ratio
of the blend times will be:
1
) )l
1/ni ni-1 nZ-1
81 KI
-
n ni n2 Egn(5.31)
_ ...
2 K1IN
2
If subscript 1 refers to the true values of 0, K and n for a particular fluid and subscript
2 refers to the measured values of K and n and the predicted value of 0, calculated
from those values, then:
K, = eK K2 Eqn(5.32)
...
and:
nl = e,, n2 Egn(5.33)
...
is
If the measured value of n correct but is
there an error in K, the relationship between
the true and predicted blend times will be:
Egn(5.34)
=e/"6K2 ...
The greater the degree of pseudo-plastic behaviour exhibited by the fluid, the greater
1/n, and the greater the error in the predicted blend time.
the value of the exponent,
For example:
83
if eK = 1.1 and n=0.9, then: 0, = 1.11 02
if eK = 1.1 and n=0.5, then: 01 = 1.21 02
Similarly, if the measured value of K is correct but there is

an error in n, the
relationship between the true and predicted blend times will be:
G Z+CN
K G" n2 C. nl 8 Egn(5.35)
12
L, ...
if e = 1.1 and n=0.9, then: e= K-0.1 A2.02 e

2
if e = 1.1 and n=0.5, then 8= K-o. '8 A3.6 8
z
This analysis shows that errors in the measurement of the power law constant and index
are magnified when the fluid is highly pseudo-plastic and accurate measurement of
these parameters is more important than it is in Newtonian blending.
5.4.4 Data Correlation
The scatter of the data about the correlation of the pseudo-plastic blend time data is
twice as wide as the scatter of Newtonian data about their correlation.
The reason for this is most probably due to errors in the estimation of the fluid
viscosity at the vessel wall, N.,,,.These errors are caused because the method used for
the estimation of the fluid velocity impinging on the baffles is based on measurements
made by Schwartzberg & Treybal (1968) of the fluid velocities produced by a Rushton
turbine in the turbulent regime. No account of the effect of the fluid viscosity on the
velocity has been included because there does not appear to be a rigorous way of doing
this.
Wong & Huang (1988) measured the pumping capacity of Rushton turbines in
Newtonian fluids and found that it is constant when Re >2x 104 and reduces
significantly as the flow regime becomes transitional. Koutsakos et al. (1990) measured
the pumping capacity of Rushton turbines in pseudo-plastic fluids and found that it is
Re > 104 it
and reduces in the range 60 < Re < 104 with Fl -- Re-0.2.
constant when
84
Although there is this disagreement about the way pumping capacity and flow near the
impeller change with changing fluid viscosity, there is no information on the way the
viscosity affects flow in the regions of the vessel away from the impeller.
Also, assumptions have been made about the velocity of the fluid after it has been in
contact with the baffle and the effect of impeller type on the fluid velocity near the
baffles.
In order to overcome the need to make assumptions about the fluid velocities impinging
on the baffles and the effect of impeller type, measurements of the flow in this region
of the vessel must be made. This could be done using laser-Doppler anemometry and
would prove, or disprove, the validity of the model proposed in this thesis for the
blending of viscous, pseudo-plastic fluids.
5.5 CONCLUSIONS
The conclusions which can be drawn from the analysis of the pseudo-plastic blend time
data are:
i) the pseudo-plastic data have been analyzed by estimating the viscosity at the
wall of the vessel. This has been done because this is the region of the vessel
where the overall blending in the is
vessel controlled.
ii) the wall viscosity has been estimated by carrying out a torque balance between
the impeller and the wall of the vessel also taking into account the pressure drop
due to the fluid flow impinging on the baffles. This calculation allows the shear
stress in the fluid at the wall to be estimated from:.
1n - 0.0638 p 2
VST ...
Eqn(5.22)
1.622 T3
iii) knowledge of the fluid's power law constant, K, and index, n, the viscosity
with
is estimated from:
) Eqn(5.23)
W_K ,(W_, ...
85
Equation (5.3) was used to find the relationship between the wall shearrate and
wall shear rate.
Assumptions have been made regarding the fluid velocity impinging the baffles,
on
required to estimated the pressure exerted on the baffles. The work of Schwanzberg
& Treybal (1968), who measured fluid velocities in the turbulent
regime, has been used
to estimate the velocities in the viscous regime. This is an area requiring further work.
When the data taken in the turbulent regime were analyzed in this way, the following
conclusions were drawn:
i)
the pseudo-plastic rheology does not affect the blending of the fluid. All the
Newtonian and pseudo-plastic, turbulent blend time data are correlated by:
5.47
Po "' Re = Eqn(5.22)
Fo ...
The standard deviation of the constant of proportionality is 13.6%.
ii)
This correlation is based on data taken with two pitched blade turbines of T/3
and T/2 diameter, a flat blade turbine of T/3 diameter and Lightnin A310 and
Chemineer HE3 hydrofoils of T/2 diameter. In all experiments the impeller was
positioned T/3 above the base of the vessel.
ii) This relationship can be re-arranged to show that NO is constant and its value
is dependent on the impeller type.
iv)
Individual probe blend times; measured beneath the impeller, halfway between
the shaft and vessel wall and behind the baffle are approximately equal.
The data taken in the viscous regime were also analyzed by substitution into equation
(5.22) and the conclusions drawn were:
i) two mixing regimes exist with the boundary between them occurring at 1/Fo
103.
ii) for 1/Fo > 103 NO = constant

86
for 1/Fo < 103 NO oc Re`
iii) when 1/Fo < 103 the pseudo-plastic blend time data can be correlated by:
200
Po' Re Eqn(5.36)
VF-o ...
The viscous Newtonian and pseudo-plastic correlation constants

are not
statistically different when compared with the "Student's t-test" at the 95%
significance level.
When 1/Fo < 103, all the blend time data taken in the viscous regime can be
correlated by:
193
Pow Re Eqn(5.26)
To ...
ii) This correlation is also based on data taken with two pitched blade turbines of
T/3 and T/2 diameter, a flat blade turbine of T/3 diameter and Lightnin A310
and Chemineer HE3 hydrofoils of T/2 diameter. In all experiments the impeller
was positioned T/3 above the base of the vessel.
111 This relationship can be re--arranged to show that NO is inversely proportional

to Reynolds number.
iv) Individual probe blend times diverge with the longest measured behind the
baffle and the shortest beneath the impeller. The degree of divergence increases
with decreasing Reynolds number.
The two correlations can be re-arranged to show the dependence of blend time on
power input per unit mass, impeller geometry and scale.

87
The conclusion drawn from the turbulent, pseudo-plastic data is identical to that drawn
from the Newtonian data, and is given by:
for 1/Fo > 103 ea1T T2 Egn(4.28)

...
eD
The conclusion drawn from the viscous, pseudo-plastic data is based on the a similar
relationship between blend time and power input deduced from the Newtonian data:
for 1/Fo < 103 1T jf3 W

ea T-z Egn(5.27)
eDp ...
This relationship shows that, when compared at constant power input per unit mass,
geometry and scale, the blend time is proportional to the wall viscosity of the fluid.
Since the wall viscosity is a function of the agitator shaft torque, an impeller with a
high power number will achieve the shortest mixing time when compared at constant
geometry (Dfl.
This conclusion is different to that drawn for Newtonian fluids where impellers of the
same diameter operated at the same power input, geometry and scale achieve the same
blend time. The economic implications of this conclusion are examined in chapter 7
of this thesis.
88
CHAPTER 6
POWER MEASUREMENTS
6.1 INTRODUCTION
It is important to measure power consumed by an agitator during blend time

measurements since this is the way in which their efficiency as blending devices is
compared. The impeller that achieves a desired blend time for the minimum power
input is the most energy efficient and will have the lowest running costs. It is also
important to consider the torque delivered by an agitator since this determines the
mechanical strength of the agitator gearbox, shaft and impellers and therefore the capital
cost of the equipment.
Also, it has been shown in the previous chapter that the power input and torque can be
used to determine the viscosity of the fluid in the baffle/wall region of the vessel and
it is the viscosity of the fluid here that controls the blending duty in the vessel as a
whole.
The method used to measure the impeller power at the 0.609 and 1.830 m scale was
identical to that used by Muskett (1987) and used the system first described by Pollard
(1983). No work has been done as part of this project to improve or modify the power
measurement technique.
The method is briefly described in section 6.2 and the definition of power number is
discussed in section 6.3. The results are presented in section 6.4 and discussed in
section 6.5, including comparison with the results reported by other workers. Finally,
the conclusions are given in section 6.6.
6.2 METHOD
The method used to determine the power consumption of the impellers tested involves
measurement of the shaft rotational speed and torque. Having

the simultaneous
the torque and speed, the power is calculated from:
measured
Eqn(4.41)
P= 2it AN...
89
The rotational speed of the 0.609 and 1.830 m vessel shafts was measured using optical
tachometers and a toothed wheel fixed on the impeller shaft. The measurements were
accurate to 1 rpm at the 0.609 m scale and 0.5 rpm at the 1.830 m scale.
The shaft torque at the 0.609 and 1.830 m scale in both the Newtonian and pseudo-
plastic fluids was measured using semi-conductor strain gauges mounted directly on the
shaft and located above the surface of the fluid under test. Semi-conductor strain
gauges were used because they are more sensitive than foil gauges. Low noise slip
rings were used to transmit the strain gauge sinal from the rotating shaft to the
stationary pick-ups on the agitator frame The signal was filtered, amplified and then
sampled by a PDP 11/23 computer.
The strain gauges were calibrated by applying known loads to a lever arm attached to
the shaft. The shaft of the 1.830 m vessel was calibrated in place with the force applied
to the lever arm measured with a load cell. The shaft of the 0.609 m vessel was
calibrated by removing it from the vessel, mounting it horizontally and adding weights
to the lever arm. Once the calibration procedure had been completed, tests were carried
out with an impeller which would give a known result. The relationship between the
voltage output and applied torque was linear.
The data were analyzed by subtracting the PDP 11 computer's digital output with the
shaft stationary from the output with the shaft rotating and averaging the readings.
Muskett (1987) found that six windows of 2048 data points sampled at 100 Hertz were
sufficient to produce an accurate average value. The filter unit was set at one quarter
of the sampling frequency (on low pass) to remove high frequency noise. The
arrangement was set from the experience reported by Pollard (1983). The impeller
power number was calculated by a sub-routine of the sampling program.
Muskett (1987) reported that the accuracy of the torque measurements was within 2%
and, allowing for error in speed measurements, the power consumption of the agitator
recorded was accurate to within 3%.
At the 0.305 m scale the power consumption was estimated for the measurements made
in viscous Newtonian fluids. The estimation was made using a power number design
(Palmer (1987)). The impeller shaft rotational speed was measured using a hand-
guide
held optical tachometer.
at the 0.305 m scale, was measured in pseudo-plastic fluids

The power consumption,
90
using a "Coesfeld Rheosyst" viscometer which gave a direct reading of the impeller
rotational speed and the torque. The output from the viscometer was a voltage which
was sampled by a Tulip personal computer which was also programed to calculate the
impeller power number from the data. It was purchased after the Newtonian tests had
been completed which is why the power numbers were estimated using the power
number design guide.
6.3. POWER NUMBER
Having measured the torque and rotational speed, the power consumption of the
impeller is calculated from equation (4.41). The power number of the impeller, Po, can
then be calculated from:
Po =P Egn(6.1)
...
pN3D5
The power number can be considered as a drag coefficient (King (1985)), the higher its
value, the higher the power input for a particular combination of impeller diameter and
rotational speed. The power number is determined by several factors:
i) the impeller geometry, ie. the type of impeller and the angle, width,
thickness, length and number of the blades. These variables have been
comprehensively. discussed by Muskett (1987) and for this reason they

are not discussed here.
ii) the vessel geometry, ie. the type and number of baffles, the shape of the
vessel base, the position of the impeller relative to the base and surface
the fluid. None of these factors were varied for the power or blend
of
time measurements.
iii) the fluid viscosity. The flow regimes (turbulent, transitional and
laminar) are defined in terms of the relationship between power number
and Reynolds number.
between Reynolds number and the flow regimes can

The relationship power number,
be summarised by:
91
Turbulent regime: Re>104 Po = constant
Transitional regime: 10<Re<104 Pooc Rea
Laminar regime: Re < 10 Po oc Re-'
The power number - Reynolds number relationship in the turbulent and laminar regimes
is well defined and is the same for all impellers. The relationship in the transitional
regime depends on impeller type: for some it remains at the turbulent value as Reynolds
number decreases and the laminar regime is reached, ie a=0. For others power
number changes as Reynolds number decreases and it may increase or decrease.
The measurements of power made in this study have concentrated on the transitional
regime where the blend time measurements have been made. No power measurements
have been made in the laminar regime.
6.4 RESULTS
The results are given in tables 6.1 - 6.4 with the power number measured in water,
reported by Mackinnon (1987) also included for comparison. The mean power number
in each fluid tested is given with the relative standard deviation. The power numbers
are approximately equal in all the fluids and rises slightly at the high viscosity end of
the range.
Power numbers are plotted versus Reynolds number in figures 6.1 6.4. The definition
-
used to calculate the viscosity of the pseudo-plastic fluids is that given by
of shear rate
Metzner & Otto (1957), ie:
Egn(2.23)
y= KN ...
The value of x used was 11 as reported by Metzner & Otto and others. The definition
of Reynolds number is therefore:
pND2_p N2-n D2 Eqn(6.2)

ReMO =K ...
(11 N)"-' K 11^'
92
Power number was also plotted versus Reynolds number derived from the
wall shear
rate, as discussed in the previous chapter, but the best correlation between the
Newtonian and pseudo-plastic data were obtained with the Metzner & Otto Reynolds
number.
6.5 DISCUSSION
6.5.1 Shear Rate and Reynolds Number
As discussed above, the best agreement between the Newtonian and pseudo-plastic data
was obtained using Metzner & Otto's definition of shear rate. This underestimates the
shear rate in the turbulent regime since it does not take account of the turbulent stresses
in the fluid (Fawcett (1990)). But, at the high end of the Reynolds number range the
power number is constant in Newtonian and pseudo-plastic fluids so it is impossible to

determine which is the correct method of defining shear rates by simply comparing
plots of power number in the two types of fluid.
At the lower end of the Reynolds number range studied, the turbulent stresses would
be reduced and, although power number may rise or fall slightly before reaching the
laminar regime, it does not change significantly from its turbulent value. Power number
must vary sufficiently for a comparison of different shear rate estimation methods to be
made.
The best agreement between the Newtonian and pseudo-plastic power numbers was
obtained using the Metzner & Otto (1957) definition of shear rate, as discussed above.
The lowest Reynolds numbers at which power numbers and blend times have been
measured is approximately 200. This is the Reynolds number at which the power
to
number starts rise significantly and there are very few data points to compare at this
value. Many other workers have used the Metzner & Otto definition of shear rate to
correlate their power data and this is discussed in the following section.
6.5.2 Comparison with Other Workers
Many workers have measured power numbers for the same types of impeller as studied
in this research.
93
Axial flow impellers have been studied by Landau & Prochzka (1964), Havas et at.
(1978b) and Shiue & Wong (1984) among others.
Landau & Prochazka (1964) measured the power number of a marine propeller in a
baffled vessel and found that its value in the turbulent regime was 0.30. As Reynolds
number was decreased by increasing the fluid viscosity, the power number started to
rise at a Reynolds number of around 2x 103.
Havas et al. (1978b) measured power numbers for standard, S. E. M. and VEGYTERV
propellers and found that power number is constant for Re > 300 and starts to rise
slightly at Reynolds numbers below this value. The turbulent power numbers of the
various propellers varied between 0.098 and 0.80 depending on the diameter, clearance
and propeller type. The power number was constant at Reynolds numbers greater than
300 and increased as Reynolds number decreased below this value.
Shiue & Wong (1984) also studied the VEGYTERV propeller and found that, for DIT
= 0.325, the turbulent power number was 0.67. This agreed very well with the value
reported by Havas et al. (1978b). The value at which power number started to rise in
the transitional regime was around 3x 103, a factor of 10 higher than reported by Havas
et al.
The turbulent power numbers reported agree well with the power number measured for
the axial flow, hydrofoil impellers studied in this project which were 0.30 in the
turbulent regime. Landau & Prochazka (1964) and Shiue & Wong (1984) both found
the boundary between the turbulent and transitional regimes occurred at a Reynolds
number of between 2x 103 and 3x 103 which is a factor of 10 higher than the value
reported by Havas et al. (1978b) and in this thesis.
The three workers referred to above also studied pitched blade turbines and generally
the same results were reported, ie. the Reynolds number at the boundary between the
turbulent and transitional regime was similar to that found for the propellers studied.
Landau & Prochzka (1964) found that the power number was 1.45 in the turbulent
Havas et al. (1978b) found that it was around 1.45, depending on the clearance
regime;
impeller the vessel base and Shiue & Wong (1984) found that it was 1.74
of the above
for a four-bladed turbine and 1.20 for a two bladed-turbine. These values agree well
in this project with the power number of the T/3 turbine (w/D =
with those measured
0.2) measured at 1.60 at the 0.609 m scale and increasing to 1.80 at the 1.830 m scale.
for the T/2 turbine (w/D = 0.13) with the power number
A similar effect was observed
94
measured at 1.10 and 1.28 at the 0.609 and 1.830 m scales respectively.
There is less information in the literature on flat blade turbines of the type studied in
this project, even though they are widely used. The bulk of work carried out with
radial flow impellers has been done with Rushton turbines on which the blades are
to
attached a disc. These are more commonly used in gas dispersion duties and are not
commonly found in blending applications. Shiue & Wong (1984) studied a turbine with
curved flat blades which had a power number of 1.56 in the turbulent regime, with Re
>5x 103. The major difference between the measurements reported in this thesis and
those in the literature is the behaviour of the power number in the transitional regime.
Figure 6.3 shows that the power number of the flat blade turbine was constant over the
range of Reynolds numbers studied. Shiue & Wong found that the power number of
the curved blade turbine falls in the transitional regime reaching a minimum value of
about 0.8 at Re = 103 and rising as Reynolds number reduces further. In fact the power
number curve of the flat blade turbine is very similar to those of the pitched blade
turbines studied here and reported in the literature.
6.6 CONCLUSIONS
The conclusions which can be drawn from the results of the power measurements are:
i) power number is constant over most of the range of speeds and fluids tested.
It rises slightly at the low end of the range of Reynolds numbers studied.
ii) the power number of the pitched blade turbines was observed to rise on scale-
up. This may be due to slight geometrical differences between the two scales.
iii) Reynolds number, the best agreement between the

when plotted versus
Newtonian and pseudo-plastic data was obtained using the Metzner & Otto
(1957) definition of shear rate. Reynolds number is given by:
N2-' D2
Egn(6.2)
ReMO -p ...
K 11"-'
trends observed in their behaviour agree very well

iv) the power numbers and the
with those reported in the literature.
95
CHAPTER 7
ECONOMIC ANALYSIS
7.1 INTRODUCTION
The results reported and discussed in chapters 4 and 5 relate the blend time in an
agitated vessel to the power input, torque, impeller-vessel geometry (D/T ratio), fluid
physical properties and scale of operation. The objective of the work reported in this
is
chapter to take that technical data and extract economic information on which types
of impeller will give the most cost-effective operation over the working life of an
agitator.
The capital cost of an agitator can be related to the torque it must deliver since this
determines the mechanical strength of the gearbox, shaft and impellers. The greater the
torque, the greater the mechanical strength and, hence, the greater the capital cost.
Similarly, the running cost of an agitator can be related to the power required to achieve
the desired blend time and the annual cost of maintenance. Maintenance costs can be
related to the capital cost and are added to the power cost to give a total running cost
for a years operation.
The method used to carry out the economic analysis is described in detail in section 7.2
and the results of a specific example are given in section 7.3. The results are discussed
in section 7.4 and conclusions are given in section 7.5.
7.2 METHOD
The operating cost of an agitator can be divided into two parts:
i) the capital cost which is the financial outlay required to purchase the
equipment from a manufacturer.
ii) is the combination of the cost of buying power

the running cost which
and the cost of maintaining the equipment.
In the economic analysis, the capital cost has been considered as a one-off payment at
life. The running cost has been calculated for each
the start of the agitator's operating
96
year, as the sum of the cost of the electricity required to drive the motor and the cost
of maintenance, as a fraction of the capital cost, discounted to give a Net Present Value
at the start of the operating life.
7.2.1 Capital Cost
The capital cost of an agitator is related to the torque which the agitator will deliver.
Rigden (1991) has provided data relating the approximate selling price of agitators to
the torque. The data is based on an agitator fitted with a single impeller having all
parts in contact with the fluid constructed of 316 Stainless Steel. The prices also
include a cartridge mechanical seal fitted to the drive. A multiplicative regression has
been carried out on the data, which is given in table 7.1, to give the following
relationship between capital cost (in pounds sterling), Co, and torque (in Nm), A.
ap,
Co = 488 A 4M Egn(7.1)
cap ...
The correlation co-efficient, r2, of the regression is 97.74%.
The torque is related to the power consumption and rotational speed of the impeller by
re-arranging equation (4.41):
PopN2D5 Eqn(7.2)
A=P= ..
2it N 2n
7.2.2 Running Cost
The running cost of the agitator is made up of two components:
i) the maintenance cost.
ii) the power cost.
The maintenance cost of equipment used in the chemical industry can be estimated as
fraction its (Weaver & Bauman (1974)). The fraction can vary
a the capital cost
between 4 and 10% depending on the environment in which the equipment is situated
it requires to keep it working. Agitators are often situated
and the amount of attention
in hostile environments and it has been assumed that the maintenance cost is at the
97
upper end of the scale, ie:
CO,,, = 0.10 Cocap Egn(7.3)

n ...
The power required to drive an agitator is given by re-arranging equation (6.1):
P=PopN3D5 Egn(7.4)
...
Having calculated the power consumption of the agitator, this can be related to the
if
annual power cost the number of hours the agitator runs each year, Hpy, and the cost
of a unit of electricity, Cokes,,,are known, ie:
Co Hrs CokWh 10-5

=P x Egn(7.5)
pow ...
The factor 10-5must be introduced because the power, P, is calculated in Watts and
must be converted to kiloWatts and the cost of a unit of electricity, Cokwh,is given in
pence per kiloWatt-hour and must be converted to pounds per kiloWatt-hour.
The total running cost for one years operation can be calculated by combining equations
(7.3) and (7.5):
Corun = Comain + Copow Eqn(7.6)

...
The total cash flow, Cf, for the agitator during its operating life, Yrs, will be:
Cf = Co Yrs Egn(7.7)
run ...
One of the methods of assessing the value of an investment is to discount the cash flow
back to the start of the project to obtain a Net Present Value, Npv. This is done by
multiplying the cash flow by a discounting factor. The discounting factor, Df, used
assumes that the cost of capital (the interest rate, Ir) is constant over the life of the
equipment and is described by Peters & Timmerhaus (1968):
(1 - exp (-Ir Yrs)) Eqn(7.8)

Df -1 ...
Ir Yrs
98
The net present value of the agitator's running cost, NpvQ, can now be calculated from:
Npv, = Cf Df = Co, Yrs Df Egn(7.9)

un un ...
Having calculated the capital and discounted running costs of the agitator at the start
of its operating life the total net present value, Npvto,, of the agitator can be calculated
by combining equations (7.1) and (7.9):
Npvlot = CO + Npvrun Egn(7.10)

cap ...
7.2.3 Analysis
The equations given in the preceding sections relate the cost of agitation to the power
consumption and the shaft torque. These two parameters have been calculated for a
hypothetical blending duty in fluids exhibiting different degrees of pseudo-plasticity
using the "tk-Solver" computer program.
"tk-Solver" is a declarative, rule-based programming language which allows the user

to work directly with a particular set of equations without having to program them in
a particular sequence. Complex relationships, such as the one between the blend time
in an agitated vessel, the agitator power and torque and the running and capital costs
of the equipment, can be investigated quickly without resorting to time consuming
computer programing and mathematics.
The analysis of the blend time data in chapters 4 and 5 has shown that a large impeller
will achieve a certain blend time for less power input than a small one so the impeller
diameter has been fixed at D= T/2.
The parameter which is a function of impeller type is power number, Po, so this has
been varied between 0.3 and 3.0 in ten steps of 0.3. This corresponds to the range of
power numbers studied in the blend time experiments with the power number of the
hydrofoils 0.3, the flat blade turbines equal to 3.0 and the pitched blade
equal to
turbines equal to 1.0 and 1.70.
The blend time was set at 100 seconds and the diameter of the vessel and fluid depth
5.0 operating volume of 91.3 m3.

at m, giving an
99
The viscosity of the fluid was estimated by calculating the shear stress in the fluid at
the wall of the vessel, a described in section 5.2. The effect of the pressure drop due
to the fluid impinging on the baffles has been ignored in this analysis because its effect
is small relative to the shear stress term. So the shear stress is given by:
Tw=A Egn(7.11)
1.622 T3 ...
The shear rate is then calculated from equation (5.3) and the viscosity from equation
(5.23).
The power law index, n, has been varied between 0.4 and 1.0 in steps of 0.1 and the
power law constant, K, selected so that the viscosity at the wall, .,, is 2.00 Ns m"2
when agitated by an impeller with power number equal to 1.0. The combinations of
constant and index are given in table 7.2.
Having set the impeller diameter, required blend time, fluid physical properties and
scale of operation, the program calculated the impeller speed required to achieve this
blend time for each value of power number by solving:
193
Po' Re Eqn(5.26)
...
Fo
D2 Egn(7.12)
where: Re =PN ...
-1w
e Egn(7.13)
and: Fo _
= '" ...
p T2
It has been
The only other information required relates to the economic analysis.
for 8000 hours a year (since this assumption is
assumed that the agitator will operate
in of equipment in the chemical industry), its life will
commonly used the assessment
be 11 years, the interest rate will be 11% (Ir = 0.11) and the cost of a unit of electricity
be 6 kiloWatt-hour (Eastern Electricity (1991)).

will pence a
100
A full list of the equations used in the "tk-Solver" program and the
values assumed for
each variable are given in appendix 1.
7.3 RESULTS
The equations listed in the previous section have been solved by the "tk-Solver"
program giving a list of capital cost, discounted running cost and total cost versus
power number for the seven fluids. The results are presented as plots of capital cost
and the net present value of the running and total costs versus power number in figures
7.1 - 7.7.
The capital cost of the agitator for all the fluids is around k 30 40 and contributes
-
between 20 and 25% of the total cost. The capital cost is lower for the low power
number impellers and rises as power number increases. The maintenance cost also
increases since this is calculated as 10% of the capital cost.
In the Newtonian fluid (K = 2.00, n=1.0) the total net present value rises from k 142
to k 164 as power number rises from 0.3 to 3.0, shown in figure 7.1. The power cost
is constant and the increase in running costs is caused by the increased torque
increasing the maintenance costs. So, in the Newtonian fluid, there is a financial
advantage in using the lowest power number impeller because it minimizes the capital
cost of the agitator and it is k 13 cheaper than the agitator with the highest power
number impeller.
The total net present value in the first pseudo-plastic fluid (K = 2.67, n=0.9) also rises
from k 151 to k 160 but the net present value of the running costs, Npv, falls from
k 121 to k 118 as the effect of increasing the impeller's power number reduces the
viscosity of the fluid and makes the blending duty easier. This is plotted in figure 7.2.
Again, the lowest power number impeller minimizes the total net present value but the
difference between this and the agitator with the highest power number is reduced
because the running costs are reduced.
Figure 7.3 shows the results for the next fluid (K = 3.55, n=0.8). The total net present
value changes from k 156 to k 152, a difference of k 4, as the power number
increases and the increase in capital cost due to the increasing torque is more or less
balanced by the decreasing power costs due to the reduced apparent viscosity of the
fluid.
101
Figures 7.4,7.5,7.6 and 7.7 show the effect of further increases in the fluids' degree
of pseudo-plasticity. The greatest difference between the total net present value for an
agitator with a low and high power numbers is found in the most pseudo-plastic fluid
(K = 11.22, n=0.4) where the total net present value falls from k 189 for a power
number of 0.3 to k 131 for a power number of 3.0, a difference of k 58, shown in
figure 7.7.
7.4 DISCUSSION
7.4.1 Comparison of Experimental and Economic Results
The results of the blend time measurements in Newtonian fluids show that, when
compared at equal power input, geometry (D/7l and scale, all the impellers tested give
the same performance, ie. achieve the same blend time. An impeller with a low power
number will have to run at a higher speed in order to input the required power and
would, therefore, operate at lower torque than an impeller with a high power number
it
and was suggested that, as a consequence, savings could be made in the capital cost
of the agitator. A similar examination of the blend time measurements in pseudo-plastic
fluids show that there is an advantage in using an impeller with a high power number
because, for a given power input, it would produce a lower viscosity making the fluid
easier to blend. The analysis of the costs of a hypothetical blending duty carried out
in the previous section show that the experimental results can be translated into
financial benefits.
The capital cost of the agitator varies between about k 30 and k 40 as the power
number of the impeller increases in all the fluids examined. In a real situation it is
more likely that the capital cost would be more or less constant as the range of torques
may be accommodated by one size of gearbox. If the torque exceeded the limit of one
it
gearbox would be to
necessary use the next size up in which case there would be a
step in the capital cost curve.
The net present value of the running costs varies much more and ranges between k 92
k 156. The largest difference is found in the most pseudo-plastic fluid and the
and
lowest, in the Newtonian fluid.
is in
The difference in the total net present value also greatest the most pseudo-plastic
increase in capital cost is
fluid but it is smallest in the fluid with n=0.8 where the
less by the decrease in running costs.
more or cancelled out
102
So, in general the experimental and cost analysis agree that there
are advantages in
using a high power number impeller in pseudo-plastic fluids provided that If
n<0.8.
the fluid exhibits slight pseudo-plasticity, using an impeller with a low power number
may produce a cheaper overall agitator than an impeller with a high power number.
7.4.2 Power Number
The results predict that using an impeller with an even higher power number would
reduce the cost of agitation even further. To test this, the results have been extrapolated
outside the range of power numbers tested.
In the most pseudo-plastic fluid (K = 11.22, n=0.4) when the power number is 3.0,
the capital cost of the agitator is k 38.6, the net present value of the running costs is
k 92.2 and the total net present value is k 130.8. If the power number is increased
to 5.0, the capital cost rises slightly to k 39.2, the running costs fall to k 65.1 and the
total costs fall to k 104.3. Similarly, if the power number is further increased to 7.0,
the capital cost rises to k 40.1, the running costs fall to k 60.0 and the total costs fall
to k 100.1.
So, increasing the power number from 3.0 to 7.0 will reduce the total cost of agitation
by k 30.7 over the life of the plant. This can be achieved by using multiple impellers
or one wide blade impeller such as a gate or paddle. As mentioned in the previous
chapter, two Japanese agitator manufacturers have started marketing wide blade paddle
impellers for viscous blending duties and the cost advantages predicted by this analysis
should be obtained if they are used in highly pseudo-plastic fluids.
7.5 CONCLUSIONS
The conclusions which can be drawn from the cost-benefit analysis are:
i) the results of the economic analysis support the findings of the experimental
ie. blending of Newtonian fluids is achieved most cost-
programme of work,
low power number impeller while blending of pseudo-plastic
effectively using a
fluids should be carried out using a high power number impeller.
ii) for the hypothetical example discussed, increasing the power number from 0.3
to 3.0 in a Newtonian fluid increased the total net present value of the agitator
by k 22 from k 142 to k 164.
103
iii) for the hypothetical example discussed, increasing the power from 0.3
number
to 3.0 in a fluid with power law index, n, of 0.4 reduced the total net present
value of the agitator by k 58 from k 189 to k 131.
iv) the results predict that pseudo-plastic fluids should be agitated with impellers of
high power number. The results have been extrapolated for the most pseudo-
plastic fluid and show that if the power number is increased from 3.0 to 5.0, the
total net present value of the agitator is reduced by k 26 from k 141 to k
104. If the power number is increased from 3.0 to 7.0 greater savings are
possible with the total net present value of the agitator is reduced by k 31 from
k 141 to k 100.
v) the results predict that multiple impellers or wide blade gates or paddles should
be used to blend pseudo-plastic fluids because they have a high shaft torque and
significantly reduce the running costs of the agitator.
104
CHAPTER 8
CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK
8.1 INTRODUCTION
The purpose of this chapter is to summarize the conclusions of the work

reported in the
previous chapters of this thesis. The conclusions of the blend time measurements and their
analysis are discussed in section 8.2 The recommendations for future work are given in
section 8.3.
8.2 CONCLUSIONS
8.2.1 Newtonian Blend Times
Blend times were measured in standard geometry vessels (see figure 3.1) at three scales,
0.305,0.609 amd 1.830 m, using three impeller types commonly found in the chemical
industry. These were pitched and flat blade turbines and hydrofoils. The fluids were
Newtonian and ranged in viscosity between 10 and 500 cP. Also, the data reported by
Mackinnon (1987), taken in water, were re-analyzed.
The blend time measurement technique used three conductivity probes placed in regions
of differing
agitation intensity and the individual probe blend times were recorded and a
combined Root Mean Square blend time calculated.
It was found that each impeller has a critical Reynolds number, above which NO is constant
and below which NO increases with decreasing Reynolds number. The lower the impellers
power number, the higher the criticial Reynolds number. Also, when Re > Re,,;, the
,
individual probe blend times were approximately equal. When Re < Recritthe individual
probe blend times diverge with the shortest measured beneatht the impeller and the longest
behind the baffle.
The data for all the impellers were combined and correlated by three dimensionless groups.
These were power number, Po, Reynolds number, Re and Fourier number, Fo.
105
The data could be divided into two groups. The turbulent data were taken at 1/Fo > 103
and the viscous data taken at 1/Fo < 103.
The turbulent data were correlated by:
5.20
Po '13Re = Egn(4.12)
Fo ...
Similarly, the viscous data were correlated by:
183
Po', " Re = Eqn(4.22)
...
Fo
Both equations were re-arranged to show that the blend time is determined by the power
input per unit mass, impeller to vessel diameter ratio and the scale of the vessel.
In turbulent blending:
1I T
ea T2 Egn(4.28)
...
ED
This shows that, for a given power input and impeller geometry, blend time increases with
increasing scale.
In viscous mixing:
` T ...
Egn(4.29)
e1a_ eD T-Z
This shows that, for a given power input and impeller geometry, blend time decrease with
increasing scale which is a surprising result.
106
The analysis of the effect of power input on blend time shows that there are potential
savings to be made in the capital cost of an agitator if an impeller with a low power
number is chosen. The reason for this is that it will run at a higher speed than an impeller
with a high power number to deliver the same power to the fluid and will therefore have
a lower torque on its shaft. Less mechanical strength is required to deliver the torque so
potential savings may be made in the capital cost of the agitator gearbox and shaft.
There are no savings to be made in running cost because a certain power will be required
to achieve a given mixing time, whatever impeller type is used.
8.2.2 Pseudo-plastic Blend Times
The same impellers and vessels were used to study blending in pseudo-plastic fluids. The
fluids used had a minimum power law index, n, of 0.55 and the power law constant ranged
from 0.014 to 1.18.
The Newtonian blend time measuremnts had shown that the overall blending in the vessel
was controlled by the time taken to achieve homogeneity behind the baffle so it was
proposed that the viscosity in this region controlled the blending process. The apparent
viscosity of the fluids was determined by estimating the average shear stress in the fluid
at the surface of the vessel wall and base and this was done by taking an equation given
by Bird et al. (1960) relating the torque to the tangential velocity gradient of the fluid at
the vessel wall and base and the pressue drop due to the fluid impinging on the baffles to
the shaft torque. Assumptions were made in order to estimate the velocity of the fluid at
the baffle using an equation proposed by Schwartzberg & Treybal (1968). The apparent
viscosity of the fluid was calculated for each data point.
The shear stress at the wall of the vessel is given by:
n- 2 Egn(5.22)
iw1.1 0.0638 p VST ...
622 T
where:
107
kST D2
vST
_N Egn(5.21)
(T 2H) 1/3 ...
When the data were analyzed in this way identical trends to those exhibited by the
Newtonian data were observed, ie. at Re > Rec, NO is constant and the individual
it, probe
blend times are approximately equal. When Re < Recril, NO increases with decreasing
Reynolds number and the individual probe blend times diverge.
The turbulent data were correlated by:
5.47
Po '13 Re = Egn(5.22)
Fo ...
Similarly, the viscous data were correlated by:
200
Po 1/3Re = Egn(4.22)
...
Fo
In the same way as the Newtonian data, both equations were re-arranged to show that the
blend time is determined by the power input per unit mass, impeller to vessel diameter ratio
and the scale of the vessel.
In turbulent blending:
ea T2 Egn(4.28)
_T ...
D
This shows that, for a given power input and impeller geometry, blend time increases with
increasing scale.
In viscous mixing:
108
81Ttw T-2 Egn(5.27)

_Dp
In turbulent mixing the same rules regarding the relationship between

power input and
blend time apply but in viscous mixing there is an effect of the fluid's
viscosity at the wall
of the vessel. An impeller with a high power number, and shaft torque,
will produce a
lower viscosity and therefore, a shorter blend time for a given power input.
Thus there are savings in running cost to be made if a high power number impeller is
used
to blend a pseudo-plastic fluid.
8.2.3 Economic Analysis
The blend time correlations were programmed into the "tk-Solver" package with rules
relating the capital cost of the agitator to its torque and the running cost to its power and
maintenance requirement.
This shows that, for pseudo-plastic fluids, significant savings in the total operating cost of
an agitator are possible if a large power number impeller is used. The increased capital
cost due to the higher torque is offset by the reduced power and running cost required to
achieve the desired blend time.
8.3 RECOMMENDATIONS FOR FUTURE WORK
8.3.1 Vessel Geometry
All the blend time and power measurements reported in this thesis have been carried out
in dished base vessels of 0.305,0.609 and 1.830 m diameter fitted with four plate baffles
T/12 wide, mounted T/60 off the vessel wall. In all cases the fluid depth was equal to the
diameter. This is shown in figure 3.1.
The experiments show that the greatest increase in blend time, as the fluid viscosity
increases, occurs behind the baffles. This indicates that one way of improving the overall
blending in the vessel would be to reduce the baffling by, for example, reducing the baffle
width or removing them altogether.

109
The use of multiple impellers for blending viscous fluids has been tested by Metzner
et al.
(1961) and was found to be more effective than single impellers, especially for
small
diameter impellers.
Thus, a new area for blend time measurement would be to compare the standard
geometry
vessel's blending performance with vessels of other geometries including multiple impellers
and those with reduced baffling.
8.3.2 Fluid Physical Properties
There are two areas which would benefit from study of the effect of the fluids' physical
properties on blend time. These are other rheological properties and fluids with different
viscosities and densities.
8.3.2.1 Rheological Properties
Only two rheological properties were studied in this research programme. They were
Newtonian and pseudo-plastic. Very few viscous fluids exhibit a Newtonian rheology and
is
pseudo-plasticity the commonest non-Newtonian rheology found in the chemical industry.
It was important to measure blend times in Newtonian fluids since this gave a base-line
against which the non-Newtonian properties of the pseudo-plastic fluids could be compared.
Among the more common rheologies are visco-elasticity which is encountered in polymer
processing and Bingham plastics, or yield stress fluids, which are often high concentration
slurries such as titanium dioxide and coal. In both cases, knowledge about the best type
of impeller to use is important as additives are often blended into the slurries for processing
reasons.
The blend time measurement technique used in this study could be applied to many other
types of fluid provided care was taken to ensure that a large stagnant zone did not form
around the conductivity probe.

110
8.3.2.2 Physical Property Differences
One of the commonest mixing problems in the chemical industry is the introduction of a
small quantity of fluid into a larger quantity with significantly different viscosities and/or
densities. The aim of the study should be to identify the best type of impeller to use for
this application and to find the best place to introduce the additive.
8.3.3 Measurement of Flow
In the pseudo-plastic data analysis it was assumed that the volumetric flow and flow
patterns were unaffected by the viscsoity of the fluid. The fact that the pseudo-plastic data
give a worse fit to the correlation than the Newtonian data shows that this assumption is
probabley wrong. Very little work has been done to measure the change in pumping
capacity with viscosity, especially in pseudo-plastic fluids and no work has been done to
measure the velocities near the vessel wall, base and the baffles.
It would be possible to use laser-Doppler anemometry to measure the flow and turbulence
in regions of the vessel away from the impeller and this would provide useful data to use
in the re-analysis of the blend time data reported here and would provide insights into the
phenomena such as the pick-up of solids from the base of a
mechanisms of other mixing
in
vessel solid-liquid mixing applications.
111
REFERENCES
Beckner, J. L. & Anchor agitated systems: power input with Newtonian

and pseudo-
Smith, J. M. plastic fluids,
Trans.I. Chem.E., vol. 44, no. 6, pp. T224 T263,1966.
-
Biggs, R. D. Mixing rates in stirred tanks,
A. I. Ch.E.J., vol. 9, no. 5, pp. 636 640,1963.
-
Bird, R. B., Stewart, W. E. Transport Phenomena
& Lightfoot, E. N. Ch. 6, p. 205, Wiley International Edition, John Wiley & Sons Inc.,
New York, 1960.
Blasinski, H. & Rzyski, E. Mixing of non-Newtonian fluids with turbine,

propeller and paddle
agitators,
1nt.Chem. Eng., vol. 16, no. 4, pp. 751 754,1976.
-
Blasiiiski, H. & Rzyski, E. Mixing of non-Newtonian fluids with anchor, screw and ribbon
mixers,
Int. Chem. Eng., vol. 18, no. 4, pp. 708 711,1978.
-
Boger, D. V. Separation of shear thinning and elastic effects in
experimental
rheology,
Rheology (vol. 1) pp. 195 218, Astarita, Marrucci and Nicolais
-
(eds.), Plenum Publishing Corp., New York, 1980.
Bourne, J. R. The mixing of powders, pastesand non-Newtonian fluids,

Chem.Engr., pp. CE202 CE216, September 1964.
-
Bourne, J. R. An analysis of the flow produced by helical ribbon impellers,
Trans. I. Chem. E., vol. 47, pp. T1I 1969.
-,
Bourne, J. R., Power and heat transfer to agitated suspensions:use of heat flow
Buerli, M. & calorimetry,
Regenass, W. Chem. Eng. Sci., vol. 36, pp. 782 784,1981.
-
Bowen, R. L. Agitation intensity: key to scaling up flow-sensitive liquid systems,

Chem.Engng., pp. 159 - 168,8 March 1985.
Bowen, R. L. Unravelling the mysteries of shear sensitive mixing systems,

Chem.Engng., pp. 55 - 63,9 June 1986.
Brennan, D. J. & Impeller mixing in vessels: experimental studies on the influence of

Lehrer, I. H. some parameters and the formulation of a general mixing time
equation,
Trans. I. Chem. E., vol. 54, pp. 139 - 152,1976.
Brito-de la Fuente, E. Mixing and circulation times in rheologically complex fluids,

Leuliet, J. C., Choplin, L. & Fluid Mixing IV, I. Chem.E. Symposium Series no. 121, pp. 75 - 96,
Tanguy, P. A. Bradford, England, 11 - 13 September 1990.
Bryant, J. & Sadeghzadeh,S. Terminal mixing times in stirred tanks,

Proc. 4`' Euro. Conf. on Mixing, paper B4, Noordwijkerhout, Holland,
BHRA, 27 - 29 April 1982.
112
Calderbank, P. H. & The prediction of power consumption in the agitation of non-

Moo-Young, M. B. Newtonian fluids,
Trans.I. Chem.E., vol. 37, pp. 26 33,1959.
-
Calderbank, P. H. & The power characteristics of agitators for the mixing of Newtonian
Moo-Young, M. B. and non-Newtonian fluids,
Trans.I. Chem.E., vol. 39, no. 5, pp. 337 347,1961.
-
Carreau, P.J., Mixing of visco-elastic fluids with helical ribbon agitators,

Patterson, I. & Yap, C. Y. Can.J.Chem.Eng., vol. 54, pp. 135 142,1976.
-
Chavan, V. V., Arumugam, M. On the influence of liquid elasticity on mixing in a vessel agitated by
& Ulbrecht, J. a combined ribbon-screw impeller,
A. I. Ch.E.J., vol. 21, no.3, pp. 613 - 615,1975.
Chemineer Liquid Agitation,

Reprinted from Chem. Engng., M`Graw Hill Inc., New York, 1976.
Cheng, D. C. -H. & The effect of agitator geometry and speed on yield in batch reactors,
Tookey, D. J. Proc. 2"d Euro. Conf. on Mixing, paper Al, Cambridge, England,
BHRA, 30 March -1 April 1977.
Corpstein, R. R., Dove, R. A. Stirred tank reactor design,

& Dickey, D. S. Chem. Eng. Prog., pp. 66 - 74, February 1979.
Coyle, C. K., Hirschland, H. E., Mixing in viscous liquids,

Michel, B. J. & Oldshue, J. Y. A. I. Ch.E.J., vol. 16, no. 6, pp. 903 - 906,1970.
Demange, A. L., Batch mixing times of viscous pseudo-plastic liquids,

Forsyth, T. H. & Focht, L. G. Elastomerics, pp. 33 - 37, April 1977.
Dickey, D. S. Program chooses agitator,

Chem.Engng., pp. 73 - 81,9 January 1984.
Eastern Electricity Tariffs for the supply of electricity,

Eastrn Electricity plc., 15tApril 1991.
Elson, T. P. Mixing of fluids possessinga yield stress,

Proc. 6' Euro. Conf. on Mixing, pp. 485 492, Pavia, Italy, BHRA,
-
24 - 26 May 1988.
Elson, T. P. The growth of caverns formed around rotating impellers during the
mixing of a yield stress fluid,
Chem.Eng.Commun., vol. 96, pp. 303 - 319,1990.
Fawcett, N. J. S. The mean flow in the discharge of pitched blade turbines,

FMP Interim Report 1046, October 1989.
The turbulence generated at the discharge of pitched blade turbines,

Fawcett, N. J. S.
FMP Interim Report 1059, January 1990.
A technical and economic evaluation of impellers for use on the

Flagg, P. L., Hill, D. J.,
Molineux, K. J. & Sampson, N. viscous regime,
Mixing, A3, Cambridge, England,
Proc. 2"d Euro. Conf. on paper
BHRA, 30 March -1 April 1977.
113
Ford, D. E., Mashelkar, R. A. Mixing times in Newtonian and non-Newtonian fluids,

& Ulbrecht, J. Process Tech.Int., vol. 17, no. 10, pp, 803 807,1972.
-
Ford, D. E. & Ulbrecht, J. Influence of rheological properties of polymer solutions
upon mixing
and circulation times,
Ind.Eng.Chem., ProcessDes.Dev., vol. 15, no. 2, pp. 321 326,1976.
-
Fox, E. A. & Gex, V. E. Single phase blending of liquids,
A. I. Ch.E.J., vol. 2, no. 4, pp. 539 544,1956.
-
Garrison, C. M. How to design and scale mixing pilot-plants,
Chem.Engng., pp. 63 70,7 February 1983.
-
Godleski, E. S. & Smith, J. C. Power requirements and blend times in the agitation of pseudo-plastic
fluids,
A. I.Ch.E.J., vol. 8, no. 5, pp. 617 620,1962.
-
Gray, J. B. Batch mixing of viscous liquids,
Chem.Eng.Prog., vol. 59, no. 3, pp. 55 59,1963.
-
Greene, H. L., Carpenter, C. & Mixing characteristics of an axial impeller with Newtonian and non-
Casto. L. Newtonian fluids,
Proc. 4' Euro. Conf. on Mixing, paper D1, Noordwijkerhout,
Holland, BHRA, 27 - 29 April 1982.
Hall, K. R. & Godfrey, J. C The effect of fluid properties and impeller design on the performance
of helical ribbon mixers,
Chemeca '70, pp. 111 - 127, Butterworths, 1971.
1-Iavas,G., Sawinsky, J. & Investigation of the homogenisation efficiency of the screw agitator,
Dedk. A. helical ribbon agitator, and the multi-paddle agitator in the mixing of
high viscosity Newtonian liquids,
Period.Polytech., vol. 22, pp. 317 - 330,1978 (a).
Havas, G., Sawinsky, J., Investigation of the homogenisation efficiency of various propeller
Dedk, A, & Fekete, A. agitators types,
Period. Polytech., vol. 22, pp. 331 - 343,1978 (b).
Hemrajani, R. R., Chang, M., Mixing and heat transfer in non-Newtonian coal liquids,
Florez, M. R. & Patel, R. D. Proc. 4' Euro. Conf. on Mixing, paper D2, Noordwijkerhout,
Hickman, A. D. Agitation, mixing and mass transfer in simulated high viscosity

fermentation broths,
PhD Thesis, University of Birmingham, 1985.
Hiraoka, S. & Ito, R. Simple relationship between power input and mixing time in turbulent
agitated vessel,
J.Chem.Eng.Japan, vol. 10, no. 1, pp. 75 - 77,1977.
Hcker, H., Langer, G. & Power consumption of stirrers in non-Newtonian liquids,

Werner, U. Ger.Chem.Eng., vol. 4 pp. 113 - 123,1981.
Hoogendoorn, C. J. & Model studies on mixers in the viscous flow region,

Den Hartog, A. P. Chem.Eng.Sci., vol. 22, pp. 1689 - 1699,1967.
114
Ide, Y. & White, J. L. Rheological phenomena in polymerisation reactors:

rheological
properties around agitators in polystyrene-styrene solutions,
J.App. Polymer Sci., vol 18, pp. 2997 3018,1974.
-
Johnson, R. T. Batch mixing of viscous liquids,

Ind.Eng.Chem., ProcessDes.Dev., vol. 6, no. 3, pp. 340 345,1967.
-
Kamiwano, M., Yamamoto, K. Mixing performance of various agitators,
& Nagata, S. Kagaku Kogaku, vol. 31, pp. 365 372,1962.
-
Kelkar, J. V., Mashelkar, R. A. On the rotational visco-elastic flows around simple bodies and
& Ulbrecht, J. agitators,
-
Khang. S. J. New scale-up and design criteria of stirrer agitated batch and flow
mixing vessels,
PhD Thesis, Oregon State University, 1975.
Khang, S. J. & Fitzgerald, T. J. A new probe and circuit for measuring electrolyte conductivity,
Ind. Eng. Chem. Fund., vol. 14, pp. 208 213,1975.
-
Khang, S. J. & Levenspiel, O. New scale-up and design method for stirrer agitated batch mixing
vessels,
Chem. Eng. Sci., vol. 31, pp. 569 577,1976.
-
Khang, S. J. & Levenspiel, O. The mixing rate number for agitator stirred tanks,
Chem. Engng., pp. 141 - 143,11 October 1976.
King, R. Mixing in the process industries,

Hamby, Edwards & Nienow (eds.), ch. 14, p. 253, Butterworths,
London, 1985.
Komori, S. & Murakami, Y. Turbulent mixing in baffled stirred tanks with vertical blade
impellers,
A. I. Ch.E.J., vol. 34, no. 6, pp. 932 - 937,1988.
Koutsakos, E., Nienow, A. W. & Laser anemometry study of shear thinning fluids agitated by a
Dyster, K. N. Rushton turbine,
Fluid Mixing IV, I. Chem. E. Symposium Series no. 121, pp. 51 - 73,
Bradford, England, 11 - 13 September 1990.
Kramers, H., Baars, G. M. &A comparative study of the rate of mixing in stirred tanks,
Knoll, W. H. Chem. Eng. Sci., vol. 2, pp. 35 - 42,1953.
Landau, J. & Prochzka, J. Studies on mixing XI:experimental methods for following the
homogenisation of miscible liquids in rotary mixers,
Coll. Czech. Chem. Commun., vol. 26, pp. 1976 - 1990,1961.
Studies on mixing XIV: homogenisation of miscible liquid in the

Landau, J., Prochzka, J.,
Vclavek, V. & Fort, I. viscous region,
Coll. Czech. Commun.,
Chem. vol. 28, pp. 279 - 292,1963.
Landau. J., & Prochzka, J. Studies on mixing XVII: relation between the energy consumption
and rate of homogenisation,
Coll. Czech.Chem.Commun., vol. 29, pp. 1866 - 1876,1964.
115
Mackinnon, C. P. Research review: comparative mixing times,

FMP Report 029, April 1987.
Magnusson, K. Effektbehovets beroende av den effektiva viskositeten

vid omrring
av strukturvisksa vtskor,
IVA, vol. 23, no. 2, pp. 86 91,1952.
-
M`Manamey, W. J. A circulation model for batch mixing in agitated
vessels,
-
Mersmann, A., Einenkel, W. D. Design and scale-up of mixing equipment,
& Kppel, M. Int. Chem.Eng., vol. 16, no. 4, pp. 590 603,1976.
-
Metzner, A. B. & Otto, R. E. Agitation of non-Newtonian fluids,
A. I. Ch. E. J., vol. 3, no. 1, pp. 3- 10,1957.
Metzner. A. B., Feehs, R. H., Agitation of viscous Newtonian and non-Newtonian fluids,
Ramos, H. L., Otto, R. E. A. I. Ch. E. J., vol. 7, no. 1, pp. 3-9,1961.
& Tuthill, J. D.
Metzner, A. B. & External flows of visco-elastic materials: fluid property restrictions on

Astarita, G. the use of velocity sensitive probes,
A. I. Ch. E. J., vol. 13, no. 3, pp. 550 - 555,1967.
Middleton, J. C. Measurement of circulation within large mixing vessels,

Proc. 3`d Euro. Conf. on Mixing, paper A2, York, England, BHRA,
4-6 April 1979.
Moo-Young, M., Tichar, K. & The blending efficiencies of some impellers in batch mixing,
Dullien, F. A. L. A. I. Ch.E.J., vol. 18, no. 1, pp. 178 - 182,1972.
Muskett, M. J. Gas-liquid dispersion in stirred tanks: scale-up and agitator

comparison,
PhD Thesis, Cranfield Institute of Technology, 1987.
Nagata, S., Nishikawa, M., Power consumption of mixing impellers in pseudo-plastic liquids,
Tada, H. & Gotoh, S. J.Chem. Eng. Japan, vol. 4, no. 1, pp. 72 - 76,1971.
Nienow, A. W, & Elson, T. F. Aspects of mixing in rheologically complex fluids,

Ch.E.R.D., vol. 66, pp. 5- 15,1988.
Norwood, K. W. & Flow patterns and mixing rates in agitated vessels,

Metzner, A. B. A. I. Ch.E.J., vol. 6, no. 3, pp. 432 - 437,1960.
Nouri, J. M. & Whitelaw, J. H. Flow characteristics of stirred reactors with Newtonian and non-
Newtonian fluids,
A. I.Ch.E.J., vol. 36, no. 4, pp. 627 - 629,1990.
Novk, V. & Rieger, F. Homogenisation with helical screw agitators,

Trans.I.Chem.E., vol. 47, pp. T335 - T340,1969.
Novak, V. & Rieger, F. Homogenisation efficiency of helical ribbon and anchor agitators,
Chem.Eng.J., vol. 9, pp. 63 - 70,1975.
116
Novak, V. & Rieger, F. Influence of vessel to screw diameter ratio on the efficiency of screw
agitators,
-
Oldshue, J. Y. Fluid Mixing Technology,

Ch. 3, pp. 62 McGraw Hill Publishing Company, New York,
-62,
1983.
Palmer, M. H. Design guide for impeller power numbers,

FMP Report 028, April 1987.
Parkin, S. P. (ed. ) M`Graw-Hill Dictionary of Scientific and Technical Terms, 3rd

edition,
McGraw-Hill Book Company, New York, 1984.
Peters, M. S. & Process design and economics for chemical engineers, 2ndedition,
Timmerhaus, K. D. Ch. 5, p. 179, McGraw-Hill Book Company, New York, 1968.
Pollard, J. & Kantyka, T. A. Heat transfer to agitated non-Newtonian fluids,

Trans.I. Chem.E., vol. 47, pp. 721 - 727,1969.
Pollard, G. J. Flooding and aeration efficiency in standard stirred vessels,

Int. Symposium on Mixing, paper C4, Mons, Belgium, Faculte
Polytechnique de Mons, 21 - 24 February 1978.
Prtner, R., Langer, G. & Homogenisation of non-Newtonian fluids in stirred tanks,

Werner, U. Proc. 6' Euro. Conf. on Mixing, pp. 521 - 526, Pavia, Italy, BHRA,
24 - 26 May 1988.
Prochzka, J. & Landau, J. Studies on mixing XII: homogenisation of liquids in the turbulent
region,
Coll. Czech. Chem. Commun., vol. 26, pp. 2961 - 2973,1961.
Raghav Rao, K. S. M. S. & Liquid phase mixing in mechanically agitated vessels,

Joshi, J. B. Chem. Eng. Commun., vol. 74, pp. 1- 25,1988.
Reavell, B. N. Practical aspects of liquid mixing and agitation,

Trans. I. Chcm. E., vol. 29, pp. 301 - 309,1951.
Rieger, F. & Novak, V. Power consumption of agitators in highly viscous non-Newtonian

liquids,
Power consumption for agitating visco-elastic liquids in the viscous

Rieger, F. & Novak, V.
regime,
Rielly, C. D. & Britter, R. E. Mixing times for passive tracers in stirred tanks,
Proc. 5' Euro. Conf. on Mixing, paper 36, Wurzburg, W. Germany,
BHRA, 10 - 12 June 1985.
Rigden, M. P. Private Communication, 1991.
Ruszkowski, S. The FMP method for measuring mixing times,

FMP Report 008, February 1985 (a).
117
Ruszkowski, S. A comparison of mixing time and mixing efficiencies for a

range of
impellers in a 0.61 m diameter vessel,
FMP Report 009, October 1985 (b).
Sano, Y. & Usui, H. Interrelations among mixing time, power number and discharge
flowrate number in baffled mixing vessels,
J.Chem.Eng.Japan, vol. 18, no. 1, pp. 47 52,1985.
-
Sano, Y. & Usui, H. Effects of paddle dimensions and baffle conditions on the
interrelations among discharge flowrate, mixing power and mixing
time in mixing vessels,
J.Chem.Eng.Japan, vol. 20, no. 4, pp. 399 404,1987.
-
Schwartzberg, H. G. & Fluid and particle motion in turbulent stirred tanks: fluid motion,
Treybal, R. E. Ind. Eng. Chem. Fund., vol. 7, no. 1, pp. 1-6,1968.
Shamlou, P. A. & Power consumption of helical ribbon mixers in viscous Newtonian

Edwards, M. F. and non-Newtonian fluids,
Chem. Eng. Sci., vol. 40, no. 9, pp. 1773 1781,1985.
-
Shiue, S. J. & Wong, C. W. Studies in homogenisation efficiency of various agitators in liquid

blending,
Can.J.Chem.Eng., vol. 62, pp. 602 - 609,1984.
Short, D. G. R. & Unsolved problems of industrial mixing,

Etchells, A. W. Proc. 4' Euro. Conf. on Mixing, paper Al, Noordwijkerhout,
Silvester, R. S. Mixing time and power requirements for non-Newtonian fluids,

Fluid Mixing II, I. Chem. E. Symposium Series no. 89, Bradford
England, 3 -5 April 1984 (a).
Silvester, R. S. Mixing of non-Newtonian media: a technical review,

BHRA Report G0128,1984 (b).
Solomon, J. Mixing, aeration and rheology of highly viscous fluids,

PhD Thesis, University College London, 1980.
Solomon, J., Nienow, A. W. & Flow patterns in agitated plastic and pseudo-plastic, visco-elastic
Pace, G. W. fluids,
Fluid Mixing, I. Chem.E. Symposium Series no. 64, pp. Al - A13,
Bradford, England, 24 - 26 March 1981.
Su, Y. S. & Holland, F. A. Agitation and mixing of non-Newtonian fluids - part 1,

Chem.Proc.Eng., pp. 77 - 84, August 1968 (a).
fluids 2,
Su, Y. S. & Holland, F. A. Agitation and mixing of non-Newtonian - part
Chem.Proc. Eng., pp. 110 120, September 1968 (b).
-
Tebel, K. H. & Zehner, P. Mixing of power law fluids in loop reactors,

Proc. 5`' Euro. Conf. on Mixing, paper 32, Wurzburg, W. Germany,
BHRA, 10 - 12 June 1985.
Ulbrecht, J. Mixing of visco-elastic fluids by mechanical agitation,

Chem.Engr., pp. 347 - 367,1974.
118
Ulbrecht, J. Influence of rheological properties upon the

process of mixing,
Verfahrenstechnik, vol, 9, no. 9. pp. 457 463,1975.
-
Van de Vusse, J. G. Mixing by agitation of miscible liquids part I,

Chem.Eng.Sci., vol. 4, pp. 178 200,1955 (a).
-
Van de Vusse, J. G. Mixing by agitation of miscible liquids part II,

Chem.Eng.Sci., vol. 4, pp. 209 220,1955 (b).
-
Vonken, R. M., Circulation flow and mixing at low Reynolds numbers,
Goetzee, W. J. J. & Beek, W. J. Int. Symposium on Mixing, paper A5, Mons, Belgium, Faculte
Polytechnique de Mons, 21 24 February 1978.
-
Wang, K. & Yu, S. Heat transfer and power consumption of non-Newtonian fluids in
agitated vessels,
Chem.Eng.Sci., vol. 44. no. 1, pp. 33 40,1989.
-
Weaver, J. B. & The chemical engineers' handbook, 5' edition,

Bauman, H. C Perry & Chilton (eds.), sect. 25, p. 25-30, M`Graw-Hill Kogakusha,
New York, 1974.
Wichterle, K. & Wein, 0. Threshold of mixing non-Newtonian liquids,

Int. Chem. Engng., vol. 21, no. 1, pp. 116 - 120,1981.
Wichterle, K., Kadlec, M., Shear rates on turbine impeller blades,

2k, L. & Mitchka, P. Chem. Eng. Commun., vol. 26, pp. 25 - 32,1984.
Wichterle, K., Zak, L. & Shear stresses on the walls of agitated vessels,
Mitschka, P. Chem. Eng. Commun., vol. 32, pp. 289 - 305,1985.
Wong, C. W. & Huang, C. T. Flow characteristics and mechanical efficiency in baffled stirred tanks
with turbine impellers,
Proc. 6' Euro. Conf. on Mixing, pp. 29 - 34, Pavia, Italy, BHRA, 24
26 May 1988.
-
Zlokarnik, M. Eignung von ruhren zum homogenisieren von flussigkeitsgemischen,

Chem. Ing. Tech., vol. 39, no. 9/10, pp. 539 - 548,1967.
2
+ - = + + +
, 3 3 i,
3
Q) z z z
C O - N 00 O - N O -
3E ^
- -
c
,
m
0
b4 c^
Q0 O
-
v ON
'
b
cn
'C7
VU
C
ri
CD CD V) CD r c: > 00
tai E_ a+ r, )
C
rn
- p p kn O> cV
UE N c N - M -- N - N
-
. CD O O O O O O O O O O 3

C'
Z
z
E
3
z
v G O GT O kI
a) Ê O O M O O O M O O O O
M 00 M 00 M M
O O O
O O - O O O --
"C C M
'9
u CD 8
r-
E r
w
E
N bA
y
U
c Q> M N `
-
i: z - "iz
O 12
Z
Table 3.2
Probe and Injector Positions
Axial position Radial position
Injector T/25 above fluid surface T/10 from shaft axis
Probe 1 T/50 below impeller T/8 from shaft axis
Probe 2 T/4.5 below fluid surface T/4.7 from shaft axis
Probe 3 T/3 below fluid surface T/2.2 from shaft axis
Table 3.3
Power Law Constants and Indices of Pseudo-plastic Fluids
Natrosol Grade Concentration Constant K Index n

(wt. /wt. water) (Ns" m-2) (-)
250 MR 0.25 0.014 0.88
" 0.50 0.068 0.86
of 0.75 0.258 0.79
250 HHR 0.25 0.092 0.80
it 0.50 1.180 0.55

Table 4.1
Newtonian Blend Time Results
T13 Pitched Blade Turbine
T N 0l (S) 2 93 (S) BRMS (S)

(S)
(m) (cP) (s'1) (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 500 12.7 9.6 9.5 77.8 59.4

(25) (17) (47) (48)
it 14.0 8.5 10.8 51.0 43.7

(12) (22) (48) (48)
to 15.5 7.3 8.7 41.6 34.3

(19) (15) (48) (46)
of 100 5.87 10.0 14.8 77.5 68.6
(11) (17) (10) (11)
it 6.98 6.7 10.1 69.0 61.1

(10) (17) (11) (14)
" 8.23 6.7 9.3 45.5 39.3

(19) (17) (11) (14)
0.609 3.67 14.8 14.4 29.7 28.0

(6) (13) (28) (27)
to 4.38 12.4 13.6 25.8 24.0

(7) (35) (38) (43)
to 5.05 9.7 10.7 26.0 18.0

(12) (24) (64) (27)
it 40 1.51 37.9 38.6 69.7 64.9

(20) (27) (24) (25)
of 3.00 12.8 15.5 23.3 23.5

(16) (22) (19) (10)
" 4.82 6.7 9.4 11.6 11.0

(14) (17) (15) (17)
it 10 1.82 20.5 22.2 23.0 23.6

(10) (5) (9) (8)
" 3.65 9.8 10.4 10.8 10.8

(16) (18) (18) (17)
4.78 7.4 8.1 8.0 7.9

it it
(16) (17) (13) (14)
33.2 60.8 90.6 86.9

1.830 300 1.38
(23) (21) (2) ( 8)
27.8 29.4 29.4 30.0

tv 40 1.37
(12) (12) (17) (11)
20.2 20.0 21.4 20.6

1.98
(13) (4) (15)
11)
(
Table 4.2
T/2 Pitched blade turbine
T 1 N 01 (S) 02 (S) 03 (S) ORMS (S)
(m) (cP) (s"') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 500 5.03 11.1 9.4 165 133

(4) (12) (7) (18)
of 6.33 8.1 7.8 116 87.0

(7) (23) (22) (16)
it 7.25 8.9 8.0 78.8 67.7

(17) (16) ( 8) (15)
100 2.98 10.2 11.6 44.4 41.5

(16) (23) (8) (23)
it 3.53 7.5 8.8 22.9 22.3

(14) (18) (64) (60)
it 4.17 6.8 7.7 25.1 21.6

(10) (11) (10) (19)
0.609 2.12 14.3 13.9 16.0 15.2

( 8) (13) (20) (18)
it 2.52 12.4 11.4 13.0 12.5

(10) (17) (17) (10)
2.92 10.1 10.3 12.3

(11) (27) (28)
It 40 1.50 16.7 18.4 23.4 21.7

(11) (17) (24) (19)
it It 2.67 6.5 7.1 10.9 9.5

(29) (20) (42) (29)
it 2.92 6.5 8.2 12.9 11.5

(40) (20) (20) (13)
3.17 6.6 6.8 11.4 9.1

of
(30) (22) (36) (17)
0.83 21.3 22.2 28.7 26.9

to 10
(14) (12) (26) (10)
10.7 12.1 14.4 14.0

It 1.50
(10) (16) (12) (13)
2.17 8.6 8.6 9.9 10.3

of It
(18) (21) (19) (15)
6.6 6.6 7.8 8.0

it to 3.00
(16) (20) (16) (15)
Table 4.3
T13 Flat blade turbine
N 01 (S) 02 (S) 03 (S) RMS (S)

T
(m) (cP) (s'') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.609 100 1.70 28.4 32.2 202 168

(16) (22) (33) (33)
2.55 21.6 20.5 88.7 71.1

(51) (18) (23) (29)
3.33 11.0 11.8 39.7 30.5

(18) (14) (40) (34)
40 1.82 20.9 21.7 22.9 21.8

(10) (7) (8) (5)
3.83 8.8 10.3 8.7 10.4

(17) (24) (14) (16)
10 1.67 16.3 21.0 25.5 23.6

( 8) (16) (28) (25)
3.33 8.1 8.8 8.3 8.7

(9) (19) (11) (11)
Table 4.4
T/2 Lightnin A310 hydrofoil
0 N 01 (S) 02 (S) 03 (S) RMS (S)

T
(m) (cP) (s"') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.609 200 1.63 98.8 127 240 218

(58) (47) (20) (24)
it 3.30 13.7 18.9 99.3 87.4

(47) (60) (15) (14)
175 2.50 31.8 36.6 66.7 58.8

(30) (28) (30) (22)
100 1.67 29.3 29.5 97.4 91.1

(37) (36) (14) (15)
3.37 23.2 16.9 47.9 43.4

(40) (35) (34) (43)
75 3.17 11.9 14.7 26.8 23.3

(21) (11) (12) (17)
40 1.68 21.8 19.6 61.9 60.8

(36) (26) (69) (58)
.. 3.30 7.3 12.0 12.6 13.0

(41) (11) (21) (15)
" 4.62 5.8 7.6 9.8 9.4

(29) (24) (48) (29)
Table 4.5
NO versus Re Regression Results
Impeller z r2 (%) F-ratio Data Points
T/3 4PBT -0.989 93.09 163 13

T/2 4PBT -1.068 86.51 65 11
T/3 4FBT -1.259 84.71 23 5
T/2 A310 -0.836 78.95 27 8
Table 4.6
Correlation Regression Results
Impeller y r2 (%) T F-ratio

F Data Points
T/3 4PBT -0.497 98.46 769 13
T/2 4PBT -0.434 95.78 228 11
T/3 4FBT 94.74 73 5

-0.425
T/2 A310 94.22 115 8
-0.488
All 96.49 990 37
-0.464
Table 4.7
Comparison of Constants
Impeller kx RSD (%) k0.5 RSD (%)
T/3 4PBT 179.5 10.98 177.7 10.99
T/2 4PBT 238.8 21.80 186.1 25.22
263.2 12.65 182.0 15.85

T/3 4FBT
199.9 14.98 190.2 15.18

T/2 A310
222.5 16.26 183.5 17.40

All
Table 5.1
Pseudo-plastic Blend Time Results
T CN (%) N e, (S) 02 (S) 83 (s) ORMS (S)
(m) Grade (s"') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 0.50 8.33 81.4 89.2 768 740

250HHR (20) (13) (10) (10)
if 10.0 17.5 18.9 670 601
(12) (33) ( 7) (7)
of 0.25 3.33 18.0 21.1 265 240
250HHR (12) (14) (22) (20)
is 5.00 10.2 18.2 91.6 83.8

(15) (29) (33) (36)
0.609 0.50 4.17 38.8 37.9 193 170

250HHR (14) (31) (3) (4)
if 5.50 28.5 28.7 66.0 58.5

(16) (25) (11) (8)
it 0.25 3.00 17.9 18.3 39.1 29.7

250HHR (15) (18) (22) (13)
" 0.75 3.33 34.7 30.3 135 115

250MR (14) (14) (10) (24)
to 5.33 20.9 20.3 49.5 43.1

(26) (24) (18) (24)
it 0.50 2.08 23.5 29.0 103 90.6

250MR (10) (21) (29) (20)
it 2.92 18.7 20.3 44.0 40.2

(13) (15) (17) (20)
" 3.75 13.4 13.7 30.3 29.6

(17) (17) (12) (19)
0.25 2.00 17.4 22.8 20.6 21.9

250MR (16) (11) (21) (10)
it 2.67 12.3 14.6 17.3 17.9

(8) (16) (26) (19)
1.830 0.50 1.67 42.0 74.4 73.2

-
250HHR (26) (13) (12)
if 2.50 25.4 28.4 28.1

-
(15) (6) (12)
Table 5.2
T12 Pitched blade turbine
T Cox (%) N 0 (s) 02 (S) 03 (S) ORMS (S)
(m) Grade (s"1) (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 0.50 4.17 26.4 40.5 216 190

250HHR (13) (22) (19) (18)
of 5.83 5.9 9.1 32.6 27.2

(43) (25) ( 8) (8)
it 0.25 2.00 12.4 13.0 95.2 82.0

250HHR (21) (16) (14) (15)
it 3.67 4.3 7.7 24.2 19.9

(20) (13) (25) (23)
0.609 0.50 2.50 18.2 22.0 104 90.8

250HHR (8) (7) (17) (19)
of 4.00 5.9 7.9 13.4 10.9

(4) (30) (22) (29)
it 0.25 0.92 15.4 18.2 88.5 79.1

250HHR (14) (13) (16) (19)
if 2.00 4.9 8.0 16.8 14.9

(4) (16) (17) (19)
it 0.75 2.00 28.3 34.3 56.0 63.3

250MR (20) (15) (22) (10)
it 2.50 19.3 24.4 25.5 25.0

(23) (30) (19) (26)
it 3.00 18.5 18.6 21.2 21.6

(18) (19) (21) (17)
It 0.50 1.25 21.7 24.0 29.2 28.4

250MR (16) (17) (17) (17)
it to 2.08 14.8 16.6 21.2 17.5

(12) (13) (13) (10)
it 2.92 10.7 12.9 11.9 13.4

(20) (20) (18) (19)
0.25 1.00 22.1 22.6 25.4 24.4

to
(10) (15) (12) (11)
250MR
1.33 14.8 17.3 19.1 18.5

to to
(9) (9) (19) (17)
1.67 12.8 15.2 14.4 15.5

to
(13) (18) (14) (12)
Table 5.2 continued
T/2 Pitched blade turbine
T CN (%O) N 9, (S) 02 (S) 03 (S) BRMS

(s)
(m) Grade (s"') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
1.830 0.50 1.00 30.6 45.2 44.3

-
250HHR (12) (9) (10)
1.25 - 25.1 27.3 26.2

(34) (11) (9)
Table 5.3
T13 Flat blade turbine
CN (%) N 81 (S) 82 (S)

T 03 (S) ORMS (S)
(m) Grade (s"') (RSD (Io)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 0.50 8.67 6.3 7.3 111 89.6

250HHR (24) (15) (25) (27)
It 10.33 5.3 8.3 163 127

(13) (39) (19) (18)
it 0.25 3.33 9.9 11.7 140 121

250HHR (27) (14) (12) (13)
if 5.00 7.1 12.5 91.1 73.7

(11) (22) (16) (21)
0.609 0.50 4.00 9.9 13.5 39.3 28.0

250HHR ( 5) (25) (18) (16)
If 0.25 2.00 10.9 16.5 82.4 77.4

250HHR (19) (19) (20) (18)
3.50 3.6 6.4 15.3 13.1

(43) (27) (23) (21)
0.75 3.00 21.4 21.1 43.2 37.0

250MR (24) (15) (18) (17)
4.00 11.5 13.2 15.6 15.6

(16) (20) (29) (23)
5.00 11.1 12.9 14.9 14.3

(34) (10) (19) (10)
0.50 2.08 16.2 16.5 23.0 18.9

250MR (14) (12) (14) (19)
2.92 12.2 13.4 16.7 16.7

(12) (18) (18) (17)
3.75 9.5 9.8 11.8 11.8

(12) (10) (11) (12)
0.25 1.33 19.3 22.6 27.4 26.8

250MR (14) (17) (19) (16)
2.00 14.6 19.5 22.2 22.0

(15) (18) (23) (22)
10.9 11.4 11.5 11.8

2.67
( 8) (13) (19) (11)
Table 5.4
T/2 Hydrofoils
T CN (%) N el (S) 2 03 (S) ORMS

(s) (s)
(m) Grade (s'') (RSD (%)) (RSD (%)) (RSD (%)) (RSD (%))
0.305 0.50 7.50 53.1 51.5 292 283

HE3 250HHR (8) (10) (15) (14)
it 9.17 16.5 14.3 205 160
(19) (40) (21) (16)
it 0.27 3.67 18.1 28.8 117 111

250HHR (18) (24) (7) (6)
it 5.33 11.1 14.7 93.5 84.1

(28) (28) (22) (29)
0.609 0.50 3.33 23.8 25.8 71.7 61.0

A310 250HHR (17) (17) (17) (17)
It 5.00 39.9 28.8 47.9 44.3

(14) (15) (14) (19)
it 0.25 2.00 22.6 21.5 82.0 61.0

250HHR (28) (10) ( 5) (18)
to 3.00 14.5 13.5 17.7 16.8

(13) (13) (13) (16)
it 0.75 3.00 25.9 28.8 38.9 36.5

250MR (16) (22) (8) (12)
it 4.00 24.0 26.5 31.6 29.3

(26) (24) (7) (15)
" 5.00 15.7 19.4 23.2 21.2

(16) (15) (11) (11)
" 0.50 2.08 16.6 21.8 25.0 24.7

250MR (20) (8) (16) (16)
" 2.92 20.3 23.3 27.3 23.4

(12) (12) (12) (6)
3.75 11.2 14.6 15.0 14.7

it
(18) (9) (21) (9)
1.33 24.3 29.8 39.4 38.4

if 0.25
(24) (24) (27) (27)
250MR
16.8 20.3 19.9 18.4

1.67
(23) (15) (15) (12)
2.00 15.8 16.7 16.8 17.0

"" "
(19) (24) (16) (17)
Table 5.5
NO versus Re Regression Results
Impeller x r2 (%) F-ratio Data Points
T/3 4PBT -0.636 73.41 29 11

T/2 4PBT -0.842 84.94 74 14
T/3 4FBT -0.894 67.23 26 13
T/2 A310 -0.960 64.21 24 14
Table 5.6
Correlation Regression Results
Impeller y r2 (%) F-ratio Data Points
T/3 4PBT -0.583 94.76 182 11
T/2 4PBT 96.29 338 14

-0.515
T/3 4FBT 90.80 119 13
-0.483
T/2 A310 84.89 74 14
-0.431
All 89.71 463 52
-0.453
Table 5.7
Comparison of Constants
Impeller kX RSD (%) ko,5 RSD (%)
141.4 21.09 193.7 24.68

T/3 4PBT
171.9 18.89 183.5 19.43

T/2 4PBT
30.40 175.8 31.17
T/3 4FBT 188.6
30.16 206.6 38.68

T/2 A310 265.7
32.19 200.0 38.36

All 237.2
Table 6.1
Power Numbers
T (CP)/ Po RSD
(m) CN (/0) (-) (%)
0.305' 500 1.54 1.47

A 100 1.41 2.04
to 0.50 HHR 2.53

-
it 0.25 HHR 1.57 6.08
0.609 100 1.51 0.14
40 1.62 2.62
10 1.60 0.58
"B 1 1.56 -
"c 0.50 HHR 1.25 2.38
"c 0.25 HHR 1.15 3.80
"c 0.75 MR 1.21 0.94
"c 0.50 MR 1.23 1.06
"c 0.25 MR 1.20 3.14
1.830 300 1.65 -

it 40 1.77 -
B 1 1.80 -
2.6708 1 1.80 -
A: Power number estimated from Palmer (1987).
B: Power number data from Mackinnon (1987).
C: Blade angle = 41.

Table 6.2
Power Numbers
T (cP)/ Po RSD
(m) CN (q0) (-) (%)
0.305A 500 1.72 2.27
A 100 1.45 0.74

it 0.50 HHR 1.22
-
" 0.25 HHR 1.08 -
0.609 100 1.10 0.23
40 1.00 0.45
10 1.10 4.72
"B 1 1.12 -
0.50 HHR 1.03 3.99
0.25 HHR 1.00 0.21
0.75 MR 1.03 0.45
0.50 MR 0.99 1.22
0.25 MR 1.05 0.41
1.8308 1 1.28 -
2.670k 1 1.28 -
A: Power number estimated from Palmer (1987).
B: Power number data from Mackinnon (1987).

Table 6.3
Power Numbers
T/3 Flat Blade Turbine
T (cP)/ Po RSD
(m) CN (%) (-) (%)
0.305 0.50 HHR 2.77

-
to 0.25 HHR 2.97 1.76
0.609 100 2.71 0.48
40 2.70 0.58
10 2.72 0.75
"A 1 2.85 -
0.50 HHR 3.23 -
if 0.25 HHR 3.02 0.87
0.75 MR 3.03 1.90
0.50 MR 2.83 1.11
0.25 MR 2.96 2.13
1.830A 1 2.98
A: Power number data from Mackinnon (1987).

Table 6.4
Power Numbers
T/2 Hydrofoils
T (cP)/ Po RSD
(m) CN (%) (-) (%)
0.305 0.50 H14R 0.55 11.6
it 0.25 HHR 0.33 0.86
0.609 200 0.41 3.13
175 0.33
-
100 0.30 0.71
75 0.29 -
40 0.29 1.39
"A 1 0.29 -
of 0.50 HHR 0.26 2.67
0.25 HHR 0.25 2.83
0.75 MR 0.28 2.15
0.50 MR 0.25 0.46
0.25 MR 0.30 0.51
A: Power number data from Mackinnon (1987).

Table 7.1
Selling Price (Capital Cost) versus Shaft Torque
(from Rigden (1991))
Selling Price Torque

(k) (Nm)
8 250
12 1000
18 2000
30 4800
50 12000
Table 7.2
Power Law Constants and Indices of Fluids in Economic Analysis
Constant K Index n
(Ns" m-) (-)
2.00 1.0
2.67 0.9
3.55 0.8
4.74 0.7
6.31 0.6
8.42 0.5
11.22 0.4
ice
T/60--
-r
3.1 Vessel Arrangement

W
3.2 Pitched blade turbine

I III
WI
3.3 Flat blade turbine

3.4 Hydrofoil
6mm
Screei
Cable
E
E
Epoxy
0
Body CV)
1/2 "point
mm Coiled
electrode screening
electrode
3.5 Conductivity Probe

U)
O
-a
0
C
Qi cU
U
U
E
(i v O
N .X
v X X w
w EO O 0
v
v w v :
"
nN 3 N LA
- -0 3
> 3: E0EE
'E 0) 0
Qo o
Eoo U' N cv
,-m
Ln -- F- t- F-- - `-
`
' T 1V J
/ `
V
2 C
JJ
_O -0 O
"

CL CL
4Y
0
Z.
Z)
m
0
z.
Q.
0
J
ID
M
0
0
Cy)
W
N
U
LC
NJ
ca
u
I ' c.
."
t7
CD
In
Ln
NO
oo Ln 0
In
SHH 0) asuodsaj japes

O
In
r--
Lf)
N
F"
0 n.j
o n.
Lj
' to
C
N
V
(1)
E
F-
N-
CC
U
.r
a
F"
0
N
LU
N
LD In -It M CD
Mio 501) DDUDIJDA50i
N
w
T- -
i
L
i 'Cl
Nm
E if
LO
eK
t-- --
"O
a 000 -~ ci
LLL
C .C 3
>> .p
L Doa
Ex
X 4.d *+O C3K
Z
-9
ly,
> "b
+UK o L
_ o
o E
f
M v1 C
L, r
CC '
O
(ai O C
O
>'TJ
. y
V
CG
O
S
E9Y cl
3
o 0 +*
aZ
I
06+ O
y
-O
0
M 1 0
17
0
f
oCP +1
i 'KI
0 00 o
t'1
00
0 0
0
0
Ti
TI 'OAT
iii
w
.._,_. _ _.___
____. .ti....
__
__.______.
__.. ____
_ ___
__.._.
j__,___,.
. _.
____ _.
_..
_. ,
cu m
L G1 D 0 tD
_ L L L
C " 1 T .J
Yl
0
L ` O
4i .Z
E
0 O
aN
O0 0
I -A-
0
0
o o
o
o
f f i o
O O c,
0 Q
0 Q cf
O
Je
I
iz
__ ___ r
ED t] Ll i !1J
000 ,.
cLLL
t
a a. n.
' tc +a
e*
i
IQ
le* jo
O+
J0
9) * L
i
ri
0*1
r
i fl
cC ' Ò
IN
O is
CL
r- 0+ * i
ICI " +-V s
y
oa EIK
O -K '0
iO
0 O

_ _.
__.
0 rl
pO 0
QO 0
OD
ÒN
1\
it]
i
t
S
L
C2
cu m
.i
jo r) Ui
0 0 0 ---1 i
L L L
=O
3
=O >'
L. C31-*
Df
0
- O
_ L
+ O
G -
ON
4 c
'' O
= di
> -a
-M
oo +
00 0 :r+
i
i 1O
z o8
1
1
{O
++ 0
r---
C1
0 +Z
e i
lT
1 0
f O
t___ --_ - -j - ---- ---

0 0 O a
0 O 0 `
0 O
O Ti
!env
N
Ui
-- T- --
1 1i 1
F- E I i
Ci ++
1
N
Co
rn
C
0
c
c
0 0 O 0 C 0
1 1 I
O "-'
E
o
'O
0
O
3
EU (ja 3
+ 0 1 11 co 3
.Q
3
3 O
UD = 0
0
m
:e <u
r 3 0
0
m
E
3
LV
GC 'l7
ift
X
CL
X
I- b
00 j .,
z f
_-J 0
i0
i
N
14 + -}-,
i
{O
i
O
o 0 0
O 0 0 vi
O 0 Ì
0 ri
Ni
SJION
1.1 J
CY)
0 CD
cn (D
0 0
co
oC
o
C
0 0 o
O 0 p
y 3

Z) m + O C tl x
c
(A y
L i
<U M ...
3O L

CD
3O E
C
LV
"C
i
X O
C
X iO
O
X
0
a1
11 11x
o= V
i
fl

00
0
z O
oQ 0
0
00 1
+
+ppa
10
O
i
ppo
o
ENS01V
_W
A to
ili
ii
3
O
3
CG C
L 3 0
L 3 0
.
0
3 0 .CD
0
3
TI
CC -' r
0
0
R3 '
X 0 w
0
G4 Tl
x
OO
11
O
a.
OO
U
0
z 0
0
Tl
cT
14
0
---lo ti
0 ii
opO
0po vi
DO
0
s14HBNV
iii
------ _ - --_
i
1 i
lD
C
W
O
3
3
LM
o
G Q
jo L
ZC 4J
y o
LM u
i
I
C
cC '
i o
.... o
o
*w l=
_
w' o
O
rte, C 11 V
Oy
z O
iO
-0
0
___-J
-KI
O0OO
OOO `
pO `
O
S!V8ON
ID
W
J.. _.. ----T- - _
1
i
1
43
io o
3 r 10 O
t 3
3 iO
3 1
O
3 O
3 J0
m
.c . O
1
ftX
J3c
O
eO
X
.2
- !1O
uyam, t
L
L .
a
U M o
cu 0
Z
4.
'G "
C 0

E
'n rn o
m
0
..... C +f

a,,,
O F:7 -
co
co
o , C
+
z -}-
$ v
iv c c
E
o
co
wo o
wo
14
U fA
+ O II X
OQ
N (p O O0
w W O OO
O
O
ti
0
99 OJ
ci
to
-- --r---- .__. _._i. _.__ -__

F-
i
A
o
---- 33 0
3 0
3 io
3
0
3 0
o
o

X
x
x
o
0
1
0
L
,}
1i
N O
O
O
---
4 C co
r.
c.
O Ol
O
N
Î
+l
+
ci
Q
N to 0 0
W 1t! 0 0 Q
0
0 0 rf
0
OH
C1,0J
LO
P.
1 - 3
io
o
3
'o
o
o
X3:
3 0
C
o
3 o
_H
L o
x
O

x
cC
L
11 - o0
L
O u j1
UM
E
E- 2 io 0
= ù
cm %K
o
v=
4.
o 0
3
o ,
N-
az CD
CD
0
0
c
c
Ti
M E
0
R
0
n
i-_-__ .-- p
O
N o 0 O O
W W 0 0 O cl
vi vi o 0
0 vi
0
EIrd
to
llJ
c1
r--
3
g
i0

--
+r
i
0
U 0
B 0
cC +
L
.
V 0
JN 0
0
ri
m
a c O
00 i
Oa 3 Co
rn
C
z 0
c
c
I 1
E j
0 ..
q?
0
++
_ -
o 0 0 0
N LU 0
uJ 0 0 0
o 0 0 `l
o
o -
cl
am
Elid
CD
_. ,. - -- - - --- .__.. T_
i+
q% O
r O
_0
.0
i 0
O
{ O
c02
1 O
O
O
c:i S'I
a*Wb
O "- o
0
-0
,K
9I

C
0
- 0
o
}
.
0. Z
''! 0
C
,
In
H F- o
O'<
co C) O
LL
E ti ti a

0 +
L LJ S
_J O
LU
O O
0 O
i'
iJ 1
O O
0 0
1 o d 0
O O Ti
o
C'I
ar c/ld
R r-.
_
.
L)
cl
C/)
ca
O tl
4l
CC
)
t-
o i _
vi O 4J

-4
7j ter
,
b
i C G/]
4- L
1 1 0
L
C.
L
C, )
.
cr..
0
II 0
Cam. A.
cl
vi
ti..
V)
b
C/?
c3
Q3
V)
T /6
fT
2
5.2 Vessel Geometry assumed for estimation of the Wall Shear Rate
N
L. rf cu m
W W
A A A
a)
W
r 0 0
-p = L L L
3
C + 0 CSK
v O
h
- L
I
0. 0
Ic U 0+* i
Z3 cC
C3'
r.. r
O b 40 (
0
a s 0 E
0 0 0
o
0 +*
vMi ( 0*
C-k
0
W
zi c IT)
al 0) Q)
CC .0D .0
v - 000
LL
Oo
>> y
aaQ.
v -0 i i
i *+o
L=
0
t...__ H o L
.
z
yN
C
O
L c
Q

CC -0' !O O
C
0'* 0
"a u i
T_.. C0
ACK O
"C
C
cC id
M.
4..
o -C
o +
Oa
f
Z. 120 8 +**O
r-
4
-
O -w -o -i
IT 00
ui
f O 0
r O
-. -
0 OO '
O
O 0 O
O 0
O ì
vi
Vv
rl
T_____ _. _ .... _f . _....._ __ __.--- ---_ _ _
w
t
to
I
o w c)
D D A W
0 0 0
r
L L L
-' CC
a a a
i
!
:r: + o
i O
ei, r,
O
O L
X O
0
yM 'cl
O
d*
c
La 'G
f C
0
O
=u O :4
O
o+'
c. O
O= OIK
ew O O-W
Oc
a. i
0}* - 0
0 0
0 0
o
In i 0 +c r
In
o+
O + 0
L.
Q o
n
o 0 oO
0 0 O '
o O
o
i
'eng
N
x cu m
Gil
D D D W
L
La a 0 0 0
w_ C. L L
E" D a a
? : + 0
f L.
C4 M GJ
L "C `
; C
'C7
O
C
"C u 1 OF
0+*
04*
o
"
0
o
o
o
ì 'eng
;
W
---
rn 0
- E
ci co m ,+ A
L
n
OD
rn
0
M cc cc
M Cl Cl c
^ -
cv
U C
=
2
S
2
N
=
=
_
z
N O
to
LO
N
=
=
m
iC c0
i
O O 0 O p
C; O Ti
O .G
I

E
se A n x 0
O
> CL i
O
1
L
CD 0
H
cu
3 __- O
O
LM
3
i
1
j
cv c i O
i V
O x
O
o-v OM
ou O
as 11 1
O
XO a
O
J O
N
tn i
L
0
1A
o0 0
L
cl
.. ___- _J
oQ
0
0 o
0
vi
SJV8ONV
N
1n
g
rn O
++
0
N
L
m
.C c
0
c
v =
S
Z
S
i
T
2
S
= m Y 3 Ito
p
C 0
Ui LO 0 N
LO O 1A 0
i t6 i
UZ U' r n N 3
N
GQ 0 0 0 0 ai E
-0 r. 0
i
O
C z- 0
3
i O +
3
LLB a E
u (aj n n x 0 3
10 3
3
L
0
3 0
oL
o .
LN
3 o
V
+
C Oc

C "C3
O
u
O
o 0
o
p; a
o r1
o
o
a 0
0
0
rl
00
tn
T-i
O O o 0
O O 0 ri
o 0 rl
O
ri
SHMON
N
i
rn
_
L
i
li izz
OD
o)
c
0
I [ c
i I
2 C[ c 1 to
c m
2 2
N 0
1 uJ
N
(iz 'r N t A N
d 0
CJ 0
E -H ri
O 0
cu
3 i
0 i
X t' I
E
R (Jj n w x o 3 3 s
I
Gn L
M 3
3 IO
Cif r. + 0
3 0 L
y O
LM
Oi:
3 E
i C
i
C
+ 1
GC GC O
li "0 O
iO O C
----- 0 v
L`
Om
D
as X o
R
X .Q
V --- o
ON
14
i
r
O
' O

-1
Q00O
O0OO
O
swNenl
N
c1121
O
L
h
Co
01
ID
Ii: In
.ri
0
0 L.
0 gi
0 .0
0
C
ri
Q r
.! V

C
L 0 a
0
C r.+
0
G
cj
0
0
o "v 0
ri
as
O
0
0
1
IKi
o000
o00
o0 cl
o cl
cl S WON
W
TI
0
0
O
v
O
O
3 O
O
.2
W -0
L .
0
0
0
"Zi
en
+'
"d
0
li 0
CG
0
-
r- t3
X,
TI O
limw
0
N-i
rH
ro
kti
Ti
(D 0 0 0 0
W O O O
LJJ
vi O O O `'
0 p +
O v'
vi
a9 c1d
W
r -r-- ri' O
33
___.
_ O
O
3 O
3
i i
v jO
C
I O
=L tO
i
CC
4
L
L a-
O
,a 1o
.
_._._
N
+t
1
+ 0
CA xA

O
O
L
O 1
a
E Cl) to co
o
0 O
Co
N m
vc
0c
CC cI cc 2 c[ rt Q aC
v x 2 z z z m
z0 z z z z m m Y
-tv N Ln o N
N 2
at LO N Ln 0
A
-0 > T
O O cj
O O O O
c O
o- ;
A
Z
ta+
O
E o +
A n x
>,
cn
o O o
r. ID 0
O p o
W W
O
O
O
auFI1d
to
33 t
3o
CC
0
3
+, L -Jo
Z 4.4
I
L. y
"a
O
O
cv O
0+O
XX 0
a
0
Il
0E<
Fr= +X
J+A
ln rn
- 27
0 LOJ
M =
c
[ei
fl
2
Q
S
Q
S
fL
2
Q [L
m
ac 0
0
me
2 S S S ;
6 N z 3 R
U=C7 0 b c1
On IN
0 ln
O IN
O
O 0 O
E
.E2
0
M
mM
zo
rf
(0 0 0 0 0
N
w 0 0 0 0
0 t-I
0
0 Ti
TI
a
Eird
W
. _4___ 1--j-
r--- r
O
0
0
._ o
o
33 O
M
7
GT.
O
L O
O
E
0
c o o
--J
-- M
LQ OO
O =_
"C
+x 0
C'am' x Z
-0 N
1- r,
CG E0
I
'101aI}
cam.
I +
'
0
tC
+n
r+
C
aQ
m
tC
N
ter..
CL Q Q CL Q Q Lt ac
0
S = Z 0C
0 Z: m
O i1 S .
ro 3
n N
p p N 4e fd
U Cri to N to
O O -0
D O O
`
cE 0
"
tT
m ro
zo
kr)
1 11-1@)1.1
Cl)
R-c.111+1; i
r
0
N (0 0 0 0 0
w w 0 0 0 Irl
Ti Iri 0 0 0
Izi
0 0
0
Iti
rl
ady
eirod
to
1 c 'o
O
0
z 1O
9.
I
ea oi
O=
+ U
3, z 11 '0
.L CC
i__
1O
L
V 1
i
a
C
od .. _..
%.,
i
0
w= lLJ
O
N=h 11
3
o CL
T!,
C
7 m cl m m y m
z N
LL Z z
j
a Z z
4i
a m
L r+
pV fN
fcC
oa .
00 'tim (S
Cl)
za X
CD co co co D cO `
In m a a L1r
g z 0,11 c
Itow
I
I m

0 U + M A
x iR
E
i. J 0
__ _.... -. ---
O p O
N to 0
p p O
i; U; 0 O O
r, 0 0
0 '; 1
0
aly
lid
5.16 "Max-Blend" Impeller by Sumitomo Heavy Industries
rr
Cl)
3 l1J
,C O
So- 3
C C r.
>4
w .0 6
L 33
0
0
0
3 +* 0
u
'G
_ C
C. C
>,
v
CC
rT
Q
Loi
I o
o
.jagtunu aaMod
R
E
Y
--
T
L
G
a +o .
L
o, O '
..,
O
o
,a
0. Q
raquinu aaMod
.
.'

0
E
Cl)
0 lll
CD
CY)
LO 3O O
a 3.
`
coin O
C
W
t.
cw o
L. en
r`r
O C
s
;.
r,
r
...
r,
w
Q
O
x}
O0
O
.iagtunu .aaMod
fl,
lU
L
U)
M
+
t
L w Y f
x
cn "" LO
-
O
0
}( 0
oK o p
M CD
L
G Q
L.
'II
y N
C
C
+n 0
Q
p
1-
,iaquinu Ja%Od
If)
Ui
v N
L
ID v
E
" C
cam. 1 L
4J
3
TI
U - V L
N1
O >'
i
a 0
Op0 l1 O
ì Cu
x
(spunod) Iso3
flTLT
ON r1T rI r rTJ- ir
C5
z
N
-- H- ff f.;,
14- T U
L H
t
L L
- _J
a
r '
TT il., W
L j
1
l L
in
O
_ y-
1
LI 1
-j
1 C
i
L
0 v
0
a
r
L
4 .
* 3
0
a,
--
N O c
U
N Cl U
ti
Fl-
E
Ct U
P-N
0
d
2
U
D I1 1
O -1111) 0 in o
x
...
(spunod) lso3
O C
S
04 U
U
O
Z!

In E
in c U
n
r--
E i
t V -
L N- EL CL
i
O 1
L 2 H
L
L Qd
H F4- LL
E
O L
L
W
IT_
O F 14-
0
1
Z.. I
r
M N- L'1
N
O
L_ i
o 0
11) O
00Cu
x (spunod) jsoj
In
-T'-T -7-1
m
a
U
1
}i W_
! T
a
in
N
C?
.
L t- ch
m
C2,
i L
to GJ
.Q
iI
O ct
L
O
3
O 0
1
.a a
a
'Tr
N
0
O
0 0
0
0 In 0
cl o00
(spunod) Jso3
In
EJ _ -1--r_-;--1 -- ---T
110
-J w
ri
-H m
-i
NC in
L
4- w -;
1"
1
L
lj W
if-
a H'
L
(Z,1
L
In Gi
1:?
G
ct
C
f L
Gl
0 Irl
a
t ti:
N
0
0
00o n o
ri cu
x (spunod) JsoZ)
__F__r-
C
m
N
UN-0
E
1 --1
II C
1I In
rO ) r
a r -
*.i I
1
L b- LL 1,
p
1 -1
011%
0
0 0 o to 0
Ti N
x
Ai-O (spunod) jsoj
in
d' -
-
-1
m
N
N a:
I
In
N
E
E
ti
C
L
3
0
0 Tl ca
a
N In
N
0
f U _
fir- 1 f
0
OOO i O
X
(spunod) 3soa
APPENDIX 1
ECONOMIC ANALYSIS "tk-Solver" Equations
VARIABLE SHEET
St Input Name- Output- Unit Comment
Cost-Benefit Analysis Calculation
tk-Solver Program
Version 1.7 25/09/91
1000 RHO kg m_3 Fluid Density

5 K Pa sn Power Law Constant
NI - Power Law Index
.7
3 PO - Power Number
L N 71857833 s -1 Impeller Speed
.
D 2 m Impeller Diameter
RE 1385.0576 - Reynolds Number
L RATIO D/T Ratio
.4
5 T m Vessel Diameter
V 91.314991 Vessel Volume
100 TM s Blend Time
TAUw 38.911568 Nm -2 Wall Shear Stress
Rw 18.750483 s -1 Wall Shear Rate
MUw 2.0752301 Ns m- 2 Wall Viscosity
OFO 120.46857 Reciprocal Fourier Number
-
L P 35619.972 W Power
EPS 39007804 W/kg Power per Unit Mass
.
L TORK 7889.3204 Nm Shaft Torque
L CAPTCO 37649.23 Pound St. Capital Cost
6 UNITCOS Pence/ kWh Cost per Unit of Power

8000 HRSPYR Hours Hours per Year Operation
POWCOS 17097.587 Pound St. Annual Power Cost
MAINCO 1505.9692 Pound St. Annual Maintenance Cost
TOTFLO 1493676.1 Pound St. Total Cash Flow
11 IR % Nominal Interest Rate

11 LIFE Years Operating Life
DISFAC 58000225 - Discount Factor
. Cost
L DIRUCO 118691.15 Pound St. Discounted Running
L TOCOST 156340.38 Pound St. Total Life Cost
RULE SHEET
S Rule
* P=PO*(N 3)*(D-5)*RHO
* TORK=(PO*(N 2)*(D 5)*RHO)/(2*PI())
* RE=RHO*N*(D 2)/MUw
* OFO=(RHO*(T 2))/(TM*MUw)
* TAUw=TORK/((T 3)*1.622)
* Rw=(TAUw/K)^(1/NI)
* IF Rw < 400 THEN MUw=K*(Rw (NI-1)) ELSE MUw=K*(400 (NI-1))
* RATIO=D/T
* (PO (1/3))*RE=182*(OFO 0.5)
* V=0.165*((T/0.609) 3)
* EPS=P/(RHO*V)
* CAPTCO=488*(TORK 0.4843)
* POWCOS=HRSPYR*P*UNITCOS/100000
* MAINCO=0.04*CAPTCO
* DISFAC=(1/((IR/100)*LIFE))*(1-(EXP(-(IR/100)*LIFE)))
* DIRUCO=(MAINCO+POWCOS)*LIFE*DISFAC
* TOTFLO=(POWCOS+DIRUCO)*LIFE
* TOCOST=CAPTCO+DIRUCO

R Grenville Thesis 1992 PDF

Uploaded by

Copyright:

Available Formats

R Grenville Thesis 1992 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

R Grenville Thesis 1992 PDF

Uploaded by

Copyright:

Available Formats

What is the thesis about?

What is the thesis about?

What techniques were used to measure blend times?

What techniques were used to measure blend times?

-

CRANFIELD INSTITUTE OF TECHNOLOGY

SCHOOL OF MECHANICAL ENGINEERING

DEPARTMENT OF FLUID ENGINEERING AND INSTRUMENTATION

Academic Year 1991 - 1992

BLENDING OF VISCOUS NEWTONIAN AND

Supervisor: Prof. M. Sanderson

Re > Rec,; NO = constant

Re < Recrij NO a Re-'

CHAPTER 1 GENERAL INTRODUCTION 1

CHAPTER 2 LITERATURE REVIEW 4

4.2.1 Instrumentation Chain 48

CHAPTER 5 BLENDING OF PSEUDO-PLASTIC FLUIDS 68

CHAPTER 6 POWER MEASUREMENT 88

CHAPTER 7 ECONOMIC ANALYSIS 95

CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS FOR

Figure No. Title

3.1 Vessel Arrangement

4.1 Typical Conductivity Response

6.1 Plot of power number versus Reynolds number

7.1 Plot of Costs versus power number -K=2.00, n=1.0

Table No. Title

3.1 Impeller Dimensions

4.1 Newtonian Blend Time Results - T/3 Pitched blade turbine

5.1 Pseudo-plastic Blend Time Results - T/3 Pitched blade turbine

6.1 Power Numbers - T/3 Pitched blade turbine

7.1 Selling Price (Capital Cost) versus Shaft Torque

Symbol Meaning Units

Greek Symbol Meaning Units

cx Kinematic fluid property m2 s"'

Other symbols used in this thesis are defined in the text

cases knowledge of the blend time is much more important.

1.2 NEWTONIAN AND NON-NEWTONIAN FLUIDS

An agitator rotating in a moderately viscous fluid is usually operating in the transitional

Non-Newtonian fluids are characterised by a non-linear shear stress shear rate

This research project is concerned with the blending of shear-thinning, or pseudo-

1.3 LAY OUT OF THE THESIS

the experimental objectives of this research.

Chapter 3 describes the equipment, instrumentation and test fluids in

Chapter 7 describes a cost-benefit analysis of the research presented in this thesis.

A brief catalogue of non-Newtonian mixing reviews is followed by a review of agitator

2.2 BLENDING IN THE TURBULENT REGIME

i) Concentration differences are measured using conductivity probes. The

the way in which the blending process takes place.

where m is the number of circulations.

ie. NO is a constant for a given impeller.

numbers at the jet nozzle.

0.3s T1.4 Ho.s

Bryant & Sadeghzadeh (1982) measured circulation times in a vessel of 1 m3 volume

using a neutrally buoyant flow-follower with an internal radio transmitter. An aerial

and the standard deviation of t by:

and the standard deviation of the mean by:

i) taken at one, small scale (Rielly & Britter (1985)).

iv) drew conclusions regarding the performance of batch mixing agitators

proportional to Reynolds number raised to the -1/6 power.