Mixing. Theory and Practice-Academic Press (1986) PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 313

MIXING

THEORY A N D PRACTICE
EDITED BY

VINCENT W. U H L
DEPARTMENT OF CHEMICAL ENGINEERING
UNIVERSITY OF VIRGINIA
CHARLOTTESVILLE, VIRGINIA

J O S E P H B. G R A Y
ENGINEERING DEPARTMENT
Ε. I. DUPONT DE NEMOURS AND COMPANY, INC.
WILMINGTON, DELAWARE

VOLUME III

1986

A C A D E M I C PRESS, I N C .
Harcourt Brace Jovanovich, Publishers
Orlando San D i e g o N e w York Austin
Boston London Sydney Tokyo Toronto
COPYRIGHT © 1 9 8 6 BY A C A D E M I C P R E S S , I N C .
ALL RIGHTS RESERVED.
NO PART O F THIS PUBLICATION MAY B E R E P R O D U C E D O R
T R A N S M I T T E D IN A N Y FORM O R BY A N Y M E A N S , E L E C T R O N I C
OR M E C H A N I C A L , INCLUDING PHOTOCOPY, RECORDING, OR
ANY INFORMATION STORAGE A N D RETRIEVAL SYSTEM, WITHOUT
P E R M I S S I O N IN WRITING F R O M T H E P U B L I S H E R .

ACADEMIC PRESS, INC.


Orlando,Florid a 3 2 8 8 7

United Kingdom Edition published b\


ACADEMIC PRESS INC. (LONDON) LTD.
2 4 - 2 8Ova lRoad ,Londo n N W 1 7 D X

Library o f Congress Cataloging in Publication Data


(Revised for vol. 3)

Uhl, Vincent W., ed.


Mixing : theory and practice.

Includes bibliographies.
1. Mixing. I. Gray, Joseph B. (Joseph Burham),
Date joint ed. II. Title.
TP156.M5U43 660.284292 65-26039
ISBN 0 - 1 2 - 7 0 6 6 0 3 - 9 (alk. paper)

PRINTE
D IN THE UNITE
D STATE
S OF AMERIC
A

8 6 87 8 8 8 9 9 8 7 6 5 4 3 2 1
Preface
Since the appearance of Mixing: Theory and Practice, Volume II, in 1967,
there has been m a r k e d i m p r o v e m e n t in theoretical concepts a n d knowledge
of mixing subject areas related to industrial-scale operations. Therefore, it is
timely that several of these areas be treated with the c o m b i n a t i o n of in-depth
scrutiny a n d reduction to practice which were the hallmarks of the earlier
two volumes.
Five topics have been selected for V o l u m e III. C h a p t e r 12 o n agitation of
particulate solid - liquid mixtures represents a needed extension of Chapter 9
in V o l u m e II. In the interim, the viewpoint has changed from o n e based
almost entirely o n practical experience to o n e based on empirical correla-
tions.
C h a p t e r 13 on turbulent radial mixing in pipes is a new topic. T h e effects
of jets a n d baffles o n such mixing are treated in detail. T h e m a n y technical
publications o n this topic have n o t been s u m m a r i z e d as extensively hereto-
fore.
C h a p t e r 14, almost entirely from the work of its author, Ivan Fort, d e m o n -
strates t h a t theoretical analysis a n d experimental confirmation can be car-
ried o u t successfully for predicting h y d r o d y n a m i c characteristics a n d s o m e
process results in mechanically agitated vessels. Axial-flow impellers a n d
low-viscosity fluids are involved in this work. In view of the complex geome-
try of the flow patterns in such e q u i p m e n t , this represents a tour de force.
A comprehensive d e v e l o p m e n t of approaches a n d r e c o m m e n d e d prac-
tices for scale-up of agitated liquid e q u i p m e n t is presented in Chapter 15.
T h e m e t h o d s delineated therein provide a useful guide for reducing the risk
of scale-up a n d scale-down catastrophies.
In C h a p t e r 16 o n the mixing of particulate solids, the topics discussed in
C h a p t e r 10 of V o l u m e II are expanded. This chapter is m a r k e d not only for
its lucid exposition of the fundamental concepts a n d measures of the quality
of mixing b u t also for its explanation of the m e c h a n i s m s of mixing a n d
segregation. It also presents well a n emerging i m p o r t a n t d e v e l o p m e n t — t h e
c o n t i n u o u s mixing of solids.
Accelerated progress is expected in the next ten years, particularly because
of the i m p a c t of improved measuring techniques a n d the use of c o m p u t e r
calculation m e t h o d s for various tasks such as experimental data processing
a n d for process a n d e q u i p m e n t design. This includes the c o m p u t a t i o n s
necessary to exploit complex models, for which the m e t h o d is discussed in
C h a p t e r 14. But there is another, rather extensive source of progress: it is the
trend toward carrying o u t tests in industrial-scale e q u i p m e n t . T h e benefits
here are presaged somewhat by material in the scale-up chapter.

ix
CHAPTER 12

Agitation of Particulate
Solid-Liquid Mixtures
Joseph B . Gray* f

Engineering Department
Ε. I. duPont de Nemours and Company, Inc.
Wilmington, Delaware 19898

James Y. Oldshue
Mixing Equipment Company, Inc.
Rochester, New York 14603

I. Introduction

S o l i d - l i q u i d systems are the m o s t c o m m o n of those processed in im­


peller-stirred tanks. Examples of such systems in the chemical process a n d
related industries include (a) c o a l - w a t e r slurries, (b) suspensions of ion-ex­
1

change resins, (c) paper p u l p slurries, (d) polymer dispersions from polymer­
ization reactions, (e) sugar crystal slurries, a n d (f) paint pigment, clay, or
starch slurries. M a n y examples also occur in various ore-processing indus­
tries.
T h e reasons for processing p a r t i c l e - l i q u i d systems in mixing e q u i p m e n t
are m a n y : (a) to p r o m o t e chemical reactions between particulate solids a n d
liquids, (b) to obtain relatively uniform concentrations of particulate solids
in liquids, (c) to p r o m o t e particle dissolution or crystal growth, a n d (d) to
obtain a uniform particle concentration in a n effluent stream when a t a n k is
emptied.

* Retired senior consultant from E.I. duPont de N e m o u r s and Company, Inc.


t Present address: Beechwood Consultants, Inc., Wilmington, Delaware 19810.
1
A slurry is a suspension o f particulate solids in a continuous, liquid phase. The two-phase
mixture is sufficiently fluid to be readily circulated by an impeller in a tank.

1
Mixing: Theory and Practice, Vol. Ill Copyright © 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
2 J o s e p h Β. Gray and J a m e s Y. Oldshue

T w o widely different types of p a r t i c l e - l i q u i d systems are involved in the


discussions of this chapter: low-concentration, near-free-settling particle sus­
pensions a n d high-concentration slurries that possess hindered settling char­
acteristics. W h e n particles in a slurry have settling rates that approach those
of single particles in quiescent liquid, a n d the particles have n o effect o n
nearby particles, the particles are said to be free settling. As particle concen­
tration is increased, particle collisions increase a n d the slurry becomes m o r e
viscous t h a n the liquid phase that suspends the particles. At still higher
concentrations, such slurries m a y develop pseudoplastic ( 0 5 , chapter 15) or
B i n g h a m plastic properties. T h e t e r m hindered settling is used to identify the
settling characteristics of high-concentration p a r t i c l e - l i q u i d systems.
T h e types of e q u i p m e n t c o m m o n l y used for mixing operations involving
s o l i d - l i q u i d systems are described in this chapter. T h e effects of system
property, e q u i p m e n t design, a n d operating variables o n the uniformity of
suspended particle concentrations are discussed. Criteria for judging the
particle suspension performance of mixing e q u i p m e n t are defined. S o m e of
the difficulties involved in sampling p a r t i c l e - l i q u i d systems are described.
Design factors such as m e t h o d s of selecting appropriate impeller speeds
a n d power for p a r t i c l e - l i q u i d systems are discussed for suspension of free-
settling particles in batch-operated cylindrical tanks. T h e i m p o r t a n t topic of
scale-up of m o d e l tests is included.
F o r an operation in which a slurry of free-settling particles is fed contin­
uously to a n impeller-stirred tank, the particle concentrations within a n d
leaving the t a n k are affected by e q u i p m e n t design, p a r t i c l e - l i q u i d system
properties, and, @f course, the operating variables. M e t h o d s of coping with
s o m e of the difficulties involved are discussed.
Problems caused by agitation of hindered settling p a r t i c l e - l i q u i d systems
are described a n d suitable types of e q u i p m e n t defined. E q u i p m e n t scale-up
a n d m e t h o d s of selecting impeller speed a n d power are also discussed.
Finally, mass transfer between liquid a n d free-settling solid particles is
considered in detail. T h e effects of e q u i p m e n t design variables a n d operating
variables o n mass transfer are described. Dissolution a n d mass transfer-lim­
ited chemical reactions are typical mass transfer operations in s o l i d - l i q u i d
systems. Although crystallization involves mass transfer between particles
a n d a liquid, it is n o t included in this chapter.

II. Particle Settling in Quiescent Liquids


T h e settling behavior of particles a n d the flow characteristics of p a r t i c l e -
liquid systems can be described in t e r m s of several parameters. Particle
diameter, shape, density, concentration a n d liquid density a n d viscosity are
i n d e p e n d e n t variables that determine particle settling velocity a n d the slurry
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 3

rheology (flow properties). Slurry rheology, e q u i p m e n t dimensions, a n d


impeller speed d e t e r m i n e the slurry flow regime (laminar or turbulent) that
will be obtained w h e n a specific e q u i p m e n t geometry is used to process a
p a r t i c l e - l i q u i d system. These e q u i p m e n t , operating, a n d system parameters
also d e t e r m i n e the pattern of transient a n d average particle concentrations
t h a t will exist in a specific piece of e q u i p m e n t a n d the types of problems that
arise (poor concentration uniformity, p o o r circulation, particle breakdown).

A. FREE-SETTLING PARTICLE VELOCITIES


T h e settling velocity ^ of a spherical particle can be estimated from a plot
of the drag coefficient C versus particle Reynolds n u m b e r Όρΐ^ρ^μ, such as
D

Fig. 1 ( 0 2 , 0 5 , chapter 5). Also shown in Fig. 1 are settling velocity correla­
tion lines for irregular-shaped, crushed particles. M e t h o d s of estimating the
particle settling rate for nonspherical particles are discussed briefly by Sa-
kiadis ( S I , p . 65) a n d Oldshue ( 0 5 , chapter 5). Correlation lines for disk-
shaped a n d cylindrical particles are presented by Lapple a n d Sheppard (L3).
However, particle shape is sometimes neglected a n d spheres are assumed.
T e r m i n a l or free-settling velocity of particles can be calculated by trial a n d
error using drag coefficient - Reynolds n u m b e r correlation lines like those in
Fig. 1 a n d the equation

u =[4gD Ap/(3p C )Y<


t p L O
2
(1)
where g is the gravitational acceleration, D the spherical particle diameter,
p

the particle terminal (free) settling velocity, p the liquid density, A ρ the
L

Ke)p
F I G . 1. Drag coefficients C for spherical particles ( 0 2 ) . [From Oldshue ( 0 2 ) . Copyright
D

(1969) American Chemical Society.]


4 J o s e p h Β. Gray and J a m e s Y. Oldshue

particle density m i n u s liquid density, a n d C D the drag coefficient (dimen-


sionless).

B . PARTICLE SETTLING IN MULTIPARTICLE SUSPENSIONS


Particle interaction occurs at relatively low particle concentrations. T h e
m a g n i t u d e of the effect of particle interactions on settling rates can be illus­
trated by the equation [from M a u d e a n d W h i t m o r e ( M l ) ]

^M = (l " Ο " (2)


where is the particle settling velocity in a slurry, 14 the single-particle
settling velocity, c the v o l u m e fraction of solid particles, a n d m = 4.65 for
sv

Stokes' law settling a n d 2.33 for N e w t o n ' s law settling. For N e w t o n ' s law
settling, ujt^ = 0.98 if c = 0.01 a n d 0.89 if c = 0.05.
sv sv

C. TYPES OF PARTICLE-LIQUID SYSTEMS


P a r t i c l e - l i q u i d systems processed in impeller-stirred tanks m a y be di­
vided into groups such as the following:

1. Low-concentration particle-liquid systems with particle settling rates


< 2 . 5 m m / s e c , low viscosity, and Newtonian rheology. T h e particles in
these systems settle so slowly that they are relatively easy to suspend uni­
formly in impeller-stirred tanks. Slurries of such small particles can be
treated like a single-phase liquid w h e n designing appropriate mixing equip­
m e n t . Discussion of the mixing of such slurries has not been included in this
chapter [see Hicks et ah (H2) a n d Oldshue ( 0 5 , chapter 4)].
2. Low-concentration, low-viscosity particle-liquid systems with particle
free-settling rates from 2.5 m m / s e c (0.5 ft/min) to 100 m m / s e c . Slurry
rheology is near N e w t o n i a n . Impeller Reynolds n u m b e r s are > 10 . Particles
4

t h a t settle at velocities from 2.5 to 100 m m / s e c are frequently encountered in


industrially used slurry systems ( 0 5 , chapter 5).
3. Low-concentration systems with particles that settle at velocities > 100
m m / s e c (20 ft/sec). Particles with such high settling rates are very difficult
to suspend ( 0 5 , chapter 5) since they readily b e c o m e stagnant on the b o t t o m
even at high liquid-circulation rates a n d velocities in an impeller-stirred
tank.
4. Moderately pseudoplastic slurries with particles that have free-settling
velocities less than 2.5 m m / s e c and in which particle concentration is not high
enough for a yield stress to exist. Such slurries are not discussed in this
chapter since they behave like single-phase systems [see Oldshue ( 0 5 , chap­
ter 15)].
5. High-concentration particle-liquid systems with 50% or more of the
ultimate particle concentration approached when prolonged settling occurs in
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 5

a stagnant liquid. Such systems have a yield stress or highly pseudoplastic


rheology (described in Section V of this chapter). If the yield stress is > 30 Pa
(300 d y n / c m ) , the particle - liquid system m a y be outside the useful range of
2

propeller- or turbine-stirred t a n k s because of stagnant slurry in parts of the


t a n k r e m o t e from a rotating impeller.

W h e n any p a r t i c l e - l i q u i d system is allowed to settle until the concentra-


tion of particles in the settled layer increases n o further, a highest or ultimate
concentration is reached. In the range from 50 to 100% of this ultimate
concentration, slurries possess a viscosity t h a t is considerably higher t h a n the
viscosity of the slurry liquid phase. At such high particle concentrations,
slurries m a y have a yield stress. This is the lowest stress that m u s t be applied
to initiate particle m o t i o n . T h e yield stress of a p a r t i c l e - l i q u i d system is
affected n o t only by particle size a n d particle concentration b u t also by size
distribution a n d particle shape. A slurry of particles with irregular particle
shapes a n d a spread in particle sizes has a higher yield stress t h a n a slurry of
spheres of the same diameter.
T h e r e is n o sharp transition between free-settling a n d hindered-settling
behavior as particle concentration is increased because the degree of interfer-
ence between particles changes gradually as particle concentration increases.
W h e n impellers are used for suspension of low-concentration, free-settling
particles in low-viscosity (1 cP) liquids, the flow regime in the t a n k is usually
turbulent. If the particle concentration for the same particle - liquid system is
increased until pseudoplastic or Bingham plastic properties exist, then the
impeller-induced slurry flow m a y be turbulent only near the impeller a n d
l a m i n a r or s o m e t i m e s stagnant in parts of the slurry that are r e m o t e from the
impeller.

III. Suspension of Free-Settling Particles in


Batch-Operated, Impeller-Stirred Tanks

A. PARTICLE BEHAVIOR IN IMPELLER-STIRRED TANKS


In a stirred tank, the forces acting o n particles are (a) gravitational w h e n
there is a density difference between liquid a n d particle; (b) inertial d u e to
rotational m o t i o n a r o u n d the impeller axis; (c) viscous a n d inertial due to
drag of liquid o n the particle surfaces when the particle moves relative to its
adjacent liquid; a n d (d) frictional between the surfaces of colliding particles.
Gravitational force m o v e s particles toward the b o t t o m of the tank, b u t
centrifugal force m o v e s liquid radially away from the impeller axis. Drag
forces o n particles t e n d to m o v e particles wherever the liquid goes. These
drag forces also t e n d to modify the liquid flow directions a n d turbulent
velocity fluctuations.
6 J o s e p h Β. Gray and J a m e s Y. Oldshue

T h e suspension of particles at the b o t t o m of a n impeller-stirred t a n k is


probably d u e to turbulent fluctuations near the b o t t o m , where there are n o
average u p w a r d liquid velocities ( C I , M 4 ) . A similar situation exists at the
u p p e r surface of the liquid in the t a n k w h e n impeller speeds are high enough
to carry particles to that elevation. U p w a r d vertical liquid velocities at loca­
tions between the t o p a n d b o t t o m of the liquid induce the upward m o t i o n
t h a t moves particles t h r o u g h o u t most of a n impeller-stirred tank.

B . LIQUID VELOCITIES
Schwartzburg a n d Treybal (S2) showed that, outside the discharge stream
from a rotating impeller, the average liquid velocities a n d also the turbulent
velocity fluctuations are proportional to the impeller speed a n d the square of
the impeller diameter, a n d inversely proportional to the cube root of the t a n k
v o l u m e . F r o m this relationship it can be deduced that the average a n d
fluctuating velocities are proportional to the peripheral velocity of the im­
peller for geometrically similar, different-size e q u i p m e n t . If rotational speed
is changed, the average a n d fluctuating velocities are changed proportionally
according to Musil (M4).

C . PARTICLE SETTLING VELOCITIES


IN IMPELLER-STIRRED TANKS
Although particles t e n d to follow the liquid direction in a n impeller-stirred
tank, they also t e n d to fall toward the b o t t o m of the t a n k when the particle
density is greater t h a n the suspending-liquid density. This settling of parti­
cles m a y have a significant effect o n the impeller speed a n d power required
for satisfactory slurry processing. Free-settling velocities less t h a n 2.5 m m /
sec (0.5 ft/min) are a n exception; they m a y have n o effect in determining the
rotational speed a n d power needed to meet particle suspension require­
m e n t s . Settling velocities in slurries involved in various chemical process
industries frequently range from 2.5 to 100 m m / s e c (0.5 to 20 ft/min).
Particle settling velocities greater t h a n 100 m m / s e c (20 ft/min) m a y require
high power to meet processing requirements. If the particle concentration is
high enough to cause hindered settling, the particle settling velocity is re­
duced a n d becomes less i m p o r t a n t . Such cases are discussed in Section V of
this chapter.
Schwartzberg a n d Treybal (S2) found that the settling velocities of parti­
cles in an impeller-stirred vessel u n d e r turbulent conditions were 30 to 60%
of the settling velocities of the same particles in quiescent liquid. T h e vertical
velocity c o m p o n e n t s of the particles a n d the liquid m o t i o n relative to the
t a n k were measured by streak photography. T h e differences between these
particle a n d liquid velocities are n o t the slip velocities described in the
following.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 7

Similar values of the ratio of turbulent to quiescent settling velocities were


obtained by Kriegel a n d Brauer (K5) for turbulent liquid flow in a pipe. They
found that the settling velocity u of particles in turbulent slurry flow was a
T

function of the pipe Reynolds n u m b e r N a n d the settling velocity of the


Rc

s a m e particle in quiescent liquid as shown in the equation


t/ M=19(7V )- /
T R e
1 2
(3)
for N = 10 , w / W t = 0.6 a n d for N = 10 , w / W t = 0.2. These results for
Re
3
x Re
4
T

pipes are similar to those for impeller-stirred tanks.

D . PARTICLE-LIQUID SLIP VELOCITIES


Slip velocity is a vector difference between particle a n d liquid velocities.
T h e particle moves relative to the liquid. A slip velocity can also exist when
turbulent-liquid velocity fluctuations occur even though the average veloci-
ties of the particle a n d liquid are the same. Slip velocities are i m p o r t a n t in
mass transfer between liquid a n d suspended particles, which is discussed
later in this chapter.
N i e n o w a n d Bartlett (N6) found that slip velocities of particles in turbine-
stirred baffled vessel were approximately twice the terminal settling veloci-
ties of the particles in quiescent liquid. They estimated slip velocities from
N i e n o w a n d Miles's mass transfer data (N7) for approximately 2300-//m-di-
a m e t e r spherical N a C l particles in N a C l - w a t e r solutions.

E . PARTICLE CONCENTRATION PATTERNS


W h e n particles are suspended in a liquid by fluid m o t i o n induced by a n
impeller in a baffled cylindrical tank, particle concentration gradients exist
vertically a n d radially. Unless the impeller speed is very high, particle con-
centrations are higher at the b o t t o m t h a n at the t o p of the slurry in the tank.
Herringe ( H I ) carried out some tests that illustrate various concentration
patterns in a n impeller-stirred tank. H e studied the suspension in water of
solid particles in a narrow size range, using impellers with six 45° pitched
blades in flat-bottom baffled vessels with diameters u p to 1 m ( 3 9 | in.).
T o r q u e m e a s u r e m e n t s for the impellers a n d slurry systems studied were used
as a basis for calculating a n effective slurry density pr* at the impeller from the
equation
N = P 2ng T /p*N D
c Q
2 5
{ (4)
T o r q u e m e a s u r e m e n t s in water were used to calculate the impeller power
n u m b e r N in Eq. (4). A p a r t i c l e - l i q u i d mixture density p was calculated
P m

from the particle density p a n d the liquid density /? using the equation
p L

A = <W>p + U ~0/>L (5)


8 J o s e p h Β. Gray and J a m e s Y. Oldshue

where c is the v o l u m e fraction of particles in the slurry (the ratio of particle


sv

v o l u m e to slurry volume).
A ratio p*/p was calculated for each of several impeller speeds. Plots such
m

as Fig. 2 were obtained in which various types of particle concentration


patterns can be identified. If in region A of Fig. 2 at low impeller speeds all
particles are o n the b o t t o m a n d the impeller is above the settled particles, the
reduced impeller clearance increases the impeller power. If in region A the
impeller is in the settled particles, power is higher because of the drag of
the settled particles o n the impeller blades. As rotational speed is increased in
region B, partial particle suspension occurs, a n d there are particle deposits
between baffles. Because of these deposits, baffling of liquid rotational m o ­
tion is ineffective a n d partial liquid rotation a n d reduced gross circulation
occur. In region C, a further impeller speed increase suspends m o r e particles
a n d increases slurry density a n d impeller power. A further rotational speed
increase in region D reduces power, because slurry density is lowered at the
impeller. A m a x i m u m in the curve of p*/p versus rotational speed was
m

observed by Herringe ( H I ) to occur at the rotational speed N^ at which n o 9

particles rested o n the t a n k b o t t o m m o r e t h a n 1 sec. Musil et al (M3) noted a


similar m a x i m u m near JV in a plot of particle concentration versus rota­
js

tional speed.
At impeller Reynolds n u m b e r s (Ν .\ greater t h a n 10 , particle concentra­
Κ(
4

tions fluctuate at nearly every location in a t a n k containing a near-Newto-

1.0

£« -

Pm

0.6 . A Β C D

Ν s

10 4
105

F I G . 2. Effect of impeller speed or (N \ Re o n effective slurry density at an impeller or impeller


power. [From Herringe ( H I ) . ]
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 9

n i a n slurry. Such fluctuations can be observed visually through the b o t t o m


a n d sides of transparent plastic or glass-walled tanks. However, a n average of
t h e concentrations at each location for long t i m e intervals should be con-
stant.
Particle concentration fluctuations are a function of e q u i p m e n t geometry,
impeller speed, a n d t h e m a g n i t u d e a n d spread of particle settling velocities.
W h e n t h e turbulence scale is as small as t h e impeller blade width [probably at
100 < C/V )i < 1000], particle concentration fluctuations are small. H o w -
Re

ever, w h e n the turbulence scale approaches the t a n k diameter [probably at


( ^ R e ) i> 10 L particle concentration fluctuations are large. T h e turbulence
5

scale increases w h e n impeller rotational speed or the ratio of impeller d i a m e -


ter to t a n k d i a m e t e r increases.
Musil (M4) noted that particles suspended in a t a n k at the rotational speed
Nj occupied only part of the liquid, sometimes only one-third. T h e liquid
S

above the slurry layer m o v e d slowly a n d contained almost n o particles.

F. PARTICLE SUSPENSION EQUIPMENT


Vertical cylindrical t a n k s are c o m m o n l y used for particle suspension
operations. T h e t a n k s are equipped with vertical baffles next to the t a n k
cylindrical wall a n d either a pitched-blade, axial-flow impeller or a vertical-
blade, radial-flow impeller. A n example of a n impeller in a baffled t a n k is
shown in Fig. 3. N o t e t h a t t h e impeller is located at the axis of the cylindrical
t a n k . T h e t a n k b o t t o m m a y be flat, b u t is preferably dished; or a t r u n c a t e d
c o n e can be used (M3). F o r small-scale installations, a propeller is often used
o n a n off-center sloped shaft as shown in Fig. 4. Here, the m o t o r power is 3 h p
(2.2 k W ) or less. F o r applications with power greater t h a n 3 h p (2.2 k W ) , the
contoured-blade airfoil type of pitched-blade impeller shown in Fig. 5 is
being increasingly used instead of flat pitched blades. F u r t h e r information of
e q u i p m e n t geometries r e c o m m e n d e d for agitation of particle suspensions in
t a n k s is presented in Section J.

G . SYSTEM PROPERTY AND EQUIPMENT DESIGN


VARIABLES
A n u m b e r of variables influence particle concentration uniformity in an
impeller-stirred vessel. In addition t o impeller speed a n d particle settling
velocity, p a r a m e t e r s such as impeller type, vessel geometry, baffle geometry,
ratio of impeller t o vessel diameters, n u m b e r of impellers, a n d impeller
locations affect the particle concentration pattern.

H. SUSPENSION PERFORMANCE CRITERIA


H o w well particles are suspended in a liquid by fluid m o t i o n in a t a n k m a y
be j u d g e d by several m e t h o d s that d e p e n d either on visual observation of
10 J o s e p h Β. Gray and J a m e s Y. Oldshue

1 Ν I

F I G . 3. Typical impeller in a baffled cylindrical tank.

particle behavior or o n m e a s u r e m e n t of particle concentrations in various


parts of the p a r t i c l e - l i q u i d suspension in the tank.

1. Rotational Speed for a Desired Particle Behavior


Several m e t h o d s for judging performance of particle suspension equip­
m e n t involve finding a n impeller rotational speed at which a desired particle
behavior is observed w h e n looking into a t a n k with a transparent wall or
window.
a. Fillet Formation. F o r m a t i o n of stable or p e r m a n e n t settled-particle
deposits (fillets) m a y occur at the intersection of a t a n k flat b o t t o m a n d a
vertical cylindrical wall. It is often less costly to permit s o m e of the particles
to form fillets t h a n to use a higher impeller speed a n d power to eliminate the
fillets [Oldshue ( 0 2 , 0 5 , chapter 5)].
b. Adequate Particle Motion. O t h e r criteria that can be used to judge
adequacy of particle suspension involve finding a n impeller speed above
which particle m o t i o n at or near the t a n k b o t t o m is judged to be satisfactory.
T h r e e such examples follow.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 11

F I G . 4. Off-center, sloped-shaft propeller in a cylindrical tank.

(1) No particles rest on the bottom more than 1 sec. T h e impeller speed
at which this condition is m e t is labeled 7V in this chapter. At this particle
js

suspension condition, the particles t h a t stay o n the b o t t o m longest are


usually located directly beneath the impeller or the baffles. M a n y particle
suspension investigators have used this criterion (see Table I). T h e maxi-
m u m in a particle concentration versus rotational speed curve used by Musil
et al. (M3) is related to the often used criterion " n o particles resting on the
b o t t o m m o r e t h a n 1 sec." Musil's m a x i m u m occurs at a slightly lower
rotational speed.
12 J o s e p h Β. Gray and J a m e s Y. Oldshue

F I G . 5 . Lightnin A 3 1 0 pitched blade impeller. (Courtesy of Mixing Equipment Co., a unit of


General Signal.)

(2) Complete on-bottom motion. All particles adjacent to the b o t t o m of


the t a n k are m o v i n g with n o fillets or stagnant particles at other locations o n
the b o t t o m ( 0 3 ) .
(3) Complete off-bottom suspension. All of the particles are moving
with s o m e vertical velocity when they are near the t a n k b o t t o m ( 0 3 , 0 5 ,
chapter 5).

2. Particle Concentration Uniformity


A second category of m e t h o d s for judging the performance of particle
suspension e q u i p m e n t involves removing samples from various locations in
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 13

Table I

Particle Suspension Performance Criteria

Investigator Measuring method Suspension criterion

Baldi etal ( B l ) Visual N o particles o n bottom more than 1 sec


Bohnet and Niesmak (B4) Light beam Relative standard deviation of concentration
attenuation
Bourne and Sharma (B7) M a x i m u m in particle N o particles o n bottom more than 2 sec
concentration vs
rpm plot
Conti et al (C2) Visual N o particles o n bottom more than 1 sec
Einenkel ( E l ) N o t described Particle concentration variance
Einenkel and Merseman (Ε2) Visual Slurry layer at 90% of liquid height
Herringe(Hl) Visual N o particles o n bottom more than 2 sec
Kneule and Weinspach (K2) Visual N o particles o n bottom more than 1 sec
Kolar(K3) Light beam Same attenuation at two elevations in tank
attenuation
Kotzek et al (K4) Visual N o particles o n bottom more than 1 sec
Lamade ( L I ) Visual N o particles o n bottom more than 1 sec
Musil et al (M3) Conductivity M a x i m u m in concentration vs rpm curve
Narayanan et al ( N l ) Visual N o particles o n bottom more than 1 sec
Nienow (N3) Visual N o particles o n bottom more than 1 sec
W e i s m a n and Efferding ( W l ) Visual Height o f slurry interface above impeller
Zwietering(Zl) Visual N o particles o n bottom more than 1 sec

a t a n k containing particles suspended in a liquid a n d measuring the particle


concentrations of these samples.
First, it should be noted t h a t there is n o way to be sure t h a t a sample has the
s a m e particle concentration as the p a r t i c l e - l i q u i d mixture in a stirred t a n k
from it was removed. T h e presence of a sampling device such as a t u b e affects
t h e liquid flow pattern, a n d this in t u r n affects the local distribution of
particle concentrations. Both the orientation a n d size of a sampling device
change liquid velocities a n d directions near the sample withdrawal point.
T h e angle t h a t a t u b e m a k e s with the liquid flow direction is n o t constant
since there are r a n d o m velocity a n d flow direction fluctuations in turbulent
liquid flow in a n impeller-stirred tank. Also, the rate of withdrawal of a
sample in a sampling t u b e can change the particle concentration in the
sample t o values different from those in the particle suspension from which
the sample was withdrawn. T h e magnitudes of these effects have n o t been
defined. F o r further details, see Aeschbach a n d Bourne ( A l ) , Bourne a n d
S h a r m a (B6, B7), Nasr-el-din et al (N2), R u s h t o n ( R l ) , a n d Stevens a n d
Davitt (S7). M e t h o d s of defining particle concentration uniformity are de­
scribed in the following.
14 J o s e p h Β. Gray and J a m e s Y. Oldshue

a. Percent Suspension. T h e percent solids at a sampling point divided by


the percent solids in the whole t a n k is called the percent suspension ( 0 3 , 0 5 ,
chapter 5). F o r example, if a sample has 3 5 % solids in a t a n k which contains
40% total solids, the percent suspension is 35/40 or 87.5% at the location
where the sample was taken. This measure of particle suspension m a y also be
applied to limited particle size ranges. F o r example, if there is 18% in the
— 20 + 40 mesh particle size range, a n d a sample has 20% solids in this size
range, the percent suspension of the —20 + 40 mesh fraction is 20/18 or
111% at the location where the sample was taken.
b. Concentration Profiles. Concentration profiles for each of several par­
ticle size fractions m a y be used as an illustration of particle suspension
performance. T h e percent solids for each of several particle size ranges is
d e t e r m i n e d for samples taken at selected elevations in the tank. W h e n tur­
bulent flow patterns are present, there are fluctuations of such concentra­
tions a r o u n d average values. T h e percent suspension for ranges of particle
sizes calculated as shown in the preceding paragraph m a y be used in these
particle concentration profiles. A n example is shown in Fig. 6 ( 0 3 ) . See also
Bohnet a n d N i e s m a k (B4) a n d Einenkel ( E l ) .
c. Complete Uniformity. If percent suspension is used as a measure of
particle concentration, complete uniformity m a y be defined as all of the

2oo y

0 ιι » >- ^
0.2 0.4 0.6 0.8 1.0
SAMPLE HEIGH T/ LIQUI D HEIGH T
F I G . 6. Particle concentration profile example. A n axial-flow impeller of 2 3 0 m m (9 in.)
diameter was used in a 7 6 0 m m (30 in.) diameter tank with 16% total solids. Impeller power was
105 W (0.14 hp). [From Oldshue ( 0 3 ) . Reproduced by permission of the American Institute of
Chemical Engineers.]
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 15

samples having 100% suspension ( 0 1 , 0 5 , chapter 5). If all samples have the
same size distribution a n d concentrations as the average size distribution
a n d concentration in the whole tank, a condition equivalent to complete
uniformity exists. T h e averages can be calculated from the masses (weights)
of liquid a n d each size fraction placed in the tank. Such a condition can, at
best, only be a p p r o a c h e d b u t never really achieved. F o r this reason the
following definition is m o r e practical.
d. Nearly Complete Uniformity. This condition m a y be said to exist
w h e n increases in impeller speed c a n n o t p r o d u c e a significant change in the
local particle size concentration next to the upper surface of the suspension.

3. ' 'Scale of Agitation' 'for Particle Suspension


Gates et al. ( G l ) described various types of particle behavior in suspen­
sions a n d identified each type by a n u m b e r between 1 a n d 10. This m e t h o d of
delineating particle a n d suspension behavior is presented in Table II. T h e
types of particle behavior that correspond roughly to some of those described
in Sections ΙΙΙ,Η, 1 a n d III,H,2 are noted in the right-hand c o l u m n of Table
II.

I. PARTICLE SUSPENSION CORRELATIONS


A n u m b e r of investigations have been m a d e using cylindrical, baffled, or
flat- or dished-bottom vessels with pitched-blade or vertical-blade impellers.
T h e e q u i p m e n t types a n d sizes used in s o m e of these studies are listed in
Table III. T h e p a r t i c l e - l i q u i d systems used a n d their properties are listed in
Tables IV a n d V. As shown in Table I, the particle suspension criterion used
for most of these studies was " n o particles resting o n the vessel b o t t o m for
m o r e t h a n 1 sec." T h e investigators w h o used this criterion each found the
lowest impeller speed for which t h e criterion was met.
F o r m a n y of these studies, the various investigators correlated the vari­
ables which influence 7V by a n e q u a t i o n such as
js

A* = / ( / * / / > L ) W ^ ^ (6)
where μ is the liquid viscosity, p t h e liquid density, A ρ the particle density
L

m i n u s liquid density, D the particle diameter, D the impeller diameter, c


p { s

the solid-particle concentration, Z) t h e vessel or t a n k diameter, a n d Z the


T c

clearance between the impeller blades a n d the vessel b o t t o m .


T h r e e correlating equations selected as typical are shown in Table VI. T h e
three experimental studies o n which these equations were based have the
s a m e particle suspension criterion a n d e q u i p m e n t geometry. T h e ranges of
system p a r a m e t e r s a n d e q u i p m e n t dimensions used by each investigator
overlapped those used by the others.
F o r a c o m p a r i s o n of the impeller speeds N predicted by these equations,
j s
16 J o s e p h Β. Gray and J a m e s Y. Oldshue

Table II

Definitions of "Scale of Agitation" a

Scale of Roughly equivalent


agitation Definition criteria

1 to 2 Scales of agitation 1 and 2 characterize applications re­


quiring low degrees of suspension to achieve process
results. At a scale of agitation o f 3 agitators will
* produce on-bottom m o t i o n of all the particles in Complete on-bottom
the vessel motion
* permit formation o f particle fillets which are peri­ Fillet formation
odically suspended
3 to 5 Scales of agitation 3 to 5 characterize most chemical pro­
cess industry particle suspension applications. At a
scale of agitation of 3 agitators will
* suspend all the particles completely off the bot­ Complete off-bottom
t o m of the tank suspension
* provide uniform particle concentrations to at
least one-third of the slurry batch height
* be suitable for slurry drawdown at low exit-nozzle
elevations
6 to 8 Scales of agitation 6 to 8 characterize applications in
which the particle suspension approaches uniformity.
At a scale of agitation of 6 agitators will
* provide particle concentration uniformity from
the bottom to 95% of the fluid batch height
* be suitable for slurry draw-off at elevations up to
80% of the fluid batch height
9 to 10 Scales of agitation 9 and 10 characterize applications in
which the particle suspension uniformity is the highest
practical. At a scale of agitation of 9 agitators will
* provide particle concentration uniformity from Nearly complete
the bottom to 98% of the fluid batch height uniformity
* be suitable for slurry draw-off by means of an
overflow at the top of the tank

a
From Gates et al. ( G l ) . Reprinted by special permission by McGraw-Hill, Inc., N e w York.
C o p y r i g h t © 1976.

values of system a n d e q u i p m e n t parameters were selected to be within the


ranges of the experimental studies on which the equations were based. T h e
values of these parameters a n d the calculated N values from each equation
js

are given in Table VII. T h e calculated values of 574, 709, a n d 778 r p m are
within ± 1 5 % . Since P*D\, power would deviate as m u c h as 1.15 or 3

± 50%. Larger deviations can be expected if values of the system a n d equip­


m e n t parameters are outside the scope of the experimental studies because
the exponents of the variables in the three equations used are not the same.
Bohnet a n d N i e s m a k (B4) calculated impeller speeds using nine correlat-
Table III

Equipment Parameters for Particle Suspension Experiments

Investigator D T (mm)* B o t t o m shape Impeller type Di/D T Z /D


C T

Balaietal (Bl) 122, 190, 229 Rat Eight disk-supported blades 0.2-0.33 0.5-1.5
Bohnet and Niesmak (B4) 290 Bat Propeller or four flat pitched 0.34 0.17; 0 . 3 4
blades
Bourne and Sharma (B7) 172 Flat or contoured Propeller 0.47 0.18
172 Propeller 0.47 0.18
122 (draft tube)
Conti et al. (C2) 130, 190 — Eight disk-supported blades 0.25, 0.37 0.05-0.5
Einenkel ( E l ) 365, 7 9 0 Dished Propeller 0.3 0.3
Einenkel and Merseman (E2) 190, 365, 790 Dished Propeller 0.3 0.3
0
Herringe(Hl) 150, 300, 1000 Flat Six 4 5 blades
Kneule and Weinspach (K2) 370 Flat, dished, conical, and Propeller or turbines 0.2-0.4 0.3-0.8
hemispherical
0
Kolar(K3) 1 6 5 , 2 1 0 , 235 Flat Propeller or four 4 5 blades 0.16-0.32 0.11-0.33
Kotzek et al (K4) — Dished Propeller — —
Lamade (LI) — Dished Four 45 ° blades — —
Narayanan et al ( N l ) 114, 141 Flat Eight vertical blades — —
N i e n o w (N3) 140 Flat Six disk-supported blades 0.26-0.52 0.14-0.2
Weisman and Efferding ( W l ) 140, 2 3 8 , 2 8 9 Flat Six vertical blades 0.18-0.43 0.05-0.6
Zwietering ( Z l ) 154, 192, 2 4 0 , 290, 4 5 0 , 6 0 0 Flat, dished, or 120° c o n e Propeller, two vertical blades 0.25-0.4 0.25-0.4
0.2-0.5 0.05-0.5
Six disk-supported blades 0.17-0.7 0.14-1

a
25.4 m m ο 1 in.
Table IV

System Parameters for Particle Suspension Experiments

3 3
Investigator Solid phase Liquid phase /> (g/cm )
p /> (g/cm )
L c , v o l u m e fraction
s v McP)

Baldi etal. ( B l ) Sand Water 5-550 2.65 1 0.0008-0.008 1-3


Bohnet and Niesmak (B4) Polymer Water 1150 1.05 1 0.002-0.02 1
Glass Water 630, 700 2.48 1 0.003-0.10 1
Bronze Water 100 8.85 1 0.002 1
Bourne and Sharma (B7) Glass Water 200-1000 2.64 1 <0.02 1
Conti et al (C2) Glass Water 440 2.65 1 0.0004 1
Benzoic acid Water 3400 1.2 1 0.0004 1-10
Phthalic anhydride Water 3400 1.5 1 0.004 1-10
Einenkel ( E l ) Glass Water 200-630 2.87 1 0.05-0.25 1
Einenkel and Merseman (E2) Glass Water 80-250 2.48 1 0.008-0.25 1
Herringe ( H I ) Sand Water 20-1200 2.65 1 0.06-0.24 1
Gravel Water 2500-5000 2.86 1 0.06-0.24 1
11 and 12 Water 130-170 4.47 1 0.06-0.24 1
Kneule and Weinspach (K2) Sand Water 700-6400 2.63 1 0.0008-0.11 1
Glass Water 100-10000 2.95 1 0.0007-0.1 1
Iron Water 1000-4000 7.7 1 0.0003-0.02 1
Lead Water 1250-8000 11.1 1 0.0002-0.02 1
Kolar (K3) Acrylic Water 570-1640 1.15 1 0.003-0.02 1
Glass Kerosene 130-630 2.98 0.8 0.002-0.1
Narayanan et al ( N l ) Quartz Water 120-680 2.63 1 0.01-0.08 1
N i e n o w (N3) Ballotini glass Water 150-600 2.48 1 0.0004-0.04 1
Weisman and Efferding ( W l ) Thorium oxide Water <4 9.7 1 0.08-0.24 1
Glass Water 45, 140 2.6 1 0.04-0.36 1-20
Zwietering ( Z l ) Sand Water 130, 300, 800 2.65 1 0.002-0.07 1
NaCl Water

a 6
1 0 / / m o 3 9 . 3 6 in.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 19

Table V

Particle Settling Velocities and Reynolds N u m b e r s for Particle


Suspension Experiments

Investigator ^(mm/sec)* (N )p
Re

Baldi etal. (Bl) 0 . 0 2 -- 8 0 0 . 0 0 3 -- 1 0 0


Bohnet and N i e s m a k (B4) 20- -110 20- -200
Bourne and Sharma (B7) 35- -166 20- -440
Conti et al (C2) 70- -240 80- -1,200
Einenkel ( E l ) 30- -120 20- -220
Einenkel and Merseman (E2) 4- -30 1- - 2 0
Herringe ( H I ) 0.3- -600 0.02--9,000
K n e u l e and Weinspach (K2) 9--1,800 3 - -160,000
Kolar (K3) 2 0 - -120 12- - 2 3 0
Narayanan et al ( N l ) 12- - 1 2 0 4- -220
Nienow (N3) 15- -90 6--130
W e i s m a n and Efferding ( W l ) 0 . 1 - -15 0 . 0 0 3 - -5
Zwietering ( Z l ) 1 2 - -120 4 - -250

a
305 m mο l ft.

ing e q u a t i o n s developed by various investigators. T h e calculated rotational


speeds deviated from —56 t o + 2 5 0 % from Bohnet a n d N i e s m a k ' s experi­
mentally d e t e r m i n e d speeds. T h e p a r a m e t e r values inserted in the equations
were those from B o h n e t a n d N i e s m a k ' s tests. T h e large deviations of the
calculated rotational speeds were probably d u e t o using test parameters
outside the range of those used in t h e tests o n which t h e n i n e correlating
equations were based. In addition, t h e s a m e criterion for satisfactory particle
suspension was n o t used in all of t h e n i n e studies. U n d e r these conditions,
agreement a m o n g calculated impeller rotational speeds is n o t expected.

Table VI

Typical Particle Suspension Correlating Equations for Propellers

Equation Reference

(Nf p Dfn/gu Apc y^


s L ss sv = 3\(DfN /v)-° is
0M
Einenkel ( E l ) Fig. 9 line for
σ = 0.95ο
2
particles not
more than 1 sec on bottom
JV = (3ΐ£/7τ) ·
js
0 308
v 0 0 7 7
(A/?//7 ) - L
0 3 0 8
J Dr a 7 7
(c vWss) -
S
0 3 0 8

(D D N nplp)
{ p iS = 5.5[Z) ^(A/>//? )(/? y//) ] (A/^ )
3
L 1
2 05
p
0 5
(F )m
0 2 5
Table 2 in Kneule and Wein­
spach (K2)
^Vjs = (5.5g°yn)(Ap/p )-^Dv ^° L
0
m
25

Nis = (7.2^°- )v°HA/?//?J


45 045
i)r a85
(100Af /M ) p L
013
Z)0- 2
Zwietering ( Z l )
20 J o s e p h Β. Gray and J a m e s Y. Oldshue

Table VII

Sample Calculations o f iV js

Values of variables
Z / Z ) = 0.25
C T
g = 9.82 m / s e c ο 32.2 ft/sec 2 2

A = 91.5 m mο 0.300 ft M /M = 0.1316 (mass of particles)/(liquid mass)


P L

D = 305 m mο 1.00 ft
T
u = 0.021 m/secο 0.07 ft/sec
s

D = 0.200 m mο 0.00787 in.


p u = 0.018 m/secο 0.059 ft/sec
ss

c = 0.05 (particle volume)/(slurry volume)


s v
p = 1000 k g / m ο 62.4 lb/ft
L
3 3

F = 0.1163 (mass of particles)/(slurry mass)


m p = 2500 k g / m
p
3

μ = 0.001 kg/m-sec ~ 1 cP
Calculated impeller speeds N
0
is

Einenkel ( E l ) equation 578 rpm


Kneule and Weinspach (K2) equation 778 rpm
Zwietering ( Z l ) equation 707 rpm

a
Equations are in Table VI.

Table VIII

Effects of Eq. (6) Particle Suspension Variables o n Rotational Speed iV js

Eq. (6) variable, Eq. (6) exponent

Investigator D ,c
p &p/p ,
L b c ,e
s μ, a

Baldieffl/. ( B l ) 0.14 0.4 0.13 0.17 -1 -0.9


Bohnet and Niesmak (B4) 0.3 0.3 0.08 -0.8
Einenkel ( E l ) 0.17-0.67 0.5 0.3 — -1.7 -0.6
Einenkel and Mersemann (E2) 0.15 0.5 0.3 0.08 -0.8
Gates et al. ( G l ) f l
0.13-0.31 0.13-0.2 — 0-0.1 -1.6 -0.8
Herringe(Hl) 0.3 0.4 0.18 — -0.7
Kneule and Weinspach (K2) 0 to-0.2 0.5 0.25 0-0.1 -1.7 -0.5
Kolar(K3) -1.6 -0.7
Kotzek et al. (K4) 0.21 0.5 0.18 — -1 -0.7
Lamade (LI) 0.21 0.5 0.15 — -0.7
Narayanan et al. ( N l ) <0.5 <0.5 <0.3 — -1.0
N i e n o w (N3) 0.21 0.4 0.12 —
Weisman and Efferding ( W l ) 0.17 0.5 0.17 — -1 -0.7
Zwietering ( Z l ) 0.2 0.5 0.13 0.1 -0.85
(propeller) -0.85
-1.5(6-blade
turbine)

a
For an impeller with six 45° pitched-blade impeller in baffled dished-bottom tanks.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 21

In Table VIII are given the exponents obtained in the 16 experimental


studies listed in Tables III - V. T h e e x p o n e n t s for each study a n d the effects of
their associated variables are valid for the ranges of e q u i p m e n t a n d system
p a r a m e t e r s o n which they were based. Since there is a wide divergence of the
e x p o n e n t s for s o m e of the variables in these studies, a wide divergence in the
effects of variables o n JVj, occurs. T h e effects of these variables are described
in the following paragraphs.

1. Effect of Particle Diameter


T h e values of the e x p o n e n t c for particle diameter D range from — 0.2 to
p

0.67 in Table VIII. F o r a positive value of this exponent, an increase in


particle size increases N^. W h e n Dickey ( D l ) expressed the suspension data
of G a t e s et al. in the form of Eq. (6), he found that the settling velocity
e x p o n e n t h a d values from 0.13 to 0 . 3 1 .
Einenkel (Ε 1) found t h a t the particle diameter exponent is 0.67 when the
particle Reynolds n u m b e r (N ) is < 10 a n d 0.17 w h e n (N ) > 100. H o w ­
Ke p Re p

ever, there appears to be n o correlation between the exponent c a n d the


particle Reynolds n u m b e r in the data of Tables V a n d VIII.
A n increase in N- is expected as D is increased since a higher settling
}S p

velocity should require a higher liquid velocity to suspend particles. T h e


small values of the e x p o n e n t c show t h a t particle size has only a small effect
o n JV . A similar small effect of particle size o n the velocity needed for
js

suspension of particles in horizontal-pipe slurry flow is discussed by Sakiadis


(SI, p . 47).

2. Effect of Particle-Fluid Density Difference


As shown in Table VIII, the values of the e x p o n e n t b for Ap/p range from L

0.1 to 0.5. Dickey ( D l ) reported values from 0.1 to 0.2 for this exponent
w h e n he expressed the suspension data of Gates et al. in the form of Eq. (6).
Positive values of this e x p o n e n t are expected since a n increase in the
particle-fluid density difference increases particle settling velocity.

3. Effect of Particle Concentration


T h e values of the e x p o n e n t e for particle concentration, shown in Table
VIII range from 0.1 t o 0.3. Dickey ( D l ) reported exponent values from 0.04
to 0.16. W h e n particle concentration is low e n o u g h t h a t nearby particles d o
n o t interfere with each other, a low e x p o n e n t is expected. However, when
interference between particles is significant, a negative e x p o n e n t is expected.
Increasing particle concentration would t h e n decrease the impeller speed
needed to m e e t the criterion for adequate particle suspension. Einenkel (Ε 1)
presented s o m e data t h a t showed this negative effect for particle v o l u m e
fractions above 0.15.
22 J o s e p h Β. Gray and J a m e s Y. Oldshue

4. Effect of Fluid Viscosity


W a t e r was used in nearly all of the systems listed in Table IV. Liquid
viscosity was either near 1 cP ( 1 0 ~ Pa-sec) or h a d a less t h a n tenfold range of
3

values. T h e e x p o n e n t a for fluid viscosity ranged from 0 to 0.17. Low expo­


n e n t values would be expected w h e n (N ) > 10 a n d fluid flow a r o u n d the
Re p

particles is turbulent, or in the l a m i n a r - t u r b u l e n t transitional range of


1 0 < ( 7 V ) < 1000.
R e p

5. Effect of Changing Impeller-to-Vessel-Bottom


Clearance
Widely divergent effects of changing impeller clearance o n the impeller
rotational speed needed for particle suspension have been reported. Oldshue
( 0 2 ) found that the suspension criterion affected this relationship (see Sec­
tion 111,1,9).
Zwietering (Ζ 1) also showed that the impeller type affects the way in which
clearance changes influence JV . F o r a m a r i n e propeller, decreasing
js Z /D C T

decreases iV since js was found to be proportional to ( Z / Z > ) - when c x


0 2

0.25 < Z /DC < 0.4 a n d 0.16 < DJD < 0.4. H e found n o effect of clear­
T T

ance for disk-supported turbine blades when 1 . 4 < Z / Z ) < 1 a n d C X

0.17 < D /D < 0.7. F o r a two-bladed paddle, decreasing Z /D


{ T decreases C T

since the ( Z / D ) exponent was found to be 0.3 when ZJD = 0.25,


C T X

0.05 < Z /DC < 0.5, a n d 0.2 < D /D


T < 0.7. T h e exponent of Z /D
{ T is C T

probably smaller for the propeller t h a n for the turbine a n d paddle, because
changing the impeller-to-bottom clearance probably does n o t change the
flow pattern m u c h for a propeller b u t does for the turbine a n d paddle agita­
tors.
N i e n o w (N3), Baldi et al. ( B l ) , a n d Conti et al. (C2) found that the
impeller speed needed for particle suspension decreased as clearance was
decreased between a turbine with disk-supported blades a n d the flat b o t t o m
of a baffled cylindrical vessel. N i e n o w a n d Miles (N7) observed the same
trend for a turbine with six disk-supported vertical blades a n d for an impeller
with two vertical flat blades.
Kolar (K3) studied the effect of impeller clearance, impeller diameter,
vessel diameter, a n d other variables o n particle suspension. His suspension
criterion was the lowest impeller speed for which the average light absorption
was the s a m e for photocells at elevations D /4 a n d 3 D / 4 above the vessel
T T

b o t t o m . For a propeller, he found t h a t the rotational speed to m e e t his


particle suspension criterion was decreased by decreasing impeller clearance
if DJD > 0.2 a n d was increased if DJD < 0.2. F o r a n impeller with four
T T

45° pitched blades, he found t h a t the impeller speed for suspension was
decreased by decreasing impeller clearance, with little effect o n this relation­
ship of changing the ratio DJD . In Kolar's tests, 0.16 < DJD < 0.32 a n d
T T

0.11 < Z / D < 0.33.


C T
1 2 . Agitation of Particulate S o l i d - L i q u i d Mixtures 23

K n e u l e a n d Weinspach (K2) included changes in the clearance between


the impeller a n d the vessel b o t t o m in their study of particle suspension. F o r a
propeller in a baffled dished-bottom vessel, they found n o effect of changing
Z /D
C T within the range 0.2 to 0.8.
T h e diversity of the effects of Z /D o n the rotational speed needed to
C T

obtain satisfactory particle suspension is related to the effect of changing


impeller-to-bottom clearance o n flow pattern. F o r a n impeller with six disk-
supported blades in a baffled flat-bottom vessel, N i e n o w (N3) observed that
the flow pattern h a d a strong horizontal stream at the mid-elevation of the
impeller blades. W h e n Z /D C > ^, the stream split into two parts at the
T

vessel wall, o n e flowing u p the wall a n d the other down. T h e latter stream
t h e n t u r n e d inward along the b o t t o m of the vessel a n d returned to the
suction side of the impeller. At Z /D <C all of the impeller discharge
T

stream sloped d o w n to the b o t t o m of the vessel before it reached the cylin­


drical wall, where it t u r n e d u p . F o r this type of flow pattern 7V was lower js

t h a n t h a t for the Z /DC > % flow pattern.


T

C o n t i et al. (C2) presented further data o n the effects described by N i e n o w


(N3). T h e y showed that the transition from two large horizontal vortex rings
above a n d below the impeller to o n e below the impeller occurred w h e n
Z ID
Q T = 0.22. A transition in power n u m b e r also occurred at the same
value of Z /D . C T Power n u m b e r s were found to be lower w h e n
Z /D
C T < 0.22.
T h e decrease in impeller speed found by Baldi et al. (Β 1), C o n t i et al. (C2),
N i e n o w (N3), a n d N i e n o w a n d Miles (N7) is the effect o n e would expect as
impeller-to-bottom clearance is decreased. However, Oldshue ( 0 2 ) a n d
Zwietering ( Z l ) found that the impeller type influenced the relationship,
K o l a r (K3) found a n effect of DJD o n the relationship, a n d K n e u l e a n d
T

Weinspach (K2) found n o effect. These divergent observations are probably


related to the effects of changing impeller clearance o n flow patterns, ac­
cording t o N i e n o w (N3).

6. Effect of Changing the Impeller-to-Vessel-Diameter


Ratio
As shown in Table VIII, values of the e x p o n e n t g for the ratio DJD T

ranged from — 0.8 to — 1.7. F o r a negative exponent, N decreases as DJD is is T

increased. In contrast to this trend, K n e u l e a n d Weinspach (K2) found a


m i n i m u m in a plot of power (and rotational speed) for particle suspension
versus DJD at DJD = 0.3. T h e y found n o effect of impeller type or bot­
T T

t o m shape o n this relationship. Oldshue ( 0 1 ) also found such a m i n i m u m at


DJD = 0.3 for a pitched-blade turbine. Bohnet a n d N i e s m a k (B4) found
y

that DJD = 0.5 provided m o r e uniform particle concentrations through­


T

o u t a propeller-stirred vessel t h a n higher or lower values of DJD . T

Zwietering ( Z l ) found t h a t the value of the e x p o n e n t g is affected by


24 J o s e p h Β. Gray and J a m e s Y. Oldshue

impeller type. F o r example, g for a propeller is between — 0.8 to — 0.9 a n d for


a six-bladed turbine is — 1.5.
Kolar (K3) found that changing Z /D C changes how DJD affects N- for a
T T }S

propeller b u t n o t for a n impeller with four 45° blades. T h e exponent g for


DJD changes from — 1.56 to —2.6 w h e n Z /D
T C is changed from 0.33 to
T

0.11.
T h e diversity of the effects of DJD o n the rotational speed needed to
T

obtain satisfactory particle suspension is difficult to explain, b u t it is proba­


bly related to the effect of DJD o n the flow pattern, as discussed above in the
T

paragraphs o n the effects of changing the impeller-to-bottom clearance.


T h e r e m a y be an interaction between impeller clearance a n d impeller d i a m ­
eter. At a different clearance between the impeller blades a n d the vessel
b o t t o m , the effect of impeller diameter is different.

7. Effect of Impeller Type on Particle Suspension


Since the type of impeller used for particle suspension has a strong effect
o n the liquid flow pattern in a stirred baffled tank, one would expect that the
impeller type would affect the rotational speed a n d power needed for ade­
q u a t e particle suspension.
Using multiple impellers should have only a small effect, if any, on 7V . js

Adding additional impellers will have little effect o n the liquid velocity at the
b o t t o m of the t a n k if the distance between impellers is greater t h a n
the impeller diameter a n d the clearance between the lower impeller a n d the
b o t t o m r e m a i n s the same. However, if suspension performance is judged by
the height of the interface between slurry a n d clear liquid, the n u m b e r of
impellers will affect the impeller rotational speed to obtain a desired slurry
interface height (see W e i s m a n a n d Efferding ( W l ) ) . Similar statements can
be m a d e in regard to the effect of changing slurry inventory in a stirred tank.
S o m e experimental comparisons of different impeller types have been
m a d e . Zwietering ( Z l ) a n d Kolar (K3) each provided a basis for c o m p a r i n g
the particle suspension ability of different impeller geometries. K n e u l e a n d
Weinspach (K2) a n d Oldshue ( 0 3 ) also m a d e such comparisons.
F o r DJD < ^, Zwietering ( Z l ) found that a m a r i n e propeller required a
T

lower iV t h a n did a R u s h t o n turbine of the same diameter. Since the power


js

n u m b e r s for propellers are m u c h lower t h a n those for turbine impellers, this


m e a n s that the power needed by the propeller was less t h a n the power for the
turbine. K n e u l e a n d Weinspach's results (K2) are consistent with this rela­
tionship.
Kolar (K3) carried out a particle suspension study in which the suspension
performances of a m a r i n e propeller a n d a n impeller with two 45 ° pitched flat
blades were c o m p a r e d . T h e particle a n d liquid properties a n d the DJD a n d
T

Z /D
C T values were the same. Kolar used light absorption at two elevations in
the particle suspension as a suspension criterion. In contrast to Zwietering's
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 25

Table IX

Properties o f Particles U s e d in Comparison o f Particle Suspension by


Axial-Flow and R a d i a l - R o w Impellers 0

Mesh +100 - 1 0 0 + 120 - 1 2 0 + 140 - 1 4 0 + 170 -170


Percent 9 14 47 14 24
Settling velocity, 25 m m / s e c ο 5 ft/min
Particle density, 3.9 g / c m 3

Particle concentration in water, 30%


Slurry density, 1.3 g / c m 3

a
From Oldshue ( 0 3 ) . Reproduced by permission of the American Insti­
tute o f Chemical Engineers.

results, roughly the s a m e power was needed. T h e different effects of changing


impeller type found by Kolar a n d Zwietering m a y be d u e to their using
different particle suspenion criteria.
A n o t h e r c o m p a r i s o n of axial-flow a n d radial-flow impellers was m a d e by
O l d s h u e ( 0 3 , 0 5 , chapter 5), w h o used a 438 m m (17± in.) diameter flat-bot­
t o m t a n k with four 38 m m ( 1 ^ in.) wide baffles a n d a slurry level equal to the
t a n k diameter. T h e particles used are those described in Table IX. Slurry
sample tubes [9.5 m m (f in.) diameter] were located as shown in Fig. 7.

S A M P L E L O C A T I O N S- P L A N V I E W
F I G . 7. Sample withdrawal locations for data in Table XI and Fig. 8. Arrows indicate direc­
tion of slurry flow into the sample tubes. Sampling elevations were 0 . 2 , 0 . 4 , 0 . 6 , 0.8, and 0.95
times the slurry depth in the tank. [From Oldshue ( O l , 0 3 ) . Reproduced by permission o f the
American Institute of Chemical Engineers.]
26 J o s e p h Β. Gray and J a m e s Y. Oldshue

20j ' r

t£151 -

co5 L
4I l Iι I I I I J
I

45 6 8 I O1 52 03 04 05 0
P/V,Hp/100 0 GA L
F I G . 8. Comparison of particle suspension uniformity with a propeller, a 45° pitched-blade
impeller, and a turbine with back-curved blades (1 h p / 1 0 0 0 gal.ο 197 W / m ) . See Fig. 7 for
3

sample locations and Table IX for particle properties. Axial-flow impeller with DJD = 0.35 in
T

a baffled 438 m m (17± in.) diameter tank (ZJD = 1). [From Oldshue ( 0 3 ) . Reproduced by
T

permission of the American Institute o f Chemical Engineers.]

Twenty samples were taken for each test, a n d the standard deviation was
calculated. T h r e e types of 152 m m (6 in.) diameter impellers were used: a
m a r i n e propeller, a four-bladed impeller with 45° pitched blades, a n d a n
impeller with six back-curved vertical blades. T h e results in Fig. 8 show that
the propeller a n d the 45° pitched-blade impeller h a d roughly the same
suspension performance at the same power. However, the curved-blade
turbine required at least twice the power of the pitched-blade impellers for
the same particle concentration standard deviation.
Oldshue ( 0 3 ) carried out other particle suspension tests in which impeller
type, vessel b o t t o m shape, a n d baffle radial width were changed. These tests
were m a d e in a 457 m m (18 in.) diameter vessel using water containing 30%
a l u n d u m particles, which h a d a 25 m m / s e c (5 ft/min) average settling veloc­
ity. Particle a n d liquid properties are s u m m a r i z e d in Table IX. O n e impeller
was 127 m m (5 in.) in diameter a n d had four 45° pitched blades. T h e other
impeller, also 127 m m (5 in.) in diameter, had six disk-supported vertical
blades. Test results are s u m m a r i z e d in Table X . For use as a particle suspen­
sion criterion, a n impeller rotational speed (and the corresponding power)
was selected at which the particle suspension appeared to be visually uniform
in a transparent tank. F o r a slurry with a low 25 m m / s e c (5 ft/min) settling
velocity, a n o n u n i f o r m suspension could readily be distinguished visually
from a uniform suspension.
T h e axial-flow impeller required considerably less power t h a n the radial-
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 27

Table X

Effect of Impeller and Tank Geometry o n Particle S u s p e n s i o n * 6

Impeller relative power

Bottom shape Baffle radial width Pitched blades Vertical blades

Flat D /\2
T 1.0 3.5
Dished D /\2
T 1.3 2.5
Rat D /24
T 1.7 2.5

a
From Oldshue ( 0 3 ) . Reproduced by permission of the American In-
stitute of Chemical Engineers.
* Particle properties are in Table IX.

flow impeller for the tests in Table X . However, changing from a flat to a
dished b o t t o m increased the power a n d rotational speed needed by the
axial-flow impeller, b u t decreased t h a t needed by the radial-flow impeller.
W h e n the baffle radial width was divided by two in the flat-bottom vessel, the
p o w e r for suspension was increased for the axial-flow impeller a n d was
decreased for the radial-flow impeller.
N i e n o w a n d Miles (N7) used three impeller types in a n experimental study
t h a t included changes in DJD , vessel size, a n d clearance between the im-
T

peller a n d the vessel b o t t o m . T h e impeller speed for particle suspension was


affected by all four of these parameters b u t in ways that are difficult to
describe.

8. Effect of Bottom Shape


Table X shows Oldshue's c o m p a r i s o n s of power to suspend particles for
flat- a n d dished-bottom vessels. H e found t h a t a n axial-flow impeller with
four 45 ° pitched blades required 30% m o r e power in a dished-bottom vessel
t h a n in a flat-bottom vessel. However, a radial-flow impeller required 4 0 %
m o r e power in a flat-bottom vessel t h a n in a dished vessel.
K n e u l e a n d Weinspach (K2) found t h a t the power to suspend particles in a
flat-bottom vessel for a radial-flow flat-blade impeller was m o r e t h a n five
times the power needed for the s a m e impeller in a dished-bottom vessel. O n
the other h a n d , the power needed for the s a m e impeller in a vessel with a
hemispherical b o t t o m was 50% of t h a t for the dished-bottom vessel.
T h a t b o t t o m shape has a n effect o n impeller power a n d rotational speed
required to obtain satisfactory particle suspension is expected because of the
effect of shape o n the liquid flow direction a n d velocity adjacent to the
b o t t o m . F o r a radial-flow impeller in a flat-bottom tank, the liquid velocity
at the b o t t o m of a vertical cylindrical wall would be less t h a n the velocity at
t h e r o u n d e d corner of a dished b o t t o m .
B o u r n e a n d S h a r m a (B6, B7) carried o u t particle suspension tests in a
28 J o s e p h Β. Gray and J a m e s Y. Oldshue

FIG. 9. Propeller and draft tube. Baffles not shown. [Adapted from Bourne and Sharma
(B7).]

propeller-agitated, cylindrical, baffled, dished-bottom t a n k with a draft tube


(see Fig. 9). A coaxial cone directly below the propeller diverted the propeller
discharge stream radially at the vessel b o t t o m . T h e flow was then changed
from horizontal to vertical by the dished b o t t o m at its intersection with the
vertical vessel wall. T h e cross-sectional areas of the draft tube a n d of the
a n n u l u s between the draft t u b e a n d the vessel wall were equal. W h e n 0 . 3 - m m
particles with a settling rate of 60 m m / s e c were used, a propeller speed of 320
r p m was required to prevent particles from resting o n the b o t t o m m o r e t h a n
2 sec. A propeller a n d flat-bottom t a n k of the same diameters as those used in
the draft t u b e a n d c o n t o u r e d - b o t t o m tests required 450 r p m . T h e rotational
speeds needed for particle suspension in these two designs b e c a m e nearly the
s a m e w h e n 1-mm particles were used a n d the settling velocity was 170
m m / s e c (33 ft/min).

9. Effect of Particle Suspension Criterion


Several investigations were m a d e in which the particle suspension criteria
were different from the often-used rotational speed for n o particles resting o n
the b o t t o m m o r e t h a n 1 sec (see Table I).
T h e type of criterion used to evaluate particle suspension was found by
Oldshue ( 0 2 ) to affect the relationship between impeller-to-bottom clear­
ance a n d power (or impeller speed) for suspension. As shown in Fig. 10,
w h e n the suspension criterion was complete uniformity, decreasing the
clearance decreased the power (and rotational speed) required. However,
w h e n a n off-bottom suspension criterion was used, decreasing the clearance
increased the power. T h e data in Fig. 10 were obtained for a 203 m m (8 in.)
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 29

ll • , , 1
0.10. 20. 30. 50. 81. 0
Z
c/ i
D

F I G . 1 0 . Effect o f particle suspension criterion and impeller off-bottom position o n impeller


power for particle suspension. 3 0 % alundum slurry with 2 5 m m / s e c ( 5 ft/min) average settling
velocity. Axial-flow impeller (DJD T = 0 . 4 5 ) in a baffled tank {ZJDT= 1 . 0 ) . [Oldshue ( 0 2 ) .
Copyright ( 1 9 6 9 ) American Chemical Society.]

d i a m e t e r impeller in a 432 m m (17 in.) diameter tank. A 30% a l u n d u m


slurry was used with particles (specific gravity of 4) that h a d a 25 m m / s e c (5
ft/min) settling velocity.
O l d s h u e ( O l ) carried o u t a particle suspension test at a n impeller speed
a n d power t h a t provided complete off-bottom suspension. In these tests, he
used a 438 m m (17.25 in.) d i a m e t e r t a n k with four baffles a n d a 438 m m
(17.25 in.) d e e p slurry stirred by a 152 m m (6 in.) diameter impeller with four
45 ° pitched flat blades. T h e particle concentration was 30 mass percent, a n d
the particle settling rate was 25 m m / s e c (5 ft/min). Samples were taken at
five elevations above the t a n k b o t t o m for each of the locations shown in Fig.
7. Percent suspension results are given in Table XL These data show that
complete off-bottom suspension is equivalent to a relative standard devia-
tion of 0.12 for the specific e q u i p m e n t a n d p a r t i c l e - l i q u i d system used.
Einenkel ( E l ) measured particle concentrations at 13 locations in pro-
peller-agitated baffled tanks. T h e variance of each set of 13 concentration
values was used as a m e a s u r e of particle concentration homogeneity. A
variance of 0.95 was stated to be equivalent to all particles in m o t i o n with
n o n e r e m a i n i n g for m o r e t h a n 1 sec o n the b o t t o m . This relationship is
probably applicable only to the specific e q u i p m e n t a n d p a r t i c l e - l i q u i d sys-
t e m used.
30 J o s e p h Β. Gray and J a m e s Y. Oldshue

Table XI

Equivalent Values o f Complete Off-Bottom Suspension


and Standard Deviation of Percent Suspension Data ( 0 1 )

Percent suspension when height of


sample tube off bottom ZJD is T

Sample point* 0.055 0.5 0.75 0.85 0.945

1 — 104 104 91.5 83


2 95.5 109 105 97.5 89
3 118 113 104 99 95
4 94.5 109 105 105 85
5 117 110 95 91 65

Average o f all samples 99.3% suspension


Standard deviation for all
samples 11.9
Relative standard deviation 11.9/99.3 = 0.12

a
See Fig. 7.

Practically nothing is k n o w n a b o u t the effects of the particle suspension


criterion o n the correlating relationships such as Eq. (6), a n d obtaining
particle concentration data o n which to base such relationships is very te­
dious.

10. Effect of Equipment Size on Particle Suspension


F o r geometrically similar different-sized e q u i p m e n t , a n y dimension can
be selected for use in evaluating the effect of e q u i p m e n t size o n particle
suspension performance. Such information o n the effect of e q u i p m e n t size
provides a relationship for scale-up of m o d e l tests to larger mixing equip­
m e n t . Only the experimental basis for such particle suspension scale-up
relationships will be discussed here. Additional information will be pre­
sented on the design of e q u i p m e n t for particle suspension operations later in
this chapter (see Section III,J).
Table VIII shows values of the exponent d for the variable D in Eq. (6). {

These values range from —0.5 to —1.0. T h e negative sign denotes that N js

decreases as e q u i p m e n t dimensions, represented by D , increase. Zwietering


{

(Ζ 1) found n o effect of changing the impeller type o n the — 0.85 value of the
e x p o n e n t of D that he obtained in his studies. This implies that the exponent
x

of D is i n d e p e n d e n t of e q u i p m e n t type. A — 0.67 exponent corresponds to a


{

constant ΡIV ratio for geometrically similar e q u i p m e n t operating in the


turbulent flow regime.
In part, at least, the spread in values of the exponent d m a y be d u e to an
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 31

effect of particle size. Herringe ( H I ) showed that d increased from — 1 to


—0.4 w h e n the average particle size was increased from 0.02 to 1 m m . T h e
—0.5 value of d obtained by K n e u l e a n d Weinspach ( K 2 ) is based o n suspen-
sion of 1- t o 10-mm particles m a d e of glass, iron, a n d lead. Most of the
studies in Table VIII with values of d less t h a n —0.7 involved < 1-mm glass
or sand particles in water. These data support the effect of particle size o n the
e x p o n e n t d shown by Herringe ( H I ) .
B u u r m a n et al. (B8) carried o u t particle suspension tests in baffled tanks of
widely different diameters (0.48 a n d 4.3 m ) . T h e y used commercially avail-
able sand at 15% particle concentration in water. A narrow particle size range
was used with a 157 m average diameter. A D = 0.4Z> impeller with four
{ T

45° pitched flat blades a n d Z = D /3 was used. T h e rotational speed at


c T

which n o particles rested o n the b o t t o m m o r e t h a n 1 sec was used for the tests
in the smaller tank. A n ultrasonic D o p p l e r velocity meter was used in the
larger t a n k to find the impeller rotational speed that m e t the equivalent
criterion (a sharp transition between a stationary a n d a m o v i n g 1-cm-thick
sand layer o n the b o t t o m of the larger tank). In a n o t h e r series of tests, particle
concentrations were measured for samples taken at various heights in the
vessels by drying 1 - a n d 7-liter samples from the smaller a n d larger tanks a n d
weighing t h e solids.
T h e e x p o n e n t d in the relationship 7V °c Df was found to be 0.67 for tests
js

using t h e particles n o t resting o n the b o t t o m m o r e t h a n 1 sec criterion. In


these tests t h e impeller blade thickness was proportional to the impeller
diameter. This e x p o n e n t is equivalent to scaling u p at the same P/V.
W h e n the particle suspension performance criterion was the same frac-
tional height Z /DH T of the particle concentration h o m o g e n e o u s zone, the
rotational speed N to achieve the s a m e Z /D was found to be proportional
H H T

to D]~ on
for the two vessels tested. T h e m e t h o d of judging particle suspen-
sion performance, then, has a n effect o n the scale-up relationship.
B u u r m a n et al. (B8) also consider the effect of impeller blade thickness o n
scale-up of particle suspension in stirred tanks. T h e y s u m m a r i z e some theo-
retical a n d experimental d a t a t h a t support their tentative conclusion that
blade thickness m u s t be proportional t o impeller diameter for N « D^ js to
2/3

b e a n appropriate m e t h o d of scale-up. T h e y believe that part of the scatter in


values of the e x p o n e n t d is d u e to blade thickness deviating from geometric
similarity.

11. Comments on Effects of Variables on Particle


Suspension
F r o m the t r e m e n d o u s a m o u n t of information in the references cited,
s o m e general conclusions are possible. These are restricted to particle sus-
pension u n d e r turbulent flow conditions in impeller-stirred baffled cylindri-
32 J o s e p h Β. Gray and J a m e s Y. Oldshue

cal tanks or vessels. Increasing particle diameter increases moderately.


Increasing p/p has a larger effect o n iV . Both increasing liquid viscosity a n d
L js

increasing particle concentration increase iV slightly. Increasing the ratio of


js

liquid depth to t a n k diameter ZJD increases N . Increasing the ratio DJD


T js T

decreases N . Increasing the ratio Z /D


is C T usually increases iV . Impeller
js

geometry (blade angle, blade shape), t a n k b o t t o m shape (flat, dished, coni­


cal), a n d e q u i p m e n t size all affect 7V . Even the suspension criterion has a
js

significant effect on 7V . In some cases, interactions a m o n g operating, equip­


js

m e n t , a n d system parameters were found to occur, a n d when a p a r a m e t e r


was changed, the effect of a n o t h e r p a r a m e t e r o n iV was changed. O n e should
js

keep in m i n d that the effects of these variables o n impeller power are m u c h


greater t h a n their effects o n iV , since Ρ <* Ν .
js
3

J. EQUIPMENT DESIGN
Here, the factors that should be considered in the design of agitated cylin­
drical vessels for suspension of free-settling particles in batch operations are
discussed. T h e p r o d u c t i o n rate desired a n d the batch cycle t i m e determine
the vessel capacity. Of course, the batch cycle t i m e depends o n the require­
m e n t s of the processes or operations taking place, such as chemical reaction,
dissolution of particles, a n d heat transfer, a n d on the operating d e m a n d s
such as t i m e to fill the vessel a n d discharge its contents.

1. Vessel, Baffle, and Impeller Geometry


Impeller-stirred baffled cylindrical vessels a n d tanks are c o m m o n l y used
for suspending free-settling particles in liquids. Ratios of vessel straight-side
length to vessel diameter often range from one-half to one. R o u n d e d bot­
t o m s such as elliptical A S M E (American Society of Mechanical Engineers)
flanged a n d dished, or standard dished, vessel heads are preferred to flat
b o t t o m s to reduce the tendency of particles to deposit at the intersection of
the vessel b o t t o m a n d the cylindrical vessel wall. F o u r to eight radially a n d
vertically oriented baffles are used with a radial baffle width to fV of the
vessel diameter. A gap between the baffle a n d the vessel wall equal to the
vessel diameter divided by 30 to 70 should be used. Ordinarily, the baffles
extend from the lower end of the vessel straight side to the highest expected
liquid level.
W h e n less t h a n 3 h p (2.2 k W ) is needed for satisfactory particle suspension
in a vessel, m a r i n e propellers are often used without baffles. T h e agitator
shaft in such cases is located off-center in the t a n k a n d 10 to 15° from
vertical. F u r t h e r details are described by Bates et al. (B3). Diameters a n d
clearances are usually within the ranges £ < DJD < % a n d 1 < Z /D T < 3,
c i

respectively.
F o r agitator drives greater t h a n 3 h p (2.2 k W ) , impellers are used coaxially
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 33

in baffled cylindrical vessels. T h e ranges of diameters a n d clearances are


usually within ^ < DJD < i a n d i < Z / Z < ^, respectively.
T C L

F o u r or six 45 ° pitched flat blades are often used. However, impellers with
three specially c o n t o u r e d pitched blades that have a decreasing angle from
horizontal progressing from the shaft to the blade tips are becoming m o r e
prevalent. Such a n impeller is shown in Fig. 5. Probably, a n impeller with
this design achieves m o r e satisfactory particle suspension t h a n a pitched
flat-blade impeller at the same power c o n s u m p t i o n because it produces a
higher flow rate a n d a m o r e nearly axial discharge stream with lower turbu-
lence t h a n a flat pitched-blade impeller does.
T h e e q u i p m e n t geometries t h a t have often been used for particle suspen-
sion in agitated vessels m a y n o t be o p t i m u m for the lowest power c o n s u m p -
tion that achieves satisfactory suspension. B o u r n e a n d S h a r m a (B6, B7)
showed t h a t the particle suspension performance of a propeller in a baffled
cylindrical vessel can be i m p r o v e d by using a propeller in a draft tube a n d a
dished b o t t o m to which a coaxial cone has been added o n the vessel b o t t o m
below the propeller as shown in Fig. 9. This cone diverts the propeller
discharge stream from vertical at the propeller to horizontal at the vessel
b o t t o m . W i t h such a geometry, the liquid velocities in the vessel are directed
to obtain strong vertical c o m p o n e n t s ( 0 5 , chapter 20, Section C), a n d the
scale of the largest turbulent m o t i o n in the vessel is reduced. A n evaluation of
whether such particle suspension e q u i p m e n t would be economical has n o t
appeared in the technical literature. T h e higher e q u i p m e n t cost of the draft
t u b e a n d special vessel b o t t o m would have to be at least balanced by the
power cost savings which provide a desired payoff period.
T o prevent particle settling o n the b o t t o m when a t a n k is emptied, verti-
cal-blade impellers m a y be used with a blade-to-bottom clearance as small as
practical. F o r such cases, the lower edges of the impeller blades m a y be
shaped to m a t c h the vessel b o t t o m a n d the impeller diameter specified to be
within i < DJD < \.
T

2. Selection of Impeller Rotational Speed


T h e rotational speed used for particle suspension in impeller-stirred t a n k s
d e p e n d s o n (a) e q u i p m e n t parameters such as geometry a n d size, (b) system
p a r a m e t e r s such as particle size, density, a n d concentration a n d liquid vis-
cosity a n d density, a n d (c) the suspension criterion which m u s t be satisfied.
Lyons (L6, p . 256) described a m e t h o d for selecting rotational speed based
o n a 3.5 m / s e c (700 ft/min) impeller peripheral velocity in turbine-agitated
vessels. Gates et al ( G l ) developed tables to predict impeller power needed
for particle suspension. See also Oldshue ( 0 5 , chapter 5). In m a n y cases,
scaleup of m o d e l tests is used to predict the rotational speed needed in large
equipment.
34 J o s e p h Β. Gray and J a m e s Y. Oldshue

a. Prediction of Impeller Speed by a Correlation. Gates et al. ( G l ) de­


scribed a m e t h o d of selecting power a n d rotational speed for particle suspen­
sion in cylindrical t a n k s with four baffles a n d o n e or m o r e pitched-blade
impellers. N o experimental basis was given for the published tables of data,
b u t a conservative prediction of power a n d rotational speed is claimed. T h e
pitched-blade impeller m a y have four blades with a projected vertical height
equal to DJ1, or six blades with a projected vertical height equal to D /S. {

T h e following equation relating rotational speed, impeller diameter D , {

t a n k diameter D , design corrected settling velocity w , a n d K can be de­


T d 7

rived from data published by Gates et al. ( G l ) :

N m = K^D^u^/Df 33
(7)

where K is a constant (see Table XII), u is the design settling velocity of


7 d

particles in feet per m i n u t e , a n d N the impeller rotational speed in revolu­


m

tions per m i n u t e .
Values of K are listed in Table XII with a description of the physical
7

significance of s o m e of the numerical values of agitation intensity which


Gates et al. called "scale of agitation." T h e design particle settling velocity u d

is related t o free-settling or terminal velocity by the equation u = f^, d

where f is a function of the particle concentration as shown in Table XIII.


w

T h e following equation can be derived from Eq. (7) when N = g P/ P c

p N Df
h
3
= 1.6, Ρ = 0 . 8 5 P , p = 62.4 lb/ft , a n d D = Z :
m h
3
T L

Pm = 7.6 Χ 1 0 " 11
K ' u° N°^ V
0
7
57
d
57 55 lA3
(8)

where P m is the m o t o r power in horsepower a n d F t h e liquid volume, gal.

[Note: F o r converting SI units to those shown for the t e r m s given in Eqs. (7)

Table XII

Scale of agitation and K Values for Eq.


7 (l) a

Scale of
agitation K X 10~
7
8
Description

1 0.29 All the particles are in motion. At the intersection of the flat vessel
bottom and the cylindrical vessel wall, there may be moving fillets
of particles which are periodically suspended.
3 1.11 All the particles with a settling velocity less than u are completely off d

the vessel bottom, and the particle concentrations are relatively


uniform to at least one-third o f the slurry height.
6 2.7 Particle concentrations are uniform for 95% o f the slurry height.
9 12.6 Particle concentrations are uniform for 98% of the slurry height.

From Gates et al. ( G l ) . Modified by special permission by McGraw-Hill, Inc., N e w York.


a

C o p y r i g h t © 1976.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 35

Table XIII

Correction Factors f for Calculating Design Settling w

Velocity u from Settling Velocity uf- d


b

Solids, wt. % 2 5 10 15 20 25
Factor f w 0.8 0.84 0.91 1.0 1.1 1.2

a
u =f u .
d Vf t

b
From Gates et al. (G1). Modified by special permis­
sion by McGraw-Hill, Inc., N e w York. Copyright ©
1976.

a n d (8), the following unit equivalents are applicable: 1 k W ο 0.746 h p ,


1 m / s e c ο 196.8 ft/min, 1 m ο 264 gal.] 3

F r o m Eq. (7), P/V* N D , a n d T oc P/N, o n e can derive the following


3 2
Q

relationships w h e n DJD a n d u are constant:T d

P/Voc Ζ>Γ<>.25 oc J)-0.25 α J/-0.083 (9)

T/ Q FOC / ) 0 . 5 0 α £0.50 α Τ/0.167 ( 1 0 )

F r o m the above relationships, o n e m a y infer that the tables of Gates et al.


(G1) are closer to constant Ρ I F t h a n they are to constant T /V. F o r example, Q

if the vessel v o l u m e is multiplied by 1000, P/V is multiplied by


(lOOO)" = 0.56 a n d r / F i s multiplied by ( 1 0 0 0 )
0083
Q = 3.2. 0 1 6 7

F r o m Eq. (7) o n e m a y derive the following relationship when D a n d u T d

are constant:
N Dt**Kjm (11)
If D is adjusted to keep K constant w h e n a different N
{ 7 m is used, t h e n
Α α Λ Γ -1/2.33 ( 1 2 )

Since T oc N D\ a n d 7V= N / 6 0 , substituting Λ ^


Q
2
m
1 / 2
· 3 3
for D in the t o r q u e
{

proportionality yields the relationship


T ocN-° Q
145
(13)
Since Ν has a small exponent, T does n o t change m u c h at the same F a n d Q

D w h e n Ν is changed a n d D is adjusted to keep the same value of K . (The


T { 7

s a m e particle suspension performance, or Gates scale of agitation.) For


example, twice the rotational speed multiplies the t o r q u e by 2 ~ = 0.90. 0 1 4 5

Such nearly constant torques at the same V, the same u , a n d the same K d 7

(the s a m e scale of agitation) are found in the tables of Gates et al. ( G l ) .


T h e use of Zwietering's e q u a t i o n (see Table VI) to calculate iV has been js

r e c o m m e n d e d by C h a p m a n et al. ( C I ) for propellers, 45° pitched-blade


turbines, a n d turbines with vertical disk-supported blades. T h e y carried out
tests in baffled t a n k s with diameters from 0.28 to 1.8 m (11 to 72 in.).
36 J o s e p h Β. Gray and J a m e s Y. Oldshue

W h e n particle suspensions flow like single-phase near-Newtonian fluids


a n d particles settle at less t h a n 2.5 m m / s e c (0.5 ft/min), the m e t h o d of
selecting agitator power a n d impeller speed described by Hicks et al. ( H 2 )
might be considered. T h e y define a bulk flow velocity u in a pitched-blade
h

impeller-stirred t a n k as the impeller discharge flow rate or p u m p i n g rate q p

divided by the t a n k cross-sectional area nD\/A [see also ( 0 5 , chapter 4)]:

(14)

Tables of r e c o m m e n d e d m o t o r power are provided by Hicks et al ( H 2 ) as a


function of t a n k volume, fluid viscosity, a n d bulk flow velocity.
b. Scale- Up of Model Tests. W h e n the vessel a n d agitator geometry or the
particle properties are outside the scope of a n available correlation for pre­
dicting impeller speed or power, a m o d e l of a proposed large-scale agitated
vessel m u s t be used to find experimentally the lowest speed t h a t meets the
desired particle suspension criterion a n d other process a n d operating re­
quirements.
T h e scale-up relationship

WJi/WJi = [(Α) /(Α)ι]'


2
(15)
m a y be derived from Eq. (6) w h e n e q u i p m e n t dimensions are geometrically
similar a n d all other variables are constant except D a n d iV . x js

Results of experiments o n the effect of e q u i p m e n t size described in Section


ΙΙΙ,1,10 show that the value of the exponent d for use in Eq. (15) is uncertain.
T h e use of d = — 1 m a y involve too m u c h risk a n d — f is conservative. T h e
value of d derived from Eq. (7), which is based o n the data of Gates et al. (G1),
is —0.75. C h a p m a n et al. ( C I ) found a n exponent o f — 0 . 7 6 . T h e use of
d = — 1 is equivalent to constant T or constant ND for geometrically simi­
Q {

lar e q u i p m e n t a n d turbulent flow of slurry in the vessels. T h e use of d = — \


corresponds to constant ΡIV. A n evaluation of the consequences of unsatis­
factory particle suspension in a scaled-up vessel is helpful in selecting an
appropriate value of the exponent d. Extensive experience will reduce un­
certainty a n d risk in scale-up.

3. Impeller Power in Slurries


T h e power c o n s u m p t i o n of a n impeller in a slurry can be predicted from
the curves of impeller Reynolds n u m b e r versus power n u m b e r reported in
m a n y literature sources; see (B3) a n d ( 0 6 ) . Selection of a n appropriate
viscosity a n d density for the particle suspension is a key factor in successful
power prediction. F o r a low-concentration, free-settling particle suspension
that has a uniform concentration t h r o u g h o u t the slurry v o l u m e , the slurry
viscosity is the same as the suspending-liquid viscosity, a n d the slurry density
can be estimated from the percent solids a n d the densities of the particles a n d
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 37

the liquid. If the particle concentration is high enough to increase the two-
phase system viscosity above the suspending-liquid viscosity, see Section
V , F for m e t h o d s of measuring the two-phase system viscosity.
Impeller power can differ from that calculated by using a n average slurry
density. If the impeller is rotating at such a low speed that particles are n o t
suspended as high as the impeller, the density of the liquid is appropriate for
calculating power. However, if the impeller provides only partial suspension
of the particles, there will be a higher concentration of particles in the slurry
a r o u n d the impeller t h a n the average particle concentration of the whole
slurry in the vessel. T h e n the power will be higher t h a n that calculated from
the average slurry density. These effects are illustrated in the plot of p*/p m

versus (Ν .) in Fig. 2. Impeller power is proportional to the slurry density /?*


Κξ {

or the Fig. 2 ordinate.


If all particles are settled o n the b o t t o m of a vessel a n d the impeller-to- ves­
sel-bottom clearance is significantly different from that which would exist if
the particle concentration were uniform everywhere in the slurry, this
changed clearance m a y have a n effect o n impeller power. F o r a flat-bladed
t u r b i n e with disk-supported blades, reduced clearance reduces impeller
power. However, for a turbine with hub-supported open pitched blades,
reduced clearance increases impeller power (B3).

K . GENERAL COMMENTS ON FREE-SETTLING PARTICLE


SUSPENSION
T h e suspension of solid particles in liquids is a complex operation that is
influenced by m a n y e q u i p m e n t design, p a r t i c l e - l i q u i d system, a n d operat­
ing parameters. Although these parameters have been intensively studied in
small-scale e q u i p m e n t , the results of different investigators d o n o t agree well
in m o s t cases. T h e technical literature has n o t provided a n adequately strong
basis for extrapolating with confidence the results of these experimental
studies in small [often < 460 m m (18 in.) diameter] vessels to the prediction
of particle suspension performance in the large, > 3 m (10 ft) diameter t a n k s
frequently used in the chemical industry.
E q u i p m e n t configurations used o n a n industrial scale are often a c o m p r o ­
mise between higher-cost m o r e complex geometries that are required for
i m p r o v e d particle suspension performance a n d simpler designs that cost
less. T h e cost of impeller power is lower for the former t h a n the latter.
A single impeller with four flat pitched blades is often used for agitation of
a p a r t i c l e - l i q u i d system in a cylindrical baffled t a n k w h e n the m o t o r is
larger t h a n 3 h p (2.2 k W ) . A dished t a n k b o t t o m permits use of a lower
impeller speed a n d achieves lower power cost. Use of impellers with special
c o n t o u r e d pitched blades (see Fig. 5) also lowers power cost.
T h e same P/Vis a conservative basis for scaling u p the performance of
38 J o s e p h Β. Gray and J a m e s Y. Oldshue

e q u i p m e n t for agitation of solid-particle-liquid systems in impeller-stirred


tanks. Extensive scale-up experience m a y permit use of a less conservative
basis in some cases.

IV. Continuous Flow of Free-Settling Particle


Suspensions through an Impeller-Stirred Tank

A. INTRODUCTION
C o n t i n u o u s mixing operations are n o t carried out as frequently as batch
operations involving p a r t i c l e - l i q u i d systems in the chemical process in­
dustry. Preparation of a polymer solution is a n example of a n operation that
is sometimes c o n t i n u o u s . In part, the infrequent use of c o n t i n u o u s slurry
mixing operations m a y be d u e to the difficulties in obtaining uniform parti­
cle concentrations which are described in the following paragraphs.

B . STEADY-STATE PARTICLE CONCENTRATIONS


Consider an agitated t a n k through which a suspension of free-settling
particles in a wide size range flows at steady, equilibrated conditions. Even
t h o u g h the impeller circulates the suspension within the tank, gravitational
force tends to separate the particles a n d liquid. F o r this reason, the particle
size distribution at each elevation in the t a n k will n o t be the same as that
entering the t a n k when the impeller speed is less t h a n that needed for c o m ­
plete concentration uniformity. In spite of this, the concentrations of se­
lected narrow particle size ranges will be the same in the entering a n d leaving
streams as required by material balances at steady-state conditions.
U n d e r the conditions described in the preceding paragraph, h o l d u p times
for each particle size range will be different a n d will not be the same as that
calculated from the slurry inventory in the t a n k divided by the total slurry
feed rate. F o r example, the inventory of the largest particle size range in the
t a n k divided by the flow rate of particles in this size range through the t a n k
will be larger t h a n the inventory divided by the flow rate for all particle sizes if
the impeller speed is n o t high enough to carry the largest particles to the t o p
of the slurry in the tank.
A n example of a particle size distribution in a continuously fed agitated
t a n k with a b o t t o m exit is given in Fig. 11, where the percent solids for each of
four size fractions is shown as a function of zone height above the t a n k
b o t t o m . N o t e that the feed has 30% total solids; therefore, so m u s t the exit
stream. T h e concentration of fraction A, which has the largest particles, falls
to 0% at zone 4, Β falls to 0 at zone 6, a n d C falls to 0 at zone 9. T h e total solids
in each zone drops from 30% at the b o t t o m to 6% in zone 9, where the
particles are all fraction D . If the exit were m o v e d to a higher level such as
1 2 . Agitation of Particulate S o l i d - L i q u i d Mixtures 39

501 1 1 1 1 1 1
r

Z 4 0-
Ο
<

IHEIGH T ABOV E BOTTOM ,ZON ENUMBE R


'DRA W OF F
F I G . 11. Typical particle concentration distribution for complete off-bottom suspension.
Draw-off at the b o t t o m of the tank. Total particles in slurry feed, 30%; fraction A, — 2 0 4- 4 0
mesh, 4%; fraction B, - 4 0 + 8 0 mesh, 12%; fraction C , - 8 0 + 100 mesh, 8%; fraction D , + 1 6 0
mesh, 6%. [From Oldshue ( 0 1 ) . ]

zone 3, all particles except those in fraction D would accumulate until their
concentrations at the new exit m a t c h e d those in the feed.

C . EFFECTS OF OPERATING CHANGES ON PARTICLE


CONCENTRATIONS
If the feed-stream particle size distribution, the impeller rotational speed,
or t h e exit stream location is changed, the exit particle size distribution will
differ temporarily from that in the feed. T h e n a new equilibrium particle size
distribution in the t a n k a n d equal exit a n d feed particle concentrations will
gradually be approached.
A n example of a changed particle concentration distribution for a higher
exit location a n d a higher impeller speed is shown in Fig. 12. N o t e t h a t exit
particle concentrations m a t c h those for the feed; also note t h a t the concen­
trations are higher in the b o t t o m zone t h a n the concentrations for the test in
Fig. 11 at a lower impeller speed. Increasing t h e impeller speed is desirable
before m o v i n g the exit t o a higher level in order to avoid excessive solids
a c c u m u l a t i o n in a dead region at the b o t t o m of the tank.
A fluctuating particle size distribution t h r o u g h o u t the t a n k a n d in the exit
stream m a y occur if the impeller speed is reduced so that it is n o t high enough
t o m o v e the larger particles t o t h e exit. Particles will t h e n a c c u m u l a t e until
t h e average flow rate of large particles leaving the t a n k equals the flow rate of
particles in the same size range entering the tank.
40 J o s e p h Β. Gray and J a m e s Y. Oldshue

50

0 1 2 3 4 5 6 7 8 9 10
HEIGHT ABOVE BOTTOM, ZONE NUMBER
F I G . 12. Typical particle concentration distribution for complete off-bottom suspension.
Draw-off at the third zone above the bottom. Same feed as that in Fig. 11. [From Oldshue ( 0 5 ) . ]

D . PARTICLE CONCENTRATION CHANGES AT A TANK


EXIT
If particles are m o r e dense t h a n the liquid a n d the slurry velocity increases
as the slurry approaches a n exit pipe, then the particle concentration will be
lower in the exit t h a n in the tank. T h e liquid entering the pipe is accelerated
m o r e rapidly t h a n the denser particles by the pressure gradient at the en­
trance of the exit pipe. T h e inverse of this effect will occur if the velocity in
the exit is less t h a n the velocity of liquid a n d particles approaching the exit.
R u s h t o n ( R l ) studied the effect of the ratio of exit velocity u to the Q

velocity u approaching the exit of a vessel when these velocities are in the
{

same direction. D a t a were obtained for 102, 305, 457, 610, a n d 1219 m m
diameter baffled vessels (25.4 m mο 1 in.). A n impeller with six disk-sup­
ported blades a n d ZJD = \ was used. Uniform particle concentrations
i

t h r o u g h o u t the 305 m m diameter vessel were claimed for 450 r p m . T h e


liquid suspension height was the same as the vessel diameter. A blade-to-
b o t t o m clearance Z /D C= $ was used. T h e feed stream entered at the center
T

of the vessel b o t t o m , a n d the exit was at the side of the vessel at the midplane
of the impeller.
F o r sand a n d glass beads ranging in size from 100 to 250 μτη a n d for 1 to
20% concentration by volume, the following equations were found to corre­
late the effect of uju o n the particle concentration
{ in the exit a n d c in the si

liquid approaching the entrance of the exit port for the slurry stream leaving
the vessel:
At uju x < 1, cjc n = K {u lu )-1-0.14
X6 0 x
(16)
k-0.087 (17)
At uju x > 1,
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 41

T h e value of K is 1 for a n exit which is sharp-edged a n d extends Z> /40


l6 T

radially inward from the vessel wall. T h e value of K is 1.25 for an exit which
l6

is flush with the vessel wall.


If uju = ^ a n d K = 1, the particle concentration in the outlet is pre-
{ 16

dicted by Eq. (16) to be 1.2 times t h a t in the vessel. If uju = 3 a n d K = 1,


x l6

the exit concentration calculated by Eq. (17) is 0.91 times that in the vessel.
Stevens a n d Davitt (S7) showed that particle migration toward the higher-
velocity liquid approaching a n exit pipe occurred in the velocity gradients
t h a t exist in the liquid approaching the exit pipe of a continuously fed
impeller-stirred t a n k containing particles with the same density as the liquid.
This migration resulted in as m u c h as a 5% higher particle concentration in
the exit t h a n in the tank. In other tests, the exit was perpendicular to the
liquid flow direction in the t a n k near the exit. In these tests the exit particle
concentrations were 2 to 2 5 % lower t h e n the particle concentrations in the
tank.

E. SELECTION OF EQUIPMENT GEOMETRY AND


OPERATING CONDITIONS
T h e m e t h o d s of predicting particle concentrations in a n agitated t a n k that
are described in the preceding paragraphs are incompletely developed. Nev-
ertheless, s o m e guidance can be provided in selecting e q u i p m e n t a n d oper-
ating conditions.
In s o m e operations, particle concentrations in an impeller-stirred t a n k
m u s t be kept within narrow limits. A n example might be the c o n t i n u o u s
preparation of a polymer solution in two steps. In the first step, particulate
solids a n d cold solvent are fed to a n impeller-stirred tank. In the second, a
c o n t i n u o u s slurry stream from the slurry t a n k is fed to a jacketed pipe to heat
t h e slurry a n d dissolve the polymer. In such cases, e q u i p m e n t geometry a n d
operating conditions should be selected t o achieve nearly uniform suspen-
sion of all particle sizes in the t a n k a n d isokinetic withdrawal of slurry from
the tank. T h e n particle concentrations everywhere in the t a n k will b e c o m e
close to those in the feed. Unfortunately, this is difficult to achieve. De-
scribed in the following are some m e t h o d s of approaching this goal of parti-
cle uniformity.
R u s h t o n ( R l ) a n d S h a r m a a n d D a s (S3) carried out tests in cylindrical
baffled t a n k s agitated by vertical-blade turbines. These studies help to define
h o w isokinetic removal of a p a r t i c l e - l i q u i d stream can be accomplished for
such tanks. However, data have n o t been published t h a t can be used for
selection of a n exit pipe location to obtain isokinetic withdrawal for im-
pellers with pitched blades.
R u s h t o n r e c o m m e n d e d using a n exist velocity equal to the impeller dis-
charge stream velocity at the t a n k exit. This velocity, u , is related to impeller
x

diameter, rotational speed, a n d radial distance r f r o m the impeller periphery


42 J o s e p h Β. Gray and J a m e s Y. Oldshue

Table XIV

Values of Constant K ls in Eq. (18)*

DJD T 0.20 0.20 0.25 0.25 0.33 0.33 0.40 0.40


Blades 4 6 4 6 4 6 4 6
Kx% 0.91 1.08 0.93 1.10 0.95 1.13 0.97 1.15

From Rushton ( R l ) . Reproduced by permission of the American


a

Institute of Chemical Engineers.

to the exit by the equation:


u = K ND /r
{ ls
2
(18)

where the constant K is a function of the n u m b e r of impeller blades a n d the


ls

ratio D /Dj is shown by the n u m b e r s in Table X I V .


{

Particle suspension studies carried o u t by Aeschbach a n d B o u r n e (A 1) a n d


by B o u r n e a n d S h a r m a (B6, B7) showed that a propeller a n d draft tube can
aid in obtaining uniform particle concentrations in a baffled, dished-bottom
vessel (see Fig. 9). Adding a n inverted coaxial c o n e o n the vessel b o t t o m also
helped i m p r o v e concentration uniformity. Intermittent withdrawal of parti­
cle suspension helped in obtaining the s a m e particle concentrations in the
withdrawn suspension as in the vessel.
Listed below are m e t h o d s that will help obtain uniform particle suspen­
sion in a n agitated t a n k a n d isokinetic withdrawal for applications in which
the same particle concentration is wanted in the exit stream a n d t h r o u g h o u t
t h e agitated tank. Unfortunately, the published information needed for de­
sign is incomplete a n d experimental work is needed.
(1) Select a n impeller, baffle, a n d t a n k geometry for which the suspen­
sion velocity does not fluctuate m u c h because of large-scale turbulence
u n d e r the conditions needed for uniform suspension a n d for which this
velocity can be predicted. Examples include a flat-bladed turbine in a cylin­
drical baffled t a n k ( R l ) a n d a propeller a n d draft tube as shown in Fig. 9.
(2) Use a n impeller speed for which relatively uniform particle concen­
trations are achieved in the tank.
(3) Select a n exit location at which the suspension velocity a n d direction
can be predicted. Use a n exit diameter for which the effluent velocity equals
the suspension velocity approaching the exit. Orient the exit pipe parallel to
the suspension flow direction. If a vertical-blade turbine is used, orient the
exit radially in the t a n k wall at the same elevation as the impeller m i d p l a n e
a n d m i d w a y between the t a n k baffles. T h e exit should be sharp-edged a n d
extend radially from the t a n k wall a distance equal t o D /40. T h e feed stream T

should enter at the center of the t a n k b o t t o m . If the particle concentrations


1 2 . Agitation of Particulate S o l i d - L i q u i d Mixtures 43

are n o t uniform t h r o u g h o u t the tank, locate the exit in a region in the t a n k at


which the particle size distribution is close to that of the feed stream.
(4) Select an effluent stream d i a m e t e r a n d flow rate for which the ef-
fluent stream velocity equals the suspension velocity u approaching the exit.
{

Use Eq. (18) a n d Table X I V to estimate u . {

F. GENERAL COMMENTS
T h e relationships between particle concentrations a n d the operating vari-
ables that affect t h e m in impeller-stirred t a n k s through which a free-settling
particle slurry flows continuously are n o t well developed in the technical
literature. F u r t h e r experimental work is needed to obtain satisfactory an-
swers to s o m e of the problems t h a t arise in predicting particle concentrations
in such tanks.

V. Batch Mixing of High-Concentration Particle


Suspensions

A. INTRODUCTION
A variety of solid-particle a n d liquid systems are used in industrial opera-
tions at high e n o u g h concentrations that (impeller- or gravity-induced) par-
ticle m o v e m e n t is greatly hindered by the proximity of nearby particles.
U n d e r these conditions the viscosity of the suspension is higher t h a n t h a t of
t h e liquid part of the suspension. If the viscosity of such a suspension in-
creases as shear rate decreases, it is pseudoplastic. If a m i n i m u m stress is
required to start m o v i n g the suspension, it has the Bingham plastic property
of a yield stress.
Examples of suspensions in which particle m o t i o n is hindered by adjacent
particles include (a) the fiber suspensions involved in m a k i n g paper, (b) coal
slurries used for hydraulic transportation in pipes, (c) various mineral slur-
ries such as iron ore, a n d phosphate slurries used in fertilizer manufacture,
a n d (d) sugar suspensions generated in sugar crystallizers. N o n - N e w t o n i a n
slurries with finer particles are involved in m a k i n g paint pigments, clay
products, a n d starch.

B. PROBLEMS CAUSED BY HIGH PARTICLE


CONCENTRATION
At low particle concentrations at which free settling occurs, the impeller
speed is often d e t e r m i n e d by a need to keep particles m o v i n g o n the b o t t o m
of the t a n k or by a need to obtain relatively uniform particle concentrations
t h r o u g h o u t the tank. As particle concentration is increased, interference
between particles develops a n d the rotational speed needed to meet either of
44 J o s e p h Β. Gray and J a m e s Y. Oldshue

these conditions b e c o m e s lower. However, w h e n particle concentration is


increased enough for pseudoplastic or Bingham plastic properties to de­
velop, then to avoid stagnant regions in the tank, the impeller speed m a y
have to be increased above that needed for the free-settling or lightly hin­
dered settling slurries.
If circulation of a particle suspension in a t a n k is stopped, the particles
settle a n d the concentration reaches a n ultimate settled concentration at the
b o t t o m of the tank, which is a function of particle size, shape, a n d density.
Such settled particles m a y be difficult to resuspend when the settled suspen­
sion is highly pseudoplastic or has a high yield stress ( K l , 0 5 , chapter 5).
Injection of a stream of process liquid or gas by a pipe or lance near the
impeller blades can be used to aid resuspension of settled slurries. Higher
impeller t o r q u e is needed during resuspension t h a n for keeping particles
suspended.
If a cornstarch suspension is allowed to settle, the high-concentration
settled layer is dilatant. T h e shear stress to m o v e such a settled slurry in­
creases as the shear rate is increased. A n impeller in such a settled layer can be
m o v e d only very slowly if at all.
Fluid velocity gradients a n d shear rates are higher near a rotating impeller
t h a n in other parts of a tank. W h e n mixing pseudoplastic or Bingham plastic
liquids, the viscosity in the region of low shear rate is higher t h a n the viscosity
near the impeller blades. T h e higher viscosity retards fluid flow. Fluid veloci­
ties n o t only m a y b e c o m e small b u t also will be zero where the particle
suspension shear stress is less t h a n its yield stress. T h e n zero-velocity static
regions are created where the baffles a n d the t a n k cylindrical wall meet, or at
the intersection of the t a n k b o t t o m a n d the cylindrical wall. In some cases,
liquid m a y m o v e only near a rotating impeller a n d shaft a n d be stagnant
everywhere else in the tank.
W h e n e v e r fluid in a n agitated t a n k is slow-moving, heat transfer with the
t a n k wall, coil, or baffle surface will be retarded. If the objective of the
agitation is t o obtain concentration uniformity (pH, pigment color, reactant
concentrations, etc.), this will also be harder to achieve because of the lower
fluid velocities in low-shear-rate regions in a pseudoplastic or yield stress
fluid.

C . EQUIPMENT FOR MIXING HIGH-CONCENTRATION


PARTICLE SUSPENSIONS
A d e q u a t e m e t h o d s of selecting mixing e q u i p m e n t geometries which are
satisfactory for quantitatively described, n o n - N e w t o n i a n rheologies such as
those exhibited by high-concentration particle suspensions are not found in
the published technical literature. T h e ranges of n o n - N e w t o n i a n rheological
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 45

behavior which can be processed satisfactorily by various e q u i p m e n t types


such as propellers, turbines, a n d helical screws have n o t been described in
quantitative t e r m s involving plots of shear stress versus shear rate, equip­
m e n t d i m e n s i o n ratios (such as DJD , Z /D ,
T { a n d Z / Z ) ) , a n d impeller
T c x

speed.
Propellers, pitched-flat-blade impellers, vertical-blade turbines, a n d
toothed-disk impellers are used to circulate pseudoplastic a n d yield stress
fluids in cylindrical vessels. However, the practical consistency limit (viscous
resistance to deformation by shear stress) a n d yield stress limit for these
impeller designs have n o t been defined. Higher rotational speed, increased
DJD (but n o t m o r e t h a n 0.66), and, often, m o r e impellers increase the
T

allowable consistency that can be processed without stagnant fluid regions.


Baffles are n o t needed to avoid a vortex in a cylindrical t a n k if (N \ < 10. Rc

F u r t h e r m o r e , there m a y be stagnant fluid in their vicinity [Lyons (L6, p.


243)].
If stagnant slurry can be tolerated in parts of the t a n k r e m o t e from the
impeller, a lower rotational speed a n d power can be used t h a n those needed
to attain fluid m o t i o n everywhere in the tank.

D . MIXING PERFORMANCE CRITERIA


W h e n the particle concentration in a suspension is high enough that
particle m o t i o n relative to the liquid is hindered by the close proximity of the
particles to each other, t h e n particle settling occurs slowly. Particle suspen­
sion is n o t a p r o b l e m a n d process performance is j u d g e d by other criteria
appropriate to the process objectives, which m a y involve heating, cooling,
dissolution, chemical reaction, ingredient concentration uniformity, t e m ­
perature uniformity, etc.
In s o m e operations, such as those involving paper p u l p suspensions a n d
coal slurries w h e n n o chemical reaction is taking place, as m u c h as 10 to 2 5 %
stagnant slurry can be tolerated. F o r such operations, the fraction of the
solid-particle inventory which is stagnant m a y be a n i m p o r t a n t process
criterion, a n d the impeller rotational speed m u s t be at least high enough to
keep the stagnant fraction below the largest tolerable value.

E . IMPELLER SPEED SELECTION


T h e relationship between impeller speed a n d suspension v o l u m e in m o ­
tion was studied by Wichterle a n d W e i n (W2) for a particle suspension with a
yield stress or a highly pseudoplastic rheology. T h e y used t a n k s with a square
cross section which h a d 180,240, a n d 300 m m sides (305 m mο 12 in.). T h e
impeller diameters were 60, 80, 100, a n d 180 m m . Six disk-supported verti­
cal flat blades were used in s o m e tests a n d four 24, 30, or 45° pitched flat
46 J o s e p h Β. Gray and J a m e s Y. Oldshue

blades in others. Pseudoplastic bentonite a n d lime slurries were used for


which

τ = Κ γ, ι9
η
0J<K <27 l9 Pa-sec", 0.05 < A < 0.07 (19)

T h e following e q u a t i o n was derived by Wichterle a n d Wein:

P/I? = ^ρ(1.8/^ ) (^τ /ρ ) /Α 20


3
0 81
1/2
(20)
where Ρ is the impeller power, L the v o l u m e of the cubical t a n k used, N the 3
P

impeller power n u m b e r (5.0 for t h e turbines used), K a n empirical d i m e n - 20

sionless constant derived for the condition that all of the cubical slurry
v o l u m e is in m o t i o n , τ the shear stress at γ = 0 obtained by straight-line
0

extrapolation of a log-log plot for the slurry used, a n d p the slurry density. sl

If g P/p N Df
C sl is substituted for N in Eq. (20), the following relationship
3
P

c a n be derived by rearranging t h e variables:

{p N Dllg T ){DJLY
sx
2
c Q = (LZ/K20) 2
(21)

F o r a t u r b i n e with six disk-supported blades, K = 0.6 a n d the right-hand 20

side o f E q . (21) is 9.
If complete m o t i o n is obtained everywhere in a small cube-shaped tank,
c o m p l e t e m o t i o n would also be obtained in a large, geometrically similar
t a n k with the s a m e fluid properties, p a n d τ , at the same impeller periph­ sl 0

eral velocity.
If Ρ is eliminated from the two equations P = 2uTqN and
Ρ = N p N Df/g ,
P sl
3
a n d the resulting equation is divided by the v o l u m e of
c

the cube-shaped t a n k V o n o n e side a n d D o n the other, the following


T

e q u a t i o n is obtained:

2n(T /V )
Q T = N / ? s i ( N A ) (A / Lflgc
P
2
(22)

F o r geometric similarity, the same N , a n d the s a m e ND ,T /V is constant. P { Q T

A similar study of m o t i o n induced by a n impeller in highly pseudoplastic


fluids a n d other fluids with a yield stress was m a d e by S o l o m a n et al (S6).
they used a 0.29-m-diameter cylindrical vessel with four baffles a n d a turbine
with six flat disk-supported blades. Impeller diameters were D /2 or D /3. T T

X a n t h a n g u m a n d Carbopol 940 were used, for which the Bingham plastic


characteristics are described by:

r= z +Kr
y 23 (23)
Values of T for X a n t h a n a n d Carbopol were 8.3 a n d 19 Pa, respectively.
y

Values of K were 50 a n d 17 Paisec)", a n d values of η were 0.13 a n d 0.24. (1


23

P aο 10 d y n / c m ) . 2

A s s u m i n g that the shear stress at a radius R of the spherical b o u n d a r y


between m o v i n g a n d stationary fluid was equal t o the yield stress, S o l o m a n et
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 47

al. derived:

T Q = n 2
R \ (24)
Since the v o l u m e V of the sphere is
s 4nR /3, 3

T /V Q s = 3nr /4 y (25)
If Ρ is eliminated from the equations Ρ = 2nT N Q and Ρ = N p N Df/g ,
P sl
3
c

the resulting equation is

2nTQ = N p N D /g
P sl
2
{
5
c
(26)
Eliminating T from Eqs. (24) a n d (26) a n d rearranging the variables yields
Q

the equation below, which is similar to Eq. (21) from Wichterle a n d Wein.

(p^Df/g^m/R) 3
= 2π /Ν
3
Ρ
(27)

T h e same scale-up relations, constant ND a n d r , can be deduced from { Q

Eq. (21) as from Eq. (27). T h e former equation is applicable t o a cube-shaped


t a n k with all fluid in m o t i o n . T h e latter equation is applicable to a spherical
b o u n d a r y between m o v i n g a n d stagnant fluid in a cylindrical baffled tank.
F u r t h e r work is needed to provide data that can be used to predict rotational
speed for complete m o t i o n in unbaffled t a n k s of industrially useful sizes.
Scale-up using the same impeller peripheral velocity is r e c o m m e n d e d in the
interim.
F o r suspensions in which s o m e hindered settling takes place (low pseudo-
plasticity a n d n o yield stress), impeller rotational speed m a y be determined
from blending, heat transfer, or mass transfer requirements. Model tests m a y
be needed t o find a rotational speed which meets all processing require­
ments.

F. IMPELLER POWER PREDICTION


Skelland (S5) a n d Bates et al. (B3) s u m m a r i z e d published m e t h o d s of
predicting power for impellers rotating in pseudoplastic, dilatant, a n d yield
stress fluids. Briefly, the m e t h o d consists of first multiplying the impeller
rotational speed by a constant [10 for turbines with disk-supported blades
to obtain a n average shear rate. Second, the viscosity corresponding to this
average shear rate is read from a plot of viscosity versus shear rate for the
slurry u n d e r consideration. Third, the impeller Reynolds n u m b e r ΰ Νρ /μ 2
81

is calculated. If this Reynolds n u m b e r is greater t h a n 200 for most impeller


types, the viscosity estimation m e t h o d described in the first two steps above
is not needed since viscosity has little influence o n the power n u m b e r . Third,
the power n u m b e r N is read from a power n u m b e r - Reynolds n u m b e r plot
P

for the impeller geometry being considered. F o u r t h , the impeller power is


calculated from Ρ = N p N Df/g .P sl
3
If the impeller's power is k n o w n a n d its
c
48 J o s e p h Β. Gray and J a m e s Y. Oldshue

rotational speed is wanted, a trial a n d error procedure should be used in


which different values of TV are assumed until t h e calculated power equals the
given value of power.
V i s c o s i t y - s h e a r rate plots for slurries m a y require special m e t h o d s for
measuring viscosity if particles a n d liquid t e n d t o separate in concentric
cylinder, capillary, or pipe (A2) viscometers d u e to migration in velocity
gradients, or if they separate d u e t o gravitational force. A viscometer with a
slurry-circulating impeller should be used ( 0 4 ) for such two-phase systems.
Stormer, Brookfield, H a a k e , a n d other t o r q u e - r o t a t i o n a l speed measur­
ing viscometers can be equipped with turbine, paddle, anchor, or other
impellers such as those described by Langer a n d W e r n e r (L2). F o r these
impellers, numerical values of the p r o d u c t (NpXN^ a n d of the constant K y

in the equation γ = Κ Ν have been published (L2). A n example is


γ

(Νρ)(Ν ς\ = 71 a n d K = 10 for a turbine impeller with six disk-supported


Κ y

blades in a cylindrical unbaffled t a n k for which Z /D C T = DJD = 0.33.


T

A shear rate can be calculated for each value of Ν used from the equation
γ = Κ Ν. Viscosities can be calculated from (Ν )(Ν )
γ Ρ Κ& {= g P/pN Df
c
2
= Κ
a n d Ρ = 2nNT for each N, a n d γ calculated from γ = Κ Ν to provide data
Q γ

for the μ-γ plot m e n t i o n e d in the preceding paragraphs.

VI. M a s s Transfer between Liquid and Free-Settling


Solid Particles

Rotating impellers are often used in cylindrical vessels or tanks to p r o m o t e


t h e transfer of dissolved molecules to or from the surfaces of solid particles.
Dissolution, leaching, a n d ion exchange are examples of such operations. In
s o m e cases, a solid particulate catalyst is used to p r o m o t e a chemical reac­
tion. O t h e r examples involve fermentation, wastewater treatment, polymer­
ization, a n d crystallization.
T h e discussion in the following paragraphs is applicable to batch opera­
tions in which the rate of transfer of molecules between the surface of free-
settling particles a n d the bulk of the liquid phase is limited by turbulent
diffusion. T h e major function of the impeller is to provide a liquid flow
pattern which suspends the solid particles. This flow pattern p r o m o t e s uni­
formity of the concentration of the transferred molecular species t h r o u g h o u t
the bulk liquid phase. In s o m e operations, the higher shear stress near the
impeller blades is used t o increase the particle dissolution rate, b u t this aspect
is n o t discussed. Mass transfer in p a r t i c l e - l i q u i d systems that have particle
concentrations high enough to involve hindered settling is also n o t dis­
cussed.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 49

A. EQUIPMENT
T h e types of e q u i p m e n t used for dissolution a n d other mass transfer
operations for low-viscosity slurries containing free-settling particles are
those described in Section III,F a n d J . Pitched-blade impellers are c o m -
m o n l y used in baffled vertical cylindrical tanks.

B . OPERATING AND DESIGN VARIABLES


T h e s o l i d - l i q u i d system parameters a n d the e q u i p m e n t operating a n d
design variables which have a n effect o n s o l i d - l i q u i d mass transfer rates are
the same as those discussed for particle suspension in Section III,G, except
that here molecular diffusivity D is a n additional system parameter.
m

C . PROCESS PERFORMANCE CRITERION


T h e mass transfer coefficient k, defined by

V(dc/dt) = kAAc (28)

is a n often-used mass transfer criterion. In this equation Kis the solid - liquid
m i x t u r e v o l u m e , A is the solid-particle surface area, c is the transferred
molecular species concentration in the bulk of the liquid, a n d Ac is the
concentration difference between the bulk of the liquid a n d the liquid adja-
cent to the particle surface.

D . MASS TRANSFER COEFFICIENT CORRELATIONS


Examples of equations for correlating the system a n d e q u i p m e n t variables
t h a t affect s o l i d - l i q u i d mass transfer coefficients are given in Table X V .
Applicability limits are also s u m m a r i z e d in this table. T h e ranges of values of
the e q u i p m e n t a n d system used to develop each of these correlating equa-
tions are given in Tables X V I a n d X V I I . In m o s t cases a relatively narrow
range of each p a r a m e t e r was investigated.
T h e effects of each p a r a m e t e r or variable o n the mass transfer coefficient
are s u m m a r i z e d in Table XVIII for the correlating relationships given in
Table X V . N o t e that decreasing particle size, increasing A/?//? , increasing L

molecular diffusivity, decreasing viscosity, a n d increasing impeller power


increase the mass transfer coefficient k. T h e correlating equations in Table
X V are based o n a relatively n a r r o w range of e q u i p m e n t sizes, a n d so k
appears to be i n d e p e n d e n t of e q u i p m e n t size. In most cases impeller speed
does not appear as a correlating variable. However, in all cases the impeller
speed m u s t be larger t h a n JV for the correlating equations in Table X V to be
js

valid.
Table X V

Typical Mass Transfer Correlating Equations

0
Equation Applicability limits* Reference

2 2 2
kD /2D p m = 0.55[(Dl/4D )(eg /vy*Y» m c D (eg y' l4v*
c < 1 Eq. (7.1) in Batchelor (B2)
2 2 2
N = D (eg y< /4D vV ^
Pe c m 1
Small rigid spheres
1
kD /D p m = 2 + 0 . 4 4 [ ( Z ) ; / v ) / 2( v / Z ) f 3 8] pl in 0 . 0 1 < % < 5 , 31 <D < P 1950μτη, Eq. (17) in Levins and Glastonbury (L5)
2 2 1/2
!? = (!/! + w + w ) s
0.04 < /) ι;/ν < 800 ρ Eq. (15) in Levins and Glastonbury (L5)
1 3 3 1 23
0 35 Eq. (8) in Levins and Glastonbury (L5)
E
u = 0.93(ν//) )[(€^) / Ζ)ί/ /ν] · (Α/^τ) · ρ

Fig. 9 in Levins and Glastonbury (L4)

50
and u are functions of Δ/> and D
s p

2
kD /D p= 2 + m \A0(D u /vy (v/D y» p s m
Eqs. (1) and (32) in Miller (M2)
D ujv
p > 030(Dlg Aplp^yViy/DJ-w
1 239
u = 0.000644w (607V) -
s t
Eq. (30) in Miller (M2)
2
u = 4gD Ap/(3p C ) p L D
Eq. (26) in Miller (M2)
C =f[(N ) ]
O Re p
This chapter, Fig. 1
2 1 3 5 0 0 8 2 5 7 5 0 5
kDp/D ^ v / = 0.11[D*I(4 Χ 1 0 - ) ] ( ^ Δ / ? Κ ) ° · Ζ ) θ · / ν · Eqs. (13) and (16) in N i e n o w ( N 5 )
D>
p l500//m
*Dp/An =2 Fig. 1 in N i e n o w (N5)
4 3 1 3 3
fc = 1.03 X 1 0 " m / s e c
js
(eDpv ) ' < 0.01 and v/D = 1 0 m N i e n o w and Miles ( N 7 )
4 3 0 27
(kD /D )(v/D )-^
p = 0.154(ei) /v ) -
m m
D = 2 2 3 0 μτη
p Fig. 7 in Sicardi et al (S4)
1100 <D < 6 1 0 0 / / m , A/?//? = 0.27
P L

4 3 3
4 0 < (€Ώ /νψ < 2 Χ 10
ρ
a
D * = particle diameter, m.
b
*inn
1000n μτηο
//m^ η0.039 in
m o in.
Table XVI

0
Typical Values of Experimental Mass Transfer Equipment Parameters

Impeller types Bottom shape DT (mm)* DJD T ZJD { ZJD T Reference

Six disk-supported blades Flat 143, 286 1_ 2 0.2 1 N i e n o w and Miles ( N 7 )


4 4 i-i
Hat 143, 286 1_ 2 0.18 1
Four 45°-pitched blades 4 4 Μ
T w o vertical blades Flat 143, 286 w 0.25 Μ 1
Four vertical flat blades Flat 128, 190, 2 2 8 0.19-0.56 0.16-0.50 1 Sicardi et al (S5)
with no draft tube

51
Eight 45 "-pitched blades
with draft tube
Four vertical flat blades Dished 152, 305, 686 0.667 0.188 0.25 1.5 Miller (M2)
Flat 126, 2 5 0 0.25-0.60 1 0.27 0.86-0.95 Levins and Glastonbury (L5)
Six vertical flat blades; six 8
back-curved blades; six 0.95
45°-pitched blades;
marine propeller
Four disk-supported blades Hat 120 0.50 0.20 i 1 N i e n o w (Ν4)
Six disk-supported blades Flat 140 0.26-0.53 0.20 1_1 1
3 7

a
All vessels were cylindrical with baffles.
b
1000 m m ο 39.37 in.
Table XVII

Typical Values of Experimental, Mass-Transfer System Parameters

6 9

PP v X 10 PL A„X10
3 3 2 3 2
Solid phase Solid shape (μτη) (kg/m ) (g/cm ) Liquid phase (m /sec) (kg/m ) (m /sec) Reference

4
NaCl Spheres 2230 2160 < 3 X 10" Water 1 1000 N i e n o w and Miles
(N7)
4
Benzoic acid Cylinders 1100-1600 1266 <10" Water 1 1000 Sicardi et al. (S4)
Benzoic acid Cylindrical 4000 1266 Water 1 1000 Miller (M2)
pellets
Metal, salts, Near spherical 31-1900 800-8480 Water, 0.6-17 790-1450 0.05-3 Levins and
plastic methanol, Glastonbury (L5)
ethanol, aq.
solutions
K S0
2 4
Spheres 2000 2660 80% saturated N i e n o w (Ν4)
aq. K S 0
2 4

Salts Various 195-9000 1530-2660 Aq. salt solutions 1 1030-1090 0.6-2 N i e n o w (Ν4)

1000 μτη ο 0.039 in


3 3
1000 k g / m ο 62.4 lb/ft
2 2
1 m / s e c o 10.8 ft /sec
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 53

Table XVIII

Values of Exponent s in Proportionality Relating Mass Transfer Coefficient k to Variable X

V An V Di/DT e Reference

-0.33 2
3 -i 0.33 Batchelor (B2)
-0.2 0.62 >-0.7 <0.2 <0.2 Levins and Glastonbury (L5)
— —
>-0.25
-0.17 —
<*
1
4
<!
f
- i
— 1
6
_
— —
Miller (large-particle equations) (M2)
N i e n o w (Ν5)
0.08 — — 2
3
— 1
2
— 0.27 Sicardi et al. (S4)

E. PREDICTION OF MASS TRANSFER COEFFICIENTS


T h e equations in Table X V can be used to predict k if the p a r a m e t e r values
inserted in the equations are within the ranges employed in the tests o n
which the equations were based (see Tables X V I a n d XVII) a n d within the
applicability limits in Table X V . Examples of the results of such calculations
are presented in Table X X for p a r a m e t e r values listed in Table X I X . T h e
calculated values of k range from 5.1 X 10~ to 11 X 10~ for 1500-//m 5 5

particles for the five equations in Table X V . T h e value of k is 11 X 10~ for 5

150-//m particles in the three equations used. T h e agreement between these


calculated results is reasonable within the limits imposed earlier in this
paragraph.
However, predicted values of k from m a n y correlating equations that have
been described in published technical literature d o n o t agree well. T h e c o m ­
parisons given by Boon-Long et al (B5) a n d Yagi et al (Y1) show as m u c h as
a n order of m a g n i t u d e spread of k values. Such a wide spread shows that
further experimental work is needed o n mass transfer between solid particles
a n d a suspending liquid in impeller-stirred vessels. Even though values of k
predicted from s o m e equations agree reasonably well, why predictions from
other equations d o n o t agree is n o t understood. For this reason, m e t h o d s of

Table XIX

Parameter Values U s e d in Sample Calculations of Mass Transfer Coefficient k a

C = 0.6
D g = 1 kg = m / N - s e c
c
2
μ = 1 0 " kg/m-sec
3

A = 0.10 m N= 5.3 rps<> 318 rpm v= 10" m /sec


6 2

D = 10" m/sec
m
9
N = 5.0 (disk turbine)
P /? = L 1000 k g / m
3

D = 1.5 Χ 1 0 " m
p
3
N =W Sc
3
p =
p 2000 kg/m 3

D = 0.30 m
T e = 0.35 W / k g ο 1.78 h p / ( 1 0 0 0 gal.)

Μ m ο 3.28 ft
1 kg/m-sec — 1 cP
1000 k g / m ο 62.4 lb/ft
3 3
54 J o s e p h Β. Gray and J a m e s Y. Oldshue

Table X X

Sample Calculations of Mass Transfer


Coefficient k Using Table X V Equations

kX 1 0 5

(/mi) (m/sec) Equation source

1500 5.1 Batchelor (B2)


150 11 Batchelor (B2)
1500 8.1 Levins and Glastonbury (L5)
150 11 Levins and Glastonbury (L5)
1500 10 Miller (M2)
150 11 Miller (M2)
1500 9.3 Nienow (Ν5)
1500 11 N i e n o w and Miles (Ν7)
1500 5.0 Sicardi et al. (S4)

predicting k are n o t well enough established for engineers to be confident of


such predictions. Probably, a major factor contributing to p o o r prediction is
inadequate definition of the applicability limits of the published correlating
equations. T h e equations might be used inadvertently for system a n d equip­
m e n t p a r a m e t e r values outside the ranges for which the equations were
derived.
In n o n e of the correlating relationships of Table X V does the size of the
e q u i p m e n t appear as a correlating parameter. T h a t these equations are appli­
cable to prediction of k in large stirred t a n k s should n o t be inferred because
they have n o t been demonstrated to be applicable to vessels outside the
narrow size ranges shown in Table X V I .
Yagi et al (Y1) described the effect of increasing particle concentration o n
mass transfer between small particles a n d liquid. T h e y found t h a t increasing
the concentration of 25- to 145-^m-diameter cation-exchange resin particles
or 25-μπι C a ( O H ) particles could decrease the mass transfer coefficient as
2

m u c h as a n order of magnitude.

F . SOLID-LIQUID MASS TRANSFER EQUIPMENT DESIGN


In the following paragraphs several factors are discussed which should be
considered in the design of impeller-stirred baffled cylindrical t a n k s for
batch operations in which mass transfer occurs between a liquid a n d solid
particles freely suspended in that liquid.

L Vessel Baffle, and Impeller Geometry


T h e geometries described in Section III,J, 1 for particle suspension in
liquids are appropriate w h e n mass transfer occurs between the particles a n d
liquids in such systems.
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 55

2. Impeller Rotational Speed


F o r mass transfer between freely suspended solid particles a n d liquid, the
impeller speed should be at least high enough that particles rest on the
b o t t o m of the t a n k n o m o r e t h a n 1 or 2 sec (N > A y . M e t h o d s of estimating
the rotational speed a n d impeller power to satisfy this just suspended crite­
rion are described in Section III,J,2 a n d 3. Mass transfer is impeded if less
t h a n this rotational speed or impeller power is used.

3. Mass Transfer Prediction


Use of N i e n o w ' s equation in Table X V is suggested for estimating the
mass transfer coefficient k at the just-suspended condition with particles
larger t h a n 1500 pm. T h e value of k from this equation should be indepen­
d e n t of the impeller type according to N i e n o w a n d Miles (N7). There is an
u n k n o w n degree of risk involved in using Nienow's equation for large tanks,
because the relation was derived for a small, 120 m m (5 in.) diameter vessel.
F o r particles less t h a n 1500 pm, Levins a n d Glastonbury's equations in
Table X V are suggested as a basis for estimating k.
If m o d e l tests are used as a basis for predicting mass transfer in large,
impeller-stirred baffled cylindrical tanks, scale-up at the same P/V a n d
e q u i p m e n t geometry will probably provide the same mass transfer coeffi­
cient k in the large t a n k as in the small o n e if n o particles remain o n the
b o t t o m m o r e t h a n 1 or 2 sec. Although there is n o published confirmation of
this scale-up relationship, it m a y be inferred from scale-up of the just sus­
p e n d e d criterion.

4. Mass Transfer Calculations


If the slurry volume, the mass transfer coefficient, a n d various other sys­
t e m p a r a m e t e r values are k n o w n , a concentration versus t i m e curve for the
transferred molecular species in the liquid can be calculated by using the
equations:

Vdc = kA(c*-c)dt (29)


A/A 0 = DllDls (30)

4> = ^ Λ (31)
MJM^D\ID\\ (32)
M = p - Vc (33)

where A is the total area of all suspended particles at time t, c the concentra­
tion of transferred molecular species at t i m e c* the concentration of the
transferred molecular species in the liquid next to the particle surface, D the p

particle diameter, k the mass transfer coefficient, M the total mass of all
p
56 J o s e p h Β. Gray and J a m e s Y. Oldshue

suspended particles, S the surface area per unit mass of particles, t the time,
w

a n d Κ t h e slurry v o l u m e . T h e subscript 0 stands for zero t i m e .


If the variables, A, A , D , D^, a n d M are eliminated from Eqs. ( 2 9 ) -
0 p p

(33), we obtain e q u a t i o n

Vdc = k(c* - c)(Mpo - Vcf* M j? 2


dt (34)

A curve of concentration versus t i m e can be calculated if Eq. (34) is


changed to the form

V(c n+l - c ) = k(c* -


n ΟίΛ/ρο - c V) <* Mtf
n
2
SJit^ x - Q (35)

If the values of V, c at t i m e t , c*, M^, k, a n d


n n are inserted, t h e n c at n + i

time t n + lcan be calculated. T h e differences (t — t ) should be very small.


n+l n

T h e s a m e concentration versus t i m e curve d e m o n s t r a t e d in a small vessel


can be calculated for a large vessel if the mass transfer coefficient k, the initial
particle size distribution a n d 5 ^ , the initial undissolved solid-particle con­
centration A/po/F, a n d the initial dissolved solid concentration c are the 0

same in the two vessels.

VII. General Comments on Agitation of Particulate


Solid-Liquid Mixtures

A wide range of flow properties are found in p a r t i c l e - l i q u i d systems.


These flow properties d e p e n d on particle a n d liquid properties, particle size,
a n d particle concentration.
Particle concentration uniformity is relatively easy to obtain w h e n the
free-settling velocities are less t h a n 2.5 m m / s e c (0.5 ft/min) a n d particle
concentration is low enough that free settling is not hindered by adjacent
particles. Particle concentration uniformity is difficult to obtain w h e n the
free-settling velocities are greater t h a n 100 m m / s e c (20 ft/min) a n d particle
concentration is low enough that free settling is not hindered by adjacent
particles.
T h e r e are m o r e t h a n 15 equations in the published literature that correlate
the variables that affect the rotational speed for suspending particles in
liquids using rotating impellers in baffled cylindrical tanks. T h e rotational
speeds iVj predicted by these equations to suspend particles so that n o n e rests
s

o n the vessel b o t t o m m o r e t h a n 1 sec show a wide spread. This spread


probably arises because of differences in e q u i p m e n t geometry (B8), particle
diameters (or settling rates), a n d particle concentrations. In addition, inter­
actions a m o n g operating, e q u i p m e n t , a n d system p a r a m e t e r s occur. W h e n a
p a r a m e t e r is changed, the effect of a n o t h e r p a r a m e t e r o n iV is changed. js

F u r t h e r experimental work is needed to define the relationships between


12. Agitation of Particulate S o l i d - L i q u i d Mixtures 57

impeller rotational speed for satisfactory particle suspension a n d the vari­


ables that affect it.
Of t h e m e t h o d s described for predicting the impeller rotational speed for
particle suspension in large tanks, scale-up of model tests at the same P/V
a n d e q u i p m e n t geometry (including blade thickness) is suggested as a con­
servative m e t h o d . Experience is helpful in reducing the risk involved in such
scale-ups.
W h e n a c o n t i n u o u s flow of free-settling particles occurs into a n d o u t of a n
impeller-stirred tank, the particle size distributions at all locations in the t a n k
except the exit will nearly always be different from that entering the tank.
F u r t h e r experimental work is needed t o i m p r o v e the m e t h o d s of obtaining
exit particle concentrations in order to obtain concentrations are closer to
those in the tank.
F o r high-concentration slurries that are highly pseudoplastic or have a
high yield stress, i m p r o v e d m e t h o d s of avoiding stagnant regions would be
useful. A d e q u a t e m e t h o d s of selecting e q u i p m e n t types for such slurries are
n o t available.
T h e published mass transfer correlating relationships are based o n tests in
relatively small e q u i p m e n t , a n d the applicability limits of some of these
relationships are not well defined. Scale-up of m o d e l tests at the same P/Vis
probably a n adequate basis for predicting impeller power a n d rotational
speed in large tanks. F u r t h e r experimental work is needed in 4 m (1000 gal.) 3

a n d larger t a n k s to define h o w this relationship should be modified, if neces­


sary, to provide increased confidence in predicting mass transfer between
liquid a n d suspended particles in such large tanks.

List of Symbols

β, b, exponents in Eq. (6)


A solid-particle surface area (L ) 2

C D
drag coefficient (dimensionless)
c concentration of transferred molecular species in bulk of liquid phase ( M / L ) 3

c* concentration of transferred molecular species in liquid at particle surface


(M/L )3

particle concentration in slurry [various units in Eq. (6), and in Tables VIII and
XVII (see cited references)]
v o l u m e fraction o f particles in slurry approaching entrance of the exit stream
(dimensionless)
v o l u m e fraction of particles in exit stream slurry (dimensionless)
^sv v o l u m e fraction of particles in slurry (ratio of particle v o l u m e to slurry volume)
D pipe diameter (L)
A impeller diameter (L)
A„ molecular diffusivity ( L / T )
2

A, particle diameter (L)


D* particle diameter (m)
58 J o s e p h Β. Gray and J a m e s Y. Oldshue

D T tank or vessel diameter (L)


F mass fraction of particles in a slurry
m

f Table ΧΙΙΙ correction factor (dimensionless)


w

g gravitational acceleration ( L / T ) 2

g force equation F = Ma/g , dimensional constant, where F is force, Μ is mass, a


c c

is acceleration, and g has the values and dimensions


c

1 kg-m/N-sec 2
32.17 lb-ft/lb -sec f
2

1 slug-ft/lb -sec
f
2
9.81 kg-m/kgf-sec 2

1 gram-cm/dyn-sec 2

Κ constant numerical value of product (NpXN^ when (NRe\ < 10


Κγ constant in equationγ = Κ , Ν
k s o l i d - l i q u i d system interphase mass transfer coefficient (L/T)
/c solid - liquid system interphase mass transfer coefficient at rotational speed iV
js js

(L/T)
L length (L)
M mass of liquid (M)
L

M mass of particles (M)


p

m exponent in Eq. (2) (dimensionless)


Ν impeller rotational speed (1 / T )
N impeller rotational speed (rev/min)
m

N is impeller rotational speed that suspends particles so that none stay on vessel
bottom more than 1 sec (rev/sec)
N power number, = g P/p N*Df
P c (dimensionless)
L

iVp Peclet number, = Z ) e / 4 v


e (dimensionless)
2 1 / 2 3 / 2

N
Re pipe Reynolds number, = Dupjp (dimensionless)
(A^ReJi impeller Reynolds number, = DfNpJp (dimensionless)
( ^ R e )p Table V particle Reynolds number, = D^pjp (dimensionless)
JVsc Schmidt number, = v/D (dimensionless) m

η exponent in Eqs. (19) and (23)


Ρ impeller power (FL/T) (hp in Fig. 8)
Pm motor power in Eqs. (7) and (8) (hp)
q impeller pumping rate ( L / T )
p
3

R radius (L)
r
radial distance in Eq. (18) (L)
£ surface area per unit mass of particles ( L / M )
w
2

s exponent in Table XVIII (dimensionless)


T agitator shaft torque (FT)
Q

t time
u fluid velocity (L/T)
u bulk flow velocity (impeller discharge rate divided by tank cross-sectional area)
h

(L/T)
u design settling velocity of particles in Eqs. (7) and (8), ft/min (w = f u )
d d w t

u effective relative velocity in Table X V ( M / T )


E

u particle velocity approaching exit of tank in Eqs. (16) - (18) (L/T)


x

u particle velocity in tank exit stream in Eqs. (16) and (17) (L/T)
0

w particle-liquid slip velocity due to inertia difference between solid and liquid
s

(L/T)
w particle settling velocity in particle suspension in stirred tank (see Einenkel's
ss

equation in Table VI) (L/T)


«t particle terminal (free) settling velocity (L/T)
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 59

particle settling velocity in slurry (L/T)


U particle settling velocity in turbulent liquid (L/T)
T

V liquid v o l u m e in Eq. (8) and Fig. 8, gal.; slurry volume in Eqs. (28), (29), (33)
(L ) 3

sphere v o l u m e (L ) 3

vT
tank v o l u m e (L ) 3

V vector resultant velocity in Levins and Glastonbury equations in Table X V


(L/T)
Zc
clearance between impeller blades and vessel bottom (L)
z, impeller blade height (L)
z liquid depth (L)
L

z height of sample tube off vessel bottom (L)


s

Ac transferred molecular species concentration difference between bulk of liquid


phase and liquid adjacent to particle surface ( M / L ) 3

€ turbulence intensity ( L F / M T )
7 fluid shear rate (1/T)
ρ fluid viscosity ( M / L T )
V kinematic viscosity, = p/p ( L / T ) L
2

Ρ* slurry density at impeller calculated by Eq. (4) ( M / L )


3

Pm slurry density calculated by Eq. (5) ( M / L )


3

PL fluid density ( M / L ) 3

PP particle density ( M / L ) 3

Psi slurry density ( M / L )


3

a 2
variance of ratio o f local particle concentration divided by average particle
concentration (dimensionless)
τ shear stress ( F / L ) 2

shear stress at τ = 0 obtained by straight-line extrapolation of a τ versus ν plot


for the slurry used
yield stress ( F / L ) 2

References
(Al) Aeschbach, S., and Bourne, J. R., Chem. Eng. J. 4, 2 3 3 (1972).
(A2) Alves, G. E., Boucher, D . F., and Pigford, R. L., Chem. Eng. Prog. 4 8 , 385 (1947).
(Bl) Baldi, G., Conti, C., and Alaria, E., Chem. Eng. Sci. 3 3 , 21 (1978).
(B2) Batchelor, G. K., J. Fluid Mech. 9 8 , Part 3, 6 0 9 (1980).
(B3) Bates, R. L., Fondy, P. L., and Fenic, J. G., "Mixing: Theory and Practice" (V. W. U h l
and J. B. Gray, eds.), Vol. I, Chap. 3. Academic Press, N e w York, 1966.
(B4) Bohnet, M., and Niesmak, G., Ger. Chem. Eng. 3 , 57 (1980).
(B5) Boon-Long, S., Laguerie, C , and Couderc, J. P., Chem. Eng. Sci. 3 3 , 813 (1978).
(B6) Bourne, J. R., and Sharma, R. N . , Chem. Eng. J. 8, 2 4 3 (1974).
(B7) Bourne, J. R., and Sharma, R. N . , Paper B3, First European Conference o n Mixing and
Centrifugal Separation, Cambridge, England. B H R A Fluid Engineering, Cranfield, Bed­
ford, England, September 1974.
(B8) Buurman, C , Resoort, G., and Plaschkes, Α., Paper 5, Fifth European Conference on
Mixing, Wurzburg, Germany, 1 0 - 1 2 June 1985. B H R A Fluid Engineering, Cranfield,
Bedford, England, 1985.
(CI) C h a p m a n , C . M . , N i e n o w , A . W . , C o o k e , M . , a n d M i d d l e t o n , J . C , Chem. Eng. Res. Des.
6 1 , 71 (1983).
60 J o s e p h Β. Gray and J a m e s Y. Oldshue

(C2) Conti, R., Sicardi, S., and Specchia, V., Chem. Eng. J. 22, 247 (1981).
(Dl) Dickey, D . S., "Agitation Insights." Chemineer, Inc., Dayton, Ohio, February 1981.
(El) Einenkel, W., Ger. Chem. Eng. 3 , 118 (1980).
(E2) Einenkel, W., and Merseman, Α., Verfahrenstechnik 11(2), 9 0 (1977).
(Gl) Gates, L. E., Morton, J. R., and Fondy, P. L., Chem. Eng. 8 3 , 144 (24 May 1976).
(H1) Herringe, R. Α., Paper D 1 , Third European Conference On Mixing, University of York,
England, 4 - 6 April 1979. B H R A Fluid Engineering, Cranfield, Bedford, England.
(H2) Hicks, R. W., Morton, J. R., and Fenic, J. G., Chem. Eng. 8 3 , 102 (1976).
(Kl) Kipke, K., Ger. Chem. Eng. 6, 2 6 4 (1983).
(K2) Kneule, F., and Weinspach, P. M., Verfahrenstechnik 1(12), 531 (1967).
(K3) Kolar, V., Collect. Czech. Chem. Commun. 26, 6 1 3 (1961).
(K4) Kotzek, R., Liepe, F., Langhans, G., and Weissgarber, H.,Mitt. Inst. Chem. Anal. 9(2), 53
(1969).
(K5) Kriegel, E., and Brauer, H., VDI-Forschungsh., 32(515) (1966).
(LI) Lamade, S., Verfahrenstechnik (Mainz) 11(2), 72 (1977).
(L2) Langer, G., and Werner, U., Ger. Chem. Eng. 4, 226 (1981).
(L3) Lapple, C. E., and Shepherd, C. B., Ind. Eng. Chem. 3 2 , 605 (1940).
(L4) Levins, D . M., and Glastonbury, J. R., Trans. Inst. Chem. Eng. 5 0 , 32 (1972).
(L5) Levins, D . M., and Glastonbury, J. R., Trans. Inst. Chem. Eng. 50, 132 (1972).
(L6) Lyons, E. J., "Mixing: Theory and Practice" (V. W. Uhl and J. B. Gray, eds.), Vol. II,
Chap. 9. Academic Press, N e w York, 1967.
(Ml) Maude, A. D . , and Whitmore, R. L., Br. J. Appl. Phys. 9, 477 (1958).
(M2) Miller, D . M., Ind. Eng. Chem. Process Des. Dev. 10(3), 365 (1971).
(M3) Musil, L., Vlk, J., and Jiroudkova, H., Chem. Eng. Sci. 3 9 , 621 (1984).
(M4) Musil, L., Chem. Eng. Sci. 39, 629 (1984).
(Nl) Narayanan, S., Bhatia, V. K., Guha, D . K., and Rao, Μ. N., Chem. Eng. Sci. 24, 223
(1969).
(N2) Nasr-el-din, H., Shook, C. Α., and Esmail, Μ. N., Can. J. Chem. Eng. 6 2 , 179 (1984).
(N3) Nienow, A. W., Chem. Eng. Sci. 23, 1453 (1968).
(N4) Nienow, A. W., Can. J. Chem. Eng. 15, 248 (1969).
(N5) Nienow, A. W., Chem. Eng. J. 9(2), 153 (1975).
(N6) N i e n o w , A. W., and Bartlett, R., Paper B l , First European Conference o n Mixing and
Centrifugal Separation, Cambridge, England. B H R A Fluid Engineering, Cranfield, Bed­
ford, England, September 1974.
(N7) Nienow, A. W., and Miles, D., Chem. Eng. J. 15, 13 (1978).
(01) Oldshue, J. Y., "Unit Processes in Hydrometallurgy" (Μ. E. Wadsworth and F. T. Davis,
eds.), p. 33. Gordon & Breach, N e w York, 1964.
(02) Oldshue, J. Y., Ind. Eng. Chem. 61(9), 79 (1969).
(03) Oldshue, J. Y., Paper presented at the First Pacific Chemical Engineering Congress,
AIChE and Soc. Chem. Eng. Japan, Kyoto, Japan, 1972.
(04) Oldshue, J. Y., Chem. Eng. Prog. 11, 95 (1981).
(05) Oldshue, J. Y., "Fluid Mixing Technology." McGraw-Hill, N e w York, 1983.
(06) Oldshue, J. Y., and Gray, J. B., "Chemical Engineers' Handbook" (R. H. Perry and D . W.
Green, eds.), Sect. 19. McGraw-Hill, N e w York, 1984.
(Rl) Rushton, J. H., AIChE-Inst. Chem. Eng. Symp. No. 10, p. 1 (1965).
(51) Sakiadis, B. C , "Chemical Engineer's Handbook" (R. H. Perry and D . W. Green, eds.),
6th ed., Sect. 5. McGraw-Hill, N e w York, 1984.
(52) Schwartzberg, H. G., and Treybal, R. E., Ind. Eng. Chem. Fundam. 7, 1 (1968).
(53) Sharma, R. N., and Das, H. C. L., Collect. Czech. Chem. Commun. 4 5 , 3293 (1980).
(54) Sicardi, S., Conti, R., Baldi, G., and Cresta, R., Paper D 2 , Third European Conference on
12. Agitation of Particulate S o l i d - L i q u i d Mixtures 61

Mixing, University of York, England, 4 - 6 April 1979. B H R A Fluid Engineering, Cran­


field, Bedford, England.
(55) Skelland, A. H. P., " N o n - N e w t o n i a n Flow and Heat Transfer," Chap. 9, p. 331. Wiley,
N e w York, 1967.
(56) Soloman, J., Elson, T. P., and N i e n o w , A. W., Chem. Eng. Commun. 1 1 , 143 (1981).
(57) Stevens, J. D . , and Davitt, J. P., Ind. Eng. Chem. Fundam. 12(3), 2 6 3 (1974).
(Wl) Weisman, J., and Efferding, L. E., AIChEJ. 6, 419 (1960).
(W2) Wichterle, K., and Wein, O., Int. Chem. Eng. 2 1 , 116 (1981).
(Yl) Yagi, H., Motouchi, T., and Hikita, H., Ind. Eng. Chem. Process Des. Dev. 2 3 , 145
(1984).
(Zl) Zwietering, Τ. N., Chem. Eng. Sci. 8 ( 3 / 4 ) , 2 4 4 (1958).
CHAPTER 13

Turbulent Radial Mixing


in Pipes
Joseph B. Gray*f
Engineering Department
Ε. I. duPont de Nemours and Company, Inc.
Wilmington, Delaware 19898

I. Introduction

A turbulent flow pattern in a fluid passing through a pipe or tube provides


a mixing action that m a y be used to (a) erase radial concentration or temper­
ature differences, (b) keep solid particles suspended in a liquid or gas, (c) mix
ingredients for a chemical reaction, (d) disperse a liquid in a n o t h e r immisci­
ble liquid, (e) disperse a gas in a liquid, or (f) p r o m o t e heat transfer between a
pipe wall a n d a fluid. In s o m e cases, feed jets are used to inject ingredients
into the pipe. Distributors with multiple holes m a y be used t o subdivide the
feed streams a n d reduce the scale of the initial nonuniformities. Baffles or
vanes m a y be inserted in the pipe t o p r o m o t e turbulence. This chapter is
concerned with mixing to obtain uniform radial concentrations or tempera­
tures.

II. Types of Pipe Mixers


M o s t m e t h o d s for bringing two fluid c o m p o n e n t s , A a n d B, together to
m i x t h e m in a pipe involve injection perpendicular to the mixing pipe axis.
Examples of such m e t h o d s are illustrated by the devices shown in Figs. 1 a n d
2. However, parallel injection of feeds can be used as shown in Fig. 3 a n d
tangential injection as shown in Fig. 4.
F o r large conduits, pipes, ducts, or open channels, multiple injection of
o n e or b o t h of the feed streams m a y be used. Parallel flow of feeds a n d tail
* Retired senior consultant from Ε. I. duPont d e N e m o u r s and Company, Inc.
t Present address: Beechwood Consultants, Inc., Wilmington, Delaware 19810.

63
Mixing: Theory and Practice, Vol. Ill Copyright © 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
64 J o s e p h Β. Gray

IB
FIG. 1 . Side injection tee.

I.
4

FIG. 2. Opposed-flow tee.

f
FIG. 3. Injector mixer.

IB £

FIG. 4. Tangential mixer.


13. Turbulent Radial M i x i n g in Pipes 65

pipe m i x t u r e is illustrated in Figs. 5 a n d 6. Distributing devices for perpen­


dicular multiple injection of o n e feed into the other feed are shown in Figs.
7 - 9 . Such mixers could be considered to be multiple small side-tee mixers
arranged in parallel with a c o m m o n mixing pipe. Multiple parallel mixing
tubes can also be used as shown in Fig. 10. A multiple parallel-layer feed
distributor has been described by Kletenik ( K 3 , C3). This is shown in Fig. 11.
Stream-dividing vanes or guiding surfaces m a y be used to p r o m o t e mixing
in a pipe d o w n s t r e a m of feed injection. T h e Kenics, Ross, Sulzer, a n d Light-
nin® types of such mixers are shown in Figs. 1 2 - 1 5 . These devices divide
a n d redirect the fluid to p r o m o t e mixing radially or perpendicular to the pipe
axis.
Multiple baffles perpendicular to the tail pipe axis are used to p r o m o t e
large-scale turbulence. T w o types are shown in Figs. 16 a n d 17.

Γ^ γ

III
θ
FIG. 5. Multiple parallel feed injectors. [Adapted from Baxendale et al. (B2).]

•t.«
FIG. 6. Multiple parallel feed tubes. [Adapted from Vassilatos and Toor (VI).]

FIG. 7. Branched feed distributor. [Adapted from Simpson (S6).]


66 J o s e p h Β. Gray

IB
t,
>B

FIG. 8. Multiple radial jet distributor. [Adapted from Simpson (S6).]

Γ 3 .
i

IB IB
FIG. 9. Multiple inverse radial jet distributor. [Adapted from A h m e d (A2) and Simpson
(S6).]

_ JL _ II
_ JL_ _ II

Γ 3

S E C T I ON Y -Y

FIG. 10. Multiple parallel tees and mixing tubes.


1 3 . Turbulent Radial M i x i n g in Pipes 67

FIG. 11. Folded-sheet distributor. [Adapted from Kletenik (K3).]

(a) (b)

(c)

FIG. 12. Kenics Static Mixer®, (a) Right-hand element; (b) left-hand element; (c) Static
Mixer unit. [Courtesy of Kenics Division, Chemineer, Inc., North Andover, Mass.]
68 J o s e p h Β. Gray

FIG. 1 3 . Ross L P D mixer. [Courtesy of Charles Ross & Son Co., Hauppage, N . Y . ]
13. Turbulent Radial M i x i n g in Pipes 69

FIG. 1 4 . K o c h type S M V static mixing elements (licensee of Sulzer Bros.). [Courtesy of K o c h


Engineering Company, N e w York, N . Y . ]

FIG. 1 5 . Lightnin® Inliner™ Mixer. [Courtesy of Mixing Equipment Company, a unit of


General Signal ( M i l ) . ]
70 J o s e p h Β. Gray

FIG. 16. Trapezoidal baffles. [From Hartung and Hiby (H7).]

FIG. 17. Orifice and annulus baffles.

III. Mixer Performance

A. MIXING PERFORMANCE CRITERIA

If two miscible fluid streams, A a n d B, are injected into a pipe, a concen­


tration pattern is set u p d o w n s t r e a m of the injection points. O n a plane close
t o the feed entrances a n d perpendicular to the tail pipe axis, a wide spread in
feed c o m p o n e n t concentrations exists from all A to all B. T h e concentrations
also fluctuate at points in this plane where turbulent mixing of A a n d Β has
started. Farther downstream, where A a n d Β have spread across most of the
pipe, point concentrations in a pipe cross section still fluctuate widely a n d
the average concentrations at each point are different. T h e point concentra­
tion fluctuations decrease a n d the spread of average c o m p o n e n t concentra­
tions in a pipe cross section decreases as the fluid moves along the pipe.
Eventually n o detectable concentration differences can be found.
Mixing performance for this type of operation can be expressed in several
ways.

( 1 ) Number ofpipe diameters downstream of the feed component injec­


tion required for the spread of highest and lowest concentrations to become
less than a desired value at all points in a tailpipe cross section. Standard
deviation σ or some function of σ such as intensity of segregation m a y be
used instead of a concentration spread as a measure of mixture uniformity
(B8, p . 59; H6).
( 2 ) Time for fluid to meet the mixing criterion described in item (1)
above. This time is the distance required divided by the average fluid ve­
locity.
( 3 ) Number of pipe diameters or time required for a chemical reaction
between components A and Β to reach a desired degree of completion. This
fraction completion of reaction is the fraction completion of mixing,
1 — (σ /σ ), when the reactant ratio is o n e ( K 2 , Η 1 2 , T 3 ) . T h e subscripts ο
0 {

a n d i are for the mixed a n d u n m i x e d streams, respectively.


13. Turbulent Radial M i x i n g in Pipes 71

C o m m o n l y used uniformity criteria are based o n the standard deviation cr,


which is expressed by
η ll/2

σ= Σ (x -x) /("-D
k
2
(1)

where x ,x ,x ,
l 2 3 . . . ,x are values of variables such as c , c , j c , a n d j c ;
n a b v a vb

χ = Σ£ = 1 x^/w is the average value of x \ c , c are the concentrations ofk a b

c o m p o n e n t s A a n d B; a n d x , x are t h e v o l u m e fractions of c o m p o n e n t s A
va yb

a n d B.
T h e variation coefficient σ/χ has been used as a m e a s u r e of uniformity.
See T a u s c h e r a n d Streiff(Tl), G e r a n d Holley (G3), H i b y ( H I 2 ) , a n d Del-
vigne (D4).
A n o t h e r often used m e a s u r e of uniformity is the intensity of segregation I S

(Dl):
^=^/[x v a (l -xj] (2)
T h i s criterion has been used by H a r t u n g a n d H i b y (H7), H i b y ( H I 2 ) , a n d
Stenquist a n d K a u f m a n (S10). T h e d e n o m i n a t o r in Eq. (2) is the standard
deviation squared w h e n n o mixing has occurred:

(3)
F o r Eqs. (2) a n d (3), t h e influent streams are p u r e A a n d p u r e B. If four
parts of o n e feed stream for which ; c = 1 are t o be mixed with o n e part of a
va

second stream for which x = 0, t h e n x = 0.80 a n d σ\ = 0.80 X


ya v a

0.20 = 0.16 ( D 1 , F 2 ) .
E a u a t i o n s (2) a n d (3) m a v be rewritten as

Ι* = σΙ/σΙ (4)
since t h e n u m e r a t o r in Eq. (2) is σ^. O n t h e basis of Eq. (4), I m a y be S

interpreted as a m e a s u r e of the mixing performed between a mixer entrance


a n d exit (G5). T h e standard deviation σ for partially mixed entering streams
{

m a y be used in Eq. (4).


T h e ratio of effluent a n d influent standard deviations σ /σ , which is called 0 {

t h e relative standard deviation, has been used as a measure of mixer per­


formance by H i b y ( H I 2 ) , Keeler et al ( Κ Ι , K 2 ) , Singh a n d T o o r (S8), a n d
T a u s c h e r a n d Streiff ( T l ) .
If b o t h sides of Eq. (3) are divided by x j , multiplied by o\, a n d the t e r m s
a

rearranged, we obtain

σ /χ
0 νΛ = (σ /σ )[(1
0 ϊ - x^/Xy^ 12
(5)
which relates t h e variation coefficient σ /χ 0 νΛ a n d the relative standard devia­
tion ο Jo..
72 J o s e p h Β. Gray

F r o m Eqs. (4) a n d (5) [see also (G5) a n d ( H I 2 ) ] ,

^o/^va={/s[(l/^va)-l]} 1 / 2
(6)

T h e relative standard deviation σ /σ is a m e a s u r e of h o w m u c h mixing is 0 {

accomplished b u t n o t of the exit stream concentration uniformity (W2). T h e


variation coefficient < 7 / x , however, is a m e a s u r e of the exit stream con­
0 v a

centration uniformity. Hiby ( H I 1) stated that σ /χ = 0.01 might be a n 0 νΛ

a d e q u a t e exit stream uniformity for m a n y purposes. In s o m e cases, this


might be as large as 0.05 according to Streiff (SI2).
T h e intensity of segregation equivalent to a value of a / x = 0.01 depends 0 v a

o n the value o f x . F o r x = 0.5, I = 10~ a n d for x = 0 . 0 1 , I = 1 0 ~ a s


v a v a S
4
v a S
6

calculated from Eq. (6). As x is decreased, 7 m u s t be decreased to obtain the


v a S

s a m e ( J / x , or m o r e mixing is needed ( G 5 , W 2 ) .
0 v a

If c r / x = 0.005 is w a n t e d for a stream leaving a mixer, the value of σ /σ


0 va α {

needed w h e n c o m p o n e n t A at 0.01 m / s e c a n d c o m p o n e n t Β at 0.03 m / s e c3 3

are fed to the mixer can be calculated as

x v a = 0.01/(0.01 + 0.03) = 0.25

ol=^{x -xfl{n-\) k

k=l

= [(1.0 - 0.25) 100 + (0 - 0 . 2 5 ) 3 0 0 ] / ( 4 0 0 - 1) = 0.188


2 2

ο- = 0.43

σ = 0.005x
α va = 0.005 X 0.25 = 0.00125

aja i = 0.00125/0.43 = 0.0029

F o r t h e calculation of σ , the choice of k is arbitrary. Preferably, k should be


{

greater t h a n 100.
If the 0.01 m / s e c stream is 6 5 % c o m p o n e n t A a n d the 0.03 m / s e c stream
3 3

is 2 5 % , t h e calculation of aja^ is similar:

x v a = (0.01 X 0.65 + 0.03 X 0.25)/(0.01 + 0.03) = 0.350

σ\ = [(0.25 - 0.35) 300 2

+ (0.65 - 0 . 3 5 ) 1 0 0 ] / ( 4 0 0 - 2
1) = 0.030

σ = 0.173 ;

σ = 0 . 0 0 5 X 0 . 3 5 = 0.00175
α

σ / σ = 0.00175/0.173 = 0.010
0 ;

F o r the latter example, not as m u c h mixing is needed as for the former.


If t h e flow patterns a n d turbulence characteristics are not influenced by
13. Turbulent Radial M i x i n g in Pipes 73

ingredient concentration, t h e intensity of segregation at the exit of several


mixer types in series is the p r o d u c t of values of I for each mixer used when
s

mixing the s a m e ingredient streams in t h e s a m e flow ratio. Similarly, the


relative standard deviation of several mixers in series can be calculated from
the p r o d u c t of the relative standard deviations of the individual mixers.
In addition to a concentration uniformity criterion such as standard de­
viation, there is a n o t h e r mixing criterion, the scale of the concentration
nonuniformities, or D a n c k w e r t ' s scale of segregation (D1). This criterion is a
m e a s u r e of the distance between high-concentration c l u m p s or regions. If
c o m p o n e n t s A a n d Β are mixed without molecular diffusion, the scale of
inhomogeneity is reduced b u t c o m p o n e n t s A a n d Β r e m a i n separated. M o ­
lecular diffusion is needed t o intermix A a n d Β o n a molecular scale. A
m i x t u r e of A a n d Β which m a y appear to be uniform because of a relatively
large sample size for concentration m e a s u r e m e n t s can be n o n u n i f o r m for
small, near molecular scale samples if molecular diffusion is slow.

B. PERFORMANCE MEASURING METHODS

Several measuring m e t h o d s have been used to obtain data for mixer per­
formance evaluation using t h e criteria previously described. Hiby ( H I 2 ) ,
Hill ( H I 3 ) , a n d R o u g h t o n a n d C h a n c e (R4) give detailed descriptions of
m a n y of these m e t h o d s . Unless the same measuring m e t h o d a n d sample size
are used to evaluate mixing e q u i p m e n t performance, comparisons of the
results of different investigators m a y be meaningless.

(1) Temperature after a c i d - b a s e mixing can be used to calculate the


fraction completion of the reaction, which is the same as the fraction c o m ­
pletion of mixing for a very fast reaction between stoichiometrically equal
reactants. T h e measured t e m p e r a t u r e difference between feed streams of the
s a m e t e m p e r a t u r e a n d the fluid at s o m e location in the mixing pipe is
proportional to the extent of reaction. T h e t e m p e r a t u r e rise expected for
complete reaction can be calculated from the heat of reaction a n d fluid
specific heats. See Hiby ( H I 2 ) a n d O t t i n o ( 0 2 ) .
(2) Temperature w h e n streams at different temperatures are mixed
(Wl).
(3) Electrical conductivity w h e n mixing a tracer electrolyte such as NaCl.
See Keeler ( K l ) a n d Stenquist a n d K a u f m a n (S10).
(4) Thermal conductivity. See Fejer et al. (F3).
(5) Light absorption by visual or photoelectric cell m e a s u r e m e n t s when
mixing a tracer dye ( H 7 , C 2 , Μ 1 0 , D 2 , R5) or fluids of different refractive
index.
(6) S a m e as item (5) b u t using acid, base, a n d a pH-sensitive color indi­
cator (H7).
74 J o s e p h Β. Gray

IV. Turbulent Mixing with N o Flow Modifiers in Pipe

This section is concerned with mixing in pipes or tubes in which a fluid is


m o v i n g in fully developed turbulence with n o disturbances such as those
introduced by feed stream injection or by baffles or elbows.

A. TURBULENCE THEORY

A s u m m a r y of the use of isotropic turbulence theory to predict mixing in a


pipe is presented by Brodkey (B8). Equations are given for calculating the
t i m e constant τ in Eq. (7) from fluid viscosity, fluid density, molecular
diffusivity, pipe length, fluid velocity, a n d the root-mean-square (rms) veloc­
ity fluctuation.
I. = er* (7)
Equation (7) gives a relationship between the measure of concentration
uniformity I a n d t i m e after a tracer is introduced coaxially at the same
S

velocity as the fluid in the center of a pipe.


W h e n the turbulence characteristics of a fluid flowing in a pipe change
radially or axially, prediction of turbulent mixing becomes m o r e difficult.
F o r fluid flow in a pipe, the r m s velocity fluctuation is a function of pipe
radius. T h e t i m e constant τ can be calculated as a function of pipe radius.
Gegner a n d Brodkey (G2) have d o n e this, a n d the comparison of calculated
a n d experimental results is again good. Brodkey (B8, p p . 101 - 1 1 0 ) discusses
the effects of feed injection on turbulence characteristics a n d mixing in a
pipe d o w n s t r e a m of multiple small coaxial feed tubes. M o r e rapid mixing is
observed experimentally a n d also calculated from turbulent velocity p a r a m ­
eters near the feed injection tubes t h a n is found m a n y pipe diameters down­
stream.

B. MATHEMATICAL MODELS OF TURBULENCE

In the b o o k " T u r b u l e n t Mixing in Non-Reactive a n d Reactive Flows,"


edited by M u r t h y ( M l 5 ) , various authors describe mathematical models of
turbulence t h a t can be used for predicting average a n d m e a n square fluctua­
tions of velocity, concentration, a n d t e m p e r a t u r e in m a n y e q u i p m e n t ge­
ometries. Constants a n d functions in the equations used were derived by
experiments or theoretically. Numerical solutions to these equations were
obtained by computers. D . B. Spalding, in his chapter in M u r t h y ' s book,
stated that " t h e turbulence model approach, coupled with m o d e r n c o m p u t ­
ing techniques, is already successful in predicting, with speed a n d economy,
turbulent-flow p h e n o m e n a in m a n y fields of engineering a n d of the natural
e n v i r o n m e n t . " H e also stated that m u c h theoretical a n d experimental work
needs to be d o n e to improve the crude b u t effective turbulence models that
have been used so far.
13. Turbulent Radial Mixing in Pipes 75

Patterson ( P I ) described t h e application of mathematical models t o var-


ious mixing operations. H e said that t h e "application of full turbulence
simulation models t o mixing in pipes, jets, static turbulent mixers, rivers a n d
the a t m o s p h e r e requires m u c h further development, b u t would be possible,
at least in principle."
Beek a n d Miller (B3, B7, B8, p p . 8 5 - 8 7 ) calculated t h e concentration
fluctuation intensity decay for a turbulent fluid in plug pipe flow [see also
Simpson (S7)]. F o r a gas, they found that increasing the n u m b e r of nondis-
turbing feed streams should decrease t h e distance t o obtain a desired inten-
sity of segregation. F o r a liquid, this effect was predicted to be m u c h smaller.
F r o m further calculations, it was predicted that increasing Reynolds n u m b e r
should have n o effect o n t h e distance t o obtain a desired segregation intensity
for gases b u t should significantly decrease t h e distance for liquids (B7, S6).

C. M I X I N G BY T U R B U L E N C E IN A P I P E

Bischoff a n d Levenspiel (B6) described mathematical eddy diffusion


models of tracer dispersion in a turbulent fluid flowing in a pipe a n d experi-
m e n t a l m e t h o d s for finding eddy diffusion coefficients € . Values of e /Du
R R

found by various investigators are in t h e range 0.0018 < e /Du < 0.0022 at K

iV = 1 0 a n d 0.0010 < e /Du < 0.0020 at N = 10 .


Re
4
R Re
5

Davies (D3) derived t h e following equation for calculating radial eddy


diffusion constants for turbulent flow of fluids in pipes from theoretical
relationships:
e /Du
R = 0.005/Nll 25
(8)
F o r Reynolds n u m b e r s of 1 0 a n d 10 , e /Du has values of 0.0016 a n d
4 5
R

0.0012 calculated from Eq. (8), which are close to the values s u m m a r i z e d by
Levenspiel a n d Bischoff (L2). F r o m G e r a n d Holley (G3), a n equation simi-
lar to Eq. (8) can be derived, b u t 0.006 appears instead of 0.005 a n d 0.104
instead of 0.125. These n u m b e r s show good agreement.
J o r d a n (J2) described m e t h o d s of solving t h e partial differential equations
that describe eddy diffusion of a fluid stream for various tracer injection
m e t h o d s a n d stream cross-sectional shapes. Equations were derived for
tracer concentrations at a n y point in a rectangular, square, or circular duct as
a function of distance d o w n s t r e a m from a n instantaneous or continuous
tracer source, which m a y be a point, a ring, or a point moving r a n d o m l y in a
plane perpendicular to t h e m a i n stream flow direction. Fischer et al (F4)
described similar m e t h o d s for obtaining tracer concentrations in a rectangu-
lar open channel. A n alternative finite-difference m e t h o d was described
briefly by Evans (E2) for circular pipes a n d continuous, point or ring, tracer
injection.
T h e following equation from Evans (E2) will be used to derive a relation-
ship between t h e tracer concentration difference at a circular pipe center a n d
76 J o s e p h Β. Gray

t h e pipe wall a n d the distance from tracer injection (continuously at a point


at the pipe center):

(9)

where c is the tracer concentration at a point (L, r); c is the average tracer
concentration for all points in the plane, which is perpendicular to the pipe
axis a n d which passes through the point (L, r); this average concentration is
equal to the tracer flow rate divided by the m a i n stream flow rate including
tracer; J is a zero-order Bessel function [described, e.g., in Chapter VI of
0

Reddick a n d Miller (R1); λ has values for which J (k ) = 0 [see Reddick a n d


η x n

Miller ( R l ) ] ; r is the radial distance of point (L, r) from the pipe axis divided
by R; L the distance d o w n s t r e a m from the tracer injection point; e the radial R

eddy diffusion constant; R the pipe radius at the wall; a n d u the average m a i n
stream fluid velocity.
At the pipe wall a n d axis,

dJ (A„r)/dr = 0
o or -A„7,(A„r) = 0
W h e n r = 1, the values of λ for which Ji(A ) = 0 are used in calculating c/c;
η n

these values are 3.8317, 7.0156, 10.1735, 13.3237, etc. W h e n r = 0,


J\(K ) 0 l Jo(K )
r
=
1-0, a n d the values ofλ to use for / ( ^ )
a n (
r =
those η 0
a r e

for which ^(A,,) = 0, i.e., 3.8317 etc.


Concentration values c^/c a n d c /c at the pipe axis a n d wall, respectively, w

were calculated from Eq. (9) for various values of Le /R u a n d are s u m m a ­ R


2

rized in Table I. T h e difference (c^ — c )/c is a straight-line function of w

Le /R u R for semilog coordinates.


2

Interpolation of values in Table I yields the values of Le /R u in Table II K


2

for the concentration spreads listed.


If the fluid velocity u a n d e are known, the pipe length L/D to obtain a
R

desired concentration spread can be calculated, for example, if

(Cax- cj/c = 0.001


and
e Z>M = 0 . 0 0 1 5 '
R

then
Le /R u
R
2
= 4Le /D u R
2
= 0.62 (from Table II)
and
L/D = 0.62/4 X 0.0015 = 103

1
A typical value from Eq. (8) at W Re = 15,000.
13. Turbulent Radial M i x i n g in Pipes 77

Table I

Values o f — c for Selected Values of


w Le /R u
R
2

Le /R u
R
2
0.1 0.2 0.3 0.4 0.5 0.6

cjc 2.5012 1.3276 1.0753 1.01735 1.00400 1.000921


cjc 0.4517 0.8684 0.9697 0.99311 0.99839 0.999629
(C -
ax Cj/C 2.05 0.459 0.105 0.024 0.0056 0.001292

Table II

Values of Le /R u R
2
for Selected Values o f ( c ^ - cj/c

(c^-cj/c 1.0 0.1 0.01 0.001 0.0001


Le /R u
R
2
0.15 0.30 0.46 0.62 0.77

Table III presents mixing performance data for straight circular pipe into
which fluids are fed in ways t h a t d o n o t disturb the flow pattern. Mixing was
j u d g e d by two m e t h o d s , using (1) the standard deviation of a measured
variable proportional to tracer concentration was used a n d (2) a radial eddy
diffusivity. These two m e t h o d s can be related by using Table I or II. For
example, at a standard deviation of 0.001, the 3σ spread is approximately
3 X 10~ . T h e spread in concentration between the pipe axis a n d the wall
6

(^ax "~ O / c is approximately equal to 3σ. See Hiby ( H I 2 ) . A straight-line


extrapolation of the n u m b e r s in Table II, using semilogarithmic coordinate
paper, yields Le /R u) R = 1 . 0 1 at a ( c ^ — c ) / c spread of 3 X 10~ . For
2
w
6

e /Du = 0 . 0 0 1 , L/D = 250 a n d for e /Du = 0.0025, L/D = 100. T h e ex­


R R

perimentally d e t e r m i n e d pipe diameters L/D in Table III to get a 0.001


standard deviation of concentrations ranged from 45 to 120.
T h e mixing performance d a t a in Table III show t h a t o n e can expect t o
need long distances w h e n fully developed, equilibrated turbulent flow is used
for mixing. If there is a density difference between the two fluids being
mixed, mixing is inhibited a n d lengths for mixing to a desired uniformity are
increased (D4).
Starting with Eq. (9), G e r a n d Holley (G3) derived a relationship between
dimensionless mixing t u b e length L/D a n d the variation coefficient ojc for
concentrations in a plane perpendicular to the mixing tube axis w h e n a
tracer is added at the t u b e axis:

L/D = O . S O A ^ / s w / / ) 1 7 2
log(2.37c/a ) 0

where / i s the actual F a n n i n g friction factor (SI) a n d the s m o o t h wall


F a n n i n g friction factor. This equation predicts values of L/D, which range
Table I I I

Mixing in a Pipe with Parallel and Equal Feed and Mixing Tube Velocities

Pipe
Mixing diameters
Tracer distributor Tracer Main-stream Mixing pipe pipe to mix or Measured Mixing
type fluid fluid diam (cm) uju m N Rc
(e*/Du) variable criterion Reference
3
Coaxial from one 3.2 mm Aq. radioac­ Water 10.56 0.95 5X 10 80* Scintillation <7 /<7ί =ο 0.01 Clayton et al.
4
i.d. tube tive tracer 8X 10 123 count (CI)
3
One 3.2 mm i.d. tube 4.8 Aq. radioac­ Water 10.56 0.95 5X 10 120* Scintillation ( 7 ^ = 0.01 Clayton et al.
4
mm from pipe wall tive tracer 8X 10 145 count (C7)
3
Four 3.2 mm i.d. tubes 4.8 Aq. radioac­ Water 10.56 0.95 5X 10 58* Scintillation ο
σ /σ ί = 0.01 Clayton et al.
4
mm from pipe wall tive tracer 8X 10 64 count (C7)
3
Four 3.2 mm i.d. tubes 33.3 Aq. radioac­ Water 10.56 0.95 5X 10 51* Scintillation ( 7 ^ = 0.01 Clayton et al.
4
mm from pipe axis tive tracer 8X 10 54 count (C7)
4
One 1.59 mm coaxial tube C 0 or H
2 2 Air 8.0 10 (0.0014) Thermal Eddy Flint tffl/.(F6)
5
10 (0.00075) conductiv­ diffusiv­
ity ity
4
One 0.83 mm coaxial tube Aq. KC1 Water 7.62 10 (0.0014) electrical Eddy Flint etal. (F6)
5
10 (0.00075) conductiv­ diffusiv­
ity ity
4
3.18 mm radial width Natural gas Air 15.24 4.4 X 10 (0.0025) Specific Eddy Lynn et al.
annulus gravity diffusiv­ (L3)
ity
4
Single 6.35 or 9.6 mm tube C 0 or H
2 2 Air 15.24 or 3 1.2 X 10 (0.0026) Tracer cone. Eddy Towle and
4
at pipe axis 2.5 X 10 (0.0022) by Orsat diffusiv­ Sherwood
4
> 5 . 7 X 10 (0.0018) ity (T6)
4
Pipe divided by a plane 0.02 Ν H S 02 4 0.02 Ν 1.9 10 45* Indicator 7 = 0.01
S Hartung and
through pipe axis NaOH 68* fluores­ / = 0.001
s Hiby (H7)
cence
intensity
4
Coaxial from one 2.2 mm Aq. dye Water 15.24 10 0.0028 Dye cone. Eddy Evans(E2)
6
i.d. tube 10 0.0013 diffusiv­
ity
5
One 0.64 cm i.d. tube Aq. NaCl Water 15.24 1.0 10 109* Electrical tf /c = 0.05
o Ger and
adjacent to mixing tube 154* conductiv- ο
σ / Γ = 0.01 Holley (G3)

* Additional data in cited reference.


*FromEq. (10).
1 3 . Turbulent Radial M i x i n g in Pipes 79

from equal t o one-fifth of t h e experimental values of L/D for a desired ajc


obtained by E v a n s (E2), Clayton et al. (CI), a n d G e r a n d Holley (G3). This
e q u a t i o n is n o t r e c o m m e n d e d because of its p o o r ability t o predict mixing in
pipes w h e n tracer is a d d e d coaxially at t h e same velocity.
G e r a n d Holley (G3) also developed t h e equation

L/D = 2 0 . 5 7 V i ° ( / s w / / )
R
1/2
log(2.40c/<r )
o (10)

which relates effluent radial uniformity ajc a n d mixing t u b e length L/D for
tracer injection adjacent a n d parallel t o t h e mixing t u b e wall. Experimental
values of L/D are within 10% of t h e values calculated from Eq. (10) for data
from Clayton et al. (C7) a n d G e r a n d Holley. It is r e c o m m e n d e d that Eq. (10)
n o t be used outside the Reynolds n u m b e r range 5 Χ 1 0 to 5 Χ 10 or the 3 5

standard deviation range 0.01 < ajc < 0.05.

V. Mixing in Pipes with Coaxial Feed Injection


and Unequal Feed Velocities

A n e x a m p l e of coaxial feed injection is shown in Fig. 3. W h e n the axial


stream A has a higher velocity t h a n t h e m a i n stream B, t h e higher-velocity
stream entrains fluid. W h e n the entrained fluid is m o r e t h a n stream Β
provides, part of t h e partially mixed stream d o w n s t r e a m of the stream A feed
t u b e flows backward or u p s t r e a m t o m e e t this d e m a n d . This modification of
t h e flow pattern in a pipe mixes streams A a n d Β in a shorter distance t h a n
would equal velocities for streams A a n d B. As might be expected, mixing
occurs m o r e rapidly in the first 3 t o 10 mixing pipe diameters t h a n farther
d o w n s t r e a m , where the flow pattern again approaches undisturbed turbu­
lent flow. Velocity a n d concentration patterns in such mixers were studied
a n d described by Tufts a n d S m o o t (T8).
In addition t o distance from t h e axial e n t r a n c e port, variables that affect
mixing uniformity are the flow rate, specific gravity, viscosity, a n d area for
flow for each of the two streams. N o comprehensive correlation of the effect
of these variables o n effluent concentration uniformity has been published.
Table IV gives examples of mixing performance. Values of pipe length for
mixing range from 250 d o w n t o 3.5. C o m p a r i s o n s of the results of different
investigators are difficult t o m a k e because of the different mixing criteria
used. Mixing d a t a in addition t o those in Table IV are presented by Alpinieri
(A4), Fejer et al. (F3), a n d H e d m a n a n d S m o o t (H8).
Swirling or rotating t h e center (primary) stream has been found to alter the
flow pattern a n d i m p r o v e mixing. This modification of coaxial mixers has
been described by Syred a n d Beer (SI5), Fejer et al. (F3), a n d Merkel (M7).
Table IV

Unequal-Velocity Coaxial Tube Mixing

Mixing tube
Inner tube Outer tube Inner tube Outer tube Mixing tube diameters Measured Mixing
fluid A fluid Β diam (cm) diam (cm) Reynolds no. to mix variable criterion Reference
5
Air Air 2.5-20 25 2-8 = 5 X 10 30-6* Temperature Constant Ahmed (Al)
axial temp.
5
NaCl soln. Water 0.165-1.65 9 1-30 2.7 Χ 10 to 10-150* Dye tracer Dimension­ Delvigne(D4)
3
9 X 10 less
standard
deviation
Aq. H S 0 2 4
Aq. NaOH 0.3-0.7 1.9 3-19 — 10.5-3.5* Fluorescence / = 0.001
s Hartung and
intensity Hiby (H7)
3
0.01 NHC1 Water 0.6 1.6 5.5 2.4 Χ 10 to - 5 for Electrical Fraction of Kramers (K4)
3
8 X 10 F=0.02 conductiv­ final
ity conductiv­
ity,^
3
0.03 Ν HC1 0.03 Ν NaOH 0.6 1.6 5.5 2.4 X 10 to - 7 for* Electrical Fraction of Kramers (K4)
3
8 X 10 F = 0.02 conductiv­ final

80
ity conductiv­
ity, F
3
Aq. acid Base or iodine -0.06 0.1 3 5 Χ 10 to 6 to 12 Temperature Distance to Neitchev et
3
thiosulfate 8.4 Χ 10 or color reach al. (Nl)
adiabatic
temp, or to
lose color
4
C 0 and air
2
Air 1.58 5.25 3.6 5 X 10 249 Specific (Axis tracer Reed and
gravity cone.)/ Narayan
(primary (R2)
stream
tracer
cone.)
5
Air H 2
1.33 13.3 13.5-50 0.46 Χ 10 to Function* of H cone.
2 Equal axial Schulz (S3)
5
1.5 Χ 10 uju b and wall
(16-20) H cones.
2
3
Aq. acid Aq. base 0.18 0.32 2.1 3.5 Χ 10 19* 7 = 0.01
S Singh and
69* 7 = 0.001
S Toor(S8)

* Additional data in reference cited.


*See(Kl,p. 133).
Table V

Multiple Feed Streams Parallel to Tail Pipe Axis with Equal Feed Stream Velocities Greater Than Tail Pipe Velocity

Exit
Tail pipe Feed stream tube
Distributor diam. Reynolds Exit stream Measured Mixing diam.
type Fluids (mm) number Wfeed/Wp Reynolds number variable criterion to mix Reference

3 3 3
14 tubes, Aq. acid and 3.18 2 X 10 4.65 3.5 X 1 0 t o 6.7 Χ 1 0 — / = 0.001
s 3.2* Singh and
0.394 m m base T o o r (S8)
i A
l.Q. 3 3
3

14 tubes, 6.76 2 X 10 4.62 3.5 X 1 0 t o 5 . 4 X 1 0 3.2


0.84 m m
i.d.
3 3 4
188 tubes, 31.75 2 X 10 2.86 5.6 Χ 1 0 to 2 Χ 1 0 — 1.1
1.37 m m
i.d.
4 4
188 tubes, Aq. acid and 28.5 1500-2400 3.3 1.2 X 1 0 a n d 2 X 1 0 T e m p , rise I = 0.0001
s 2.9* M a o and

81
1.14 m m base 2.5 Toor ( M l )
i A
l.Q. 3 3

14 tubes, Aq. acid and 3.18 2000-3100 4.65 3.5 Χ 1 0 to 5 Χ 1 0 T e m p , rise 97% reacted 5 Toor and
0.394 m m base Singh (T5)
i A
l.Q. 3 3

14 tubes, Aq. acid and 6.76 2000-3100 4.65 3.5 Χ 1 0 to 5 Χ 1 0 5


0.84 m m base
i.d.
4
100, 1.32 Aq. acid and 31.75 3700 5.8 1.5 Χ 1 0 Temperature 95% 0.9 Vassilatos
m m i.d. base conversion 2.2 and T o o r
tubes (VI)
3
24 parallel 0.05 Ν HC1 5.6 60-450 2 7 X 10 — Visual color 1.8 Kletenik (K3)
4
0.1 m m and 0.05 Ν 5 X 10 disappear­ 0.14
sheets NaOH ance
separated
by 0.1 m m
(see Fig. 11)

a
Additional data in cited reference.
82 J o s e p h Β. Gray

VI. Mixing of Multiple Feed Streams Parallel


to the Mixing Pipe Axis

Figure 6 shows a m e t h o d for dividing two feed streams into m a n y parts by


using multiple parallel small tubes that terminate in a mixing pipe of m u c h
larger diameter. This mixer geometry provides a mixing pattern that requires
only a few mixing pipe diameters to achieve a high degree of concentration
uniformity. T h e required n u m b e r of mixing pipe diameters shown in Table
V ranges from < 1 to 5. There is little incentive to use such a complex mixer
geometry unless very short times are required a n d the desired concentration
uniformity c a n n o t be obtained by a lower-cost, simpler design.
T h e distance to mix, expressed as the n u m b e r of feed jet diameters of
mixing t u b e length L/D to obtain a desired 7 , correlates with feed jet
} S

Reynolds n u m b e r N . for (L/D )N^


Rc } > 200.

/ = 3.52 Χ
s 10 /[(L/D^NU*]
7 345

T h e use of mixing tube diameter D is n o t appropriate for this type of mixer. If


the n u m b e r of feed jets is increased in proportion to D , the same mixing
2

tube length is needed to obtain a desired value of 7 . S

A n ingenious multiple-parallel-feed mixer was devised by Kletenik (K3).


It consists of folded metal sheets as shown in Fig. 11 a n d produces alternate
thin layers of two feeds at the upstream e n d of a mixing tube. A very short
distance to mix was found.

VII. Tee Mixers

A pipe tee provides a simple m e t h o d of bringing together two fluid streams


for mixing. O n e stream m a y pass straight through the tee a n d the other enter
perpendicularly at o n e side as shown in Fig. 1. O r the two streams m a y enter
coaxially opposed a n d leave perpendicular to the entering direction as shown
in Fig. 2.
Mixing performance data for tee mixers are presented in Tables VI a n d
VII. T h e entering stream a n d mixing t u b e diameters m a y be the same or
different. For all of these design variations, mixing takes place in shorter
distances t h a n it does in a pipe with undisturbed turbulent flow. A side tee
with the same diameters for entering a n d leaving streams mixes in a shorter
distance t h a n opposed-flow streams with the same diameter a n d flow rate.
Relatively short distances to mix are obtained w h e n small high-velocity
opposed coaxial streams enter perpendicular to the mixing tube (see Henzler
in Table VII).
In addition to geometry, the mixing distance to obtain a desired radial
uniformity of concentration or t e m p e r a t u r e depends o n the uniformity cri­
terion used, the ratio of side-stream t o main-stream flow rates a n d velocities,
Table VI

Side-Injection Tee Mixers

Mixing
Straight-through Stream A Mixing tube tube
Side-feed feed diam. and stream Β Mixing tube diam. Measured Mixing
fluid A fluid Β (cm) diam. (cm) uJuB
to mix variable criterion Reference

3 4
Air and Air 0.64-3.8 4.45 1 to 6 4 X 1 0 t o 1.8 Χ 1 0 2-3 T i 0 smoke
2
Visual smoke Chilton and
TiCl 4 cone, (by uniformity Genereaux
vapor eye) (C6)
4
Aq. NaCl Water 0.32 15.24 6 6 X 10 105* Electrical a / c = 0.01
o Ger and
0.158 12 105* conductiv­ Holley
0.079 24 105* ity (G3)
5
3 ~ 6 ° C water 5 0 - 7 0 ° C water 0.48, 1.0 12.5 2.5 Χ 1 0 4-5 Temperature (T -T )/
m p M o z h a r o v et
al ( M l 4 )
= 0.01
4
19%C0 , 2
Air 1.58 5.25 2.7 4.6 Χ 1 0 10 C 0 cone.
2 Approx. R e e d and
81% air calc. from equal C 0 2 Narayan
specific cones, at (R2)
gravity pipe axis
and
periphery
4
0.5 Ν H N Q 3
0.5 Ν N a O H 0.635 0.635 1 10 7 Temperature 97% o f final Swanson
4
0.635 0.635 1 2.1 Χ 1 0 6.4 temp, rise (S14)
4
0.635 0.635 1 4 X 10 6

fl
From Eq. (10).
Table VII

Opposed-Flow Tee Mixers

Mixing
Streams A Mixing Mixing tube
and Β diam. tube diam. tube diam. Measured Mixing
Feed A Feed Β (cm) (cm) w /w
B p
to mix variable criterion Reference

3
H or C 0
2 2 N or air
2 1.0 5.7 16.3 16.3 >1.5X 10 1.5 Thermal χ = 0.95* Henzler ( H 9 )
3
1.0, 1.42 5.7 16.3 8.0 >1.5X 10 2.1 conductiv­
3
1.42 5.7 8.0 8.0 >1.5X 10 2.5 ity
4
Aq. NaCl Water 4.2 4.2 0.5 0.5 1.7 Χ 10 43 Electrical Standard Laimer ( L I )
6
conductiv­ deviation
ity = 0.05
6
19%C0 , 2
Air 1.58,5.25 5.25 2.1 0.82 4.6 Χ 1 0 4 C 0 cone.
2 Equal tracer Reed and
81% air cone, at Narayan
axis and (R2)
pipe
periphery
3
2NHC1 2 Ν NaOH 0.185 0.185 0.5 0.5 5.5 Χ 1 0 20 Temperature 97% of final Roughton
temp, rise and
Milliken
(R5)
4
0.5 Ν H N Q 3 0.5 Ν N a O H 0.635 0.635 0.5 0.5 10 15 Temperature 97% of final Swanson(S14)
4
4 X 10 11 temp, rise

a
See Section X.
b
Based o n concentration divided by feed stream concentration difference.
13. Turbulent Radial M i x i n g in Pipes 85

t h e ratio of specific gravities of the two feed streams, a n d the mixing tube
Reynolds n u m b e r a n d surface roughness.
Edwards et al (Ε 1) studied mixing of h e l i u m a n d nitrogen in a 2.3-cm-di-
a m e t e r t u b e d o w n s t r e a m of a sharp-edged hole in the side of the tube.
Typical Reynolds n u m b e r s were 1 0 for the tube flow a n d 1 0 for the side
3 4

stream. T h e r m s concentration fluctuations divided by average concentra­


tion, ajc, were calculated from helium concentration m e a s u r e m e n t s . Mix­
ing uniformity as j u d g e d by the relative standard deviation ajc was found to
be a function of the t u b e length L/D a n d the m o m e n t u m ratio Μ = pud/ s
2 2

p uD.
m
2
A correlating equation of the form ajc = K(L/D)~~ can be derived
2 l

from the data obtained. Values of the constant Κ are tabulated below for
various values of M.

Μ 0.37-0.46 1.9 18-32


L/D 2-20 2-10 2-10
Κ 1.0 0.42 0.076

This correlating equation shows that the t u b e length to mix to a desired value
of ajc decreases as the m o m e n t u m of the side stream is increased. N o
o p t i m u m m o m e n t u m ratio was observed.
T u c k e r a n d Suh (T7) studied mixing by confined 3.2- to 9.5-mm imping­
ing jets that were b o t h perpendicular to a 2 2 . 2 - m m mixing tube similar to the
a r r a n g e m e n t in Fig. 2. T h e y found t h a t turbulent mixing occurred at feed
stream Reynolds n u m b e r s as low as 150. U n d e r these conditions, the mixing
t u b e flow pattern changed from t u r b u l e n t to l a m i n a r as large-scale eddies
decayed.
F o r n e y a n d K w o n (F8) developed a n equation relating the ratio of side-
stream a n d main-stream velocities to the ratio of side-stream a n d m a i n ­
stream diameters for o p t i m u m mixing. This can be written as

QjQ m = (DJD){a + [a 2
+ 0.6(DJD)Y< } 2
(11)

where a = 0.055, q is the side-stream v o l u m e flow rate, q the entering


s m

m a i n - s t r e a m v o l u m e flow rate, D the side-stream diameter, a n d D the


s

m a i n - s t r e a m a n d entering stream diameter. E q u a t i o n (11) is applicable for


a n y value of DJD < 1.0 a n d for N . > 9000. If N . < 9000, Forney a n d
Re Ke

K w o n suggested multiplying qjq from Eq. (11) by (TV ./9000) .


m Re
1/2

E q u a t i o n (11) approaches qjq = 0.7S(DJD)


m for DJD =1.0
15
and
qJq = Q.\\{DJD)
m for DJD = 0. These expressions c o m p a r e favorably
with the empirical equations

qJq =lA9{DJDy-
m
Table VIII

O p t i m u m Side-Stream Diameter for Various Mixer Geometries

Mixer type Optimum diameter ratio (DJD) Eq. no. Parameter limits Reference

Side tee (Fig. 1) DJD = t.5(qJqJ (12) qjq < 0.0026


m O'Leary and Forney ( O l )
DJD = 6A[(qJq )L/D](L/Df" m (13) qjq < 0.0026
m Eqs. (11) and (13) in ( O l )
2<L/D< 10
M
DJD = OM(qJq f m
(14) qjq > 0.0026
m O'Leary and Forney ( O l )
Side tee (Fig. 1) DJD = i(qJq ) m (15) qjq < 0.002
m G u v e n and Benefield (G6)
DJD = OJ&(qJq f" m (16) 0M2<qJq < 1 m G u v e n and Benefield (G6)
Multiple central jets (Fig. 8) DJD = 9.\(qJq \[DI(D-D )} m i (17) 9jQ < 0.002
m
G u v e n and Benefield (G6)
DJD - n.2(qJqJ [D/(D 2- 2D,)] (18)
Oblique tee with side-stream angle φ (see Fig. 18) DJD = &.5(qJqJ[co^ - 90°)]" (19) qjq <0.0l
m Fitzgerald and Holley (F5)
90° <φ< 150°
L/D<7
2 6
Oblique tee with side-stream angle φ (see Fig. 18) DJD = 0 . 7 8 ( / ) ° " [ c o s ( ^ -
9 s 9 m 90·)]-* (20) 120° < φ < 150° Maruyama et al. (M5)
/i >0.06
i s m

L/D< 1
Tangential side-stream jets DJD = (2.2/n)(qJq ) m (21) / , < 0.058
9 s 9 n Maruyama et al. (M4)
DJD=[(0.67/n)(qJq )]o« m
(22) « / > 0.058
s 9 m
1 3 . Turbulent Radial M i x i n g in Pipes 87

for 0.1 < DJD < 1.0 from M a r u y a m a et al (M2) a n d

qJqm = 0.\25{DJD)

for DJD < 0.02 from G e r a n d Holley (G3).


F o r n e y et al (F7, F 8 , F 9 , O l ) a n d G u v e n a n d Benefield (G6) developed
relationships for estimating the o p t i m u m ratio of side-stream to m a i n ­
stream d i a m e t e r DJD for a desired ratio of side-stream t o main-stream flow
rate qjq m in a side-entering tee of the type shown in Fig. 1. E q u a t i o n s that
relate these d i a m e t e r a n d flow ratios are given in Table VIII. Predictions of
DJD from Eqs. (12) a n d (15) in Table VIII agree within 6%, a n d predictions
from Eqs. (14) a n d (16) agree within 20%.
E q u a t i o n (13) is a modification of Eq. (12), which includes t h e effect of
L/D o n t h e o p t i m u m DJD. It is based o n tests m a d e by O'Leary a n d Forney
( 0 1 ) . T h e s a m e DJD is calculated from Eqs. (12) a n d (13) w h e n L/D = 4.2.
T h e o p t i m u m DJD for t h e oblique side-tee mixer in Fig. 18 is a function of
t h e injection angle φ, which is measured relative t o the main-stream pipe
axis. Fitzgerald a n d Holley (F5) obtained d a t a that can be correlated by Eq.
(19) in Table VIII for qjq m < 0.01 a n d 90 < φ < 150°. F o r φ > 9 0 ° , the
side-entering fluid has a n u p s t r e a m velocity c o m p o n e n t in the main-stream
flow direction. F r o m Eq. (19) o n e can see t h a t the o p t i m u m DJD b e c o m e s
smaller as φ is increased.
M a r u y a m a et al (M5) also studied the effect of the flow rate ratio o n the
o p t i m u m DJD. F o r 120 < φ < 150°, Eq. (20) in Table VIII correlated the
d a t a obtained. O n e should note that Eq. (20) predicts that DJD increases as
φ increases, in contrast t o Eq. (19), which predicts the opposite effect. This
difference m a y be explained by Forney's c o m m e n t (F7) that M a r u y a m a
optimized DJD using concentration m e a s u r e m e n t s close t o the injected side
stream, a n d Fitzgerald a n d Holley (F5) optimized the diameter ratio using
c o n c e n t r a t i o n m e a s u r e m e n t d a t a obtained m u c h farther downstream.
G e r a n d Holley (G3) developed a n e q u a t i o n relating effluent radial uni­
formity σ/c a n d mixing t u b e length L/D for side-tee mixers:

L/D = 26.9A^i°(/sw//) 1 / 2
log(C c/<7)
t
(23)
where / i s t h e F a n n i n g friction factor (SI), the smooth-walled tubing

FIG. 18. Oblique side-stream mixer.


88 J o s e p h Β. Gray

friction factor, a n d C a dimensionless constant which is a function of the


t

m o m e n t u m flux ratio ( f t / / i ) ( i I ) / i A ) ' T h e experimental values of L/D


m m
2

are within 10% of the values calculated by this equation. In their experi­
m e n t s , G e r a n d Holley used the following ranges of variables: 0.0052 <
DJD < 0.021, 0.00022 < qjq < 0.0034, a n d 2 < uju
m < 32. F o r predic­
m

tion of L / D , G e r a n d Holley's equation should not be used outside these


ranges. In addition, values of σ/c < 0.005 should not be used, n o r should
calculated values of L/D < 30. T h e relatively large values of L/D obtained by
G e r a n d Holley are probably d u e to the small flow rates of the side streams
that they used.
G e r a n d Holley also used a numerical m e t h o d of integrating the differen­
tial equations for radial turbulent diffusion of a tracer in pipes, using con­
stants derived empirically. T h e average deviation between experimental a n d
calculated lengths to mix for a specified σ/c was less t h a n 3%; the largest
deviation was 12%.
O'Leary a n d Forney ( O l ) developed a n equation that can be used to
calculate the m a x i m u m tracer concentration c in the pipe cross section atM

any distance downstream from tracer injection in a perpendicular radial side


stream such as Β in Fig. 1:
c /c =\.15{DJD)(L/D)-w
M s (24)
where c is the m a x i m u m tracer concentration at a selected pipe cross
M

section, c the tracer concentration in the entering side stream, D the side-
s s

stream diameter at the pipe wall, D the pipe diameter, L the distance down­
stream from the center of the side-stream entrance, a n d DJD < 0.022.
In one series of tests with DJD = 0.014 a n d c = 0.003 for a m e t h a n e s

tracer in air, the completely mixed tracer concentration c was estimated to be


4.4 X 10~ . Typical values of c a n d c/c calculated from Eq. (24) for se­
6
M M

lected values of L/D are:


L/D 20 40 60 65
c X
M 10 6
10.0 6.3 4.8 4.5
c/c M 0.44 0.70 0.92 0.98
O n the basis of these c/c values, the L/D to mix the m e t h a n e tracer a n d
M

air side stream with the m a i n stream is m u c h greater t h a n 10 since c/c = 1 M

for complete mixing.


M a r u y a m a et al (M2) measured air temperatures at various radial loca­
tions in a pipe at four or five distances from L/D = 2 to L/D = 12 down­
stream from injection of a heated side stream (like Β in Fig. 1). A standard
deviation was calculated for each set of radial temperature data a n d plotted
against L/D. T h e slopes of the standard deviation - L/D curves, after a length
4 < L/D < 12 downstream of stream Β injection, approached the slope of
13. Turbulent Radial M i x i n g in Pipes 89

Table IX

Jet-Swirled Side-Stream Mixers

Mixer geometry Φ* (deg) DJD ujum QjQm L/D

T w o jets (Fig. 19) 45 0.042 11 0.019 5 0.008


30 0.042 15 0.026 5 0.004
0 0.042 11 0.019 5 0.0065
45 0.071 7 0.035 5 0.005
30 0.071 8 0.040 5 0.002
0 0.071 7 0.035 5 0.003
45 0.113 5 0.064 5 0.002
30 0.113 4 0.051 5 0.0009*
0 0.113 4 0.051 5 0.0015
One jet (like Fig. 19) 0 0.042 11 0.019 5 0.03
0 0.042 11 0.019 10 0.0065*
T w o jets (Fig. 19) 0 0.042 11 0.019 2.5 0.03
0 0.042 11 0.019 5 0.0065

a
Extrapolation of Maruyama (M6), Fig. 17.
b
Extrapolation o f Maruyama et al. (M6), Fig. 9.

the s a m e type of plot for equilibrium turbulent mixing in a pipe with n o flow
modifiers.
T h e ratio of standard deviations leaving a n d entering a pipe for which
L/D = 10 is 0 . 9 5 or 0.60 (see Table X I X ) . O n the basis of these results, o n e
10

m a y infer t h a t the increased turbulence a n d mixing d u e to injecting a side


stream die o u t in 4 to 12 pipe diameters for the velocity ratios 2 < uj
u < 1 0 a n d the diameter ratios 0.096 < DJD < 0.255 t h a t were used.
m
2 2

M a r u y a m a et al (M6) studied side-stream a n d main-stream mixing for


the mixer in Fig. 19. O n e or two side streams were used with the angle </> = 0, s

30, a n d 45° a n d with the angle at which the edge of the side stream enters

FIG. 19. Jet-swirled side-stream mixer. [From Maruyama et al. (M6).]

2
From Fig. 7 in (M2).
90 J o s e p h Β. Gray

Table X

Side- Main­
stream stream
Side-stream Main-stream diam. diam. Mixing tube
Side streams fluid A fluid, Β (cm) (cm) diam. (cm)

Six streams (see Fig. 9) Air Air 3.7 20.5 20.5

Hemispherical 0.1 M H C 1 0.1 Μ tris(hydroxy- 0.10 2.8 and 3.2 m m radii


annulus with 4 methyl)aminometh- 0.025 for spheres
radial feeds for A ane
and 4 for Β (see
cited ref.)
15 stream A jets and Acid Base 0.11 — 0.60
15 stream Β jets
(see Fig. 20)
5 stream Β jets and 5 0.02 Ν 0.02 Ν N a O H 0.60 — 1.9
stream A jets H SQ2 4

perpendicular to
mixing tube
2 stream A and 2 0.01 Ν 0.01 Ν base 0.5 — 1.0
stream Β jets acetic
perpendicular to acid
mixing tube
10 stream A and 10 0.01 Ν 0.01 Ν base 0.22 1.0
stream Β jets acetic
perpendicular to acid
mixing tube

a
Standard deviation divided by feed stream temperature difference.

tangential to the main-stream pipe wall. Typical mixing performance results


are presented in Table IX.
T h e ratio oJo is the mixed-stream t e m p e r a t u r e standard deviation at a
x

selected value of pipe length L / D divided by the standard deviation for the
u n m i x e d streams at L / D = 0. This measure of mixer performance is a func­
tion of the side-stream to main-stream flow rate ratio qjq ; the pipe length m

L / D ; a n d the geometric parameters of n u m b e r of jets, angle of injection φ Β

relative to the pipe radius passing through the center of the side stream, a n d
ratio of side-stream t o main-stream diameters DJD. T h e values of DJD in
Table I X are t h e o p t i m u m values for the qjq ratios shown.
m

O n the basis of the data in Table IX, the o p t i m u m angle φ is 30° for φ 5 5

values from 0 to 180° a n d qjq from 0.02 to 0.06. Also, two side streams at
m

φ = 0° mix better t h a n one side stream.


8
13. Turbulent Radial M i x i n g in Pipes 91

Multiple Radial Side-Stream Mixers

Mixing
tube Diameters Measured Mixing
u /u
B p
to m i x variable criterion Reference

3 1 1 0 to 1 0
5 6
2-3 Temp. St. d e v . =
a
A h m e d (A2)
0.02
— At hemi­ Temp. 99% of temp. Berger et al.
3 m/sec 3 m/sec sphere exit rise (B5)

1 1 9 X 10 to4
1.5 Color Chance (C4)
1.5 Χ 1 0 5 disappear­
ance
«1 «1 8 X 10 3
5.2 Indicator 7 = 0.001
S
Hartung and
fluores­ Hiby (H7)
cence
intensity
1 1 340 12 Visual color Indicator Smith (S9)
intensity color
change

1 1 190 12 Visual color Indicator Smith (S9)


intensity color
change

T h e o p t i m u m DJD for a one-jet-swirled mixer in Fig. 19 when


<f) = s i n " ( 1 — DJD) has been studied by M a r u y a m a et al. (M4). Equations
s
l

(21) a n d (22) in Table VIII can be used to calculate DJD w h e n qjq a n d the m

n u m b e r of side streams η are k n o w n .

VIII. Multiple Side Stream Mixers


T w o mixer types in Table X m a y be considered multiple-tee mixers in
parallel (see Figs. 9 , 1 0 , a n d 20). T h e radial side-stream mixer of H a r t u n g a n d
H i b y (H7) has all feed streams injected radially adjacent to a flat closed end
of the mixing tube. In this mixer, the feed streams are perpendicular to the
mixing tube axis. Smith's mixer (S9) is nearly the same as the radial feed
mixer of H a r t u n g a n d Hiby. In such mixers, a zone of turbulence is induced
92 J o s e p h Β. Gray

FIG. 20. Multiple-tee mixer. [Adapted from Chance (C4).]

at the feed end of the mixing tube. T h e diameters to mix range from 1.5 to 12,
depending o n the geometry, m e t h o d of judging mixture uniformity, a n d
mixing tube Reynolds n u m b e r . T h e low Reynolds n u m b e r tests of Smith
(S9) show that feed injection can induce as complete mixing in a normally
l a m i n a r mixing t u b e as d o higher turbulent Reynolds n u m b e r s .
T h e mixers in Table X I are similar to those in Table X b u t have the feed
streams tangentially located to swirl the fluids being mixed. T h e mixing tube
lengths required range from roughly the same to longer.
T h e mixers of Hartridge a n d R o u g h t o n (H3), Milliken (M10), R o u g h t o n
a n d Milliken (R5), a n d C h a n c e (C4) were used in studies of reaction kinetics
over a span of 50 years. Very rapid mixing a n d small size characterize the
mixers these m e n developed.
Three tangentially fed mixers were evaluated by L a i m e r ( L I ) . In o n e type,
two coaxial feed streams are fed radially into a spherical c h a m b e r with a
v o l u m e equal to a length of 150 or 8 exit tube diameters. In a second type,
o n e feed stream enters a sphere tangentially, a n d the other feed is injected
radially into the higher-flow-rate feed at a location approximately four en­
trance tube diameters upstream of the spherical chamber. T h e third type has
a cylindrical c h a m b e r instead of a spherical one. T h e cylinder v o l u m e is
equal to a length of 3, 7, or 16 exit t u b e diameters. Exist stream relative
standard deviations ranged from 0.013 to 0.057 for the spherical mixer a n d
0.013 to 0.020 for the cylindrical mixer.
G u v e n a n d Benefield (G6) described a design procedure for a multiple
radial jet mixer like the one shown in Fig. 8. They specified two rings of holes,
with each ring injecting eight jets radially outward from a coaxially located
side-stream distributor pipe. T h e o p t i m u m ratio of jet diameter to m a i n ­
stream pipe diameter, DJD, for the first a n d second rings of jet streams are
given by Eqs. (17) a n d (18), respectively, in Table VIII. In these equations D x

is the outside diameter of the coaxial distributor pipe. In Eqs. (17) a n d (18) q s

is the flow rate through a single jet a n d q is the entering main-stream flow
m

rate. T h e design procedure can be modified for a single ring of jets.


Table XI

Tangential Side-Stream Mixers

Mixing Mixing Pipe


Feed tube tube diams. Measured Mixing
Distributor type Fluid A Fluid Β diam. (cm) diam. (cm) u /u
A p N Re to mix variable criterion Reference

4
T w o fluid A and two fluid 0.01 Ν 0.01 Ν HC1 0.26 0.628 (1) 2.5 Χ 1 0 «1 Visual color Indicator Hartridge
Β jets in a 0.79 c m NaOH intensity color and
diam. by 0.26 m m cyl. change Roughto
mixing chamber with a (H3)
0.42 c m diam. exit orifice
3
T w o fluid A and two fluid Aq. HC1 NaOH 0.05 0.113 «1.3 3 X 10 1.2 Light 97% of final Milliken
3
Β jets (Fig. 4) 1.1 X 1 0 3.5 absorption temp, rise (M10)
3
T w o fluid A and two fluid 2 Ν HQ 2 Ν NaOH 0.05 0.11 1 1.9 Χ 1 0 2.4 Temp. 97% of final Roughton
Β jets (Fig. 4) temp, rise and
Milliken
(R5)
T w o fluid A and two fluid 0.02 Ν acetic 0.01 Ν 0.25 1.0 1 470 12 Visual color Indicator Smith (S9)
Β jets (Fig. 4) acid base intensity color
change
4
Four fluid A and four fluid 0.5 Ν H N Q 3 0.5 Ν 0.124 0.397 1 3.7 Χ 1 0 0.9 Temp. 97% of final Swanson
4
Β jets o n face of a 120° NaOH (square) 1.8 Χ 1 0 1.0 temp, rise (S14)
cone (Fig. 21)
94 J o s e p h Β. Gray

Table XII

Mixing
Stream A Stream Β tube
Fluid Fluid diam. diam. diam.
Type of mixer A Β (cm) (cm) (cm)

Trapezoidal baffles (see Fig. 0.02 Ν H S 0 2 4 0.02 Ν N a O H 1.9 1.9 1.9


16) (semicircle) (semicircle)
Sulzer, 5 elements + 2 Aq. NaCl Water 0.2 3.5 3.5
diam. of empty pipe (see tracer (coaxial)
Fig. 14)
Sulzer, 2 S M V 6 - 4 5 Acid Base — — 3.63
elements (Fig. 14) with 1
diam. empty pipe
between elements
Sulzer, 2 S M V elements H o t flue gas Cold air (-25) 45 45
(Fig. 14), each \ pipe
diam. long
Kenics, 8 elements + 1 0 . 5 Aq. NaCl Water 0.15 1.6 2.35
diams. of empty pipe (see Tracer (coaxial)
Fig. 12)
Kenics elements (Fig. 12) 0.02 Ν H S 0 2 4 0.02 Ν N a O H 1.9 1.9 1.9
(semicircle) (semicircle)
Single orifice 20.3 c m diam., Hot air Cold air (Two pipe (Two pipe 61
30.5 c m diam. quad­ quad­
rants) rants)
90° elbow with a (IAD) Aq. H S 02 4 Aq. N a O H 1.9 1.9 1.9
radius of curvature (semicircle) (semicircle)
located at 6.3 diam. from
feed entry

a
Additional data in cited reference.
b
Based o n concentration divided by feed stream concentration difference.
c
See also (G5).

IX. Baffle, Orifice, Elbow, Grid, and Screen Mixing


Various devices have been added to pipe to p r o m o t e turbulent mixing by
creating turbulence in the wakes of surfaces that change the direction of flow
or increase the scale of turbulence. Kenics (Fig. 12), Ross (Fig. 13), K o c h
(Sulzer) (Fig. 14), Lightnin® (Fig. 15), a n d other motionless mixers divide
the fluid stream a n d alter the flow pattern. Baffles a n d orifices such as those
in Figs. 16 a n d 17 set u p large eddies in the wakes downstream of the edges
across which fluid moves. Wire screens or grids of tubing initiate increased
smaller-scale local turbulence in the wakes of the wires or tubing.
13. Turbulent Radial M i x i n g in Pipes 95

Baffle, Orifice, and Elbow Mixers

Mixing
tube Diams. Measured mixing
ujup w /w
B p
to mix variable criterion Reference

1 1 8 X 10 3
3.0 Fluorescence / =io-
s
2
Hartung and
8 X 10 3
3.6° intensity / =10"
s
3
Hiby (H7)
«1.3 «0.7 2 X 10 4
5 (elements) Electrical Std. dev.* Laimer (LI)
+ 2 (empty conductivity = 0.014
pipe)
— 1.1 X 1 0 4
Visual color Indicator color Streiff ( S l l )
intensity disappear­
ance

2 X 10 3
2 Temperature <7 M = 0.1
o
Tauscher and
to 1 0 s
Streiff ( T l ) c

«2.5 0.6 4 X 10 3
8 (elements) Electrical Std. dev.* Laimer (LI)
+ 1 0 . 5 (empty conductivity = 0.013
pipe)
1 1 8 X 10 3
5.5 Fluorescence 7 = 10"
S
2
Hartung and
8 X 10 3
7.2 intensity 7 = 10"
S
3
Hiby (H7)
1 1 105
Temperature σ Μ = 0.02
0
Faison et al
4 X 10 4
4i aja = 0.04 i
(Fl)

1 1 « 3 X 10 3
62 Indicator (7 = 0.0001)
S
Hiby ( H l l )
(98 with n o fluorescence
elbow) intensity

H a r t u n g a n d H i b y (H7) showed t h a t 3 to 3.6 mixing t u b e diameters were


needed for the baffled mixer in Fig. 16 to reach a n intensity of segregation I s

of 0.01 to 0.001 (see Table XII). T h e feed streams entered at equal velocities
t h r o u g h adjacent halves of the pipe separated by a plane passing through the
pipe axis. A range of Reynolds n u m b e r s from 4,500 to 17,000 was covered in
t h e tests, b u t n o effect of Reynolds n u m b e r s was found o n a plot of I against s

mixer length.
L a i m e r ( L I ) , Streiff (SI 1, S I 2 ) , a n d Tauscher a n d Streiff ( T l ) evaluated
Sulzer (Koch) mixing elements (Table XII). In Laimer's work, approxi­
mately seven mixing t u b e diameters were used between feed injection a n d
96 J o s e p h Β. Gray

conductivity m e a s u r e m e n t . O n e , two, three, four, or five Sulzer elements


were used. Each element was o n e pipe diameter in axial length. Values of the
relative standard deviation ranged from 0.090 for o n e mixing element fol­
lowed by six e m p t y pipe diameters, to 0.040 for three elements, to 0.014 for
five elements followed by two e m p t y pipe diameters. It is n o t apparent why
L a i m e r needed m o r e Sulzer (Koch) elements t h a n Streiff or Tauscher.
H a r t u n g a n d Hiby (H7) a n d L a i m e r ( L I ) reported tests o n Kenics ele­
m e n t s . Laimer's tests required m o r e elements t h a n H a r t u n g a n d Hiby's tests
(Table XII) for a given degree of mixing.
Hartridge a n d R o u g h t o n (H2) used a slit in a p o r o u s diaphragm followed
by a n elbow a n d a loose wire spiral t o p r o m o t e turbulence in a diluted
bloodstream in a study of the kinetics of carbon m o n o x i d e displacement of
oxygen in hemoglobin. These investigators (H3) also used a n orifice at the
exit of a mixing c h a m b e r with tangential feed streams (see Table XI).
Henzler (H9) used a n orifice d o w n s t r e a m of a coaxial jet a n d a radial jet to
p r o m o t e mixing (see Table XIV).
H i b y ( H l l ) found t h a t the presence of a single 90° elbow at 6.3 pipe
diameters from feed entry reduced the distance to mix from 98 pipe d i a m e ­
ters for straight pipe to 62 pipe diameters (Table XII).
Simpson (S6, S7, p p . 2 8 9 - 2 9 5 ) described a baffle a n d orifice mixer for
pipe in which two types of baffles were used. F o r o n e type, three fractional-
circular holes were cut with equally spaced hole centers at the pipe periphery.
F o r the other type, a flat plate was used with o n e circular hole centered in the
pipe as shown in Fig. 2 1 . T h e axial distance between adjacent baffles was two
pipe diameters. T h r e e or five baffles were used (to obtain adequate mixing,
presumably). T h e two baffle types were used alternately in a sequency of
baffles. N o mixing performance data were given by Simpson.
Faison et al ( F l ) evaluated the mixing performance of circular orifices
a n d orifice target c o m b i n a t i o n s using air in a 61-cm-diameter duct (see
Table XII). Standard deviations of upstream, σ a n d downstream, σ , t e m ­
ί9 0

perature m e a s u r e m e n t s were used to evaluate mixing performance. After


three duct diameters d o w n s t r e a m of a 20.3-cm-diameter circular orifice
centered in the duct, I = 0.001 to 0.0016 for airflow rates between 8.5 a n d
s

40 m / s e c . F o r the 30.5-cm a n d 40.6-cm orifices, σ /σ = 0.15 a n d 0.28 a n d


3
0 {

7 = 0.022 a n d 0.078, respectively.


S

W h e n a single disk (20.3, 30.5, or 40.6 c m in diameter) was placed o n e

FIG. 2 1 . Orifice and target mixer. [Adapted from Simpson (S7, pp. 2 8 9 - 2 9 5 ) . ]
Table XIII

Mixing by Grids and Screens

Mixing
Stream A Stream Β tube Mixing
Fluid Fluid diam. diam. diam. tube Length Measured Mixing
Type of mixer A Β (cm) (cm) (cm) uju p u /u
B p N Re to mix variable criterion Reference
3
Square grid of 18 Aq. Water 18 gauge 10.2 10.2 «1 -1 2.4 X 1 0 L/L
M = 65" Electrical Keeler et
2
gauge tubes in a NaN0 3 tubing for grid con­ 3 Χ 10" al. (Kl,
0.71-cm sq. ductiv­ K2)
array. ~ ID ity
downstream is a
sq. weave wire
screen with 0.40
cm openings and
0.23 cm wire
Two stream A and 0.02 Ν 0.01 Ν 0.5 None 1.0 1 1 254* L/D= 12 Visual Indicator Smith
two stream Β jets acetic alkali color (S9)
perpendicular to acid change
mixing tube (400
hole electron
microscope
screen «0.2 cm
downstream)
10 stream A and 10 0.02 Ν 0.01 Ν 0.22 None 1.0 1 1 84* L/D= 12 Visual Indicator Smith
stream Β jets acetic alkali color (S9)
perpendicular to acid change
mixing tube (400
hole electron
microscope
screen =0.2 mm
downstream)
4 e
Fluid A was Aq. NaCl Water 0.053 5.1 5.1 -1 -1 1.5X10 to L/D = 20 x Electrical 0
<7 Μ = Stenquist
4 2
distributed by 4 5 X 10 L/A = 50* con­ ΙΟ" and
holes in a grid of ductiv­ Kauf­
3
six 0.64-cm- ity 3 Χ 10" man
diam. tubes on (S10)
1.27-cm centers
(see Fig. 22)
a
Additional data in cited reference.
* See Table VIII for no screen.
98 J o s e p h Β. Gray

duct diameter d o w n s t r e a m of a 30.5-cm orifice, the mixing performance of


the orifice was not i m p r o v e d b u t was m a d e worse.
Mixing data were obtained by Keeler et al (Κ 1, K2), using a square grid of
h y p o d e r m i c needles for tracer injection a n d a square-weave wire screen with
4 0 % o p e n area a b o u t OAD downstream. T h e y used the ratio L/L M as a
m e a s u r e of the distance to obtain a desired tracer uniformity; L is the axial
distance from the wire screen, a n d L is the opening dimension between
M

screen wires (Table XIII).


Stenquist a n d K a u f m a n (S10) studied the mixing of a n N a C l tracer in­
jected into water flowing in a 5 1 - m m - d i a m e t e r pipe (see Table XIII). F o r
tracer injection, o n e grid of parallel small tubes was arranged in a plane
perpendicular to the pipe axis, a n d immediately adjacent a second grid of
parallel tubes was placed at right angles to the first grid a n d in a plane
perpendicular to the axis of the circular cross-section pipe (see Fig. 22).
Tracer injection holes were drilled in the d o w n s t r e a m tubes at locations
diametrically opposite to where the upstream tubes touched the d o w n s t r e a m
grid tubes. Electrical conductivity readings were m a d e by using a probe at the
pipe axis a n d various distances d o w n s t r e a m from tracer injection.
Y a m a m o t o et al ( Y l ) , Miyairi et al ( M l 2 ) , a n d Sato et al (S2) describe
the results of tests o n tracer mixing d o w n s t r e a m of turbulence-generating
grids.
F o r grid or screen mixers, the ratio L/L is m o r e appropriate t h a n L/D for
M

correlating mixing data. If the same grid tube diameter, t u b e spacing, a n d


tracer hole diameter a n d spacing are used, increasing the mixing tube d i a m e ­
ter will n o t change the distance L to reach a desired concentration uniform­
ity. Changing the grid tube diameter without changing D, however, will
change the distance L to mix.
Smith (S9) used a screen containing square holes with L = 0.45 m m M

approximately 0.2 m m d o w n s t r e a m of multiple radial feed streams to pro­


m o t e further mixing. In all his tests the total flow rate of the feed stream was

FIG. 22. Hollow-tube feed distributor. [Adapted from Stenquist and Kaufman (S10).]
1 3 . Turbulent Radial M i x i n g in Pipes 99

adjusted to obtain disappearance of a color indicator at L/D = 12. H e found


t h a t lower acid a n d base flow rates could be used w h e n the screen was
present. F o r comparison, see Smith in Table X , where the 7V values are Re

higher t h a n those in Smith's tests in Table XIII for 4 a n d for 20 jets. M o s k o -


witz a n d B o w m a n ( M l 3 ) used a n electron microscope grid to p r o m o t e
mixing.

X. Miscellaneous Types of Pipe Mixers

Included in this section are a variety of mixers studied by Henzler (H9),


which use c o m b i n a t i o n s of feed streams parallel a n d perpendicular to the
mixing pipe (see Figs. 23 a n d 24). In s o m e cases, one or two streams are
tangential to the pipe. In all cases, a 5.7-cm-diameter mixing tube was used.
Either air a n d carbon dioxide or hydrogen a n d nitrogen were mixed. Ther-
m a l conductivity m e a s u r e m e n t s were m a d e to determine the time-averaged
composition of gas flowing past a sampling t u b e of 1 m m cross section,2

which could be m o v e d across the mixing t u b e at each of several distances L


from feed stream injection.
T h e r m a l conductivity m e a s u r e m e n t s were used to calculate (a) the vol-

OR
FIG. 2 3 . Coaxial and radial and coaxial tangential jet mixers. [From Henzler (H9).]

c b d

0.7 D

FIG. 24. Parallel-feed radial and tangential jet mixers. [1


[From Henzler (H9).]
100 J o s e p h Β. Gray

u m e fraction JC of tracer c o m p o n e n t ( C 0 or H ) in a v o l u m e element, (b)


v 2 2

the average value x of x after complete mixing, a n d (c) the largest v o l u m e


v y

fraction difference Δ χ observed in a n axial distance L. T h e uniformity


ν

criterion used was


χ=1-Αχ /[χ (\-χ ψ ν ν ν
2

χ was calculated for each of several values of L a n d the ratio (L/D) found 095

for which χ = 0.95. T h e value of 7 equivalent to # = 0.95 is 0.0003 for S

Ax = 3 σ . Values of L/D for other values of χ can be calculated from


y 0
3

(L/D)/(L/D) . 0 95 = 1 + 0.68 log[0.05/(l -χ)]


Tables X I V - X V I list dimensions, conditions of operation, a n d mixing
t u b e lengths L/D to obtain a value of χ = 0.95. Figures 23 a n d 24 show the
locations of the feed tubes a, b , c, d, a n d e for which the lengths L , Z^, L , a c

etc. are given in the tables as a multiple of the pipe diameter D . T h e feed tube
diameters d , d etc. are also listed in the tables as multiples of the mixing
a b9

pipe diameter. In addition, /? , fo, etc. are the densities of feed streams a, b ,
a

etc.; # , q , etc. are the v o l u m e flow rates of streams a, b , etc., a n d N is


a b Rc

shown for the mixing pipe. T h e value in parentheses is the probable lower
limit of N for which the L/D values are valid.
Re

Design Μ in Table X I V is a mixer that has stream a entering the mixing


t u b e coaxially for a n axial distance L = 1 2 5 D = 7.13 c m . Stream c (shown
a

in Fig. 23) enters radially at L = 0.65D = 3.71 c m from the beginning of the
c

mixing tube. Streams b, d, a n d e shown in Fig. 23 were n o t used in the design


Μ mixer.
Various c o m b i n a t i o n s of feed tube diameters a n d feed stream densities
were used in tests 1 - 8 . C o m p a r i s o n of tests 1, 2, a n d 4 shows that L/D
becomes smaller as the coaxial stream velocity is increased by a smaller
diameter for stream a. C o m p a r i s o n of tests 2 a n d 3 or tests 4 a n d 5 shows that
decreasing the side-stream velocity shortens the distance to mix slightly.
T h e short distance to mix of test 1 m a y be d u e to the high velocity of
stream a entraining m o r e fluid t h a n stream b can provide. This large jet
e n t r a i n m e n t sets u p a backflow of fluid which mixes effectively.
T h e type Q l mixer differs from type Μ by having a shorter coaxial tube
length, which reduces L/D (compare tests 1 a n d 3 5 , 2 a n d 3 6 , 4 a n d 3 8 , 5 a n d
39). T h e results of tests 3 a n d 37 show a reversed trend c o m p a r e d to the
others.
W h e n a n orifice plate is added d o w n s t r e a m of the feed stream tubes as
shown in Fig. 23 at h, a shorter distance to mix is obtained. C o m p a r e tests 39

3
See Hiby's Eq. (9) in ( H I 2 ) .
Table X I V

Coaxial and Radial Jet Mixers'*

Test A>,c,d,e
3 3 3 c
Type Design no. Streams* LJD LJD LOK/D dJD dJD (kg/m ) (kg/m ) QjQc ^ReXlO- L/D

Coaxial + side tee Μ 1 a,c 1.25 0.65 0.078 0.175 1.2 1.84 1 (1-5)* 1.9
Μ 2 a,c 1.25 0.65 — 0.175 0.175 1.2 1.84 1 (1.5) 3.8
Μ 3 a,c 1.25 0.65 — 0.175 0.25 1.2 1.84 1 (1.5) 2.5
Μ 4 a,c 1.25 0.65 — 0.25 0.175 1.2 1.84 1 6 5.9
Μ 5 a,c 1.25 0.65 — 0.25 0.25 1.2 1.84 1 6 4.8
Μ 6 a,c 1.25 0.65 — 0.175 0.175 1.84 1.2 1 (1.5) 2.8
Μ 7 a,c 1.25 0.65 — 0.175 0.175 0.084 1.16 1 4 8
Μ 8 a,c 1.25 0.65 — 0.175 0.175 1.16 0.084 1 (1.5) 2.2
Coaxial + tee Q1 35 a,c 0.5 0.65 — 0.078 0.175 1.2 1.84 1 (1.5) 1.8
Q1 36 a,c 0.5 0.65 — 0.175 0.175 1.2 1.84 1 (1.5) 2.4
Q1 37 a,c 0.5 0.65 — 0.175 0.25 1.2 1.84 1 3.5 3.0
Q1 38 a,c 0.5 0.65 — 0.25 0.175 1.2 1.84 1 5 3.3
Qi 39 a,c 0.5 0.65 — 0.25 0.25 1.2 1.84 1 3 4.0
Coaxial + tee + orifice MB 45 a,c 1.25 1.1 0.8 0.25 0.25 1.2 1.84 1 (1-5) 1.8
Q1B 46 a,c 0.5 0.65 1.3 0.25 0.25 1.2 1.84 1 (1.5) 1.5
a
Derived from Henzler (H9).
* See Fig. 2 3 .
C
D = 5J cm.
d
( ) Probable lower limit for which L/D is constant.
Table X V

0
Coaxial and Tangential Jet Mixers

Test Pa
6 3 3 3 c
Type Design no. Streams LJD LJD LJD dJD d ,JD
d (kg/m ) (kg/m ) Qj(Q + Q )
a t
Λ^ΧΙΟ" L/D

Coaxial + tangential MD1 13 a,d 1.25 1.1 — 0.175 0.175 1.2 1.84 1 (1.5)* 5.1
feed
Coaxial + two MT1 15 a,d,e 1.25 1.1 1.1 0.175 0.175 1.2 1.84 1 (1.5) 2.3
tangential feeds MT1 16 a,d,e 1.25 1.1 1.1 0.175 0.25 1.2 1.84 1 (1.5) 3.3
MT1 17 a,d,e 1.25 1.1 1.1 0.25 0.175 1.2 1.84 1 (1.5) 4.0
MT1 18 a,d,e 1.25 1.1 1.1 0.25 0.25 1.2 1.84 1 (1.5) 5.3
MT1 19 a,d,e 1.25 1.1 1.1 0.175 0.175 0.084 1.16 1 (1.5) 8.1

102
MT1 20 a,d,e 1.25 1.1 1.1 0.175 0.175 1.16 0.084 1 (1.5) 2.2
MT1 21 a,d,e 1.25 1.1 1.1 0.175 0.175 1.2 1.84 4 (1.5) 2.1
MT1 22 a,d,e 1.25 1.1 1.1 0.175 0.25 1.2 1.84 4 (1.5) 3.1
MT1 23 a,d,e 1.25 1.1 1.1 0.25 0.175 1.2 1.84 4 (1.5) 3.4
MT1 24 a,d,e 1.25 1.1 1.1 0.25 0.25 1.2 1.84 4 (1.5) 3.9
MT1 25 a,d,e 1.25 1.1 1.1 0.175 0.175 1.84 1.2 4 (1.5) 2.2
MT1 26 a,d,e 1.25 1.1 1.1 0.175 0.175 1.2 1.84 0.25 (1.5) 7.8
MT1 27 a,d,e 1.25 1.1 1.1 0.175 0.175 1.84 1.2 0.25 (1.5) 5.2

α
Derived from Henzler (H9).
b
See Fig. 24.
c
D = 5.7 c m .
d
( ) Probable lower limit for which L/D is constant.
Table XVI

Parallel Feed, Radial, and Tangential Jet Mixers*

Test A 4a
6 3 3 3
Type Design no. Streams LJD L^/D LJD LJD dJD dJD d ,JD
c (kg/m ) (kg/m ) QjQh Qc + Qd JV X 10"
Re
L/Z>

Five parallel MS 29 a,b 1.25 1.1 — — 0.079 0.175 — 1.2 1.84 1.0 — 4 2.9
feeds + side MS 30 a,b 1.25 1.1 — — 0.079 0.25 — 1.2 1.84 1.0 — 5 3.8

103
tee MS 31 a,b 1.25 1.1 — — 0.079 0.175 — 1.84 1.2 1.0 — 5 2.0
MS 32 a,b 1.25 1.1 — — 0.079 0.25 — 1.84 1.2 1.0 — 5 2.8
Five parallel MST 33 a,c,d 1.25 — 1.1 1.1 0.078 — 0.175 1-2 1.84 — 1.0 5 5.1
feeds + two MST 34 1.25 — 1.1 1.1 0.078 — 0.175 1.84 1.2 1.0 5 3.3
tangential

a
Derived from Henzler (H9).
b
See Fig. 24.
C
D=5J cm.
104 Joseph Β. Gray

a n d 46 in Table X I V . Test 45 shows a smaller L/D t h a n test 8, b u t the


geometries differ slightly since L is n o t the same.
c

T h e mixer types listed in Table X V have tangential a n d coaxial feed tubes


as shown in Fig. 2 3 . C o m p a r i s o n of tests 2 a n d 13 shows that L/D is smaller
for the radial feed tube device. However, the distances of feed tubes c a n d d
from the start of the mixing t u b e are not the same. C o m p a r i s o n of tests 13
a n d 15 shows that two tangential feeds, which are opposed in their direction
of mixing tube fluid rotation, have a smaller L/D t h a n o n e tangential side-
stream feed tube.
Increasing the coaxial feed velocity by decreasing the feed tube diameter
decreases L/D, as shown by tests 16 a n d 18. Tests 17 a n d 18 demonstrate that
increasing the tangential feed velocity by decreasing the tangential t u b e
diameters decreases L/D. W h e n b o t h coaxial a n d tangential tube diameters
were decreased, as in tests 15 a n d 18, the smallest L/D of this group of four
tests at equal coaxial a n d side-stream flow rates was obtained.
If the ratio of coaxial to tangential feed t u b e flow rates is increased with n o
change in mixer dimensions, the values of L/D are decreased (compare tests
15, 2 1 , a n d 26, 16 a n d 22, 17 a n d 2 3 , 18 a n d 24, or 25 a n d 27).
T h e mixer types in Table X V I have multiple parallel feeds rather t h a n the
o n e coaxial feed of the mixers in Tables X I V a n d X V . T h e parallel a n d side
feed tube locations of the mixers in Table X V I are shown in Fig. 24. T h a t five
parallel feeds require a longer distance to mix t h a n a single coaxial feed is
shown by test 15 (Table X V ) a n d test 33 (Table XVI). C o m p a r i s o n of tests 29
a n d 30 shows that a higher velocity for the side stream reduces L/D. Tests 31
a n d 32 show the s a m e effect.

XL Pressure Loss for Pipe Mixers

Pressure d r o p data for c o n t i n u o u s mixers are used as a basis for selecting


the p u m p head a n d power that will m o v e the feed streams through the mixer
a n d its auxiliary piping. Both pressure d r o p a n d mixing performance data
are needed for comparisons of investment a n d operating costs of different
designs for the same flow rates, the same pressure drop, a n d the same effluent
stream radial uniformity.

A. PRESSURE LOSS

Pressure loss in pipe mixers is d u e to several factors: (a) fluid friction losses
in tubes or holes that accelerate the feed streams to their injection velocities,
(b) changes in the direction a n d velocity of the fluid w h e n it enters the mixing
t u b e or pipe, (c) fluid friction at the mixing device surfaces, (d) inefficient
conversion of mixing t u b e kinetic or velocity head to static head, a n d (e)
elevation changes.
13. Turbulent Radial Mixing in Pipes 105

B. MIXER POWER BALANCE

A s s u m e that two fluids, A a n d B , at flow rates vv a n d w a n d velocities w a b a

a n d u pass adiabatically through the mixer shown in Fig. 25. A n energy per
b

u n i t t i m e or power balance for this mixer is:

wJlPiJPu + gZJZc + u\J2g c + J(c )UT p l& - T )] T

+H ^[Ab/Ab + gZlh/gc + «5b/2 + J(c ) (T & p lb lb ~ r)] T

= Κ + w )[A>/A> + g Z / & + u J2g


b 6
2
c + / ( c ) ( r - T )] p 6 6 t (25)
where ρ is pressure, ρ is fluid density, Ζ is elevation, 7 is t h e mechanical
equivalent of heat, c is specific heat, a n d Τ is temperature. T h e subscripts
p

1 - 6 refer t o locations in Fig. 2 5 . T h e symbol T denotes a n arbitrary refer­ T

ence temperature.
There are four types of power in the various parts of Eq. (25). T h e s u m of
the wp/p t e r m s is the power associated with the pressure differences between
the entering a n d leaving streams. T h e s u m of the gZ/g t e r m s is t h e power c

associated with elevation difference between locations 1 a n d 6. T h e s u m of


the wu /2g t e r m s is t h e power associated with differences in velocity be­
2
c

tween locations 1 a n d 6. T h e s u m of the w(Jc T) t e r m s is t h e power asso­ p

ciated with t h e irreversible frictional losses between locations 1 a n d 6. If


Ρ la Ρ lb Pe> Z = Z
= =
= Z , w = u = u,
l a l b (c )i = (c ) = (c ) ,
6 la ih 6 p & p lb p 6

Pi* = Pib, Τ = T , a n d w = w Eq. (15) becomes


ΪΛ l b a b9

{P\-Pe)lp = Jc {T -T )
p 6 l

U n d e r these conditions, the energy used between locations 1 a n d 6 per unit


mass of fluid flowing is (/?, — p )/p or Jc (T — T ). T h e temperature change
6 p 6 x

(T —T )
6 x between 1 a n d 6 for t h e mixer is {p — p )/p(Jc ). T h e average x 6 p

rate of t e m p e r a t u r e change in t h e mixer is {p — p )/pjec , where θ is t h e x 6 p

t i m e for fluids t o pass from location 1 t o location 6 in Fig. 25^ T h e average


power per unit v o l u m e between locations 1 a n d 6 is (p — ρ )/θ. T h e h o l d u p l 6

la 2 a3 a

lb 2 b3 b 45
FIG. 2 5 . Mixer for power balance.
Table XVII

Mixer Pressure Loss

Measure-
ment Fluid Fluid
Mixer type locations A Β " /"p
A
Z>(cm) N /{L/D)
VH Reference
4
Pipe — — 1.27 10 0.036 Sakiadis
7
30.5 10 — 0.013 (Fig. 5-28
in SI)
3
Pipe — 0.02 Ν 0.02 Ν 1 1 1 1 1.9 8 X 10 — 0.033 Hartung
H S0 2 4 NaOH and Hiby
(H7)
3
Opposed-flow tee (Table — H or
2 N or
2 16.3 16.3 1 0.07 5.7 > 1 . 5 Χ 10 1-10 — Henzler
3
VII and Fig. 2) co air 8 8 1 to 5.7 > L 5 X 10 (H9)
2

4
Opposed-flow tee (Table Aq. NaCl Water 0.5 0.5 1 1 4.2 1.7 Χ 10 2 Laimer (LI)
VII and Fig. 2)
4
Multiple tangential feeds Feed tank to Aq. NaOH Aq. HC1 -1 «1 1 1 6.3 2.5 Χ 10 10 — Hartridge
and orifice (Table XI) mixing and

106
tube exit Roughton
(H3)
3
Five stream A jets and — Aq. NaOH Aq. H S 02 4 -1 -1 1 1 1.9 8 X 10 5.9 — Hartung
five stream Β jets and Hiby
perpendicular to a (H7)
mixing tube

5
Single 40.6-cm orifice Less than 24 Hot air Cold air 1 1 1 -1 61 1.1 X 10 7.5 Faison et al.
4
Single 30.5-cm orifice in. before 9 X 10 41 (Fl)
4
Single 20.3-cm orifice and after 7 X 10 176
orifice

3
Trapezoidal baffles (Fig. — Aq. H S 0 2 4 Aq. NaOH 1 1 1 1 1.9 8 X 10 6.6 Hartung
16 and Table XII) and Hiby
(H7)

Orifice Simpson
Fract. open area = 0.76 0.25 — (S5)
Fract. open area = 0.60 1.0 —
Fract. open area = 0.38 5.0 —
4
Cylindrical-tube grid « 1 0 cm Aq. NaCl Water «1 -1 — 1 5.1 1.5 X 10 to — — Stenquist
4
(Fig. 22) before 5 X 10 and
(Tube spacing)/(tube grid and 1.0 Kaufman
diam.) = 4 120 cm (S10)
(Tube spacing)/(tube after 3.3
diam.) = 2
3
Pipe 90° elbow + 62 — — — — — — — - 3 X 10 3.2 — Hiby ( H l l )
pipe diam. (elbow
radius = 1 AD). See
also Table XII
3
U-bend + 75 pipe diam. — — — — — — — - 3 X 10 2.6 — Hiby ( H l l )
(U-radius = 5.5Z)). See
also Table XII
3
Kenics (Fig. 12 and — Aq. H S 0 2 4
Aq. NaOH 1 1 1 1 1.9 8 X 10 — 0.49 Hartung
Table XII) and Hiby
(H7)
3
Kenics (Fig. 12 and 3.5 cm Aq. NaCl Water -2.5 -0.6 — 1 2.35 4 X 10 — 2.0 Laimer (LI)
Table XII), 8 elements before
tracer and
9.8 cm

107
after
Kenics
3
Lightnin® Inliner™ (Fig. — — — — — — — — > 2 X 10 — 2.5-7 Mixing
15) Equip.
Co. ( M i l )
3
LPD mixer (Fig. 13) — — — — — — — — > 2 X 10 — 5.7 Chas. Ross
Co. (R3)
4 a
Sulzer mixer (Fig. 14) 32 cm Aq. NaCl Water 0.9 0.7 0.05 1 3.5 2 X 10 — 8.4 Laimer ( L l )
(type not identified) before
tracer and
4.3 cm
after elem.
3
Sulzer type SMV mixer — — — — — — — > 2 X 10 — ~4 Williams
(Fig. 14)
_ (W2)
3
Sulzer type SMV mixer — co 2
Air — — — -1.5 45 > 2 X 10 — 3.8 Tauscher
(5 plates) and Streiff
(Tl)
Sulzer Type SMV mixer 6.8
(9 plates)
a
See also Buergi et al. (B9).
108 Joseph Β. Gray

t i m e θ is n o t the same as t h e mixing time, which is t h e pipe length required


for t h e desired extent of mixing divided by t h e fluid velocity in t h e pipe, θ
includes t h e t i m e t o accelerate fluids for injection into t h e tail pipe a n d
decelerate after t h e e n d of the mixing pipe.
T h e power Ρ to p u m p fluids A a n d Β from locations 1 a a n d 1 b t o location 6
is given by t h e equation

Ρ= wj>JPu + WbAb/Ab (26)

where / ? a n d p are gauge pressures a n d the t w o fluids are at 1 a t m o n t h e


l a lb

suction side of the p u m p .


T h e power P^ t o m i x fluids A a n d Β can be defined as t h e power con­
verted t o heat between locations 3 a n d 5 if the length from 4 t o 5 is n o longer
t h a n that needed t o complete t h e mixing desired:

(27)

C. EXPERIMENTAL PRESSURE LOSSES


Table X V I I presents pressure loss data for several types of mixers. Straight
pipe (with n o baffles or jet flow pattern modifiers), tee mixers, orifices, a n d
various baffled mixers are included. T h e pressure losses are reported in t e r m s
of the n u m b e r of velocity heads Λ ^ , which is defined by
kplp-iglgcWwWllg) (28)
Here, u is t h e fluid velocity in t h e mixing t u b e a n d u /2gis the velocity head.
2

Also, g Δp/gp is t h e friction head lost, H. F o r straight pipe with n o baffles or


c

jet flow pattern modifiers:

7Vvh = 4 / ( L / Z ) ) (29)

where / i s t h e F a n n i n g friction factor (S1). F o r straight a n d for baffled mixing


tubes in which t h e pressure d r o p is the same per unit mixing t u b e length, t h e
head loss is reported in t e r m s of N^/iL/D) which is t h e n u m b e r of velocity
heads per diameter of tail pipe length.
T h e n u m b e r of velocity heads reported for a n opposed-flow tee by Henzler
(H9) covers a wide range, from 1 t o 1 0 . T h e lowest value occurs when t h e
m o m e n t u m fluxes wu/g of the t w o streams are equal.
c

F o r Kenics mixing elements, C h e n (C5) gives a n equation for calculating


pressure loss:
Nyn/iL/D) = 3.6fK N i°
OT R (30)
where Κ is a constant which is a function of pipe diameter a n d type of
ΟΎ

Kenics element. Values of Κ should be obtained from the manufacturer.


οτ
1 3 . Turbulent Radial M i x i n g in Pipes 109

F o r the Lightnin® mixing elements shown in Fig. 15 (Μ 11), the following


e q u a t i o n can be used:

N^HL/D) = 266/< WRe°0 6 8 1


(31)
where is viscosity in centipoise.
F o r Sulzer (Koch) SMV-type mixing elements (B9, W 2 , SI3), the follow­
ing equation can be used:

N /(L/D)
VH = 4fD/d e li
2
(32)
where d is a hydraulic diameter, which can be supplied by the manufac­
H

turer, a n d e is the fraction of the pipe n o t filled by mixing element corrugated


sheets.
Coaxial flow mixer pressure changes m a y be estimated by the calculation
m e t h o d described by Sakiadis (SI) for ejectors a n d venturi tubes. Experi­
mentally d e t e r m i n e d data m a y be found in A h m e d ( A l ) , Mikail (M8), a n d
Schulz (S3).
Miller (M9) developed correlations for predicting the head lost in 90 ° pipe
tees of circular cross section. T h e n u m b e r of velocity heads lost is presented
in Table XVIII for values of the flow ratio qjq a n d DJD calculated from
m

Eq. (14). T h e (N^X values are the velocity heads for side stream to exit pipe
flow. T h e {Νγγ^η values are for the entering m a i n stream to the exit stream.
T h e head lost can be calculated by multiplying the n u m b e r of velocity heads
by the exit stream velocity head.

Table XVIII

Side-Tee Mixer, Velocity Heads Lost*

qjqm 1/(1 + qjq )s DDS AJA (Ν^\ (Λ^η) ,

0.05 0.048 0.134 0.018 <0.4 0.1


0.10 0.091 0.203 0.041 0.5 0.15
0.20 0.167 0.308 0.095 0.8 0.25
0.30 0.231 0.393 0.155 1.0 0.35
0.40 0.286 0.467 0.219 1.1 0.4
0.50 0.333 0.534 0.286 1.1 0.45
0.60 0.375 0.596 0.355 1.0 0.5
0.70 0.412 0.654 0.428 0.8 0.53
0.80 0.444 0.709 0.502 0.8 0.55
0.90 0.474 0.760 0.578 0.7 0.55
1.00 0.500 0.810 0.656 0.7 0.55

a
For use with the mixed stream velocity head u /2g. Table 2

derived from Miller (M9) and Eq. (14) in Table VIII.


110 J o s e p h Β. Gray

D. S C A L E - U P O F PRESSURE LOSSES IN M O D E L S

F o r turbulent flow, using the s a m e velocity a n d the same fluid density a n d


viscosity in geometrically similar large a n d small e q u i p m e n t will predict a
pressure d r o p for the large e q u i p m e n t which is slightly m o r e t h a n the pres­
sure d r o p measured for t h e large e q u i p m e n t . T h e n u m b e r of velocity heads
measured in a m o d e l is usually higher t h a n the n u m b e r of velocity heads for a
geometrically similar larger device.

XII. Mixing Fluids with Different Densities

A difference in density between two fluids m a k e s turbulent mixing m o r e


difficult in plain pipe, b u t with various confined jet mixer geometries it
generally has a relatively small effect o n mixing. Examples of the effect of a
density difference o n mixing are described in the following paragraphs.

A. HORIZONTAL PIPE

If t h e fluids t o be mixed differ in density, they t e n d to stratify in a horizon­


tal pipe. W h e t h e r fluids of different densities will mix or stratify depends not
only o n their densities b u t also o n their velocities, viscosities, a n d the mixing
pipe angle from horizontal.
F o r a horizontal pipe with feeds entering parallel t o the pipe axis at the
s a m e velocity, Hiby (Η 11) found that liquids will stratify if N < 15, whereFT

N =u p/(DgAp)
FT
2
(33)

H o w large the densimetric F r o u d e n u m b e r N m u s t be for complete free­


FT

d o m from stratification was n o t defined. F o r 20 < N < 250, values of 7V


FT Re

between 8,000 a n d 12,000 provide the shortest distance to effect complete


mixing.
G e r a n d Holley ( G 3 , p . 154) r e c o m m e n d that the jet densimetric F r o u d e
n u m b e r be greater t h a n 2500 t o avoid stratification.
Bakke a n d Leach ( B l ) developed a n N criterion for mixing a layer of
FT

lower-density fluid entering at t h e u p p e r edge of a rectangular duct with the


higher density fluid in the rest of the duct. They found that N should be FR

greater t h a n 100 for mixing to take place a n d that at N less t h a n 40 n o


FT

mixing would take place. F o r N greater t h a n 100, Bakke a n d Leach devel­


FT

oped a m e t h o d for predicting the rate of mixing from the rate of increase in
the thickness of the lighter-fluid layer or the distance t o obtain a specified
layer thickness. These calculation m e t h o d s are valid for horizontal or sloped
ducts.
According to H a r l e m a n ( H I ) , the interface in stratified flows is stable
w h e n the densimetric F r o u d e n u m b e r N is less t h a n one. At values greater
FR
13. Turbulent Radial M i x i n g in Pipes 111

t h a n this, interfacial waves between the strata break a n d some mixing occurs
between the stratified layers.
It m a y be inferred from a criterion given by T u r n e r (T9) for stratification
in estuaries t h a t stratification in a pipe will occur for N < 1.5 a n d mixing
FR

will occur for N > 15. Simpson (S7, p p . 2 8 9 - 2 9 5 ) placed these limits at
FT

0.77 a n d 7.7, respectively.


F u r t h e r work needs to be d o n e o n mixing of different-density fluids. At
present, the conditions a n d geometry needed to avoid difficulty c a n n o t be
specified confidently.

B. SULZER (KOCH) MIXING ELEMENTS

F o r Sulzer type S M V mixing elements, Tauscher a n d Streiff ( T l ) recom-


m e n d N > 20 to avoid stratification. M e a s u r e m e n t s which support this
FT

r e c o m m e n d a t i o n are presented by Tauscher et al (T2).


T h e appropriate diameter for D in N is the hydraulic diameter d , which
FT H

can be supplied by the mixing element manufacturer. T h e velocity u should


be the velocity through the mixing elements, which is higher t h a n the veloc-
ity through the e m p t y pipe.

C. MIXERS WITH FEEDS PARALLEL AND


PERPENDICULAR TO THE M I X I N G TUBE

G e r a n d Holley (G3) m a d e s o m e tests with different-density feed streams.


A small tracer stream was injected radially into a pipe (Table VI). T h e y
found t h a t density differences between the tracer a n d m a i n streams had a
negligible effect o n mixing of the t w o streams if the F r o u d e n u m b e r for the
side stream was greater t h a n 2500. For this type of F r o u d e n u m b e r , the
velocity a n d diameter of the small radial side stream were used to calculate

H e n z l e r ( H 9 ) included density differences between the fluids mixed as a


variable in his study of mixers with feed streams parallel a n d perpendicular
to the mixing pipe. Tests 2 a n d 6 in Table X I V (and also tests 7 a n d 8) show
that a shorter distance to mix is obtained by injecting the higher-density fluid
in the coaxial feed tube. T h e s a m e observation was m a d e for tests 19 a n d 20
a n d tests 26 a n d 27 in Table X V . T h e test pairs 29 a n d 3 1 , 3 0 a n d 32, a n d 33
a n d 34 in Table X V I also d e m o n s t r a t e this effect.

XIII. Cavitation

W h e n pressure in a flowing liquid becomes less t h a n the vapor pressure


d u e to (a) friction loss, (b) centrifugal force in a rotating liquid, or (c) a n
increase in liquid velocity p r o d u c e d by reducing the area for flow, vaporiza-
tion occurs a n d small vapor bubbles are formed. T h e rapid collapse of these
112 Joseph Β. Gray

vapor bubbles produces noise a n d can d a m a g e the adjacent surfaces in metal


or plastic e q u i p m e n t . W h e n a cavitating mixer is used in reaction kinetics
studies in which optical measuring m e t h o d s are used, the bubbles interfere
with the m e a s u r e m e n t s .
C h a n c e (C2, C 3 , C4) described m e t h o d s for avoiding such vapor forma­
tion. Degassing the solutions to be mixed sometimes helps reduce the n u m ­
ber a n d size of bubbles formed by cavitation. Higher pressure in a mixer will
eliminate cavitation if this change m a k e s all pressures in the mixer greater
t h a n the vapor pressure of the liquid. R o u n d i n g sharp edges near high-veloc­
ity flow m a y eliminate small low-pressure regions a n d cavitation therein.
Avoiding rotational m o t i o n in a mixer eliminates a low-pressure core at the
axis of the mixer exit tube.

XIV. Pressure and Flow Instability

U n d e r some conditions, the feed stream flow rates a n d pressures fluctuate


in a way that changes the feed stream flow ratio. Such ratio variations will
travel d o w n s t r e a m without m u c h attenuation since axial mixing is poor.
Rudinger (R6) stated that a vibration of several cycles per second took
place when a 3.2-mm-diameter side stream of gas a n d particles was injected
into a 114-mm square duct carrying air. Reed a n d N a r a y a n (R2) observed
periodic large flow-ratio fluctuations after a 0.08-m head change in the 7-m
water fluid head of the feed streams flowing to a mixing tee. G a u b e ( G l )
stated that vortices can activate resonant pressure vibrations in gas flowing in
a pipe into which a side stream is injected. If the vortices shed downstream of
the injected gas stream have a frequency near the natural vibration fre­
q u e n c y of the gas c o l u m n d o w n s t r e a m of the side stream, periodic pressure
fluctuations will occur.
J o h n s t o n a n d Stewart (J 1) derived equations relating feed flow rate ratio to
pressure d r o p across feed control valves a n d mixer pressure d r o p . T h e y
showed for a Y-shaped mixer that the feed control valve pressure d r o p should
be high c o m p a r e d to the mixer a n d tail pipe pressure d r o p to i m p r o v e the
stability of the feed flow rate ratio when a feed stream supply pressure
fluctuates. T h e y also showed that a venturi (eductor) mixer provides better
flow ratio stability t h a n a Y-shaped mixer.
Pressure d r o p in the feed streams before mixing d a m p s flow or pressure
disturbances (S7, p . 288). W h e n the pressure d r o p in the pipe a n d fittings
d o w n s t r e a m of a mixer is high c o m p a r e d to the pressure d r o p in the pipe a n d
fittings upstream of a mixer a n d the two feed stream pressure drops are
u n e q u a l , a change in d o w n s t r e a m pressure d r o p will change the feed stream
flow ratio. F u r t h e r m o r e , a change in the flow rate or pressure of o n e feed
stream will have less effect o n the flow rate of the other feed stream when the
13. Turbulent Radial M i x i n g in Pipes 113

after-mixer pressure drops are low. Simpson (S5) stated that the pressure
d r o p in control valves should be at least 30% of the total friction pressure
d r o p in pipe, mixer, a n d pipe fittings.

X V . Pipe Mixer Design

T h r e e steps are involved in designing a mixer to obtain radial concentra-


tion or t e m p e r a t u r e uniformity after two streams are mixed: (a) specification
of the process requirements, (b) selection of o n e or m o r e types of mixers that
are candidates for meeting the process requirements, a n d (c) defining the
functional relationships a m o n g the design a n d operating variables.

A. PROCESS REQUIREMENTS

T h e process performance requirements a n d design criteria that m u s t be


considered in selecting the d i m e n s i o n s a n d geometry of feed injection de-
vices a n d baffles for a pipe mixer are discussed in the following paragraphs.

1. Ingredient Flow Rates


T h e highest, average, a n d lowest flow rates a n d the longest allowable t i m e
for mixing d e t e r m i n e the areas for flow a n d the cross-sectional dimensions.
These r e q u i r e m e n t s also strongly influence the type of mixer selected. T w o
or m o r e mixer sizes m a y be required in parallel to meet the range of flow rates
required.

2. Ingredient Temperatures, Vapor Pressures, Densities,


and Viscosities
V a p o r pressure determines what the pressure m u s t be to avoid cavitation
w h e n velocity is high. Density a n d viscosity affect the lowest velocity which
will mix as j u d g e d by the Reynolds n u m b e r criterion for turbulence. T h e
density a n d viscosity ratios for the feed streams have a strong influence o n
mixer geometry a n d the ratios of feed stream to mixing tube velocities. T h e
m o r e the density or viscosity ratio deviates from a value of one, the m o r e
desirable are baffles, cross-flow jets, a n d high feed stream velocity ratios.

3. Longest Allowable Mixer Length and Mixing Time


T h e space into which a n installed mixer m u s t fit m a y be limited. T h e t i m e
to obtain a desired radial uniformity in the effluent stream m a y be specified,
or b o t h length a n d t i m e m a y have highest allowable values. These restric-
tions have a strong influence o n the selection of mixer geometry a n d feed
stream pressures. Shorter distances a n d mixing times can be obtained by
using higher velocities, b u t this requires larger pressure drops. T h e use of
baffles or multiple feed jets shortens the distance a n d t i m e to mix.
114 Joseph Β. Gray

A short mixing time, i.e., a few seconds, is needed for stable control of p H .
A transit t i m e as long as a m i n u t e between ingredient injection a n d p H
m e a s u r e m e n t in a pipe mixer can result in unstable p H control.
Relatively complete mixing before reaction takes place is needed for cases
in which simultaneous or consecutive reactions between partially mixed
ingredients yield a n u n w a n t e d product distribution. Mixing times as short as
0.01 to 0.001 sec can sometimes m e e t this need.

4. Effluent Stream Uniformity


T h e radial (concentration or temperature) variation coefficient σ/c m u s t
be below a n allowable highest value for s o m e operations. If n o value is
specified, 0.01 m a y be chosen; this is considered an adequately low effluent
uniformity in m a n y applications (W2). If the selected value c a n n o t be
reached within the longest mixer length or mixing t i m e allowable, a higher
velocity or a m o r e complex mixer design using baffles or multiple jets is
necessary.

5. Mixer Pressures
T h e process steps d o w n s t r e a m of a mixer determine the mixer effluent
pressure. T h e mixer design a n d dimensions a n d the ingredient flow rates
d e t e r m i n e what the mixer pressure loss will be. This loss, in turn, determines
the pressure loss across the feed stream flow control valve for adequate
control of the ingredient flow rate. All of these items a n d friction losses in
connecting pipes affect the pressure that m u s t be developed at the feed p u m p
discharge.
In s o m e cases, the pressure developed by a n existing p u m p will dictate the
pressure d r o p available for a mixer, a n d this, in turn, will influence mixer
design by imposing the highest allowable pressure d r o p for the mixer. T h e
higher costs t h a t a c c o m p a n y higher p u m p pressures provide an incentive to
keep mixer pressure losses low for high-flow-rate mixers. In such cases, a
tapered pipe at the discharge of a mixer can change velocity head to static
head a n d thereby lower p u m p i n g cost.
T h e need to avoid cavitation where a high-velocity stream m a k e s a direc­
tional change or in a jet m a y dictate the pressure at such locations. W h e n a
liquid vapor pressure reaches the static pressure, cavitation can occur. Pre­
venting cavitation is desirable to avoid pitting of adjacent mixer surfaces a n d
the production of high-frequency flow variations, which m a y cause p o o r
mixing.

6. Flow Rate Ratio Variations


Both the a m p l i t u d e a n d the range of frequencies of flow ratio fluctuations
should be restricted to keep axial concentration or temperature variations as
1 3 . Turbulent Radial M i x i n g in Pipes 115

small as radial variations. T h e a m p l i t u d e of flow ratio variations should be


n o larger t h a n the spread in concentrations or temperatures that a specified
σ/c will permit. T h e period of such σ/c fluctuations should probably be n o
higher t h a n one-fourth of the h o l d u p t i m e of the largest well-mixed vessel
d o w n s t r e a m of the pipe mixer so that averaging of the axial variations can
occur.
Large variations in feed flow rate can p r o d u c e backflow of a n o t h e r ingre­
dient into the feed channel. If a particulate solid or gelatinous product is
being m a d e , such flow fluctuations can lead to plugging of an ingredient
channel.

7. Reynolds Number
A mixing t u b e Reynolds n u m b e r larger t h a n 3000 will ensure turbulent
mixing in a n open pipe w h e n n o disturbance is introduced by the feed
streams. Higher Reynolds n u m b e r s m a y be needed when the ingredient
density a n d viscosity ratios deviate from o n e in the open pipe. Reynolds
n u m b e r s lower t h a n 2000 can be tolerated w h e n a feed jet or baffle intro­
duces a disturbance (see Η 1 0 , S9, T7). Tests in models m a y be needed to
select a n appropriate Reynolds n u m b e r to ensure mixing when ingredient
density a n d viscosity ratios deviate from o n e a n d w h e n a Reynolds n u m b e r
less t h a n 3000 appears to be necessary. T h e need for turbulence affects mixer
geometry a n d dimensions.

8. Froude Number
Stratification of ingredients with different densities m a y occur when they
are fed into open horizontal pipes coaxially at the same velocity as the fluid in
the d o w n s t r e a m mixing pipe. However, if the mixing pipe velocity is high
enough, vertical velocity c o m p o n e n t s of turbulent eddies can mix the differ­
ent-density liquids. A densimetric F r o u d e n u m b e r iV = pu /ApDg can be
Fr
2

used as a criterion to ensure such mixing in o p e n pipes. Hiby (Η 11) suggested


N > 15 a n d N > 8000 to ensure adequate mixing.
FT Re

9. Avoiding Erosion
A need to avoid erosion m a y impose a limit on fluid velocity a n d also
influence the selection of construction material. Tests with the fluid system
a n d a m o d e l of the proposed e q u i p m e n t can be helpful in detecting erosion
a n d finding a way to eliminate it.

10. Freedom from Fouling


In s o m e cases, high fluid velocity m a y reduce fouling by gelatinous or solid
reaction products. Use of Teflon® plastic for mixer construction m a y help
reduce such fouling.
116 Joseph Β. Gray

Β . INITIAL M I X E R SELECTION

M a n y types of mixer geometries are available to meet a specific set of


process requirements a n d design criteria. These were described briefly in
Section II of this chapter. In some cases, adequately long pipe is available in a
process to mix two streams brought together by an ordinary pipe tee. M o r e
often, w h e n mixing pipe length is limited, multiple feed jets or baffles are
needed to meet process requirements. Usually, m o r e t h a n one type t h a t is
satisfactory can be found.
T h e simpler, lower-cost mixer types from which to choose include (a) pipe
with n o baffles a n d with two streams fed to the opposed branches of an
ordinary pipe tee or o n e stream fed perpendicular to the other in such a tee as
shown in Figs. 1 a n d 2, (b) coaxial feed streams as shown in Fig. 3, a n d (c)
multiple stream distributors such as those in Figs. 7 a n d 8. Several types of
baffles are often a d d e d d o w n s t r e a m of coaxial feed injection as shown in
Figs. 1 2 - 1 7 . T h e multiple-stream feed distributing devices shown in Figs.
4 - 6 , 8 - 1 1 , 20, a n d 2 3 - 2 6 are m o r e costly b u t mix in shorter lengths t h a n
the simpler types.
Mixers effective for feed streams of widely different densities were tested
by Henzler (H9). In these mixers, some small feed jets were directed perpen­
dicular a n d s o m e parallel to the mixing tube axis at velocities 16 to 160 times
the mixing tube velocity. Such mixers are shown in Figs. 23 a n d 24. T h e feed
stream density ratios ranged from 1.5 to 14.
Several mixer types are often used for the study of the kinetics of fast
reactions. These mixers are small a n d have multiple high-velocity feed
streams. Small dimensions are used because the pressure d r o p to obtain the
s a m e t i m e to mix in a large mixer becomes excessively high. Dimensions are
limited by the need to obtain turbulence in the mixing tube, which m a y be as
small as 1 m m . T h e design of such mixers was first described by Hartridge
a n d R o u g h t o n (H5). M o r e recently, C h a n c e (C3, C4) described in detail the
design criteria used for a similar mixer. Mixing times of less t h a n 0.001 sec to
obtain 9 5 % completion of a neutralization reaction were demonstrated by

A
FIG. 2 6 . Tangential feed mixer. [Adapted from Swanson (SI4).]
13. Turbulent Radial M i x i n g in Pipes 117

these a n d other investigators. Examples of such very rapid mixers are shown
in Figs. 4, 11, 20, a n d 26.
Mixers t h a t a p p e a r suitable for large-scale applications include those
shown in Figs. 16,17, a n d 21 a n d the multiple-jet distributors shown in Figs.
7 - 1 0 a n d 22. Kenics (Fig. 12) a n d Sulzer (Fig. 14) mixers have also been
m a d e in large sizes. If a large mixer has a high enough velocity, a conical
expansion section can be used t o change velocity head t o static head (H4)
a n d thereby reduce t h e p u m p power a n d head required.

C. DESIGN AND OPERATING VARIABLE RELATIONSHIPS

Graphical correlations or equations are needed to relate s o m e of the design


a n d operating variables. T h e following types of functional relationships
m u s t be obtained from literature references or m o d e l tests:

aJa { = cf)[(L/D\N ]
Re (34)

(^VH)a,b = 0[(^),^ ] R e (35)

(36)

where d a n d d are feed stream diameters for c o m p o n e n t s A a n d B. S o m e


a b

selected values of σ /σ a n d Nyn are given in Table X I X .


0 {

Often t h e Reynolds n u m b e r has only a small effect o n aja at t h e s a m e


{

value of L/D. F o r s o m e geometries σ /σ increases slightly as N increases


α { Ke

(C7, D 4 , H 7 , S8), a n d for others it stays the same (B3, H 7 ) or decreases


slightly (B3, E2, F l , N l , R 5 , S8, SI 4). If n o information is available for the
geometry being considered, m o d e l tests can be carried o u t to find the corre­
lation expressed by Eq. (24). According t o Ajmera et al (A3), gas mixing can
be substituted for liquid mixing (or t h e reverse) to obtain this relationship.
A s s u m i n g t h a t increasing N has n o effect o n the relationship between ajc
Re

a n d L/D is reasonable.
A n alternative m e t h o d of defining a relationship between effluent uni­
formity a n d mixing t u b e length for a n o p e n pipe involves the use of eddy
diffusivity. T h i s is described in Section IV.
T h e n u m b e r of velocity heads Ny^ is either unaffected or decreased
slightly by a n increased Reynolds n u m b e r . Again, if n o d a t a are available to
define t h e relationship expressed by Eq. (35), model tests are needed. If the
s a m e N is used in geometrically similar m o d e l a n d prototype e q u i p m e n t by
Rc

adjusting viscosity, the values will be the s a m e at the s a m e densities a n d


velocities.
O n l y partial information exists t o define Eq. (36). O'Leary a n d Forney
( O l ) a n d G u v e n a n d Benefield (G6) developed t h e equations relating DJD
a n d qjq
m in Table VIII for a tee mixer of t h e type shown in Fig. 1. These
equations d o n o t include t h e effects of viscosity a n d density o n t h e o p t i m u m
118 J o s e p h Β. Gray

Table X I X

Mixer Performance Values

Type of mixer A^Re J V vh N /(L/D)


yH

Baffles, trapezoidal 0.38 8 X 10 3


6.6
[Fig. 16, Tables XII and
XVII (H7)]
Coaxial
[Fig. 3, Table IV ( A l ) ]
w / " b = 3, L/D = 10.5 0.032
— 10
— —
5
a

u /u = 19, L/D = 3.5 0.032


— 10 5
A B

Elbow and pipe (L/D = 62) 0.010


— 3 X 10 3.2 —
3

[Tables XII and XVII, ( H 1 1 ) ]


Grid and pipe (L/D = 20) 0.010
— 10 1-3
4

[Fig. 2 2 , Tables XIII and


XVII (SI3)]
Kenics mixer
— 0.63 > 2 X 10 1-3
3

[Fig. 12, Tables XII and


XVII (C5)]
Multiple-parallel-feed mixer 0.032
— — —
[Fig. 6, Table V for L/D = 3
andw /w = 3-5(S8)] f e e d p

Multiple-radial-feed mixer 0.032


— 8 X 10 6
3

(Tables X and XVII for


L/D = 5.2, Hartung and
Hiby)
Orifice and pipe (L/D = 4.5)
[Tables XII and XVII ( F l ) ]
0.02
— 10 180 —
5
A>rifice
/A> = i
A> fice 0.04
— 10 40 —
5
/A> = i
ri

Pipe ( n o flow modifiers)


— 0.95 10 to 10 0.04-0.01
4 7

[Tables III and XVII,


Hartung and Hiby)
Tee with opposed feeds 0.017
— 2 X 10 1-10
3

(L/D = 2)
[Table VII (H9)] (EG
Sulzer m i x e r — T y p e S M V ( T l )
Five plates
— 0.2 2 X 10 4
3

N i n e plates
— 0.1 2 X 10 — 7
3

(<7 A7i) for L/D = χ is


0 (aJayL)

DJD. O'Leary a n d F o r n e y ( O l ) found n o effect of pipe or j e t Reynolds


n u m b e r w h e n t h e former exceeded 4000 a n d t h e latter 6000. T h e y also
found t h a t decreasing t h e side-stream density t o 17% of t h e m a i n s t r e a m
density h a d only a small effect o n t h e o p t i m u m mixing conditions. Opti­
m u m feed stream diameters have n o t been d e t e r m i n e d for t h e oblique,
tangential, a n d multiple jets listed in Table VIII.
13. Turbulent Radial M i x i n g in Pipes 119

O t h e r equations required for design relate (a) velocity, t u b e diameter, a n d


fluid flow rate; (b) friction head lost, fluid velocity, a n d the n u m b e r of
velocity heads for each feed stream; (c) t i m e for mixing, distance to mix, a n d
fluid velocity; a n d (d) effluent a n d feed stream densities. If a relationship
between feed a n d effluent viscosities is needed, laboratory tests m u s t be
made.

D. PIPE M I X E R DESIGN CALCULATIONS

A n example of the use of the design equations described in Section X V , C


is given in the following paragraphs for several types of mixers. T h e process
r e q u i r e m e n t s are listed in part A of Table X X . A 6 liter/sec (approximately
95 gal./min) stream m u s t be mixed with a 1.5 liter/sec (approximately 24
gal./min) stream, which contains a catalyst at 10.0% concentration. T h e
effluent variation coefficient required is less t h a n 0.010. F o r the flow rates
a n d catalyst concentration given, the variation coefficient ajc for the un­
mixed feed is 2.0. T h e ratio of o u t p u t to i n p u t standard deviations that the
mixer m u s t provide is 0.0050.
T h e performance parameters for four mixer types are given in part Β of
Table X X . (aJa\ =L/D 1} is the ratio of o u t p u t to i n p u t standard deviations for
a mixer length of o n e pipe d i a m e t e r (see Section III,A). CNVHW>=I) is the
fluid friction head loss for o n e pipe diameter (see Section X I , C ) . T h e values
chosen for Table X X , although based o n published literature, m a y differ
from those provided by a mixer manufacturer. In all cases, coaxial feed
addition at the same velocity is used.
E q u a t i o n s for calculating mixer length, friction head loss, mixer diameter,
mixing t i m e , a n d power are given in part C of Table X X . Reynolds n u m b e r is
a s s u m e d t o have n o effect o n Λ ^ η or aja , a n d N values are large enough
{ FT

(greater t h a n 100) to ensure t h a t the feed density difference does n o t retard


mixing.
F o r o n e set of calculations, a 10-m fluid friction head loss limit was
assumed, a n d the mixing pipe diameters a n d lengths were calculated for four
mixer types. T h e results are given in Table X X I . T h e diameters ranged from
31 to 65 m m . Mixing times ranged from 0.06 to 0.33 sec. A sample calcula­
tion is given below Table X X I .
F o r a second set of calculations, a 5 0 - m m constant mixing pipe diameter
was assumed. T h e results in Table X X I I show friction head losses from 1.5 to
29 m a n d mixing times from 0.05 to 1.3 sec. Mixers such as trapezoidal
baffles (Fig. 16) have a high head loss. A larger mixing pipe diameter can
reduce this if a longer mixing t i m e is satisfactory.
F o r a third set of calculations, a 0.15-sec mixing t i m e limit was assumed.
T h e results in Table X X I I I show mixer diameters from 24 to 71 m m a n d
head losses from 2.8 to 29 m .
If a head loss < 11 m a n d a mixing t i m e < 2 0 msec is required, the Sulzer
Table X X

Mixer Design Example

A. Specifications

Stream Feed A Feed Β Effluent

Flow rate ( m / s e c ) 3
6 X 10~ 1.5 Χ 1 0 " 7.5 X 1 0 ~
3 3 3

Viscosity (Pa sec) 10~ ΙΟ" 3


10" 3 3

Density ( k g / m ) 1050
3
1100 1060
Tracer cone. (wt. %) 0 10.0 2.0
Unmixed-feed variation coefficient [from Eq. (1)]
<7i = [600(0 - 0 . 0 2 0 ) 4- 150(0.10 - 0 . 0 2 0 ) ] / / ( 7 5 0 - l ) ' = 0.040
2 2 ! 2 1 2

^ = 0 . 0 4 0 / 0 . 0 2 0 = 2.00
Effluent variation coefficient required: GJC< 0.010
Required relative standard deviation: aja = 0 . 0 1 0 / 2 . 0 = 0.0050 l

B. Mixer performance parameters

Koch Trapezoidal
Mixer type Pipe Kenics (Sulzer) baffles

( f f
o M ) ( L / D = |) 0.95 0.63 0.21 0.38
(Avh)(L/Z)=1) 0.02 2 3.8 6.6

C. Mixer design equations

(*o/*i) Q=unD /4 2

= (L/D)(N \ „ VH L/D l) '«.» = L/u


Η = (N )u /2g
VH
2
P= gQpH/g c

Table XXI

Mixer Design Example: 10 m Head Loss

Koch Trapezoidal
Kenics (Sulzer) baffles
Pipe (Fig. 12) (Fig. 14) (Fig. 16)

L/D 103.3 12 3.5 6


u, (m/sec)1 9.74 2.86 3.84 2.22
D(m) 0.0313 0.0578 0.050 0.0655
Urn) 3.2 0.69 0.017 0.39
(mix (sec) 0.33 0.24 0.050 0.177
H(m) 10 10 10 10
P(W) 779 779 779 779
Sample calculation: Kenics mixer
0.0050 = 0 . 6 3 ^ ) , L/D = 11.47 (Use L/D = 12)
(L/D)(N \ ^ = 1 2 X 2 = 24
yH L/D

H = 10 = 2 4 w / 2 X 9.:8, u = 2.86 m/sec


2

Q = 0.0075 = 2.86 nD /4, D = 0.0578 m 2

L = 0.0578 X 12 = 0.694 m
'mix = 0 . 6 9 4 / 2 . 8 6 = 0.243 sec
Ρ = gQpH/g = 9.8 X 0.0075 X 1060 X 10/1 = 779 W
c
13. Turbulent Radial M i x i n g in Pipes 121

Table XXII

Mixer Design Example: 0.050-m-Diameter Pipe

Koch Trapezoidal
Kenics (Sulzer) baffles
Pipe (Fig. 12) (Fig. 14) (Fig. 16)

L/D 103.3 12 3.5 6


u (m/sec) 3.82 3.83 3.80 3.81
D(m) 0.050 0.050 0.050 0.050
Urn) 5.2 0.60 0.175 0.30
'mix (sec) 1.35 0.156 0.046 0.079
H(m) 1.5 18 9.8 29
P(W) 120 1400 764 2280
Sample calculation: pipe
Q = 0.0075 = WTT0.050 /4 u = 3.82 m / s e c
2

σ /σ 0 { = 0.0050 = 0.95<^> L/D = 103.3


L = 1 0 3 . 3 X 0 . 0 5 = 5.16 m
^ i x = L/u = 5 . 1 6 / 3 . 8 2 = 1 . 3 5 sec
N VH = 0.02(L/D) = 0.02 X 103.3 = 2.07
Η = N ^ / l g = 2 . 0 7 Χ 3 . 8 2 / 2 X 9 . 8 = 1.54 m
2

P = gQpH/g c = 9 . 8 X 0 . 0 0 7 5 X 1 0 6 0 X 1.54/1 = 120 W

Table XXIII

Mixer Design Example: 0.15-sec Mixing T i m e

Koch Trapezoidal
Kenics (Sulzer) baffles
Pipe (Fig. 12) (Fig. 14) (Fig. 16)

L/D 103.3 12 3.5 6


u (m/sec) 16.5 3.94 1.73 2.48
D(m) 0.024 0.049 0.074 0.062
L(m) 2.48 0.59 0.26 0.37
'mix (sec) 0.15 0.15 0.15 0.15
H(m) 29 19 2.0 12.4
P(W) 2240 1480 159 970
Sample calculation: trapezoidal baffles
σ /σ
0 { = 0.005 = 0.38 ^>, (
L/D = 5.48 ( U s e L/D = 6)
0.15 = L/u
Q = unD /4 2
= 0.0075 = (L/0.15)TT(L/6) /4, 2
L = 0.372 m
D = L/6 = 0.372/6 = 0.0062 m
u = L/t = 0 . 3 7 2 / 0 . 1 5 = 2 . 4 8 m/sec
mix

A V H = {L/DWvnXu^ = 6 X 6.6 = 39.6


Η = N u /2g VH
2
= 39.6 Χ 2 . 4 8 / 2 X 9.8 = 12.4 m
2

Ρ = gQpH/g c = 9.8 X 0.0075 X 1060 X 12.4/1 = 9 6 6 W


122 J o s e p h Β . Gray

(Koch) a n d trapezoidal baffle mixers (Table X X I ) meet these requirements.


W h i c h should b e chosen d e p e n d s o n t h e s u m of the investment a n d t h e
present value of the operating costs. T h e nearest larger standard pipe d i a m e ­
ter should b e used for the mixer selected.
G a s mixing by a side tee in a pipe has been selected for a second example of
side tee a n d simple pipe mixer designs. T h e objective in this example is a n
estimate of the mixing time, pipe length, a n d fluid friction head losses t o
obtain a desired value of0.050 for the mixed stream variation coefficient σ/c.
R o w rates, pressures, a n d t e m p e r a t u r e s for two gas streams t o be mixed are

Table X X I V

Gas Mixer Design Example*

A. Specifications

NH 3 Air Effluent

Flow rate (kg/hr) 1,820 27,470 29,290


Temp. ( ° C ) 270 270 270
Pressure (kPa) 795 795 —
Pressure (std. atm.) 7.85 7.85 —
Molecular weight 17 29
V o l u m e flow rate ( m / h r )
3
608* 5,380 5,988
N H (vol. fract.)
3 1.0 0 0.102
Density ( k g / m )3
3.0 5.11 4.89
Unmixed-feed variation coefficient [from Eq. (3); see also Eq. 10 in (F2)]
{σ = [608(1 - 0 . 1 0 2 ) + 5380(0 - 0 . 1 0 2 ) / ( 6 0 8 + 5 3 8 0 - l ) ] ' = 0.302
2 2 1 2

ajc = 0 . 3 0 2 / 0 . 1 0 2 = 3.0
Effluent variation coefficient required: ajc = 0.050
Relative standard deviation required: aja = 0 . 0 5 0 / 3 . 0 = 0.0167
{

B. Mixer performance parameters (derived from Table I X )

( < 7 M W > = 1 0) =
o
0 0 0 6 5

( * o M W > = 5) = 0 . 0 3 0

Mixer type Side tee Pipe

( ^ V v hL 0.3< -

(N \ VH °.8' -
( A V hW D — 0.02
C . Mixer Design Equations

°J°i = (oJa-MiU) Q= unD /4 2

NVH = (L/D)(N \ yH L/D= 1} t mix = L/u


H=(N )u /2g
WH
2
P = gpqH/g c

a
Adapted from Buergi et al. (B9).
b
( 1 8 2 0 kg/hr)(22.4 m a t m / k g mole)(kg m o l e / 1 7 kg)(543/273)(101.3)
3

kPa/atm)( 1/795 kPa) = 6 0 8 m / h r . 3

c
For DJD = 0.22 in Table XVIII.
13. Turbulent Radial M i x i n g in Pipes 123

given in part A of Table X X I V . T h e variation coefficient for the u n m i x e d


feed is 3.0. T h e ratio of the variation coefficients, 0.0167, is the relative
standard deviation which is required of the mixing device.
T h e proposed mixing pipe diameter is 400 m m . T h e o p t i m u m side-stream
d i a m e t e r estimated from Eq. (14) in Table VIII is 88 m m . Velocities, Reyn­
olds n u m b e r s , etc. calculated for these diameters are given in Table X X V .
T h e F r o u d e n u m b e r is probably high enough for adequate mixing of the two
different-density feed streams even t h o u g h it is slightly less t h a n the magni­
t u d e of 2500 suggested by G e r a n d Holley (G3).
T h e pipe length needed to m i x for σ /σ = 0.0167 (see Table X X I V ) m a y
0 {

be estimated by using the one-jet mixer performance values in Table I X . For


φ = 0° a n d DJD = 0.042, σ /σ = 0.03 when L/D = 5 a n d 0.0065 when
δ 0 {

L/D = 10. Interpolation of these values yields L/D = 7 for aja = 0.0167. {

Probably w h e n DJD = 0.22, L/D will be less t h a n 7 for aja = 0.0167, since{

the data in Table IX for two jets show that σ /σ decreases as DJD is in­
0 {

creased at the same L/D. O n e jet would probably behave similarly. T h e use
of L/D = 7 for DJD would therefore be conservative. T h e pipe length, time,
friction head losses, a n d power for mixing are s u m m a r i z e d in Table X X V I
for this side-tee mixer.
Design m e t h o d s are described by Simpson (S7, p p . 289 - 295) for a baffled
mixer a n d several multiple feed jet mixers. G e r a n d Holley (G3) give a
detailed description of a m e t h o d for designing a tee mixer for use in measur­
ing water flow rate in a large water pipe with a tracer a n d in mixing chlorine
a n d water. O t h e r examples are presented by Buergi et al (B9), G u v e n a n d
Benefield (G6), Tauscher a n d Streiff ( T l ) , a n d Williams (W2).
A cost estimate m a y be m a d e for each mixer type by using the same
process r e q u i r e m e n t s for feed rates, effluent uniformity, a n d pressure loss.
Such costs should include m o d e l studies, if any were m a d e , a n d design,
construction, a n d operating costs. Unfortunately, the costs of m o d e l studies
a n d multiple designs can easily exceed the potential savings that an o p t i m u m
design might achieve. Seldom can such costs be justified. C o m p r o m i s e s m u s t
t h e n be m a d e .

Table X X V

Side-Tee Mixer Velocity Parameters

Side stream Main stream Mixed stream

Diameter ( m m ) 88 400 400


Velocity (m/sec) 27.8 11.9 13.2
— — 10 6

N = (Dup/μΥ
Re

N =u p/DgAp
Fr
2
2100 — 103

α
¥οτμ = 2.ΊΧ 10" Pasec.
5
124 Joseph Β. Gray

Table XXVI

Side-Tee Mixer Performance

Mixer length: L = l X 0.40 = 2.8 m


a

T i m e to mix: t = 2.8/13.2 = 0.2 sec


m

Friction head losses:


Side stream Main stream Mixed stream
Velocity (m/sec) 27.8 11.9 13.2
Friction head (m) 7.1* 2.7 C
\2
d

Mixing power: Ρ = Σ qgpH/g c

P = [(608 X 9.8 X 3.0 X 7.1 + 5380 X 9.8 X 5.11 X 2.7 + 5988


X 9.8 X 4.89 X 1.2)]/3600
P=330W

a
(log 0.0167 - log 0.0065)/(log 0.030 - log 0.0065) = (L/D - 1 0 ) / ( - 5).
* / / = 0 . 8 X 1 3 . 2 / ( 2 X 9 . 8 ) = 7.1 m.
s
2

c
H = 0.3 Χ 13.2 /(2 X 9.8) = 2.7 m.
m
2

* i / = 7 X 0.02 X 13.2 /(2 X 9 . 8 ) = 1.2 m.


p
2

XVI. Applications of Pipe Mixers

O n e of the most intensively a n d extensively studied applications of turbu­


lent confined-jet mixers has been the kinetics of chemical reactions. By using
high velocities in small-diameter mixers, a few milliliters of reactants can be
mixed in milliseconds. Examples of these small rapid mixers are described by
Hartridge a n d R o u g h t o n (H3), Milliken (M10), R o u g h t o n a n d Chance (R4),
Caldin ( C I ) , Gibson a n d Milnes (G4), Berger et al (B5), a n d Chance (C4)
(see Figs. 4 a n d 20). S o m e mixing performance data are given in Tables X
a n d XI.
T o o r et al ( T 3 , T 5 , M l , S8, V I ) used multiple parallel feeds for injecting
reactants into a mixing tube to elucidate the role of turbulent reactant con­
centration fluctuations in predicting reaction rates when the reaction t i m e is
equal to or less t h a n the t i m e of the mixing process (see also Table V).
T u r b u l e n t mixing of reactants is a c o m m o n l y used industrial application
of confined jets a n d mixing pipes. Sheeline (S4) used a modified Hartridge -
R o u g h t o n device for mixing a l u m i n u m a n d a l u m i n u m perchlorate slurries
to m a k e rocket fuel. Beer a n d Chigier (B4) described a wide variety of jet
devices for mixing fuel gas a n d air in combustion applications.
Jet mixers are often used in wastewater treatment. Vrale a n d Jorden (V2)
a n d G u v e n a n d Benefield (G6) described a side-stream mixer (Fig. 1) for
coagulation a n d flocculation agents. W a t e r chlorination a n d p H adjustment
are other examples (G3). W h e n the pipe has large diameter, multiple-ingre­
dient injection streams as shown in Fig. 7, 8, or 10 m a y be used to reduce the
pipe length to obtain complete mixing. Baffled mixers such as those shown in
1 3 . Turbulent Radial M i x i n g in Pipes 125

Figs. 12 - 1 5 are also often used. A n o t h e r relatively large-scale mixing opera-


tion using these baffled mixers is the blending of petroleum products such as
gasoline, jet fuel, diesel fuel, or heating oils.
Baffled mixers a n d feed jets can be used t o shorten t h e distance to obtain
radial t e m p e r a t u r e or concentration uniformity w h e n representative sam-
ples or m e a s u r e m e n t s m u s t be obtained in large pipes or ducts. G e r a n d
Holley (G3) a n d J o r d a n (52) describe m e t h o d s for using tracer mixing to
measure fluid flow rates.

XVII. Research Needs

In general, data are inadequate for the design of m o s t devices for p r o m o t -


ing t u r b u l e n t mixing in pipes. F u r t h e r mixing performance a n d pressure
d r o p data are needed for selected types of mixers, using the same m e t h o d s of
m e a s u r e m e n t a n d the same mixing performance criterion. Such mixing
devices include (a) coaxial feed streams (Fig. 3) with widely different veloci-
ties, (b) coaxial feed streams as in (a) b u t with the high-velocity stream
injected coaxially in the throat of a venturi (J3), (c) the Ross L P D mixer (Fig.
13), (d) side-tee mixers (Fig. 1), (e) a folded-sheet multilayered feed mixer
(Fig. 11), (f) a baffle a n d orifice mixer described by Simpson (Fig. 21), (g) the
multiple radial feed mixer described by C h a n c e (C4) a n d shown in Fig. 20,
a n d (h) multiple parallel p a t h mixers (Figs. 5, 6, a n d 10).
T h e effects of feed stream velocity, density, a n d viscosity ratio changes o n
mixing should be included in such studies. Mixing tube diameter, fluid
density, fluid velocity, a n d fluid viscosity should be varied in geometrically
similar e q u i p m e n t so t h a t the effect of Reynolds n u m b e r can be evaluated
a n d scale-up m e t h o d s can be demonstrated.
Mixing with mixing t u b e Reynolds n u m b e r s as low as 100 should be
studied further. Smith (S9, C4) d e m o n s t r a t e d mixing at 200 in a multiple
side stream mixer. Berger et al (B5) found that mixing occurred at N = 7 5 . Re

See also T u c k e r a n d Suh (T7) a n d Henzler (H10).


G e r a n d Holley (G3) described mixing tests in which a tracer was injected
continuously a n d coaxially at the s a m e velocity as fluid in a pipe. Large
deviations in distances to obtain a desired effluent tracer concentration
uniformity were found by t h e m a n d other investigators. These deviations
were n o t satisfactorily explained.

XVIII. General Comments


A t u r b u l e n t fluid in a pipe will m i x a n o t h e r stream in a radial direction a n d
eventually achieve a relatively uniform concentration distribution if the
densities a n d viscosities of the fluids are nearly the same. In this chapter, it is
126 Joseph Β. Gray

shown that m u c h shorter distances to mix can be obtained by modifying the


flow pattern by using baffles or high-velocity feed stream jets or by dividing
o n e or b o t h feeds into several streams. Shortening the t i m e a n d distance to
obtain radial uniformity increases the pressure d r o p and, consequently, the
power to mix. If the feed streams differ appreciably in density or viscosity, a
m o r e complex geometry a n d higher feed stream velocities are needed, which
also result in higher pressure d r o p a n d power to obtain a radially uniform
effluent. For s o m e mixer geometries, the time, distance, pressure d r o p , a n d
power to obtain this uniformity can be estimated. Unfortunately, the data
available for m o s t mixer types are n o t adequate for the design procedures
described.
T h e following references merit detailed study by those w h o w a n t to delve
further into radial mixing in pipes: C h a n c e (C4), Fischer et al (F4), Forney et
al ( F 7 , 0 1 ) , G e r a n d Holley (G3), H a r t u n g a n d Hiby (H7), Simpson (S6, S7,
p p . 2 8 8 , 3 2 4 ) , Streiff ^ al (B9, S 1 1 , S12, Τ 1 , T2), a n d T o o r et al ( M 1 , S8, T 3 ,
T4, T5, VI).

List of Symbols
c concentration [(mass)/(length) ] 3

c m a x i m u m tracer concentration in a main-stream pipe cross section


M

c specific heat [(heat)/(mass)(temperature)]


p

d Sulzer (or Koch) element hydraulic diameter (length)


H

d, D tube or pipe diameter (length)


Dj outside diameter of the coaxial feed distributor pipe in Fig. 8
/ Fanning friction factor (dimensionless)
F fraction of final electrical conductivity (dimensionless)
g local gravitational acceleration [(length)/(time) ] 2

g dimensional constant in equation (force) = (mass)(acceleration)/g . Values and units


c c

are

Force Mass Acceleration gc

Ν kg m/sec 2
1 m kg/N sec 2

lb f
lb ft/sec 2
32.2ftlb/lb sec
f
2

kg f kg m/sec 2
9.8 m kg/kg s e c f
2

lb f
slug ft/sec
2
1 ft slug/lb s e c
f
2

Η friction head lost (length)


7 S intensity of segregation [see Eq. (2)]
J mechanical equivalent o f heat [(force)(length)/(heat)]
L length
L M opening dimension in Keeler's square-grid distributor (Table XIII) and in Smith's
screen (S9) (see Section IX) (length)
η number o f measurements in Eq. (1) and number o f side streams in Eqs. (21) and (22)
N Fr u p/Dg Ap (densimetric Froude number) (dimensionless)
2

N Re Dup/μ (Reynolds number) (dimensionless)


N wli number of velocity heads (dimensionless)
13. Turbulent Radial M i x i n g in Pipes 127

ρ pressure [(force)/(length) ] 2

Ap pressure loss due to fluid friction [(force)/(length) ]


{
2

Ρ power [(force)(length)/(time)]
q flow rate [(length) /(time)] 3

r radius (length)
t time
Τ temperature
u velocity (length/time)
w flow rate (mass/time)
x v o l u m e fraction
v

xya v o l u m e fraction of c o m p o n e n t a or A
Δχ greatest difference in measured values of x in a v o l u m e in which mixing intensity χ is
ν v

calculated
x m e a n value of x
v y

Ζ height above a reference level (length)


χ mixing intensity, = 1 — AxJ[x (l v — x )] v
l/2

e v o l u m e fraction pipe not filled by mixing-element corrugated sheets


€ R radial eddy diffusion coefficient [(length) /(time)] 2

θ m e a n residence time (or v o l u m e divided by flow rate) (time)


μ viscosity [(mass)/(length)(time)]
ρ density [(mass)/(length) ] 3

Ap density difference [(mass)/(length) ] 3

σ standard deviation [see Eq. (1)]


τ constant in Eq. (7) (time)
φ oblique side-stream injection angle relative to main-stream flow direction (see Fig. 18)
φ side-stream injection angle relative to the main-stream pipe radius that passes through
Β

the center o f the side-stream entrance (see Fig. 19)

SUBSCRIPTS
a, b, c c o m p o n e n t or stream A, B, C
ax axial
i entering
ο leaving
J jet
ρ mixing pipe
m entering main stream
s entering side stream
t distributor tube
w wall
1, 2, 3 locations in equipment

References

(A 1) A h m e d , S. R., VDI-Forschungsh. 547, 18(1971).


(A2) A h m e d , S. R., Gas Warme Int. 25(1 &2), 7 (1976).
(A3) Ajmera, P., Singh, M., and Toor, H. L., Chem. Eng. Commun. 2(3), 115 (1976).
(A4) Alpinieri, L. J., AIAA J. 2(9), 1560 (1964).
(Bl) Bakke, P., a n d Leach, S. J., Appl. Sci. Res. 15, 97 (1965).
(B2) Baxendale, D . N . , H o m e , W., and Williams, Α., J. Inst. Fuel 4 7 , 139 (1974).
(B3) Beek, J., and Miller, R. S., Chem. Eng. Prog. Symp. Ser. 55(25), 23 (1959).
128 J o s e p h Β. Gray

(B4) Beer, J. M., and Chigier, Ν . Α., "Combustion Aerodynamics." Wiley, N e w York, 1972.
(B5) Berger, R. L., Balko, B., and Chapman, H. F., Rev. Sci. Instrum. 3 9 , 4 9 3 (1968).
(B6) Bischoff, Κ. B., and Levenspiel, O., Chem. Eng. Sci. 17, 2 4 5 , 257 (1962).
(B7) Brodkey, R. S., "Mixing: Theory and Practice" (V. W. U h l and J. B. Gray, eds.), Vol. I,
Chap. 2, pp. 7 4 - 7 8 . Academic Press, N e w York, 1966.
(B8) Brodkey, R. S., "Turbulence in Mixing Operations" (R. S. Brodkey, ed.), Chap. II.
Academic Press, N e w York, 1975.
(B9) Buergi, R., Tauscher, W. Α., and Streiff, F. Α., Chem. Ing. Tech. 5 3 , 39 (1981).
(CI) Caldin, E. F., "Fast Reactions in Solution," Chap. 3. Wiley, N e w York, 1964.
(C2) Chance, B., J. Franklin Inst. 2 2 9 , 455 (1940).
(C3) Chance, B., Eisenhardt, R. H., Gibson, Q. H., and Lonberg-Holm, Κ. K., eds., "Rapid
Mixing and Sampling Techniques in Chemistry," p. 5 1 . Academic Press, N e w York,
1964.
(C4) Chance, B., "Techniques of Chemistry" (G. G. H a m m e s , ed.), Vol. VI, Part II, pp.
6 - 2 7 . Wiley, N e w York, 1973.
(C5) Chen, S. J., Pressure D r o p in the Kenics Mixer, Bull. K T E K - 2 , Kenics Corp., North
Andover, Mass. (1978).
(C6) Chilton, Τ. H., and Genereaux, R. P., Trans. AIChE 25, 102 (1930).
(C7) Clayton, C. G., Ball, A. M., and Spackman, R., Dispersion and Mixing during Turbu­
lent R o w o f Water in a Circular Pipe, United K i n g d o m A t o m i c Energy Authority
Research Group Rept. (1968).
(D1) Danckwerts, P. V., Appl. Sci. Res. Sect. A 3 , 279 (1952).
(D2) Danckwert, P. V, Chem. Eng. Sci. 7, 116 (1957).
(D3) Davies, J. T., "Turbulence Phenomena." Academic Press, N e w York, 1972.
(D4) Delvigne, G. A. L., Paper presented at a seminar of the International Center for Heat
and Mass Transfer, Dubrovnik, Yugoslavia, August 1976. Hemisphere Publishing
Corp. (series o n Thermal and Fluid Engineering), Washington D.C., Vol. I, pp. 391 -
4 0 3 , 1977.
(Ε 1) Edwards, A. C , Sherman, W. D . , and Breidenthal, R. E., AIChE J. 3 1 , 516 (1985).
(E2) Evans, G. V., Trans. ASME J. Basic Eng. 3 4 , 6 2 4 (1967).
(Fl) Faison, Τ. K., Davis, J. C , and Achenbach, A. R., Performance of Square-Edged
Orifices and Orifice-Target Combinations as Air Mixers, U.S. Dept. Commerce, Natl.
Bur. Stand., Bldg. Sci. Ser. 12 (24 N o v e m b e r 1967).
(F2) Fan, L. T., Chen, S. J., and Watson, C. Α., Ind. Eng. Chem. 62(7), 53 (1970).
(F3) Fejer, Α. Α., Hermann, W. G., and Torda, T. P., Factors That Enhance Jet Mixing,
Aerospace Res. Lab. Rept. ARL 69-0175, Illinois Inst, of Technology, Chicago
(October 1969).
(F4) Fischer, Η. B., List, E. J., K o h , R. C. Y., Imberger, J., and Brooks, Ν . H., "Mixing in
Inland and Coastal Waters," Chaps. 2 and 3. Academic Press, N e w York, 1979.
(F5) Fitzgerald, S. D . , and Holley, E. R., ASCE J. Hydrol. Div. 107, 1 1 7 9 ( 1 9 8 1 ) .
(F6) Hint, D . L., Kada, H., and Hanratty, T. J., AIChE J. 6, 325 (1960).
(F7) Forney, L. J., "Encyclopedia of Fluid Mechanics," ( N . P. Cheremisinoff, ed.), Vol. II,
Chap. 32. G u l f P u b l . , 1986.
(F8) Forney, L. J., and K w o n , T. C , AIChE J. 25(4), 6 2 3 (1979).
(F9) Forney, L. 1 , and Lee, H. C , AIChE J. 28, 9 8 0 (1982).
(Gl) Gaube, E., Chem. Ing. Tech. 5 1 , 14 (1979).
(G2) Gegner, J. P., and Brodkey, R. S., AIChE J. 12(4), 817 (1966).
(G3) Ger, A. M., and Holley, E. R., Turbulent Jets in Crossing Pipe Flow, Hydraul. Eng. Ser.
30, U n i v . of Illinois, Urbana ( U I L U - E N G - 7 4 - 2 0 2 0 ) (1964). See also J. Hydrol. Div.
Proc. Am. Soc. Civ. Eng. 102(6), 731 (1976).
(G4) Gibson, Q. H., and Milnes, L., Biochem. J. 9 1 , 161 (1964).
13. Turbulent Radial M i x i n g in Pipes 129

(G5) Grosz-Roll, F., Int. Chem. Eng. 20, 542 (1980) [from Aufbereit. Tech. 20, (1979)].
(G6) G u v e n , O., and Benefield, L., J. Am. Water Works Assoc. 7 5 , 358 (1983).
(HI) Harleman, D . R. F., "Handbook of Fluid D y n a m i c s " (V. L. Streeter, ed.), Chap. 26.
McGraw-Hill, N e w York, 1961.
(H2) Hartridge, H., and Roughton, F. J. W., Proc. R. Soc. London Ser. Β 9 4 , 336 (1922).
(H3) Hartridge, H., and Roughton, F. J. W., Proc. R. Soc. London Ser. A 104, 376 (1923).
(H4) Hartridge, H., and Roughton, F. J. W., Cambridge Philos. Soc. Proc. 2 2 , 4 2 6 (1924).
(H5) Hartridge, H., and Roughton, F. J. W., Cambridge Philos. Soc. Proc. 2 3 , 4 5 0 (1926).
(H6) Hartung, Κ. H., and Hiby, J. W., Chem. Eng. Sci. 26, 488 (1971).
(H7) Hartung, Κ. H., and Hiby, J. W., Chem. Ing. Tech. 44(18), 1051 (1972).
(H8) H e d m a n , P. O., and Smoot, L. D . , AIChEJ. 2 1 , 372 (1975).
(H9) Henzler, H. J., "Untersuchungen z u m Homogenisieren v o n Flussigkeiten oder Gasen,"
V D I Forschungsheft 587, pp. 1 - 6 0 . VDI-Verlag G m b H , Dusseldorf, 1978.
(H10) Henzler, H. J., Chem. Ing. Tech. 51(1), 1 (1979).
(HI 1) Hiby, J. W., Verfahrenstechnik4(12), 538 (1970).
(HI2) Hiby, H. W., Chem. Ing. Tech. 5 1 , 7 0 4 (1979 [also in Int. Chem. Eng. 2 1 , 197 (1981)].
See also Fortschr. Verfahrenstech. 17, 137 (1979).
(HI3) Hill, Α. V., Proc. R. Soc. 116, 185 (1934).
(Jl) Johnston, A. K., and Stewart, D . B., Ind. Eng. Chem. Process Des. Dev. 3 , (1964).
(J2) Jordan, D . W., Q. J. Mech. Appl. Math. 14(2), 2 0 3 (1961).
(J3) Jung, R., "Die Berechnung und A n w e n d u n g der Strahlgeblase," V D I Forschungsheft
4 7 9 , 1 - 3 2 . VDI-Verlag G m b H , Dusseldorf, 1960.
(K1) Keeler, R. N . Mixing and Chemical Reactions in Turbulent Flow Reactors, Univ. Calif.
Radiation Lab. Rept. 7852, Livermore, Calif. (1964).
(K2) Keeler, R. N . , Petersen, Ε. E., and Prausnitz, J. M., AIChE J. 1 1 , 221 (1965).
(K3) Kletenik, Υ. B., Russ. J. Phys. Chem. 37(5), 638 (1963).
(K4) Kramers, H., Chem. Eng. Sci. 8, 45 (1958).
(LI) Laimer, F., Fortsche. Ber. VDI, part 7, no. 4 0 (March 1976).
(L2) Levenspiel, O., and Bischoff, Κ. B., "Advances in Chemical Engineering" (Τ. B. Drew,
J. W. H o o p e s , and T. Vermeulin, eds.), Vol. 4. Academic Press, N e w York, 1963.
(L3) Lynn, S„ Corcoran, W. H., and Sage, Β. H., AIChE J. 3 , 11 (1957).
(Ml) M a o , K. W., and Toor, H. L., Ind. Eng. Chem. Fundam. 10, 192 (1971).
(M2) Maruyama, T., Suzuki, S., and Mizushina, T., Int. Chem. Eng. 2 1 , 205 (1981) [from
Kagaku Kogaku Ronbunshu 5(5), 437 (1979)].
(M3) Maruyama, T., Mizushina, T., and Shirasaki, Y., Kagaku Kogaku Ronbunshu 7, 215
(1981).
(M4) Maruyama, T., Hayashiguchi, S., and Mizushina, T., Kagaku Kogaku Ronbunshu 8,
327(1982).
(M5) Maruyama, T., Watanabe, F., and Mizushina, T., Int. Chem. Eng. 22, 287 (1982).
(M6) Maruyama, T., Mizushina, T., and Hayashiguchi, S., Int. Chem. Eng. 2 3 , 707 (1983).
(M7) Merkel, F., "Der Chemie-Ingenieur" (A. Eucken and M. Jakob, eds.), Vol. 4, Chap.
XXIII, p. 200. Akademische Verlagsgesellschaft, Leipzig, 1934.
(M8) Mikhail, S., J. Mech. Eng. Sci. 2, 59 (1960).
(M9) Miller, D . S., "Internal Flow, a Guide to Losses in Pipe and D u c t Systems." British
Hydrodynamics Research Association, Cranfield, Bedford, England, 1971.
(M10) Milliken, G. A. Proc. R. Soc. London Ser. A 155, 2 6 9 (1936).
(Mil) Mixing Equipment Co. Bull. B564. Mixing Equipment Co., Rochester, N e w York
(Ί978).
(Μ 12) Miyairi, Y., K a m i w a n o , M., and Y a m a m o t o , K., Kagaku Kogaku (Chem. Eng. Jpn.) 3 4 ,
1315 (1970). See also Int. Chem. Eng. 1 1 , 344 (1971).
(M13) Moskowitz, G. W., and B o w m a n , R. L., Science 1 5 3 , 4 2 8 (1966).
130 Joseph Β. Gray

(Μ 14) Mozharov, Ν . Α., Chikilevskaya, Α. V., and Kormilitsyn, V. L, Thermal Eng. 19(3) 42
(1972).
( M l 5 ) Murthy, S. Ν . B., "Turbulent Mixing in Non-Reactive and Reactive Flows." Plenum,
N e w York, 1975.
(Nl) Neitchev, V. Z., Detchev, G. D., and Boyadjiev, L. Α., / Phys. E: Sci. Instrum. 3 , 722
(1970).
(01) O'Leary, C. D . , and Forney, L. J., Ind. Eng. Chem. Process Des. Dev. 24(2), 3 3 2 ( 1 9 8 5 ) .
(02) Ottino, J. M., AIChE J. 27, 184 (1981).
(PI) Patterson, G. K., "Turbulence in Mixing Operations" (R. S. Brodkey, ed.), Chap. 5.
Academic Press, N e w York, 1975.
(Rl) Reddick, H. W., and Miller, Ε. H., "Advanced Mathematics for Engineers," 3rd ed.
Wiley, N e w York, 1955.
(R2) Reed, R. D . , and Narayan, B. C , Chem. Eng. 8 6 , 131 (4 June 1979).
(R3) Ross Corp. Bull M 3 7 6 , "Ross Motionless Mixers." Charles Ross & Son, Co., Haup-
pauge, N e w York, 1980.
(R4) Roughton, F. J. W., and Chance, B., "Techniques of Organic Chemistry," Part II,
"Rates and Mechanisms of Reactions" (S. L. Friess, E. S. Lewis, and A. Weissberger,
eds.), 2nd ed., Chap. XIV, pp. 7 0 3 - 7 9 1 . Interscience, N e w York, 1963.
(R5) Roughton, F. J. W., and Milliken, G. Α., Proc. R. Soc. London Ser. A 1 5 5 , 2 5 8 (1936).
(R6) Rudinger, G., Turbulent Mixing and Combustion, Air Force Office of Scientific Re­
search Rept. A P O S R - T R - 7 5 - 1 6 6 9 , Boiling Air Force Base, National Technical Infor­
mation Service A D - A 0 2 0 6 9 4 / 6 S T (November 1974).
(51) Sakiadis, B. C , "Perry's Chemical Engineers' Handbook" (R. H. Perry, D . W. Green,
and J. O. Maloney, eds.), 6th ed., Sect. 5. McGraw-Hill, N e w York, 1984.
(52) Sato, Y., Kamiwano, M., and Y a m a m o t o , K., Preprint for 34th annual meeting of the
Society of Chemical Engineers, Japan, C 2 1 0 (1969).
(53) Schulz, R. J., A n Investigation of Ducted, Two-Stream, Variable-Density, Turbulent-
Jet Mixing with Recirculation, Arnold Eng. Dev. Center Rept. A E D C - T R - 7 6 - 1 5 2 , Ar­
nold Air Force Station, Tennessee (January 1977).
(54) Sheeline, R. D . , Chem. Eng. Prog. 6 1 , 77 (1965).
(55) Simpson, L. L., Chem. Eng. 75(13), 196 (1968).
(56) Simpson, L. L., Chem. Eng. Prog. 70, 77 (1974).
(57) Simpson, L. L., "Turbulence in Mixing Operations" (R. S. Brodkey, ed.), Chap. VI.
Academic Press, N e w York, 1975.
(58) Singh, M., and Toor, H. L., AIChE J. 20, 1224 (1974).
(59) Smith, Μ. H., Biophys. J. 13, 817 (1973).
(510) Stenquist, R. J., and Kaufman, W. J., Initial Mixing in Coagulation Processes, Environ­
mental Protection Agency Rept. EPA-R2-72-053, Acc. N o . 7 3 - 0 3 5 1 1 , PB-213902 ( N o ­
vember 1972).
(511) Streiff, F. Α., Paper C2, Third European Conference o n Mixing, B H R A Fluid Engineer­
ing, Cranfield, Bedford, England, ( 4 - 6 April 1979).
(512) Streiff, F. Α., Chem. Ing. Tech. 5 2 , 520 (1980).
(513) Sulzer Brothers Limited, "Sulzer Static Mixing," Winterthur, Switzerland, 1979.
(514) Swanson, W. M., unpublished data from internal communication, Ε. I. du Pont de
N e m o u r s & Co. (1958).
(515) Syred, N., and Beer, J. M., Combust Flame 2 3 , 143 (1974).
(Tl) Tauscher, W. Α., and Streiff, F. Α., Chem. Eng. Prog. 79, 61 (1979).
(T2) Tauscher, W., Streiff, F., and Buergi, R., VGB Kraftswerkstechnik 60(4), 290 (1980).
(T3) Toor, H. L., AIChE J. 8(8), 70 (1962).
13. Turbulent Radial Mixing in Pipes 131

(T4) Toor, H. L., "Turbulence in Mixing Operations" (R. S. Brodkey, ed.), Chap. Ill, p. 133.
Academic Press, N e w York, 1975.
(T5) Toor, H. L., and Singh, M., Ind. Eng. Chem. Fundam. 12, 448 (1973).
(T6) Towle, W. L., and Sherwood, Τ. K., Ind. Eng. Chem. 3 1 , 457 (1939).
(T7) Tucker, C. L., and Suh, N . P., Polym. Eng. Sci. 20, 875 (1980).
(T8) Tufts, L. W., and Smoot, L. D . , J. Spacecr. Rockets 8, 1183 (December 1971).
(T9) Turner, J. S., "Buoyancy Effects in Fluids," p. 158. Cambridge Univ. Press, London,
1973.
(VI) Vassilatos, G., and Toor, H. L., AIChE J. 1 1 , 666 (1965).
(V2) Vrale, L., and Jorden, R. M., / . Am. Water Works Assoc. 6 3 , 52 (1971).
(Wl) Walker, R. E., and Kors, D . L., Multiple Jet Study, NASA CR-121217 (June 1973).
(W2) Williams, G. D . , Process Eng. 61(6), 85 (June 1980).
(Y1) Y a m a m o t o , K., Kamiwano, M., and Sato, Y., Preprint for the 34th Annual Meeting of
the Society of Chemical Engineers, Japan, vol. 2, C 3 1 3 (1969).
CHAPTER 1 4

Flow and Turbulence in Vessels


with Axial Impellers
Ivan Fort*
Department of Chemical Engineering
Prague Institute of Chemical Technology
166 28 Prague 6, Czechoslovakia

I. Introduction

This chapter demonstrates that a theoretical analysis a n d experimental


confirmation can be carried o u t for the h y d r o d y n a m i c characteristics of
s o m e mechanically agitated process vessels. T h e systems considered all h a d
axial impellers with t h e shaft o n the symmetrical vertical axis in a cylindrical
t a n k equipped with radial baffles (S3, S4, U l , N l ) . T h e fluids were of low
viscosity a n d therefore the flow regime tended to be turbulent.
T h e axial impellers studied were the propeller a n d the pitched-blade tur-
bine. These have been shown to be especially efficacious for what have been
t e r m e d flow-sensitive operations. Such operations comprise homogeniza-
tion (or blending) of miscible liquids, suspension of solids, a n d b o t h heat
transfer a n d mass transfer to surfaces. T h e y include batch a n d c o n t i n u o u s
operations, h o m o g e n e o u s reactions, crystallization, a n d dissolution of the
solid phase. T h e t e r m flow-sensitive used to describe such processes suggests
t h a t they d e p e n d mainly o n the bulk flow conditions of the mixed charge.
Therefore a knowledge of the flow characteristics is basic to securing a
quantitative description of the operations. T h e flow information includes
streamlines, velocities of fluids, a n d turbulence characteristics, all as func-

* Present address: Department of Chemical and Food Process Equipment Design, Czech
Technical University, 16607 Prague 6, Czechoslovakia.

133
MIXING: THEORY AND PRACTICE, VOL. Ill Copyright © 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
134 Ivan Fort

tions of geometry. T h e special techniques required to measure these flow


variables are also briefly described.
T h e goal addressed in this chapter seems to have been first considered in
the survey of velocity fields by G r a y ( G l ) . T h u s this t r e a t m e n t serves n o t
only to extend this purpose, b u t also to completely describe the velocity of
axial impellers in baffled tanks; in fact, it comprised one facet of the author's
theoretical a n d experimental studies which are u n i q u e for their correlation
of these two approaches a n d their completeness.
H e r e the emphasis is o n the quantitative expression of the flow mecha­
nism of b o t h the whole system a n d local points, e.g., in the region of the
rotating impeller a n d at the system boundaries (at the t a n k b o t t o m a n d
walls). F o r the latter a knowledge of the velocity field is indispensable for a
description of the rate of heat transfer between the mixed charge a n d its
boundaries. It will be demonstrated that the predicted behavior of individual
subregions is always in agreement with the flow properties found experi­
mentally for the whole system. These last results were presented in the form
of relations between dimensionless n u m b e r s that characterize the d y n a m i c
a n d kinematic quantities (local, m e a n , a n d fluctuation velocity; volumetric
flow rate; axial force o n the t a n k b o t t o m ; etc.), as well as the system proper­
ties (diameter a n d speed of the impeller; viscosity a n d density of the charge).
Dimensionless n u m b e r s are also used to describe the rate of mass transfer
between the phases of the charge in the vessel, the rate of heat transfer to the
vessel surfaces, a n d the homogenization rate of miscible liquids.

II. The System: Axial High-Speed Impeller with Radial


Baffles
T h e agitated system is a cylindrical vessel with a flat b o t t o m provided with
at least three b u t usually four baffles equally spaced at the wall with their
u p p e r ends extending above the liquid surface (see Fig. 1). In the vessel axis
a n axial high-speed impeller (propeller or pitched-blade turbine as shown in
Figs. 2 a n d 3) rotates. T h e relative size of this impeller expressed as D/T does
n o t exceed 0.5. Also, the relative distance of the horizontal plane of the
blades above the b o t t o m of the tank, h /T, is never m o r e t h a n 0.5, a n d the
2

relative distance of the height of the liquid expressed as HjΤequals one. This
is t e r m e d a square t a n k configuration. T h e t a n k contents are N e w t o n i a n
h o m o g e n e o u s liquids. In addition, these general simplifying assumptions
apply:
(1) T h e e q u i p m e n t is axially symmetrical. (All cross sections in which
the vessel lies are the same.)
(2) T h e flow in the whole system is fully turbulent.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 135

FIG. 1. Cylindrical tank with an axial impeller and radial baffles.

FIG. 2. Propeller mixer with constant pitch (/? = £>); tanδ = 0.31 SD/d hl a = 17°40',
r lr=(\ 1£
136 Ivan Fort

(b)
FIG. 3 . Pitched-blade impellers, (a) a = 2 4 ° , h = 0 . 2 Z ) ; (b) a = 4 5 ° , h = 0.2D.

(3) T h e regime of the flow is quasi-stationary. 1

(4) T h e fluid system is closed for mass transfer relative to the surround­
ings. T h e fluid boundaries are the vessel b o t t o m a n d sides, the baffles, and, of
course, the free liquid surface.
(5) T h e flow velocity is zero at t h e wall a n d o n the surface.
F o r t h e system u n d e r scrutiny the coordinate system of the e q u i p m e n t is
formed by cylindrical coordinates, n a m e l y r, φ, a n d z. In Fig. 1, the origin of
the coordinate system is located at the intersection of the cylindrical axis of
s y m m e t r y of the vessel with the b o t t o m plane. T h e coordinate ζ is parallel to
this axis a n d oriented upward.

III. Convective Flow of the Vessel Contents


T h e convective flow is characterized in t e r m s of the time-averaged flow
characteristics—pressure a n d velocity v e c t o r s — a t each point. This is irre­
spective of the instantaneous flow behavior. H e r e reference to Fig. 4 is

1
Quasi-stationary refers to the time-average velocity field.
14. Flow and Turbulence in Vessels with Axial Impellers 137

·-<
1 1 —

ι Λ

( /

FIG. 4. Typical liquid circulation pattern for an axial impeller with radial baffles. Primary
flow ( ) and induced flow ( ). [From Fort (F3).]

helpful. In particular, the following quantities are used to describe the con-
vective flow of t h e charge ( F 3 , F 8 , P2):

(a) T h e p u m p i n g capacity of the impeller, or the so-called primary flow


rate K . This is t h e liquid flow rate t h r o u g h a circle in a plane at the lower
P

edge of the impeller blades with a diameter equal to that of the impeller, i.e.,
t h r o u g h the so-called rotor region of the impeller.
(b) T h e total volumetric flow rate of the mixed liquid V . This is the c

liquid that passes d o w n t h r o u g h a circular plane b o u n d e d by the zero-aver­


age radial a n d vertical velocity points Ρ in Fig. 4.
(c) T h e i n d u c e d flow rate V . This is that portion of the total flow rate
E

t h a t does not circulate t h r o u g h the rotor region. It results from the m o m e n ­


t u m transfer between the p r i m a r y flow a n d the fluid that surrounds it.

These three flow rates are used in the dimensionless flow rate n u m b e r s
defined as
P r i m a r y flow rate n u m b e r : K = V /ND
P P
3
(1)
Total flow rate n u m b e r : K c = V /ND
C
3
(2)
I n d u c e d flow rate n u m b e r : K = V /ND
E E
3
(3)
T h e n obviously this relation holds:
K = K -K
E C P (4)
F o r the calculation of the volumetric flow rate through a n axial impeller, the
circulation flow m o d e l of the vessel contents can be employed ( F l , F 3 , P2).
138 Ivan Fort

T h e validity of this m o d e l is based on these four other "C" simplifying


assumptions:
C I . T h e region of the mixed charge is completely filled by the primary a n d
induced flows; i.e., there are n o static regions.
C2. Mass transfer by macroscale m e c h a n i s m s takes place between the
p r i m a r y a n d induced flow regions.
C3. T h e probability that a n arbitrary particle of the mixed charge belongs
to the primary or induced flow rate is determined by the part of the volumet­
ric flow rate which corresponds to the partial stream considered a n d to the
s u m of volumetric flow rates of b o t h flow types.
C4. T h e r a n d o m residence t i m e of a particle carried by any of the streams
at the given flow rate has, in the region where the flow takes place, a m e a n
value expressed by the v o l u m e of the region u n d e r scrutiny divided by the
volumetric flow rate of the given stream.
U n d e r these assumptions, the volumetric flow rate through the impeller
m a y be calculated from
VP=V/T P (5)

where V is the overall v o l u m e of the mixed charge a n d τ is the so-called


Ρ

m e a n t i m e of p r i m a r y circulation, i.e., the m e a n - t i m e interval between two


successive passages of the traced liquid particles through the rotor region.
T h e tracer particle can be modeled by a suitable indicating particle (S2)
consisting of three m u t u a l l y perpendicular circular plates m a d e of Silon
0.2 m m thick a n d 6 m m in diameter (see Fig. 5). This particle was described

FIG. 5. Tracer particle. [From Steidl (S2).]


14. Flow and Turbulence in V e s s e l s with Axial Impellers 139

by Steidl (S2). A particle with such a geometry will m o v e along with the
liquid v o l u m e in the space it occupies. W h e n choosing the n u m b e r of pas­
sages of the index particle through the rotor region, which is needed for
calculation of the m e a n t i m e of p r i m a r y circulation, it is necessary to take
into consideration the distribution function of t h e quantity τ , which was Ρ

found t o be of the form (F3)

/ ( τ ρ ) = ί 0 ,
Γ
τ ρ <
; ρ
" ' " „ ^ (6)
t m exp[—m(r —T P Pmin )J, ~ ^ -
(see Fig. 6a) so that the first m o m e n t of this function equals

τ Ρ = Ι Τ ρ / ( τ )rfr = ^ + Ρ P T Pmin (7)

T h e model of the frequency function of t h e t i m e of p r i m a r y circulation


corresponds t o the c o m b i n a t i o n of a n ideal mixer in t h e space (outside the
rotor region) connected in series with a section of piston flow (inside t h e rotor
region) (L3). Accordingly, the batch m i x e d system can be considered as a
flow system in which t h e outlet recycles t o t h e inlet; this whole flow system is
consistent with a s s u m p t i o n C4 above.

IV. Convective Flow and Turbulence for Axial Impellers

T h e rotating impeller is the source of convective flow of the mixed charge.


O n t h e basis of studies of t h e velocity field in t h e liquid stream below a
rotating axial high-speed rotary mixer ( C I , F 3 , F4, F 1 0 , F l 1), t h e following
two sets of equations m a y b e written for t h e radial m e a n - t i m e velocity a n d
intensity of turbulence ( F 6 , F13):

re(0;r >: c w n = 2nNkr (8a)

(0 */w
!
e = c/r (8b)

re<r ;r ):
c e w^ = 2nNC/r (9a)
(0 1 / 2
/ w „ = ^r (9b)
N o t e that r is the p o i n t at which the value of
c is a m a x i m u m (this can be
seen in Fig. 29). T h e r e is a n analogy t o t h e R a n k i n e vortex ( N l ) for the
stream leaving the blades of the rotating impeller as well as in the induced
flow, as applied to the axial c o m p o n e n t of the liquid. T h e constants c, C, a n d
Κ used in Eqs. (8a,b) a n d (9a,b) are d e p e n d e n t on t h e size a n d type of the
axial impeller a n d are obtained experimentally. T h e quantity k used in Eq.
(8a) d e p e n d s only o n the type of axial impeller. F o r the turbine with blades
140 Ivan Fort

FIG. 6. (a) Circulation time frequency function of the indicating particle. [From Fort (F3).]
(b) Effect of blade inclination angle a o n flow rate number K ;0: two blades («„ = 2), Δ : three
P

blades (w = 3), + : four blades (n = 4), • : six blades (n = 6). [From Medek and Fort
B B B

( M 2 , M3).]
14. Flow and Turbulence in V e s s e l s with Axial Impellers 141

inclined to the horizontal plane at a n angle a , this relation has been shown to
hold (F3):

k = cot a (10a)

F o r a propeller with a constant slope of the helix, where ρ = nD, k is


defined in (F6) as

k=n/n (10b)

But here, because of t h e lift effect of t h e a e r o d y n a m i c shape of the blades


(II), Eqs. (8a) a n d (9a) for the propeller mixer b e c o m e

vv„ = InNkr + 2πΝφ (8a')

ϊν„ = InNC/r + 2πΝφ (9a')

Because Eqs. (8a') a n d (9a') m a y be transformed t o the form of Eqs. (8a) a n d


(9a) by introducing

^ax,norm= W „" 27ίΝφ (11)

only Eqs. (8a) a n d (9a) are discussed further. O n this basis, the following
e q u a t i o n s are derived for t h e m a x i m u m velocity ν ν at t h e corresponding
ωc

coordinate r : c

w^ = 2nN(kCY'
c
2
(12a)
r = (C/ky<
c
2
(12b)

so t h a t the coordinate of location r , at which the m e a n - t i m e velocity reaches


0

t h e value vv^c/2, equals


r = 2r = 2(C/ky*
0 c (13)

If Eqs. (8a) a n d (9a) are transformed t o a dimensionless form, the following


relations are obtained:
r G (0; r >: h
c ^ / ^ C = 2r/r 0(14a)

r G <r ; r . ) : W a x
c / * W = r /2r
0 (14b)
E q u a t i o n s (14a) a n d (14b) are i n d e p e n d e n t of the size a n d type of the axial
rotary impeller. But at t h e s a m e t i m e they satisfy t h e conditions t h a t at the
origin they are equal t o zero, for r = r they reach m a x i m u m values equal to
c

one, a n d for r = r they are equal t o ±. F r o m Eqs. (8b) a n d (9b) for the
0

root-mean-square values of t h e fluctuation velocity c o m p o n e n t , by use of


Eqs. (8a) a n d (9a), the following e q u a t i o n s are obtained:

r G (0; r >:c (yC) m


= 2nNkc (15a)

r G <r ; r )\
c m (0 1 / 2
= 2nN/KC (15b)
142 Ivan Fort

Table I

Pitched-Blade Impeller* Radial Profi : Mean Velocity and Velocity


h

Fluctuations

D C/nDN
(mm) D/T (mm) (mm) w^nDN r0
2r /D
0

58.0 1
5 16.2 21.8 0.747 43.6 1.50 0*2) */iVc
!
0.152
72.5 1
4 17.8 25.6 0.702 51.2 1.41 0.132
96.7 1
3 26.2 35.6 0.738 71.0 1.47 0.152

* Six 45° blades; R e > 10 , h /T 4


2 = i, Τ =290 m m ; four baffles, b/T= 0.10.
* Fort et al (F8).

W i t h respect to Eq. (12b) a n d to t h e equality

Κ=φΙ (16)

Eq. (15b) m a y be arranged to the form

r e ( r ; r >:
c e (<*) 1 / 2
= InNKc (15b')

T h u s for the given impeller type t h e root m e a n square of the fluctuation


velocity is i n d e p e n d e n t of the position along t h e radial axis r, a n d is depen­
d e n t only o n the impeller speed N. Simultaneously, the relative intensity of
turbulence, which is defined as

(0 1 / 2
/Wax,c = c(k/CY< 2
[r e (0; r >] e (17)
holds. This is a quantity practically i n d e p e n d e n t of the size of a given type of
axial rotary impeller. These conclusions are supported by experiments car­
ried o u t with a six-bladed pitched impeller with a = 45 ° (see Fig. 3), for
which the p a r a m e t e r s are given in Table I. T h u s the liquid stream u n d e r the
rotating axial impeller can be expressed by the radial dimensionless profile of
the m e a n - t i m e velocity corresponding to the velocity profiles in the mixing
z o n e of classical free jets ( A l ) a n d further by a constant value of the relative
intensity of turbulence (w^) /v?ax,c- 1/2

V. Pumping Capacity of Axial Impellers

T h e p u m p i n g capacity of a n axial high-speed rotary impeller V is o b ­ P

tained by integration of the velocity profile = w^r) over the impeller


radius D/2 (F6Y
Γ D/2 _
V F = 2π I w^r dr (18)
Jo
By this integration of velocity profiles described by Eq. (8a ) in the range 7
14. Flow and Turbulence in V e s s e l s with Axial Impellers 143

r Ε ( 0 ; r ) a n d of Eq. (9a') in the range r G ( r ; D/2) we obtain


c c

V = 4n N[k(r D/2
P
2 2
c - $r\) + φ £ / 8 ]
2
(19)

If these dimensionless quantities are introduced,

k x = 2rJD (20a)

and
k 2 = φ/D (20b)
Eq. (19) can be rearranged into the dimensionless form

V /ND*
P = K = 4n [k(k /S
P
2 2
- k]/l2) + jfc /8] 2 (21)

T h e p u m p i n g capacities of the axial high-speed impellers are d e p e n d e n t o n


t h e shape, size, a n d location of the given impeller. They were determined
experimentally by m e a s u r e m e n t s of the m e a n t i m e of the primary circula­
tion τ ( F 3 , F4, F 8 , P2). T h e results were expressed in the form of a n
Ρ

exponential relation
K = A Re~°(D/T) (h /T)
P
a
2
b
( R e > 1000) (22)

for different impeller types. T h e specific exponents a n d constants in Eq. (22)


are given in Table II for the three types of axial impellers illustrated in Figs. 2
a n d 3. It has been shown ( M l , M 2 ) t h a t the p r i m a r y flow rate n u m b e r K also P

depends o n the n u m b e r of impeller blades according to the relation

K = 0.572«g-
P
295
(a = 4 5 ° ; D/TG i ) , n G<2; 1 0 »
B (22a)

F o r impellers with pitched blades at different angles a this d e p e n d e n c e is


expressed graphically in Fig. 6b. F r o m a n inspection of this graphical repre­
sentation of K = f(a\ it can be seen t h a t the value of K increases for
P P

Table II

Pumping Capacity of Axial Flow I m p e l l e r s 00

Impeller A a b

Propeller (p = D) c
0.592 0.15 ~0
Three-pitched-blade paddle (a = 2 4 ° ) 0.387 -0.13 -0.06
Six-pitched-blade paddle (a = 4 5 ° ) 1.014 0.21 -0.17

a
From Fort (F3); Fort and Sedlakova (F4); Fort et al. (F8); Por-
celli and Marr (P2).
b
R e > 1000; four radial baffles with b/T=0A.
c
In the parlance of mixing, w h e n ρ = D for a propeller, it is said to
have a square pitch.
144 Ivan Fort

a G ( 1 0 ° ; 4 5 ° ) , changes only slightly for α G ( 4 5 ° ; 6 0 ° ) , a n d decreases for


a G ( 6 0 ° ; 9 0 ° ) . T h e latter decrease is caused by the p r e d o m i n a n t radial
c o m p o n e n t of the impeller p u m p i n g capacity that appears for this range of
angle a ( M 3 ) .
T h e experiments o n which Eq. (22) a n d (22a) are based were carried o u t by
using oriented Pitot tubes ( F 5 , F10) as well as tracer particles. F r o m this work
it follows first that for Reynolds n u m b e r s greater t h a n 1000 the primary flow
rate n u m b e r PK is i n d e p e n d e n t of the Reynolds n u m b e r . This is in agreement
with the effect of the Reynolds n u m b e r o n dimensionless characteristics,
such as the Euler (power) n u m b e r a n d the dimensionless homogenization
time, in a turbulent flow regime (S3, U 1 ) . Second, the supplementary p u m p ­
ing effect of the c o n t o u r e d impeller blades m u s t be considered. For a
pitched-blade turbine the lifting effect of the blades can be neglected; i.e., the
value of the quantity φ a n d consequently the value of the parameter k is 2

zero. T h e part of the p u m p i n g capacity of the propeller resulting from the


c o n t o u r e d propeller blades account for 30 to 50% of the total p u m p i n g
capacity; the rest of the p u m p i n g is d u e to the blade pitch. Here the value of
the p a r a m e t e r k is in the range 0.03 to 0.05 (F6, F10).
2

T h e point at which the quantity reaches the m a x i m u m value on the


radial ray at the discharge from the blades of a n impeller is located by the
radial distance r . This point depends first o n the impeller type, then on the
c

n u m b e r of its blades « , a n d finally o n the angle a of the slope of the blades.


B

F o r the three-blade pitched turbine (a = 24°) the value of k does not quitex

reach 0.24, while for the six-blade pitched turbine (a = 45°) this value is
a b o u t 0.75, as can be calculated from the data in Table I. By increasing the
value of a the value of k is decreased in accordance with the decreasing
2

radius of the region of the potential core of the R a n k i n e vortex in the stream
2

leaving the impeller blades. W i t h a n increase in the n u m b e r of blades, the


value of K as well as the radius r increases.
P c

VI. Total and Induced Flows in an Impeller-Stirred Vessel


Total volumetric flow rate V consists, as has been noted, of the p u m p i n g
c

capacity of the impeller a n d that part of the flowing charge that receives its
m o m e n t u m not from the blades of the impeller d u e to turbulent or molecu­
lar friction b u t from the b o u n d a r y layer of the primary flow. A description of
this induced flow a n d its behavior in the whole mixed system can be obtained
by construction of the field of streamlines in the system for irrotational
(nonviscous) flow, which are represented by the Laplace equation. T h e

2
For a Rankine vortex all the fluid rotates at the same angular velocity.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 145

streamlines n o t only define the velocity field at a n arbitrary point of the


charge b u t also furnish essential information o n possible changes in the
geometry of the system to p r o m o t e a flow in the vessel t h a t produces nearly
equal velocities in all parts of the vessel. This m e a n s that turbulence intensity
a n d the rate of mixing b e c o m e h o m o g e n e o u s in space (L2).
N o w we shall consider this flow as expressed by streamlines for the regions
defined in Fig. 1 for a baffled vessel with a n axial impeller, denoting by V the x

region below the impeller a n d V the region above the impeller. Besides
u

a s s u m p t i o n s C I - C 5 (given in Section II), assumption T l ( T denotes total


flow) also applies:
Tl. T h e flow in the region V (i = I, II) is irrotational.
t
3

T h e following dimensionless quantities are defined by

Dimensionless radial coordinate: R = r/T (23a)


Dimensionless axial coordinate: Ζ s s /H Z (23b)

Dimensionless axial c o m p o n e n t of the local m e a n velocity:

W n = w /nDN
n (24a)

Dimensionless radial c o m p o n e n t of t h e local m e a n velocity:


W^w^/nDN (24b)

Dimensionless stream function: Ψ = Ψ / Μ ) 3


(25)
T h e continuity equation for this system m a y then be written in the d i m e n ­
sionless form ( K 2 )
djRW^) d(RW )_ ax

—u^ +
—dz 0 ( 2 6 )

a n d the field of streamlines in regions V a n d V m a y be expressed by the


x u

Laplace e q u a t i o n ( K 2 ) in dimensionless form



2x
d¥ 1 <9Ψ 2x

^ — + —— - — — = 0 (27)
dZ dR
2
RdR 2 1
'
F o r the proposed geometric a r r a n g e m e n t of the mixed system,
H= Τ (28)
or a square t a n k is used. T h e n the dimensionless stream function Ψ, in

3
However, the flow in the cylindrical region between regions V and V cannot be regarded as l n

irrotational owing to the location of the rotor region in this space.


146 Ivan Fort

accordance with the transformation Eqs. ( 2 3 a ) - ( 2 5 ) , is given by

^rad = (l/n)(D/T) (l/R)


2
ΘΨ/dZ (29a)

^ a x = ~(l/n)(D/Tf(l/R) δΨ/dR (29b)


T h e partial differential equation [Eq. (27)] can be solved for b o t h regions V Y

a n d V with appropriate b o u n d a r y conditions, i.e., from the k n o w n values of


u

the stream function or its derivatives. A solution is sought for the b o u n d a r y


conditions given by
Ψ Ξ Ο (Z=Z b t = 0;i?e<0;0.5» (30a)

Ψ Ξ 0 ( Ζ = Z = 1; R e < 0 ; 0 . 5 »
s (30b)

Ψ = 0 (R = R y/ = 0.5; Ζ e (0; 1)) (31)


Equations ( 3 0 a ) - ( 3 1 ) are m a t h e m a t i c a l expressions of assumptions C4 a n d
C5 (Section II). T h e b o u n d a r y conditions for the surfaces S, (i = I, II)
through which the liquid enters a n d leaves regions V (i = I, II) are given by t

Ψ, = ψ , ( ϋ ) ( Ζ , = const, / = I, II) (32)


which are obtained from the experimental radial profiles = W^R)
{Z = const, i = 1, II) in the cross sections S a n d S a n d further from the
t Y u

relations

Ψ,(Λ) - Ψ,0Κ ) = n(T/D )


ρ
2
Γ W^(R)R dR ( Z , = const, i = I, II)(33)
JRP

where the value of the dimensionless stream function at the point R is set p

equal to zero
Ψ , ( Λ ) Ξ= 0
Ρ (Z = const, / = I, II)
t (34)
T h e coordinate R determines that position o n a n arbitrary profile
p

W^jiR) (Zj = const) where in the vicinity of the axis of s y m m e t r y the axial
velocity c o m p o n e n t reaches zero; i.e., with regard to Eq. (29b) at point
R = R the function has a m i n i m u m . Equation (33) t h u s expresses the radial
p

profile of the dimensionless stream function Ψ . Similarly, the relation for the
axial profile of this function can be written as

Ψ / Ζ ) - Ψ / Ζ * ) = n(T/D) R 2
\ Z
W^jiZ) dZ
Jz ht

(RjEi(R ;R yjp w =1,2, 3, . . . ) (35)


where, if Eqs. (30a) a n d (30b) are observed, it reduces to
Ψ/Ζ*) = Ψ/Ζ.) = 0 (Rj G (R ; R );j p w = 1, 2, 3, . . . ) (36)
14. Flow and Turbulence in V e s s e l s with Axial Impellers 147

By use of Eqs. (33) a n d (35) together with Eqs. (34) a n d (36), the axial a n d
radial profiles of the dimensionless stream function can be calculated by
numerical integration. F r o m the k n o w n value of the quantity Ψ along the
b o u n d a r i e s of each of the regions V a n d K , the field of streamlines in both
l n

regions can be found by numerical solution of the partial differential equa­


tion [Eq. (27)]. This provides a sequence of coordinates corresponding to
a n y value of the stream function satisfying the b o u n d a r y conditions [Eqs.
( 3 0 a ) - ( 3 1 ) ] . So we have the so-called first b o u n d a r y or Dirichlet problem,
which can be solved for the subject system by some of the standard n u m e r i ­
cal m e t h o d s using a c o m p u t e r ( R l ) . T h e flow pattern in the entire mixed

FIG. 7 . Apparatus for measuring distribution of the local velocity; Τ = 0 . 2 9 0 m; H =


0 . 2 9 0 m ( F 5 ) . Detail A, Probe: S = 6 0 ° ; U - l , U - 2 , and U - 3 are manometers.
148 Ivan Fort

system, including the cyclindrical section between regions V a n d V , can be Y u

d e t e r m i n e d by graphical or numerical interpolation of the function Ψ =


Ψ ( ϋ , Ζ ) over the region in question. By this m e a n s the field of streamlines in
a charge mixed by a n axial impeller can be defined.
Experimental data for the solution of Eq. (27), i.e., the velocity profiles
W^j = W^/R) (i = I, II) in the cross sections Si a n d S (see Fig. 1), were u

obtained from m e a s u r e m e n t s with a three-hole oriented Pitot tube for the


system with inside diameter Τ = 0.290 m , filled to height Η = 0.290 m with
distilled water (see Fig. 7). M e a s u r e m e n t s were m a d e in various regions of
the system with impellers having different relative sizes as characterized by
D/T(F5, F10, P I ) . Radial profiles of the dimensionless axial velocity c o m ­
p o n e n t W across the cross sections Si a n d S are presented together with the
n u

results of experiments in Figs. 8 a n d 9 for the six-pitched-blade impeller with


a = 45 °. It can be seen from the plot of these profiles that there is a m a r k e d
b u n c h i n g together of the streamlines in the stream above the impeller for
d o w n w a r d flow into the horizontal plane of symmetry of the impeller. These
streamlines then e x p a n d somewhat, b u t then contract again as they t u r n
u p w a r d at the vessel wall.
In a relatively large portion of the cross section Si, that is, between the
downward- a n d upward-flowing streams, the axial velocity is very low. T h e
smaller radius of this area, [see Eqs. (8a) a n d (9a)], depends o n both the
impeller blade n u m b e r n a n d the blade inclination angle a (M3). In d i m e n -
B

FIG. 8. Radial profile of dimensionless axial c o m p o n e n t of the local velocity vector in cross
section II above impeller. Six 45° pitched blades; h /T=\\ 2 four radial baffles, b/T'=0.1;
Re Χ 1 0 " = ( O ) 5.2, (O ) 7.8, ( · ) 10.4. [From Fort et al. (F10).]
4
14. Flow and Turbulence in V e s s e l s with Axial Impellers 149

1.0
(c)

t
* Ί
0.5
2 ·**

30 60 90

FIG. 9. (a) Radial profile o f dimensionless axial c o m p o n e n t o f the local velocity vector in
cross section I below impeller. Six 4 5 ° pitched blades; h /T=\\
2 four radial baffles, b/T=0A\
R e Χ 1 0 " = (O) 5.2, (3) 7.8, ( · ) 10.4. [From FoHetal. (F10).] (b) Effect o f number o f impeller
4

blades o n C . [From Medek and Fort ( M 3 ) . ] (c) Effect o f blade angle a o n C ; 1 denotes six
T x

blades o n impeller, 2 denotes three blades o n impeller. [From Medek and Fort (M3).]
150 Ivan Fort

sionless form this radius is


C =2rJT
T (36a)

F o r the pitched-blade impeller (D/TE ( i ; ^)) these dependencies are ex­


pressed in Fig. 9a,b. P a r a m e t e r C c a n be considered constant with sufficient
r

precision for the range of experiments carried out (n G (2; 10)), while the
B

lower the angle a the higher the value of C . Nevertheless, in the range
T

a G (30 °; 60 °)) the value of C c a n be considered constant a n d therefore the


r

reference area where the velocity of the charge is very low can be considered
i n d e p e n d e n t of b o t h the size a n d the blade inclination of the pitched-blade
impeller.
Solutions of the partial differential Eq. (27) for the subject arrangements of
the mixed system are plotted in Figs. 10 - 1 2 (F12); the field of streamlines
represents the flow pattern of the total flow rate of the charge for the system
being examined. Here, the n u m b e r s o n the individual curves are the values
of the stream function Ψ for each of the streamlines. T h e dashed lines
represent the planes S a n d S a n d also provide the boundaries for the regions
l n

V a n d V . Figures 13 a n d 14 give the radial a n d axial profiles for the quantity


l n

FIG. 10. Streamline field in a cylindrical vessel with an axial impeller and radial baffles. Six
45° pitched blades, D/T = ±, h /T=\\
2 four radial baffles, b/T = 0.1. [From Fort et al. (F12).]
14. Flow and Turbulence in V e s s e l s with Axial Impellers 151

FIG. 11. Streamline field in a cylindrical vessel with an axial impeller and radial baffles. Six
45° pitched blades, D/T=\, h /T=
2 J; four radial baffles, b/T=0A. [From Fort et al. (F12).]

Ψ for o n e relative size of the impeller which represents b o t h b o u n d a r y


conditions (Eq. 32) a n d a c o m p a r i s o n of experimental a n d calculated pro­
files of this quantity in different regions of the system. F r o m these results it
was found t h a t the m e a n deviation of the solution of Eq. (27) obtained for
regions V a n d V a n d the actual shape of the stream function does n o t
1 u

exceed 20% for the given a r r a n g e m e n t of the system; also, the accuracy of the
velocity field as d e t e r m i n e d by the experimental m e t h o d already described is
within ± 10%.
Inspection of the results in Figs. 1 0 - 1 2 demonstrates that the shape of the
curves is affected primarily by t h e geometric dispositon of the mixed system.
T h e field of streamlines is deformed d u e to the source of the convection flow
(the impeller), a n d the deformation is greater the smaller the relative size of
the impeller. At the same t i m e the m a x i m u m value of the stream function
increases in a given cross section for a given axial distance from the vessel
b o t t o m . T h e following relation holds ( F 8 , F 1 0 , F12, P I ) :

^max^const-eWrr 1
(37)
T h e p a r a m e t e r Q decreases with increasing distance above or below the
152 Ivan Fort

FIG. 12. Streamline field in a cylindrical vessel with an axial impeller and radial baffles. Six
45° pitched blades, D/T= \, h /T=
2 \\ four radial baffles, b/T= 0.1. [From Fort et al (F12).]

horizontal plane with the coordinate Ζ = h /T. T h e flow pattern is affected


2

primarily in the region below the plane of the impeller (region because
the radius of the conical b o u n d a r y through which the liquid passes from
v o l u m e V into V significantly changes with the relative size of the impeller,
u l

Z)/r(see Figs. 1 0 - 1 2 ) . However, this factor affects n o t only the streamlines


in the d o w n w a r d flow b u t also the streamlines at the b o t t o m a n d in the
u p w a r d flow, where a relatively smaller impeller causes closer proximity of
the same Ψ-valued streamlines. T h e turning radii of the streamlines below
the plane of a rotating impeller d o n o t appear the same for all streamlines.
Practically all the liquid changes its direction from u p w a r d to downward, i.e.,
by 180°. However, in doing so o n e part t u r n s gradually, i.e., between two
t u r n s by a b o u t 90 °, a n d flows horizontally along the b o t t o m , while the other
part follows paths that t u r n uniformly from the original into the reverse
direction. O n the other hand, it follows from Figs. 1 0 - 1 2 that the shape of
the flow pattern above the plane of the impeller (region V ) is not signifi­ u

cantly affected by the relative size of the impeller D/T, because the distribu­
tion of streamlines across the cross section S is practically independent
u

(F10, P I ) of the ratio D/T (sec Fig. 8).


14. Flow and Turbulence in V e s s e l s with Axial Impellers 153

Ι '
04 —

(A)
?H
0-2

...·»»*. β.

ι
o
(b)
04 - ~ ο ο
k Λ
σ

Φ · · Φ Φ
02 Φ
Φ
Φ

o . ~Λ Λ
..···
© |
Φ

( C)
04 _

Φ € Φ Φ Φ
· · · · Φ
0-2
Φ Φ
Φ
Φ Φ
o. - - « β
|
(d)
04 -

0-2 _

0 o°°o8888S 8
« „.
FIG. 13. Stream function radial profiles. Six 45° pitched blade paddle mixer, D/T = $,
h /T=$;
2 four radial baffles, ft/Γ — 0 . 1 ; O : determined from experimental velocity field; ·:
solution o f Laplace equation for V (i = I, II); 3: boundary condition for Laplace Eq. (27) and
t Vt

(i = I, II). (a) Ζ = 0.380, (b) Z „ = 0.310, (c) Z , = 0.207, (d) Ζ = 0.0345. [From Fort et al
(F12).]

T h e region with the higher energy level in the circulating charge is below
the b o d y of the impeller. H e r e the flow pattern is m u c h m o r e affected by the
d i a m e t e r of the impeller t h a n in the v o l u m e above the impeller, the region
with the lower energy level in the circulating charge. T h e velocity field in the
mixed fluid is t h u s strongly affected by the region with the higher energy level
in the circulating charge.
This behavior in t e r m s of the energy level of the system has also been
studied (F10); it has been d e t e r m i n e d experimentally that the spatial distri­
b u t i o n of the rate of dissipation of mechanical energy per unit of v o l u m e of
the mixed liquid (see Fig. 15) €, (ι = I, II, m) can be expressed in dimension-
154 Ivan Fort

ι ι 1 II I
(α) (d)

-
ο ·
ο .·
- ο .·
ν *

°* ι ι ι
II
1 1 1
I
(b)
(β)

- -
- 0 . ·

1 1 1 • ι ι ι
1 1 1 I I I
(c)

tf)
- ο ·


. 0 1 1 1
0050-1 001 5Ο 00 5ΟΧ )0-1 5
Ζ[-1
FIG. 14. Stream function axial profiles. Six 45°° pitched .· blades, D/T=$, h /T = \\ four 2

radial baffles, b/T = 0.1; O : determined from experimental velocity field; · : solution of Laplace
equation for V . Values ofR: (a) 0.15; (b) 0.20; ( c ) ,β*
x 0 . 2 5 ;ι (d) 0.30;
ι (ιe ) 0 . 3 5 ; (f) 0.45. [From Fort et
al. (F12).]

less form by
Po*; = €, VJpN^D 5
(i = I, II, m) (38)
P 0 / = (P-e V )/pN*D* m m (39)
// = P o , / P o
h (40)
where € is the rate of energy dissipation per unit volume (newton meters per
f

second per cubic meter, or watts per cubic meter), V the v o l u m e of a speci­ t

fied part of the fluid (/ = I, II, or m in Fig. 15) (cubic meters), Ρ the impeller
power input, a n d P o = P/pN D , a n d e V is the power dissipated in im­
3 5
m m

peller region V (watts) (see Fig. 15).


m

Table III gives values of the preceding dimensionless quantities for an


impeller with six 45°-pitched blades corresponding to the conditions of the
streamline field in this e q u i p m e n t . T h e value of the hydraulic efficiency ^ of h

this six-bladed impeller is within the range 60 to 70%, depending o n the D/T.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 155

FIG. 15. Fluid energy dissipation zones for an axial impeller in cylindrical vessel with radial
baffles. [From Fort et al. (F10).]

T h e hydraulic efficiency gives t h e percentage of the power i n p u t that is


dissipated in t h e v o l u m e of the fluid outside t h e rotor region, F , shown in m

Fig. 15. Since P o / P o , » Ρ ο / Ρ ο , in Table III, t h e rate of dissipation of


zI ζ Π

mechanical energy in a unit v o l u m e of t h e charge u n d e r t h e plane of the


impeller V greatly exceeds t h e rate of dissipation of energy in a unit of
l

Table III

Spatial Distribution o f Fluid Energy Dissipation in Vessel with Axial Mixer and Radial
Baffles***

1h Po /Po
Z I t Po Z I I /Po t

D/T R e Χ ΙΟ" 4
Po t (%) Po ZI Po z I I Χ 10 2
Po z m (%) (%)

1
5 5.2 1.22 71.6 1.13 8.7 0.52 92.7 7.3
7.8 1.23 72.4 1.17 6.5 0.51 94.1 5.9
10.4 1.18 69.4 1.09 8.8 0.59 92.3 7.7
1 5.2 1.10 64.6 1.00 10.2 0.62 90.9
4 9.1
7.8 1.12 65.8 1.03 8.8 0.60 92.0 8.0
10.4 1.18 63.4 1.09 9.2 0.54 92.4 7.8
1 5.2 1.12 65.8 1.01 11.1 0.56 90.0 10.0
3
7.8 1.08 63.4 0.94 13.9 0.59 87.4 12.6
10.4 1.07 62.3 0.94 13.0 0.60 88.2 11.8

From Foft et al. (F10).


a

Six-pitched-blade paddle mixer, a = 4 5 ° ; h /T=


b
1/4; four radial baffles, b/T
2 =0.1.
V /V=
l 28.0%, V /V=
u 68.5%, VJV= 3.5%, P o = 1.70 . c

From Kvasnicka ( K 6 ) .
c
156 Ivan Fort

Table IV

Comparison of Experimental and Theoretical


Ratios of Dissipation R a t e s * ' fl c

D/T (*ΐ/€„) βχρ


(€ι/€„) Λ

1 34.40 38.00
5
1 31.30 28.30
4
1 19.50 17.20
3

From Fort et al (F10).


a

Pitched-blade i m p e l l e r w i t h s i x 4 5 blades;/z /7" =


b 0
2

1/4; Re > 10 ; four radial baffles, b/T= 0.1.


4

€j is for v o l u m e V and e is for V .


c
x n n

v o l u m e V . This difference can be explained by the following from ( B l ) :


u

etVJg = &vP)*jVate d = l Π) (41)


where the m e a n kinetic energy of the v o l u m e unit of the stream entering
through the / t h cross section into the / th v o l u m e is defined for volume V a n d x

cross section S b y x
4

2g
= 2π
" JJr
Γ
ro 0 (2ngr f n
(42a)

a n d for volume V a n d cross section S


a n by
Cti:
= 2π (42b)
2g }nu
Jrni 2ng{TVA-rl ) x

If we assume that the local drag coefficients, or friction factors, for a


change of flow direction by 180° in volumes V a n d V are nearly equal Y n

ξι = ξη (43)
then the ratio of theoretical rates of energy dissipation per unit v o l u m e in
volumes V a n d V is
Y u

(6ι/€ ) = π Λ V ^ U / V ^ U (44)

T h e ratio of e a n d e was also calculated from the experimentally deter­


l u

m i n e d velocity profiles [ ^ ( r ) ] ! ! a n d [ ^ ( r ) ] ^ (index 1 is for the d o w n w a r d


a n d index 2 for the upward flow direction).
In Table IV good agreement is shown between the theoretical results from

4
The radial coordinate r (i = I, II) characterizes the location o f the turn where the liquid
n

stream o n the radial ray changes from an upward to a downward direction.


14. Flow and Turbulence in V e s s e l s with Axial Impellers 157

Eq. (44) a n d the experimental d a t a in Table III (see Table IV, where for
several relative impeller sizes D/T the m e a n values are given.)
T h e spatial distribution of the rate of dissipation of the mechanical energy
in a u n i t of t h e liquid is closely related to the liquid flow pattern. T h e regions
where the energy dissipation rate is small take substantially longer to achieve
a specified homogeneity t h a n regions with intense turbulence.
W h e r e viscosity reduces the effect of eddy diffusivity o n the rate of mass
transfer, the i s o h o m o c h r o n e field (lines with constant local mixing rate or
homogenization) is very close t o the streamline field lines with constant
values of the stream function Ψ (see Figs. 1 0 - 1 2 ) . T h e region of insufficient
mixing at the wall a n d at the vessel axis in the vicinity of the liquid surface
(L2) is practically identical with the dead region t h r o u g h which n o stream­
line is passing at the selected streamline distributions. A n o t h e r dead region
situated below the impeller along the vessel axis is m o r e interesting with
respect to suspension of the solid phase; it is the region where a sediment
accumulates (S3, S4, U 2 ) .
T h e streamlines in region V above the rotating axial impeller in Fig. 15
u

can be m o r e uniformly distributed by m e a n s of the draft t u b e located above


the impellers in Fig. 16. T h e impeller rotates in the cylindrical part of the
draft t u b e a n d its sense of rotation is such t h a t the liquid is directed toward
the b o t t o m . Draft t u b e d i m e n s i o n s are given in Fig. 16. Such a draft tube
should be positioned in the t a n k o n t h e basis of the knowledge of the stream­
line field without the draft t u b e .
Figure 17 shows the streamline field in region V above the draft t u b e (see
u

Fig. 16), d e t e r m i n e d by the numerical procedure previously described using


Eqs. ( 3 3 ) - ( 3 5 ) o n t h e basis of knowledge of the radial profile of the d i m e n ­
sionless axial c o m p o n e n t of the m e a n velocity in cross section S . Figure
u

17 shows that the density of streamlines in region V is considerably greater


u

t h a n t h a t for the system without the draft t u b e (see Fig. 12). A quantitative
c o m p a r i s o n of flow velocities for these conditions with those obtained with­
o u t a draft t u b e shows that the volumetric flow rate through region V n

increased by a n order of m a g n i t u d e with respect to the standard condition


(F15). In addition, the region contains a convective flow created by the
impeller itself, for which the turbulence intensity is, o n the average, higher
t h a n the turbulence intensity in the induced flow. Consequently, the inten­
sity of the mass a n d heat transfer in the subject v o l u m e will also increase,
because the value of the eddy viscosity in this region will be higher. At the
s a m e time, the draft t u b e can provide additional surface for heat transfer,
e.g., in a crystallizer. Since the degree of homogenization of miscible liquid is
directly related t o the distribution of the streamlines (L2), the installation of
a draft t u b e in region V considerably improves the spatial homogeneity of
u

the m i x e d change. Instead of a cylindrical draft tube, a conical one can be


158 Ivan Fort

t.

FIG. 1 6 . Cylindrical vessel with an axial impeller and draft tube. Six 4 5 ° pitched blades,
D/T=\, H/T= 1, 7 ^ / 7 = 0 . 3 6 7 , T /T=
3 V2/2, h' /T = ± h'/T=
2 0.067, hJT=\.

OA OA
FIG. 17. Field of streamlines in region V for a cylindrical vessel with an axial impeller and
n

draft tube. Six 45° pitched blades, D/T=\. Streamlines are numbered 1 - 1 0 ; corresponding
values of the stream function Ψ are: ( 1 ) 0 , (2) 0.002, (3) 0.005, (4) 0.01, (5) 0.015, (6) 0.025, (7)
0.035, (8) 0.05, (9) 0.07, (10) 0.1. ( ) Calculated streamlines; ( ) supposed course of
streamlines. [From Fort et al. (F15).]
14. Flow and Turbulence in V e s s e l s with Axial Impellers 159

Table V

Ratios of Induced Flow to Primary H o w for Axial Flow


Impellers*

Impeller

Propeller (p = D) 0.196* (T2


- D )/D
2 2

Three-pitched-blade turbine ( a = 2 4 ° ) 0.185* (T2


- D )/D
2 2

Six-pitched-blade turbine (a = 4 5 ° ) 0.113* (T2


— D )/D
2 2

a
h /T(E ( 1 / 5 ; 1/2); R e > 10 ; four radial baffles, b/T=
2
4
0.1
(F8,P1).
* These factors correspond to values o f Β in Eq. (45).

installed ( H 4 ) . This type of draft t u b e is c o m m o n l y used in industrial crys-


tallizers.
Fort (F3) a n d Porcelli a n d M a r r (P2) confirmed the qualitative descrip­
tion regarding the ratio of the p r i m a r y a n d induced flow rates. Conclusions
can n o w be d r a w n a b o u t streams with higher a n d lower mechanical energy
dissipation from the ratio V /V or from their d e p e n d e n c e o n the geometry
P E

of the e q u i p m e n t . This ratio is expressed by


V =
E B[(T -D )/D ]V
2 2 2
P (45)
This relation corresponds t o the a s s u m p t i o n t h a t the induced flow (in the
cross section in the impeller plane) occurs in the a n n u l u s located next to the
cylindrical region for the p r i m a r y flow with the same diameter as the im­
peller. Table V gives values of the constant Β for several types of axial-flow
impellers (see Figs. 2 a n d 3) obtained from experimental studies ( F 8 , Ρ1). O n
basis of the data given in this table, it can also be inferred that the turbine with
three 24°-pitched blades will have essentially the same flow pattern as the
propeller (p = D). This, together with the established identical p u m p i n g (see
Table II), homogenization (F8), a n d power i n p u t ( K 6 ) characteristics of
b o t h mixer types, leads to the conclusion t h a t it is m o r e appropriate to use
the simpler, easier to manufacture, three-bladed turbine t h a n the propeller.

V I I . Flow at the Bottom of the T a n k


T h e boundaries of the region V at the b o t t o m of the t a n k are shown in
bt

Fig. 18: t h e surface of the cylinder of radius R , a n n u l a r area 5 , with radii R


0 0

a n d i ? j , area S = S (R, Z) situated between the circles with coordinates


2 2

(R j , Z ) a n d (R , Z ) , a n n u l a r area S with radii of the b o u n d a r y circles R


{ 3 3 3 3

a n d R , a n d the corresponding part of the b o t t o m , wall, a n d adjacent radial


k

baffles. In the region V t o which o u r study is limited, liquid flows so that it


ht

enters the region t h r o u g h a n n u l a r area 5 \ a n d leaves it t h r o u g h a n n u l a r


area S . 3
160 Ivan Fort

- "Η
FIG. 18. Region V at the vessel bottom.
bt

T h e following additional ( F for flow) simplifying assumptions are m a d e


for V . bi

F l . T h e v o l u m e V exchanges mass with its surroundings only through


ht

areas S a n d S .
x 3

F2. T h e flow is considered t o be irrotational.

Fort et al. (F14) showed t h a t t h e dimensionless equations [Eqs. ( 2 3 a ) -


(25)] for region V can be transformed by applying the continuity e q u a t i o n
bt

[Eq. (26)] a n d the Laplace equation [Eq. (27)]; t h e n the dimensionless


stream function Ψ is d e t e r m i n e d by using Eqs. (29a) a n d (29b) a n d applying
t h e equality [Eq. (28)], namely, H=T. E q u a t i o n (27) m a y be solved in the
region V for the b o u n d a r y conditions
ht

Ψ = 0 (R = R ;Z<E(0\Z )
o x (46a)

Ψ = 0 (R = R ;Z<Ek (0; Z » 3 (46b)

Ψ=·0 (Z = 0 ; i ? G < ^ ; ^ »0 (47)

E q u a t i o n s ( 4 6 a ) - ( 4 7 ) express the m a t h e m a t i c a l validity of assumption F l .


F o r cross-sectional areas 5 , (/ = 1, 3) through which liquid enters a n d leaves
14. Flow and Turbulence in V e s s e l s with Axial Impellers 161

region V , the b o u n d a r y conditions are given by the function


bt

Ψ, = Ψ,(Λ) (Ζ,- = const; / = 1, 3) (48)

which m a y be d e t e r m i n e d by integration of (29a) a n d (29b)

Ψ , ( Λ ) = n(T/D) 2
Γ W^ (R)R
X dR, (Zl = const; R Ε (R ; 0 R ))
x

JRO (49a)

Ψ ( Λ ) = - n(T/D)
2
2
Γ W^ (R)R 3 dR ( Z = const; R Ε (R ;
3 k R ))
3

jR k (49b)
where for (Ζ,· = const; ζ = 1, 3) values d e t e r m i n e d experimentally in
areas S a n d S are substituted. Since continuity equation [Eq. (26)] holds
x 3

a n d assumptions C I a n d F l apply, we have

*¥{R Z )
u l = V{R ,Z )
3 3 = V mzJi (50)

a n d t h e projection of plane S i n t o the plane (R, Z) is the geometric locus of


2

m a x i m u m values of the function

Ψ*ΒΧ = Ψ ^ 0 Κ , Ζ ) = const (R Ε (R , t i? >;


3 Ζ Ε Ζ,,;Ζ »
3 (51)
while the shape of the curve, which is the considered projection of the plane
S , m u s t be determined. E q u a t i o n (51) represents the b o u n d a r y condition
2

for the given region of a r g u m e n t s which, together with Eqs. ( 4 6 a ) - ( 4 8 )


provides a basis for solution of the partial differential equation [Eq. (27)].
This solution, i.e., the first b o u n d a r y problem, for the subject system m a y be
d e t e r m i n e d by s o m e standard numerical m e t h o d s with the help of a c o m ­
p u t e r ( R l ) , as was d o n e for the d e t e r m i n a t i o n of the flow pattern for the
entire mixing system. T h e n the solution is the function Ψ = Ψ(Λ, Ζ ) in the
volume V . bt

T h e b o u n d a r y conditions, Eqs. (49) a n d (51), for the solution of the La­


place e q u a t i o n [Eq. (27)] m a y be d e t e r m i n e d from the radial profile of axial
pressures acting o n the vessel b o t t o m . These axial pressures result from the
flow of the charge t h r o u g h the region V . Such information concerning this
bt

v o l u m e d e p e n d s o n the shape of the curve = / ( r ) = p^Jj), which was


experimentally obtained ( F 2 , F7) a n d is represented by curves in Fig. 19.
O t h e r simplifying assumptions for the definition of the v o l u m e V are: bt

F3. Liquid enters a n d leaves v o l u m e V (see Fig. 18) through cross-sec­ bt

tional areas S a n d S only in the axial direction.


x 3

F4. T h e second power of the velocity in each of the cross-sectional areas S {

a n d S is directly proportional to the static pressure o n the vessel b o t t o m in


3

the axial (vertical) projection of the considered point.


F5. Forces d u e to gravity acting o n mass are negligible.
162 Ivan Fort

FIG. 19. Radial profile of axial pressures acting o n the vessel bottom for propeller (p = D\
D/T= 2 h /T= | ; four radial baffles, b/T = OA. Curves correspond to values of Re X 1 0 ~ o f 4

( 1 ) 7 . 1 1 , ( 2 ) 7.98, (3) 8.30, ( 4 ) 9 . 7 5 , (5) 10.65, (6) 11.52, (7) 12.50, (8) 13.40. [From Fort and
T o m e s (F2).]

Boundaries of the region V are then located at these characteristic points


bt

of the diagrams for axial pressures (Fig. 19):


(1) Radial coordinate R (inside radius of region V ) is given by the
0 hi

location of point A o n the profile of axial pressures where the derivative


dPax/dr changes from approximately zero to a nonzero value.
(2) Radial coordinate R (outside radius of cross section S ) is deter­
{ x

m i n e d by the location of point Β where the axial pressure is zero.


(3) Radial coordinate R (inside radius of cross section S ) is given by the
3 3

location of point C o n the profile of the axial pressures where the value of the
function is also zero.
(4) Radial coordinate R (outside radius of region V ) is given by the
k ht

external boundaries of the axial profile of the radial pressures. It has a


constant value
R = 0.5
k (52)
14. Flow and Turbulence in V e s s e l s with Axial Impellers 163

because the external b o u n d a r y of region V is identical t o the inside wall of ht

t h e vessel.

Regarding a s s u m p t i o n s F 3 a n d F4, t h e cross-sectional area for flow at R x

in Fig. 18 after the flow direction changes at the vessel b o t t o m , can be


considered identical t o the cross-sectional area S . Correspondingly, the x

cross-sectional area for flow at i ? , before t h e flow direction change d u e t o


3

t h e vessel b o t t o m , can be considered identical t o cross-sectional area S . 3

Heights Z a n d Z of cross sections S a n d S above t h e b o t t o m can b e


x 2 x 3

d e t e r m i n e d from t h e k n o w n q u a n t i t i e s R , R ,R , a n d R . Radial profiles


0 X 3 k

of t h e dimensionless axial velocity c o m p o n e n t W^** W^R) ( / = 1. 3)


can be d e t e r m i n e d in cross sections S a n d S from the corresponding p r o ­
x 3

files of axial pressures at the b o t t o m across the regions being considered.


If t h e angular change of flow direction caused by t h e plane b o t t o m equals
π/2 radians, because of a s s u m p t i o n F4 we can write

2flPax(r)r dr = 2npw Jj)r 2


dr (53)

If we define the dimensionless pressure at t h e vessel b o t t o m by

^ax = Pjp{nDNf (54)

t h e n by use of t h e definitions in Eqs. (23a) a n d (24a), Eq. (54) can be


rearranged t o

W (R)^[P (R)]
n u
l/1
(55)
In this way a basis can be developed for the solution of Eqs. (49a) a n d (49b)
for t h e dimensionless stream function Ψ in cross sections S a n d S . Maxi­ x 3

m u m values of the flow function are reached at points R = R a n d x

R = R . T h e curve Ψ , ^ = Ψ ^ Λ , Ζ ) t h u s represents t h e projection of the


3

wall of t h e stream conduit, i.e., part of t h e b o u n d a r y of region F . b t

T h e radial profile of the q u a n t i t y p^ between points Β a n d C(see Fig. 19) is


different from t h a t in the regions below the cross sections S a n d S . H e r e the x 3

flow reoriented after t h e t u r n of the charge at the vessel b o t t o m is considered


a n d the radial profile p^ = p^ir) indicates the average energy conditions in
this stream. T h e subject region can be subdivided into two subregions: the
b o u n d a r i e s of the first are formed by cylindrical areas having radii R (or r ) x x

a n d R (or r ) a n d by the corresponding part of the area S = S (R, Z ) , a n d


2 2 2 2

t h e b o u n d a r i e s of the second are formed by cylindrical areas having radii R 2

(or r ) a n d R (or r ) in Fig. 18 a n d by t h e corresponding part of the area


2 3 3

S = S (R, Z ) . In t h e first of subregion of oriented flow of the charge along


2 2

the b o t t o m (between points R a n d R ) contraction of t h e cross-sectional


x 2

area for flow by diminishing its axial coordinate is considered, while in t h e


second subregion (between points R a n d R ) expansion of t h e cross-sec-
2 3
164 Ivan Fort

tional area for flow along the b o t t o m resulting from a n increase in the value
of the axial coordinate is considered. If point R = R , where the quantity 2

o n its radial profile reaches its m i n i m u m , contraction of the flow area is


m a x i m u m . For b o t h these defined subregions of oriented flow along the
vessel b o t t o m , the Bernoulli equation for the real liquid a n d the continuity
equation (B2, K2) can be written. T h e system of these equations can be
solved for the quantity Z(R) ( Ψ = * F ) , i.e., for the height of the stream max

flowing along the vessel b o t t o m at the point i£, by use of assumptions C I - 3 ,


F l , a n d F5 (F14). In this way we obtain

R Ε (R ,R ):
l 2 Z(R) = Z R W /W(R)R X X X (56a)
W{R) ={{m )W +{WA)W\
x

+ (1 + ξ)[ W\ - 2ΡΜ]Υ )/( /2


1 + ξ) (56b)
R Ε (R ; R ):
2 3 Z(R) = Z R W /W(R)R
3 3 3 (57a)

W(R)=W -P (R)/W
3 ax 3 (57b)

E q u a t i o n (56a) expresses the relationship between velocity W(R) a n d height


Z(R) at locations along the curve S = S (R, Z) shown in Fig. 18. It is valid x 2

for the range i? Ε (R ;R ). Equation (57a) is similar to Eq. (56a) b u t is valid


X 2

for the range r Ε (R ;R ). E q u a t i o n (56b) was obtained from the Bernoulli


2 3

equation for real liquids. In the derivation of Eq. (57b) friction forces were
considered negligible.
F o r the contraction in the conduit, from (K3),

£ = 0.45 (58)

Quantities W a n d W are the average liquid velocities in the cross-sectional


x 2

area of the cylinder jacket of radius R a n d height Ζ , , or of radius R a n d x 3

height Z .3

W i t h regard to the validity of assumptions F3 a n d F4, the velocities m e n ­


tioned are equal to average velocities across the cross-sectional areas S or S x 3

(see Fig. 18). Therefore W a n d W can be calculated from the absolute


x 3

values of total axial forces ί a n d f ^ acting in cross-sectional areas S a n d


β 1 x

S , which are calculated by integration of profiles of radial pressures across


3

the vessel b o t t o m in Fig. 19 corresponding to cross sections S a n d S . T h u s x 3

we have

W = (ί
l ω1 l{[p{nDNm[n(R\ - tfg)]}) '
1 2
(59)

= (^/{[pinDNmWRl - Rl)])) i/2


(60)

E q u a t i o n s (59) a n d (60) represent the forces with which the liquid acts o n the
b o t t o m in regions of cross-sectional area S a n d S . There, according to t 3
14. Flow and Turbulence in V e s s e l s with Axial Impellers 165

a s s u m p t i o n F 3 , the fluid m a k e s a directional change of π / 2 radians in b o t h


cases.
In experiments, most attention has been paid to the distribution of axial
pressures at the vessel b o t t o m . These distributions were measured directly
(F2) by measuring total pressures in holes situated in the vessel b o t t o m (see
the a p p a r a t u s used in Fig. 20). T h e suitability of the proposed model, ex­
pressed quantitatively by the given field of streamlines a n d also the m o d e l for
drawing the field of streamlines in the whole system described earlier, was
proved by use of results from direct m e a s u r e m e n t s of the velocity field in the

HI

( b ) ^ '
FIG. 20. Apparatus for measuring distribution of axial pressures o n the vessel bottom, (a)
Overall arrangement; (b) view from above; (c) pressure taps.
166 Ivan Fort

Φ 1τ

_J» mcinorneters
manometers

•B

FIG. 2 1 . Apparatus for measuring local velocity distribution at the vessel bottom by means of
oriented Pitot tubes; B, bottom o f the vessel; P, support slab; 1, 2 , 3 , types o f directional Pitot
tubes used.

charge next to the b o t t o m . Experimental velocity a n d pressure measure­


m e n t s were m a d e with the oriented Pitot tube apparatus shown in Fig. 2 1 .
T h e characteristics of the flowing liquid were measured at different geomet­
ric positions a n d for the specific sizes of axial impellers used. T h e physical
properties of the fluid—dynamic viscosity a n d d e n s i t y — w e r e also mea­
sured. In the experiments, the speed of the impeller was such that the flow
regime was turbulent.
F o r the system in question, the distribution of streamlines in the region V ht

above the vessel b o t t o m was found by experimental m e a s u r e m e n t s of the


distribution of axial pressures o n the vessel b o t t o m . T h e results are given in
Figs. 2 2 - 2 4 . Here, beside the dimensionless radial (R) a n d axial ( Z ) coordi­
nates, the streamlines are also drawn, i.e., curves with a constant value of the
stream function Ψ . T h e n u m b e r o n each streamline gives the value of the
stream function corresponding to it.
F r o m a comparison of the field of streamlines determined in this way with
the results of direct m e a s u r e m e n t of the velocity field by use of oriented Pitot
tubes ( F 5 , K4), it m a y be concluded (F14) that the streamlines in region V bt

can be plotted from the measured radial profiles of axial pressures acting o n
t h e vessel b o t t o m with a n average accuracy better t h a n 2 1 % . T h e accuracy
with which the experimental radial profiles were obtained was considered to
be n o worse t h a n ± 10%.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 167

T h e shape of the region V so discovered corresponds to the proposed


bt

m o d e l ; i.e., it is formed by a hollow cylinder b o u n d e d at the b o t t o m by part


of the vessel, laterally by the walls a n d the adjacent radial baffles, a n d at
the t o p by the a n n u l u s through which the liquid enters (Si) a n d leaves (S ) 3

the reference region a n d also by the area of second order characterizing the
successive contraction a n d expansion of the stream flowing through the

FIG. 2 2 . Field of streamlines in region V calculated from the radial profiles of axial pressures
ht

acting o n the vessel bottom. Six-pitched-blade impeller (a = 45°); four radial baffles, b/T =
0.1; D/T= R e = 1.39 Χ ΙΟ ; Λ / Γ = (a) ±, (b) ±, (c) 0.418.
5
2
168 Ivan Fort

FIG. 22 (continued)

region V . F r o m the inside this region is limited by the cylindrical shell with
bt

radius R in which is situated the vessel axis, where the flow of the charge is
0

irregular. W i t h changes in the liquid kinematic viscosity the shape a n d size of


t h e v o l u m e V together with the flow intensity in this region change signifi­
bt

cantly a n d in a systematic way (see Figs. 22 - 24). As the kinematic viscosity


of the charge increases, the size of the v o l u m e V also becomes larger (see
bi

Figs. 2 2 - 2 4 also).

FIG. 2 3 . Field of streamlines in region V calculated from the radial profiles of axial pressures
bt

acting o n the vessel bottom. Six-pitched-blade impeller ( a = 45 °); D/T = four radial baffles,
b/T= 0.1; Re = 7.76 Χ 10 ; h /T=
4
2 (a) ±, (b) ±, (c) 0.418.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 169

(b)
0.14

0.12

0.10

0.08
Ι

Ν 0.06

0.04

0.02

0
0.5 0.4 0.3 0.2 0.1
_ R [ - ]

FIG. 2 3 (continued)

T h e effect of increasing the liquid kinematic viscosity o n the flow pattern


at the b o t t o m can be explained in t e r m s of greater dissipation of mechanical
energy transformed into heat in the liquid. Correspondingly, the m e a n flow
velocity decreases at a constant volumetric flow rate so that the cross-sec­
tional area m u s t be larger t h a n t h a t for a lower-viscosity liquid. At a lower
m e a n velocity a n d lower values of the velocity gradient across the subject
region, the rate of dissipation of mechanical energy both increases d u e to the
increased liquid viscosity a n d decreases d u e to the decreased specific kinetic
energy of the liquid. So the flow pattern in a n agitated system is significantly
170 Ivan Fort

affected by the material properties of the h o m o g e n e o u s liquid being mixed


while the overall quantities characterizing the mixed system, e.g., its power
i n p u t or specific power input, essentially d o n o t change. In the space below
the rotating axial impeller in turbulent flow, a n average of 60% of the power
i n p u t of the mixers is dissipated (F10) (see Table III), although this v o l u m e
represents only a b o u t one-fourth of the entire charge volume. This spatial
distribution of dissipation will therefore decisively affect the overall balance

0.14 - (Α)

0.12· -©•^βρ -Φ— -o -j

0.10

0.08·

Ν 0.06-
Λ \ \ \ \ ^ 0125 y
0.04- \ \ \ "^9^ o.ioo

0.02· Ov ^ < " ^ _ _ ^ 0 Ό 5 03


^^^_^-<Γ Χ
' [

O- 0.025°—
0 1

L J

FIG. 24. Field o f streamlines in region V calculated from the radial profiles o f axial pressures
bt

acting o n the bottom. Six-pitched-blade impeller (a = 45 °); D/T= ^; four radial baffles, b/T =
0.1; R e = 3.37 Χ 10 ; h /T = (a) i , (b)
4
2 (c) 0.418.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 171

FIG. 2 4 (continued)

of mechanical energy in the entire system. T o meet the goal of dissipating a


m i n i m u m of energy, as the kinematic viscosity of the charge increases, the
shape a n d size of region V should be modified.
bt

T h e effect of the relative distance of the mixer from the vessel b o t t o m h /T 2

is the same for all impeller sizes considered (for example, see Fig. 24); ash /T 2

increases, t h e liquid velocities decrease markedly in region Fj*. F r o m the


results of experiments (F7), the liquid velocities are proportional to
(T/h ) .
2
0A0

W i t h larger impellers the areas S a n d S (see Fig. 18) increase, a n d the


{ 3

coordinates Z a n d Z increase as well. T h e n the area of the plane S corre­


x 3 2

spondingly decreases. Also, the flow of the charge becomes greater with
increasing values of t h e ratio h /T since m o r e a n d m o r e i n d u c e d flow is
2

pulled d o w n with the p r i m a r y impeller-discharge flow as h /Tis increased. 2

Finally, the boundaries of regions 5Ί a n d & merge. T h u s the condition

R region
is satisfied w h e n thelmax 2 (42/2)R
= R of= negative
R pressures
k
3 = = (V2/4)T
at the b o t t o m (see Fig. 19)
(61)

disappears a n d boundaries of b o t t o m areas o n which the forces resulting


from the first a n d second t u r n s merge; t h e n the liquid layers flowing in the
d o w n w a r d direction are continuously reversed into the u p w a r d one, without
a n y contraction of streamlines. Further, the total flow rate at the b o t t o m
decreases to m a i n t a i n the validity of the continuity equation for the liquid
flowing in the region F . T h u s its force o n the vessel b o t t o m also decreases
b t

(H2).
In Fig. 25 the plot represents the preceding consideration, which is ex­
pressed by the d e p e n d e n c e A = A ( R e ) (Ri = Ri^J to correspond to
2max 2max
172 Ivan Fort

I- 2r m a x -j Η

Τ -

FIG. 25. Effect of Reynolds number o n the flow pattern below an axial-flow impeller in a
cylindrical vessel with radial baffles. Six-pitched-blade impeller (a = 45 °); D/T=%\ h = / ( R e ) ;
2

R c = ( - · - · · - · · - ) 10.0 Χ 10 , ( 4
) 6.66 Χ 10 , ( ) 3.33 Χ 10 , (
4 4
)
1.00 X 10 .4

the results of published experiments (F7). In this figure the limiting condi­
tion of the geometric parameters, for a given Reynolds n u m b e r , is expressed
by the m a x i m u m distance of the plane of symmetry of the impeller from the
b o t t o m , h . T h e size of the region in the vicinity of the vessel axis in Fig. 25
2max

where the flow is n o t oriented (i.e., its radius i ? ) , with increasing distance of
0

the impeller from the b o t t o m , increases with the successive transfer of


streamlines into the region V a n d the corresponding change in their direc­
bx

tion. T h e size of this dead region t h u s grows with increasing v o l u m e of


the region V a n d decreasing flow velocity inside this region (see also Sec­
bt

tion VI).
T h e influence of D/Tis represented by these different streamline densities
in the v o l u m e V a n d by the shape of region V . T h e streamline density a n d
bt bt

t h u s the velocities at the b o t t o m increase as ratio D/T decreases [see Eq.


(37)]. T h e n the possibility of the overall volumetric flow rate at the b o t t o m
exceeding the p u m p i n g capacity of the mixer is determined not only by the
impeller size b u t also by its distance from the b o t t o m . Also, the shape of
14. Flow and Turbulence in V e s s e l s with Axial Impellers 173

region V d e p e n d s o n the ratio D/T with the corresponding change in the


bt

d i a m e t e r of the jet from the rotor region (see Fig. 22). Areas 5Ί a n d S (see 3

Fig. 18) b e c o m e larger with increasing values of D / r ( t h e radius R j , for point


Β in Fig. 19, is proportional to the sixth root of D/T) a n d region S dimin­ 2

ishes correspondingly. T h e height of region V decreases as well if D/T


bt

decreases. T h u s the smaller the relative size of the mixer, the smaller the
d i m e n s i o n s of the region V , while the region below the surface S (where
bt 2

flow at the b o t t o m prevails) is expanding.


T h e region of the cylinder with radius R also decreases with a reduction in
0

the value of the ratio D/T A smaller impeller t h u s guarantees increased


velocity in the vicinity of the b o t t o m with a simultaneous decrease of the
dead region (cylinder of radius R ), while a larger impeller p r o m o t e s a
0

greater v o l u m e of flow b u t the velocities are lower. F r o m the facts m e n t i o n e d


in this paragraph it follows t h a t the flow of the agitated charge at the vessel
b o t t o m is rather complex, a n d even in the turbulent flow regime for the
agitated charge there is a significant influence of its viscosity o n the flow
pattern in this region. At the same t i m e the geometric arrangement of the
system (diameter of the impeller a n d its height above the b o t t o m ) affects the
velocity field in this region.

V I I I . Axial Force of an Axial Impeller

In this section the force interaction a m o n g impeller, vessel, a n d agitated


liquid will be discussed. In particular, axial high-speed rotary impellers
create significant axial force (thrust) and, therefore, special problems. T h e
rotation of the impeller creates a field of axial forces by which the flowing
liquid acts o n the impeller a n d the vessel walls. D e t e r m i n a t i o n of such forces
a n d their c o m p o n e n t s is relatively easy a n d can be d o n e without interfering
with the liquid velocity field. Such investigations can be related to the power
i n p u t of the mixer or to the flow rate of liquid in the vessel. So far a theoretical
analysis (S1) has been m a d e a n d the results of a large n u m b e r of experiments
(S3, S4, U 1 ) concerning the tangential force c o m p o n e n t a n d shaft t o r q u e are
available. These results are utilized to d e t e r m i n e the power i n p u t of a n
impeller. Investigation of other force c o m p o n e n t s can also provide valuable
information o n the impeller power inputs a n d flow rates (S1). T h e axial force
c o m p o n e n t ( F 2 , F 7 , Η 1 , H 2 ) for axial impellers can provide valuable infor­
m a t i o n o n flow in the system.
F o r a n axial impeller a n d radial baffles in a cylindrical flat-bottom vessel
(see Fig. 1), the vertical fluid force acting o n the vessel can be expressed as the
change in vessel weight (see Fig. 26) (F7).
T h e force f^ originates w h e n the fluid jet streaming d o w n w a r d from the
rotor region deviates from its direction a n d flows along the vessel b o t t o m .
174 Ivan Fort

FIG. 2 6 . Axial force c o m p o n e n t for mixing with an axial-flow impeller coaxially located in a
cylindrical vessel with radial baffles.

Force appears in the region where the liquid flowing along the vessel
b o t t o m changes its direction a n d starts flowing u p the cylindrical vessel
walls. Consistent with the discussion of the flow of the fluid along the b o t t o m
area, S in Section VII, the force
2 is the pressure force acting o n this
corresponding area. T h e force f „ results from internal friction in the liquid,
4

especially in the b o u n d a r y layer at the vessel wall a n d at the baffles.


T o d e t e r m i n e the relation between the velocity field of the fluid at the
vessel b o t t o m a n d the axial c o m p o n e n t of the system, the development for
the field of streamlines in the region V can be employed, b u t in a simplified
bt

form. F o r liquid flow in the e q u i p m e n t in Fig. 1, the following assumptions


(A for axial force) m u s t be m a d e in addition to general assumptions 1 - 5
m a d e earlier:
A l . T h e liquid flow causing force is the same as that causing force ί . ω 3

Both the b o t t o m areas o n which forces and are acting are the same.
A2. T h e c o m p o n e n t of force can be neglected in comparison to c o m p o ­
nents f^, f^, a n d
Values of c o m p o n e n t s f a n d ί are derived from an impulse theorem.
nl Μ 3

T h e vertical m o m e n t u m c o m p o n e n t of the free liquid stream changes in the


regions where a π/2-radian t u r n takes place [see also Eqs. (59) a n d (60)] (F7):
f^=P(V ) /nr]
m
2
(62)
14. Flow and Turbulence in Vessels with Axial Impellers 175

^=P{V 3?MT*IA-rl)
U (63)
T h e volumetric flow rates V a n d V act o n the areas S a n d S (see Fig. 18)
bn bt3 x 3

a n d create the force c o m p o n e n t s ί a n d f^, respectively. Force f ^ is


ω 1

related t o the flow along the b o t t o m of the vessel between the locations of the
t u r n s already noted, i.e., between the regions in which forces ί a n d ί act. ω 1 ω 3

In region S , where
2 acts, there is a contraction of the cross-sectional area
(and t h u s greater liquid velocity) d u e to the action of neighboring liquid
layers. T h e following equation can be written for f^:
f 2 = 2Yp(V a) /n(r -r )
aX
1
b
2 2
3
2
(64)
where Yis a n empirical coefficient a n d r (R ) a n d r ( R ) are defined in Fig. x x 3 3

18. T h e final form of a n equation relating t h e axial force c o m p o n e n t in the


mixed system ί a n d the volumetric flow rate at the vessel b o t t o m V can be
ω bt

obtained by c o m b i n i n g Eqs. ( 6 2 ) - ( 6 4 ) while taking into account the simpli­


fying assumptions A l a n d A2:

f
" ^f"L 1-8Λ? J (65)

This equation characterizes the convective flow at the vessel b o t t o m a n d


relates the volumetric flow rate F a n d the axial c o m p o n e n t ί of the
b t ω

resulting force. T w o parameters appear in this equation: radius r (or R ) of x x

t h e circular b o t t o m region in which the first reversal of flow takes place (see
Fig. 18) a n d the value Γ from Eq. (64). These parameters depend only o n the
e q u i p m e n t geometry a n d the physical properties of the system. Let the
volumetric flow rate at the vessel b o t t o m be expressed by the dimensionless
n u m b e r K or by the m a x i m u m value of the stream function
bt at the
b o t t o m [defined by Eq. (50)]. T h e n ,
4 Ξ
VJND 3
= 2rf¥ max (66)
E q u a t i o n (64) can be arranged in the dimensionless form

F
" =
p N W n K
* \ T ) R l
[ 1-8*? J ( 6 ? )

This equation relates the dimensionless n u m b e r s a n d K . F o r a given bt

value of F^, Eq. (67) can be used to calculate the quantity K (or Ψ ^ ) a n d bt

vice versa.
T h e various variable relationships which are d e p e n d e n t on flow condi­
tions in the fluid can be expressed by the general relation
/ = U(Re) (D/T) (h /T)y
v x
2 ( R e > 1.0 Χ 10 ) 4
(68)
In Table VI the d e p e n d e n t variable / is considered successively to be the
176 Ivan Fort

Table VI

Axial Force Effects of Axial-Flow Impellers 0

Impeller υ V X y

Propeller (p = D) 0.169 ~0 -0.38


Six-pitched-blade paddle 0.328 ~0 ~0 -0.40
(a = 45°)
Propeller (p = D) K or (27r^
bt max ) 0.188 ~0 -0.80 -0.27
Six-pitched-blade paddle K or (27r^
ht max ) 0.118 ~0 -1.07 -0.69
(a = 45°)
Propeller (p = D) Y 1.81 X 1 0 ~ 3
0.32 -1.33 0.63
Six-pitched-blade paddle Y 4.99 Χ 1 0 " 7
0.57 -2.76 -3.37
(a = 4 5 ° )
Propeller (p = D) Ri 0.417 -0.04 0.27 0.16
Six-pitched-blade paddle 0.293 -0.02 0.16 0.23
(a = 4 5 ° )

a
Re > 10 ; four radial baffles, b/T=0A
4
(F7).

variables F^, K Y, a n d R . Table VI also gives the exponents v, x, a n d y


bu x

a n d coefficient U calculated from the experimental data for some types of


axial impellers in the e q u i p m e n t in Fig. 1 (F7). T h e results in Table VI shows
that a n d K (or Ψ ^ ) are independent of the Reynolds n u m b e r ; this
bt

agrees with the results of other studies describing the behavior of turbulent
fluids in stirred t a n k s ( N 1 , S3, U 1 , U2).
O n the other h a n d , in agreement with the dissipation by the stream V of a bt

significant portion of the mechanical energy to heat in the region V , the bt

quantities Y a n d R in Eq. (67) characterizing the flow in the region of


{

contraction at the b o t t o m are quite d e p e n d e n t on the ratio of inertial a n d


viscous forces (see Table VI, exponent v). This was discussed a n d explained
below Eq. (65). It is also interesting to note that the dimensionless axial force
F^ for the two types of impellers in Table VI is i n d e p e n d e n t of the value of
ratio D/ Τ (exponent χ — 0) a n d also that the effects of ratio h /Ton F^ for 2

b o t h these mixers are nearly identical (exponents y are nearly the same).
These conclusions have been confirmed experimentally (H2), also for an
impeller with two 45 °-pitched blades a n d for a three-blade m a r i n e propeller
(according to the C h e m i n e e r Co.).
T h e volumetric flow rate of the mixed liquid at the vessel b o t t o m is
proportional to the p u m p i n g capacity of the mixer V . O n dividing Eqs. (68) P

a n d (22), the ratio of the volumetric flow rate at the vessel b o t t o m V a n d the bt

quantity V is obtained for the given system. F o r the three-blade propeller


p

mixer (p = D\ this relation is


Vu/K = 0.319(Ζ)/Γ)-°· (Λ /Γ)-°· 95
2
27
( R e > 1.0 Χ 10 ) 4
(69)
14. Flow and Turbulence in V e s s e l s with Axial Impellers 177

a n d for the six-pitched-blade paddle mixer (a = 45°) the relation is


KJVp = 0A\7(D/T)- *(h /T)-°
l2
2
52
( R e > 1.0 Χ 10 ) 4
(70)
These relations indicate that the volumetric flow rate at the vessel b o t t o m
can be smaller or larger t h a n the p u m p i n g capacity of the impeller, depending
o n the e q u i p m e n t geometry. T h e occurrence of induced flow explains how
the flow rate at the b o t t o m can be greater t h a n the impeller p u m p i n g rate V .
p

This agrees with the preceding data given for the field of streamlines. T h e
existence of induced flow u n d e r the impeller plane can be seen in Figs.
1 0 - 1 2 . T h e streamlines of induced flow in Figs. 10 - 1 2 are those that d o not
pass through the impeller blades. T h e k n o w n streamline fields at the b o t t o m ,
especially below t h e area S , show the effect of Reynolds n u m b e r o n Y a n d
3

i?! in Eq. (67) a n d m u s t be t a k e n into consideration in calculating the


dimensionless axial force that results from the use of axial-flow impellers.
T h e effect of R e m u s t also be t a k e n into account in calculating the total
volumetric flow rate at the vessel b o t t o m .

IX. Flow at the Vessel W a l l

Liquid flow at the vessel wall exhibits typical behavior. But this flow is
affected by the streamline field in the b o t t o m region from which the liquid
enters the space near the wall a n d by the conditions in t h e wall region (A2,
F9). These conditions include the effect of a solid b o u n d a r y o n adjacent
liquid, significant liquid force, a n d the existence of a n oscillating liquid
b o u n d a r y layer. However, it is n o t possible here to assume the validity of the
a s s u m p t i o n C I (Section II), n a m e l y that of axial symmetry of the flow, since
there is a significant effect of radial baffles situated at the walls.
Figure 27 shows a n experimentally based idealized flow pattern at the wall
between two neighboring radial baffles. In Table VII the distribution of the
dimensionless m e a n velocity vector Wis given together with the c o m p o n e n t
W in a plane tangential to the cylindrical wall at selected points between two
x

adjacent baffles in a vessel with Τ = 290 m m . Figure 28 presents plots of the


radial profiles of the dimensionless absolute value of the m e a n local velocity
vector W&\ several points o n the vessel wall determined experimentally by
measuring the radial profiles of the total a n d static pressures in this region
(F9).
T h e radial profiles of the m e a n local velocity vector at the vessel wall are
the s a m e as the velocity profiles for liquid flow next to a solid body. Very
close to the wall a velocity gradient exists, a n d at 3 to 6 m m from the vessel
wall flow velocity can be considered constant; i.e., the velocity field is not
affected by the neighboring solid body. But this velocity field changes from
point to point, especially in the u p p e r part of the vessel (see Table VII). T h e
178 Ivan Fort

FIG. 27. Flow pattern at the vessel wall between two adjacent baffles. Propeller (/? = D\
D/TG <1/5; 1/3 ), h /T=
2 1/3; y, = - 6 ° , y = - 2 2 % y - 7 5 ° , y = 2 1 7 ° .
2 3 4

absolute value of the m e a n local velocity vector decreases toward the liquid
surface; this is also confirmed by the shape of the streamlines observed for the
m i x e d system (see Figs. 1 0 - 1 2 ) . T h e velocity in the lower half at the wall a n d
its direction are practically constant; this is also in agreement with the veloc­
ity profile observed for the fluid inside the vessel. T h e shape of velocity
profiles at the vessel wall changes along the wall; at the location where the
velocity is greater the slope of the velocity profile is also greater.
Fluid velocities are directly proportional to the first power of the rotational
speed of the impeller. T h e velocity field at the baffled vessel wall w h e n a n
axial-flow impeller is used can be considered two-dimensional as the value of
the angle φ (see Table VII) between the m e a n local velocity vector a n d the
plane tangential to the cylindrical vessel wall in the given point determined
by individual measuring points is n o t greater t h a n 25 °, a n d in the lower half
of the wall is less t h a n 10°. Only at a few points (locations 12 a n d 1 4 - 1 9 ) is
the value of this angle significantly greater. In the region of positions 12 a n d
14 two streams are in contact (upward a n d d o w n w a r d ) ; in the region of
position 19 at the chosen direction of rotation the effect of a n eddy appears
near t h e baffle. T h e flow pattern at t h e wall between t w o neighboring baffles
(see Fig. 27) is given by the m e a n p a t h of a liquid particle entering the region
14. Flow and Turbulence in V e s s e l s with Axial Impellers 179

Table VII

Velocity Distribution along the Vessel Wall


between T w o Adjacent Baffles' '*
1

Position Φ
in Fig. 27 (deg)< W

1 2.0 0.139 0.139


2 8.0 0.145 0.142
3 6.5 0.156 0.152
4 6.0 0.146 0.143
5 6.5 0.153 0.150
6 6.5 0.112 0.109
7 6.5 0.133 0.130
8 6.0 0.122 0.120
9 6.5 0.115 0.113
10 5.5 0.111 0.109
11 7.5 0.077 0.075
12 66.0 0.083 0.027
13 8.0 0.078 0.075
14 8.0 0.045 0.043
15 158.0 0.043 0.036
16 25.0 0.049 0.042
17 13.0 0.025 0.021
18 167.5 0.034 0.028
19 42.5 0.040 0.023
20 168.5 0.014 0.011
21 167.5 0.019 0.016

a
From Fort et al (F9).
b
Propeller (p = D)\ D/T=\\ h /T = ^ R e >
2

5.0 Χ 10 ; four radial baffles, b/T = 0.1.


4

c
Angle between the vector of the local velocity and
the tangential plane o n the vessel wall.

at the wall. According to this pattern, the flow at the wall can be divided into
t w o parts. In t h e lower half the flow is axial (i.e., practically vertical) u p w a r d
with a small effect of mixer rotation (it is eliminated by t h e baffles). In the
u p p e r half of the wall, the flow pattern is m o r e complicated because of the
effect of the practically still l i q u i d - g a s interface a n d t h a t of negative pressure
b e h i n d t h e baffle. In t h e u p p e r a n d lower parts of t h e cylindrical wall, in
agreement with the velocity field for the whole system, the absolute values of
the velocity vector differ significantly. T h e value of the m e a n velocity vector
Win Table VII is markedly lower in t h e u p p e r half of the t a n k t h a n in the
lower half ( n u m b e r s of positions less t h a n 13) because t h e liquid flow direc­
tion has already been reversed by the suction effect of the impeller, i.e.,
simultaneously d o w n w a r d a n d in line with t h e t a n k axis. This is o n e of the
180 Ivan Fort

0.36
I I
(a)
ο ο
0.32

0.28
"8
08 • •
-

Ι
0.16
I
(*)

β 88-
W 0.12 - •ο
0-08

ΊΟ
0-W 1
I ο
ο ο
0-10 _ ο •
• •
0-06 β· I i
U
r, mm
6

FIG. 28. Radial profile of dimensionless absolute value of local velocity vector at the vessel
wall. Six-pitched-blade impeller ( a = 45°);D/T= \\h /T= \\ four radial baffles, b/T= 0.1; O:
2

R e = 6.3 Χ 10 ; · : R e = 9.4 Χ 10 . (a) Position 2 (Fig. 27), (b) position 8 (Fig. 27), and (c)
4 4

position 14 (Fig. 27).

explanations for the increase in flow velocity between the m e a s u r e m e n t


points 2 1 , 18, a n d 15, i.e., contrary to the flow direction in the lower part of
the vessel.

X. Relation between Total Fluid Circulation and


Homogenization Rate in a Stirred Tank
T h e total flow rate in a vessel, V is the impeller p u m p i n g rate V p i the
c p u s

fluid entrained by V [see Eqs. (1) - (4)]. Rate V accounts for all circulation
E c

loops in the entire liquid volume. O n e such circulation loop consists of a


p a t h followed by a chosen liquid particle between one reversal of the vertical
c o m p o n e n t of m o t i o n of the particle a n d the second of two m o r e such
reversals (P2).
In a miscible liquid, a sample exists with v o l u m e Δ V having a n initial
concentration c of a dissolved solid c o m p o n e n t t h a t is different from the
0

initial concentration c in the vessel. Such a sample should be small c o m ­


0

pared to the charge v o l u m e b u t several orders of magnitude larger t h a n the


dimensions of molecules. This sample, whose physical properties d o not
differ significantly from those of the initial fluid in the vessel, is added to the
fluid in the vessel. If the intensity of convective flow (mass flux) is sufficient
14. Flow and Turbulence in V e s s e l s with Axial Impellers 181

a n d the initial mass of the added sample is small, it moves with the same
velocity as the fluid that s u r r o u n d s it. T h u s we can consider this sample to be
a n arbitrary particle of the liquid that is circulating in the vessel. After adding
the sample, mass transfer of the dissolved substance takes place between the
sample a n d the fluid in the vessel. A s s u m e that inside the sample the turbu­
lence intensity is constant a n d that an unsteady turbulent diffusion of the
dissolved substance takes place there. Outside the sample, turbulent mass
transfer also occurs. T h u s we assume t h a t inside the sample a m o t i o n of
clusters of solution molecules takes places so that the gradient of the m e a n
velocity of the liquid m e d i u m inside the v o l u m e A F h a s a negligible value.
T o be able to investigate m o r e closely the relation between the total circu­
lation a n d the rate of homogenization of the dissolved substance in the
mixed charge, let us consider a coordinate system firmly b o u n d with the
sample. Further, several R assumptions are m a d e below a b o u t the sample
a n d the concentration of the dissolved m a t t e r introduced into the system
with the sample (F8):

R l . T h e sample occupies a spherical space Δ V of constant radius R.


R2. T h e m o t i o n of the dissolved m a t t e r inside the sample can be described
in t e r m s of constant turbulent diffusion; i.e., the coefficient of eddy diffusiv-
ity in the sample v o l u m e is n o t a function of position, time, a n d concentra­
tion; also, other effects are negligible.
R3. T h e concentration distribution of the dissolved m a t t e r in the sample
a n d o n its surface is spherically uniform.
R4. At the beginning (i.e., at the m o m e n t w h e n the sample is added to the
vessel) the sample is h o m o g e n e o u s with respect to the concentration of the
dissolved matter; this is true for the liquid in the stirred vessel as well.
R5. At a very long t i m e after the addition of the sample, concentrations of
dissolved m a t t e r in the vessel a n d in the sample d o n o t differ.
R6. T h e rate of mass transfer of the dissolved c o m p o n e n t between the
sample a n d t h e liquid in the vessel is proportional t o the concentration
difference o n the sample surface a n d the concentration averaged in all of the
liquid in the vessel. T h e mass transfer coefficient is n o t a function of position,
time, a n d concentration.

N o w four "definition relations" are given:

Total circulation time τ is the t i m e interval between two consecutive


α

up-to-down reversals ( m a x i m a ) of the sample, or two consecutive down-to-


u p reversals ( m i n i m a ) :

(71)
182 Ivan Fort

where w is the absolute value (modulus) of the m e a n velocity vector in the


p a t h a n d / is the path length of sample m o v e m e n t . T h e integral in Eq. (71) is
curvilinear along the path of the sample between two points G a n d G in t i+l

which the vertical projection of the p a t h passes through two consecutive


m a x i m a or two consecutive m i n i m a . T h e m e a n t i m e of total circulation can
be t h e n written in the form

t c - j£ t d (72)

where Μ is the n u m b e r of consecutive m a x i m a or m i n i m a .


The degree of homogeneity of the vessel contents is given by

m m i^LZSK ( 7 3 a )

^0 C
k

where

(c(t)) = ψ j c(V,t)dY (73b)

the initial concentration is defined as


c= 0 (c(t = 0)) (73c)

a n d the final concentration of the m a t t e r in the charge is

(73d)
Q u a n t i t y c(T 1) is the instantaneous concentration of the dissolved m a t t e r
9

(averaged over the liquid v o l u m e V, which does n o t include the sample).


Degree of sample homogeneity is given by the same quantities defined in
the preceding paragraph for the sample, i.e., by

<c'(/)>
s
-£vl '( >» AV
c r dv ( ? 4 a )

c' = (c'(t = 0))


0 (74b)
c' k = (c^-^oo)) (74c)

c*(t\ = < ('»-<* c/ ( 7 4 d )

^ _
r'
Co — r'
k c

Homogenization time t is the interval between the t i m e w h e n the sample


h

is added to the charge a n d the t i m e when the sample reaches the homogeni­
zation degree expressed by a specified value of C' : h

6 = t h and C ' ( / ) = C'(0) = C i


h (75)
14. Flow and Turbulence in V e s s e l s with Axial Impellers 183

T h e n on t h e basis of t h e definitions in Eqs. (73) a n d (74), a material balance


can be written for the dissolved m a t t e r

(c(t)) V +(c'(t)) AV = c V + c AV=


0 0 c Y + c' AV
k k (76)

where t h e integration region Δ V is the region which the added sample occu­
pies at a n y m o m e n t a n d whose v o l u m e a n d shape are constant according t o
a s s u m p t i o n R l . F r o m Eq. (76)

(c(t)) - c _ (c'(t))
k - c'k (77a)
C
0~ kC C
0~~ k C

and

C(t) = C'(t) (77b)

T h u s , t h e degree of homogeneity C(t) of the liquid in the vessel equals t h e


degree of h o m o g e n i z a t i o n C'(t) of t h e sample at t i m e /. T h e homogenization
t i m e can also be d e t e r m i n e d on t h e basis of the concentration change in the
m o v i n g sample.
By use of a s s u m p t i o n s R 1 - R 3 , concentration homogenization can be
described by t h e differential e q u a t i o n

2iM-/),[i!fi»] (,>0,0<r< R) (78)


T h e b o u n d a r y conditions in the center a n d o n the surface of the sample of
radius R are

q ^ - 0 , c"(0,„*»
dr
(79)
D c + k [c\K
c t) - (dt))] = 0

T h e initial condition based o n a s s u m p t i o n R4 has the form

c'(r,0) = c' 0 (80)


W i t h regard to Eq. (76) a n d a s s u m p t i o n R5, the relation for the quantity
(c(t)) in the second e q u a t i o n of Eq. (79) can be written

(c(t)) =c' k + AV/V[c' - k (c'(t))] (81)

W h e n this relation is substituted into the corresponding relation of the


system of Eq. (79) a n d t h e t e r m AV/Vis neglected in c o m p a r i s o n t o c'(R, t),
a new q u a n t i t y can be defined

X(r,t) = c'(r,t)-c k (82)


184 Ivan Fort

By differentiation of Eq. (78) with the initial a n d b o u n d a r y conditions [Eqs.


(79) a n d (80)], it can be written in the form

S[rXjr 1)] 9 _ fd (rX(r,t)]]


2

dt
c
l dr 2
J
(t>0;0<r<R) (83)

(84)
D c
d
- ^ + k [X(R,t)]
c = 0

X(r, 0) = c' - o c'k (85)

T h e m e a n integral value of the function X(r, t), d u e to the sample v o l u m e


a n d with regard t o Eqs. (77b) a n d (82) equals (F8)

where
L ,s 6Βΐ /μ}(μ
2 2
+ B i - Bi)
2
(87)

and

Bi - k R/D c c (88)

F o r the given degree of homogeneity C , Eq. (86) determines the homogeni- h

zation t i m e θ = t . T h e relation between the homogenization t i m e a n d the


h

m e a n t i m e of total circulation can be expressed if several further partial steps


are taken; these lead to a simplification of Eq. (86). First, a quantity J is
defined, which represents the ratio of the homogenization t i m e a n d m e a n
circulation t i m e .

J=e /x c (89)

Next, only the first t e r m on the right-hand side of Eq. (86) is considered,
which is correct for / > 0 a n d Bi —»0. O n t h e basis of a study by Prochazka
a n d L a n d a u (P3) values D are in the range 1 0 ~ t o 1 0 " m / s e c , a n d the
c
4 2 2

quantity R can be considered ~ 10~ m according t o L a n d a u a n d Prochazka


2

( L I ) . O n the basis of several types of dimensionless equations from t h e b o o k


by Sterbacek a n d T a u s k (S4), the mass transfer coefficient k can be consid­ c

ered to be in the range 10~ t o 10~ m/sec. T h u s , by simple calculations the


2 4

validity of the a s s u m p t i o n concerning the behavior of the quantity Bi can be


14. Flow and Turbulence in V e s s e l s with Axial Impellers 185

verified. F r o m the sample a n d e q u i p m e n t geometry, this relation applies:

R = ωΤ ,
2 2
where ω ~ (AF/r) ' 2 3

If Eq. (86) is arranged o n the basis of these considerations, the following


relation is obtained, using / = 1:
C(0) = L , εχρ[-μ (ΰ /ωΤ )
2
0
2
J T]C (90)

By dimensional analysis, the relation for eddy diffusivity m a y be obtained:


D ~ (T/DY ND
c
2
[Re > 1.0 Χ 10 ] 4
(90a)
which characterizes the transfer of dissolved m a t t e r inside the sample. This
d e p e n d e n c e was reported earlier (P3), with mass transfer in the whole charge
expressed by use of the mass transfer coefficient. 5

It was found experimentally by Fort et al ( F 8 , F14) a n d Porcelli a n d M a r r


(P2) that the m e a n t i m e of total circulation depends o n the quantities iVand
T/D as shown in

τ ~ (l/N)(T/Dy
€ [Re > 1.0 Χ 10 ] 4
(90b)

Finally, by substituting Eqs. (90a) a n d (90b) into Eq. (90), we obtain


0/ T c = J ==
d(T/D)K
log[L,/C(0)] [Re > 1.0 Χ 10 ] 4
(91)

F r o m this it can be concluded that the homogenization time θ is directly


proportional to the m e a n t i m e of total circulation T , with the proportional­
c

ity coefficient being a function of the e q u i p m e n t geometry a n d of the re­


quired degree of homogeneity C(0) for the liquid system.
Table VIII gives values of the e x p o n e n t κ a n d the constant d in Eq. (91) for
three types of axial impellers. T h e reported values were d e t e r m i n e d with an
estimate of their standard deviations obtained from published results o n
total circulation (F8) a n d t i m e dependencies of the homogenization of mis-
cible liquids ( K 6 , P3). In Table VIII the values of d a n d κ for the impellers
studied roughly agree. Such agreement can be explained by the assumptions
m a d e , which d o n o t correspond to the actual conditions in the system being
mixed. O n e example is the m a r k e d variance in the rate of dissipation of
mechanical energy at different points in the vessel; a n o t h e r is the obviously
incorrect a s s u m p t i o n that the coefficient of eddy diffusivity D has a constant c

value t h r o u g h o u t the system. Nevertheless, for the agitated system consid­


ered as a whole, the relation between the m e a n t i m e of total circulation a n d
the homogenization t i m e is u n a m b i g u o u s l y determined if the general as-

5
In Prochazka and Landau (P3) the mass transfer coefficient k is proportional to (NDY,
c

where the exponentλ ~ 1.


186 Ivan Fort

Table VIII

Dependence of the H o m o g e n i z a t i o n T i m e o n the M e a n


T i m e o f Total Circulation"'*^

Impeller d <7d
κ

Propeller (p = D) 1.11 0.253 0.35 0.12


Three-pitched-blade paddle 1.39 0.202 0.28 0.12
(a = 2 4 ° )
Six-pitched-blade paddle 1.12 0.221 0.25 0.08
(a = 45°)

* From Fort et al (F8).


b
h /T=i;
2 R e > 1.0 Χ 10 ; four radial baffles, b/T= 0.1.
4

c
d is a constant in Eq. (91); κ is the exponent of (T/D) in Eq.
(91); σ signifies standard deviation of respective values.

s u m p t i o n s m a d e concerning the flow of the mixed charge are fulfilled. T h e n


the h o m o g e n i z a t i o n t i m e is a multiple of the total circulation time; also, the
larger the ratio T/D the smaller the degree of homogeneity possible in a given
t i m e . T h e homogenization t i m e of miscible liquids in the whole system is
t h u s d e t e r m i n e d for t h e given degree of homogeneity C o n the basis of the
k n o w n t i m e of total circulation of t h e charge a n d t h e e q u i p m e n t geometry.
T h e value of p a r a m e t e r L is d e t e r m i n e d by the initial condition, Eq. (80)
x

lim L = 1 x (92)
*—ο

But this state concerns only the liquid next to the added sample. T h e condi­
tion at a n o t h e r point in a n industrial-size tank, in which the quantity x can c

reach quite a high value, significantly differs from that at the point where the
sample is added. T h e value of p a r a m e t e r L at a n arbitrary point in the t a n k
l G

(characterized by the point G) can be obtained from the t i m e dependence of


the homogenization process for the considered point. This can be expressed
o n the average by the relations 6

C (d)=l
G (0<? C G ) (93a)

C (d) = L
G XG exp[-J(l/d )(D/Tr] x (93b)
where x is the m e a n t i m e of total circulation between the point where the
CG

sample is added a n d the chosen point G. F r o m t h e system of Eqs. (91) a n d


(93) t h e value of p a r a m e t e r L can be obtained
lG

LlG = exp[(T C G / T ) / ( l / ^ )(D/TT]


c (94)

6
T h e value of the constant d =
x 2303d.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 187

F o r a n actual t i m e at different points in the liquid, the same degree of


h o m o g e n e i t y will n o t be attained b u t there will be a certain t i m e lag resulting
from t h e so-called rise t i m e , i.e., the part of the total circulation t i m e during
which t h e circulating sample travels from the point of its addition to the
proximity of t h e p o i n t being considered. T h e m a x i m u m value of L is then 1G

L Xmax = txp[(l/d )/(D/Tr]


x (95)

N o w h o m o g e n i z a t i o n in all t h e liquid in a vessel is considered, for which


t h e simplifying a s s u m p t i o n s C 1 - C 5 (Section II) a n d R 1 - R 6 (Section X )
were introduced. T h e t i m e d e p e n d e n c e of the degree of homogenization at
a n arbitrary p o i n t in the liquid is given by t h e system of Eqs. (93). F o r the
average degree of homogeneity of all the liquid in a vessel, it is necessary to
k n o w the m e a n value of the q u a n t i t y C (6) t h r o u g h o u t the entire v o l u m e
G

a n d from it t o calculate from Eq. (93b) t h e corresponding value of the


q u a n t i t y L . But this m e a n s that o n e m u s t k n o w the values of the quantities
l G

m e n t i o n e d that d e p e n d o n t h e location in t h e system at a given t i m e . By


applying t h e m o d e l introduced earlier for convective transfer of the concen­
tration pulse in the charge t h r o u g h t h e loops of total circulation, instead of
t h e average value of t h e concentration t h r o u g h o u t the volume, it is possible
t o consider the concentration m e a n value for the m e a n t i m e over total
circulation. T a k i n g into a c c o u n t Eqs. (77b) a n d (93b), t h e degree of h o m o g e ­
neity of the whole mixed system can be given by

C(0) = i - Γ Q(0, T ) dx
~c Jo C c

=
iio ' ^' H^) ] '
CLiG( cG)e J K d cG (96)

T h e n substituting from Eq. (94) a n d integrating with the defined value of


I i m a x w e obtain
C(0) = d (T/Dr(L
x Xmax - 1) exp[-J(l/d )(D/Ty]x (97)

T h i s equation can be rearranged t o the form of Eq. (91), or


J=d(T/D)« \og[d (T/Dy(L
x Xmax - 1)/C(0)] ( R e > 1.0 Χ 10 ) 4
(98)

F r o m Eq. (98) the value of p a r a m e t e r L in Eq. (90) averaged for the whole
x

system, b u t excluding t h e sample v o l u m e AV, equals

(L )
x = d (T/Dr(L
x Xmax - 1) ( R e > 1.0 Χ 10 ) 4
(99)

O n t h e basis of this relation, t h e general solution of t h e diffusion equation for


unsteady-state mass transfer from the sample t o the surrounding liquid m u s t
be modified to be applicable t o t h e industrial size of t h e liquid system in
188 Ivan Fort

which the convective mass transfer takes place. This m e a n s that the value of
quantity L [see Eq. (92)] h a d to be changed to t h a t given by Eq. (99), which
x

already includes the effect of e q u i p m e n t geometry a n d implicitly the circula-


tion inside it. Also, in accordance with results published by Prochazka a n d
L a n d a u (P3), the value of L for this unit c o m e s o u t t o equal n o t o n e b u t
x

t w o , which is significantly different. Experiments have n o t yet been d o n e in


7

a large industrial tank; t h u s these considerations c a n n o t be proved. But the


assumptions used are considered reasonable a n d invoked with s o m e confi-
dence.

XL M e c h a n i s m of H e a t Transfer between the Liquid and


t h e Vessel Bottom or W a l l

T h e flow of liquid directed by an axial impeller toward the b o t t o m a n d the


wall of a cylindrical vessel can be expressed as a sequence of impact a n d wall
turbulent flows (see Fig. 29) ( F l 3) as determined below. A n axially symmet-
rical jet, shown by I in Fig. 29, results from the rotation of the impeller
blades; this is also described by Eqs. (8) a n d (9) or Eqs. (14) a n d (16). This free
jet begins at the lower edges of the impeller blades; then it changes into a
so-called impact jet (II) with a change in its direction, concurrent with a
modification of its character. This is a c c o m p a n i e d by the formation of a
conical region in the vicinity of the vessel axis where the flow is practically
suppressed, called a dead region. T h e b o t t o m of the cone has radius r . After sp

the change of direction of the stream at the b o t t o m it is possible to consider


the so-called radial wall jet (III), for which the velocity profile is affected by
the shape of the profile in the preceding free jet a n d by the existence of the
b o t t o m solid wall, which forms a b o u n d a r y for the stream. T h e size of the
b o t t o m region through which some of the streams flow depends first o n the
distance of the impeller from the b o t t o m , next on its relative size D/T, a n d
t h e n o n the value of the Reynolds n u m b e r . N o w the action of the sidewall
further t u r n s the impact jet (IV). Of course, the character of the stream is
again governed by the presence of a solid (now side) wall a n d by the radial
velocity profile (see Figs. 8 a n d 9). T h e source of this profile can be found in
the preceding history of the impeller effluent stream, especially in the shape
of the profile at the exit from the blades of the rotating impeller. T h u s
another, this t i m e axial, a n n u l a r wall jet forms ( V ) (see also Figs. 27 a n d 28).
F o r consideration of the m e c h a n i s m of heat transfer, the flowing liquid
b o u n d e d by the b o t t o m surface is divided into two sectors: that in the vicinity
of the stagnant points (within coordinate r ) , where the character of the flow
s p

7
The results given in Table VIII were obtained for the given value of parameter ( L , ) , viz.
< L , ) = 2.0.
14. Flow and Turbulence in V e s s e l s with Axial Impellers 189

FIG. 2 9 . Flow model below the horizontal plane of symmetry of an axial-flow impeller in a
cylindrical vessel with radial baffles. Identification of "jets" in agitated liquid: I, free axially
symmetrical jet; II, axially symmetrical impact jet at the bottom; III, axially symmetrical radial
wall jet at the bottom; IV, axially symmetrical impact jet at the wall; V, axially symmetrical axial
wall jet at the wall.

is n o t purely turbulent, a n d t h a t b e y o n d the ring of stagnant points, where


purely t u r b u l e n t transfer obtains ( D l , D2). Simultaneously, in t h a t part of
t h e b o t t o m where the radial coordinate is less t h a n the value r , there is a
sp

considerable effect of natural convection. T h e size of this region (also the


location of the points r ) depends considerably on the relative size of the
s p

impeller. It also depends o n the distance of the impeller from the b o t t o m a n d


o n the flow conditions characterized by the impeller Reynolds n u m b e r (see
the discussion of the effect of k i n e m a t i c viscosity o n the flow pattern at the
b o t t o m in Section VII).
In the b o t t o m region, where the heat transfer m e c h a n i s m can be consid-
ered fully turbulent, it can be assumed t h a t the axial profile is fully developed
in the vicinity of the wall. F r o m the results of studies of wall jets by A b r a m o -
vich (Al), it was found t h a t the " o n e - s e v e n t h " law resulted for this region,
with velocity profiles similar to the transformations (14a) a n d (14b).
It was found from experiments by Fort et al ( F l 1, F l 3) a n d K u d r n a et al
( K 4 ) that the turbulence intensity of the flow between the rotating axial
impeller a n d the b o t t o m was practically i n d e p e n d e n t of the relative size of
the impeller a n d the vessel, i.e., D/T (see Table I X ) . This turbulence can also
be assumed to exist in the radial wall stream at the vessel b o t t o m , whose
k i n e m a t i c description fully corresponds to that of Eqs. (14a), (14b), a n d (17)
190 Ivan Fort

Table IX

Average Turbulence Intensity in the Stream


between the Axial-Flow Impeller and the Vessel
Bottom 0

Place D/T [(w y> /w]


2 2
st

Below the impeller 1 33.6


5

i 30.0
1 36.6
3
At the vessel bottom 1 32.8
5
1
4 36.2
1 29.0
3

From Fort ( F l 1); Fort et al (F13); Kudrna et al


a

(K4).

for t h e free turbulent j e t at t h e discharge from t h e blades of a rotating


impeller.
F o r t h e rate of heat transfer between t h e b o t t o m surface a n d t h e liquid
where r > r ^ , with a n assumption of constant t e m p e r a t u r e in b o t h regions,
we c a n write t h e dimensionless relation for heat transfer in t h e turbulent
b o u n d a r y layer ( K l )

Nu (r) = h {r){r
T hx - r )/k
sp = 0.0292Re?- Pr* 8

(r>r ,Re >


9 T 1.0 Χ 1 0 ) 5
(100)
where t h e Reynolds n u m b e r R e is defined as r

Re (r) = i v ( r - r ) / v
r c s p (101)

H e r e the characteristic velocity is the m a x i m u m velocity vv^ of the impeller c

discharge jet [see Eq. (17)]; this is the quantity used t o calculate the intensity
of turbulence of the subject stream. This concept is in agreement with t h e
a s s u m p t i o n (see Section I V ) regarding t h e preservation of t h e turbulent
structure of streams which are mutually interconnected.
Since t h e quantity is directly proportional t o t h e peripheral velocity
c

of the blades of the rotating impeller (F10) (see t h e fifth c o l u m n in Table I)


a n d t h e quantity (r — r ^ ) is also directly proportional t o t h e vessel diameter
T, Eq. (100) c a n be written in t h e form
h (r)T/ks
ht Nu(r) = const(r)(r/Z)) Re°- Pr" a8 8

(r>r s p , R e > 1.0 Χ 1 0 ) 4


(102)
which holds for t h e given relative distance of the impeller above t h e vessel
b o t t o m h IT. T h e value of the exponent of the Prandtl n u m b e r Pr, based o n
2
14. Flow and Turbulence in V e s s e l s with Axial Impellers 191

experiments o n liquid mixing in t u r b u l e n t flow ( H 3 , U l ) , is:

σ = 0.45 (103)

W i t h regard to heat transfer, the region next t o the vessel cylindrical wall
can be considered a c o n t i n u a t i o n of the turbulent region at the b o t t o m . In
particular, with regard to the additional intensity of turbulence of the stream
resulting from its reversal at t h e b o t t o m of the cylindrical wall, it is possible to
use as the characteristic velocity the quantity νν for calculation of the heat ω c

transfer coefficient at turbulent flow along the s m o o t h plate in Eq. (100).


Also, the origin of the b o u n d a r y layer can again be situated (as in the case of
flow at t h e b o t t o m ) at the point r . F o r calculation of the m e a n heat transfer
sp

coefficient for the whole vessel wall, / j , the exponents of the Reynolds a n d
w s t

Prandtl n u m b e r s can be considered identical to those given in the relation for


calculation of the local Nusselt n u m b e r , Eq. (100). But for calculation of the
m e a n heat transfer coefficient for the whole b o t t o m , h , the values of these h >st

Table X

Heat Transfer between Liquid and the Vessel Wall and Bottom

A. Local Heat Transfer Coefficient at the Vessel Bottom*'*

r/T D/T r /T
sp R e exponent

0.100 1
5 0.050 0.50
0.100 1
4 0.080 0.40
0.100 i 0.090 0.30
0.375 1
5 0.050 0.85
0.375 i 0.080 0.85
0.375 i 0.090 0.90

B. Overall Heat Transfer Coefficient'^

Impeller D/T R e exponent

At the vessel bottom


Propeller (p = D) 0.40 0.58
Two-pitched-blade paddle ( a = 4 5 ° ) 0.40 0.63
At the vessel wall
Propeller (p = D) 0.40 0.77
Two-pitched-blade paddle (a = 45 °) 0.40 0.78

From Fort et al (F13).


a

Six-pitched-blade paddle (a = 45°); h /T=


b
1/4; four radial
2

baffles, b/T= 0 . 1 ; R e e <8.0 Χ 10 ; 2.0 X 10 >. 3 5

From Kupcik (K5).


c

h /T
d
2 = b four radial baffles, b/T=0.1; Ree<8.0X10 ; 3

2.0 X 10 >. 5
192 Ivan Fort

exponents will vary; for instance, in the b o t t o m region where the radial
coordinate r = r , the m e c h a n i s m of heat transfer will correspond to that for
sp

the l a m i n a r b o u n d a r y layer, where the value of the e x p o n e n t for both R e a n d


Pr is ±.
Experimental studies of the local (F13) as well as the m e a n (K5) heat
transfer coefficient for the b o t t o m a n d the vessel wall provide the results
given in Table X . F r o m this table it is obvious that there is good agreement
between the exponents of R e for the propellar a n d the pitched-blade paddle.
T h e same R e e x p o n e n t produces the same effect of the turbulent velocity
field o n the rate of heat transfer between the vessel wall a n d the mixed liquid.
T h e heat transfer m e c h a n i s m at the vessel b o t t o m is m u c h m o r e complicated
t h a n that at the vessel cylindrical wall, where at R e > 1.0 Χ 1 0 the t u r b u ­
4

lence can be considered fully developed. O n the other hand, the conditions at
the vessel b o t t o m are quite d e p e n d e n t o n the size a n d location of the axial
impeller because of the relatively complicated flow pattern in this region.
Therefore these conditions m u s t be taken into account when determining
b o t h the local values a n d the m e a n values of the heat transfer coefficient.
Moreover, the situation at the b o t t o m can be complicated by the location of
a n inlet or outlet of a reaction mixture (for a mixed h o m o g e n e o u s reactor),
which is usually situated at the axis of the b o t t o m symmetry; then the
velocity field in its vicinity could be significantly affected. It m a y be con­
cluded ( K 5 , S3, U 1 ) that the exponent of the Reynolds n u m b e r can be taken
equal to 0.6 in the case of the m e a n heat transfer coefficient between the
b o t t o m a n d the charge (see Table X ) , a n d the e x p o n e n t of the Prandtl
n u m b e r can be taken as 0.45 as given by Eq. (103).

X I I . Conclusions

It was d e m o n s t r a t e d that for agitated liquid systems the h y d r o d y n a m i c


characteristics can be modeled a n d the results can be confirmed by the
considerable experimental data in the literature. T h e study, although exten­
sive, dealt only with axial impellers a n d agitation of low-viscosity fluids in
baffled tanks. T h e strictly h y d r o d y n a m i c portion dealt not only with the
m e a n velocity field b u t also with the fluctuating velocities in critical parts of
the vessel, e.g., in the vicinity of the impeller a n d next to the vessel sidewalls
and bottom.
T h e h y d r o d y n a m i c description included m a i n flow characteristics in rela­
tion to the impeller, that is, the p u m p i n g capacity of the impeller, induced
volumetric flow, a n d flow at the b o t t o m of the vessel.
T h e analysis of the circulation of agitated liquids for the systems investi­
gated was applied to two process results: the course of homogenization of
14. Flow and Turbulence in V e s s e l s with Axial Impellers 193

miscible liquids, a n d heat transfer. F o r the latter, heat transfer between the
vessel flow a n d b o t h the vessel sidewalls a n d b o t t o m were considered.
T h e chapter provided a comprehensive view of the hydrodynamics within
a vessel agitated with axial impellers a n d then applications to two well-stud­
ied process results: homogenization (blending) a n d heat transfer. This work
suggests h o w such a n approach might be extended to other types of agitated
systems a n d applied to less studied b u t i m p o r t a n t critical operations such as
solids suspension, crystallization, dissolution, extraction (two-phase), a n d
the reaction of b o t h h o m o g e n e o u s a n d heterogeneous systems. Specific goals
of such a detailed study, b e y o n d t h e description a n d understanding of sys­
tems, are the optimization of designs a n d the diagnosis of unsatisfactory
operation.

List of Symbols

b Baffle width (radial), m


c Concentration, k g / m 3

D Impeller diameter, m
D c
Eddy diffusivity, m / s e c
2

d h
H u b diameter, m
di Diameter of ith part of the force action o f the mixed charge o n the vessel bottom, m
f Force, kg m / s e c
2

G Specification o f the point in the mixed charge


g Acceleration due to gravity, m / s e c
2

Η Fluid depth in the vessel, m


h Coefficient of heat transfer, vessel liquid, ( J / m sec K); blade width of impeller, m
2

h' Height of cylindrical part o f the draft tube, m


h\ Height of conical part of draft tube, m
h 2
Height of horizontal plane of symmetry of impeller blades above tank bottom, m;
height of lower base of draft tube above tank bottom, m
k Thermal conductivity of liquid, J/(m sec K)
Κ Mass transfer coefficient, m/sec
ι Path of sample circulation, m; angular coordinate, deg
Μ N u m b e r of elements in the set
Ν Frequency of revolutions o f the impeller, s e c - 1

η Propeller pitch ratio, p/D


"β N u m b e r of blades per impeller
Ρ Power drawn by impeller shaft, kg m / s e c ; probability
2 3

Ρ Propeller pitch, m (pitch is defined as the distance an impeller would advance for each
revolution); pressure, kg/(m s e c )
2

R Sample radius, m
r Radial distance from axis o f symmetry (coordinate), m
r 0
Radial distance from axis o f symmetry, where w = ^w , n axc m
S Cross-sectional area, m 2

Τ Tank or vessel diameter, m


Diameter o f lower (cylindrical) part o f draft tube, m
194 Ivan Fort

Γ 3 Diameter o f upper base of draft tube, m


t Time, sec
V Vessel volume, m 3

V m V o l u m e of rotor region, m 3

V bx V o l u m e of region at vessel bottom, m 3

Τ Vessel v o l u m e without sample volume, m 3

AV Sample volume, m 3

V c Total flow rate of mixed charge, m / s e c 3

V E Induced flow rate o f mixed charge, m / s e c 3

V P Pumping capacity or primary flow rate of mixer, m / s e c 3

w Local velocity, m/sec


X Relative concentration defined by Eq. (82), k g / m 3

ζ Axial (vertical) distance in vessel (coordinate), m


a Angle between plane o f an inclined impeller blade and plane in which the impeller
rotates, deg
β Initial angle between plane of a propeller blade and plane in which the propeller rotates,
deg
y Angle between local velocity vector and axial co-ordinate, deg
δ Apex angle of Pitot tube, deg
e Rate of energy dissipation in v o l u m e unit, kg/(m s e c ) 3

Θ Homogenization time, sec


μ R o o t o f characteristic Eq. (83)
ν Kinematic viscosity, m / s e c 2

ρ Density, k g / m
3

τ Circulation time o f liquid particle, sec


τ Ρ Mean time of primary circulation [see Eq. (6)], sec
φ Angle between local velocity vector and tangential plane, deg; lift coefficient of pro­
peller blade, m
ψ Stream function, m / s e c 3

DIMENSIONLESS NUMBERS
Bi Biot number, k R/D
c c

C Degree of homogeneity, [ (c(t)) — c ]/(c — c ) k 0 k

Po Power number, P/pN D 3 5

Po t Power output number, (P - e V )/pN D m m


3 S

Po z / Rate of energy dissipation in /th v o l u m e unit, € V /pN D i i


3 5

F a x Axial force, iJpN D* 2

J Ratio of homogenization time and m e a n total circulation time, 0 / t c

K bx Flow rate of vessel bottom, V /ND bt


3

K P Primary flow rate number, V /ND P


3

K c Total flow rate number, V ND C


3

K E Induced flow rate number, V /ND E


3

C T Dimensionless diameter, defined as [Eq. (36a)]


Ρ Dimensionless pressure, p/p(nDN) 2

Pr Prandtl number, cvp/k


R Dimensionless radial coordinate, r/T
Re Reynolds number, ND /v 2

Re r Reynolds number for radial wall jet at vessel bottom, (vv^ ) ( r — r ) / v c sp

W Dimensionless velocity, w/nDN


Ζ Dimensionless axial (vertical) coordinate, z/H
// h Hydraulic efficiency of impeller, P o / P o t
14. Flow and Turbulence in V e s s e l s with Axial Impellers 195

ξ Local drag coefficient


Ψ Dimensionless stream function, ψ/ND*

SUBSCRIPTS
ax Axial
b,bt Related to the vessel bottom
C Related to the total circulation
c Corresponding to the m a x i m u m value
exp Experimental
G Related to point G in the mixed charge
h Homogenization
Element o f a set
J Element of a set
k Final value
m Related to the impeller region, i.e., between S, and .Si, (see Fig. 1)
max Maximum
min Minimum
norm Normalized
Ρ Related to the pumping capacity o f the impeller
ρ Initial (space)
ρ Particle
rad Radial
s Related to the liquid surface
sp Related to the stagnant point
st Averaged over (cross-sectional) area
th Theoretical
χ Related to the plane tangential to the cylindrical surface formed by the inner surface o f
the vessel wall
w Related to the vessel wall
0 Initial (time)
00 Infinity (very far from the origin)
1 Related to v o l u m e V below the mixer
l

II Related to v o l u m e V above the mixer


n

1 Related to first turn o f the mixed charge at the vessel bottom


2 Related to flow at the mixed charge along the vessel bottom
3 Related to second turn o f the mixed charge at the vessel bottom

SUPERSCRIPTS AND OTHER SYMBOLS


Time-averaged value (mean time value)
' Fluctuation value
( ) Averaged value over the v o l u m e
() Limits of the variables in the interval, e.g., in Eq. (8a); ( m e a n s that the lower limit of the
interval does not contain the first value (or zero in the example) and ) m e a n s that the
upper limit o f the interval contains the second value (r in the example).
c

References
(A 1) Abramovich, G. N . "Theory of Turbulent Jets" (in Russian). Gos. Izd. Fix.-mat. Liter,
M o s c o w , 1960. (Translated into English by Scripta Technica, M I T Press, Cambridge,
Mass., 1963.)
(A2) Askew, W. S., and Beckmann, R. B., Ind. Eng. Chem. Process Des. Dev. 5 , 2 1 8 (1966).
196 Ivan Fort

(BI) Bird, R. B., Chem. Eng. Sci. 6, 123 (1956).


(B2) Bird, R. B., Stewart, W. F., and Lightfoot, Ε. N . "Transport Phenomena." Wiley, N e w
York, 1960.
(CI) Cooper, R. G., and Wolf, D . Can. J. Chem. Eng. 4 5 , 197 (1967).
(Dl) Donaldson, C. D . , and Snedeker, R. S. J. Fluid. Mech. 4 5 , 281 (1971).
(D2) Donaldson, C. D . , Snedeker, R. S., and Margolis, D . P., J. Fluid Mech. 4 5 , 4 7 7 (1971).
(Fl) Fort, I., Sci. Pap. Inst. Chem. Technol. Prague K l , 25 (1967).
(F2) Fort, I., and T o m e s , L., Coll. Czech. Chem. Commun. 3 2 , 3520 (1967).
(F3) Fort, I., Coll. Czech. Chem. Commun. 3 2 , 3 6 6 3 (1967).
(F4) Fort, I., and Sedlakova, V., Coll. Czech. Chem. Commun. 3 3 , 836 (1968).
(F5) Fort, I., Podivinska, J., and Baloun, R., Coll. Czech. Chem. Commun. 3 4 , 959 (1969).
(F6) Fort, I., Coll. Czech. Chem. Commun. 3 4 , 1094 (1969).
(F7) Fort, I., Eslamy, M., and Kosina, M., Coll. Czech. Chem. Commun. 3 4 , 3673 (1969).
(F8) Fort, I., Valesova, H., and Kudrna, V., Coll. Czech. Chem. Commun. 3 6 , 164 (1971).
(F9) Fort, I., Vlcek, J., Cink, M., and Hruby, M., Coll. Czech. Chem. Commun. 3 6 , 1546
(1971).
(F10) Fort, I., Neugebauer, R., and Pastyrikova, H., Coll. Czech. Chem. Commun. 3 6 , 1769
(1971).
( F l 1) Fort, I., Coll. Czech. Chem. Commun. 3 6 , 2 9 1 4 (1971).
(F12) Fort, I., Grackova, Z., and Koza, V., Coll. Czech. Chem. Commun. 37, 2371 (1972).
(F13) Fort, I., Hruby, M., and Mosna, P., Coll. Czech. Chem. Commun. 3 8 , 1737 (1973).
(F14) Fort, I., Koza, V., and Grackova, Z., Coll. Czech. Chem. Commun. 3 8 , 3 0 7 4 (1973).
(F15) Fort, I., Jaroch, O., and Hostalek, M., Coll. Czech. Chem. Commun. 4 2 , 3555 (1977).
(G1) Gray, J. B., "Mixing: Theory and Practice" (V. W. U h l and J. B. Gray, eds.), Vol. I, Chap.
4. Academic Press, N e w York, 1966.
(HI) Hixson, A. W., and Baum, G. J., Ind. Eng. Chem. 3 4 , 194 (1942).
(H2) Hruby, M., and Zaloudik, P., Chem. Prumysl 15, 4 6 9 (1965).
(H3) Hruby, M., Paper presented at the 4th International CHISA Congress, Prague, 1972.
(H4) Hostalek, M., and Fort, I., Coll. Czech. Chem. Commun. 5 0 , 9 3 0 (1985).
(II) Isay, W. M., "Propellertheorie—Hydrodynamische Probleme" (in German). Springer
Verlag, Berlin, 1963.
(Kl) Knudsen, J. G., and Katz, D . K., "Fluid D y n a m i c s and Heat Transfer." McGraw-Hill,
N e w York, 1958.
( K 2 ) Kotchin, Ν . E., Kibel', Ν . E., and Roze, Ν . V., "Theoretical Hydrodynamics," Vol. I (in
Russian). Gos. Izd. Fiz.-mat. Liter., Moscow, 1963.
( K 3 ) Kramers, H., "Physische Transporterschijnsalen" (in Dutch). Technische Hogeschool,
Delft, 1958.
( K 4 ) Kudrna, V., Fort, I., Eslamy, M., Cvilink, J., and Drbohlav, J., Coll. Czech. Chem.
Commun. 3 7 , 241 (1972).
( K 5 ) Kupcik, F., Thesis (in Czech), Research Institute for Macromolecular Chemistry, Brno,
1970.
( K 6 ) Kvasnicka, J., Thesis (in Czech), Research Institute for Chemical Machinery (CHEPOS),
Brno, 1967.
( L I ) Landau, J., and Prochazka, J., Coll. Czech. Chem. Commun. 26, 1976 (1961).
(L2) Landau, J., Prochazka, J., Vaclavek, V., and Fort, I., Coll. Czech. Chem. Commun. 28,
279(1963).
(L3) Levenspiel, O., and Smith, W. K., Chem. Eng. Sci. 6, 227 (1957).
( M l ) Medek, J., and Fort, I., Chem. Prumysl 29, 62 (1979).
( M 2 ) Medek, J., and Fort, I., Chem. Prumysl 3 0 , 6 2 2 (1980).
( M 3 ) Medek, J., and Fort, I., Coll. Czech. Chem. Commun. 46, 963 (1981).
14. Flow and Turbulence in V e s s e l s with Axial Impellers 197

(Nl) Nagata, Sh., "Mixing, Principles and Applications." Wiley, N e w York, 1975.
(PI) Pastyrikova, H., and Fort, I., Sci. Pap. Inst. Chem. Technol. Prague K4, 41 (1971).
(P2) Porcelli, J. V., and Marr, G. R., Ind. Eng. Chem. Fundam. 1, 172 (1962).
(P3) Prochazka, J., and Landau, J., Coll. Czech. Chem. Commun. 26, 2961 (1961).
(R1) Robertson, J. M., "Hydrodynamics in Theory and Application." Prentice-Hall, London,
1965.
(51) Standart, G., Coll. Czech. Chem. Commun. 2 3 , 1163 (1958).
(52) Steidl, H., Chem. Listy 5 2 , 839 (1958).
(53) Strek, F., "Mixing and Mixers" (in Polish), Chap. 3. Wydawnictvo Naukowo-tech-
niczne, Warszawa, 1971.
(54) Sterbacek, Z., and Tausk, P., "Mixing in Chemical Industry." Pergamon, London, 1965.
(Ul) Uhl, V. W., "Mixing: Theory and Practice" (V. W. U h l and J. B. Gray, eds.), Vol. I,
Chap. 5. Academic Press, N e w York, 1966.
(U2) Uhl, V. W., and Gray, J. B., eds., "Mixing: Theory and Practice," Vol. II. Academic
Press, N e w York, 1967.
CHAPTER 15

Scale-Up of Equipment for


Agitating Liquids
Vincent W . Uhl
Department of Chemical Engineering
University of Virginia
Charlottesville, Virginia 22901

John A. Von Essen


Philadelphia Mixers Division
Philadelphia Gear Corporation
Palmyra, Pennsylvania 17078

I. Introduction

T h e i m p o r t a n c e of scale-up for mixing operations is evident from the


profuse literature o n the subject. T h e tendency has been to oversimplify the
scaling process. This has led to several egregious misconceptions that deserve
attention at the outset. A n example is the b r o a d adoption of the constant
power per unit v o l u m e rule, apparently first noted in the literature by Buche
(B5) in 1937. Despite its irrevelance t o m o s t c o m m o n mixing applications
— b l e n d i n g a n d solid s u s p e n s i o n — i t has gained the status in some quarters
of a universal rule. A n o t h e r mistaken n o t i o n is that there is a specific or a
m i n i m u m power r e q u i r e m e n t to achieve a given mixing result. Power re-
q u i r e m e n t is a function of the efficiency of the impeller a n d the t o r q u e at
which it is operated. Large, high-efficiency impellers operated at high torque
(low speed) t e n d to require extremely low power, as little as a n order of
m a g n i t u d e less t h a n standard industrial practice, albeit at the penalty of the
m o r e expensive high-torque drives.
Scaling-up designs are also frequently fraught with uncertainty; often the
results are unpredictable a n d even surprising. Although some operations
199
MIXING: THEORY AND PRACTICE, VOL. Ill Copyright © 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
200 Vincent W . U h l and John A. Von Essen

have gone directly from bench scale to commercial size, e.g., the Shell hot-
acid alklation process (M2), except for clearly defined operations, for which
there is scale-up experience, this leap from laboratory tests to full scale is very
risky. Examples of abortive scale-ups are legion. T o further emphasize the
capricious n a t u r e of scaling, it is of interest to note that Hall's electrolytic
process for the p r o d u c t i o n of a l u m i n u m did n o t work o n a small scale, b u t
w h e n larger e q u i p m e n t was used the process proved successful ( H I ) .
O n e can identify several factors that contribute to the complexity of scal-
ing in mixing operations. First, there is the varied nature of the operations,
each with its particular process requirements; examples are blending, sus-
pension of solids, heat transfer, m a n y types of mass transfer operations,
reaction, emulsification, a n d dispersion. In addition, there are processes in
which several of these operations govern, either in concert or in series. T h e n
there is the range of properties of the materials being processed, for example,
mobile to high-viscosity liquids, a n d either single or multiple phases. Next
there is the complex fluid mechanics of any agitated liquid system; this can
be c o m p r e h e n d e d somewhat through the principles of similitude. A n d fi-
nally there is the nebulous aspect that we shall t e r m the " a p p r o a c h , " that
d e p e n d s n o t only o n the technology of concern b u t also the sophistication of
the technologists, their experience, a n d the current scaling practice. T h e
literature demonstrates that these so-called approaches lead to a variety of
rules of t h u m b , calculation m e t h o d s , a n d extrapolation procedures in going
from laboratory tests to plant operation.
Obviously, such a potpourri calls for some delineation a n d marshaling,
n o t only to aid in understanding what occurs w h e n a process is scaled u p b u t
also to provide useful directions for good scaling practice. Unfortunately, the
prodigious literature a n d attributions to the subject seem to have served
m o r e to confound. S o m e allusions are specious, most rules are extremely
limited in application, examples give too little data a n d limited analysis; the
contribution from m o s t sources is fragmentary in c h a r a c t e r — t i d b i t s e m p t y
of calories. In this chapter, a comprehensive elucidation of scale-up p h e n o m -
e n a is a t t e m p t e d along with a n ordering of design practice. All the while it
m u s t be appreciated that the effort is almost entirely art, supported by
limited concepts a n d a kit of imperfect tools.
This t a n d e m t r e a t m e n t — u n d e r s t a n d i n g a n d practice—will be consid-
ered from several points of view. T h e potential value of the principle of
similarity will be explored. This leads naturally into the "extended principle
of similarity," a basis for empirical equations which allows scale-up for
certain specific process results. This is followed by a close look at the crucial
aspects of the fluid mechanics: circulation a n d shear, the ratio of vessel
v o l u m e to wetted surface area, a n d such. T h e n , after a consideration of
c o m m o n rules of t h u m b , such as constant power per unit v o l u m e a n d
15. S c a l e - U p of Equipment for Agitating Liquids 201

constant t o r q u e per unit v o l u m e , the "translation e q u a t i o n s " n o w in vogue


are considered in detail. These equations use criteria such as power per unit
v o l u m e a n d t o r q u e per unit v o l u m e as variables. Next, agitation intensity
a n d t h e n miscellaneous scale-up considerations will be taken u p ; these in-
clude departure from geometric similarity, c o n t i n u o u s operation, a n d reac-
tors. After general advice a b o u t laboratory a n d pilot plant scale-up, a n or-
derly p r o g r a m for design practice will be delineated.

II. Principles of Similarity

A very significant concept for scale-up is the principle of similarity, even


t h o u g h for processes of s o m e complexity it is impossible to apply directly.
F o r m o s t agitated liquid systems only fluid d y n a m i c types of similarity need
to be considered: geometric, kinematic, a n d d y n a m i c . However, for some
systems, particularly reactions, two further kinds of similarity m u s t be in-
cluded: t h e r m a l a n d chemical, which are characterized by t e m p e r a t u r e a n d
concentration differences. Reviews of the formal material o n the principle of
similarity a n d the related theory of models for process operations a n d partic-
ularly o n their application t o agitation can be found in b o t h the b o o k by
J o h n s t o n e a n d Thring ( J l ) a n d papers by R u s h t o n (R4, R5).
In this connection it is to be appreciated that the fluid flow pattern in
agitated vessels can be represented by the differential equation of flow, the
N a v i e r - S t o k e s equation for noncompressible viscous flow ( D l , R4). H o w -
ever, because of the complex geometry of mixing systems, this equation
virtually defies m a t h e m a t i c a l integration; therefore, this equation does n o t
have direct utility. If o n e falls back o n t h e m a i n t e n a n c e of pertinent types of
similitude, it is found t h a t the applications are sharply restricted. Neverthe-
less, the concepts of similarity prove useful indirectly in two ways. First, they
lead to what is t e r m e d the extended principle of similarity, which in es-
sence is the correlation of test data by empirical equations using variables
consisting of pertinent dimensionsless groups. T h e second way in which
similarity proves useful is in the extrapolation of small-scale results where
t h e bases are s o m e of the similarity concepts.
T h e conditions of similitude to be reviewed are those confined to flow
systems a n d to fluid mechanics, namely, geometric, kinematic, a n d d y n a m i c
similarity. T w o systems are geometrically similar w h e n the ratios of all
corresponding form d i m e n s i o n s are equal; see Fig. 1. In this connection it
m u s t be appreciated that the fluid flow pattern is strongly controlled by the
system geometry. Kinetic similarity refers to fluid m o t i o n ; t h u s for two
geometrically similar systems, w h e n the patterns of m o t i o n are also geo-
metrically similar a n d w h e n the ratios of velocities at corresponding points
are equal, the systems are kinematically similar.
202 Vincent W . U h l and John A. Von Essen

ι r ι,ι r
1
Ί ι

D DZ i Z2 5l 22 I1-E2
1 2 1~ T
1= 1 =

T<\ T ΤΊ T
T T T
=
2 2 2

FIG. 1. A n example of geometric similarity.

F o r d y n a m i c similarity to obtain in scale-up, as shown in Table I, all the


c o m m o n force ratios m u s t equal the same constant, e.g., (F ) /(F ) ', and { M { F

also, as a corollary, all property force ratios, such as the Reynolds n u m b e r ,


m u s t have the same value for b o t h the m o d e l Μ a n d the prototype P.
It m u s t be appreciated that kinematic a n d d y n a m i c conditions of flow are
related to the geometric form or b o u n d a r y conditions, flow velocities, always
the fluid properties of density a n d viscosity, a n d sometimes the properties of
surface tension a n d elasticity. N o t e that each of the property force ratios
given in Table I consists of both the inertia force a n d a n o t h e r force arising
from the action of a fluid property such as viscosity, specific weight (gravity),
capillarity (or surface tension), a n d elasticity. T h e inertia force is determined
by the choice of the impeller, its speed, a n d its size along with fluid density.
Naturally, these property force ratios are dimensionless n u m b e r s a n d they
each have a n a m e , e.g., the Reynolds n u m b e r . T h e other use of these d i m e n ­
sionless force ratios as variables in empirical equations has been m e n t i o n e d
a n d will be discussed later for specific process results.
T o attain complete similarity in scaling it is essential that all significant
n u m b e r s , e.g., the Reynolds n u m b e r , be m a i n t a i n e d constant as shown in
Table I, or be so large that they show that the inertia force predominates.
N o w the application of the principle of similarity will be illustrated for
impeller power in an agitated tank.
F o r the viscous regime (segment A B in Fig. 2), since n o vortex forms, only
viscous forces govern, a n d the power responses, expressed by the power
n u m b e r , is given as a function of the mixing Reynolds n u m b e r or

N p = = K
* ( ^ i r ) ' = K
^ r i w h e n N
* °< 1 0 (i)

T h e presence of baffles has n o m a r k e d effect o n this relation. However, for


15. S c a l e - U p of Equipment for Agitating Liquids 203

Table I

Similitude Relations for Fluid Flow

Geometric:

Kinematic:
*>1M _ ^2M _ _ N _ V

· · * — V T —Λ 2

V V2p lP

[the numbers in the subscripts for ν refer to correspond­


ing locations in the t w o sizes ( M and P) o f
vessel]
Dynamic:
1. The common force ratios equal a constant

(Fj)
M _ (^v)M __ ( * » ) M _ (FQ)M _ K

(Fdr (F )P
V (F ) g F (F )a P
3

2. The property force ratios for mixing, such as

FJF V = N Re = ΝΌ ρ/μ 2
Reynolds number

FJF t = N Fr = N D/g 2
Froude number

FJF a = N W e = N D ρ/σ 2 3
Weber number

are the same for the model Μ and the prototype Ρ


(if applicable), that is,

(^Re)M = ( ^ R e )p

(^ ) F r M = (iV )p Fr

(^We)M =
( ^ W e )p

c B A F F L E DD

1 0 ° ΙΟ 1
ΙΟ 2
K> 3
10 4
10 5
10 6

D Np 2

H-
FIG. 2. Power characteristics of a mixing impeller. For curve A B C D , n o vortex present,
<f) = Pg/pN D .
3 5
For curve BE, vortex present, φ = (Pg/pN D )(DN /g)~ . [Adapted from 3 5 2 n

Rushton ( R 4 ) . Reproduced by permission o f the American Institute of Chemical Engineers.]


204 Vincent W . U h l and John A. Von Essen

the turbulent regime with baffles present (segment C D in Fig. 2), inertia
forces p r e d o m i n a t e , a n d the flow resistance is d u e not to " m o l e c u l a r " viscos­
ity μ b u t to eddy viscosity. Accordingly, for a given system, regardless of size,
the power n u m b e r is approximatey constant or

N = Pg /pN*D
P c
5
= K 5 (NRe > 1000) (2)

N o t e that in b o t h cases [Eqs. (1) a n d (2)] geometric, kinematic, a n d dy­


n a m i c similitude can be attained o n scaling with a single liquid. However,
the power scales u p differently in the turbulent regime with a n unbaffled
tank, or o n e insufficiently baffled to completely suppress vortex formation;
this case is represented by segment C ' E in Fig. 2. F o r this condition a vortex is
formed because of the specific weight of the liquid or a gravity field. This
p h e n o m e n o n is described by the ratio of inertia forces to gravity forces,
which is t e r m e d the F r o u d e n u m b e r (see Table I). Here, to m a i n t a i n kine­
m a t i c a n d d y n a m i c similitude o n the large scale, a fluid m u s t be used which
has different viscous properties from that for the m o d e l unit, so that the
pertinent property force ratios can be kept constant with scale-up. Attain­
m e n t of this prerequisite will be d e m o n s t r a t e d by an example adapted from
R u s h t o n (R4, R5).
F o r the example the m o d e l a n d full-scale systems are depicted in Fig. 3.
Obviously, viscous forces are a factor, b u t the presence of a vortex a n d
torodial circulation is evidence t h a t gravity also has a significant influence,
a n d its extent is m e a s u r e d by the ratio of the depth of the vortex Y over a
c o m m o n linear d i m e n s i o n such as T, or Y/T. F o r the process results to be the
same, Yj Τ m u s t be equal for b o t h scales. T o describe this condition, b o t h the
viscous a n d gravity property force ratios m u s t be invoked, as well as the
derivation of the fluid property variables o n which relative velocities at
different points in the vessel depend. This information is displayed in Table
II for the case where gravity a n d viscous forces c o m p e t e for control. If the
r e q u i r e m e n t t h a t the property force ratios for b o t h the N ' s a n d the N 'S are
R e FT

equal is m e t as shown in Table I, d y n a m i c similarity is evident from having


Υ/ Τ equal for b o t h scales. T h e n the ratio of velocities at all locations d u e to
viscous forces m u s t be equal to the ratio of velocities d u e to gravity forces for
b o t h the m o d e l a n d the prototype, i.e., for the two sizes being tested. This is
d o n e by equating the velocity ratio relations shown in the last c o l u m n in
Table II, or

v /L
T r = (L g y*
T T

Since the force of gravity is the same o n b o t h scales, i.e., g is 1.0, this t e r m T

d r o p s o u t of the relation. T h e n

v = (Z, ) '
r r
3 2
15. S c a l e - U p of Equipment for Agitating Liquids 205

FLUID MOTIO NNO T FLUID MOTIO N SIMILA R


SIMILAR FO R SAM E WHEN FLUI D P R O P E R T I E S
FLUID P R O P E R T I E S ARE CORREC T
(a) . ( b )
φ

z/= 1. 0 z/= 1. 0
LINEAR SCALE-U P O F2
(O rL (d) φ
4-

i/=1.0 z/= 2.8 3


FIG. 3. Scale-up for dimensionally similar systems with viscous and specific weight (gravity)
forces effective: vessels are unbaffled. (a) Small mixer charged with a liquid having a kinematic
viscosity ν of 1.0; (b) system which corresponds exactly to (a); (c) larger mixer with linear
scale-up of 2 from (a); it contains a liquid having the same ν as (a) and (b); (d) the larger mixer
containing a liquid with a kinematic viscosity ν of 2.83. [Adapted from Rushton (R4). Repro­
duced by permission of the American Institute of Chemial Engineers.]

This m e a n s that d y n a m i c similarity can be achieved in two geometrically


similar systems of different size only if liquids with unlike kinematic viscosi­
ties are used in each. In Fig. 3 it is shown that to m a i n t a i n similarity for a
linear scale-up of 2, the kinematic viscosity ν in the prototype vessel m u s t be
2.83 times that in the model vessel. As noted above, this process condition is
described by the C'E segment of the power curve in Fig. 2. Other pairs of

Table II

Characterizing Force Ratios and Their Effect on Velocity Ratio in Scale-Up -* 0

Dimensionless
number Velocity
Force Force Kinematic Kinematic ratio in
characteristics dimension notation dimension Name Group scale-up

Viscosity FB/L = μ
2
L /6 2
Reynolds Lu/v « Γ
= v /L
r r

Specific weight F/L? = γ y/p = g L/Θ 2


Froude u /Lg
2

« Γ = (^r) 1 / 2

Surface tension F/L = σ σ/ρ = ω L /6


3 2
Weber w L/o)
2
u =
r
(co /L Y'
r T
2

a
Adapted from Rushton (R5).
b
F, force; L, characteristic length; θ, time; u, velocity; subscripts: 1, 2, model, plant; r, ratio of
value for plant to value for model, e.g., L = L /L .
T 2 x
206 Vincent W . U h l and John A. Von Essen

force properties can be handled in the same way; some possible c o m b i n a -


tions are suggested in Table II. For different degrees of scaling it is generally
n o t convenient or possible to change viscosity, interfacial tension, or other
properties as required by the relations from Table II to achieve similarity. In
such cases the principles of similarity might still be exploited. T w o devices
are n o t e d in the following.
If viscosity controls o n e of two effective force ratios, it is possible t o
m i n i m i z e the effect or even to render it negligible by operating at velocities
high e n o u g h to achieve the turbulent regime. If at the same t i m e it is possible
to evaluate the other force parameter, it m a y then be possible to a p p r o x i m a t e
the c o m b i n e d effect. In the event that individual parameter effects can be
evaluated by such m e a n s it can be assumed for practical purposes that the
individual forces d e p e n d e n t o n each p a r a m e t e r can be added together. This
is the technique developed by F r o u d e for taking into account the viscous a n d
gravity forces operating o n the hull of a moving ship. W h e n analyzing the
resistance of ships of different sizes a n d w h e n using models tested in towing
tanks, the m e t h o d devised by F r o u d e is used to translate the resistance of a
m o d e l of a ship to a geometrically similar m o d e l of different length. This
m e t h o d is n o t precise b u t gives satisfactory engineering results. F r o u d e as-
s u m e d that he could estimate the skin friction of the curved hull to be
approximately the same as that of a thin plank of the same length a n d the
same underwater area as the ship. By separating wave resistance a n d fric-
tional resistance in this way he could extrapolate from a m o d e l to full-scale
resistance. H e obtained data o n the drag of thin planks by towing m a n y of
various lengths a n d t h u s was able to provide the necessary friction coeffi-
cients. Later, m a n y others secured similar data. This is the practice in ship
hull design today (A2).
W h e n three force properties play a part in fluid m o t i o n it is not possible to
reproduce d y n a m i c similarity for e q u i p m e n t of any other size. It is clear
from inspection of the velocity ratios for scale-up that n o t m o r e t h a n two
ratios can be equated at a time. This rules out the possibility of exploiting
similitude relationships for m o r e t h a n two property force ratios.
W h e r e similarity can be applied, either directly as in the example de-
scribed in Fig. 3, or indirectly as in ship hull designs by a c o m b i n a t i o n of test
work a n d analysis, it is the preferred approach. This explains why so m u c h
effort is given to tests in towing basins.
This brief review provides insights into the effects of scaling o n fluid
m o t i o n s and, of course, emphasizes its limitations. A n ancillary value of this
exercise is to p r o m o t e the realization that with scaling, trade-offs are c o m -
m o n . Examples of this will be pointed out later.
T h e concepts of geometric a n d d y n a m i c similarity also suggest the use of
the property force ratios, e.g., the Reynolds n u m b e r , the F r o u d e n u m b e r ,
15. Scale-Up of Equipment for Agitating Liquids 207

Table III

Process Dimensionless Relations 0

I. For heat transfer to vessel wall or surface of coil in vessel by the methods of dimensional
analysis, it is surmised that

h=f{N,d ,p,p,C ,k,D,


0 p T)

where the variables can be arranged in rational dimensional groups to give a relation with the
dimensionless variables:

or

( Nusselt \ _ r Γ / R e y n o l d s \ / Prandtl \ / g e o m e t r i c \ 1
number/ number / ' \ n u m b e r / ' \ ratio /J

or in words

( heat transfer coefficient\ _ XI


system conductivity /
inertia force \
|_\ i
v s c o u s
force/

( m o m e n t u m diffusivity\ ,,. , . Λ

thermal diffusivity )> l*™"™* correct.on)J


This relation generally takes the form

or

Note: For this somewhat simplified development, it is assumed that the physical proper­
ties throughout the vessel liquid are constant.
II. For time of blending of a batch, by the methods of dimensional analysis, it is surmised that

θ=ΑΝ,Ώ,ρ,μ, Τ)

where the variables can be arranged in rational dimensionless groups to give a relation with
the dimensionless variables

or

( Dimensionless\ = JfReynolds\ /geometrical


blend time / ^ [ \ number / ' \ ratio /J

or in words

———
( blend time

: — I =/ I—
\ XI

τ
inertia force \ ,,.

I, (dimensional correction)
. , .1

circulation time / |_ \ viscous force / J


a
These can be developed from dimensional analysis or by writing the pertinent Navier-
Stokes equation o n a dimensionless basis.
208 Vincent W . U h l and John A. Von Essen

a n d other dimensionless groups, to correlate test data by empirical equa­


tions. As noted previously this practice has been t e r m e d the extended princi­
ple of similarity. It is shown in Table HI that besides the property force ratios,
"process dimensionless" ratios or groups are necessary, a n d " o t h e r " d i m e n ­
sionless groups are either helpful or essential. Property force ratios relate
directly to d y n a m i c similarity; e.g., the Reynolds n u m b e r correlates the
effect of fluid m o t i o n in a viscous fluid. Process dimensionless ratios relate a
process r e s u l t — m e a s u r e d or d e s i r e d — t o the difficulty of its realization.
Those for heat transfer a n d t i m e of blending are given in Table III. H e r e
" o t h e r " dimensionless n u m b e r s appear; for instance, relating t h e r m a l
transfer to m o m e n t u m transfer by the ratio of the kinematic viscosity μ/ρ
(molecular transport diffusivity) to the thermal diffusivity k/C p gives the p

Prandtl n u m b e r C^/k. A n o t h e r example is the geometric ratio D/T, which


is used to a c c o m m o d a t e a variant from geometric similitude. There are two
ways in which these "dimensionless g r o u p " variables can be derived. O n e is
by writing the N a v i e r - Stokes equation in dimensionless form, in which case
the dimensionless groups appear ( D l , R4). They can also be established by
the m e t h o d of dimensional analysis, which can be found in basic chemical
engineering texts (M3).

III. Design Correlations

Correlations of pertinent dimensionless groups have been developed from


tests on a range of scales for several specific process results. For some opera­
tions, the correlation can be used directly with confidence, e.g., the power
response of c o m m o n impellers a n d t a n k geometries. N o t e that the power
response a n d the process requirement are n o t the same; the power response is
a function of the system, i.e., the power that is d r a w n w h e n the system is
operated. Also, a specific size a n d type of impeller system will have a mini­
m u m speed of operation that will satisfy the process requirements, a n d the
power d r a w n at t h a t m i n i m u m speed m a y be considered the process power
r e q u i r e m e n t for t h a t impeller system. If the power requirement exceeds the
actual power response at the design operating speed, a mixer is said to have
failed to meet the power requirement. As previously discussed, the process
power requirement is n o t a fixed m i n i m u m , b u t can be raised or lowered by
changing impeller efficiency or the applied torque.
F o r other process results such as mass transfer, there are also caveats such
as the following:
(i) Correlations are often based only o n tests in small bench-scale or
pilot-size e q u i p m e n t , a n d their extension to commercial scale has n o t been
d o c u m e n t e d . O n e particular problem can arise in heterogeneous systems, in
15. S c a l e - U p of Equipment for Agitating Liquids 209

which the performance m a y d e p e n d o n a relation between some characteris-


tic d i m e n s i o n of the v e s s e l - i m p e l l e r system that increases a n d some physi-
cal property like crystal particle size or droplet or bubble diameter that
r e m a i n s the same with scale-up.
(ii) If there is a change in the fluid regime, e.g., from viscous to transi-
tion, or from transition to turbulence, the nature of the correlation always
changes.
(iii) A d e q u a t e performance parameters m u s t be specified, b u t often
there is n o assurance t h a t the available correlation can generate a design that
corresponds to t h e m . A n example is the diverse criteria used for adequate
blending.
(iv) T h e use of correlations that d o n o t include critical geometric ratios,
e.g., Z / r a n d D/T, presumes t h a t geometric similarity will be maintained.
But it m a y be infeasible or economically undesirable to adhere to these
constraints.
(v) Because it is impractical to include all the characteristic variables, or
variable groups, in almost all cases it is impossible to generate a
"universaF'correlation. Accordingly, the possible existence of these noni-
dentified or n o n r e c o r d e d variables m u s t be appreciated a n d allowed for.
(vi) Several fluid mechanics considerations, which are described in Sec-
tion IV, m u s t be taken into account, at least as far as is practicable.
Therefore, for any operation the process engineer m u s t develop some back-
g r o u n d from experience a n d the study of at least a few, b u t significant,
references; examples are given for the candidate operations below.
As part of the t r e a t m e n t in this section, four operations are examined:
power response or investment, heat transfer, mass transfer, a n d blending
time.

A. POWER RESPONSE
This crucial aspect of mixing e q u i p m e n t has been well covered for a wide
range of impellers, vessel types a n d sizes, a n d fluid properties. Although the
major sources should be familiar, such as R u s h t o n et al (R8), one m a y need
to review Bates et al (B2), a n d resort to Nagata ( N l ) for data, rules, a n d
approaches for u n c o n v e n t i o n a l geometries, impellers, a n d vessels a n d for
viscous fluids. Skelland (SI) should be consulted for application to n o n -
N e w t o n i a n fluids. Regardless of the wealth of data in the literature, design for
power response for special geometries a n d some n o n - N e w t o n i a n liquids
should be based o n laboratory tests a n d extrapolation with j u d g m e n t by
techniques described in Section I X .
N o t e again that the power response corresponds to power investment, or
so-called mechanical power. It m u s t be at least equal to, a n d for good prac-
210 Vincent W . U h l and John A. Von Essen

tice it should a m p l y exceed, the power requirement called for to achieve the
process result. In the event that it does not, two remedies are possible:
(i) Increase the power response by operating at a higher speed or in­
creasing the size of the impeller. This m a y be difficult d u e to e q u i p m e n t
design or strength limitations.
(ii) Decrease the power requirement by changing the impeller design.
Accurate prediction of the effect of the design changes will likely require
s o m e scale m o d e l tests.
Unfortunately, the distinction between these two designations of power is
often n o t m a d e .

B . HEAT TRANSFER
T h e general correlation of heat transfer for agitated vessels shown in Fig. 4
is a good example of the application of the extended principle of similarity.
However, heat transfer for agitated process vessels m u s t be taken u p u n d e r
two headings: heat transfer at vessel walls, a n d heat transfer to surfaces
placed in vessels such as helical coils a n d baffle-type coils.
T h e data for heat transfer to vessel walls are extensive; the literature has
been thoroughly surveyed by U h l ( U l ) a n d m o r e recently by Edwards a n d
Wilkinson ( E l , E2) for b o t h N e w t o n i a n a n d n o n - N e w t o n i a n vessel liquids.
Several correlations include vessels u p t o 5 ft (1.5 m ) in diameter. Also, the
design equations have been s u m m a r i z e d by U h l in Perry et al (P2) for

Log DfN£

FIG. 4. Correlation of heat and mass transfer coefficients, fluid properties, and fluid motion
for the turbulent regime in agitated vessels. For heat transfer ψ = (hT/k)(cp/k)~ .
p
For desorp-
tion or other mass transfer operations ψ = (A: r/D )(////?Z) )" . [After Rushton (R7). Repro­
s L L
S

duced by permission of the American Institute of Chemical Engineers.]


15. S c a l e - U p of Equipment for Agitating Liquids 211

c o m m o n impellers. T h e agreement of the correlations for a variety of im­


pellers from m o s t sources is striking a n d reassuring.
T h e references above also survey the correlations for heat transfer from the
vessel liquid to coils of b o t h the helical a n d the baffle type. Unfortunately,
often these m u s t be modified for scale-up; there appear to be five difficulties:

(i) Correlations from a given source are almost always based o n tests in
a single size of vessel.
(ii) F o r the Nusselt n u m b e r , as illustrated in Table III, either d or Γ is
Q

used for heat transfer to coils.


(iii) T h e correlations use various, b u t n o t all the same, geometric ratios,
as illustrated by the 12 sources reviewed by Marshall a n d Y a z d a n i ( M l ) .
(iv) T h e addition of a coil complicates the geometry.
(v) All data except those from ( Μ 1) are for single helical coils, whereas
in practice, particularly in critical reactions, nests of concentric helical coils
are c o m m o n . Accordingly, scale-up of a vessel with coils, particularly multi­
ple coils, m u s t be d o n e carefully, a n d m u s t be based o n some test work when
the rate of heat transfer is controlling, as in a polymeric reactor, where dead
zones m u s t be scrupulously avoided.

C. MASS TRANSFER
This operation m u s t be considered u n d e r two major headings: mass
transfer from a c o n t i n u o u s liquid phase to a discrete dispersed phase such as
crystals, a n d to a fluid dispersed phase (liquid drops or gas bubbles).
H e r e only mass transfer to a dispersed gas will be taken u p . It is an
operation c o m m o n to reactors, strippers, a n d fermentors. T h e latter include
large agitated basins for the secondary t r e a t m e n t of sewage a n d industrial
wastes. F r o m the following list from Nagata (Ν 1) of the items to be taken into
account, o n e can infer the complexity of a g a s - l i q u i d mass transfer process
in an agitated vessel:

(i) the state of dispersion, i.e., the size distribution of the bubbles,
(ii) gas h o l d u p a n d retention t i m e in the vessel,
(iii) dispersion a n d coalescence of gas bubbles,
(iv) convection currents a n d degree of back mixing, a n d
(v) mass transfer across the interface of gas a n d liquid.
Accordingly, the p h e n o m e n a taking place are complex, so m u c h so that
m a n y aspects of this operation are incompletely understood at present.
S o m e relief from this imbroglio can be expected from a stream of work by,
a m o n g others, v a n ' t Riet a n d Smith (V2), N i e n o w et al (N4), a n d War-
moeskerken a n d Smith ( W l ) , w h o deal with flooding, loading, a n d recircu-
212 Vincent W . U h l and John A. Von Essen

lation of the system, the degree a n d extent of the dispersion, a n d microflow


phenomena.
In any case, understanding of a g a s - l i q u i d operation is fostered by defin­
ing the u n i q u e regime of the operation, that is, whether the system is flooded
or loaded [see Oldshue ( 0 4 , Chapter 7) a n d Hicks a n d Gates (H2)]. If the
system is to be loaded, which is likely, with increased agitation intensity, the
gas bubbles b e c o m e smaller a n d m o r e uniformly distributed t h r o u g h o u t the
liquid a n d the h o l d u p of gas in the liquid phase increases. If the primary
process concern is gas mass transfer, for the loaded conditions the mass
transfer rate is governed by the degree of dispersion a n d the extent of the
h o l d u p , which is a strong function of the power intensity P/Vof the system.
T h e mass transfer rate in the well-dispersed regime has been often correlated
by equations of the general form

k T/D = K (N Y(N y (3)


s L 6 Ke Sc

as shown in reviews by H y m a n (H6) a n d Valentin ( V I ) . This relation is


represented in Fig. 4. However, equations with other forms have been devel­
oped. Here the selection tables in Hicks a n d Gates (H2) are of practical use.
W h e n the g a s - l i q u i d mass transfer problem is of a special n a t u r e — a reac­
tion with a rate controlled somewhat by mass transfer or a biochemical
process in a n o n - N e w t o n i a n b r o t h — a pilot program is clearly indicated (see
Section IX) unless the scale-up can be based directly o n similar experience.
Two-phase l i q u i d - l i q u i d processes are in the fluid dispersed phase cate­
gory, which includes extraction operations, often multistage. For these the
considerations related to g a s - l i q u i d systems are applicable.

D . BLENDING TIME
As a simplistic approach some investigators have found that for open
impellers, a simple process relation from Table III applies, such as,
Νθ = t = h K (N ) (D/T)
7 Re
& z
(4)
Here a ~ 0 for the t u r b u l e n t regime a n d approaches — 1 for the transition
regime ( D l , Κ Ι , K2). F o r the l a m i n a r regime a again equals 0; this is sup­
ported by the information which follows from Nagata (N2). In Eq. (4) the
value of the constant K a n d the e x p o n e n t ζ d e p e n d o n the type of impeller
7

a n d the ranges of D/T and N u n d e r consideration. T h e following have been


Re

proposed for the D/T exponent for values of D/T from a b o u t 0.2 to 0.5 or
0.6.

'-2.0 for propellers ( K l )


-2.3 for R u s h t o n turbines ( K l )
ζ = -2.3 for pitched-blade turbines ( D l )
.-2.5 for pitched-blade turbines (P4)
15. S c a l e - U p of Equipment for Agitating Liquids 213

P r o b l e m s in the correlations are considerable variation, apparently due to


the p o o r correlation of the data, a n d the vari?tv of criteria used to indicate
complete blending. This was pointed o u t by Fen. . a n d F o n d y ( F l ) , o n whose
%

work the data reported in (D1) were based. Howe /er, K h a n g a n d Levenspiel
(K2) i n t r o d u c e d a m e a s u r e of concentration fluctuations which supported
their correlation.
Fortunately, blending of low-viscosity fluids takes very little t i m e a n d
energy, so rough estimates of mixing t i m e are generally adequate. If the
blending t i m e m u s t be held constant o n scale-up, it can be shown that the
power intensity can be a p p r o x i m a t e d by Ρ IV « D or V . Obviously, o n the
2 2/3

large scale it is impractical to achieve such projected power inputs.


F o r high-viscosity fluids, ΝΘ t e n d s t o be a constant; for instance, M? = 33
for the helical ribbon blender (N2). F o r a given design of impeller such a
criterion value of ΝΘ proves to be i n d e p e n d e n t of the charge material.
Also for high-viscosity fluids, if the agitator is of the circulation type, e.g., a
helical ribbon or a helix in a draft tube, Nagata found that complete mixing is
attained with three circulations of the vessel contents. Critical blending
operations, such as paint tinting, have employed 10 circulations or m o r e to
achieve a n acceptable blend.
T h e n u m b e r of impeller revolutions required to achieve o n e full circula­
tion will vary with the type a n d diameter of the mixing impeller, from a few
revolutions for a large-diameter helical ribbon to 50 or m o r e for smaller axial
turbine impellers. T h e resulting higher value of ΝΘ for smaller impellers is
consistent with the discussion regarding Eq. (4). However, there seems to be
n o scale-up p r o b l e m here provided geometric similitude is observed.

I V . Fluid M e c h a n i c s Considerations

Background for scale-up in agitated vessels includes a n understanding of


t h e several fluid flow p h e n o m e n a , s o m e of which act in concert; each can
profoundly affect scale-up. These p h e n o m e n a are
(i) the concurrent existence of mixed regimes in a vessel, namely, tur­
bulent a n d laminar, along with stagnant zones,
(ii) the variation in flow directions a n d velocities in various locations in
the agitated vessel,
(iii) the varying distribution of the power invested between shear a n d
circulation, with system geometry a n d N,
(iv) the spectrum of intensity a n d scales of turbulence in an agitated
liquid b o d y in the t u r b u l e n t regime, or in turbulent zones, a n d
(v) changes in fluid flow patterns caused by phases of different density,
e.g., flooding a n d loading regimes for g a s - l i q u i d systems; this was m e n -
214 Vincent W . U h l and J o h n A. Von Essen

tioned in Section I I I , C in connection with mass transfer between dispersed


gas a n d liquid.
In the literature these are often only alluded to, a n d w h e n they are described,
it is generally d o n e qualitatively or in a fragmentary fashion. Here the en-
deavor will be limited t o generating a n awareness of t h e n a t u r e a n d i m p a c t of
these p h e n o m e n a .

A. CONCURRENT EXISTENCE OF SEVERAL REGIMES


F o r m o s t agitated systems turbulence prevails t h r o u g h o u t the liquid body.
However, for the transition regime, best bracketed within the curve segments
BC a n d B C of the power n u m b e r correlation in Fig. 2, the fluid is t u r b u l e n t
within t h e zone s u r r o u n d i n g the impeller a n d l a m i n a r outside this space as
shown in Fig. 5a. F o r a system agitated by a n open impeller, namely, a
propeller, turbine, or paddle, where the power drawn is described by the
l a m i n a r curve segment A B in Fig. 2, the state of flow is probably described by
Fig. 5b. This m e a n s that a stagnant zone exists, especially if the liquid is
pseudoplastic, n o n - N e w t o n i a n in character. O n t h e other h a n d , if a proxim-
ity (close clearance) agitator is used, as shown in Fig. 5c, the liquid is likely to
be entirely in l a m i n a r flow. W h e n agitating pseudoplastic or Bingham plastic
liquids with open impellers, the occurrence of stagnant zones, e.g., in corners
or toroids above or below the impellers, is almost assured in the l a m i n a r a n d
transition regimes a n d likely in the wholly turbulent regimes. O t h e r t h a n
direct observation, which is usually n o t possible, the only indication of
stagnant zones is t h e c o m b i n a t i o n of the l a m i n a r regime a n d a knowledge of
the n o n - N e w t o n i a n behavior of the mix. Of course, the quality of the prod-
uct or t i m e t o mix would serve as a check for any such prediction.
As systems are scaled u p the Reynolds n u m b e r increases as the impeller

(a) (b) (c)

(
•STAGNAN TZON E
^LAMINA RREGIM E
HTURBULEN TREGIM E

FIG. 5. Flow regimes in agitated vessels: (a) flat-bladed impeller in transition regime; (b)
flat-bladed impeller in laminar regime; (c) helical ribbon impeller in laminar regime.
15. S c a l e - U p of Equipment for Agitating Liquids 215

d i a m e t e r squared or F . This m e a n s t h a t the liquid regime as defined by


2 / 3

N m a y go from the l a m i n a r to the transition or the turbulent regime. W i t h


Re

this trend, at least for o p e n impellers, the relative size of stagnant zones
would be reduced or they could disappear. If such a passage takes place with
scale-up, other m e c h a n i s m s could well operate to attain the process results
a n d different design correlations would apply. T h e caveat here is that per­
formance criteria c a n n o t be extrapolated w h e n there is a passage from o n e
fluid regime to a n o t h e r in scaling u p .

B. FLOW DIRECTION AND VELOCITY


Beyond a n awareness of the different types of regimes within the liquid
zones, s o m e conception of the flow velocities a n d patterns in various sectors
of the liquid b o d y can be pertinent. This picture can be gained from general
t r e a t m e n t s dealing with flow patterns, such as those of R u s h t o n a n d Oldshue
(R9) a n d G r a y (G3), a n d also by observation of circulating liquids in test
t a n k s with transparent walls. T h e flow can readily be tracked by dispersed
colored telltale particles which have a density a b o u t the same as that of the
vessel liquid.
Besides the properties of the liquid, the types of impellers, their n u m b e r ,
relative size, a n d location, the absolute size of the vessel, a n d the aspect ratio
or Z/T can all have a m a r k e d effect o n the flow pattern a n d the relative
velocities at different t a n k locations. In this connection it is essential to
appreciate t h a t the fluid flow streamlines t e n d to conform to patterns that are
symmetrical.
Inspection of Fig. 6 clearly d e m o n s t r a t e s the influence of geometry for a
typical system, for example,
(i) the d i m i n u t i o n in velocity at the walls a n d t o p surface,
(ii) the tendency to flow in symmetrical p a t t e r n s — o n e below the tur­
bine a n d two above, a n d
(iii) the toroids of virtually stagnant fluid in the eye of each symmetrical
pattern.
In this case the Z/T aspect of 1.2 a n d a n impeller somewhat close to the
b o t t o m — a C/Tof a b o u t 0 . 2 — p r o m o t e the formation of three circulation
patterns. F o r the system in Fig. 6 a greater C/ Τ might have been specified, or
t w o impellers properly spaced provided o n the shaft. Either would have
generated a m o r e symmetrical flow pattern a n d higher velocity or reduced
the extent of the dead zones, which would have been preferable.
T o obviate designs that have p o o r or problem flow patterns, elevation
drawings to scale such as t h a t shown in Fig. 6 should be m a d e a n d the
streamlines sketched in. Examples are tall vessels that have Ζ / Τ ratios of say
2 or 3 a n d that require m o r e t h a n o n e impeller, a n d vessels in which the
216 Vincent W . U h l and John A. Von Essen

B o f
! , e
Surface of liquid

\o.i*- — 0,5
~-*^PA

025

Vessel
wall

FIG. 6. Velocity pattern with a typical flat-bladed turbine agitator. The numbers on the
streamlines indicate the relative magnitude o f the fluid velocity based on a value of 1.0 for the
peripheral velocity of the impeller. N o t e that D/T = 0.5; C/T = 0.2; Z/T= 1.2. [From McCabe
et al. (M3). Copyright © 1985, by McGraw-Hill, Inc., N e w York.]

b o t t o m head is conical a n d which require s o m e flow c i r c u l a t i o n — t h i s is a


d e m a n d i n g situation. There are two guides for such exercises: first, a natural
sense of s y m m e t r y — o f where the streamlines should be located, a n d sec-
ond, representations of these flow patterns, to be found particularly in
sources such as G r a y (G3), Nagata ( N l ) , a n d Oldshue ( 0 4 ) .
In this connection, note that in Chapter 14 of this v o l u m e Fort provides
equations to predict the flow p a t t e r n s — f l o w velocities a n d d i r e c t i o n — f o r
axial-flow impellers in vessels with a square aspect (Z/T= 1.0). F o r some
operations, other geometries or even two or m o r e impellers would be pre-
ferred. This aspect of scaling will be considered in Section VIII.

C. SHEAR AND CIRCULATION


Analogous to the case of a p u m p , the power invested by an impeller in an
agitated vessel depends on the effective volumetric p u m p i n g rate from the
impeller a n d the head that it develops, or
P=QpAH (5)
N o w , for any given impeller in a specific geometry
Q = N NDQ
3
(6)
15. S c a l e - U p of Equipment for Agitating Liquids 217

where is the p u m p i n g n u m b e r . T h e variation of the p u m p i n g n u m b e r


with Reynolds n u m b e r is d e m o n s t r a t e d in Fig. 7 for a pitched-blade turbine
in a baffled vessel. T h e form of the individual curves is somewhat representa-
tive of those for other o p e n impeller systems; N is constant for a given
Q

geometry in the t u r b u l e n t regime a n d the values of N decrease with increas-


Q

ing D/T. In general from Eqs. (5) a n d (6) for the turbulent regime in a baffled
vessel
P = N ND p
Q
3
AH (7)

a n d from Eqs. (2), (5), a n d (6) it can be shown that

P=[N ND*)[K pN D ]
Q s
2 2
(8)

where the t e r m within the first pair of brackets corresponds to Q, the impeller
discharge rate, a n d that within the second pair of brackets, K pN D , is s
2 2

proportional to AH, the head difference between the suction a n d discharge


surfaces for a n impeller blade. This head difference results in shear a n d is
dissipated by turbulence; it is analogous to the loss of head by wall friction for
liquid flowing in a conduit.

I m p e l l re R e y n o l d
s N u m b e,r N R e

FIG. 7. Impeller pumping numbers for pitched-blade axial-flow turbines versus Reynolds
numbers at various D/Tratios in a baffled system. [From B o w e n (B4) based o n Hicks et al (H3).
Excerpted by special permission from Chemical Engineering 1976. Copyright © 1976 by
McGraw-Hill, Inc., N e w York.]
218 Vincent W . U h l and John A. Von Essen

T o explicate the shear a n d circulation aspects, several examples from the


literature will be examined. First, there is the fairly obvious effect of the
distribution of circulation Q a n d head Η for constant power in a vessel of
diameter Γ with variation of the ratio D/T This is c o m m o n l y expressed as

which can readily be derived. F r o m Eq. (6) Q ND , a n d from Eqs. (7) a n d


α 3

(8) it is a p p a r e n t that AH* N D . 2 2


T h e n Ν can be eliminated from the
resulting ratio Q/AH since from Eq. (2), for turbulent baffled flow, N<*
j)-5/3 p j t t o be m a d e by Eq. (9) is t h a t as impeller size increases, m o r e
0 n

of the constant power is expended o n flow, a n d less o n turbulence, or shear.


This is shown graphically in Fig. 8. Because the derivation of Eq. (9) ignores
effects such as t h a t of D/T on N in Eq. (6) a n d o n K in Eq. (2), it should only
Q 5

be considered as a n useful generalization.


N e x t the effect of scale-up o n fluid shear rate m u s t be understood. F r o m
the above the m a x i m u m shear rate y is proportional to AH a n d therefore
m a x

to (ND) . O n the other hand, the average shear rate in the impeller zone is
2

proportional to Ν T h e n
γ„ = Κ Ν 9 (10)

where K ~ 12 for open impellers from the data s u m m a r i z e d by Skelland


9

(SI, Table 9.3). This average shear rate was derived for use in calculating
impeller power for n o n - N e w t o n i a n systems u n d e r l a m i n a r flow conditions

ICIRCULATION, HEAD,SHEAR,
OR F L O W OR T U R B U L E N C E
FIG. 8. Distribution of power between circulation and head for constant power intensity e.
15. S c a l e - U p of Equipment for Agitating Liquids 219

by M e t z n e r a n d Taylor (M4). According t o Oldshue ( O l , 0 2 ) , velocity


gradient data obtained in the vicinity of impellers indicate that Eq. (10) is
also applicable in the turbulent regime. Skelland (SI, Chapter 9) presents a
t h o r o u g h discussion of Eq. (10), the average shear rate equation.
N o w these two relations lead t o a n i m p o r t a n t development, showing how
b o t h the m a x i m u m impeller shear rates a n d the average impeller zone shear
rates vary with scale-up. This is shown in Fig. 9, along with the relations
derived from y m a x« ( N D ) , t h a t is,
2
« D , a n d from y <* Ν, that is,
1 / 3
av

y oc D~ .
av
2/3
O r with scale-up for constant power intensity €, while m a i n ­
taining geometric similitude, Fig. 9 shows that m a x i m u m shear rate in­
creases while the average shear rate decreases. T h e increase in the m a x i m u m
shear rate can represent a significant limitation for scale-up systems which
are critically affected by the m a x i m u m fluid shear stress. Examples are
biological solids in fermentation a n d waste treatment, emulsion a n d suspen­
sion polymerization, where particle size a n d coagulation are affected, dis­
persion of pigments in coatings a n d o n fibers, crystallization, a n d floccula-

24 6 8 1 0
TANKDIAMETE RRATI O
FIG. 9 . Typical changes in m a x i m u m and average impeller zone shear rate with scale-up.
Power per unit v o l u m e and geometric similarity are maintained. [Adapted from Oldshue ( 0 3 ) .
Copyright © 1 9 6 9 Wheatland Journals Ltd., London.]
220 Vincent W . U h l and John A. Von Essen

tion (ion removal). W h e n it is necessary to attenuate the m a x i m u m impeller


zone shear rate, the impeller type is changed to o n e with a lower power
n u m b e r Λ^>, i.e., to a n impeller providing a higher Q/AH ratio, a n d / o r
increasing the Dj Τ ratio to reduce the value of (ND) . 2

So far only the m a x i m u m a n d average impeller zone shear rates (or shear
stress) have been considered. N o t e that the average shear rate throughout an
entire vessel is a n order of m a g n i t u d e less t h a n the average for the impeller
zone, a n d that the m i n i m u m shear rates, such as those in corners a n d next to
the liquid surface, are a b o u t one-fourth to one-third of the average shear rate
t h r o u g h o u t the t a n k ( 0 3 ) .

D. SPECTRUM OF TURBULENCE
T u r b u l e n c e properties of a n agitated liquid affect the microscale m o t i o n s
superimposed o n the bulk circulation of the fluid. These turbulent properties
are i m p o r t a n t in p r o m o t i n g diffusion for the complete intermingling of
molecules in blending, mass transfer between phases, and, of course, chemi­
cal reactions. Despite the considerable n u m b e r of academic investigations of
turbulence properties, the measures in practice are based o n results rather
t h a n on analysis of the m e c h a n i s m s . However, it is necessary to be aware of
a n d u n d e r s t a n d the effect of the turbulence properties at least for those
operations in which microscale effects govern.

E. PHASE DENSITY
W h e n phases have densities that are radically different, as in g a s - l i q u i d
dispersion, the flow pattern in the vessel can be controlled by the natural
tendency of the lighter dispersed phase to rise. At low mixing intensity levels,
this effect will d o m i n a t e a n d give the results different from what would be
predicted by an otherwise valid correlation. This is the case for mass transfer
between gas a n d liquid phases when the agitation level drops below that
required for complete dispersion.
As shown in Fig. 10, a typical empirical correlation for the mass transfer
coefficient can be expected to operate only at agitation levels above the point
of complete dispersion, where the impeller p u m p i n g d o m i n a t e s the flow
pattern. Below that point the gas phase controls the flow pattern in the vessel,
a n d mass transfer is less t h a n predicted by the correlations. Therefore, it is
necessary to use a separate correlation or to call on prior experience to
predict the point of complete dispersion.
W a r m o e s k e r k e n a n d Smith ( W l ) a n d N i e n o w et al. (N4) suggested that
the flooding a n d complete-dispersion transition points m a y be predicted
from a n equation of the form
Qg/ND 3
= K (N ) (D/T)
l0 FT
c d

(Π)
15. S c a l e - U p of Equipment for Agitating Liquids 221

G AS C O N T R O LS I M P E L L ER C O N T R O LS
F L OW P A T T E RN F L OW P A T T E RN

—FLOODIN G

k A
L

P/V
FIG. 10. Effect of gas flow rate o n the flooding and complete dispersion transition points for
operations where gases are dispersed in agitated liquids.

T h u s , if a g a s - l i q u i d process is to be scaled u p by the use of design


correlations, as discussed in Section III,C, it is necessary to ascertain whether
the flow regime is impeller d o m i n a t e d , e.g., from a correlation such as Eq.
(11).
A similar b u t less p r o n o u n c e d situation can occur with two liquid phases
of different densities or even with two miscible liquid c o m p o n e n t s if the
density difference is considerable. A n example is the dilution of concen-
trated caustic solution, where water is mixed with liquid-phase sodium hy-
droxide solution (specific gravity = 1.8). T h e length of t i m e t o achieve a
uniform blend will exceed all n o r m a l predictions if the agitation level is t o o
low because of disruption of n o r m a l flow patterns by the density gradient.
This situation is considered later (specifically in Table X in Section VII).

V. Common Rules of T h u m b
In their survey of scale-up for mixing e q u i p m e n t , J o h n s t o n e a n d T h r i n g
( J l ) noted the prevalence of two rules:
(i) constant power per unit v o l u m e P/V, also t e r m e d constant power
intensity e is a c o m m o n basis for scale-up; a n d
(ii) constant peripheral velocity of the impeller; which corresponds to
N D = N D is a n o t h e r widely used scale-up rule.
l l 2 2
222 Vincent W . U h l and John A. Von Essen

T h e constant power intensity rule seemed to be almost universal a b o u t 40


years ago; e.g., it was advanced for a n u m b e r of operations by Buche (B5),
Hirsekorn a n d Miller (H4), a n d Miller a n d M a n n (M5). In 1962 H y m a n
(H6) noted that the best d o c u m e n t a t e d use of the constant power intensity
rule as a scaling base was in g a s - l i q u i d systems for fermentations a n d sulfite
oxidations. It has also been found to apply to other reactors for b o t h h o m o g e ­
n e o u s a n d heterogeneous reactions. These are shear-sensitive operations.
This agitation action appears to account for less t h a n 20% of all fluid mixer
installations.
It will be d e m o n s t r a t e d in Section VI, o n translation equations, that for the
other m o r e t h a n 80% of applications, use of the P/V constant rule, which
n o w is thought to be a demolished m y t h , would provide m o r e t h a n adequate
power i n v e s t m e n t — i n fact, a n excessive i n p u t of mechanical energy.
In Fig. 9 it was seen that the peripheral velocity or tip speed determines the
m a x i m u m fluid shear rate (and shear stress). For this reason it governs the
formation of emulsions. But this criterion gives a result that corresponds to
the constant t o r q u e intensity τ or constant t o r q u e per unit v o l u m e T /V Q

rule, which has b e c o m e p r o m i n e n t in the past decade. It was first emphasized


by Connolly a n d Winter ( C I ) because it controlled the c o m m o n flow sensi­
tive operations such as blending of miscible liquids, suspension of solids,
dissolution, a n d heat transfer, all of which d e p e n d o n the bulk circulation of
the vessel liquid. In fact, constant τ has been used to successfully scale u p a
coal slurry suspension operation from 50 gal. to 8 million gal. T h e direct
correspondence of these t w o rules, involving m a x i m u m fluid stress a n d τ,
can be seen by modifying the power relation Eq. (2) for turbulent baffled
agitation to express the t o r q u e 2nT as P/N. T h e n
Q

N= P 2nT gJpN D
Q
2 5
(12)

or since F<* K L , n where L is a linear scale factor such as Τ or D, for


3

geometrically similar systems

2nT /K D
Q n
3
= N pN D /K g
F
2 2
n c (13)

F r o m this the correspondence of constant t o r q u e intensity a n d constant


peripheral velocity is evident for turbulent flow conditions.
Connolly a n d W i n t e r ( C I ) also showed that the c r i t e r i o n — a p p a r e n t liq­
uid superficial velocity—calls for results that correspond to the rule de­
scribed above, namely, constant τ or constant ND. A p p a r e n t liquid superfi­
cial velocity, or average bulk velocity i^, has proved to be a valuable
definition in design, particularly for suspension of solids. T h e apparent
superficial velocity is found from

% = QIA (14)
15. S c a l e - U p of Equipment for Agitating Liquids 223

where A is the horizontal cross-sectional area of a cylindrical vessel, or (π/4)


Γ . T h e superficial velocity is qualified by t h e word " a p p a r e n t " because the
2

actual velocity m u s t be greater t h a n that c o m p u t e d by Eq. (14). F o r instance,


with a n axial impeller the flow directed toward t h e b o t t o m of the vessel m u s t
r e t u r n in t h e opposite direction; therefore the actual average bulk velocity in
each direction o n the average w o u l d be a b o u t twice the v c o m p u t e d from b

Eq. (14). F o r other impellers, such as radial types, t h e physical situation


c a n n o t be conceived in such simplistic terms, b u t the definition in Eq. (14) is
widely applied. N o t e that for each of these materially similar rules

N*l/L (15)

if geometric similarity is m a i n t a i n e d . Therefore, with scale-up at constant


ND,

e = P/V* (N D )/L
3 5 3
oc [(l/L) L ]/L 3 5 3
oc \/L (16)

T o grasp the significance of this relation, it m u s t be realized that while the


e q u i p m e n t capacity is scaled proportional t o L , the area of t h e vessel a n d its
3

wetted fittings such as vessel baffles increase as L , a n d that the absolute fluid 2

velocities are constant, i.e., k i n e m a t i c similitude obtains. Accordingly, the


equal fluid velocities or equal fluid m o t i o n s , reacting at corresponding
points o n the wetted surfaces of t h e vessel, are resisted with equal unit force
per u n i t area o n these surfaces. F r o m this, Eq. (16) follows for scale-up based
o n equal t o r q u e intensity a n d its two related scaling criteria.
F o r all the rules a n d criterion previously considered geometric similitude
has been tacitly assumed. But there appears to be an exception for the rule of
c o n s t a n t t o r q u e intensity. It was first stated by V o n Essen (V3) a n d later
s u p p o r t e d by t h e following derivation from Bowen (B4). It is a p p a r e n t from
Eq. (12) t h a t

TQ = N p(ND) D /2ng
P
2 3
c (17)

T h u s , by rearranging Eq. (6) a n d multiplying t h e n u m e r a t o r a n d d e n o m i n a ­


t o r of t h e t e r m o n t h e right side by Γ , o n e has 2

According t o Bowen, from information in Fig. 7 for a pitched-blade turbine


operating in a baffled vessel for t h e t u r b u l e n t regime, this relation holds:

N Q = 0A3/{D/T) ' 1 2
(19)

If the equivalent for N Q above is substituted into Eq. (18),

(20)
0.43 \ D ) Τ 2
224 Vincent W . U h l and John A. Von Essen

N o w if this identity for ND is squared a n d substituted into Eq. (17) for (ND) , 2

a n d the left-hand side is divided by V while the right-hand side is divided by


the t e r m for its equivalent v o l u m e (π/4)Τ Ζ for a vertical cylindrical vessel,2

o n e has
T /V=
Q lA0N pQ /g (Z/T)T<
P
2
c (21)

N o w Nj Q/T from Eq. (14) a n d the fact that A = (π/4)Τ . Accordingly for
α 2 2

a specific value of Ν with a pitched-blade turbine, a fixed Ζ / Τ ratio, a n d a


τ

specific liquid. T /Vcan be independent of all other variables. It is remark­


Q

able that for this interpretation the D/T t e r m s in Eq. (21) cancel, a n d it
h a p p e n s because of the particular correlations used between N a n d D/T in Q

Eq. (19). A n o t h e r source of actual data for a n axial impeller is given in Fig.
11. According to V o n Essen (V3), based o n this a n d other data the t o r q u e
intensity o n scale-up can be assumed to be relatively insensitive to D/T for
axial impellers, at least within the n o r m a l D/!Toperating range of0.25 - 0.50.
F o r the flat-blade disk turbine there appears to be n o m a r k e d effect of D/T
o n the value of the impeller discharge coefficient N mq = N ND where q is q q
3

the impeller discharge rate according to Revill (R2). H e concluded from his
study of 15 literature sources t h a t the values of N lie in the range 0 . 6 - 0 . 9 5 q

for D/T values of 0 . 2 - 0 . 5 when the C/T ratios are 0 . 2 - 0 . 5 . In fact Revill
r e c o m m e n d s t h a t N = 0.75 be used for design purposes. However for scale-
q

u p , what is i m p o r t a n t is the value of N calculated from Eq. (6) using Q, the


Q

effective volumetric p u m p i n g rate, which is the total circulation generated


by the s u m of discharge a n d entrained flow; it applies in the analysis above
for the pitched-blade turbine. F r o m the m a n y pertinent published studies,

LU
σ
DC
Ο
01ι ι ι ι Iι ι ι ι Iι ι ι ι Ι ι ι ι ι Iι ι 'ι I
0.250.3 00.3 50.4 00.4 5
IMPELLERT OTAN KDIAMETE RRATIO ,D/ T
FIG. 11. Torque requirements for mixing 51 % uranium sandstone slurry in a 5-gal. vessel as a
function of impeller to tank diameter ratio. Vessel was baffled and agitated with a pitched-blade
turbine (P4).
15. S c a l e - U p of Equipment for Agitating Liquids 225

there is s o m e indication that for the flat-blade disk turbine the exponent for
D/T in a relation with the form of Eq. (19) would be a b o u t 1.0. This
m e a n s with scale-up for a flat-blade disk turbine a change in D/T would
affect T IV. N o t e that N a n d N as defined here are c o m m o n l y confused in
Q Q q

the literature.
At this point it should be recognized, at least for the turbulent regime, that
either o n e or the other of the two different types of fluid m o t i o n s generally
serves to attain the process result:
(i) shear, expressed as turbulence or head, in which case the energy
investment intensity is measured by power per unit v o l u m e Ρ/ V, or e;
and
(ii) circulation, in which case the energy investment intensity is mea­
sured by t o r q u e per unit v o l u m e T /V, or τ.
Q

Each of these generalizations calls for a qualifier. W h e n the a t t a i n m e n t of a


process result is determined by P/V, the controlling turbulence m u s t be
a c c o m p a n i e d by sufficient circulation to draw the vessel charge periodically
within the vicinity of the impeller so t h a t all portions are subjected to the
average a n d m a x i m u m impeller shear rate as indicated in Fig. 9. O n the other
h a n d , w h e n circulation governs, as for flow-sensitive operations, the circula­
tion m u s t be a c c o m p a n i e d by sufficient shear to accomplish microscale
blending, as molecular diffusion alone m a y n o t be adequate.

VI. Translation Equations

W h e r e possible, as described in Section III, scale-up is based on reliable


empirical equations in which the variables are dimensionless groups such as
those set forth in Tables I a n d III. In a n o t h e r approach, which has b e c o m e
p o p u l a r in recent years, translation equations are introduced, a n d scale-up
correlations are based o n the mixing rules of t h u m b given in Section V,
namely, e a n d τ, a n d also the "speed r a t i o " N /N . In the translation equa­
x 2

tions these criteria are expressed as functions, in particular power functions,


of a scale ratio such as D /D or sometimes V /V for the mixing operations
2 l 2 l

c o m m o n l y scaled u p .
F o r the P/V basis the equations are

(22)

(22a)

where the subscripts 1 a n d 2 c o n n o t e small scale a n d large scale, respectively.


226 Vincent W . U h l and John A . Von Essen

T h e equations for t h e T /V criterion are


Q

3x
(TQJV) _T _ 2 2 (
(23)

(23a)

H e r e note that t h e p r i m e e x p o n e n t χ applies t o t h e v o l u m e ratio, because, as


discussed in Section V, T /F Q usually can b e scaled o n the basis of v o l u m e for
axial impellers, even when geometric similitude is n o t maintained. T h e third
correlating criterion is t h e speed ratio; it is expressed only in terms of t h e
diameter ratio D /D 2 L because for its use dimensional similitude m u s t be
strictly adhered t o .

N /N =f(D /D )
2 l 2 L = (D /D )>
2 I (24)

T h e earliest effort t o develop a general translation equation for scale-up for


a n u m b e r of mixing operations was apparently that by P e n n e y ( P I ) . H i s
translation equations use e or P / F [ s e e Eqs. (22) a n d (22a)] a n d , as shown in
Table IV, are listed for u p t o seven operations for b o t h t h e l a m i n a r a n d
t u r b u l e n t regimes. In his comprehensive survey, Tatterson ( T l ) added t h e
T I F r u l e a n d the speed ratio a p p r o a c h t o five of these operations considered
Q

by Penney. However, Tatterson's tabulation (given in Table V) is for t h e


t u r b u l e n t regime only a n d presumably for baffled systems. It m a y be recalled
that Connolly a n d W i n t e r (C1) proposed t h e T /Frule. It was first extended
Q

to a range of operations by V o n Essen (V3). T h e speed ratio basis for scale-up


was introduced by R u s h t o n (R6), b u t extended a n d popularized by R a u t z e n
et al(R\).
T h e form of the equations in these tables shows that (P/V) /(P/V) = 1 2 L

for equal mass transfer in two-phase systems a n d that (T /V) /(T /V) = 1
Q 2 Q L

for equal fluid velocities a n d the corresponding equal tip speeds, as indicated
in Section V. Also, it c a n be shown that N /N = 1 for constant blend time.
2 l

Tatterson's survey (Τ 1) served to dispel confusion a b o u t the three bases for


these translation equations. However, this approach c a n b e extended to
include relations such as those depicted in Fig. 2 for power a n d Fig. 4 for
transport processes. F o r instance, t h e "process results" are expressed as a
power function of the Reynolds n u m b e r , i.e.,

(25)
where t h e corresponding e x p o n e n t is a.
T h e exponents a, n, x, a n d y for t h e four translation equations are dis­
played in Table VI for t h e five operations in Table V. F r o m Fig. 12 o n e can
gain a better appreciation of the effect of the wide variation in t h e values of
the exponents of the translation equations, in particular Eq. (22a).
15. S c a l e - U p of Equipment for Agitating Liquids 227

Table IV

R e c o m m e n d a t i o n s by Penney (1971) (PI)

Laminar or viscous mixing, Turbulent mixing,


impeller Reynolds number < 300 impeller Reynolds number > 300

Criteria Scale-up procedures Criteria Scale-up procedures

1) Equal heat (P/V) 2


la) Equal mass (P/v)2

transfer per and heat (P/V),


unit transfer co­
volume efficients to
particles,
bubbles, or
drops
lb) Equal
bubble or
drop

;~(c~(tr
diameter
2) Equal heat (P/v) 2) Equal heat (P/v)
~ ι
2 2

transfer (P/V)i transfer (P/v),


coefficients coefficients
on
stationary
surfaces
3) Equal
blend time
(P/v)
(p/n
2
~ ι
3) Equal
blend time
(P/V)
(P/v),
2

(C~(0
4) Equal tip
speed
(P/v)
{P/V),
2

(c-(r 4) Equal tip


speed
(P/v)2

(P/V), (C"~(r
5) Equal
impeller
Reynolds
(P/V)i
(P/V)i
5) Equal
impeller
Reynolds
(P/v)2

(P/v), (r*-(r
(r~(r
number number
6) Equal (P/V) 2

Froude (P/v),
number
7) Equal (P/v)2

solids sus­ (P/v),


pension

By algebraic m a n i p u l a t i o n it can be shown that all four rules generate the


same numerical answer for scaling u p a given operation. For example, the
translation e q u a t i o n for equal blend time, using Eq. (22) with exponent
y = 2 from Table V or VI, is
(P/V) 2 (D V
2
Table V

Equivalent Scale-Up Procedures for Impeller Reynolds N u m b e r > 300°

Penney (PI) V o n Essen (V3) Rautzen et al. ( R l )


Eq. (22) Eq. (23) Eq. (24)
Target
operations Criteria Scale-up procedures Criteria Scale-up procedures Criteria Scale-up procedures

1 Equal blend time (P/V) 2 (D V


2 Equal blend time (Τ /ν)ΰ 2 (D V
2 Equal blend time
^ ~ 1
(T /v\
Q \DJ N,
(ρ/η \DJ 045
2 Equal Froude number (P/v) 2 /£> \ 2
Equal surface effects (Tg/V) 2 (D \>
2 Equal surface behavior N 2 (D V<»
2

(PJV) X \DJ (T /v),


Q \DJ N, \Dj
066
3 Equal mass transfer Equal dispersion (T IV\Q (D \™ 2 Equal mass transfer N 2 /D \-
2

coefficients to (p/n (T /V)


Q I \DJ (rate) AT, \DJ
particles, drops, or
bubbles; equal
bubble or drop
diameter
05
4 Equal solids suspension 055 Equal dilute solids (T IV)
Q 2 /£) \ 2 Equal solids suspension N2 /D \-°"
2
(P/V)i (D Y 2

suspension (T IV\
Q \Dj
(p/n \DJ
5 Equal tip speed l Equal fluid m o t i o n (T /V)
Q 2 Equal liquid motion
(P/v) 2 I DY 2 N 2 2
//) y>
(P/V), \DJ (velocity) (velocity)

a
From Tatterson ( T l ) .
15. S c a l e - U p of Equipment for Agitating Liquids 229

Table VI

Exponents for Translation Equation

Translation equation criteria and corresponding exponents

*,P/V τ, T /V Q
N /N
2 x

Eq. (22) Eq. (23a) Eq.(24) Eq. (25)


Target operations y JC η a

Equal blend time 2 23 0 h


12
Surface effects * - i i
14
Dispersion 0 i ~\
Dilute solids suspension La —4la
1a 6 r
Equal fluid m o t i o n - 1 .4 0 0 -1.0 1.0
(velocity) and dense solids
suspension
Geometric similitude Some Insensitive Strict Approximate
liberty t o ( B 4 , V3) (Rl)

a
Actually a range depends o n process conditions.

VOLUMEO FCHARGE ,V
FIG. 12. General representation of Eq. (22a) where the relations are considered to be power
functions. The exponent ζ is equal to y/3 as defined in Eq. (22a) and Table V. N o t e that the
values o f Κ will differ with the specific equation used for each process result.
230 Vincent W . U h l and John A. Von Essen

T h e n from Eq. (2), since Ρ « N D\ 3


and D 3

(N D /D )
3 5 3
2 (D \
2
2

{N D ID \
3 5 3
\Dj

which reduces to

N /N2 x ~ 1
which corresponds to the speed ratio relation in Table V for constant blend
t i m e . In the appendix a detailed numerical example demonstrates the corre­
spondence of the four rules. However, some approaches have advantages
over others; e.g., for axial impellers the constant τ rule is clearly superior to
other rules.
This appears to be an appropriate place to p u t forward useful c o m m e n t s
o n each of the five "target" operations listed in Tables V a n d VI.

(1) Equal blend time


(a) constant turnover time;
(b) very fast reactions ( A l ) ;
(c) liquid velocities increase with L\ therefore power requirement
increases rapidly with vessel size a n d can b e c o m e prohibitive o n
scale-up.
(2) Surface effects
(a) vortex formation or elimination;
(b) dry solids drawdown.
(3) Dispersion (constant p o w e r / u n i t volume)
(a) shear-sensitive operation;
(b) usually use a radial turbine, e.g., R u s h t o n type;
(c) gas dispersion in liquid;
(d) two-phase liquid contacting, as for extraction.
(4) Equal fluid motion (constant t o r q u e / u n i t volume)
(a) constant tip speed (peripheral velocity);
(b) flow-sensitive operations;
(c) usually use an axial impeller, such as a pitched-blade turbine, or
a hydrodynamically efficient impeller, e.g., a hydrofoil;
(d) relatively insensitive to variation in D/T ratio;
(e) m o s t c o m m o n mixing requirement;
(f) results in same relative m o t i o n t h r o u g h o u t tank;
(g) ensures some absolute velocity t h r o u g h o u t tank;
(h) for solids suspension of b o t h fine particles a n d dense slurries
(under hindered settling conditions);
(i) blend t i m e increases with t a n k diameter (note example below);
(j) applies for suspension polymerization ( A l ) .
15. S c a l e - U p of Equipment for Agitating Liquids 231

(5) Solids suspension


(a) scale-up rules can vary [also refer to item (4) above];
(b) power per unit v o l u m e usually decreases somewhat with in­
creasing t a n k size for fast-settling dilute slurries;
(c) t o r q u e per unit v o l u m e can range from constant to the t a n k
v o l u m e ratio raised to the 1/6 power or even to the 2/9 power for
solids with very rapid settling rates (this corresponds to constant
power intensity e).
F r o m the c o m m e n t s above o n solids suspension it is noted that the scale-up
rule to use depends o n the concentration of solids a n d the settling rate of the
particles. Also, as the solids concentration increases the slurry becomes m o r e
viscous. T h e reason for o n e rule, equal fluid m o t i o n , is either solids with low
free-settling velocities (less t h a n 0.5 cm/sec or 1 ft/min) or, at the other
extreme, the onset of the (fluidized) hindered settling regime. F o r these

- it
Λ 1

LU

it
CO
<

100
S O L I D S S E T T L I N G V E L O C I T Y , ft/min a

FIG. 1 3 . Choice of scale-up rule as a function o f solids concentration and settling rates for
solids suspension. The solids concentrations exclude suspension o f particles which b e c o m e
solvated such as biological cells and cellulose pulp. Superscript a indicates multiplying the value
in feet per minute by 0.508 to convert to centimeters per second and (b) indicates consulting
Table VIII for the scale-up rule to apply for a specific sector.
232 Vincent W . U h l and John A. Von Essen

limiting conditions, roughly defined as sections A a n d Β in Fig. 13, the solids


are so readily suspended that the rule for equal fluid m o t i o n controls. N o t e
that it is for solids concentrations greater t h a n a b o u t 40 wt. % a n d for low
settling velocities. T h e slurries, which are difficult to suspend, lie within
section C in Fig. 13; these follow the variable exponent solid suspension rule.
N o w a n o t h e r effect m u s t be considered: as the solids concentration rises, the
particle interactions create a n increase in the apparent viscosity of the slurry
a n d the corresponding required t o r q u e intensity, which rises asymptotically
as the slurry approaches the m i n i m u m settled bed concentration, at which
point the slurry is n o longer fluid. This is all shown in Fig. 14. T h e χ scale-up
exponents to be considered here are χ = \ for a n ideal dilute slurry a n d χ = 0

I N C R E A S E IN T O R Q U E
R E Q U I R E M E N T S DUE
T O R I S E IN A P P A R E N T
VISCOSITY AT HIGHER
PERCENT SOLIDS

S E T T L I N G AT HIGHER
PERCENT SOLIDS

° 100
WEIGHT PERCENT SOLIDS IN SLURRY
FIG. 14. Effect o f solids concentration o n torque intensity for the suspension o f solids. [From
V o n Essen (V3).]
15. S c a l e - U p of Equipment for Agitating Liquids 233

for an ideal dense slurry in the hindered settling regime. As shown by the
dashed line in Fig. 15 for a n example real slurry at intermediate concentra­
tions in the transition zone, the scale-up e x p o n e n t ordinarily ranges between
\ a n d zero, or \ > JC ^ 0. However, t h e e x p o n e n t c a n be as high as \ for
rapidly settling solids as noted previously or even higher for very rapidly
settling particles.
O n e should fully appreciate what is c o n n o t e d by these relations. F o r
instance, for equal fluid m o t i o n , for which the t o r q u e intensity τ a n d the tip
speed would be constant, the blend t i m e increases directly with vessel diame­
ter. This could apply to the d e v e l o p m e n t of a dense solids suspension in a
hindered settling regime. Therefore, if tests in a 1-ft-diameter t a n k take 15
m i n to reach the desired degree of suspension, one should expect a b o u t 15 hr
to expire before the specified degree of suspension is realized in a 60-ft-diam-
eter tank. Although this length of t i m e to attain the process result m a y appear

D I L U TE
D E N SE S E T T L ED
S L U R RY S L U R RY B ED

>

ζ
ω
H
Z

σο
UJ

>
<
UJ

10 0

W E I G HT P E R C E NT S O L I DS IN S L U R RY
FIG. 1 5 . Variation of scale-up exponent with solids concentration for the suspension of
solids. The exponents jc are as used in Eq. (23a) and Table VI. [From V o n Essen (V3).]
234 Vincent W . U h l and John A. Von Essen

surprising, it was found by Einsle (E3) t o apply approximately for fermentors


from 30 plants having capacities from 50 liters to 1000 m . H e found that
3

m o s t of the tip velocities ranged from 5 to 7 m/sec, which controls the


m a x i m u m shear that the organisms can tolerate. For these conditions it was
found that P/V* V~ , c o m p a r e d to D~ or F ~ from Table V for equal
037 l 1 / 3

tip speed, which provides reasonable confirmation of this scale-up relation.


T h e n , in line with the example above, Einsle found that θ is proportional to
τ/0.30 £)0.9 h i c h is close to θ <* D, which was used in the example above.
O R 5 w

F o r the l a m i n a r regime, relations are given in Table IV for five operating


conditions. These relations can all be derived from Eq. (1) a n d the relation
for the heat transfer coefficient for a vessel as given in Table III.
This section has two purposes. First, it summarizes the somewhat familiar
translation equations, which provide concise relations for scaling m a n y of
the m o r e c o m m o n mixing operations. Second, it serves to dispel the current
confusion between the translation equations that have different bases. H o w ­
ever, note that these translation equations hold only when the liquid regime
at b o t h scales is either l a m i n a r or turbulent. N o rules are provided for the
transition regime.

V I I . Agitation Intensity

A . MEASURES OF AGITATION INTENSITY


F o r certain applications there is latitude in the degree of liquid m o t i o n that
corresponds to the agitation intensity. This in t u r n is related to the m o d e of
investing power or t o r q u e a n d its m a g n i t u d e . S o m e operations for which a
range of agitation intensity is considered are the following:
(i) for blending, where it determines the t i m e for homogenization a n d
also takes into account the property differences of the c o m p o n e n t s to be
mixed;
(ii) for suspension of solids, where it defines the degree of completeness
of the particle distribution;
(iii) for g a s - l i q u i d operations, where the degree a n d grain of the gas
dispersion are i m p o r t a n t ; a n d
(iv) for heat transfer, where some margin of safety is prudent, e.g., for a
polymerization reactor.
F o r purposes such as these, several bases have been suggested for character­
izing agitation intensity: verbal hierarchy, impeller tip speed, power inten­
sity, a n d numerical grades or index n u m b e r s . T h e numerical indices can
correspond to superficial vertical fluid velocities or turnover rates. These
various scales are discussed in the following.
15. S c a l e - U p of Equipment for Agitating Liquids 235

Table VII

Typical Power Requirements for S o m e Mixer Operations' 1

Operation Power requirement

Blending vegetable oils 1 h p / 1 0 0 , 0 0 0 lb (0.4 k W / 1 0 0 , 0 0 0 kg)


Blending gasoline 0.3 h p / 1 0 0 0 bbl (0.014 k W / m ) 3

Clay dispersion 10 - 1 2 h p / 1 0 0 0 gal. ( 1 . 9 - 2 . 2 k W / m ) 3

Fermentation (pharmaceutical) 3 - 1 0 h p / 1 0 0 0 gal. ( 0 . 6 - 1 . 9 k W / m ) 3

Suspension polymerization 6 - 7 h p / 1 0 0 0 gal. ( 1 . 2 - 1 . 3 k W / m )


3

Emulsion polymerization 3 - 1 0 h p / 1 0 0 0 gal. ( 0 . 6 - 1 . 9 k W / m ) 3

Solution polymerization 1 5 - 4 0 h p / 1 0 0 0 gal. ( 3 - 7 . 8 k W / m ) 3

a
F r o m Baasel (BI).

Verbal hierarchies. Several have been suggested, of which the " m i l d , "
" m e d i u m , " a n d " v i o l e n t " descriptions of Weber (W2) are typical. Such
w o r d descriptions of the degree of agitation intensity are still in use today.
N o t e that they generally apply to either power intensity or level of fluid
m o t i o n (velocity) in the vessel.
Impeller tip speed. A n example of a scale o n this basis has been provided
by H o l l a n d a n d C h a p m a n ( H 5 , p . 191), with a range from 500 ft/min for very
low to 1100 ft/min for very high degree of agitation.
Power intensity. In the literature there are r e c o m m e n d e d values for
m a n y operations, often in t e r m s of horsepower per 1000 gal. S o m e power
r e q u i r e m e n t s collected by Baasel (Β 1) are given in Table VII. Bates et al. (B2)
give power intensities for m a n y other services. (It should be appreciated that
the use of these Ρ IV values for scaling m o s t process tasks will result in overly
conservative power investment.)
Numerical grades or index numbers. In 1961 a system was published by
T h e Pfaulder C o m p a n y (P3) which provided index n u m b e r s from 0.5 to 10
for levels of agitation for five c o m m o n process results. T h e scheme consisted
of charts providing agitator sizes a n d speeds for glass-lined a n d stainless steel
vessels from 5 to 4000 gal. However, n o rationale for the system was set forth.
M o r e recently ( 1 9 7 5 - 1 9 7 6 ) , engineers from Chemineer, Inc. authored a
series of articles which presented a similar scheme, t e r m e d Chemscale, 1

giving index n u m b e r s from 1 to 10 for three operations: blending a n d m o ­


tion, solids suspension, a n d gas dispersion.

T h e latter scheme can be useful for preliminary work or checking pur­


poses. T h e bases for the three operations m e n t i o n e d are presented in Table
VIII with related information. These bases were deduced from the tabulated

1
Registered trademark, Chemineer, Inc., Dayton, Ohio 4 5 4 0 1 .
Table VIII

Basis for Chemscale* Index Scheme for Agitation Intensity for Three Operations

Applicable
Operation D y n a m i c response Basis for dynamic response Impeller scale-up rule Reference

Blending and Bulk fluid velocity Bulk velocity Pitched-blade TQ/V constant Hicks et al ( H 3 )
motion* Mild agitation: Velocity 6 ft/min (N = 1) to
t turbine
violent agitation. 6 0 ft/min (JV, = 10). See
Eqs. (26) & (27) and Table X /
Solids suspension* Level of dilute From m o v e m e n t of solids at tank bottom Pitched-blade Approximate Gates et al ( G l )
,/4
solids suspension (Chemscale = 1) to virtual slurry uniformity turbine ρ/κ«φ /ΛΓ 2
(Chemscale = 10)
0
Gas dispersion Degree of gas From flooding (Chemscale = 0) to completely Flat-blade turbine Ρ/ V constant Hicks and Gates
dispersion loaded with m a x i m u m interfacial area and (H2)
recirculation (Chemscale = 10)

a
Registered trademark, Chemineer, Inc., Dayton, Ohio 4 5 4 0 1 .
b
Flow-sensitive operations.
c
A shear-sensitive operation.
d
For blending and motion the Chemscale index is s y n o n y m o u s with JV,.
15. S c a l e - U p of Equipment for Agitating Liquids 237

data o n p r i m e m o v e r power a n d shaft speed for various capacities corre­


sponding to a given Chemscale n u m b e r . T h e shaft speeds are those for
c o m m o n l y available p r i m e m o v e r speeds, e.g., 1750 a n d 1150 r p m , linked
with gear ratios standardized by the A m e r i c a n G e a r Manufacturers' Associa­
tion ( A G M A ) . As a n example of the published schedules, a n abridgement of
such a p o w e r - s p e e d tabulation from Gates et al ( G l ) is displayed in Table
I X ; it is for suspending solids with a settling velocity of 10 ft/min.
T h e basis for d y n a m i c response for blending a n d m o t i o n is equal fluid
m o t i o n , a n d for this operation Hicks et al (H3) equate the Chemscale
n u m b e r Ν of 1 through 10 linearly to the bulk fluid velocity v of 6 to 60
λ b

ft/min. Since by definition

ν* = 0/Α = 0/[(π/4)Τ ] 2
(26)

Table IX
Prime Mover Power and Shaft Speed for Solids
Suspension (for Particles with Settling Velocities of
10 ft/min) fl

Scale Equivalent v o l u m e (gal.)


of
agitation 500 5,000 15,000 100,000

1 1/350* 5/125 10/84 60/84


3/84 7.5/68 50/68
3/68 5/45 40/56
2/45 3/37 30/37
4 1/155 7.5/84 30/100 200/68
5/56 26/84 150/56
15/45 125/45
100/30
5 1/125 15/156 40/100 300/100
10/100 250/84
7.5/68 150/45
5/45 125/37
6 1/100 10/84 40/84 300/68
30/68 250/56
25/56 200/45
20/37 150/37
10 5/125 50/100 150/84
40/84 125/68
30/68 100/56
25/56 75/45

a
Abstracted from Gates et al. ( G l ) .
b
1/350 denotes nominal horsepower of drive and
shaft speed in revolutions per minute (hp/rpm).
238 Vincent W . U h l and John A. Von Essen

Table X

Intensity of Agitation N e e d e d for Blending Miscible Fluids, as Determined from Basic


Correlations 0

Specific
gravity Viscosity
variation variation, o f Suspending < 2 %
JVf, degree (ASg) c o m p o n e n t s (// /// )a b
trace solids Surface m o t i o n

2, mild <0.1 <10 2


No Fluid surface
flat but
moving
6, m e d i u m <0.6 <10 4
Settling Rippling surface
rates = 2 - 4 at low
ft/min viscosities
10, violent <1.0 <10 5
Settling Surging surface
rates = 4 - 6 at low
ft/min viscosities

a
Abstracted from Hicks et al ( H 3 ) by Bowen (B4).
b
By definition, N = Q/6A.
x

from t h e information directly above


Ν = v /6 (ft/min)
τ h (27)

N o t e that the bulk velocity v is t h e same as the so-called apparent superficial


b

velocity.
T h e need for increasing values of t h e Chemscale index t o blend c o m p o ­
nents with different specific gravities a n d viscosities, a n d also suspend solids
of increasing settling rates, is well demonstrated by the reduction by Bowen
(B4) t o a concise tabulation (Table X ) of the original process requirements of
Hicks et al (H3). F o r m o r e detailed information o n t h e Chemscale system
the reader is directed t o t h e three articles cited in Table VIII. It m u s t b e
appreciated that fluid m o t i o n a n d blending a n d the suspension of solids are
flow-sensitive operations. F o r b o t h solid suspension a n d gas dispersion, the
bases for t h e Chemscale index are descriptive (see Table VIII).
For flow-sensitive operations, including a n u m b e r of solid suspensions,
Table X I presents r e c o m m e n d e d Chemscale values, which can be used for
scale-up.

B. TURNOVER AND RELATED MEASURES


C o m m o n criteria for the performance of mixed systems are t h e rate of
t u r n o v e r Q/ V, say per m i n u t e ; its inverse t h e turnover time V/Q, which is the
t i m e t o t u r n over t h e contents of the vessel once; a n d t h e n u m b e r of t u r n ­
overs 9 Q/V, where 0 is t h e operating time.
q q
Table XI

0
Examples of Flow-Sensitive Systems in the Chemical Process Industries, Including a N u m b e r of Solids Suspensions

Chemscale Application Industry Process Operation

<1 Crude oil storage Petroleum Suspending bottom sludge and water in crude Suspension
1 or less Equalization basin Water treatment Blending to prevent concentration surges Blending
1 Fuel additive blending Petroleum Blending miscible additives Blending
1-2 Brew kettles Fermentation Beer fermentation Reaction
2 Syrup storage Sugar and starch Syrup holding tank Storage
2-3 Pigment suspension Paint Maintaining suspension Storage
2-3 Lime slurry storage Water treatment Maintaining 0 - 2 0 % solids suspension Storage
2-3 Polymer storage Polymer Maintaining emulsion Storage
2-4 Starch converter Sugar and starch Enzyme conversion Dissolving
2-5 Clay storage Ceramics Maintaining clay suspension Storage
3 Feed or holding Paint Maintaining uniformity during temporary holding Storage
3 Sugar dissolving tank Sugar and starch Dissolving dry sugar to produce syrup Dissolving
3-4 Starch storage Sugar and starch, pulp and Holding tank for suspended starch Storage
paper
3-4 Clay storage Clay for coatings Storage
Pulp and paper
3-4 Lime slurry storage Water treatment Maintaining 2 0 - 3 0 % solids suspension Blending
3-8 Blend tank Adhesives Blending ingredients Dissolving
3-10 Flash mixer Water treatment Rapid mixing of water treatment chemicals Suspension
4-5 Lime slurry makeup Water treatment Suspending slaked lime in water ( 0 - 2 0 % solids) Suspension
6-7 Lime slurry makeup Water treatment Suspending slaked lime in water ( 2 0 - 3 8 % solids) Suspension
6-8 Starch cooking Pulp and paper Starch preparation for coatings Reaction
6-10 Blunger Ceramics Breaking up and suspending ball clay Suspension
6-10 Thin and tint Paint Blending of base, vehicle, and pigments Blending
6-10 Rubber cement tank Adhesives Cutting and dissolving Dissolving
6-10 Emulsion polymerization Polymer M o n o m e r emulsions usually in water with stabilizers Reaction
8-10 Lime slaking tank Water treatment Converting CaO to C a ( O H ) 2 Reaction
8-10 Bulk polymerization Polymer Polymer is molten or soluble in m o n o m e r Reaction
8-10 Solution polymerization Polymer M o n o m e r and polymer are soluble in solvent Reaction
10 or more Starch converter Sugar and starch Acid conversion in staged c o l u m n Reaction

a
Abstracted from Gates et al (G2) by Bowen (B4) with s o m e modification by the authors.
240 Vincent W . U h l and J o h n A. Von Essen

Relations for the first two circulation criteria Q/Vand V/Q can readily be
derived for a flat-bottomed cylindrical vertical vessel. Using Eq. (27) t o m a k e
use of Ν ϊ9

Q_ 6Λί(π/4)Γ» _ 6Ν τ _6N l

V (π/4)Τ (Ζ/Τ)
3
T(Z/T) Ζ 1 }

[This use of Eqs. (26) a n d (27) t o generate Eq. (28) follows Bowen (B4).]
Accordingly, the values of these three t u r n o v e r measures are related to the
a p p a r e n t superficial velocity, u , defined by Eq. (26).
b

T h e V/Q criterion, t u r n o v e r time, defines agitation intensity, a n d it is


proposed in this sense by Bathija (B3), w h o suggests this t u r n o v e r t i m e scale
for axially positioned radial jet mixers:

(i) mild agitation — 3 to 60 m i n — f o r dye blending, neutralization,


a n d agitation of storage tanks,
(ii) m e d i u m a g i t a t i o n — 3 0 sec to 3 m i n — f o r p H control, batch mix­
ing, solids suspension, a n d heat transfer,
(iii) violent a g i t a t i o n — 1 0 t o 30 s e c — f o r g a s - l i q u i d operations a n d
p i g m e n t blending.

However, it can be shown from the above [Bowen (B4, Eq. 23)] that for any
Ni scale, t u r n o v e r t i m e should increase with a linear scale L. See Bowen (B4).
A n o t h e r c o m m o n criterion, particularly for blending, is the total n u m b e r
of turnovers 0 < 2 / Vto attain a satisfactory degree of homogeneity. A n exam­
q

ple is the three or so turnovers needed to completely blend c o m p o n e n t s in


the l a m i n a r regime w h e n using helical ribbon agitation, as m e n t i o n e d in
Section III,D. N o t e that these turnovers can be carried out in a short t i m e or
over a n extended period, depending o n the p u m p i n g rate Q. Therefore they
d e p e n d n o t o n agitation intensity b u t o n the quantity of agitator action.

V I I I . Miscellaneous Considerations

So far a n a t t e m p t has been m a d e t o e m b r a c e all pertinent considerations


within the framework used t o develop the subject of this chapter. Because of
t h e complexity a n d intricacy of this subject, it should be a p p a r e n t that this
endeavor is impossible. Accordingly, in this section pertinent ancillary
topics are developed sufficiently for the present purposes. O n e can judge
from the subheadings that the topics are truly miscellaneous in nature.

A. GEOMETRIC VARIATIONS
Because of the prevalence of vertical vessels with axially located impellers,
such as t h e system portrayed in Fig. 6, t h e t r e a t m e n t here will b e confined t o
this geometry. T h e information below is intended as a guide to good practice
15. S c a l e - U p of Equipment for Agitating Liquids 241

t h a t should be heeded if n o t meticulously followed. Violations of such good


practice are the cause of m a n y unnecessary aberrations in the operations of
process mixing installations. T h e relative geometry is generally portrayed by
ratios, particularly D/T, which has been discussed, Z/T'for the ratio of height
of liquid to t a n k diameter or aspect ratio, a n d C/Τΐοτ the distance of the
impeller from the vessel b o t t o m for o p e n impellers a n d from the wall for
close-clearance or proximity impellers. Variations in these values can have
p r o f o u n d effects o n the liquid flow patterns a n d velocities in a vessel. H o w ­
ever, designers a n d operators should be aware of other geometric factors
which affect the fluid flow. Examples are the type a n d proportions of the
impeller, e.g., W/D (height to diameter), angle of the blade, presence of a disk
with the flat-bladed turbine; baffle width a n d height; type of tank, e.g.,
vertical cylindrical, horizontal cylindrical, or square; a n d appurtenances
within the vessel: thermowells, steady bearing m o u n t s , etc. H e r e only the
p r i m a r y three geometric ratios will be considered.
Several aspects of the D/ Τ ratio are considered. In Fig. 7 a n d t h e n in Eq.
(19) it was shown that for pitched-blade turbines, the p u m p i n g n u m b e r N Q

decreases as Dj ^ i n c r e a s e s . T h e usual range of D/T for o p e n impellers is from


0.25 t o 0.50 or at t h e m o s t 0.6.
T h e point has frequently been m a d e in the mixing literature, e.g., by
H o l l a n d a n d C h a p m a n ( H 5 , Fig. 3-5) a n d Oldshue ( 0 4 , Fig. 9-9), that there
is a n o p t i m u m D/T to attain a specified process result. Unfortunately, such
sources only give indications such as that in Fig. 16. Also, since such plots
vary with the size of the vessel, it was suggested by H o l l a n d a n d C h a p m a n

FIG. 16. Examples o f power vs D/ Τ ratios for shear-sensitive (HS) and flow-sensitive opera­
tions. [From Holland and Chapman (H5). Copyright © 1962 by Reinhold Publishing Corp. All
rights reserved.]
242 Vincent W . U h l and John A. Von Essen

( H 5 , p . 58) that the o p t i m u m D/T be found by extrapolating the results at


different scales. In m a n y installations D/T ratios lower t h a n the o p t i m u m are
specified because less torque, m e a n i n g a smaller speed reducer with a lower
capital cost, does the j o b . Of course, there is a trade-off, the extra a n n u a l
expense for power. W i t h escalating expense for mechanical energy input,
there has been a trend toward higher values of D/T
W h e t h e r o n e or several impellers are used o n the shaft, the r e c o m m e n d e d
location of the lower impeller is determined by process requirements, as
brought o u t by r e c o m m e n d e d values in Table XII for the three operations
listed in Table VIII. N o t e that vessels sometimes have to operate with charges
m u c h less t h a n the design capacity; this can control the design C/ Τ ratio by
calling for a lower value. However, operation at extremely low fluid levels
can be a problem, since fluid flow patterns c a n n o t produce circulation in the
resulting " p a n c a k e " shape of the fluid batch. Normally it is h a r d to operate at
a Z/T ratio less t h a n 0.2, even with a n impeller next to the b o t t o m .
T h e values of Z/T used vary widely. F o r test work the so-called square
configuration is c o m m o n , i.e., Z/T= 1.0. However, for the turbulent re­
gime with only o n e impeller, values of Z/Tas high as 1.2 to 1.4 are frequently
acceptable for flow-sensitive operations ( G l , H3), a n d for shear-sensitive
operations Z/T= 1.0 is called for as brought out in Table XII. Vessels

Table XII

Impeller Clearance Ratios and N u m b e r of Impellers for Three Operations 0

Impeller
clearance

Viscosity N o . of Max. Z/T


Operation (cP) impellers Bottom Upper ratio* Source

Blending 25,000 1 Z/3 — 1.4 Hicks et al (H3)


and 2 Γ/3 \Z 2.1
motion 25,000 1 Z/3 — 0.8
0

2 Γ/3 \z 1.6
Solids suspension
— 1 Z/4
— 1.2 Gates et al ( G l )
— 2 Γ/4 \z 1.8
Gas dispersion
— 1 Γ/6 — 1.0 Hicks and Gates (H2)
— 2 Τ/β \z 1.8

a
Consult Table VIII for additional information regarding the three subject operations.
* M a x i m u m values of Ζ / Τ shown must be considered the limits for N e w t o n i a n fluids where
surface motion is not critical. If fluid is thixotropic or surface m o t i o n in excess o f nominal is
desired, m u c h lower Ζ / Γ limits should be employed, frequently as low as half the values given
above.
c
For blending and motion Hicks et al (H3) should be referred to for r e c o m m e n d e d baffling
and corrections to the diameter of the pitched-blade turbine when N < 700. Re
15. Scale-Up of Equipment for Agitating Liquids 243

having higher aspect ratios (Z/T) are desired for a n u m b e r of reasons in plant
operations: space limitation, s o m e staging effect for g a s - l i q u i d processes,
less weight because of t h i n n e r walls for high-pressure operations. This calls
for multiple impellers, which are discussed in the following subsection.

B. MULTIPLE IMPELLERS
Although laboratory setups for b o t h research a n d piloting almost always
have only o n e impeller, production-scale agitated systems frequently space
several impellers o n the same shaft. Services where this is c o m m o n are
reactors, particularly at high pressure; fermentors, e.g., for antibiotics; a n d
staged extraction units.
W h e n staged operation is called for it can be achieved by installing disk- or
d o n u t - s h a p e d baffles between the equidistant impellers, a n d the rule of
s y m m e t r y should be observed (cf. Section IV,B) for each stage, i.e., they
should approach a square aspect; stages are usually in the range of Z/T = \ to
1, a n d a Ζ/ΤοΐΟ.Ί to 0.8 is considered ideal. If there are n o horizontal baffles
to separate the stages the flow pattern will tend to be cellular, which provides
flows t h a t a p p r o a c h a staged system. This " a u t o m a t i c " staging is m o r e pro­
n o u n c e d w h e n the vessel charge has a high consistency, i.e., when the fluid
regime is in the l a m i n a r or at the low e n d of the transition regime. As shown
in Fig. 17, Richards (R3) dramatically d e m o n s t r a t e d the effect of the spacing
of three flat-bladed impellers o n flow patterns a n d incidentally o n power
absorption for a fermentation m a s h in a baffled vessel. If impellers are spaced
too far apart, it b e c o m e s possible for high-consistency charges, as in this case,
to have stagnant zones between the fields of influence of the impellers. If they
are t o o close together, serious interference can occur between the flow
streams from two adjacent impellers. This can have the double effect of
reduced power i n p u t a n d inadequate mixing. Figure 17 confirms the tend­
ency to a u t o m a t i c staging for equally spaced impellers.
However, the proclivity of systems with equally spaced impellers to form
closed flow patterns, in effect cells or stages, is greatly diminished when gases
are sparged into liquids. In this case, using the mass transfer rate as the
m e a s u r e of effectiveness for a given power intensity, n o m a r k e d change is
noted with added impellers a n d altered positioning according to data pre­
sented by Oldshue ( 0 4 , p p . 2 6 5 - 2 6 6 ) .
Although there is evidence t h a t m o r e work has been d o n e by operating
c o m p a n i e s to elucidate these p h e n o m e n a , to date the results have n o t been
published.

C . SEVERAL SIMULTANEOUS PROCESSES


A c o m m o n p r o b l e m in scale-up is that of contending with several pro­
cesses that proceed simultaneously. F r e q u e n t c o m b i n a t i o n s are heat transfer
a n d reaction; gas dispersion a n d mass transfer; suspension of solids a n d
DISTANCEBETWEE N
0.3D 0.5D 1.0D 1.5D 2.0D
I M P E L L E R< £

POWERDRAW N
0.76 0.76 0.73 0.91 0.90
H.P./1000 G A L L O N S

(a) (b) (c) (d) (e)


FIG. 17. Effect of spacing of flat-bladed impellers o n the flow pattern and power adsorption
for fermentor mashes. For the different arrangements, the cross sectioning indicates zones
where the mash is stagnant or virtually so. [From Richards (R3).]
15. S c a l e - U p of Equipment for Agitating Liquids 245

disssolution or crystallization; blending (homogenization) a n d reaction; a n d


heat transfer, catalyst suspension, a n d reaction. T h e c o m p a n i o n mecha-
nisms are generally rate processes, b u t they can be simply a m a t t e r of degree
or fluid mechanics, as for suspension of solids.
T h e p r o b l e m being dealt with here is what J o h n s t o n e a n d Thring (J 1) t e r m
" m i x e d regimes" a n d define as " a condition in which a n overall rate of
change is significantly influenced by two processes having incompatible
similarity criteria."
H e r e two questions p r e d o m i n a t e : first, which of the several coexisting
processes are critical or limit the operation, a n d second, which processes m a y
shift in i m p o r t a n c e with scaling. But there is a third consideration: whether
there is a favorable synergism between the coexisting processes, i.e., whether
the concurrence is beneficial or deleterious to the process result.
Knowledge of the several coexisting processes or the relations which gov-
ern t h e m is needed to c o n t e n d with p h e n o m e n a that are limiting. A c o m -
m o n l y cited example is mass transfer a n d reaction in two-phase systems. Of
course, reaction rates are extremely sensitive t o temperature, a n d their varia-
tion provides a measure for distinguishing the relative i m p o r t a n c e of the
c o m p e t i n g processes. In a n y case, the design is governed by the d o m i n a n t
process, of course at the desired scale.
O n e of several competing processes m a y shift from being inconsequential
to d o m i n a t i n g the overall process results. This can occur because of the
different rates of change of the various fluid mechanical behaviors (see Fig. 9)
a n d the effect of geometric features such as wetted area a n d liquid depth. A n
example of such a n o m a l o u s behavior occurs with fermentors. Pilot-scale
units can disperse air readily b u t are blending-limited, while plant-scale units
with air rates scaled o n a n air v o l u m e rate to liquid v o l u m e — g e n e r a l l y
expressed by V V M for air v o l u m e dispersed per m i n u t e per liquid batch
v o l u m e — a r e air dispersion-limited b u t blend readily. This m e a n s that a
specification for air loading dictated by satisfactory pilot-scale tests m a y n o t
be achievable in a commercial-size installation.
E x a m p l e s of synergistic c o m b i n a t i o n s readily c o m e to m i n d . O n e is the
beneficial effect of a n i m p r o v e d degree of solid suspension, at least to the
extent of complete suspension, o n the mass transfer m e c h a n i s m of dissolu-
tion. Actually, the e n h a n c e m e n t of solution rate beyond this degree of sus-
pension is m i n i m a l , as n o t e d in C h a p t e r 12 of this v o l u m e .
T h e i m p a c t of these various effects o n design for scale-up are discussed in
Section X,C,4.

D . BATCH CHEMICAL REACTORS


T h e reactor is frequenty the focal point, the heart, of a chemical process;
t h e n it is the critical unit. According to Olson a n d Stout ( 0 5 ) , chemical
246 Vincent W . U h l and John A. Von Essen

reactors can be categorized into single-batch units, continuous-flow stirred


trains, a n d piston-flow reactors. T h e latter two relate t o continuous-flow
systems; therefore, they will be taken u p in the next subsection.
Because of the i m p o r t a n c e a n d complexity of this subject it has been given
m u c h experimental a n d theoretical study, in fact so m u c h that its t r e a t m e n t
is b e y o n d the scope of this chapter. However, because chemical kinetics, heat
transfer, fluid a n d solid flow rates, catalyst size, a n d geometry are involved,
m u c h that has been presented above is applicable to scale-up of batch reac-
tors.
F o r relatively slow reactions, the major considerations are mixing t i m e
a n d heat transfer. Because a large batch reactor will have the same reaction
t i m e as a smaller one, it will have to be mixed in the same time, b u t the
relations in Section VI show that this m a y n o t always be possible. It is also
likely that the ratio of heat transfer area to v o l u m e for the small unit c a n n o t
be m a i n t a i n e d for a very large commercial-scale unit. This limiting situation
obtains particularly for certain types of polymerization processes because the
viscosity of the charge increases as the reaction proceeds. It becomes difficult
to m a i n t a i n the required mixing a n d rates of heat transfer.
A n o t h e r problem arises in fermentors a n d reactors producing long-chain
polymers. W i t h increased power intensity o n scale-up the higher agitator
shear levels m a y d a m a g e the organisms or disrupt long-chain polymers with
deleterious effects.
Very rapid reactions present further problems which m a k e severe design
d e m a n d s unless the mixer size can be severely restricted. Here, as pointed
out in the chapter o n chemical reactions by J. Bourne in Oldshue ( 0 4 ) , it is
critical that the reactants be rapidly mixed at the feed point.
It m u s t also be appreciated that some reactions simply d o n o t scale u p well
a n d that the yield of a specific reactant m a y be determined by factors other
t h a n the type a n d degree of agitation a n d the available heat transfer surface.
F o r background o n scaling u p batch reactors the reader is directed to three
references: J o r d a n (J2) for process development, Olson a n d Stout ( 0 5 ) for
a n excellent comprehensive s u m m a r y , a n d B o u r n e in ( 0 4 ) for a briefer b u t
m o r e recent view, particularly of rapid reactions.

E . CONTINUOUS OPERATIONS
It was m e n t i o n e d in the subsection above that c o n t i n u o u s chemical reac-
tions are carried o u t in single stirred vessels, continuous-flow stirred trains,
a n d piston- or plug-flow units. Because of the i m p o r t a n c e of this topic,
particularly in relation to chemical reactions, considerable excellent theoret-
ical a n d experimental work has been carried out. F o r two reasons the reader
will be referred to other sources for guidance a n d instruction o n scale-up: the
15. S c a l e - U p of Equipment for Agitating Liquids 247

information is too extensive to include in this chapter, a n d there are several


excellent sources for instruction a n d design guidance.
However, s o m e general information seems in order. W h e t h e r single
stirred vessels or trains of such vessels are used, if they are perfectly m i x e d —
a n ideal limiting state for which they are t e r m e d C S T R s for continuous-
flow-stirred-tank r e a c t o r s — t h e i r performance can readily be described by
theoretical relations. Actual systems can be well a p p r o x i m a t e d by models.
Systems with true piston flow are approached by pipe flow, static mixers, a n d
packed c o l u m n s .
T h e d o m i n a n t performance characteristic is residence t i m e distribution
( R T D ) , which for perfectly mixed stirred vessels follows a n exponential
relation, which is modified for a train a n d for incomplete mixing in the
vessels in the train. Residence t i m e distribution is readily measured in actual
systems by a n u m b e r of experimental techiques. F o r a train of stirred tanks,
as m o r e stages are added a n d the n u m b e r approaches 10 the R T D a p -
proaches that for plug flow.
F o r preliminary background the reader is referred to undergraduate texts
in chemical reaction kinetics a n d t h e n to Olson a n d Stout ( 0 5 ) . Certainly,
t h e up-to-date review of pipe line mixing in Chapter 13 of this v o l u m e should
be consulted. F o r a general t r e a t m e n t from the point of view of process
d e v e l o p m e n t see J o r d a n (J2). However, the m o s t comprehensive t r e a t m e n t
is the b o o k by N a u m a n a n d Buffham (N3), in particular the last chapter,
which is devoted to scale-up.

IX. Scaling Based on T e s t s

Tests should be regarded n o t only as a crutch for process design, b u t also as


opportunities for education a n d exploration. Processing limitations can be
discovered a n d idiosyncratic behavior discerned. Because of the obscure
character of s o m e operations, laboratory studies are often necessary. E x a m -
ples are suspending solids, crystallization, emulsification, reactions, a n d
c o m p e t i n g m e c h a n i s m s where they exist (see Sections VIII,C a n d X,C,4).
Small-scale tests are also useful w h e n it is difficult or inconvenient to de-
scribe the properties of the system charge. T h e preferred technique is to
e m p l o y the actual process fluid a n d / o r solids for the test, t h u s obviating
correction factors for property differences u p o n scale-up. Similarly, when
the proposed geometry is unusual, it is simpler a n d m o r e reliable to r u n tests
in a small vessel t h a t is a m o d e l of the specified geometry t h a n to devise
elaborate correction factors for the geometric differences. Studies are also
frequently u n d e r t a k e n to find the o p t i m u m agitation system: vessel aspect
a n d type, size, a n d placement of the impeller. This exercise is carried out n o t
only for new installations b u t also to upgrade a going operation.
248 Vincent W . U h l and John A. Von Essen

A. TEST PLANNING
In an example in their book, H o l l a n d a n d C h a p m a n ( H 5 , p p . 5 8 - 6 4 )
presented a setup for securing scaling data for mixing operations. It consisted
of three metal vessels of 2 . 5 , 2 0 , a n d 160 gal. capacity for a total v o l u m e scale
ratio of 64 a n d a linear scale ratio of 4. T h e units allowed power measure­
m e n t s to be m a d e a n d were equipped with jackets for heat transfer. Desirable
test e q u i p m e n t can often be approximated by units in pilot plant or semi-
works facilities of large companies. M a n y vendors m a i n t a i n laboratories to
carry o u t performance, optimization, a n d scale-up studies for potential cus­
tomers. T h e vendor m u s t learn, or discern, from the client the process results
to be achieved. These should be defined in t e r m s of measurable properties or
observable p h e n o m e n a . Examples of the latter for solids suspension are
uniformity of density within 2% between the t o p a n d b o t t o m of the vessel,
complete suspension of solids off the b o t t o m as evidenced by the disappear­
ance of a fillet in the corner of a transparent vessel, a n d a specified residence
t i m e distribution.
It is also often necessary for clients to carry out, or assist with, the experi­
m e n t a l m e t h o d s , testing techniques, etc. for their industrial specialty.
Sometimes bench-scale tests will be r u n a n d then tests in a m u c h larger
unit, which m a y have a different geometry. T h e effect of the difference in
geometry might be attenuated by use of a n index that is not very sensitive to
geometric changes, such as T /Vor τ, as suggested in Section V. However,
Q

the small-scale tests should not be limited to ones in bench-scale e q u i p m e n t


that provide desired research results; attention should be given to a reason­
able m i n i m u m v o l u m e of charge, a geometry approximating that used in
practice, a n d a power or torque i n p u t related to the specification for a
commercial installation.
Obviously, the size of the largest test unit is limited by what is on h a n d or
what can be afforded. However, there are reservations a b o u t very small test
setups. O n e concerns the vessel size. F o r solids suspensions Lyons (L1) states
that the smallest vessel should be at least 12 in. in diameter a n d the impeller
should be n o less t h a n 4 in. across. Oldshue ( 0 4 , p . 203) says that for solids
suspension the impeller blade should be two to three times wider t h a n the
largest particle. A n o t h e r reservation a b o u t very small sizes concerns the
effect o n fluid m o t i o n of the relatively high ratio of wetted area to liquid
v o l u m e ; this causes small vessels to behave as if the liquid they contain has a
m u c h higher apparent viscosity. This was manifested by the scale-up exam­
ple in Section II illustrated by Fig. 3 a n d Table II. It was shown that in most
cases a test fluid on one scale was called for which was tailored to special
property specifications. Since ordinarily it is n o t possible to find such a fluid,
the usual technique is to use the actual process fluid for test work. If solids are
15. S c a l e - U p of Equipment for Agitating Liquids 249

involved, it is preferred that they have a corresponding sieve analysis. In the


case of a slurry which is concentrated enough to be n o n - N e w t o n i a n or
presents a classification problem, there is n o reliable substitute for the actual
m i x t u r e to be used in the commercial operation.

B. TEST PROCEDURES
W h e n a testing p r o g r a m is undertaken, it is normally desirable to first
m e a s u r e the power characteristics of the system. Either power or t o r q u e m a y
b e measured directly a n d the other calculated from the relation between
power, torque, a n d speed. F o r small e q u i p m e n t it is usually easier to measure
t o r q u e directly, b u t power is m o r e readily determined for large mixers.
While it is preferred t h a t tests be r u n with m o r e t h a n one size a n d type of
impeller, appropriate choices should be m a d e to m i n i m i z e the n u m b e r of
tests. Radial-flow impellers are usually better for dispersion, a n d axial-flow
pitched-blade turbines or hydrofoils are indicated for blending a n d solids
suspension duties. Information in Section VIII,A, in particular Table XII,
should usually be helpful in selecting reasonable ranges of values of D/T,
C/T, a n d Z/T At this point it m a y also be useful to observe the effect of
varying the s p e e d s — i n fact, to experience unsatisfactory results, i.e., learn
the limits of successful operations. Of course, the search is directed toward a n
o p t i m u m design, such as the best process result or the m i n i m u m cost, i.e., an
overall c o m b i n a t i o n of a n n u a l cost of capital a n d operating expense.
T h e procedure should include recording the data for each test in as m u c h
detail as possible. S o m e items which m a y seem to be irrelevant, inconse-
quential, or even obvious m a y prove later, at the analysis a n d c o m p u t a t i o n
stage, to be very significant. It is i m p o r t a n t that pertinent samples be taken
a n d meticulously labeled. Both the laboratory technician a n d the develop-
m e n t engineer should observe what they can of the operation: homogeneity,
surface m o t i o n , m o v e m e n t of bubbles, drops, a n d solid particles. These
observations a n d conclusions should be c o m m i t t e d to a written record c o m -
plete with supporting sketches, graphs, a n d photographs.
As a concluding note, there is a caveat. Resources should not be c o m m i t -
ted to a test program unless it is necessary or promises to generate valuable
information. O n e does n o t r u n tests w h e n a b u n d a n t a n d reliable data are at
h a n d or w h e n it seems certain that the o u t c o m e will be successful.

C . COMPUTATIONS FROM TEST DATA


At this stage the data which have been recorded are critically processed.
T h e form of the c o m p u t a t i o n s is d e t e r m i n e d by the nature of the operation:
its correspondence to the standard translation equations for conditions given
in Tables IV a n d V; its relation t o available correlations in the literature; its
250 Vincent W . U h l and John A. Von Essen

comparison with in-plant data. Knowledge, thoroughness, a n d insights are


needed at this stage to exploit the potential of costly testing efforts. Examples
of such c o m p u t a t i o n s are delineated in the appendix a n d in H o l l a n d a n d
C h a p m a n ( H 5 , p p . 5 8 - 6 3 ) . In any case, the various values that are used as
correlating indices should be calculated, such as Reynolds n u m b e r (this will
also indicate the regime), P/V, a n d T /ValongQ with its analog, the impeller
tip speed, a n d Q/H, where Η is back calculated from Eq. (5). If the system
vortices, i.e., is in the turbulent regime a n d unbaffled, or if a gas phase is
involved, the F r o u d e n u m b e r should also be c o m p u t e d . T h e Weber n u m b e r
should be found if the charge is two-phase.
While processing test data, o n e should recall that smaller laboratory-scale
systems as a rule have higher p u m p i n g capacities, shorter blend times, higher
average shear rates, a n d a narrower range of overall vessel shear rates.
Realize that ideally a scale-up calculation consists n o t only of extrapola­
tion from the laboratory data, b u t also of checks from applicable design
correlations a n d information from full-size e q u i p m e n t , when available.

X. Procedures for Scale-Up

T h e material presented so far is intended to prepare a process or develop­


m e n t engineer for scaling u p an agitation system in a substantial a n d direct
m a n n e r . However, these procedures have a n o t h e r use; they provide a basis
for scale-down (model) studies to i m p r o v e a n d optimize a commercial unit,
particularly where the pilot-scale work was inadequate. T h e actual proce­
d u r e to be followed depends o n m a n y factors: background, experience, risk,
j u d g m e n t , i m p o r t a n c e of the project, quality of the data, test facilities, time,
a n d funds. Test facilities were considered in Section I X . T h e program m a y be
considered to consist of these three phases, to be u n d e r t a k e n m o r e or less
consecutively: preliminary examination, process design calculations with
tests as required, a n d consideration of ancillary aspects to refine the design.
As a consequence of the latter phase, the design might require reworking.

A. PRELIMINARY EXAMINATION
F o r e m o s t are a n understanding a n d a specification of the process result,
preferably in writing. T h e n the operation should be characterized:
(i) Is it flow-sensitive (macro-scale mixing)? Incidentally, this accounts
for most installations.
(ii) Is it shear-sensitive (micro-scale mixing)?
(iii) Are there several operations acting in concert? If so, d o they occur
simultaneously or in series, a n d which ones d o m i n a t e or control?
(iv) Does the i m p a c t of o n e operation or material p h e n o m e n o n decline
15. S c a l e - U p of Equipment for Agitating Liquids 251

in i m p o r t a n c e a n d a n o t h e r d o m i n a t e as the process is scaled up? (Examples


are heat transfer, because in a large vessel the jacket surface can be inade­
q u a t e a n d adversely affect reaction conditions, a n d residence t i m e of dis­
persed gas can increase because of increased t i m e for disengagement result­
ing from increased height of liquids o n the plant scale.)
(v) Will the n a t u r e of the liquid regime change with scale-up? (This can
be checked by the mixing Reynolds n u m b e r . F o r most impellers the regime
is l a m i n a r w h e n the mixing Reynolds n u m b e r is less t h a n 10 a n d turbulent
w h e n it is greater t h a n 1000. This distinction is indicated clearly by a plot of
N versus N for the system, e.g., Fig. 2.)
P Re

(vi) Are there any t i m e limits? (Examples are blending, batch heating,
reaction.)
(vii) Are there space constraints? (An example is use of a plant-scale
vessel of slender a s p e c t — a high Ζ / Τ— because of lack of r o o m or lower cost
as c o m p a r e d with the square configuration c o m m o n in pilot-scale tests.)
O t h e r characteristics should also be considered, such as those peculiar to
specific processes.

B. PROCESS DESIGN
A n u m b e r of procedures can be used, b u t they should be taken u p in order
of difficulty or efficacy to generate a s o u n d design. These procedures are:
(i) Use of available equations based o n the extended principle of simi­
larity (Section III).
(ii) Selection a n d use of a n applicable measure of agitation intensity
(Section VII).
(iii) Extrapolation of the result from a single laboratory- or plant-scale
test, using a n appropriate translation equation (Section IX).
(iv) Extrapolation or interpolation of test data, generally obtained o n at
least two scales (Section IX).
(v) D e v e l o p m e n t of models for special situations.
Before discussing these alternatives, s o m e advice is in order. T h e range of
uncertainty of the result should be gauged o n the basis of experience or, if
possible, by analysis. T h e n solutions should be generated by m o r e t h a n one
avenue; checks should be sought. F o r s o m e operations, for instance, refrac­
tory g a s - l i q u i d ones, there are m a n y correlations in the literature, a n d some
are better t h a n others for certain ranges of operation. In fact, a few can lead to
deficient designs. A n example is illustrated by Fig. 10. T o reduce the risk, the
insights of knowledgeable associates within the organization should be
sought; c o m p e t e n t a n d astute consultants are n o t to be eschewed during the
d e v e l o p m e n t a n d design stage. V e n d o r s can also be helpful.
252 Vincent W . U h l and John A. Von Essen

N o w each of the suggested design paths is discussed briefly.

1. Scale-up by empirical equations. F o r m a n y operations, examples are


given in Section III; performance can be c o m p u t e d with confidence for
systems having conventional geometries. However, uncertainties are intro­
duced when the geometry is unusual, e.g., heat transfer to a nest of concentric
helical coils, power d r a w n by a n agitator in a horizontal cylindrical tank,
blend time in a t a n k with a square cross section.
2. Degree of agitation intensity. Although this concept m a y apply to all
agitated systems, it has special utility for blending a n d solids suspension
because they m a y a c c o m m o d a t e or require different levels of fluid m o t i o n .
This was demonstrated in Section VII a n d intensities of fluid m o t i o n were
r e c o m m e n d e d in Table XI for specific flow-velocity-sensitive tasks.
3. Application of translation equation. Operations listed in Tables IV
a n d V can be extrapolated by use of translation equations to the plant scale
from pilot-scale data, or from actual commercial applications, preferably
similar in geometry, fluid properties, a n d process requirements.
A special case of scaling from k n o w n plant-scale applications is encoun­
tered w h e n the size (scale) is similar. Here the design will not be affected
markedly regardless of the translation equation utilized because the scale

>-
fc:
CO

ESTIMATE
D T / V FOR
Q

COMMERCIA
L UNIT
LU

σ
ο

>

LU
CC
TWO TEST RESULT
S

LOG-LOG PLOT
I L_
11 010 0100 0
BATCH SIZE, GALLON S
FIG. 18. Plot developed from test data to extrapolate torque intensity τ to attain equal process
results on a commerical scale. It is assumed that the extrapolation is a power function as
represented by Eq. (23a).
15. S c a l e - U p of Equipment for Agitating Liquids 253

does n o t change m u c h . Such direct use of d a t a is quite c o m m o n in industry,


where information o n mixers of similar size m a y be found within the c o m ­
pany.
4. Extrapolation of test data on several scales. W h e n the required action
is n o t understood, b u t test data at two or m o r e capacities such as those
plotted in Fig. 18 can be represented by translation relations, e.g., Eqs.
(22) - (24), the a p p r o x i m a t e plant-scale design can be predicted. A n example
of this procedure is given in the appendix. T h e availability of test data o n
three or m o r e scales, the largest approaching the plant scale, permits better
extrapolations, especially when the effect of scaling does n o t plot as a straight
line o n a l o g - l o g plot, i.e., does n o t correspond to a power function. W h e n
the extrapolation from the data is considerable, as suggested by Fig. 19, the
e q u i p m e n t should be overdesigned t o provide a high certainty of meeting the
performance specifications.
5. Development of models for specific situations. This approach is gen­
eral a n d extremely vague. It m a y involve physical models, in which case the
d e v e l o p m e n t in Section II should prove useful. If mathematical models are
t o be used, approaches such as those used in Chapter 14 are called for.
However, they d e m a n d a high degree of m a t h e m a t i c a l prowess, usually m o r e
t h a n o n e can afford to expend to m e e t the challenge of a single process
application.

π1 1 Γ

b>-
CO
MI w ESTIMATE
D T / V FOR

Μ
Q

COMMERCIA
L UNIT
CO

σ
in °>
D Ο

ο^
ο
oc ο

Ρ
< ε'
> £

UJ TWO TEST RESULT


S

LOG-LOG PLOT
IΙ -
Ι1 010 0100 0
BATCH SIZE, GALLON S
FIG. 19. Conjectural plots from the test data shown in Fig. 18 which are designed to indicate
the uncertainty in a commercial-scale projection based o n only two test data points.
254 Vincent W . U h l and John A. Von Essen

In the execution of the design alternatives additional pertinent factors


m u s t often be explored. These m a y p r o m p t r e c o m p u t a t i o n or additional test
work.

C . ANCILLARY ASPECTS OF DESIGN


S o m e aspects to be e x a m i n e d while carrying out the process design or
refining the preliminary designs are the following:

1. Shift in regime. In going from the pilot plant to the plant scale, the
fluid regime m a y change, e.g., from the l a m i n a r to the transition or turbulent
regime. This might have been anticipated from the r e m a r k in Section X,A. It
is obvious that in scaling a n operation for a viscous m e d i u m the regime can
change because of the e n o r m o u s impact of D o n the mixing Reynolds
2

number.
2. Optimum geometry of plant scale. It was noted in Section VIII,A that
variations in some geometric ratios, particularly D/T, Z/T, a n d C/T, can
profoundly affect the liquid flow patterns a n d velocities in a vessel. Guides to
good practice were provided in Table XII for several operations. However,
pilot-scale tests in which these parameters are varied can indicate the pre­
ferred geometry. T h e o p t i m u m could be the m i n i m u m in t e r m s of power
input, t o r q u e investment, total a n n u a l cost, or a better process result. There
are other related variables that can be explored. A n outstanding example is
the impeller type, namely, a hydrofoil versus a R u s h t o n or a pitched-blade
turbine. In Table XII, for a single impeller, m a x i m u m values for Z/Tof 1.0
to 1.4 are r e c o m m e n d e d depending o n the operation. However, the tabu­
lated ratios are n o t sacred. For a case where a low clearance, a small C/T, is
required because of operational d e m a n d s , or where the charge is highly
viscous or pseudoplastic, a Ζ / Τ of 0.8 could be excessive.
3. Multiple impellers. W h e n Z/T is in excess of 0.8 to 1.4 as noted
above, two or m o r e impellers o n the same axial shaft are generally called for.
First the clearance C/T is established o n the basis of process needs, e.g.,
suspension of solids in the vessel b o t t o m , or operational limitations, e.g., the
m i n i m u m level of liquid in the vessel. T h e n the impellers should be equally
spaced as shown for the a r r a n g e m e n t in Fig. 17d a n d discussed in Section
VIII,B. Ideally, the height of each stage should be less t h a n square, i.e. T, to
provide some overlap of the circulating streams of the upper a n d lower
impeller in a sequence a n d ensure good distribution t h r o u g h o u t the vessel.
Lyons ( L I ) r e c o m m e n d s approximately 15% overlap of the circulation pat­
terns. This is in line with the height of a stage of a b o u t 0 . 8 T n o t e d in Section
VIII,B. Interstage circulation can also be improved, e.g., for the arrangement
in Fig. 17d, by substituting axial impellers, pitched blade or hydrofoil, for
flat-bladed turbines, particularly above the b o t t o m impeller. Elevation
15. Scale-Up of Equipment for Agitating Liquids 255

sketches d r a w n to scale as m a n d a t e d in Section IV, Β are imperative at this


stage to d e t e r m i n e the m o s t favorable impeller spacing.
4. Several processes acting in concert. These were considered in detail in
Section VIII,C u n d e r the categories:

(a) several coexisting processes where o n e d o m i n a t e s the overall result of


scale,
(b) a shift in i m p o r t a n c e of competing processes with scaling, a n d
(c) a synergistic effect between t w o or several of the c o m p a n i o n phe­
nomena.

Because of the uncertainty they can introduce, processes in categories (b) a n d


(c) are to be especially discerned a n d allowed for if possible. In the absence of
experience on the large scale, b o t h can jeopardize the final result. They call
for t h o r o u g h study a n d allowance for uncertainty in the design. Unfortu­
nately, for situations described by these two classifications, the solution m a y
b e c o m e a p p a r e n t only in t e r m s of the success of the commercial operation.
5. Shear rates controlling. F o r a small proportion of agitated vessels,
such as reactors, fermentors, a n d dispersers, the shear rate is controlling.
Inspection of Fig. 9 shows how, with geometrically similar scale-up, t w o
shear rate parameters, the m a x i m u m a n d the average for the impeller zone,
diverge. F o r s o m e services it might be essential to reduce this difference.
Ways to d o so were suggested in Section IV,C.
6. Torque versus power investment. It was demonstrated above, particu­
larly in Section VI, that for flow-sensitive operations, t o r q u e i n p u t is control­
ling, a n d o n scale-up the power intensity decreases with a characteristic
length L, which serves as a linear scaling dimension. But higher torque,
which calls for a larger speed reducer, with less power i n p u t a n d hence lower
operating expense m a y cost m o r e overall t h a n a cheaper device with a higher
power i n p u t operating at higher speed b u t developing a lower torque. T h e
problem here is t o find the system t h a t produces the required process result
with the lowest overall a n n u a l cost. This trade-off was introduced in Section
VIII,A.
7. Mechanical design considerations. These can be critical a n d m u s t be
t a k e n into account before firming u p a design. T w o examples are noted here
to illustrate the kinds of trade-offs which can occur. For small units cast
impellers are c o m m o n a n d pose n o problem, b u t w h e n the unit is scaled u p a
cast impeller is relatively heavy a n d therefore can be b o t h expensive to
purchase a n d exorbitant to a c c o m m o d a t e . Also a pilot-scale u n i t operating
at high speeds, because it is small, is likely to exhibit a critical speed limita­
tion, whereas a commercial-size unit m a y not. This could cause the pilot-
plant unit to be operated at a higher D / T r a t i o , e.g., 0.5, t h a n is economically
indicated for the commercial scale.
256 Vincent W . U h l and John A. Von Essen

X I . Conclusions
Persons involved with process mixing technology, whether in research,
development, design, or operations, should have an understanding of the
n a t u r e of scale-up a n d the background it requires. Their specific functions in
research a n d development have been frequently alluded to in this chapter.
Particular tasks in which scaling c o m p e t e n c e proves useful are: specifying
pilot-scale programs, interpreting results obtained with small equipment,
preparing inquiries for a n d c o m m u n i c a t i n g with vendors, checking quota-
tions, using consultants, troubleshooting a n d analyzing plant operations.
Frequently, scale-up or scale-down (often t e r m e d modeling) tasks are
challenging a n d d e m a n d i n g because they require the orchestration of m a n y
pertinent resources with skill a n d j u d g m e n t . This is a p r i m e example of
technical synergism at work, of the art of engineering. T h e t r e a t m e n t in this
chapter emphasizes the need for knowledge, conversance with techniques,
availability of testing facilities, a n d skilled technologists. Background is
needed in hydrodynamics, the principles of similarity, a n d mixing technol-
ogy. T h e techniques include the traditional scale-up procedures which draw
o n theory, a n d actual results obtained o n both the small a n d plant scales. T h e
testing facilities m u s t be used wisely a n d efficiently. T h e competence a n d
skill of the development engineers are exhibited by their ability to define the
problem, devise approaches, a n d m a k e keen observations as well as sound
decisions.
Certain aspects of the scale-up task require special attention. O n e is the
uncertainty inherent in any engineering solution. Sometimes the uncer-
tainty is low, for instance, for blending, which requires a quantity, not a n
intensity, of agitation, as discussed in Section VII; with some extra t i m e
blending can be attained in a n inadequately sized blend tank. But in other
operations the agitation intensity can be critical—for example in reaction,
where a degree of micromixing is absolutely necessary to attain the specified
yield a n d m i n i m i z a t i o n of side reactions. A n o t h e r example would be failure
to suspend solids, resulting in "sanding i n " of the impellers, perhaps shutting
d o w n the operation, or causing e q u i p m e n t failure. These examples support
the d i c t u m ascribed to A. P. Colburn: " A n y engineering solution at best is
only approximately correct, b u t it m u s t never be absolutely w r o n g . "
It takes special attention to avert a n occasional catastrophe. As part of the
analysis, the p r o b l e m areas should be identified a n d the degree of risk evalu-
ated. It m u s t be recognized that test results are frequently less t h a n conclu-
sive, especially if they are bench-scale results intended for research a n d not
carried o u t in a suitable m o d e l for extrapolation to a commercial-size unit.
T h e n , m a n y correlations are only approximately correct at best, a n d m a y be
in error for certain conditions (see Fig. 10). But by checking the results of o n e
15. S c a l e - U p of Equipment for Agitating Liquids 257

m e t h o d against those of another, n o t only is understanding enhanced b u t


also the risk is reduced, it is h o p e d to a n acceptable limit.
T h e practice as delineated in this chapter should provide a useful guide for
sound, efficient scaling, u p or down, of mixing operations.

Appendix. Exercise to Demonstrate T h a t the Four


Translation Equations Predict the S a m e Scale-Up Result

T h e A B C C o m p a n y has obtained laboratory- a n d pilot-scale data for a


complex mixing application a n d n o w wants to evaluate a mixer offered for a
large vessel. A chemical conversion efficiency of 6 5 % is desired. Both labora­
tory- a n d pilot-scale mixers were r u n at variable speed a n d set just fast
enough to obtain 6 5 % conversion. T h e charge has a specific gravity of 1.2
a n d viscosity μ of 100 c P .
Laboratory (L) vessel 12 in. diameter X 12 in. liquid level
impeller four 45° pitched-blade turbine (PBT), 5 in. di­
ameter
speed 643 r p m
Bench (B) vessel 36 in. diameter X 36 in. liquid level
impeller four 45° P B T , 15 in. diameter
speed 258 r p m
Plant (P) vessel 8 ft diameter X 8 ft liquid level
impeller four 45° P B T , 40 in. diameter

N o t e t h a t for all three cases geometric similarities are maintained, a n d there­


fore D/T is constant a n d equal to 1/3.

SOLUTION
1. Try a dimensionless number correlation. Use ψ = y/'D = K(N ) , Re
a

which is Eq. (25), where y/' = 0.65 for chemical conversion efficiency. F o r
L

the laboratory-scale case L

£>L = A = 0.417 ft
N= 643 r p m
Sg=1.2

μ = 100 c P
T o facilitate calculation of the mixing Reynolds n u m b e r , this dimensional
form of the Reynolds n u m b e r will be used:
N Rc = 1480£> MSg///
2
258 Vincent W . U h l and John A. Von Essen

where D is in feet, Ν in revolutions per m i n u t e , for the density ρ specific


gravity (Sg) is used, a n d μ is in centipoise. T h e n

(NRX = 1480 X 1.2 X 643 X (0.417) /100 = 1986 2

Similarly for the bench scale case Β

(NRX = 1480 X 1.2 X 258 X (1.25) /100 = 7160 2

Then

" ( ^ R e ) L l g
_ ^ L

.(^Re) J B VBA*
_ Ιη(ψ'Μψ' Ρ ) Β Β _ ln(0.65 X 0.417/0.65 X 1.25) _
a
H(N )J(N ) ] Rc Rc B ln(1986/7160)

ψΡ Β Β _ 0.65 X 1.25 _
( Μ 8 6
(7160) · 0 86
^ A 1 U

T h u s , for the plant case Ρ where D = 4 0 / 1 2 = 3.33 in ψ'Ό = K(N ) P Ke


a
from
above a n d the conversion ψ' = 0.65,

0.65 X 3.33 = 3.93 X 1 0 ~ ( i V ) g 4


Re
86

or

(^Re)p
=
2.24 Χ 1 0 4
(turbulent regime)

and

1480 X 1.2 X W X (3.33) /100 = 2.24 Χ 1 0 2 4

N= 114 r p m
2. Try a power per unit volume correlation. T h e power requirement is
calculated from

P = N SgN D /(6A2
P
3 5
Χ 10 ) 7

for which the values of Sg, N, a n d D are given in the same units as in the
dimensional Reynolds n u m b e r equation above. T h e n since N = 1.62 from P

a n appropriate N versus iV correlation P Re

P = 1.62 X 1.2 X (643) (0.417) /(6.12 χ 10 )


L
3 5 7

P = 0.106 h p
L

F = 7.48^/4)71Z
L L

V = 7.48 X 0.785 Χ (1.0) X 1.0 = 5.87 gal.


L
2
15. Scale-Up of Equipment for Agitating Liquids 259

(P/*\ = 0.106 hp/5.87 gal X K T 3

W )L=
/
18.1 h p / 1 0 0 0 gal
Similarly, (P/V) B = 10.5 h p / 1 0 0 0 gal

(p/B) B = (P/v) (T /T y
L B L

1 0 . 5 = 18.1(3/1)'
j/ = -0.50

Then
(P/V) P = (P/V) (T /T )-<>*>
B P B

(P/V) P = 10.5(8/3)-°- = 6.43 h p / 1 0 0 0 gal.


50

F = 3008 gal.
P

P = 6.43 X 3008/1000 = 19.3 h p


P

Substituting these values in the power requirement equation above

Γ 6.12 X 1 0 X 19.3 7
Il l / 3
1

[ 1.62 X 1.2 X ( 4 0 / 1 2 ) . 5

TV = 114 r p m

3. Try a torque per unit volume correlation.


T Q = 63,000P/N
where Ρ is in horsepower a n d Ν in revolutions per m i n u t e . T h e n
( T ) = 63,000 X 0.106/643 = 10.4 in. lb
e L

(T /V)
Q L = 10.4/5.87 = 1.77 in. lb/gal.

Similarly,
(T /V)
Q B = 2.56 in. lb/gal.
(T /V)
Q B = (T /V) (T /T y
Q L B L

or
2.56 = 1.77(3/1)*

and
JC = 0.34
260 Vincent W . U h l and J o h n A. Von Essen

F r o m this
(T /V)
Q P = (T /V) (T /TX*
Q B P

(T /V)
Q P = 2.56(8.3) 034
= 3.57 in. lb/gal.

(T )
Q P = 3.57 X 3008 = 10,739 in. lb

Γ 971T Q I 0 5
= Γ 971 X 10,739 > 5

LWpSg£ J 5
Ll-62 X 1.2(40/12) J5

and
TV = 114 r p m

4. Try a speed ratio correlation.

N =
B N (T /T )-»
L B L

258 = 643(3/1)-»

« = 0.83

N =
L N (T /T r™*
B P B

N = 258(8/3Γ ·
L
0 8 3

N= 114 r p m

CONCLUSION
All four m e t h o d s predict t h e same plant-scale impeller speed. If D/T or
Ζ IΤ ratios were n o t identical, t h e different m e t h o d s would have predicted
different plant-scale impeller speeds, a n d in that case, as demonstrated, t h e
t o r q u e per unit v o l u m e scale-up would have been most reliable.

List of Symbols
A horizontal cross-sectional area of inside of a cylindrical vessel
Β width o f baffle
C distance from bottom o f impeller to vessel bottom
C p specific heat
d 0 outside diameter o f heat transfer coil in a vessel
D impeller diameter
DL liquid phase diffusivity
F force
g acceleration due to gravity
g c gravitation conversion factor in equation F = ma/g c

h heat transfer coefficient


Η head, a measure o f fluid shear
k thermal conductivity
k s mass transfer coefficient
15. S c a l e - U p of Equipment for Agitating Liquids 261

K, K,
x 2 . . . , Kn constants
L characteristic length
Ν impeller rotational speed
Ρ power response or invested in mixer fluid
Q impeller volumetric discharge rate
Q effective volumetric p u m p i n g rate of an impeller
Q* volumetric gas flow rate into vessel
sg specific gravity
τ vessel diameter
torque response or input to mixer fluid
«t terminal settling velocity of particle in liquid
liquid velocity
ν apparent superficial velocity, Q/A
fluid v o l u m e in vessel
V width of impeller blade (see Fig. 7)
w depth of vortex (see Fig. 3)
Y fluid depth in vessel; height of stage in multi-impeller vessel
ζ specific weight (see Table II)
y fluid shear rate, units of reciprocal time
y denotes difference, e.g., AH
Δ intensity of power input, P/V
e blend t i m e
Θ operating time related to number of turnovers
β viscosity
ν kinematic viscosity, μ/ρ
Ρ density
σ surface tension
τ intensity of torque input, T /V Q

Ψ function (see Fig. 4)


Ψ' ψ/D; used in example in appendix

SUPERSCRIPTS
α exponent for Reynolds number in various correlations
η exponent for speed ratio N /N , Eq. (24) (see Table VI)
2 x

Ρ exponent for Prandt number in Fig. 4


s exponent for Schmidt number in Fig. 4
χ exponent for torque intensity translation equation, Eq. (23a) (see Table
VI)
y exponent for power intensity translation equation, Eq. (22) (see Table VI)
ζ exponent frequently for (D/T)

SUBSCRIPTS
av average
i inertia
max maximum
Μ model, same as subscript 1
Ρ prototype, same as subscript 2
r ratio of condition 2 to condition 1
1 small scale (initial size)
2 large scale (final size)
262 Vincent W . U h l and J o h n A. Von Essen

DIMENSIONLESS GROUPS
N Froude number, DN /g
Fr
2

Ni agitation intensity number; for flow-sensitive operations corresponds to


v /(6 ft/min)
b

N N e w t o n or power number,
P PgJpN D 3 5

N impeller discharge coefficient, q/ND


q
3

N p u m p i n g number Q/ND
Q
3

Prandtl number, ( C / / ) / kp

iV mixing Reynolds number,


Re D Np/p 2

Schmidt number, p/pD L

N mixing Weber number,


W E N D p/a 2 3

t blend time number, ΝΘ


b

References

(Al) Aerstin, F., and Street, G., "Applied Chemical Process Design." Plenum, N e w York,
1978.
(A2) A n o n . , "Hydraulic Models," Manual of Engineering Practice N o . 25. American Society
of Civil Engineers, N e w York, 1942.
(B1) Baasel, W. D . , "Preliminary Chemical Engineering Plant Design," p. 115. Elsevier, N e w
York, 1976.
(B2) Bates, R. L., Fondy, P. L., and Fenic, J. G., "Mixing: Theory and Practice" (V. W. U h l
and J. B. Gray, eds.), Vol. I, Chap. 3. Academic Press, N e w York, 1966.
(B3) Bathija, P. R., Chem. Eng. 89(25), 89 (1982).
(B4) Bowen, R. L , Jr., Chem. Eng. 92(6), 159 (1985).
(B5) Buche, W., VDI Z. 3 7 , 1065 (1937).
(CI) Connolly, J. R., and Winter, R. L., Chem. Eng. Prog. 65(8), 70 (1969).
(Dl) Dickey, D . S., and Fenic, J. G., Chem. Eng. 83(1), 139 (1976).
(El) Edwards, M. F., and Wilkinson, W. L., Chem. Eng. (British), N o . 264, 3 1 0 - 3 1 9 (1972).
(E2) Edwards, M. F., and Wilkinson, W. L., Chem. Eng. (British), N o . 2 6 5 , 3 2 8 - 3 3 5 (1972).
(E3) Einsle, Α., Paper 4.13, Fifth International Fermentation Symposium, Berlin, 1976.
(F1) Fenic, J. G., and Fondy, P. L., Paper presented at Annual Meeting of the AIChE, Atlantic
City, N.J., 1966.
(Gl) Gates, L. E., Morton, J. R., and Fondy, P. L., Chem. Eng. 83(11), 144 (1976).
(G2) Gates, L. E., Hicks, R. W., and Dickey, D . S., Chem. Eng. 83(26), 165 (1976).
(G3) Gray, J. B., "Mixing: Theory and Practice" (V. W. U h l and J. B. Gray, eds.), Vol. I, Chap.
4. Academic Press, N e w York, 1966.
(HI) Hall, C. M., Metall. Chem. Eng. 9, 71 (1911).
(H2) Hicks, R. W., and Gates, L. E., Chem. Eng. 83(15), 141 (1976).
(H3) Hicks, R. W., Morton, J. R., and Fenic, J. G., Chem. Eng. 83(9), 102 (1976).
(H4) Hirsekorn, F. S., and Miller, S. Α., Chem. Eng. Prog. 4 9 , 459 (1953).
(H5) Holland, F. Α., and Chapman, F. S., "Liquid Mixing and Processing in Stirred Tanks."
Reinhold, N e w York, 1966.
(H6) H y m a n , D . , "Advances in Chemical Engineering" (Τ. B. Drew, J. W. Hoopes, and
T. Vermeulen, eds.), Vol. III. Academic Press, N e w York, 1962.
15. S c a l e - U p of Equipment for Agitating Liquids 263

(Jl) Johnstone, R. E., and Thring, M. W., "Pilot Plants, Models and Scale-Up Methods."
McGraw-Hill, N e w York, 1957.
(J2) Jordan, D . G., "Chemical Process D e v e l o p m e n t , " Part 1. Interscience, N e w York, 1968.
(Kl) Khang, S. J., and Levenspiel, O., Chem. Eng. 83(21), 141 (1976).
(K2) Khang, S. J., and Levenspiel, O., Chem. Eng. Sci. 3 1 , 569 (1976).
(LI) Lyons, E. J., "Mixing: Theory and Practice" (V. W. Uhl and J. B. Gray, eds.), Vol. II,
Chap. 9. Academic Press, N e w York, 1967.
(Ml) Marshall, J. C , and Yazdani, N., Chem. Process Eng. 5 1 , 89 (1970).
(M2) McAllister, S. H., Oil Gas J., 139 (12 N o v e m b e r 1937).
(M3) McCabe, W. L., Smith, J. C , and Harriott, P., "The U n i t Operations of Chemical
Engineering," 4th ed. McGraw-Hill, N e w York, 1985.
(M4) Metzner, Α. B., and Taylor, J. S., Am. Inst. Chem. Eng. J. 6(1), 109 (1960).
(M5) Miller, S. Α., and Mann, C. Α., Trans. Am. Inst. Chem. Eng. 4 0 , 709 (1944).
(Nl) Nagata, S., "Mixing," p. 200. Wiley, N e w York, 1975.
(N2) Nagata, S., Yamagimota, M., and Y o k o y a m a , T., Chem. Eng. (Tokyo) 2 1 , 278 (1957).
(N3) N a u m a n , Ε. B., and Buffham, Β. Α., "Mixing in Continuous Flow Systems." Wiley, N e w
York, 1983.
(N4) Nienow, A. W., Warmoeskerken, Μ. M. C. G., Smith, J. M., and K o n n o , M., Paper F l ,
Second European Conference o n Mixing, pp. 1 - 1 6 . B H R A Fluid Engineering, Cran­
field, Bedford, England, 1977.
(01) Oldshue, J. Y., Chem. Process Eng. 4 7 , 183 (1966).
(02) Oldshue, J. Y., Biotechnol. Bioeng. 3(1), 3 (1966).
(03) Oldshue, J. Y., "The Spectrum of Fluid Shear in a Mixing Vessel," Chemeca '70, pp.
9 9 - 1 1 0 . Butterworth, London, 1971.
(04) Oldshue, J. Y., "Fluid Mixing Technology." McGraw-Hill Publ., N e w York, 1983.
(05) Olson, J. H., and Stout, L. E., Jr., "Mixing: Theory and Practice" (V. W. U h l and J. B.
Gray, eds.), Vol. II, Chap. 7. Academic Press, N e w York, 1967.
(PI) Penney, W. R., Chem. Eng. 78(7), 86 (1971).
(P2) Perry, R. H., Green, D . W., and Maloney, J. O., eds., "Chemical Engineers Handbook,"
6th ed., Section 10, p. 19. McGraw-Hill, N e w York 1984.
(P3) The Pfaudler Co., Bull. 1018, "Agitator Speed-Power Calculator." Rochester, N.Y.,
1961.
(P4) Philadelphia Gear Corp., "Philadelphia Mixers: Energy Conservation in Fluid Mixing"
(15-page booklet with no identifying number). King of Prussia, Pa., 1980.
(Rl) Rautzen, R. R., Corpstein, R. R., and Dickey, D . S., Chem. Eng. 83(23), 119 (1976).
(R2) Revill, Β. K., Paper B l , Fourth European Conference o n Mixing, pp. 1 1 - 2 4 . B H R A
Fluid Engineering, Cranfield, Bedford, England, 1982.
(R3) Richards, J. W., Br. Chem. Eng. 8(3), 158 (1963).
(R4) Rushton, J. H., "Application of Fluid Mechanics and Similitude to Scale-up Problems,"
S y m p o s i u m o n Relationship between Pilot-Scale and Commercial Chemical Engineer­
ing Equipment. AIChE, White Sulphur Springs Meeting, 1 1 - 1 4 March 1951.
(R5) Rushton, J. H., Proc. 2nd Midwestern Conf. Fluid Dynamics, pp. 1 5 6 - 1 7 4 (1951).
(R6) Rushton, J. H., Chem. Eng. Prog. 47(9), 485 (1951).
(R7) Rushton, J. H., Ind. Eng. Chem. 4 4 , 2931 (1952).
(R8) Rushton, J. H., Costich, E. W., and Everett, H. J., Chem. Eng. Prog. 46(1), 395; 46(11), 467
(1950).
(R9) Rushton, J. H., and Oldshue, J. Y., Chem. Eng. Prog. Symp. Ser. No. 4 4 9 , 161 (1953).
(S1) Skelland, A. H. P., " N o n - N e w t o n i a n Flow and Heat Transfer." Wiley, N e w York, 1967.
(Tl) Tatterson, G. B., Food Technol. 25, 6 5 - 7 0 (May 1971).
264 Vincent W . U h l and John A. Von Essen

( U 1 ) Uhl, V. W., "Mixing: Theory and Practice" (V. W. U h l and J. B. Gray, eds.), Vol. I, Chap.
5. Academic Press, N e w York, 1966.
( V I ) Valentin, F. Η. H., Br. Chem. Eng. 12(8), 1213 (1967).
(V2) van't Riet, K., and Smith, J. M., Chem. Eng. Sci. 28, 1931 (1973).
(V3) V o n Essen, J. Α., Lecture Notes: Scale-up Procedures, Short Course o n Liquid Mixing.
Center for Professional Advancement, East Brunswick, N.J., 1980.
( W l ) Warmoeskerken, Μ. M. C. G., and Smith, J. M., Chem. Eng. Sci. 40(11), 2 0 6 3 (1985).
(W2) Weber, A. P., Chem. Eng. 70(18), 91 (1963).
CHAPTER 16

Mixing of Particulate Solids


J o h n C. Williams
Postgraduate School of Powder Technology
University of Bradford
West Yorkshire BD7 1DP, England

I. Introduction

Although it is o n e of the oldest industrial operations a n d plays a n impor-


t a n t part today in a very wide range of process industries, the mixing of
particulate solids is a subject t h a t has n o t been systematically studied to the
same extent as m o s t other aspects of process engineering a n d it is still n o t well
understood. D u r i n g the past 30 years there has been m u c h work d o n e at
universities in t h e study of solids mixing, b u t the results of this effort are not
yet widely applied in industrial practice. Attitudes are still frequently deter-
m i n e d by experience of liquids mixing, although it can readily be d e m o n -
strated that techniques successful in the mixing of liquids may, in certain
cases, be entirely unsuitable for the mixing of solid particles.
Research has been concerned with the application of statistical m e t h o d s to
the description of the quality of a mixture a n d the tendency of a mixture of
free-flowing particles of different properties to segregate o n handling. T h e
use of statistics in particle mixing was first established by Lacey ( L 1 , L2), a n d
it has been shown by H a r n b y (H3) that the statistical m e t h o d s developed for
dealing with the results of social surveys are applicable to the problems of
describing particle mixtures. S o m e knowledge of statistics is necessary for a
p r o p e r understanding of the mixing process, particularly in dealing with
information obtained by analyzing samples taken from a mixture. T h e aim
in this chapter is to set o u t the basic ideas which are considered to be essential
in the interpretation of such test data.
T h e fact t h a t particles of different physical properties will tend t o segregate

265
MIXING: THEORY AND PRACTICE, VOL. Ill Copyright © 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
266 J o h n C. Williams

has been given considerable attention, b u t its implications have n o t been


widely realized. T h e prevalence of such fallacies as a belief that mixing
quality will always improve if mixing t i m e is increased or that all types of
mixer are equally suitable for all mixing duties is evidence of this. It is
i m p o r t a n t in selecting e q u i p m e n t a n d designing processes, which include
solids mixing, that the causes of segregation be understood so that its conse-
quences can be minimized.
T h e further sections in this chapter deal with selecting the type of mixer to
be used for a given duty, the m e t h o d s used for specifying the performance
required from a mixer, a n d testing to ensure that the mixing quality achieved
is satisfactory.
T h e usual practice in designing processes t h a t involve solid particle mixing
is to use batch mixing, even when the rest of the process is continuous. Apart
from the e c o n o m i c advantages of using c o n t i n u o u s mixing, there are techni-
cal reasons why it should be preferred, especially when dealing with free-
flowing a n d easily segregating materials. In this chapter the literature on
c o n t i n u o u s solids mixing is discussed.
Recent work has been concerned with the mixing of cohesive materials,
where segregation does n o t occur; mixing is easier if the correct type of mixer
is used. T h e problems arising w h e n the scale of scrutiny is very small are also
discussed.
Publications to which particular reference is m a d e in this chapter are listed
u n d e r References. A m o r e general survey of the literature of solids mixing
can be found in C o o k e et al (C3).

II. Statistics of Solids M i x i n g

A perfect mixture of two or m o r e kinds of solid particles is one in which


any set of contiguous particles, removed from any part of the mixture,
contains the same proportions of the c o m p o n e n t s as the proportions present
in the whole of the mixture. In practice, such a quality of mixing c a n n o t be
achieved.
Generally the a i m is to p r o d u c e a r a n d o m mixture, that is, one in which
the probability of finding a particle of a given c o m p o n e n t is the same at all
points in the mixture a n d is equal to the proportion of that c o m p o n e n t in the
whole mixture. W h e n segregation occurs or w h e n mixing is not complete,
the particles of o n e c o m p o n e n t have a preference for being in one part of a
mixture rather t h a n in another, so that the condition for r a n d o m mixing is
n o t satisfied.
T h e quality of a mixture can be examined by removing a n u m b e r of sets of
particles from different positions in the mixture a n d determining the c o m -
position of each set of particles. In the terminology used in the literature of
16. M i x i n g of Particulate Solids 267

solids mixing, each set of continguous particles r e m o v e d is referred to as a


sample a n d the size of a sample is the n u m b e r of particles it contains. T h e
n u m b e r of sets removed in examining a m i x t u r e is the n u m b e r of samples.
In the simple case of a r a n d o m m i x t u r e of black a n d white particles in
which all particles are of the same size, the fractions of black particles in a
n u m b e r of samples will follow the b i n o m i a l distribution; therefore if P i s the
fraction of black particles in the mixture a n d a sample containing η particles
is removed, the probability that the sample will contain m black particles is

U n d e r conditions that are almost always satisfied in practical powder


mixing it can be assumed that the compositions of samples taken from a
r a n d o m mixture will be normally distributed. Since a state of r a n d o m mix­
ing is generally the aim of industrial particle mixing a n d generally represents
t h e best mixing quality that can be achieved, the properties of a r a n d o m
m i x t u r e are of particular interest; they will be discussed later.

A . ESTIMATION OF THE COMPOSITION OF A MIXTURE


W h e n the composition of a mixture is n o t k n o w n it can be estimated from
the compositions of a set of samples. If y , y , . . . ,y represent the frac­
x 2 N

tions of one c o m p o n e n t in TV samples, the estimated m e a n composition y for


the whole mixture is given by

(1)

Since this estimate is based o n the limited information obtained by sam­


pling, y will n o t in general be the true value for the composition of the
mixture. If the samples were r e m o v e d from a r a n d o m mixture the confi­
dence limits for the true m e a n μ would be given by

(2)
where t is the value of Student's t, which can be obtained from tables of
statistical functions, a n d S is the standard deviation of the sample composi­
tions.
In selecting the value of t it is necessary to specify the degree of confidence
required a n d the n u m b e r of degrees of freedom, which is o n e less t h a n the
n u m b e r of samples taken.
F o r example, if 10 samples are removed from a mixture for which y = 4 5 %
a n d S = 2%, the value of / for 9 degrees of freedom a n d 9 5 % confidence
limits is 2.26; t h e n
μ = 45 ± 2.26(2/χ/ΪΟ) = 45 ± 1.43%
268 J o h n C. Williams

T h e r e is therefore a 9 5 % probability that the composition of the mixture lies


between 43.57 a n d 46.43%.
If the composition of the mixture is already k n o w n , the Student t test can
be used to assess the probability that the sampling procedures are biased. In
practice it m a y be difficult to remove samples from a mixture without
preferentially collecting the particles of o n e c o m p o n e n t ; it is therefore advis­
able to apply a statistical test to the results to check for bias.
In the preceding example, if the true fraction of one c o m p o n e n t in the
mixture is k n o w n to be 48%, which lies outside the range 45 ± 1.43%, this
would indicate a high probability that the m e t h o d used to obtain or analyze
the samples was subject to bias a n d the cause of this bias should be identified
so that it can be eliminated.

B. STANDARD DEVIATION AS A MEASURE OF MIXING


QUALITY
W h e n samples are withdrawn from a mixture a n d analyzed they will have
different compositions, even if analytical errors are negligibly small. T h e
extent to which they differ will be a n indication of the quality of the mixture.
T h e spread in the compositions of the samples is usually characterized by
their standard deviation, which is used as a measure of the quality of the
mixture. A lower standard deviation indicates better mixing.
F o r a m u l t i c o m p o n e n t mixture containing c c o m p o n e n t s a complete
description of the mixing quality requires (c— 1) standard deviations. In
practice sufficient information can usually be obtained by selecting o n e or
m o r e c o m p o n e n t s whose dispersion is most i m p o r t a n t , or which are k n o w n
to be the m o s t difficult to disperse.
If consideration is limited to a binary mixture, two cases arise in evaluating
the standard deviation of the fractions of one c o m p o n e n t in the samples,
depending o n whether or n o t the composition of the whole of the mixture is
k n o w n . If y y ,
u 2 . . . , represent the fractions of o n e c o m p o n e n t in the Ν
samples, the standard deviation S of the sample compositions is given by
either

^^[Σ^/-^ ]/^ 2
( 3
)
where μ is the k n o w n fraction of the c o m p o n e n t in the whole mixture, or

5 = [i(X-J) ]/(JV-l)
2 2
(4)
where the composition of the mixture is not k n o w n a n d has to be estimated
as y [defined in Eq. (1)].
16. M i x i n g of Particulate Solids 269

T h e quality of mixing m a y also be characterized by the variance S of the 2

sample compositions, or by the coefficient of variation C , which is the value


v

of the standard deviation divided by the m e a n value of the composition


( C = SI μ or S/x).
v

If 10 samples are r e m o v e d from a mixture, t h e standard deviation of the


sample compositions S provides a n estimate of the true value of the standard
deviation σ. If a different set of 10 samples is r e m o v e d a n d analyzed, they will
in general give a different value for the estimate of the standard deviation.
T h e confidence that can be placed in a n estimate of the standard deviation
will be discussed later.
W h e n samples are d r a w n from a r a n d o m mixture, the sample composi­
tions will be normally distributed. It can t h e n be assumed, from the proper­
ties of the n o r m a l distribution, t h a t 95.4% of samples will have compositions
within the range μ±2σ a n d 99.7% will be within the range μ ± 3<r. For
example, if a manufacturing system produces packages containing a mixture
for which t h e m e a n p r o p o r t i o n of c o m p o n e n t A is 4 0 % a n d the standard
deviation has been found to be 1.5%, it is expected that 4.6% of the packages
p r o d u c e d will have compositions outside the range 37 to 4 3 % a n d 0.3% of the
packages will be outside the range 35.5 to 44.5%. These values apply only if
the sample compositions are normally distributed, which c a n n o t always be
a s s u m e d to be the case. Before m a k i n g this assumption the experimental
values for the sample compositions should be tested for normality; the
m e t h o d of doing this is discussed by Snedecor a n d C o c h r a n (S5).

C. PROPERTIES OF RANDOM MIXTURES


T h e properties of a r a n d o m mixture of particles are of particular interest.
Lacey ( L 1 , L2) was the first to point o u t that the quality of a mixture could be
described by finding the standard deviation of the compositions of a set of
samples. H e derived a n expression for the true standard deviation σ of a κ

r a n d o m m i x t u r e containing fractions Ρ a n d (1 — P) of two c o m p o n e n t s ; he


assumed that all particles are of the same size a n d that each sample removed
from the m i x t u r e contains η particles. T h e result was
σ =[Ρ(1-Ρ)/η]ν
κ
2
(5)
If a r a n d o m binary m i x t u r e contains a fraction 0.1 by mass of o n e c o m p o ­
n e n t a n d samples containing 10,000 particles are removed, the standard
deviation of the sample composition σ will be ((0.1 X 0 . 9 ) / 1 0 , 0 0 0 )
κ =
1/2

0.003.
F o r a system where the two c o m p o n e n t s are completely u n m i x e d the
standard deviation σ of sample composition is
0

a 0 = [W-P)] L / 2
(6)
270 J o h n C. Williams

In the above example the value of σ will be (0.1 X 0 . 9 ) = 0.3.


0
1/2

Usually all the particles in a m i x t u r e are n o t of the same size a n d each


c o m p o n e n t has a size distribution. In that case Eq. (5) still applies b u t it is
necessary to estimate the expected n u m b e r of particles in a sample from the
size distribution functions F(x) by mass of the two materials.
T h e m e t h o d of estimating the n u m b e r of particles in a sample is to plot
F(x) against l/x , where F(x) is the fraction by weight of the material less
3

t h a n size x, for each of the materials (Fig. 1). T h e shaded area to the left of the
curve, between F(x) = 0 a n d 1, gives the value of 1/x , where χ is the m e a n
3

particle size. F o r spherical particles the m e a n particle weight w is equal to


πρςΧ /6, where p is the particle density. This calculation is carried o u t for
3
s

each c o m p o n e n t . F o r a binary m i x t u r e containing fractions P a n d P of the


a b

c o m p o n e n t s the expected n u m b e r of particles in a sample of weight W is


given by

n=W(PJw a + P /w )
b b (7)
a n d from Eq. (5)

<7R = (WO 1/2 (8)

10 1 11 21 3

1 x 1 0 " ( x in c m)
3

FIG. 1. Calculation of m e a n particle mass. Shaded area to left of graph is equal to / " = 0 dF/x .
3

This equals \/x , where χ is the m e a n particle size.


3
16. M i x i n g of Particulate Solids 271

T h e m e t h o d is illustrated by the following worked example.


A r a n d o m mixture contains two c o m p o n e n t s , A a n d B. Both consist of
spherical particles a n d their proportions are 60% by weight of A a n d 40% of
B. Their size gradings are given in the following table a n d their densities are
respectively 500 a n d 700 k g / m . If a n u m b e r of samples of weight 1 g are
3

withdrawn from the mixture, what is the expected value for the standard
deviation of the composition of the samples?

Size χ (μηι) 2057 1676 1405 1204 1003 853 699 599 500 422

Fraction less than size,


m
A 1.00 0.80 0.50 0.32 0.19 0.12 0.07 0.04 0.02 0
Β — — 1.00 0.88 0.68 0.44 0.21 0.08 0
1/jc ( c m " )
3 3
115 212 361 573 991 1611 2928 4653 8000 13306

Values of F(x) are plotted against 1/x for the two materials A a n d Β (Fig.
3

1), a n d in each case the area to the left of the curve between F(x) = 0 a n d 1 is
found. T h e results are as follows:
A Β

Area t o left of curve = np /6 w = l/x s


3
891 1960
M e a n particle weight w (//g) 294 187

Then

n = 106 x
(i^ iw)+ = 4 1 8 0 for w = l 0 6μ %

and

*,-V^§F = 0.0076 or 0.76%


It has already been pointed o u t that if we remove, say, 10 samples from a
m i x t u r e a n d find the standard deviation of the sample compositions, this
gives a n estimate S of the true standard deviation σ of the mixture. If we take
a n o t h e r set of 10 samples we will obtain a different value for S. T h e practical
p r o b l e m to be considered is as follows: if we remove o n e set of samples from
the mixture a n d find the value of the standard deviation of the sample
compositions 5 , what can we say a b o u t the value of the true standard devia­
tion σ? If we are sampling from a r a n d o m m i x t u r e the compositions of the
samples will be normally distributed; the values of the estimates S obtained
by repeating o u r selection of a set of samples a large n u m b e r of times will fit a
272 J o h n C. Williams

chi-squared distribution, with a n u m b e r of degrees of freedom equal to the


n u m b e r of degrees of freedom used in calculating the standard deviation of
the compositions of the set of samples. If 10 samples are taken, the n u m b e r of
degrees of freedom will be either 10 or 9, depending o n whether the m e a n
composition of the mixture is k n o w n . Since the properties of the chi-squared
distribution are available in statistical tables, this provides information
a b o u t the spread of the estimates S of the standard deviation. Given o n e
estimate S, we can specify the u p p e r a n d lower limits within which the true
value of the standard deviation σ will lie, with any required degree of confi­
dence.
Such calculations have been carried out a n d the results are tabulated in
Table I. This shows the confidence limits of σ/S for either 90% or 9 5 %
confidence for different n u m b e r s of degrees of freedom. T h e m e t h o d of using
Table I is shown in the following worked example.
Sixteen samples are r e m o v e d from a binary mixture a n d the percentage
proportions by weight of o n e c o m p o n e n t in the samples are
41, 37, 41, 39, 45, 37, 39, 40
41, 43, 40, 38, 39, 37, 43, 40

Table I

Confidence Limits of σ/S

90% confidence 95% confidence

Degress of Upper Lower Upper Lower


freedom limit limit limit limit

1 00 0.510 00 0.446
2 4.41 0.578 6.26 0.520
3 2.92 0.620 3.75 0.566
4 2.37 0.649 2.87 0.600
5 2.09 0.671 2.45 0.625
6 1.91 0.690 2.20 0.645
7 1.80 0.705 2.03 0.611
8 1.71 0.718 1.91 0.676
9 1.64 0.730 1.82 0.689
10 1.59 0.739 1.76 0.698
15 1.44 0.775 1.55 0.738
20 1.35 0.798 1.45 0.766
25 1.31 0.815 1.38 0.784
30 1.27 0.828 1.34 0.798
40 1.23 0.847 1.28 0.822
50 1.20 0.861 1.24 0.837
100 1.13 0.897 1.16 0.878
16. M i x i n g of Particulate Solids 273

D e t e r m i n e the upper a n d lower 9 5 % a n d 9 0 % confidence limits of the stan­


d a r d deviation of the mixture.
F r o m Eq. (1), the m e a n value of the sample compositions,

Since t h e true composition of the m i x t u r e is n o t k n o w n , the estimate of the


standard deviation is found from Eq. (4):

5 =
{[ς( - °) ]/ }
/ 4 2 ι 5 1 / 2 = 2
· 3 1

F r o m Table I, for 15 degrees of freedom, the 9 5 % confidence limits for σ/S


are 0.738 a n d 1.55 a n d σ lies between 0 . 7 3 8 5 a n d 1.55.S (i.e., between 1.70
a n d 3.58); the 90% confidence limits for σ/S are 0.775 a n d 1.44 a n d σ lies
between 0.775.Sand 1.445 (i.e., between 1.79 a n d 3.33).
It is seen from Table I t h a t the estimate of the standard deviation based o n
a small n u m b e r of samples is very imprecise. F o r example, if five samples are
taken, giving four degrees of freedom, a n d the standard deviation based on
the compositions of the samples is 1.0, there is a 9 5 % probability t h a t the true
value of the standard deviation σ lies between 0.60 a n d 2.87.

D. SAMPLING FROM NONRANDOM MIXTURES


W h e n samples are withdrawn from a r a n d o m mixture, as previously dis­
cussed, the standard deviation of sample compositions is given by

o =[P{\-P)lri\w
K (9)

so that
a oc«-i/2
R ( 1 0 )

If a standard deviation is obtained for samples of a certain size, this


relationship permits the calculation of the standard deviation for samples of
a different size. If the mixture is n o t r a n d o m , this relationship n o longer
holds. Williams (W3) has shown theoretically that the effect of changing the
sample size is given by
a = L + [P(l -P)-L]/n
2
(11)

where L is a constant for a given mixture, which can be determined if the


value of σ is k n o w n for o n e value of n. Kristensen (K2) developed a m e t h o d
for evaluating «, the expected n u m b e r of particles in a sample, when the
c o m p o n e n t s d o not consist of mono-sized particles. His equation, when
274 John C. Williams

modified, gives

^(P a A , + W ( l 2 )

where P a n d P are the weight fractions of the two c o m p o n e n t s in the


a h

mixture, p a n d t h e particle densities of the c o m p o n e n t s , Wis the weight of


a

a sample, a n d vv a n d vv are the average particle weights.


a b

T h e reliability of a n estimate of the standard deviation based on sampling


has already been discussed for the case w h e n the mixture is r a n d o m . For
n o n r a n d o m mixtures the confidence limits are narrower. However, as the
mixing quality improves, the confidence limits rapidly approach those ap-
plying to a r a n d o m mixture. For a mixture of a quality likely to be acceptable
in an industrial process, the confidence limits for the value of the standard
deviation can be obtained by the m e t h o d described for a r a n d o m mixture.

III. Segregation of Particulate M a t e r i a l s

M o s t of the early research work on the mixing of solid particles was carried
out with two c o m p o n e n t s whose particles were identical in all i m p o r t a n t
properties, differing perhaps only in color. In such a case, if the mixing
process goes o n long enough, r a n d o m mixing will be achieved, different
types of mixers varying only in the speed with which r a n d o m n e s s is ap-
proached. During recent years attention has been given to systems contain-
ing particles of different properties that tend to exhibit segregation. Particles
with the same property t h e n collect together in some part of the particle
mass. F o r such a system, r a n d o m mixing is n o t a natural state. Even if the
particles are originally mixed, they will u n m i x on handling.
T h e properties that can give rise to segregation include differences in the
size, density, shape, a n d resilience (determined by the a m o u n t of a particle's
energy lost on impact) of the constituent particles. Each of these properties
can, u n d e r certain circumstances, p r o d u c e segregation, but all the available
evidence shows that difference in particle size is by far the most i m p o r t a n t . In
particular, a n d contrary to the c o m m o n expectation, density difference is
comparatively u n i m p o r t a n t . A n exception to this rule occurs in the case of
segregation in a fluidized bed, where density difference is m o r e serious t h a n
size difference. In this chapter attention will be concentrated on size segrega-
tion.
M a n y industrial problems arise from segregation. T h e most obvious is the
u n m i x i n g of free-flowing particles of different size when mixing is at-
t e m p t e d . T h e action of a mixer tends to produce r a n d o m n e s s a n d the parti-
cles resist this trend by segregation. T h e result is an equilibrium between
16. M i x i n g of Particulate Solids 275

mixing a n d segregation that sets a limit to the quality of mixing that can be
attained.
Even if satisfactory mixing is achieved in the mixer it c a n n o t be assumed
that the quality of the mix will persist u n c h a n g e d during subsequent han­
dling a n d storage.
Considerable difficulties arise in the removal from a large quantity of
segregated materials of a sample that has the same particle size distribution as
the entire charge. F o r example, samples which are removed for sieve analysis
are liable to large errors unless care is taken to avoid the effects of segregation.
In feeding materials to packaging machines, tableting presses, or other
devices in which equal volumes of material are taken in the h o p e that they
will be of equal weight, segregation causes fluctuations in the size distribu­
tion of the particles a n d this in t u r n leads to variations in bulk density. This is
o n e of the causes of variation in the weight of the contents of packets a n d
tablets.

A . MECHANISMS OF SEGREGATION
A m i x t u r e of particles of different size m a y segregate by four different
m e c h a n i s m s , which will n o w be discussed. T h e locations where they are
likely to occur in a n industrial process are also considered.

7. Trajectory Segregation
If a particle of diameter D a n d density p is projected horizontally with
s

velocity v into a fluid of viscosity η, the horizontal distance t h a t it will travel


0

after infinite time, sometimes referred to as its "stopping distance," is equal


to v p D 118 η. A particle of diameter 2D would therefore travel four times as
0 s
2

far. This m e c h a n i s m can cause segregation in a mixer in which particles are


lifted o u t of a mass of material a n d t h r o w n with the intention of scattering
t h e m across the surface. S o m e segregation will occur due to the large parti­
cles traveling farther a n d this will limit the quality of mixing that can be
achieved.
W h e n a cloud of particles is in flight, the interaction of particles with the
s u r r o u n d i n g fluid c a n n o t be assumed to be that of an individual particle.
Segregation then occurs by a different m e c h a n i s m . U n d e r these circum­
stances, fine particles m a y travel farther t h a n coarse particles a n d m a y be
spread over a wider area.

2. Percolation of Fine Particles


If a mass of particles is disturbed in such a way that individual particles
m o v e , a rearrangement in the packing of the particles occurs. F r o m t i m e to
t i m e gaps between particles will be produced, allowing a particle from above
276 J o h n C. Williams

to d r o p , a n d a particle in some other place moves upward. If the powder mass


contains particles of different size it will be easier for a small particle to fall
d o w n into the next layer t h a n a larger one, a n d so there will be a tendency for
the smaller particles to m o v e downward, leading to segregation. For this
effect to occur, it is not necessary for the fines to be so small that they can pass
through the voids between the bigger particles w h e n they are at rest; a very
small difference in size is enough for measurable segregation to take place.
Percolation can occur whenever a mixture of particles of different size is
disturbed in such a way that rearrangement of the particles occurs. This can
arise from the existence of shear within the mass, caused, for example, by
stirring or by pouring the particles into a heap. It also occurs when a mixture
is shaken, a procedure which would be effective for mixing liquids, b u t which
separates the c o m p o n e n t s when they are particles of different size. Percola-
tion can also take place when a particle bed is vibrated.
Segregation o n pouring a h e a p is of such i m p o r t a n c e that it deserves
particular m e n t i o n . It is generally k n o w n that, w h e n a mixture of particles of
different size is p o u r e d into a conical pile or heap, the larger particles tend to
r u n d o w n to the edge of the heap. T h e extent to which segregation occurs
inside the heap is, however, not generally realized. W h e n material is p o u r e d
o n t o the t o p of a heap there is a layer of particles, a few particle diameters
deep, in which there is a very high velocity gradient. This layer effectively
forms a screen through which all b u t the largest particles are able to pass to
the stationary region below. T h e ease of passage is a function of particle size
a n d this leads to very thorough size segregation in the heap. This is the most
frequent a n d serious cause of size segregation.

3. Rise of Coarse Particles on Vibration


In addition to the percolation effect, there is a n o t h e r m e c h a n i s m of segre-
gation when a mass containing particles of different size is vibrated. This can
be d e m o n s t r a t e d by placing o n e large particle at the b o t t o m of a bed of
smaller particles. O n vibration the large particle can be m a d e to rise to the
surface. This effect occurs even if the large particle is denser t h a n the fine
particles. In fact, the denser the large particle the easier it is to persuade it to
rise. This m a y be explained by the fact that the large particle causes an
increase in pressure in the region below it, which compacts the material a n d
stops the particle from moving downward. Any upward m o v e m e n t allows
fines to r u n in u n d e r the coarse particle, a n d these in t u r n are locked in
position. If the intensity of vibration is suitable the large particle will m o v e
right u p to the surface. A m o r e detailed discussion of this effect is given by
Williams (W2).
16. M i x i n g of Particulate Solids 277

4. Elutriation Segregation
W h e n a particulate material containing a wide size distribution, including
a n appreciable a m o u n t less t h a n , say, 50 μτη, is discharged into the t o p of a
h o p p e r or other container, air is displaced upward. T h e air velocity m a y
equal or exceed the terminal velocity of the fine particles, which t h e n r e m a i n
in suspension as a cloud after t h e larger particles have settled out. T h e fine
particles will eventually fall o u t of suspension a n d form a layer o n t o p of the
coarse particles. W h e n the h o p p e r is e m p t i e d a pocket of fine particles,
containing almost all the fines in the system, will appear in the emerging
stream.

B. MEASUREMENT OF SEGREGATION
T h e first detailed study of the m e c h a n i s m of segregation was reported by
D o n a l d a n d R o s e m a n (D3), w h o investigated the effect of placing a mixture
of particles of different size or density in a rotating horizontal d r u m . They
showed that, u n d e r certain circumstances, almost complete separation of the
c o m p o n e n t s could occur. First radial segregation was observed, the finer or
denser material being concentrated in a central core, parallel to the axis, with
t h e larger or less dense particles a r o u n d it. This can be explained by the
m e c h a n i s m of the " h e a p p o u r i n g " effect. T h e finer or denser particles were
sieved o u t from the highly sheared surface layer a n d accumulated in the
comparatively static a n d u n d e f o r m e d central region. O n c o n t i n u e d rotation,
axial segregation occurred a n d t h e c o m p o n e n t s separated into a n u m b e r of
alternate layers. T h e surface of the bed showed a series of stripes r u n n i n g
perpendicular t o the axis. F o r this effect to occur, the c o m p o n e n t s m u s t have
different angles of d y n a m i c repose in the d r u m .
C a m p b e l l a n d Bauer ( C I ) m a d e m e a s u r e m e n t s of the a m o u n t of segrega­
tion occurring in a rotating horizontal d r u m by fitting a n axial shaft which
rotated with the d r u m with 12 cups which, in each revolution, lifted samples
from t h e contents of the d r u m . After r u n n i n g for the required time, the
samples were r e m o v e d a n d analyzed. T h e y confined their experiments to
binary mixtures of particles of d i a m e t e r ratios u p to 1.4. Their results show
t h a t measurable segregation occurs for very small size differences b u t their
m e t h o d of collecting the samples probably underestimates the severity of the
segregation.
T h e work of several experimenters w h o have a t t e m p t e d to isolate the
individual m e c h a n i s m s of segregation is outlined in the following.
Williams a n d Shields (W9) investigated the segregation that occurred
w h e n a stream of particles containing a m i x t u r e of two sizes was fed contin­
uously o n t o a vibrated channel. T h e m a i n a i m of this work was t o determine
278 John C. Williams

the conditions u n d e r which the m a x i m u m a m o u n t of segregation occurs, as


this would i m p r o v e the performance of a vibrating screen. T h e greatest
a m o u n t of segregation was found to occur w h e n the direction of vibration
was 30° to the horizontal a n d the m a x i m u m acceleration was 6.6g. T h e
vertical c o m p o n e n t of this acceleration was 3.3g.
Campbell a n d Bridgwater (C2) carried out a series of experiments in which
the a i m was to isolate the effect of a shear field in a flowing powder mass as a
cause of segregation. T h e bed of particles was contained between vertical
walls of float glass. T h e horizontal dimensions of the bed were 13.5 X 6.35
c m . T h e bed rested o n a piston, which was driven d o w n w a r d at a controlled
speed to p r o d u c e m o v e m e n t of the bed. O n e end wall of the container was
roughened by attaching 120-grit sandpaper to it, t h u s introducing a failure
zone in the m o v i n g powder near the rough wall. T h e bed consisted of 4 - m m -
diameter glass spheres; glass spheres of other sizes were introduced as tracers
a n d their m o v e m e n t was followed photographically.
W h e n the tracer particles were larger t h a n 4 m m , n o percolation was
detected, b u t for tracer particles smaller t h a n 4 m m , percolation was ob-
served for particle diameter ratios m o r e t h a n 2. T h e width of the zone in
which percolation occurred was a b o u t 7 particle diameters ( 7 X 4 m m ) ; the
width of the shear zone was velocity-dependent, varying from 5 to 15 particle
diameters.
W h e n the tracer particles were of higher density t h a n the bed particles
(density ratio 6) a n d the particles were of the same size, n o percolation was
observed. This indicates that, u n d e r the circumstances of the test, density
difference alone did not cause segregation. W h e n the denser particles were
smaller t h a n the bed particles, a diameter ratio of 1.7 was enough to cause
segregation. Glass spheres with this diameter ratio did n o t segregate, which
showed that density difference has a slight effect o n the a m o u n t of percola-
tion occurring d u e to size difference.
It m a y be noted t h a t in the above experiments the shear plane a n d the
direction of percolation were b o t h vertical. M o r e m a r k e d percolation would
be expected to occur if the shear planes were horizontal. Scott a n d Bridg-
water (S2) studied the rate of percolation of a small particle through a bed of
closely sized larger particles in a simple shear cell in which the bed was
subjected to a uniform shear strain. By reversing the direction of m o v e m e n t
of the cell, unlimited strain could be applied. They found that difference in
particle size was the m o s t serious cause of percolation, the effect of diameter
ratio being as shown in Fig. 2. W h e n the diameter ratio of the bed particles to
the fines was less t h a n 2, reproducible results were n o t obtained. However,
measurable percolation rates were observed when the difference in particle
diameter was only a b o u t 3%. T h e y showed that the percolation velocities
observed could be explained in t e r m s of the diffusion equation.
16. M i x i n g of Particulate Solids 279

7 r

PARTICLEDIAMETE RRATI O W/d)


FIG. 2. Effect of diameter ratio o n percolation velocity. [After Scott and Bridgwater (S2).]

O t h e r experiments by Scott a n d Bridgwater (S2) showed that percolation


occurred w h e n the bed consisted of particles with a c o n t i n u o u s size grading.
Differences in particle density a n d shape, the m a g n i t u d e of the n o r m a l stress
acting o n the shear planes, a n d the shear strain rate were found to have little
effect c o m p a r e d with particle size difference.
Sugimoto a n d Y a m a m o t o (S9) used a m e t h o d similar in principle to that
of C a m p b e l l a n d Bridgwater, in which a bed of particles was allowed to m o v e
d o w n w a r d in a vertical channel of rectangular cross section. In this case,
however, all the walls were s m o o t h a n d plug flow was observed. A layer of
tracer particles was placed o n t o p of the bed and, after allowing flow, the bed
was e x a m i n e d to find the position of the tracer particles. T h e effects of size
difference a n d density difference were investigated a n d the ratio of the m e a n
displacements of tracer a n d bed particles was found to fit the equation
x /x
T B = O.S(p /p )
T B + 12(D /D )B T
k
- 1 (13)
where χ is the particle displacement, ρ the particle density, a n d D the particle
diameter. Suffixes Β a n d Τ refer to bed a n d tracer particles, respectively.
H e r e k = 1 if the tracer particles are smaller t h a n the bed particles a n d k = 2
if the tracer particles are larger.
280 J o h n C. Williams

In these experiments the bed was not subjected to shear. Percolation


occurred only because of the rearrangement of particles during plug flow.
T h e results obtained are therefore n o t consistent with those of Campbell a n d
Bridgwater.
Harris a n d H i l d o n (H5) devised a test for the m e a s u r e m e n t of segregation
in which a core flow h o p p e r is filled with a mixture from a central p o u r point,
so that segregation occurs. T h e material is then discharged into a box in
which a h e a p is formed, causing further segregation. F o u r samples are taken
from different points at the base of the heap a n d analyzed. They defined a
degree of segregation as follows. If the percentages of one c o m p o n e n t in the
four samples are proportional to a a , oc , a n d a , where a + a +
l 9 2 3 4 x 2

α + α = 100, then the degree of segregation is


3 4

2 Ια»-25| (14)
n= 1

T h e value of this measure of segregation varies from 0 (when all samples have
the same composition) to 150 (when only one sample contains the c o m p o ­
nent).
Williams a n d K h a n (W4) used a n inclined d r u m to measure the a m o u n t of
segregation taking place in mixtures of particles of different size. T h e a i m
was to devise a simple m e t h o d to detect segregation for given c o m p o n e n t s as
a n aid to mixer selection. T h e mixture is placed in the d r u m , a n d during
rotation the finer particles m o v e to the lower part of the mixture a n d the
coarse particles to the u p p e r part. T h e mixture is then cut in half in a plane
perpendicular to the axis of the d r u m a n d the compositions of the two halves
are used to calculate a coefficient of segregation C : s

"CT + "CB

where is the weight fraction of coarse particles in the t o p half a n d is


the fraction of coarse particles in the b o t t o m half; C is zero for n o segrega­
s

tion a n d 100% for total separation of the c o m p o n e n t s .


F o r binary mixtures of constant m e a n particle size, it was found that the
coefficient of segregation increased as the diameter ratio of the c o m p o n e n t s
increased u p to a value of a b o u t 4.5 (Fig. 3). W h e n the absolute size of the
particles was changed a n d the diameter ratio was kept constant, it was found
that as the m e a n particle diameter dropped below a b o u t 500 μιτι, there was a
considerable falloff in the a m o u n t of segregation, b u t with a m e a n size of 100
μτη a n appreciable a m o u n t of segregation was still detected (Fig. 4).
W h e n the same materials were mixed in a twin-shell blender (see Fig. 5),
the quality of mixing achieved, as measured by the standard deviation of
16. M i x i n g of Particulate Solids 281

100

- 80
ο
z~
ο
' 60

Ο SAND
Χ FERTILIZE
R GRANULE
S

40

Γ7 2 0

1 2 3 4 5 6
PARTICLEDIAMETE RRATI O
FIG. 3. Variation o f coefficient o f segregation with particle diameter ratio. [Williams and
K h a n (W4).]

MEAN PARTICL EDIAMETER ,(pm )


FIG. 4. Variation o f coefficient o f segregation with m e a n particle diameter. [Williams and
Khan (W4)J
282 John C. Williams

ROTATING C Y L I N D E R

ROTATING CUB E

D O U B L E CON E O B L I Q U E CON E

TWIN S H E L L ( O R V )
WITH A G I T A T O R B A R

ORBITIN G S C R E W

LODIGE M I X E R

RIBBONBLENDE R
FIG. 5. S o m e c o m m o n types of particle mixers.
16. M i x i n g of Particulate Solids 283

sample compositions, was found to follow a pattern very similar to the values
of the coefficient of segregation from the rotating d r u m tests. It was con-
cluded that in the rotating d r u m tests the material was being subjected to a
t r e a t m e n t similar to that which it would experience in a t u m b l i n g mixer, a n d
t h a t the test was therefore helpful in deciding whether such a mixer could be
used for given materials.
Williams a n d K h a n (W4) also used the inclined d r u m to investigate the
effect of adding small quantities of water as a m e a n s of reducing segregation.
F o r a free-flowing sand the addition of 2% by weight of water reduced the
coefficient of segregation from 70% to less t h a n 10%. A similar effect was
found for the quality of mixing obtained in a t u m b l i n g mixer; the addition of
1% by weight of water reduced the standard deviation of sample composi-
tions from a b o u t 30% to less t h a n 5%.
Shinohara et al (S4) considered the segregation that occurs when a mix-
t u r e of large a n d small particles flows d o w n through a converging channel. In
this case the material is subjected to shear as it flows. In their theoretical
m o d e l they assumed that the large particles formed a matrix through which
the fine particles m o v e d downward. T h e matrix was regarded as presenting
t o the fine particles a series of apertures through which they were able to flow.
T h e flow rate was predicted from m e a s u r e m e n t s of the rate of flow from a
m o d e l hopper. T h e y were t h u s able to predict the discharge rates of coarse
a n d fine particles, a n d so t o obtain the variation with t i m e of the composition
of the discharge stream. Their experimental results agreed reasonably well
with their predictions, in spite of the rather doubtful assumptions they m a d e ,
b u t this m a y be d u e to the fact t h a t in their experiments the ratio of the
diameters of coarse a n d fine particles was 14.
Lawrence a n d Beddow (L3) investigated the segregation occurring w h e n a
binary m i x t u r e was p o u r e d into a cylindrical die 50 m m in diameter a n d 35
m m high. T h e y found that difference in particle shape a n d density had a
negligible effect c o m p a r e d with particle size difference. For mixtures of
particles of two sizes the contents of the die were generally rich in fines in the
lower a n d central parts of the powder mass. A diameter ratio of 1.2 was
e n o u g h to cause considerable segregation. Also, for a given diameter ratio,
increasing the absolute size of the particles increased the degree of segrega-
tion. W h e n the size of the larger particles was m a i n t a i n e d constant a n d the
size of the smaller particles reduced, segregation increased markedly at first,
passed t h r o u g h a m a x i m u m , a n d t h e n decreased, probably because the large
particles found it difficult to m o v e t h r o u g h a bed of m u c h smaller particles. If
the p r o p o r t i o n of fine particles was increased above 60% this effect b e c a m e
m u c h m o r e m a r k e d a n d the pattern of segregation was inverted, the larger
particles being retained in the center of the die fill. T h e segregation observed
284 John C. Williams

was generally reduced by increasing the height of d r o p of material into the


die or by increasing the rate of fill.
Lawrence a n d Beddow (L3) explained these effects by assuming that "fine
particles filter d o w n through the m o v i n g powder m a s s . " It seems m o r e likely
that segregation is caused by the " h e a p p o u r i n g " effect; their results are
consistent with this explanation. If this is true, the effect could be consider­
ably reduced by filling the die in such a way t h a t the formation of sloping
surfaces was avoided.

C. REDUCTION OF SEGREGATION
Since segregation occurs primarily as a result of difference in particle size,
the difficulty of mixing two c o m p o n e n t s can be reduced by m a k i n g the sizes
of the c o m p o n e n t s as close as possible a n d also by reducing the absolute size
of the particles. It is n o t possible to give general rules regarding the effect of
particle size, b u t in most cases for particles of density a b o u t 2000 to 3000
k g / m segregation will not cause serious problems if all the particles are
3

smaller t h a n a b o u t 30 μτη. For denser particles the critical size will be lower.
F o r powders in which the particles are too small to permit segregation there is
a possibility that aggregates will form a n d that they will segregate.
Addition of a small a m o u n t of a liquid to the mixture will prevent segrega­
tion a n d allow a better mixing quality to be achieved. Advantage should be
taken of this when the process involves the addition of a liquid by adding it in
the mixer.
If o n e of the c o m p o n e n t s is very fine (less t h a n a b o u t 5 μτή) a n d the other is
comparatively large, the mixing m e c h a n i s m consists of coating the fines
o n t o the surface of the larger particles. Segregation will not occur, a n d it is
possible to achieve an ordered mix which will be m o r e h o m o g e n e o u s t h a n a
r a n d o m mixture. This type of mixing is discussed by Y e u n g a n d Hersey
(Yl).
In handling a mixture of particles of different sizes, care should be taken to
avoid situations which are likely to p r o m o t e segregation; in particular, the
process should be designed to avoid pouring the material so that it forms a
sloping surface. T h e most damaging thing that can be d o n e is to p o u r the
mixture into a core flow hopper, so that a h e a p of material forms inside the
mixture, a n d then discharge the hopper. A core flow hopper is one in which,
when material is being discharged from the hopper, d o w n w a r d flow is con­
fined to a core above the outlet, the r e m a i n d e r of the material in the hopper
being stationary. Segregation o n filling produces a high concentration of
fines in a central core above the outlet a n d this is the first material to leave the
hopper; the coarser particles, which were placed near the walls of the hopper,
discharge later. D u e to a secondary percolation effect occurring during flow
o u t of the hopper, a small pocket of fines is formed at the wall just above the
16. M i x i n g of Particulate Solids 285

outlet a n d this is the last material to leave the hopper. This segregation effect
can be reduced by using a mass flow hopper, t h a t is, o n e in which there are n o
dead spaces, every particle being in m o t i o n w h e n discharge from the outlet
occurs. However, the m o s t effective r e m e d y is to prevent the formation of
sloping surfaces while filling the hopper.

IV. Selection of M i x e r s

M a n y different types of mixers are commercially available; s o m e of the


m o r e c o m m o n types are shown in Fig. 5. In selecting a mixer for a given duty
the m o s t i m p o r t a n t consideration is to select the type of mixing m e c h a n i s m
which is best suited to the physical properties of the particles being mixed. In
particular, w h e n mixing free-flowing particles of different size, which are
liable t o segregation, a mixer should be chosen which establishes an equilib-
rium between t h e mixing a n d segregating tendencies which permits the
required mixing quality t o be achieved.

A . MECHANISMS OF MIXING
T h e classification of mixing m e c h a n i s m s usually adopted is that proposed
by Lacey (L2), w h o distinguished a n d defined these three mechanisms:
(1) diffusive mixing, which occurs w h e n particles roll d o w n a sloping
surface;
(2) shear mixing, which occurs w h e n slip zones are established in a
powder; a n d
(3) convective mixing, which occurs w h e n circulation patterns are set u p
inside a bulk p o w d e r mass.
W h e n particles t h a t have n o tendency to segregate are mixed these three
m e c h a n i s m s are all effective, b u t from the discussion o n segregation it is
a p p a r e n t t h a t w h e n particles of different size roll d o w n a sloping surface
segregation takes place, because the fine particles pass m o r e easily through
the surface layer. Similarly, w h e n slip zones are established, mixing takes
place by the interchange of particles between layers, b u t w h e n a size differ-
ence is present fine particles will m o v e m o r e easily t h a n coarse particles,
p r o d u c i n g segregation. T h e first two m e c h a n i s m s will therefore produce
segregation, rather t h a n mixing, for particles of different size. Convective
mixing involves the m o v e m e n t of comparatively large masses of particles
a n d is m u c h less p r o n e to p r o m o t e segregation.

B. CLASSIFICATION OF MIXERS
Generally m o r e t h a n o n e of these three effects will be present in a mixer,
b u t usually it is possible to identify which of the m e c h a n i s m s predominates
286 John C. Williams

in a given type. Mixers therefore can be classified according to their ability to


m i x particles which t e n d to segregate.
A n y mixer that relies on a t u m b l i n g m o t i o n , that is, one in which the
particles are in a container which rotates, is mainly using diffusive mixing.
This class would include twin shell or V mixers, rotating cubes, a n d double
cones. F o r free-flowing particles of different size there will be a large a m o u n t
of segregation in the mixer, which will severely limit the quality of mixing
achieved. Fitting baffles inside the mixer will have very little effect in reduc-
ing the segregation. Mixers which rely o n a stirring action, not accompanied
by convective m o t i o n , will also produce a n appreciable a m o u n t of segrega-
tion a n d therefore will n o t be suitable for mixing materials p r o n e to segrega-
tion. In cases where segregation is k n o w n to be a problem the mixer chosen
should be o n e which relies mainly o n a convective action, the most c o m m o n
types being the ribbon b l e n d e r — p e r h a p s the most useful general-purpose
m i x e r — a n d a n orbiting screw mixer. This general conclusion is supported
by the results of mixing tests reported in the literature by H a r n b y (H2),
A d a m s a n d Baker (A 1), A s h t o n a n d Valentin (A3), a n d Williams a n d K h a n
(W4).

C . SELECTION OF MIXERS
In selecting a mixer for a given duty it m u s t first be ascertained whether or
n o t the c o m p o n e n t s segregate. If they d o segregate, attempts should be m a d e
to adjust the particle size of the c o m p o n e n t s to reduce the effect; this can be
d o n e by bringing the size distribution of the c o m p o n e n t s as close together as
possible a n d by reducing the absolute size of the particles. If the process
involves c o m m i n u t i o n , or if particle size reduction can be tolerated, the
c o m p o n e n t s should be crushed together, since this will generally lead to good
mixing in the mill. W h e n the process requires the addition of a small a m o u n t
of liquid this should be added in the mixer, since it will render the particles
m o r e cohesive a n d thereby reduce segregation. This will m a k e the mixture
m o r e difficult to handle, b u t in cases in which segregation of the dry m a t e -
rials is severe, or in which a high quality of mixing is required, it is generally
easier to solve t h e problems of handling the wet mixture t h a n to ensure good
mixing of the dry c o m p o n e n t s . If n o n e of these steps can be taken to avoid
mixing materials that t e n d to segregate, it is i m p o r t a n t to use a mixer that
relies p r e d o m i n a n t l y o n convective mixing, preferably a ribbon blender or
a n orbiting screw mixer, a n d to design the rest of the process so as to mini-
mize the a m o u n t of segregation occurring in the subsequent handling of the
mixture.
F o r mixing particles that are too cohesive to segregate, the m a i n require-
m e n t is that the mixer provide sufficient shear forces to break u p aggregates.
A t u m b l e r mixer will generally p r o m o t e the formation of aggregates, which
16. M i x i n g of Particulate Solids 287

will prevent i n t i m a t e mixing of the c o m p o n e n t s , b u t the addition of a shaft


that carries devices for breaking aggregates, a n d that rotates at a speed m u c h
higher t h a n t h a t of the mixer shell, will greatly i m p r o v e the mixing quality. If
the particles to be mixed are liable t o aeration, this m a y lead to settling of a
layer of fine particles o n the surface of the mixture when the mixing action is
stopped. In such a case a mixer should be chosen that will reduce this effect.
T h e introduction of large shear forces m a y in some cases lead to the fracture
of particles; if this c a n n o t be tolerated mixers with high-speed impellers,
mills, a n d possibly ribbon blenders would be excluded.

V. Specification and Testing of M i x t u r e Quality

A . SPECIFICATION OF MIXER PERFORMANCE


In designing a process that involves the mixing of particulate solids a
specification should be d r a w n u p for the quality of mixing required. T o d o
this it is necessary to specify three quantities: the scale of scrutiny, the
allowable variation from the m e a n composition, a n d the frequency with
which this variation m a y be exceeded.
T h e scale of scrutiny is the smallest a m o u n t of material within which the
quality of mixing is i m p o r t a n t . F o r example, if the mixture is to be used to fill
pharmaceutical capsules, each capsule should contain the necessary propor-
tion of each c o m p o n e n t b u t it does n o t m a t t e r h o w well they are mixed
within the capsule, since mixing will occur in solution. In this case the
a m o u n t of material going into o n e capsule is the scale of scrutiny. In any
particular application the scale of scrutiny should be decided only by consid-
ering the use t o which the m i x t u r e will be put. In testing a mixture, if the
samples withdrawn are smaller t h a n the scale of scrutiny, unnecessary de-
m a n d s are being m a d e o n the mixing process; if the sample size is larger t h a n
the scale of scrutiny some of the variation that will appear in the product will
be s m o o t h e d o u t a n d it is difficult to predict confidently from the results of a
sampling test what the composition variations in the product will be.
T h e allowable variation from the m e a n composition can be decided only
by considering t h e purpose for which the m i x t u r e is being m a d e . In some
cases it m a y be prescribed by law or specified by standards. It should always
be d e t e r m i n e d by the tolerance of the process in which the mixture will be
used.
It is n o t practical to set allowable limits a n d require that the composition
of samples always lie between those limits. F o r a r a n d o m mixture, the sam-
ple compositions will be normally distributed a n d there is always a possibil-
ity, n o m a t t e r h o w small, t h a t the limits will be exceeded. It is necessary to
288 J o h n C. Williams

decide h o w often the limits m a y be exceeded a n d incorporate this in the


specification.
A typical specification might be: w h e n the mixture is subdivided into
100-g portions the proportion of o n e c o m p o n e n t in the portions shall be
between 40 a n d 4 5 % at least 99 times out of 100. This specification is usually
expressed in the form of a standard deviation of sample compositions.

B . TESTING A MIXER
In arranging to carry out tests o n a mixer to determine its effectiveness a
n u m b e r of decisions have to be m a d e . T h e problem can be formulated as a
set of questions to be answered. H o w m a n y samples are to be taken? W h e r e
will they be taken from? H o w large will the samples be? H o w are the samples
to be removed? These questions will n o w be discussed in t u r n .

(1) H o w m a n y samples are to be taken? T h e a i m of the tests is to investigate


the quality of the whole mixture by removing samples a n d using the standard
deviation of the sample compositions as a n estimate of the standard devia­
tion for the whole mixture. It has already been pointed out t h a t the reliability
of this estimate depends o n the n u m b e r of samples taken. F o r a small n u m ­
ber of samples the confidence limits of the estimate of the standard deviation
are very wide. It was shown earlier that if five samples are removed from a
m i x t u r e a n d analyzed, a n d the standard deviation of their compositions is
1% the 9 5 % confidence limits of the ratio of the true standard deviation σ to
the estimate .Sare 0.60 a n d 2.87. T h e only statement that can be m a d e a b o u t
the standard deviation of the mixture is that there is a 9 5 % probability that it
lies between 0.6 a n d 2.87%.
T h e precision of this estimate can be improved by increasing the n u m b e r
of samples. However, collection a n d analysis of a large n u m b e r of samples is
b o t h t i m e c o n s u m i n g a n d costly a n d this imposes a limit o n the n u m b e r of
samples t h a t can be taken during the routine checking of a process. There
are, however, two situations in which removal of a large n u m b e r of samples
is justifiable. T h e first occurs w h e n a mixer is being tested to establish its
suitability for inclusion in a process. Sufficient samples should then be taken
that, bearing in m i n d the confidence limits, it is established with reasonable
certainty t h a t the mixer is capable of achieving the mixing quality required in
the process specification. T h e penalty for error is that a n unsuitable mixer
will be included in t h e process. T h e second occasion when a large n u m b e r of
samples should be taken is during start-up testing of the plant. This will verify
t h a t the mixer performs satisfactorily u n d e r operating conditions a n d will
provide information a b o u t the performance of the plant which can be used
to establish suitable routine testing procedures.
In deciding h o w m a n y samples to take in these tests, reference should be
16. M i x i n g of Particulate Solids 289

m a d e to Table I, which gives the 90 a n d 9 5 % confidence limits for the


estimated value of the standard deviation, for different n u m b e r s of samples.
(2) W h e r e will samples be r e m o v e d from? In removing samples from a
m i x t u r e it is necessary t h a t every part of t h e mixer has the same probability of
being selected. This is usually achieved by using a procedure for the r a n d o m
selection of sample sites, a n d the c o m m e n t s that were m a d e earlier a b o u t the
confidence limits of the estimated standard deviation apply to this case. A n
alternative m e t h o d t h a t satisfies the condition that all parts of the mixture
are equally likely to be selected is to use pattern sampling. T h e mixture is
conceptually divided into a n u m b e r of parts of equal v o l u m e a n d a sample is
taken from a r a n d o m l y selected position in each part. T h e estimate of the
standard deviation obtained by this m e t h o d will generally be m o r e precise
t h a n t h a t obtained by purely r a n d o m selection, b u t it is n o t possible to give
a n y general rules as to the a m o u n t of i m p r o v e m e n t . Pattern sampling is the
preferred m e t h o d of selecting sample positions.
Samples should never be r e m o v e d from the surface of a mixture as there is
a considerable risk t h a t the surface will n o t be typical of the underlying
material.
(3) W h a t sample size should be used? It has already been pointed out that
whenever possible the sample size should be the same as the scale of scrutiny,
especially during mixer trial tests. If this is impossible the standard deviation
obtained can be corrected to the scale of scrutiny by using Eq. (11). If the
a m o u n t of material needed for analysis is less t h a n the sample size, great care
m u s t be t a k e n to ensure t h a t the analysis sample has the same composition as
the original sample, especially if there is any tendency for segregation to
occur. T h e only satisfactory m e t h o d of obtaining the analysis sample is to use
a rotary sample divider [Allen a n d K h a n (A2)].
(4) H o w are samples to be removed? R e m o v a l of a sample from a mixture
without introducing bias presents considerable difficulties. A scoop should
never be used to collect a sample if the c o m p o n e n t s segregate, because the
particles falling from the edge of the sample will be preferentially coarse,
leaving a high proportion of fines in the sample. T h e device most c o m m o n l y
used to r e m o v e samples is a sample thief, a long rod with o n e or m o r e cavities
in which samples can be collected. A t u b u l a r sheath fits over the rod, a n d
w h e n it is rotated the sample cavities c a n be sealed off or opened. T h e cavities
are closed w h e n the device is inserted i n t o or withdrawn from the mixture,
a n d o p e n w h e n samples are being collected. Insertion of the sampling thief
can cause t h e material near the surface of the mixture to be dragged d o w n
with it a n d to be preferentially collected in the sample.
It is so difficult to avoid bias w h e n r e m o v i n g samples from inside a mixer
t h a t whenever possible the samples should be r e m o v e d from the stream of
material leaving the mixer w h e n it is discharged. It is t h e n m u c h easier t o
290 John C. Williams

avoid bias. T h e preferred m e t h o d is then to collect the whole of the outlet


stream for a short t i m e interval as a sample. Very great care should be taken
to ensure that this condition is satisfied. This m e t h o d has the added advan-
tage that some loss of mixing quality m a y occur in emptying the mixer. It is
i m p o r t a n t t h a t the test should give the quality of the mixture leaving the
mixer, rather t h a n that which occurs inside the mixer.
Whatever sampling technique is used it is advisable to c o m p a r e the m e a n
composition of the samples with the composition of the whole mixture,
using a Student t test, to detect whether the samples are biased; the m e t h o d
was discussed earlier.

VI. Continuous Mixing of Solids


In designing a n industrial plant for the handling a n d processing of liquids,
it is usually taken for granted that a c o n t i n u o u s system will be chosen unless
there are compelling reasons for using a batch process. T h e e c o n o m i c ad-
vantages of using a c o n t i n u o u s process are n o w widely accepted a n d d o not
need to be argued here.
In processes involving the handling a n d processing of particulate solids,
even w h e n other parts of the process are continuous, it is c o m m o n for solids
to be mixed in batches a n d stored in a hopper until required. Such systems
should be e x a m i n e d critically to see if it would n o t be better to use continu-
ous mixing.
Apart from e c o n o m i c advantages, there m a y be good technical reasons for
preferring c o n t i n u o u s mixing. F o r materials that tend to segregate it is often
very difficult to obtain good mixing in a batch mixer; even if satisfactory
mixing is obtained, the subsequent handling a n d storage of the mixture will
lead to segregation, so that o n arrival at the point at which the mixture is used
it is almost inevitable that some loss of mixing quality will have occurred. In
such a case good mixing m a y be m o r e readily obtained a n d handling a n d
storage of the mixture can be avoided by using a c o n t i n u o u s mixer placed
immediately before the next stage in the process.
A change to c o n t i n u o u s mixing to some extent changes the problem, as it
is n o w necessary to ensure satisfactory controlled feeding of the c o m p o n e n t s .
However, since the technical advantages of c o n t i n u o u s mixing are most
m a r k e d with c o m p o n e n t s which t e n d to segregate, these will be free flowing
a n d the problem of controlling feed rates is t h e n generally m o r e tractable
t h a n that of obtaining good batch mixing. In addition, a c o n t i n u o u s mixer
can be used to s m o o t h out fluctuations in the feed rates, so that the require-
m e n t s of the feeders can be m o r e easily met.
C o n t i n u o u s mixing is m o r e likely to be preferred when dealing with a
small n u m b e r of c o m p o n e n t s of free flowing, segregating materials with long
p r o d u c t i o n runs.
16. M i x i n g of Particulate Solids 291

A. EQUIPMENT FOR CONTINUOUS MIXING


A simple type of c o n t i n u o u s mixer can be used w h e n the materials to be
m i x e d are free flowing a n d very p r o n e to segregation. T h e c o m p o n e n t s are
fed, o n e o n t o p of t h e other, o n to a conveyor belt, which t h e n carries a ribbon
of material like a striped toothpaste. If a n y short length of this ribbon could
be cut o u t from the stream it would contain the correct proportions of the
c o m p o n e n t s , subject to the fluctuations in the delivery rates of the feeders.
T o obtain a uniform m i x t u r e only transverse mixing of the c o m p o n e n t s in
the stream is necessary. This can be d o n e as the material falls off the con­
veyor belt by arranging that it passes t h r o u g h a set of rotating blades. T h e
m i x t u r e t h e n goes directly, without storage, to the next part of the process,
which might be a bagging m a c h i n e , a tableting press, a glass melting tank, or
other e q u i p m e n t .
Since t h e residence t i m e in this mixer is negligibly small, n o appreciable
a m o u n t of back mixing takes place a n d the quality of the mixture obtained is
therefore only as good as the feeding systems will allow. F o r materials very
p r o n e to segregation this m a y be the best m e t h o d t o use. In general, however,
the use of a c o n t i n u o u s mixer with a n appreciable residence t i m e in which
back mixing takes place s m o o t h s o u t s o m e of the fluctuations in the c o m p o ­
sition of t h e stream entering the mixer.
Apart from the work of H a r w o o d et al. (H6) a n d Schofield a n d Tookey
(SI), there is very little information available a b o u t the performance of
industrial c o n t i n u o u s mixers. S o m e general conclusions can, however, be
d r a w n from the results for batch mixers. Mixers which rely p r e d o m i n a n t l y
o n the t u m b l i n g of the ingredients are k n o w n to be unsuitable for segregating
materials in batch mixing a n d it is reasonable to suppose that this will also
extend t o c o n t i n u o u s mixing.
If the materials are not liable t o segregation, generally because they are
very fine or cohesive, any type of mixer should give satisfactory results,
provided the m e a n residence t i m e of the mixer (capacity divided by v o l u m e
feed rate) is large enough.

B. CONTINUOUS MIXER PERFORMANCE PARAMETERS


T h e t e r m s a n d symbols t h a t will be used in this discussion of c o n t i n u o u s
mixing are as follows:

V v o l u m e of material in mixer
ν v o l u m e feed rate t o mixer
θ m e a n residence time, V/v
τ reduced t i m e , t i m e / 0

Both t h e i n p u t a n d the o u t p u t streams of a c o n t i n u o u s mixer can be


characterized by a composition variance a n d a correlation coefficient. T h e
292 John C. Williams

variance is a measure of the m a g n i t u d e of the fluctuations from the m e a n


value. T h e function of a c o n t i n u o u s mixer is to s m o o t h out fluctuations in
the composition of the input stream. A measure of its ability to d o this is the
variance reduction ratio ( V R R ) , which is the ratio of the variances of the
i n p u t a n d o u t p u t streams (σ^/σΐ). This ratio is used as a measure of
the effectiveness of the mixer.
T h e correlation coefficient of a stream is a description of the order in
which the fluctuations are arranged. For points separated by a t i m e r, the
correlation coefficient R(r) is given by
R(r) = cov(x , t x )/\2LT(x )
t+r t (16)
where x represents the fraction of o n e c o m p o n e n t in the stream at t i m e t,
t

covariance cov^x,, x ) = l,[(x — x)(x


t+r t — x)]/(n — 1), a n d variance
t+r

νατ(χ ) = Σ(χ -χ) /(η-


ί ί
2
1).
T h e effect of correlation over all separation times is represented by a serial
correlation index (SCI), where

(17)

T h e value of the serial correlation index lies between plus infinity a n d — \ .


F o r r a n d o m fluctuations, SCI = 0.

C. IDEAL MIXER
T h e best that can be expected from a c o n t i n u o u s mixer is that, at any time,
its contents will be r a n d o m l y mixed. Such a mixer is referred to as an ideal
mixer because it provides the highest value of the variance reduction ratio
that can be expected from a mixer; it also has the smallest size that will
perform a given duty.
Danckwerts a n d Sellers ( D l ) have shown that for a n ideal mixer the
variance reduction ratio is given by

T o integrate this expression it is necessary to k n o w how the correlation


coefficient varies with the interval r. G o l d s m i t h ( G l ) suggested that it could
be assumed that as the interval r between the points examined increases, R(r)
falls off in a geometric progression, that is,
R(r)=jr
where \j\<\ (19)
Substituting this into Eq. (18) a n d integrating gives
V R R = 1 -(V/v)lnj (20)
where the value of j depends o n the e q u i p m e n t being used to feed the mixer.
16. M i x i n g of Particulate Solids 293

E q u a t i o n (18) can be used t o calculate the variance reduction ratio for a n


ideal mixer or the least capacity of a mixer required for a given duty.
E q u a t i o n s (18) a n d (20) apply w h e n the i n p u t a n d o u t p u t are continuous.
Beaudry ( B l ) a n d G o l d s m i t h ( G l ) discussed the effect of various types of
intermittent feeding a n d discharging of a n ideal mixer.
T h e works of K r a m e r s a n d Alberda ( K l ) a n d Walker a n d Cholette ( W l )
for liquid mixing suggest t h a t better mixing can sometimes be obtained by
using a n u m b e r of small mixers in series rather t h a n o n e mixer of the same
total capacity. It is possible t h a t this a p p r o a c h could also be applied to solid
particle mixing.

D. NONIDEAL MIXER
Generally, it c a n n o t be assumed t h a t a c o n t i n u o u s mixer will be ideal. T o
predict the performance of a nonideal mixer it is necessary to obtain infor-
m a t i o n a b o u t t h e a m o u n t of back mixing occurring in it. This is determined
by the residence t i m e distribution, which can be obtained by using a
s t i m u l u s - r e s p o n s e technique. T h e m e t h o d s m o s t convenient for studying
particulate solids mixers are the introduction of a step change, which gives
the internal age distribution / , or the introduction of a pulse, which gives the
exit age distribution E. If either distribution is obtained by experiment the
other can be obtained from the relation

E = -dI/dt
M o s t of the published papers o n the performance of c o n t i n u o u s systems
are concerned with fluids, in connection with the design of c o n t i n u o u s
reactors ( D 2 , S3), distillation c o l u m n s ( H I ) , heat exchangers (C4), etc. Le-
venspiel (L4) discussed the interpretation of residence t i m e distributions.

E. REVIEW OF THE LITERATURE


Although m a n y papers dealing with the batch mixing of particulate solids
have appeared, there seems to have been little m o v e m e n t toward the study of
c o n t i n u o u s mixing systems. Poole et al ( P I , P2) considered the c o n t i n u o u s
mixing of cohesive powders in a ribbon mixer. In their experiments they
used very accurately controlled feeders so that there were n o appreciable
fluctuations in the input. T h e p r o b l e m was t h u s reduced to mixing of the
c o m p o n e n t s in a direction perpendicular to that of the flow. T h e y investi-
gated the homogeneity of the m i x t u r e by extracting samples from r a n d o m
positions in the mixer a n d in the o u t p u t , a n d found that the degree of
homogeneity obtained was as good as in batch mixing. T h e y also found, in
mixing u r a n i a a n d thoria, t h a t if the m i n o r c o m p o n e n t was fed intermit-
tently, m a i n t a i n i n g the s a m e overall rate as in c o n t i n u o u s mixing, the h o m o -
geneity of the m i x t u r e at the outlet was the same as in the case of c o n t i n u o u s
feeding, provided the interval between feed batches was less t h a n one-fifth of
294 John C. Williams

the m e a n residence time. They obtained residence time distributions for the
mixer b u t did not investigate the relation between residence t i m e distribu-
tion a n d mixer performance.
Sugimoto a n d his coworkers (S6) a t t e m p t e d to explain the performance of
a c o n t i n u o u s horizontal d r u m mixer in t e r m s of the m e c h a n i s m s of flow a n d
particle segregation inside the mixer. T h e y found that the segregating zones
occur in the same way as h a d previously been reported in batch mixing. In
the case of a binary mixture of particles of different size, they studied the
situation in which the smaller particles are fine enough to fit into the voids
between t h e larger particles without disturbing their packing arrangements
a n d the proportion of fine particles present is higher t h a n that which can be
a c c o m m o d a t e d in the voids. T h e y then found that two zones were formed,
o n e consisting of coarse particles with fines filling the voids a n d the other
consisting of fines only. They argued that the bulk density of the former zone
would be higher t h a n that of the latter a n d that this would account for
fluctuations in the discharge rate from the mixer. Predictions of the varia-
tions of composition a n d discharge rate agree only moderately well with
experimental results.
T h e same workers (S7) suggested that a c o n t i n u o u s d r u m mixer m a y be
divided into two parts, o n e where the segregating zones had already been
formed a n d n o further changes were taking place in the axial direction a n d
the other where the zones were n o t yet completely formed. F r o m a residence
t i m e distribution they obtained a n axial dispersion coefficient a n d showed
experimentally that axial dispersion increased with increasing proportion of
the smaller particles a n d stopped when the segregating zones had been
formed; the rate of axial dispersion also increased with d r u m speed within
t h e range 1 0 - 3 0 % of the critical speed.
Sugimoto (S8) attempted to predict the residence time distribution of a
c o n t i n u o u s horizontal d r u m mixer by considering the paths followed by
different particles. T h e m o d e l he proposed was that each particle follows a
p a t h in which it alternately moves u p w a r d in a circular p a t h a n d d o w n w a r d
in a straight line in a direction parallel to the m a x i m u m slope of the surface of
t h e particle bed. D u r i n g each d r u m rotation t h e particle moves forward a
distance which depends on its position within the bed. By assuming that each
particle r e m a i n s at the same relative position in the bed in passing through
the mixer (e.g., a particle which is in the outer surface of the bed continues to
m o v e in the outer surface), the residence t i m e distribution of the particles
can be predicted. In spite of the simplicity of the model, reasonable agree-
m e n t was obtained between predicted a n d experimental results.
T h e papers referred to previously approach the problem of attempting to
predict the performance of a mixer by using a model that describes the
m e c h a n i s m s of flow a n d mixing. Such attempts are valuable in that they
16. M i x i n g of Particulate Solids 295

increase o u r u n d e r s t a n d i n g of what is going o n inside the mixer, b u t it is


unlikely t h a t they will lead to sufficiently reliable predictions of mixer per­
formance a n d of the effect of operating conditions o n the quality of mixing
produced, for use in the design or selection of mixers for particular require­
m e n t s . Even in the comparatively simple case of a rotating d r u m only partial
success can be claimed, a n d for mixers with m o r e complex flow mecha­
nisms, e.g., ribbon blenders, there would b e little h o p e of producing a satis­
factory working model.
In studying c o n t i n u o u s systems involving the mixing of liquids, the most
successful a p p r o a c h has been to use residence t i m e distributions to give
information a b o u t the effectiveness of the mixer a n d t o e x a m i n e the effect of
operating variables. This a p p r o a c h to the c o n t i n u o u s mixing of solid parti­
cles was t a k e n by Williams a n d R a h m a n ( W 5 , W 6 , W7). Their a i m was to
develop a m e t h o d for predicting the performance of a mixer from the result
of a stimulus - response test. T h e work described was confined to the study of
nonsegregating systems. T h e mixer used was an inclined rotating d r u m
which was fed intermittently every 15 sec. T h e outlet stream was e x a m i n e d
by taking 15-sec samples, the whole of the stream being collected. U n d e r
these conditions the results are the same as if the feeding were c o n t i n u o u s .
T h e effects of feed rate, d r u m inclination, a n d d r u m speed were first exam­
ined to establish the o p t i m u m conditions for future tests, using a pulse input
test to d e t e r m i n e the residence t i m e distribution. T h e effect of varying the
d r u m speed is shown in Fig. 6, where the internal age distribution is plotted
in the form log / against reduced t i m e τ. F o r a n ideal mixer the internal age
distribution is of the form: / = e~ , so t h a t a plot of log / against τ will be a
T

straight line of slope — 1 / 2 . 3 . This line is shown for comparison in Fig. 6,


which shows t h a t at low d r u m speeds the d r u m mixer departs very m u c h
from the ideal. As the d r u m speed increases, the graph approaches a straight
line of slope — 1 / 2 . 3 . At 80 r p m the graph is very nearly a straight line parallel
to the ideal mixing line b u t displaced to t h e right. A further increase in speed
leads to lines of greater slope, indicating a falloff in mixing performance.
T w o m e t h o d s were developed for predicting the performance of the mixer,
as characterized by the variance reduction ratio, from the residence t i m e
distribution. In the first, the residence t i m e distribution is plotted as shown in
Fig. 6 (log / against reduced time) a n d is a p p r o x i m a t e d by o n e or m o r e
straight lines. Using a t h e o r e m proved by D a n c k w e r t s (D2),

(21)

the V R R can be evaluated provided the correlogram R(r) can be represented


in a m a t h e m a t i c a l form. If it is assumed t h a t the correlation coefficient falls
296 John C. Williams

REDUCEDTIME , t /0o r τ

FIG. 6. Effect of drum speed o n the performance of a drum mixer. [Williams and R a h m a n
(W5).]

off geometrically with time, R(r) = j where |y| < 1, the integration can be
r

performed a n d the value of the variance reduction ratio obtained.


E q u a t i o n (21) has been used, in conjunction with the experimentally
d e t e r m i n e d residence t i m e distributions shown in Fig. 6, t o predict the
variance reduction ratio for the d r u m mixer, w h e n r u n n i n g at different
speeds. It is assumed, for the purpose of this example, that the composition
fluctuations going into the mixer can be characterized by a value of; equal to
0 . 1 1 . T h e results are given in Table II. It is seen that at low d r u m speeds the
predicted variance reduction ratio is very small, indicating that little mixing
has occurred. As the d r u m speed increases the mixing improves, the variance
reduction ratio reaching a m a x i m u m value of 13.6 at 80 r p m , c o m p a r e d with
a predicted value of 14.5 for a n ideal mixer. F u r t h e r increase in the d r u m
speed leads to a falloff in mixture quality. T h e critical speed of the d r u m was
134 r p m .
This m e t h o d has t h e disadvantage t h a t it requires the assumption of a
16. Mixing of Particulate Solids 297

Table II

Effect of D r u m Speed o n Mixer Performance*'*

D r u m speed (rpm) 20 40 50 60 80 100 120

Variance reduction ratio calculated from


Eq. (21) 2.0 2.9 5.5 10.6 13.6 12.1 8.8

a
For ideal mixing V R R = 14.5; j was assumed to be 0.11; the critical drum speed was 134
rpm.
b
Based on data from Williams and R a h m a n (W7).

m a t h e m a t i c a l form for the correlogram of the input stream. A second


m e t h o d was developed which did n o t have this limitation ( R l , W6). T h e
composition fluctuations in the i n p u t stream a n d the exit age distribution are
expressed as matrices a n d multiplied together; the product matrix gives the
composition fluctuations in the o u t p u t stream. Physically the m e t h o d as-
s u m e s t h a t the i n p u t can be divided into a n u m b e r of short t i m e elements
a n d that for each such element the fraction of o n e of the c o m p o n e n t s acts as a
pulse whose distribution in the discharge is k n o w n from a pulse i n p u t test.
Experimental m e a s u r e m e n t s were m a d e to d e t e r m i n e the form of c o m p o -
sition fluctuations leaving the mixer for various types of input, the results
being c o m p a r e d with the predicted values. T w o typical results are shown in
Figs. 7 a n d 8, which d e m o n s t r a t e close agreement. It was shown that the
V R R was i n d e p e n d e n t of the m a g n i t u d e of the i n p u t fluctuations b u t was
heavily d e p e n d e n t o n the i n p u t serial correlation index. T h e results are
shown in Table III.
T h e m e t h o d therefore permits the prediction of the composition of each
sample in the o u t p u t provided t h e residence t i m e distribution a n d the form

FIG. 7. Comparison of predicted and measured output sample compositions (input serial
correlation index = 0.1). [Williams and R a h m a n (W7).]
298 John C. Williams

ARIANCE
11 .55

D 4.05

ED 4.32

FIG. 8. Comparison of predicted and measured output sample compositions (input serial
correlation index = 0.86). [Williams and R a h m a n (W7).]

of the i n p u t are k n o w n . Actual i n p u t data can be used, without m a k i n g any


assumptions a b o u t the m a t h e m a t i c a l form of the i n p u t fluctuations.
In work by Williams a n d Richardson (W8) the effect of segregation o n the
performance of a c o n t i n u o u s mixer was investigated a n d the above m e t h o d
of predicting the form of the o u t p u t was extended. For segregating materials
the inclined rotating d r u m previously used was found to give very poor
mixing, even u n d e r the most favorable conditions. Experiments were there-
fore carried out o n a continuously operated fluidized bed mixer. T h e results
of the experiments showed that the variations in composition in the dis-
charge from the mixer arise from two causes. First, the composition fluctua-
tions in the i n p u t stream are reduced in a m p l i t u d e in passing through the
mixer, b u t some fluctuations will still appear in the o u t p u t stream; the
variance of these fluctuations can be predicted, as in the case of nonsegregat-
ing materials, using the external age distribution. Second, composition fluc-
tuations appear in the o u t p u t stream as a result of segregation in the mixer;
the variance of these fluctuations can be found by feeding the mixer with a
stream in which there are n o fluctuations a n d measuring the fluctuations in

Table III

Effect of Input Serial Correlation Index o n Variance Reduction R a t i o 0

Input serial correlation index + 4 +0.86 +0.10 -0.19 -0.44*

VRR
Predicted 1.3 2.9 9.9 16.3 500
Measured 1.3 2.7 9.8 15.6 1200

a
Based o n data from Williams and R a h m a n (W7).
b
In the experiment at correlation index —0.44 the amplitude of composi-
tion fluctuations in the output is so small that more accurate results cannot be
expected.
16. M i x i n g of Particulate Solids 299

the o u t p u t ; this variance is referred to as the steady-state variance. T h e total


variance of composition fluctuations in the o u t p u t can be predicted by
adding together the two c o m p o n e n t variances described above, as

o u t p u t variance = nonsegregating variance obtained from external age


distribution + steady-state variance

E x p e r i m e n t s with a large n u m b e r of i n p u t s of different serial correlation


index confirmed that the predicted o u t p u t variance agreed closely with the
m e a s u r e d value. Typical results are shown in Fig. 9. T h e dashed line shows
the variance predicted for a nonsegregating system a n d the full line gives
values of the predicted total o u t p u t variance, which agrees well with mea­
sured variances.

F. FUTURE WORK
T h e r e is a need for further research into the c o n t i n u o u s mixing of solid
particles for b o t h cohesive a n d free-flowing materials. Preferably the work
should be directed toward the problems of selecting suitable mixers for
different types of industrial materials, the two approaches to the subject,
mechanistic a n d s t i m u l u s - r e s p o n s e , being c o m p l e m e n t a r y . Following the
mechanistic a p p r o a c h it would be valuable to have a better understanding of
the patterns of flow a n d mixing in different types of mixer. T h e s t i m u l u s -

INPUT VARIANC
E σ-j 2

FIG. 9. Predicted mixer performance (segregating materials). [Williams and Richardson


(W8)J
300 John C. Williams

response m e t h o d should be used to examine the performance of industrial-


type mixers, giving information that would lead to optimization of their
design a n d operation.

V I I . M i x i n g of Cohesive Particulate M a t e r i a l s

A. INTRODUCTION
So far in this chapter o n the mixing of particulate solids attention has been
concentrated o n the problems caused by segregation, which is observed
w h e n handling free-flowing particulate materials containing particles of dif-
ferent sizes. It has been pointed out that as the size of the particles is reduced a
better quality of mixing can be expected, for two reasons. First, if r a n d o m
mixing is achieved, or closely approached, the standard deviation of the
composition of samples taken from the mixture is inversely proportional to
the square root of the n u m b e r of particles in a sample; for samples of a given
weight the standard deviation will therefore fall as the particle size decreases.
Second, reducing particle size reduces the severity of segregation, leading to
better mixing.
T h e other major problem in the mixing of particles arises from cohesive-
ness; as particle size is reduced, the attractive force between particles, which
is negligible in free-flowing materials, approaches a n d eventually exceeds the
weight of a particle. Segregation is then prevented, because particles are n o
longer free to m o v e relative to each other. U n d e r these circumstances mixing
is a comparatively slow process a n d m o r e energy has to be p u t into the mixer
t h a n for free-flowing materials; however, because of the absence of segrega-
tion, a state of r a n d o m mixing can be approached m o r e closely a n d there will
be n o serious loss of mixing quality in the subsequent handling a n d storage of
the mixture.
O n e solution to the problem of mixing very segregating particles is, there-
fore, to m a k e the particles m o r e cohesive, either by reducing the particle size
or by adding a small a m o u n t of liquid. T h e improved mixing quality that will
be achieved will be adequate for very m a n y industrial processes; it will,
however, have to be paid for by having a product that is m o r e difficult to
handle or perhaps by requiring the addition of a drying stage to the process.
In some applications, in which the scale of scrutiny is very small, a further
p r o b l e m arises. In a bed m a d e u p of particles sufficiently fine to be cohesive,
there is a very strong tendency to form aggregates. If we consider the case in
which two or m o r e materials, in the form of small particles, are p u t into a
mixer, the first effect of the mixing action m a y be to form aggregates, each of
which will consist almost entirely of o n e c o m p o n e n t . If the subsequent
action is to p r o d u c e a mixture of these aggregates, the resulting mixture m a y
16. Mixing of Particulate Solids 301

be satisfactory if the scale of scrutiny is large c o m p a r e d with the aggregates.


However, in situations in which the scale of scrutiny is of the same order as or
less t h a n the aggregate size, the result would be a n unsatisfactory mixture.
This would be the case, for example, in the pharmaceutical industry, in the
preparation of nuclear fuels, a n d in the mixing of pigments.

B . INTERPARTICLE FORCES
T h e n a t u r e a n d relative magnitudes of the cohesive forces are discussed by
H a r n b y et al. (H4). T h e y identify three principal types of interparticle force:
(a) forces d u e to electrostatic charging, (b) van der Waals forces, a n d (c)
forces d u e to moisture.
T h e m a g n i t u d e of electrostatic forces is very d e p e n d e n t o n the n a t u r e of
the particles a n d particularly o n their electrical conductivity. F o r n o n c o n -
ducting particles high cohesive stresses, in the range 1 0 to 10 N / m , have
4 7 2

been reported.
V a n der Waals forces are inversely proportional to the square of the
distance between the surfaces of two particles. Particle shape is also impor-
tant; for example, the force between a sphere a n d a plane surface is twice that
between t w o equal-sized spheres. V a n der Waals forces are generally several
orders of m a g n i t u d e higher t h a n electrostatic forces.
T h e effect of moisture in the a t m o s p h e r e is to p r o d u c e a layer of adsorbed
vapor o n the surface of the particles, at humidities below a critical value, a n d
to form liquid bridges at higher humidities. T h e critical relative humidity is
typically in the range 65 to 80%. T h e attractive force d u e to the adsorbed
layer m a y be a b o u t 50 times the van der Waals force for s m o o t h surfaces, b u t
roughness will reduce, a n d m a y even eliminate, the effect. T h e presence of a n
adsorbed water layer will increase the van der Waals force, since it effectively
increases the size of the particles, b u t the layer will considerably reduce
electrostatic effects by providing a conducting path. At atmospheric h u m i d i -
ties above the critical value a liquid bridge forms between two particles that
are touching or nearly touching. This gives rise to a n attractive force between
the two particles d u e to surface tension, or capillary forces. W h e n a liquid
bridge forms, t h e force between the particles increases by a n order of magni-
t u d e c o m p a r e d with those observed w h e n the moisture creates only a n
adsorbed layer o n the particles.

C . SELECTION OF MIXERS
W h e n cohesive particles are to be mixed a n d the scale of scrutiny is very
small (i.e., less t h a n a b o u t 1 c m ) , it is essential that the action of the mixer
3

will break a n y aggregates t h a t m a y form. It is n o t necessary to ensure that


aggregates d o n o t form: it is enough t h a t they have a short life, so that
aggregates will change, until eventually they have the proportions of the
302 John C. Williams

required mix. T h e mixer must, therefore, have a region in which aggregates


are b r o k e n a n d a circulation pattern which ensures that aggregates are peri­
odically fed to the breakage zone.
T u m b l i n g mixers would n o t be suitable in such applications; in fact, a
t u m b l e r would be a n effective aggregator, b u t would treat the aggregates too
gently to break t h e m . However, a t u m b l e r in which there is a high-speed
agitator bar, for example, the twin shell blender shown in Fig. 5, would be
quite suitable. Other types of mixers that would be effective include edge
r u n n e r mills, in which the aggregates would be b r o k e n between a wheel a n d a
flat plate, a n d high-shear sigma blenders of the type generally used for pastes,
in which high shear rates are generated.

D. SURFACE COATING OF PARTICLES


In s o m e situations the c o m p o n e n t s to be mixed contain particles of very
different sizes, e.g., o n e having a m e a n size of 100 μτη a n d the other a b o u t 1
μηι; a n example would be in the dispersion of a pigment a m o n g polymer
particles. In such a case segregation will not create a problem, since any fine
particle coming into contact with the surface of a large particle will be held in
position by a force greater t h a n its weight a n d will not easily be removed.
This b o n d i n g force will be higher t h a n the force between two fine particles.
T h e aim, therefore, is to break u p aggregates of fines a n d spread individual
particles over the surface of the larger particles.
This process is usually referred to as "ordered mixing," since the expected
o u t c o m e is to p r o d u c e a mixture in which the standard deviation for the
composition of samples removed from it will be less t h a n that expected from
a r a n d o m mixture. T h e properties of ordered mixtures were studied by the
late Dr. J o h n Hersey a n d his colleagues ( H 7 , H 8 , Y l , Y2), with particular
reference to mixing in the pharmaceutical industry. T h e type of e q u i p m e n t
used to bring a b o u t this type of mixing is o n e in which there is enough shear
to break u p aggregates of fines a n d enough circulation to give opportunities
for contact between coarse a n d fine particles. Again, t u m b l e r mixers with
agitator bars are suitable, as are mixers with high-speed stirrers.

V I I I . Conclusion
Considerable advances have been m a d e in the understanding of solid
particle mixing, particularly with regard to the difficulties caused by segrega­
tion, the behavior of cohesive materials, a n d the use of statistics in the
m e a s u r e m e n t a n d description of mixture quality. For the future the m a i n
need is to see this i m p r o v e d understanding of the mixing process reflected in
industrial practice.
A n a t t e m p t has been m a d e in this chapter to set out the basic statistics of
16. Mixing of Particulate Solids 303

p o w d e r mixing. A n u n d e r s t a n d i n g of these principles is essential to a n y o n e


w h o has t o specify the mixing quality required for a given process a n d set u p
procedures for testing the p r o d u c t to d e t e r m i n e whether it conforms to the
specification. It is c o m m o n for a decision to be m a d e o n the acceptance or
rejection of a mixed batch as a result of sampling tests. Because of failure to
appreciate the width of the confidence limits associated with a standard
deviation obtained from sampling tests, the decision is frequently m a d e o n
i n a d e q u a t e information; the probability that the wrong decision will be
m a d e m a y be nearly 50%.
W h e n the particle mixing in a n industrial process is not satisfactory, the
m o s t likely reason is that a n unsuitable type of mixer has been included in
t h e plant. T h e m o s t i m p o r t a n t aspect of particle mixing is to choose a mixer
t h a t is capable of dealing with the materials to be handled. A n understanding
of the principles of mixer selection is therefore vital a n d n o effort should be
spared to ensure that a suitable mixer is selected when designing a process.
M o r e consideration should be given to the possibility of using c o n t i n u o u s
mixing. This is particularly attractive for free-flowing materials t h a t are very
p r o n e to segregation, since it is easier t o m i x such materials in a c o n t i n u o u s
mixer t h a n in a batch mixer, a n d the problems of handling a n d storing the
m i x t u r e are avoided.
M o r e m e a s u r e m e n t s are needed of the performance of c o n t i n u o u s mixers
of different types. Since the performance of a c o n t i n u o u s mixer depends
primarily o n the correlation index of composition fluctuations in the input
stream, m o r e information is needed a b o u t the type of variations in feed rate
t h a t can be expected from feeders of different types; such fluctuations would
need to be m e a s u r e d for a t i m e interval very m u c h less t h a n the m e a n
residence t i m e of the mixer to be used.
F u t u r e research is likely to be concerned with the mixing of cohesive
materials a n d with the d e v e l o p m e n t of new m e t h o d s for bringing together
particles of different c o m p o n e n t s to p r o d u c e ordered mixtures of high
quality.

List o f Symbols

C s coefficient of segregation [Eq. (15)]


C v coefficient of variation (S/μ or S/x)
D particle diameter
D B diameter of bed particles
D T diameter of tracer particles
Ε external age distribution
F(x) mass fraction less than size χ
I internal age distribution
j measure of fallofF in correlation coefficient R(r) as r increases; defined in Eq. (19)
304 John C. Williams

L constant for a given mixture [Eq. (11)]


m number of black particles in a sample
Ν number of samples
η number of particles in each sample
Ρ fraction o f o n e c o m p o n e n t in a binary mixture
R{r) correlation coefficient for points separated by r, defined in Eq. (16)
S estimate of standard deviation o f sample compositions
S 2
estimate of variance
SCI serial correlation index, defined in Eq. (17)
V v o l u m e capacity of a mixer
VRR variance reduction ratio, σ /σ
2 2

ν v o l u m e feed rate
v 0 initial horizontal velocity
W weight of sample
W a proportion of coarse particles in top half of bed
Wb C proportion of coarse particles in bottom half of bed
w m e a n particle weight
X B displacement of bed particles
Χ τ displacement of tracer particles
χ particle size
y composition of samples by weight fraction
y m e a n value of sample compositions by weight
a n proportion of one c o m p o n e n t in «th sample
η fluid viscosity
θ m e a n residence time, V/v
μ true fractional composition
p B density o f bed o f particles
p s density o f solid particles
p T density of tracer particles
σ true standard deviation
σ χ standard deviation o f sample compositions going into a continuous mixer
σ 0 standard deviation of sample compositions going out o f a continuous mixer (or stan­
dard deviation for samples taken from unmixed components)
σ κ standard deviation for a random mixture
steady-state variance for a continuous mixer
τ reduced time, t i m e / 0

References
(Al) Adams, J. F. E., and Baker, A. G., Trans. Inst. Chem. Eng. 3 4 , 91 (1956).
(A2) Allen, T., and Khan, Α. Α., Chem. Eng. 2 3 8 , CE108 (May 1970).
(A3) Ashton, M., and Valentin, F. F., Trans. Inst. Chem. Eng. 4 4 , Τ 1 6 6 (1966).
(BI) Beaudry, J. P., Chem. Metall. Eng. 5 5 , 112 (July 1948).
(CI) Campbell, H., and Bauer, C , Chem. Eng. 73(19), 179 (1966).
(C2) Campbell, A. P., and Bridgwater, J., Trans. Inst. Chem. Eng. 5 1 , 72 (1973).
(C3) Cooke, M. J., Stephens, D . J., and Bridgwater, J., Powder Technol 15, 1 (1976).
(C4) C u m m i n s , J. D . , Convention o n Advance in Automatic Control, Institute of Mechanical
Engineers, Nottingham, 5 - 9 April, 1965.
( D l ) Danckwerts, P. V., and Sellers, S. M., Ind. Chem. 27, 395 (1951).
16. Mixing of Particulate Solids 305

(D2) Danckwerts, P. V., Chem. Eng. Sci. 2, 1 (1953).


(D3) Donald, Μ. B., and R o s e m a n , B., Br. Chem. Eng. 7, 749 (1962).
(Gl) Goldsmith, T. L., Statistician 16, 1 (1966).
(HI) H a m m o n d , P. H., and Barber, D . L. Α., Trans. Soc. Inst. Technol. 17, 59 (1965).
(H2) Harnby, N . , Powder Technol. 1, 9 4 (1967).
(H3) Harnby, N . , Powder Technol. 5, 155 ( 1 9 7 1 / 7 2 ) .
(H4) Harnby, N . , Edwards, M. F. Ε., and N i e n o w , A. W., eds., "Mixing in the Process Indus­
tries." Butterworth, London, 1985.
(H5) Harris, J. F. G., and Hildon, A. M., Ind. Eng. Chem. Process Des. Dev. 9, 363 (1970).
(H6) Harwood, C. F., Walanski, K., Luebcke, Ε., and Swanstrom, C , Powder Technol. 1 1 , 2 8 9
(1975).
(H7) Hersey, J. Α., Powder Technol 1 1 , 4 1 (1975).
(H8) Hersey, J. Α., Thiel, W. J., and Yeung, C. C , Powder Technol. 24, 251 (1979).
(Kl) Kramers, H., and Alberda, G., Chem. Eng. Sci. 2, 173 (1953).
(K2) Kristensen, H. G., Powder Technol. 7, 249 (1973).
(LI) Lacey, P. M. C , Trans. Inst. Chem. Eng. 2 1 , 53 (1943).
(L2) Lacey, P. M. C., / Appl. Chem. 4, 2 5 7 (1954).
(L3) Lawrence, L. R., and Beddow, J. K., Powder Technol. 2, 2 5 3 (1968).
(L4) Levenspiel, O., "Chemical Reaction Engineering," 2nd ed. Wiley, N e w York, 1972.
(PI) Poole, K. R., Taylor, R. F., and Wall, G. P., Trans. Inst. Chem. Eng. 4 2 , T 3 0 5 (1964).
(P2) Poole, K. R., Taylor, R. F., and Wall, G. P., Trans. Inst. Chem. Eng. 4 3 , T261 (1965).
(Rl) R a h m a n , Μ. Α., P h . D . thesis, U n i v . o f Bradford, 1970.
(51) Schofield, G , and Tookey, D . J., Paper presented at the Third European Conference o n
Mixing, British Hydromechanics Research Assoc., York, England, 1979.
(52) Scott, A. M., and Bridgwater, J., Ind. Eng. Chem. Fundam. 14, 2 2 (1975).
(53) Sharif, Μ., M. Sc. thesis, U n i v . of Bradford, 1966.
(54) Shinohara, K., Shoji, K., and Tanaka, T., Ind. Eng. Chem. Process Des. Dev. 9, 174
(1970).
(55) Snedecor, G. W., and Cochran, W. C , "Statistical Methods," 7th ed., Iowa State Univ.
Press, A m e s , Iowa, 1980.
(56) Sugimoto, M., Endoh, K., and Tanaka, K., Kagaku Kogaku 4 , 348 (1966).
(57) Sugimoto, M., Endoh, K., and Tanaka, K., Kagaku Kogaku 3 1 , 145 (1967).
(58) Sugimoto, M., Kagaku Kogaku 3 2 , 291 (1968).
(59) Sugimoto, M., and Y a m a m o t o , K., / . Soc. Mater. Sci. Jpn. 2 2 , 6 8 4 (1973).
(Wl) Walker, O. S., and Cholette, Α., Pulp Pap. Mag. Can., 113 (March 1958).
(W2) Williams, J. C , Fuel Soc. J., U n i v . Sheffield 14, 2 9 (1963).
(W3) Williams, J. C , Powder Technol. 3 , 189 ( 1 9 6 9 / 7 0 ) .
(W4) Williams, J. C , and Khan, Μ. I., Chem. Eng. 2 6 9 , 19 (1973).
(W5) Williams, J. C , and R a h m a n , Μ. Α., / . Soc. Cosmet. Chem. 1(21), 3 (1970).
(W6) Williams, J. C , and R a h m a n , Μ. Α., Powder Technol. 5, 87 ( 1 9 7 1 / 7 2 ) .
(W7) Williams, J. C , and R a h m a n , Μ. Α., Powder Technol. 5, 307 ( 1 9 7 1 / 7 2 ) .
(W8) Williams, J. C , and Richardson, R., Powder Technol. 3 3 , 5 (1982).
(W9) Williams, J. C , and Shields, G., Powder Technol. 1, 134 (1967).
(Yl) Yeung, C. C , and Hersey, J. Α., Powder Technol. 2 2 , 127 (1979).
(Y2) Y i p , C. W., and Hersey, J. Α., Powder Technol. 16, 189 (1977).
INDEX

Agitation intensity, 2 0 1 , 2 3 4 - 2 4 0 , 2 5 1 , 256 Circulation, see Circulation rate


bulk velocity, 2 3 7 - 2 3 8 Circulation flow model, 1 3 7 - 1 3 9
definition, 2 3 4 Circulation rate, 2 1 6 - 2 1 7 , 2 1 8 , 225
superficial velocity, apparent, 2 3 7 - 2 3 8 , for shear, m i n i m u m required, 225
240 torque per unit v o l u m e measure, 2 2 5
turnover rate, 238 Circulation time, 138, 181, 185, 186
turnovers, 2 4 0 Circulation time, frequency function, 1 3 9 -
turnover time, 238 140
Agitation intensity, measures of, 2 2 2 , 2 3 5 - Coal slurry suspension operation, 2 2 2
237 Cohesive forces between particles, 301
Chemscale, 2 3 7 - 2 3 9 Cohesive particles, mixing of, 3 0 0 - 3 0 2
impeller tip speed, 235 aggregate formation, 3 0 0 - 3 0 1
index numbers, 235 size reduction, effect of, 3 0 0
power intensity, 235 Composition of solids mixture, 2 6 7 - 2 6 8
verbal hierarchies, 235 by estimate from samples, 267
Agitation scale-up, see Scale-up; Scale-up de- Computation, scale-up test data, 2 4 9 - 2 5 0
sign correlations check against design correlations, 2 5 0
Axial-flow impeller, 134, 1 4 3 - 1 4 4 , 1 7 3 - 1 7 6 check with full-scale data, 2 5 0 , 2 5 2 - 2 5 3
pitot tubes for flow measurement, 144, lab-scale system characteristics, 2 5 0
148, 166 Continuous mixing of solids, 2 9 0 - 3 0 0
Rankine vortex, 144 advantages, 2 9 0
tracer particles for flow measurement, backmixing, 2 9 1 , 2 9 3
138, 144 cohesive powders in ribbon mixer, 2 9 3
velocity, radial average, 139 in drum mixers, 2 9 4 , 2 9 5 , 298
Axial fluid velocity ideal mixer, 2 9 2
below impeller, 164 mechanistic approach, 299
for pitched-blade turbine, 1 4 8 - 1 4 9 nonideal mixer, 2 9 3
Axial force o f axial impeller, 164, 1 7 3 - 1 7 7 problems, 2 9 0
residence time distribution, 2 9 4 - 2 9 5
s t i m u l u s - r e s p o n s e approach, 2 9 9 - 3 0 0
Bench scale versus plant scale tests, 2 0 8 , 256 Continuous reactor scale-up, see Reactors,
Blending, see Homogenization continuous, scale-up
Blending operations, scale-up, 2 0 7 , 2 1 2 - 2 1 3 , Continuous solids mixing equipment, 291
2 2 6 - 2 3 1 , 2 3 3 - 2 3 4 , 236, 238, 239 Continuous solids mixing performance,
agitation intensity, 2 3 8 , 2 3 9 feed composition fluctuation effect, 298
miscible fluids, 238 models, flow and mixing, 2 9 3 - 2 9 5
translation equations, 2 3 0 - 2 3 1 , 2 3 3 - output composition prediction, 2 9 6 -
234 298
Blending time scale-up, 2 1 2 - 2 1 3 , see also parameters, 2 9 1 - 2 9 2
Blending operations, scale-up residence time distribution, 295
high-viscosity fluids, 2 1 3 segregation in mixer, effect, 298
low-viscosity fluids, 2 1 2 - 2 1 3 steady-state variance, 2 9 8 - 2 9 9
pitched-blade turbines, 2 1 2 Convective flow for axial impellers, 1 3 9 - 1 4 2
propellers, 2 1 2
Rushton turbines, 2 1 2 D / T ratio
Bulk velocity, see Agitation intensity blending time, effect on, 208, 2 1 2

307
308 Index

D / T ratio (continued) regime trend from viscous to turbulent,


discharge coefficient, effect o n , 2 2 4 215
distribution o f circulation rate and head, Flow at tank bottom, 1 5 9 - 1 7 3
218 axial pressures acting, 1 6 2 - 1 6 6
optimum, 2 4 0 - 2 4 1 D / T effect, 172
pumping number, effect on, 217 impeller distance from tank bottom, ef-
at tank bottom, effect of flow, 172 fect of, 171
Dead region, 1 5 7 , 1 8 8 , see also Solids suspen- Laplace equation model, 160
sion stream function, model variable, 160
Design correlations for scale-up, see Scale-up streamlines defining velocity field, 166 —
design correlations 171
Dimensionless numbers, see Property force Flow at vessel wall, 177 - 1 8 0
ratios; Process dimensionless ratios. Flow rate from impeller
Discharge coefficient of disk turbine, 2 2 4 induced, 137, 144, 177
Draft tube, 157 primary, 137, 143
flow characteristics, effect on, 1 5 7 - 1 5 8 primary to induced flow rate ratio, 159
total, 137, 180
Eddy diffusion turbulence model, 7 5 - 7 7 Flow rate number
Empirical equations, source of, see also Ex- induced, 137
tended principle of similarity primary, 137
dimensional analysis, 208 total, 137
N a v i e r - Stokes equation, 2 0 1 , 208 Flow ratio, p r i m a r y - i n d u c e d , 159
Emulsion formation, 222 Flow-sensitive operations, 133
Energy dissipation in stirred vessel, 154, see Flow stream lines, 1 4 4 - 1 5 9 , 1 6 5 - 1 7 3
also Power response Froude number, 110, 115, 2 0 3 - 2 0 6
Energy levels in agitated charge, 153, 1 6 9 -
171
G a s - l i q u i d operations,
Equal fluid motion operations, 230, 231 —
flooded system, 2 1 2 , 2 1 3 - 2 1 4 , 2 2 0 - 221
232, 2 3 3 - 2 3 4
loaded system, 2 1 2 , 2 1 3 - 2 1 4 , 2 2 0 - 2 2 1
Euler number, see Power number
G a s - l i q u i d operations, scale-up
Extended principle of similarity, 200, 2 0 1 ,
agitation intensity, 2 3 4
252, 253
selection tables, 2 1 2
Geometric variation in scale-up, 2 0 1 , 2 0 8 ,
Fillet formation, 10
209, 2 4 0 - 2 4 3
Flow direction and velocity, 2 1 5 - 2 1 6 , 217
C/T, 2 4 1 , 2 4 2
geometry effect on, 215
D / T , 2 4 1 , 2 4 2 , see also D / T ratio
modeling, 2 1 5 - 2 1 6 , 217
effect o n liquid flow patterns, 2 1 5 - 2 1 6 ,
symmetrical zone formation, 2 1 5 - 2 1 6
241,242
toroid formation, 2 1 5 - 2 1 6
Z/T, 2 4 1 , 2 4 2 - 2 4 3
Flow sensitive operations, 2 2 2 , 2 4 1 , 2 5 0 , 2 5 1 ,
see also Blending operations scale-up;
Heat transfer scale-up; Solids suspen- Head, see Shear
sion operations, scale-up Heat transfer mechanism in agitated vessels,
Fluid-flow pattern model by N a v i e r - S t o k e s 188-192
equation, 201 axial impellers, 1 8 8 - 1 9 2
Fluid mechanics in scale-up, 2 1 3 - 2 2 1 see at tank bottom, 1 8 8 - 1 9 0 , 192
Flow direction and velocity; Shear; at tank walls, 1 8 8 - 1 8 9 , 191, 192
Circulation rate; Turbulence spec- Heat transfer scale-up, 210, 211
trum; Phase density, effect of agitation intensity, 2 3 4
mixed regimes, existence of, 2 1 4 - 2 1 5 to coils, 211
Index 309

difficulties with coils, 211 g a s - l i q u i d , complexity of, 211


at vessel walls, 2 1 0 - 2 1 1 limitations, 2 0 8 - 2 0 9
Hindered settling, 5 l i q u i d - l i q u i d process, 2 1 2
Homogenization, see also Blending. Mechanical power, see Power response
H o m o g e n i z a t i o n - circulation relationship, M i x e d regimes, 2 1 4 - 2 1 5 , 2 4 3 , 2 4 5 , 2 5 4 , 255
180-188 definition, 245
Homogenization o f concentration, 1 8 0 - 1 8 8 d o m i n a n t process, 2 4 5 , 255
Homogenization rate in agitated vessel, 1 8 0 - shift in dominant process with scale-up,
188 2 4 5 , 2 5 0 - 2 5 1 , 2 5 4 , 255
Homogenization time, 182, 186 Mixing o f solids, continuous, see Continuous
Hoppers for solids storage, 2 8 4 , 2 8 5 mixing of solids
core flow, 2 8 0 , 2 8 4 M o m e n t u m flux, 88
mass flow, 285 M o m e n t u m ratio, 85
Hydraulic efficiency o f impeller, 1 5 4 - 1 5 5
N a v i e r - S t o k e s equation, see Fluid-flow pat-
Ideal mixer, see Continuous mixing o f solids tern model
Impeller N e w t o n number, see Power number
cast, 2 5 5 Nonideal mixer, see Continuous mixing of
flat-blade disk turbine, 2 1 2 , 2 3 0 solids
flat-blade turbine, 2 1 2 , 2 3 6 N o n r a n d o m solid mixture properties, 2 7 3 -
pitched-blade turbine, 134, 1 4 3 - 1 4 4 , 274
159, 1 7 3 - 1 7 6 , 2 1 2 , 217, 2 3 0 , 2 3 6
propeller, 143, 1 7 6 , 2 1 2
Ordered mixing o f solids, 302
Rushton turbine, see Impeller, flat-blade
Orifice mixer, 70, 9 4 - 9 6 , 106
disk turbine
Impeller-induced flow rate
primary, 137 Particle concentration fluctuation, 9
Impellers, multiple, in scale-up, 2 4 3 Particle concentration patterns, 7
automatic staging, 2 4 3 Particle concentration profiles, 14
rule o f symmetry, 2 4 3 Particle concentration uniformity, 1 2 - 1 5
staged system, 2 4 3 Particle-liquid system, types, 4 - 5
Injector mixer, 6 4 Bingham-plastic rheology, 5, 4 4
Interparticle forces, see Cohesive forces be- dilatant rheology, 4 4
tween particles Newtonian, 4
pseudoplastic, 4, 4 4
Particle mixing, see Solids mixing
Jet mixers, radial in tanks, 2 4 0
Particle m o t i o n in impeller-stirred tanks, 5 - 9
Particle settling
Kenics mixer, 6 5 , 67, 9 4 - 9 6 , 107
multiple-particle systems, 4
K o c h mixer, 6 9 , 9 4 - 9 5 , 107, 111
quiescent liquids, 2
Particle settling velocity
Lightnin Inliner® mixer, 6 9 , 107 impeller-stirred tanks, 6
Particle suspension
Mass transfer, see Mass transfer scale-up; bottom shape effect, 27
G a s - l i q u i d operations; G a s - l i q u i d continuous flow through tanks, 3 8 - 4 3
operations, scale-up correlations, 1 5 - 3 2
Mass transfer, particle-liquid, 4 8 - 5 6 criteria, 9 - 1 5 , 2 8 - 3 0
Mass transfer scale-up, 2 0 8 - 2 0 9 , 2 1 1 - 2 1 2 density difference effect, 21
to crystals, 211 equipment-size effect, 3 0 - 3 1
to gas bubbles, 2 1 1 - 2 1 2 fluid viscosity effect, 22
310 Index

Particle suspension (continued) swirling feed stream, 79


free settling in impeller-stirred tanks, 5 - tangential feed stream, 64, 8 9 - 9 1 , 9 2 -
38 93, 99, 1 0 2 - 1 0 4
high concentration, 4 3 - 4 8 , see also Hin- tee, 64, 8 2 - 8 8 , 106, 118
dered settling thin-layer feeds, 65, 82
impeller blade thickness effect, 31 type selection, 1 1 6 - 1 1 7
impeller clearance effect, 2 2 Pipe mixing
impeller-type effect, 2 4 - 2 7 different density fluids, 1 1 0 - 1 1 1
impeller-vessel diameter ratio effect, 23 grid effect, 94, 9 7 - 9 9 , 118
particle concentration effect, 21 low Reynolds number, 85, 92
particle diameter effect, 21 Reynolds number effects, 115, 1 1 7 - 1 1 9
scale up, 3 0 - 3 1 , 33, 36, 47, see also screen effect, 94, 9 7 - 9 9
Solids suspension operations, scale-up Pipe mixing measuring methods, 73
Particle suspension equipment, 9, 4 4 electrical conductivity, 73
design, 3 2 - 3 7 light absorption, 73
impeller speed selection, 3 3 - 3 6 temperature change, 73
selection, 3 2 - 3 3 thermal conductivity, 7 3 , 99
Particle velocities Pipe mixing performance criteria, 7 0 - 7 3
free-settling particles, 3 concentration fluctuation decay, 75
Percent suspension, 14 concentration spread, 70, 7 5 - 7 7
Peripheral velocity of impeller constant, 2 2 2 fraction completion of reaction, 7 0
Phase density, effect of mixing time, 70
g a s - l i q u i d systems, 2 2 0 - 2 2 1 pipe diameters to mix, 70, 7 7 - 8 5 , 8 7 -
l i q u i d - l i q u i d systems, 221 98, 1 0 0 - 1 0 4 , 1 1 7 - 1 2 4
Pipe mixer relative standard deviation, 7 1 - 7 3 , 89,
application, 124 90, 9 5 , 9 7 , 118, 1 1 9 - 1 2 3
axial plus radial feeds, 9 9 - 1 0 1 , 111 segregation intensity, 7 1 - 7 2 , 100
axial and tangential feeds, 99, 1 0 2 - 1 0 4 standard deviation, 7 0 - 7 3 , 77, 9 1 , 95,
baffled, 6 7 - 7 0 , 9 4 - 9 6 , 107, 1 2 0 - 1 2 2 120-122
cavitation, 1 1 1 - 1 1 2 , 114 variation coefficient, 7 1 - 7 2 , 77, 79, 83,
coaxial, unequal-velocity feeds, 7 9 - 8 0 85, 87, 114
design, 1 1 3 - 1 2 4 Powder mixing, see Solids mixing
elbow, 96, 118 Power dissipation
flow instability, 112, 114 distribution, 157, 170
fluid pressure, 114 Power intensity constant, see Power per unit
head loss, 1 0 4 - 1 1 0 v o l u m e scale-up constant for scale-up
injector, 6 4 Power investment, see Power response
multiple parallel feeds, 65, 8 1 , 82, 104, Power number, 2 0 2 , 2 0 4
118 definition, 2 0 2 , 2 0 4
multiple, radial-jet feed streams, 66, 9 0 - Power number correlation, see Power re-
92, 118 sponse scale-up
multiple side-stream, 9 0 - 9 3 , 99, 1 0 3 - Power per unit v o l u m e constant for scale-up
104, 118 myth demolished, 199, 2 2 2
n o flow modifiers, 7 4 - 7 9 , 118 rule, 199, 2 2 1 - 2 2 2
oblique side stream, 87 Power, pipe mixing, 105, 108
orifice, 70, 9 4 - 9 6 , 100, 106, 118 Power, process requirement, see Power re-
pipe tee o p t i m u m dimensions, 8 5 - 8 7 , quirement
90-91 Power requirement, 199, 208, 2 0 9 - 2 1 0
pressure loss, 1 0 4 - 1 1 0 definition, 208, 2 0 9 - 2 1 0
short mixing time, 1 1 3 - 1 1 4 , 124 failure to meet, 2 0 9 - 2 1 0
Index 311

Power response, 2 0 8 , 2 0 9 - 2 1 0 , see also Reactors, batch, scale-up, 2 4 5 - 2 4 6


Power number; Power requirement polymerization processes, 246
definition, 2 0 8 , 2 0 9 - 2 1 0 rapid reactions, 246
non-Newtonian fluids, 209 shear levels, 2 4 6
versus power requirement, 2 0 8 , 2 0 9 - Reactors, continuous, scale-up, 2 4 6 - 2 4 7
210 pipe-line mixing, 247
shear and circulation rate, depends o n , Retangular duct mixing, 110
216-217 Ross mixer, 68, 9 4 , 107
in slurries, 3 6 - 3 7 , 4 7 - 4 8 Rules of t h u m b for scale-up, 2 0 0 - 2 0 1 , see
Power response scale-up, 2 0 9 - 2 1 0 also Peripheral velocity of impeller
correlations, 2 0 2 , 2 0 4 constant; Power per unit v o l u m e c o n -
Pressure loss, pipe mixer, 1 0 4 - 1 1 0 stant for scale-up; Superficial velocity
scale-up, 110 constant; Torque per unit v o l u m e
Pressure at vessel bottom, 163, 165, 166 constant
Process dimensionless ratios, 207, 208 power intensity constant, 2 0 0
for blending time, 207, 208 torque intensity constant, 201
definition, 2 0 8
empirical equations, use in, 207, 208 Sample size, solids mixture, 2 6 7 , 2 8 7 , 2 8 9 , see
for heat transfer, 2 0 7 , 208 also Scale o f scrutiny
Nusselt number, 211 Scale-down tests, 247, 2 5 0
Propeller, 143, 1 7 6 , 2 1 2 to check operation, 247
lifting effect of blades, 141, 144 to optimize operation, 247, 2 5 0
performance comparison with pitched- Scale o f scrutiny, 266, 2 8 7 , 3 0 0 - 3 0 1
blade turbine, 159 Scale-up
Property force ratios, 2 0 2 approach to, 2 0 0
definition, 2 0 2 , 2 0 3 an art, 200, 256
dynamic similarity, 202 difficulties in, 2 0 0
empirical equations, use in, 2 0 2 from laboratory tests, 201
Froude number, 2 0 3 - 2 0 6 mistaken notions, 199
Reynolds number, 2 0 2 , 206 from pilot plant tests, 201
Pumping capacity of axial impellers, 137, shear rates controlling, 255
142-144 uncertainty in, 1 9 9 - 2 0 0 , 2 5 1 , 2 5 3 , 256
Pumping number, 2 1 6 , see also P u m p i n g universal correlation, 209
rate, effective volumetric Scale-up design correlations, 2 0 0 , 2 0 8 - 2 1 3 ,
Pumping rate see also, Blending time scale-up; Ex-
impeller, 137, 142, 144, 176 tended principle o f similarity; Heat
pitched-blade turbine, 1 4 2 - 1 4 4 , 176 transfer scale-up; Mass transfer scale-
propeller, 143, 176 up; Power response scale-up
Reynolds number effect, 144 Scale-up guidelines, 2 3 0 - 2 3 1
Pumping rate, effective volumetric, 2 2 2 , 2 2 4 dispersion operations, 2 3 0
equal blend time, 2 3 0
equal fluid m o t i o n , 2 3 0 , 231
solids suspension, 2 3 1 - 2 3 4
Radial jet mixer, 66, 9 0 - 9 2 , 118 surface effects, 2 3 0
R a n d o m solids mixture properties, 266, 267, Scale-up operations
269-273 contributing factors to complexity, 2 0 0
chi-square distribution, 2 7 2 in laminar regime, 226, 2 2 7 , 2 3 4
by standard deviation, 2 6 9 - 2 7 3 Scale up, particle suspension, 3 0 - 3 1 , 33, 36,
by statistical analysis, 2 7 1 - 2 7 3 4 6 , 4 7 , see also Solids suspension oper-
Rankine vortex, see Axial flow impeller. ations, scale-up
312 Index

Scale-up procedures, 2 5 0 - 2 5 5 o n handling, 274, 275


characterize operation, 2 5 0 - 2 5 1 by heap-pouring effect, 276, 2 7 7 , 2 8 4
extrapolation from several results, 251 — mechanisms, 2 7 5 - 2 7 7
253 particle density difference, 2 7 4
extrapolation from single result, 2 5 1 , 2 5 3 by percolation, 2 7 5 - 2 7 6 , 278
o p t i m u m geometry, 2 5 4 radial segregation, 277
preliminary examination, 2 5 0 - 2 5 1 by shear plane, 278
process result specification, 2 5 0 with time o f mixing, 2 7 4 - 2 7 5
space constraints, 251 by trajectory segregation, 275
time limits, 251 unmixing, 2 7 4
Scale-up relations, see Extended principle o f by vibration, 2 7 6 , 278
similarity Segregation scale, 7 3
Scale-up rules, see Rules o f t h u m b for scale- Serial correlation index, 2 9 2 , 297
up Shear
Scaling based o n tests, 2 4 7 - 2 5 0 for circulation, m i n i m u m required,
charge properties, 2 4 8 - 2 4 9 225
purpose, 247 head measure, 225
research results versus model tests, 248 power per unit v o l u m e measure, 225
test set-ups, 248 turbulence measure, 225
vessel sizes for tests, 248 Shear rate
SCI, see Serial correlation index average impeller, 2 1 8 - 2 1 9
Segregation behavior o f solids mixtures, 2 8 1 , m a x i m u m impeller, 2 1 8 - 2 1 9 , 2 2 2
283 modification, 2 2 0
converging channel flow, 2 8 3 throughout vessel, 2 2 0
twin-shell blender, 2 8 1 , 2 8 3 variation with scale-up, 219
water addition, effect of, 2 8 3 Shear sensitive operations, 2 2 2 , 2 4 1 , 2 5 0
Segregation measurement of solids mixtures. Similarity, types o f
277-284 chemical, 201
correlation for plug flow, 2 7 9 - 2 8 0 dynamic, 2 0 1 - 2 0 2
diameter ratio effect, 278, 280, 2 8 3 fluid dynamics, 201
measures, 2 8 0 geometric similarity, 2 0 1 - 2 0 2
in plug flow, 2 7 9 - 2 8 0 kinematic, 2 0 1 - 2 0 2
in rotating drum, 280 thermal, 201
size of particle, effect of, 280, 281 Similarity principles, application of, 2 0 1 -
shear field effect, 2 7 8 208, see also Extended principle o f
o n vibrated channel, 2 7 7 - 2 7 8 similarity
Segregation reduction of solids mixtures, addition o f parameters, 2 0 6
2 8 3 - 2 8 5 , 300, 302 example, worked, 2 0 2 , 2 0 4 - 2 0 6
coating larger particles with fines, 284, extrapolation o f small-scale results, 201
302 limitations, 206
by cohesiveness increase, 300 viscosity, minimizing effect of, 2 0 6
core flow hopper, avoid flow to, 2 8 4 Simultaneous processes in scale-up, 2 4 3 , 2 4 5 ,
liquid addition, 2 8 3 , 284, 300 254, 255, see also Mixed regimes
by particle size reduction, 300 Slip velocity, 7
Segregation o f solids mixtures, 2 6 6 , 2 7 4 - 2 8 5 Sloping-duct mixing, 110
axial segregation, 277 Slurry, 1
diameter ratio effect on percolation, complete m o t i o n in, 4 6 - 4 7
278-279 Solids mixer classification, 2 8 5 - 2 8 6
difference in particle size, 2 7 4 Solids mixer selection, 280, 2 8 5 - 2 8 7
by elutriation during hopper filling, 277 for aggregate dispersion, 3 0 1 - 3 0 2
Index 313

Solids mixer testing, 2 8 8 - 2 9 0 Tank exit particle concentration change,


sample number, 2 8 8 - 2 8 9 40-41
sample removal technique, 2 8 8 - 2 9 0 Tee mixer, 64, 8 2 - 8 8 , 106, 118
sample size, 289 Tests as basis for scale-up
sample source location, 289 fluid regime change, 209
Solids mixing size of test vessels, 208, 209
batch, 266 Tests for scale-up
cohesive particles, 266 computations from test data, 2 4 9 - 2 5 0
continuous mixing, see Continuous mix- planning tests, 2 4 8 - 2 4 9
ing of solids procedures for tests, 249
fallacies, 266 Test procedures for scale-up, 2 4 9
Solids mixing mechanisms, 2 8 5 - 2 8 7 , 2 9 4 data derived from, 245
c o m m i n u t i o n effect, 286, 287 impeller selection, 249
diffusive mixing, 285, 286 vessel geometry, 245
shear, 285, 286 Torque intensity constant, see Torque per
Solids mixture quality, 2 6 6 , 2 6 8 - 2 6 9 , 2 8 7 - unit v o l u m e constant for scale-up
288 Torque per unit v o l u m e constant for scale-up,
coefficient o f variation, 269 221-225
standard deviation, 2 6 8 - 2 6 9 , 288 geometric similarity, independent of,
variance, 269 223-224
variation allowable from m e a n composi- Torque versus power investment, 255
tion, 2 8 7 - 2 8 8 Total flow in agitated vessel, 1 4 4 - 1 5 9 , 180
Solids mixture quality improvement, 3 0 0 Laplace equation model, 144
by size reduction, 3 0 0 stream function, model variable, 151,
Solids suspension 157, 160, 163, 166, 185
dead region below impeller, 157 streamlines defining velocity field, 145 —
stagnant regions, 4 4 - 4 7 152
Solids suspension operations, scale-up streamline deformation factors, 1 5 1 -
translation equations, 2 3 1 - 2 3 4 152
Spherical mixer, 92 Translation equations for scale-up, 2 0 1 , 2 2 5 -
Square tank configuration, 134 234
Statistics in solids mixing, 2 6 5 - 2 7 4 , see also definition, 225
Solids mixture quality empirical correlation basis, 2 2 6
quality of mixture, 265, 2 6 8 - 2 6 9 examples illustrating their use, 2 2 7 , 2 5 7 -
segregation measure, 265 260
Student t test, 2 6 7 - 2 6 8 , 2 9 0 exponents, range of, 2 2 6 , 2 2 9
Stratification, pipe mixer fluid, 1 1 0 - 1 1 1 linear dimension as variable, 2 2 5 - 2 2 6
Stream function power functions, 2 2 5 - 2 2 6
below impeller, 160, 163, 166 power intensity as variable, 225
Sulzer mixer, 69, 9 4 - 9 5 , 111, 118 process result as variable, 226
Superficial velocity, 2 2 2 , see also Agitation speed ratio as variable, 226
intensity torque intensity as variable, 226
Surface coating of particles, 302 v o l u m e ratio as variable, 2 2 5 - 2 2 6
Suspension percent, 14 Turbine, pitched-blade, performance c o m -
Suspension of solids, scale-up parison with propeller, 159
agitator intensity, 2 3 4 , 2 3 9 Turbulence, see Shear
Turbulence intensity for axial impellers, 139,
1 4 1 - 1 4 2 , 189
Tangential feed mixer, 64, 8 9 - 9 3 , 99, 1 0 2 - Turbulence models, 74
104 Turbulence spectrum, 2 2 0
314 Index

Turnover rate, see Agitation intensity at vessel wall, 1 7 7 - 1 8 0


Turnover time, see Agitation intensity Velocity fluctuation, 1 4 1 - 1 4 2
Velocity head loss
Ultimate particle concentration, 5 pipe mixer, 1 0 6 - 1 0 9 , 1 1 9 - 1 2 4
pipe tee, 109
Velocity profiles by Pitot tube, 148
Variance reduction ratio, 2 9 2 , 2 9 3 , 295, 297
Velocity profile from impeller, 142
Velocity, average bulk, see Superficial veloc-
Viscosity measurement, slurry, 48
ity constant
Vortex formation, 2 0 4
Velocity field, 145
measurement of, 2 0 4
at vessel bottom, 152, 1 6 6 - 1 7 3

You might also like