Geldart - Gas Fluidization Technology PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 238

GAS FLUIDIZATION

TECHNOLOGY

JOHN WILEY & SONS


Chichester . New York . Brisbane . Toronto . Singapore

;9.;i12f iJ, r}.}

C LASSIF ..
AUTOR

C-r:.'gJ::: .

V .................... EX ..__

TOMBO BC/~.L~

Contents

1 Introduction

......

D. Geldart
2

Single particles, Fixed and Quiescent Beds

11

D. Geldart
3

Characterization
D.
Geldart

of Fluidized Powders

Hydrodynamics of Bubbling Fluidized Beds

R. Clift
5
No part of this book may be produced by any means, or transmitted,
or translated into a machine language without the written
permission of the publisher.

Gas fluidization technology.


'A Wiley-Interscience publication.'
Includes index.
1. Fluidization. I. Geldart, D,
TP156.F65G37 1986
ISBN 0 471 908061

8
9

. . ..... .. . .... ..... ..... ..... . .. ... .. ... ... ... ..... .. . 97

Particle Entrainment

D. Geldart

and Carryover

. 123

Beds

155

~~~=:~s~:aration

197

Fluid Bed Heat Transfer

219

I. S.M. Botterill
.10

11

ISBN 0 471 908061


Typeset by Quadra Graphics
Printed and Bound in Great Britain

Solids Mixing

7 ~i~~;,;~ztluidized

British Library Cataloguing in PublicationDaJa:


Gas fluidization technolgy.
1. Fluidization 2, Gases
I. Geldart, D,
660,2'884292
TP156.F65

53

I. Baeyens and D. Geldart

Library of Congress Cataloging-in.Publication Data:


.Main entry under title:

. 33

Fluid Bed Drying


D. Reay

259
.

Fluid Beds as Chl;lmical Reac~ors

285

I.R. Grac~
12

Solids Transfer in Fluidized Systems

T.M. Knowlton

, 341

13

Instrumentation and Experimental Techniques

415

l.R. Grace and l. Baeyens

Contributors
SIPEF SA, Kasteel Calesberg 2120, Schoten, Belgium
Department of Chemical Engineering, University of
Birmingham, PO Box 363, Birmingham B152TT, UK
PROFESSOR ROLAND
CUff,
Department of Chemical and Process
Engineering, University of Surrey, Guildford, Surrey GU2 5XH, UK
DR
DEREK
GELDART;
Postgraduate School of Powder Technology,
University of Bradford, Bradford BD7 1DP, West Yorkshire, UK
PROFESSOR JOHN R
GRACE,
Department of Chemical Engineering,
University of British Columbia, Vancouver, V6T 1W5, Canada
DR TED M. KNOWLTON, Institute of Gas Technology, 4201 W. 36th St.,
Chicago, Illinois 60632, USA
DR
DAVID
REAY,
Engineering Sciences Division, AERE Harwell,
Oxfordshire, OXll ORA, UK
DR LADISLAV SVAROVSKY, Postgraduate School of Powder Technology,
University of Bradford, Bradford BD7 1DP, West Yorkshire, UK
DR JOSEPH YERUSHALMI, PAMA Limited, 7 Kehilat Saloniki St., Tel-Aviv
69513, PO Box 24119, Israel
DR JAN BAEYENS,

DR J.S.M.

BOTIERILL,

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Introduction

1.1

CHARACTERISTICS

OF FLUIDIZED SYSTEMS

A fluidized bed is formed by passing a fluid, usually a gas, upwards through a


bed of particles supported on a distributor. Although it is now known that
even above the minimum fluidization velocity the particles are touching each
other mostof the time, with the exception of cohesive solids the interparticle
friction is so small that the fluid/solid assembly behaves like a liquid having a
density equal to the bulk density of the powder; pressure increases linearly
with distance. below the surface, denser objects sink, lighter ones float, and
wave motion is observed. Solids can be removed from or added to the bed
continuously, and this provides many processing advantages. All fine
powders have a very large specific surface area - 1 m3 of 100 JLm particles
has a surface area of about 30,000 m2 - but in a fluidized bed the stirring
action of the gas bubbles continually moves the powder around, shearing it
and exposing it to the gas; it is this good solids mixing which gives the high
rates of heat transfer from surface to bed and gas to particle, and which is
responsible for isothermal conditions radially and axially. Compared with a
fixed bed of the same powder operated at the same bed depth and gas
velocity, the pressure drop over a fluidized bed is much smaller, and this
together with most of the characteristics described above make the fluidized
bed an attractive choice as a chemical or physical processing tool.
There are, however, disadvantages which may outweigh the attractive
features: for some applications the gas bubbles make scale-up more difficult
and provide a means whereby the reacting gases can avoid contact with the
solids; particle entrainment is almost inevitable and particle attrition and
metal surface erosion occur in regions where gas velocities are high.
The basic component required for a fluidized bed are items 1 to 4 on Fig.
1

1.1 - a container, a gas distributor, a powder, and a source of gas. The


provision of the other elements shown depends on the particular application
contemplated. For example, fluidized bed granulators may include the spray,
screw conveyor, and solids offtake pipe but not the internal cyclone or heat
transfer surfaces; acrylonitrile reactors have internal cyclones and internal
heat exchanger tubes but no spray nozzles, screw conveyors, or standpipes.

The variety of fluidized beds encountered in commercial operation is


enormous and includes powders having mean sizes as small as 15 JLm and as
large as 6 mm, bed diameters from 0.1 to 10 m, bed depths from a few
centimetres to 10 m and gas velocities from 0.01 to 3 m/s or even as high as 10
m/s for recirculating high velocity beds. The behaviour of particulate solids in
fluidized beds depends largely on a combination of their mean particle size
and density (Geldart, 1973), and it has become increasingly common to
discuss fluidized systems in relation to the so-called Geldart fluidization
diagram. Although this is to be discussed in detail in Chapter 3 a simplified
version is shown in Fig. 1.2, and this can be used to identify the 'package' of

6
5

1\

"'E

".E

'F
<>,Q.

-c
Cohesive

0
~

Aerotoble

0.5

Spoutoble

1\

Sand -like

1'\

"

I /

"'r--.

:\
\

Figure 1.2 Simplified diagram for classifying powders. according to their fluidization behaviour in air at ambient conditions (Geldart, 1973).

1.
2.
3.
4.
5.

Shell
Powder
Blower
Gas distributor
Heat exchanger
for fluidizing gas

6. Internal heating or cooling


7. External heating or cooling
8. Cyclones
9. Solids feeder
10. Solids offtake
11.Spray feed

Figure 1.1 A conceptualized fluidized bed which could be used for a wide
Iiety of applications.

fluidization characteristics associated with fluidization of any particular


powder at ambient conditions. A somewhat different 'phase diagram' has
recently been proposed by Grace (1984). This is based on an earlier diagram
proposed by Reh (1968) and can be used to broadly identify flow regimes
appropriate to combinations of gas velocity and particle properties (Fig. 1.3).
The Archimedes number, Ar (or 3/4Co Re2, as it is sometimes written),
characterizes the basic particle/gas properties and Re/Co characterizes the gas
velocity. Thus at low values of Arl13 (small particles) the processing options

Group C
powder

Dimensionless
2 1/3

(314 CeRe I

particle diameter
1/3

=Ar

=d 9 (ppPgl/fL2

}113

Figure 1.3 Regime/processingmode diagram for


grouping systems according to type of powder and
upward gas velocity used (Grace, 1984;after Reh, 1968).

are, in order of increasing gas velocity, fluidized beds, circulating beds,


transport and dilute phase transport reactors; for large particles the options
are fixed, moving/spouted beds, and fluidized beds.

Although the technique of fluidization was in commercial use as early as 1926


for the gasification of coal, it was not until the early 1940sthat its widespread
use began with the construction of the first fluid bed catalytic crackel (FCC).
Various accounts of the development of this and other fluidized bed processes
have been given (Jahnig, Campbell, and Martin, 1980;Kunii and Levenspiel,
1969) and a comprehensive account of the industrial scene up to 1967 was
given by Geldart (1967, 1968, 1969); but the most complete and carefully

researched account is that given by Squires (1982), who has been associated
with process invention and development involving fluid beds since 1946.
Squires provides a fascinating picture of how commercial and wartime
pressure, together with a compromise between daring innovation and the
need to reduce catalyst losses, steered th~ ~~~igQ.a\Va.yfromt.bebigh.velocity
(upflow) mode towarps low velocity (downflow) operation; since that time
more than 350 FCC units have been built and most are still in operation. In
the late 1940sDorr Oliver applied the technique successfully to the roasting
of sulphide ores and since that time virtually all new ore roasters have been
fluidized beds. Fluid bed dryers also made rapid progress and by the mid
1950sthe technique was well established.
However, in all these applications the degree of conversion required was
either not critical (as in FCC) or easily achievable (as in roasting and drying);
a major setback occurred when the large fluidized bed Fischer- Tropsch
synthesis plant at Brownsville in Texas fell far short of the conversions
achieved in the pilot plant. The aim was to use natural gas to manufacture
gasoline, but unlike the other processes, here bubble hydrodynamics were
critical and their .crucial role in scale-up was not properly appreciated.
Although eventually the plant came somewhere near its designed output,
with the discovery of huge quantities of oil in the Middle East, the economics
became so unfavourable that the plant was shut down in 1957.
The Sohio process for making acrylonitrile in a fluidized bed was
immensely successful; since 1960 virtually all new acrylo plants have been
fluidized beds and 50 large units are in operation throughout the world.
Undoubtedly the major success in the late 1970s and in this decade is the
Union Carbide polyethylene synthesis process. The alternative high pressure
liquid phase reactors were limited in scale of operation, whereas in the low
pressure gas phase process single fluid bed units can be built as large as
required; the better quality product and dramatic reduction in costs which are
features of this fluid bed process have ensured the demise of virtually all
competitors. Also in the 1970s and 1980s, fluidized bed combustion has
attracted much attention largely due to its relatively low temperature
operation (800 to 900C) and its ability to absorb S02 through the use of
limestone or dolomite. These features mean that NO" and 802 emissions in
the flue gases can be made acceptably low. Although such units operate at 1
to 2 mis, an atmospheric pressure fluidized bed boiler for commercial power
generation would constitute an extremely large pice of equipment. The more
compact pressurized fluidized bed combustors are therefore attractive but
commercially are still a long way off. A great many small atmospheric
pressure fluidized combustors are in use throughout the world - probably
more than 2,000 - on a variety of duties including burning plastic waste,
providing hot gases for drying grass, and raising steam for process use
(Highley and Kaye, 1983).

The history of fluidization contains examples of processes which were


developed but were either never built on a commercial scale or were built and
operated for only a short time. The Shell chlorine process in which HCI is
oxidized to chlorine was systematically scaled up (de Vries et al., 1972) using a
combination of cold models, pilot plants, and theoretical equations; as far as
is known the full-scale plants were successful, but it is believed that none is
still in operation.
Use of fluid beds for ethylene oxide production and ethanol dehydration
does not appear to have progressed beyond the pilot plant stage, and
although the Mobil methanol-to-gasoline
process has recently been demonstrated to be a technological success (Penick, Lee and Mazink, 1982),
currently the economics are attractive in only a few countries.
As with all new processes, unless there is an outstanding economic or
product quality advantage there is little enthusiasm for exchanging a wellknown process for one which offers marginal improvement while incurring a
potential scale-up risk. Where the fluidized bed will form the major part of
the overall processing cost, risk-taking may be justified, but in many
processes there are so many other unit operations before and after th~ fluid
bed that its substitution makes little impact on the unit prodJir'tost.
Many
current industrial applications of fluidized beds are sho n in Fig. 1.4.,
arranged in five categories according to predominating me hanisms. Some of
these are treated in detail in two recent books (Hetsroni, 982; Yates, 1983).

For those who wish to maintain an active interest in the field it is essential to
read the general technical press, scientific journals, and proceedings of
conferences which are partly or wholly concerned with fluidization. Most of
the well-known chemical engineering journals regularly publish papers on
fluidization,
e.g. Chemical Engineering Science, Chemical Engineering

Journal, Chemical Engineering Research and Development, American


Institution of Chemical Engineers Journal, Canadian Journal of Chemical
Engineering, Journal of Chemical Engineering Japan, etc. Powder
Technology probably publishes more papers relevant to fluidization than any
other single journal and at roughly two year intervals publishes a list of recent
papers and ongoing research projects involving the technique (Geldart, 1982,
1983). However, conferences are undoubtedly the biggest single source of
information: the Engineering Foundation (New York) has organized conferences on the subject in 1975, 1978, 1980, 1983, and 1986, and published the
60 to 80 refereed papers each time in book form; the Institute of Energy
(London) has published proceedings of conferences on fluidized combustion
since 1980 (e.g. Beer, Massimilla, and Sarafim, 1975), and proceedings of
fluidized bed combustion conferences in the United States since 1970 are also

oe
c
;'c
00
.10 J:
eu
0'"
oS

available (e.g. Elliott and Virr, 1972). The American Gas Association
regularly sponsors symposia which include papers on fluidized bed gasification (e.g. Elgin and Perks, 1974). The annual meeting of the American
Institution of Chemical Engineers (AIChE) held in November each year
usually contains four sessions on fluidization and fluid/particle systems but by
no means all appear subsequently in the Chemical Engineering Progress
symposium series. International symposia on chemical reaction engineering
are held frequently and published by the American Chemical Society and
always contain papers on fluidized systems.
The abstracting service provided by the International Energy Agency
includes sections relating to fluidization, as does the HTFS (Heat Transfer
and Fluid Flow Service, Harwell, United Kingdom), Chemical Abstracts, and
Engineering Index. It is fortunate that these services are now computerized,
making searches much less time-consuming.
The active researcher is often able to keep up with the field by personal
contact with others throughout the world; those involved in fluidization and
fluid/pahicle research in the academic world form a competitive but friendly
network, and companies frequently obtain access to recent and ongoing work
through short courses and consulting contracts.

1.4

SCOPE OF THE BOOK

This book is based on a series of three-to-four-day co ses directed by the


editor since 1968 and presented in North America, Holland, and the United
Kingdom at least once a year. All the authors have lectured on these courses
on many occasions and their approach has been fashioned partly by working
together but largely, and more importantly, by noticing the reactions and
comments of attendees and responding to their needs.
All of the contributors are experts in the field either as active researchers/
consultants or as practitioners in industry. Like the lectures in the courses
from which it is derived the book is aimed at engineers in design and plant
operation, scientists/technologists in research and process development, and
postgraduate students starting work in the field. The fluidization literature is
so enormous and so many new papers and patents are published every year
(Geldart, 1983)that it is difficult for the specialist to keep up - let alone for
the novice to gain a good perspective. The aim throughout, therefore, has
been to give the reader a clear picture of the mechanisms at work, in so far as
these are known, and to cover the subject broadly yet in sufficient detail to
enable practical calculations to be made. Where possible, working eqpations
and correlations are recommended and numerical illustrations of their use are
given in most chapters.
The physical parameters which describe the particles and characterize the
fluidized powder at or near minimum fluidization conditions are dealt with in

Chapters 2 and 3. The hydrodynamic characteristics of bubbles, slugs, and gas


distributors are discussed in Chapter 4, and their influence on mass transfer
and chemical reaction between solids and gas form the subject of Chapter 11.
Solids mixing and segregation have a considerable influence on the
operation of many fluid bed processes and are discussed in Chapter 5.
Carryover of solids by the gas is an intrinsic feature of fluid beds, and the
many factors which determine its magnitude are considered in Chapter 6,
whilst beds which are designed so that all the solids pass out with the gas and
are then reinjected are dealt with in Chapter 7; effective separation of solids
from gas is obviously essential (Chapter 8). Fluidized beds are frequently
selected as processing tools because of their excellent heat transfer properties
and because they permit controlled transfer of solids into, out of, and within
the system; these features are addressed in Chapters 9 and 12 respectively.
There are probably more fluid beds used as dryers than any other single
application and their design is outlined in Chapter 10. In spite of the fact that
much more is now known about fluidized systems it is rare for a new fluid bed
process to be designed and installed without the need for cold modelling or
pilot plant work, and the experimental techniques available are found in
Chapter 13.
Fluidization research is an exciting field full of surprises; new applications
of the technique and improvements to existing processes are continually being
made, particularly in industry, and the next 20 years will undoubtedly prove
to be as interesting and fruitful as the last 20 years.

Beer J.M., MassimilIa, L., and Sarafim, AF. (1980) Inst. Energy Symp. Ser., 1 (4), 4.
de Vries, R.J., van Swaaij, W.P.M., Mantovani, c., and Heijkoop, A. (1972). Proc.
Conf. Chern. React. Eng. Amsterdam, 70, 141.
Elgin, D.C., and Perks, H.R. (1974). Proc. 5th Synth. Pipeline Gas Symp., AGA Cat.
No. L51173, p.145.
Elliott, D.E., and Virr, M.J. (1972). Proc. 3rd. Int. Coni. Fluidized Bed Combustion,
Paper 4-1, US-EPA.
Geldart, D. (1967). Chern. Ind., 1474
Geldart, D. (1968). Chern.Ind., 41
Geldart, D. (1969). Chern. Ind., 311
Geldart, D. (19~3). Powder Technol., 1,285.
Geldart, D.(1982).Powder TechnoL., 31, 1.
Geldart, D. (1983). Powder Technol., 36, 149.
Grade, J.R. (1984). 'Gas fluidization' Course, Center for Prof. Advancement, New
Jersey.
Hetsroni, G. (Ed.) (1982) Handbook of Multiphase Systems, Chap. 8.5, Hemisphere,
Washington.
Highley, J., and Kaye, W.G. (1983). Chapter 3 in Fluidized Beds - Cornbustion and
Applications (Ed. J.R. Howard), Applied Science Publishers.

Jahnig, C.E., Campbell, D.L., and Martin, H.Z. (1980). In Fluidization (Eds J.R.
Grace and J.M. Matsen), Plenum Press, New York.
Kunii, D _, and Levenspiel, O. (1969). Fluidization Engineering, John Wiley, New
York.
Penick, I.E., Lee, W., and Mazink, J. (1982). Int. Symp. on Chern. React. Engng,
ISCRE - 7, Boston.
Reh, L. (1968). Chem-Ing-Techn., 40, 509.
Squires, A.M. (1982). Proc. Joint Meeting of Chern. Ind. & Eng. Soc. of China &
Am. Inst. Chern. Engrs, Beijing, Sept. 19-22, p.322.
Yates, J.G. (1983). Fundamentals of Fluidized-bed Chemical Processes, Butterworths.
Landor!.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Single Particles, Fixed and


Quiescent Beds

The arrival time of a space probe travelling to Saturn can be predicted more
accurately than the behaviour of a fluidized bed chemical reactor! The
reasons for this frustrating situation lie not in the inadequacies of chemical
engineers or powder technologists but rather in the complexity of defining
(and measuring) unambiguously even such fundamental parameters as the
size and size distribution, shape, and density of particles. Given that these
parameters influence explicitly and implicitly the behaviour of fixed and
fluidized beds, it is hardly surprising that most correlations give predictions
no better than 25 per cent.

For a particle of any shape other than a sphere, there are many ways of
defining its size; Allen (1981) lists twelve. Only four definitions are of interest
for packed and fluidized beds:
dp

= sieve

dv

= volume

size: the width of the minimum square aperture through which


the particle will pass;
diameter: the diameter of a sphere having the same volume as

the particle;
dsv = surface/volume diameter: the diameter of a sphere having the same
external surface area/volume ratio as the particle;
11

ds = surface diameter: the diameter of a sphere having the same surface as


the particle.
Some of these diameters are related through Waddell's sphericity factor 1/1,
defined as:
surface area of equivalent volume sphere
surface area of the particle

sieving (the most common method used for powders larger than about 75
J.'m) or by using the Coulter counter (for particles smaller than about 75
J.'m). The Coulter counter gives the volume diameter, dy, so if the sphericity
is known dsy can be estimated (Eq. 2.2). Unfortunately there is nb simple
generally accepted method for measuring the sphericity of small irregular
particles, so although values have been published (Table 2.2) they should be
regarded as estimates only. The tables show that 1/1 is between 0.64 and 1 for

= (dvlds)2
Table 2.2

Sphericities of some
common solids

Because it is now well established that the most appropriate parameter for
correlating the flow of fluids through packed beds is Sy (the external surface
area of the powder per unit particle volume), the most relevant diameter is
dsy' The sphericity 1/1, dy, and dsy can be calculated exactly for geometrical
shapes such as cuboids, rings, and manufactured shapes (Table 2.1). Most
particles, however, are irregular and their size is generally measured by
Table 2.1

Shape

Sphericities for some regular solids


Relative
proportions

;>

.c:::==

1:1:2
1:2:2
1:1:4
1:4:4

0.93
0.92
0.78
0.70

Ellipsoid

1:2:4

0.79

diameter
2 x diameter
4 x diameter
If2 x diameter
V4 x diameter

0.87
0.83
0.73
0.83
0.69

Rectangular
parallelepiped

Rectangular
tetrahedron
Regular
octahedron

Height =
Height =
Height =
Height =
Height =

1:1:1
1:1:2
1:2:2
1:1:4
1:4:4
1:2:4

Crushed coal
Crushed sandstone
Round sand
Crushed glass
Mica flakes
Sillimanite
Common salt

0.75

0.8-0.9
0.92-0.98

0.65
0.28
0.75
0.84

'" = dsJdy

Spheroid

Cylinder

\fJ

0.81
0.77
0.77
0.68
0.64
0.68
0.67
0.83

mqst materials; viewing the particles through a microscope and comparison


with Tables 2.1 and 2.2 will usually enable a realistic value of 1/1 to be
estimated. A better way to find dsy (= I/Jdv) is to use the Ergun equation
(Eq. 2.17). An attempt to compare the sieve size with the volume and
surface/volume
diameters for crushed quartz has been described by
Abrahamsen and Geldart (1980), They concluded that for materials like
quartz which has a sphericity of about 0.8:
dy

::::::

1.13dp

The average sphericity for regular figures in Table 2.1 is 0.773 (s


cent). That is:

d=

(2.3)
=

11 per

0.773

Combining Eqs. (2.3) and (2.4):


dsy

::::::0.87dp

(2.5)

Note that Eq. (2.5) is an approximation for particles which are non-spherical.
For spherical or near-spherical particles:
dv

::::::dsy ::::::dp

(2.6)

2.3

MEAN SIZE AND SIZE DISTRIBUTION


20

If a powder mass M has a size range consisting of Npl spherical particles of


size dh Npz of size dz, and so on, the mean surface/volume size:

~ = lI~X/dp)
=253~m

where x is the weight fraction of particles in each size range. When sieving is
used, dJ, dz, ... are replaced by the averages of adjacent sieve apertures, dp;,
and the equation becomes:
1

= !'x/d ;
p

~.~

This definition of mean particle size gives pr~hasis


to the important
influence which small proportions of fines have. It slibuld not be confused
with another method of characterizing a powder, the media~ dpm, which is the
size corresponding to the 50 per cent value on the graph of cumulative
percentage undersize versus size. It is, however, not directly related to the
surface/volume mean.
Equation (2.8) should not be used if the powder has an unusual
distribution, e.g. bi- or trimodal, or trimodal, or has an extremely wide size
range. Such powders generally will not behave in an homogenous way and
cannot be characterized by a single number. It is always advisable to first plot
the size distribution of the powder as the weight fraction, or percentage in a
size range, against the average size, that is x versus dp;, because a plot of the
cumulative percentage undersize can conceal peculiarities of distribution. An
example of this is shown in Fig. 2.1(a). The size distribution of the powder
given in Table 2.3 is rather unusual but this does not show up in Fig. 2.1(b).
There is no entirely satisfactory way of comparing the width of the size
distribution of two powders having different mean sizes, nor of defining how
wide a distribution is. One useful way is to specify the relative spread, uld m'
This is obtained by using the cumulative percentage undersize versus the :ize
plot, e.g. Fig. 2.1(b), and defining the spread as:
.

In the example given,

(J'

= 105 Il-m and:


(J'

pm

Sieve
aperture,
Il-m
-600
-500
-420
-350
-300
-250
-210
-180
-150

+ 500
+ 420
+ 350
+ 300
+ 250
+ 210
+ 180
+ 150
+ 125

Size
dpi
Il-m
550
460
385
325
275
230
195
165
137

105

= 270 = 0.39

dp =

100

I7d
x p;

= 253 JLITI

Weight
percentagein
range Xi

Cumulativepercentage
undersize

0.5
11.6
11.25
14.45

100
99.5
87.9
76.65
62.2
41.4

20.8
13.85
12.5
11.9
3.15

27.55
15.05
3.15

The British Standard sieve is arranged in multiples of 4\12, and this is used as
a basis in Table 2.4 to give an idea of the relative spread as judged from the
number of sieves on which the middle 70 per cent. by weight of the powder is
found.

Table 2.4

Width of size distributions


spread

Number of sieves
on which the middle
70% (approx.) of the
powder is found

Type of
distribution

1
2

0.03
0.17

3
4
5

based on relative

~.

Very narrow
Narrow
Fairly narrow
Fairly wide

0.33
0.41
0.48

6
7
9
11

0.6
0.7
> 0.8

>13

2.4

Very wide

The volume includes the voids inside the particle whether they are open or
closed pores. The particle density should not be confused with the bulk
density of the bed, PB, which includes the voids between the particles; Pp is
a 'hydrodynamic' density since it is based on the shape and volume which the
flowing gas 'sees'. For nonporous solids the particle density is equal to the
true, skeletal, or absolute density of the material, PABS, which is measured
by a specific gravity bottle or air pycnometer;
but for porous solids
Pp < PABS and cannot
be measured by the usual means. A mercury
porosimeter can be used to measure the particle density of coarse porous
solids but is not reliable for fine powders since the mercury may not penetrate
the voids between small particles.
In the petroleum industry the particle or piece density of the free-flowing
cracking catalyst is estimated indirectly by measuring the pore volume. When
a liquid with low viscosity and volatility, such as water, is added, the powder
should remain free-flowing until the liquid has filled all the open microscopic
pores. Any additional liquid then coats the external surface of each particle
causing immediate caking by surface tension (the 'caking end point'). This
method is not reliable for all porous powders and a simple alternative method
has been developed (Abrahamsen and Geldart, 1980). This is based on the
fact that the minimum packed bed voidage is virtually the same for particles
of similar size and particle shape. The bed voidage is defined as:
volume of bed - volume of particles

Extremely wide

volume of bed

PARTICLE DENSITY

This is defined as (see Fig. 2.2):

Hydrodynamic envelope
bounding.particle volume ~
= Ii!

p
p

Pp

Pa

1 - -

(2.11)

Pp

Firstly, 0.2 to 0.25 kg of a control powder c, of known particle density Ppe, is


poured into a measuring cylinder and tapped until it reaches its minimum
volume, corresponding to the maximum bulk density PaTe' The procedure is
repeated with the unknown powder x. Ideally, several control powders should
be used. If:

mass of a single particle


volume the particle would displace if its surface were non-porous

Ppx

PaTx
k --Ppc
PaTe

The empirical factor k is introduced because in practice it is not always


possible to find control powders having the same particle shape as that of the
unknown powder:

(b) Particle size - the larger the particles, the lower the loosely packed
voidage (see Table 2.6); the dense packing voidage is not as dependent
on size.

k = 1 if x and c are approximately the same shape;

= 0.82 if x is rounded or spherical and c is angular;


k = 1/0.82 is x is angular and c is rounded or spherical.
k

Table 2.6

dp, ILm

If at all possible the voidage of a packed bed should be measured in the


condition in which it will be used, but if this is impossible an estimate can
often be made for group Band D solids (see Chapter 3). Obviously? the
degree to which the bed is vibrated or tapped is very important and two
extreme conditions are used as reference points: 'loose' packing gives the
maximum possible voidage and 'dense' packing the minimum; both are based
on random packings. Facto~the
voidage are:
(a) Particle shape (see Table 2.5)

Variation of packed bed voidage with particle size for narrow size
distributions (Partridge and Lyall, 1969)

2,890

551

284

207

101

89

72

55

0.386

0.385

0.390

0.411

0.424

0.434

0.441

0.454

230

140

82

72

0.563

0.590

0.602

550

460

390

330

0.422

0.432

0.440

0.437

Table 2.5

Voidage versus sphericity for randomly


packed beds unifonnly sized particles larger than
about 500 JIolll (Brown et al., 1950)

(c) Size distribution - the wider the size spread, the lower the voidage
(see Table 2.7).
Table 2.7

0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00

0.507

the lower the sphericity t/J, the higher the voidage

Loose packing

Dense packing

0.85
0.80
0.75

0.80
0.75
0.70
0.67

0.72

0.68
0.64
0.61
0.58
0.55
0.53
0.51
0.49
0.47
0.45
0.43
0.41

Effect of size distribution on


packed bed voidage

0.432
0.469

Sand 1
Sand 2

0.63

0.59
0.55
0.51
0.48
0.450.42
0.40
0.38
0.36
0.34
0.32

(d) Particle and wall roughness - the rougher the surface, the higher the
voidage.

Through the work of Darcy and Poiseuille it has been known for more than
120years that the average velocity of fluid through a packed bed, or through a
pipe, is proportional to the pressure gradient:
t::.p

Ua-

A packed bed may be thought of as a large number of smaller tortuous pipes


of varying cross-section. A number of workers developed this approach,
notably Carman (1937), Kozeny (1927), and Ergun (1952). The reasoning
behind these equations is summarized by Allen (1981). At Reynolds numbers
less than about 1 (laminar flow) the Carman-Kozeny equation applies:
Ap

KI-' (1 - e)2

=7
sv

-'

(2.15)

to the wall and the particles now start to move with an absolute velocity lis,
then the relative velocity (interstitial) between molecules and solids becomes:
U

USLIP.i

pgUdsv

= --

(2.16)

I-'

K is generally assumed to be 180 but there are indications that this may be
correct only for narrow cuts, voidages between 0.4 and 0.5, and Reynolds
numbers 0.1 to 1. Abrahamsen (1980) found that for fine powders of mean
size 30 to 80 !-,m,K had aV~263
(s 35 per cent.) for spherical
or near-spherical particles and 291 (s 26 per cent.) for other shapes. For
Reynolds numbers greater than about 1 the Ergun equation has proved
satisfacto ry:
Ap

= p~
P

/150 (1-e)2
~

I-'Ut

d;v

+ 1.75 l-e

pgtUf}

(2.17)

dsv

The term Pt/p is a correction factor for compressibility, where p is the average
absolute pressure in the bed. Subscript 1 denotes conditions at the inlet to the
bed. Under laminar flow conditions (Re < 1) the first term on the right-hand
side dominates:
Ap
!-'U
a -

d;v

Vs

where Ys is positive if the particles move in the same direction as the gas or is
negative if they move counter-current. The corresponding superficial slip
velocity is:

Where U is the superficial or empty tube velocity and the Reynolds number is
defined as:
Re

=- -

USLIP

= e ( ~ -

Ys)

Substitution of USLIP for U1 in Eq. (2.17) enables the pressure gradient to be


calculated.
2.8

MINIMUM FLUIDIZATION

VELOCITY Umf

When gas is passed upwards through a packed bed unrestrained at its upper
surface, the pressure drop increases with gas velocity according to Eq. (2.15)
or Eq. (2.17) until, on the microscopic scale, the drag on an individual
particle exceeds the force exerted by gravity or, on the macroscopic scale, the
pressure drop across the bed equals the weight of the bed per unit area. If the
bed has been compacted or is composed of interlocked, very angular or
cohesive particles, then an excess pressure is required to free them (points C
and C' in Fig. 2.3) and they adopt a higher voidage configuration causing a
Narrow range. well mixed bed
with increasing velocity

(2.18)

In fully turbulent flow (Re > 1000) the second term dominates and:
Ap

pgU2
a -dsv

(2.19)

Note that the surface/volume size dsv is used; if only sieve sizes are available,
depending on the particle shape, Eq. (2.5) or Eq. (2.6) should be used.
2.7

OAB Fixed bed region


AE

Partial fluidization for wide size range


indicating segregation

EF

Fully fluidized region

PRESSURE DROP ACROSS MOVING BEDS

Imagine gas molecules moving between particles in a packed bed. The


average relative velocity between particles and molecules is ute when the
particles are stationary. If the molecules maintain their same velocity relative

(2.21)

Figure 2.3 Pressure drop across fixed and fluidized beds for group B
and D powders.

fall back to the theoretical pressure drop. With group Band D powders (see
Chapter 3) a further increase in velocity causes the formation of small bubbles
whose size increases with gas .velocity. The bed pressure drop begins to
fluctuate and if the bed is deep enough (H > 2D) the bubbles occupy a
substantial proportion of the cross-sectional area. These large bubbles are
called slugs and cause regular piston-like movements of the upper surface of
the bed. The average bed pressure drop then increases as shown in Fig.
2.4(b). If the gas velocity is now reduced, depending on the size distribution
of the powder, the pressure drop declines along curves 3,4, or 5 in Fig. 2.3. If
the powder has a narrow size range, curve 3 is followed; increasing the size
distribution (but maintaining the same mean size) results in curves 4 or 5
because the larger particles settle out progressively on the distributor. Points
D and E represent the minimum velocity required to fully support the solids
(though not necessarily in a well-mixed state) and are called the minimum
velocity of complete fluidization, Ucf' Umf is usually defined as the
intersection of the horizontal fluidized bed line EDF and the sloping packed
bed line OAB, but it is cl~ids
with a very wide size range these
could be drawn almost anywhere; the velocity corresponding to point E is of
practical interest but relatively little work has been done to predict it (see Sec.
2.11). For the present the best recourse is to make measurements and visual
observations.
The pressure drop across a fluidized bed is the only parameter which can be
accurately predicted:
Mg

tiPF
or tiPF

=-

cm water gauge (w.g.) =

tiPF
H

A
O.IM
A

N/m2

(2.22)

kg
m2

(2.23)

=(pp - pg) (1 -

Emf)g

(2.24)

Emf is the bed voidage at Umf and a close approximation to it can be obtained
by measuring the aerated or most packed bulk density PeLP' This is done by
pouring the powder through a vibrating sieve and allowing it to fall a fixed
height into a cylindrical cup of capacity 100 em3. The equipment is shown
later in Fig. 3.6 and for consistent results the powder should be poured
through in 20 to 30 s.
If Eqs. (2.24) and (2.17) are combined (with Pl/P = 1, AplH eliminated,
and U and E set equal to Umf and Emf, then:

~~\=\
~

Ij

::J

The group on the left-hand side is dimensionless and known as the


Archimedes number or, by some workers, the Galileo number. Wen and Yu
(1966) correlated many data in this form. They combined the numerical
constants with the voidage terms and, using the volume diameter dv instead of
dsv proposed:

Remf

PgUrnpiv
p.

Recently there have been several studies on the independent effects of


temperature and pressure on Umf In general the qualitative effects are as
predicted by Eqs. (2.27) and (2.28): in fine powders Urnf decr~ases with
increasing temperature and is hardly affected by pressure (e.g. King and
Harrison, 1982) whilst in coarse powders increased temperature causes an
increase in Umf and pressu~t
should be noted that some strange
effects can occur in beds of coarse particles of high temperatures (Botterill
and Teoman, 1980). Also, in any powder, if softening or sintering occurs
producing agglomeration, none of the equations is valid. Re-arranged, Eq.
(2.26) becomes:
Urnf

p.
= pplv

{ (1,135.7

'h}

+ 0.0408Ar) - 33.7

pgd~(pp -' pg)g


p.2

Equation (2.27) should be used for particles larger than 100 p.m (i.e. group B
and D solids) in conjunction with Eqs (2.3), (2.6) and (2.8). For particles
smaller than 100p.m, Baeyens' equation gives the best agreement with
experiments (Geldart and Abrahamsen, 1981):

small bubbles appear at the surface. These must not be confused with the
continuous channels or spouts which often appear. They resemble miniature
volcanoes which disappear when the side of the column is gently tapped.
Further increases in velocity produce, at first, a further slight increase in bed
height (see Fig. 2.6) followed by a reduction. Relatively large bubbles burst
through the bed surface periodically, causing the bed to collapse rapidly; it
then 'reinflates' slowly to collapse again as another swarm of bubbles bursts
through. Reduction of the gas velocity produces a retracing of the heightvelocity graph and finally the last bubble disappears giving once again a
quiescent bed. The average of the velocities at which the bubbles appear and
disappear is called the minimum bubbling velocity or bubble point and
generally coincides with the maximum bed height for deep beds; premature
bubbling can be caused by non-uniform distributors or protuberances in the
bed, but the maximum values of Umb have been correlated by Abrahamsen
and Geldart (1980) and found to depend on the gas and particle properties:
Umb = 2.07 exp (0.716F)

d pO.06
Po ;47
p.'

where F is the mass fraction of the powder less than 45 p.m. The numerical
constant is dimensional and SI units must be used. If F = 0.1 and the powder
is fluidized by air at ambient conditions:
Umb

= 100 d

MINIMUM BUBBLING VELOCITY Umb

Fine powders (cracking catalyst is a typical example) exhibit a type of


behaviour not found in coarse solids, namely the ability to be fluidized at
velocities beyond Umf without the formation of bubbles. The bed expands,
apparently smoothly and homogenously, until a velocity is reached at which

(2.30)

Note that Eq. (2.29) implies that Umh is not dependent on particle density, a
finding confirmed independently by Simone and Harriott (1980). It must be
stressed that Eqs (2.29) and (2.30) are valid only for powders which are fine
enough (in the main, less than 100 p.m mean size) to have values of Umf less
than Umbo If the calculations show that Urnf (from Eq. 2.28) is larger than
Umb, then the powder will start to bubble at or very slightly above incipient
fluidization and Eq. (2.29) should not be used. This is discussed further in
Chapter 3.

2.10 EFFECT ON Umb OF TEMPERATURE,


AND TYPE OF GAS

2.9

(2.29)

PRESSURE,

This is still a relatively unknown area. Equation (2.29) indicates that Umb
increases with pressure: King and Harrison's (1982) data show agreement
with the power on density of 0.06; Godard and Richardson (1966) indicate
0.1; Guedes de Carvalho (1981) an even higher dependency.
Piepers et at. (1984) give data which show that the power depends on the
type of gas. Up to p = 15 bar, Urnb for a catalyst in H2 is affected little, in N2 as
p~.13, in argon as p~.18. The reason is not clear but adsorption of the gas in

the surface of the catalyst may have played a part and it may be that not all
fine powders behave in the same way.
Omitting any sintering effects, increasing temperature
reduces Umb
according to 1/JLO347.

In materials having a wide size range, particularly those having a large mean
size, segregation by size tends to occur at velocities close to Umf for the
mixture. Although a value of Umf can be defined for the mixture based on dp
its usefulness is limited. Knowlton (1974) defined a velocity, Uef, at which all
the particles are fully supported (even though there may be segregation) and
suggested that:

1968, and Crowther and Whitehead, 1978). Recently Wong (1983) has
shown that both nand v;/vt increases with decreasing particle size below
about 60 JLm;
is the intercept of the logE-log U plot at e = 1 and Vt is
the Stokes terminal velocity.
An alternative approach based on the Carmen-Kozeny equation was
used by Abrahamsen and Geldart (1980). Combining Eqs (2.15) and
(2.24) (with Emf = E and dp = dsv) gives:

v;

(p

-----p

As we shall see in Chapter 4, almost all gas fluidized beds operate in the
bubbling regime and consist of two phases - bubbles and the emulsion (or
dense) phase. Conditions in the dense phase (gas velocity Uo and the
corresponding
voidage ED) are the subject of much discussion and
speculation, largely because it is believed that (a) most of the chemical
conversion occurs there and (b) in fine powders the equilibrium size of the
bubbles may be controlled by the voidage, ED. The dense phase also figures
in modes of flow in standpipes or downcomers.
Uo and ED are extremely difficult to measure directly in bubbling beds
though they can be predicted for quiescent fluidization of fine powders.
(a) Voidage in non-bubbling (quiescent) beds. The earliest attempts to
measure and correlate the expansion of quiescent (or particulate)
fluidized beds in gas/solid systems were made by Davies and Richardson
(1966) who adapted the approach used by Richardson for liquid/solid
sedimentation and fluidization:
U

-=~
v[
In liquid fluidization n is a function of dpiD and the terminal velocity
Reynolds number, and varies between 4.65 and 2.4. The form of the
relationship appears to hold for gas/solid systems but experimental values
of n between 3.84 and 19.7 have been reported (Godard and Richardson,

Pg)g d~
JL

--

1-E

KfU

(2.33)

Plots of all their data using 48 gas/solid systems gave a general expression
to predict the non-bubbling expansion of a bed of fine powder, namely:
~
(pp - pg)g d~
1- E
JL

where Umfi and Xi refer to fraction of size dpi' This appears to give reasonable
agreement with experim~en
at high pressures.

= 210 (U _ Umf) +

~mf

1 - Emf

(pp - Pg)g d~
/k

where Umf < U < Umbo


The standard deviation of the numerical constant is 22 per cent; and
Emfcan be found from the simple experiment to determine the aerated
voidage described earlier; Umf is obtained from Eq. (2.28) and Umb from
Eq; (2.29).
(b) Voidage in bubbling beds. The average overall voidage of a bubbling
.bed, i.e the bed expansion, is largely caused by the bubble hold-up, and is
dealt with in Chapter 4. The voidage of the dense phase (the portion of
the powder where bubbles are absent) is generally taken as being Emffor
group Band D systems, but can be significantly higher for group A
powders.
Two methods have been used to measure dense phase voidage in
bubbling fluidized beds. Rowe et a1. (1979) used X-rays and concluded
that the dense phase voidage ED was so high that it could carry gas
velocities up to 20Umf More recent studies have given values in line with
those obtained from the bed collapse technique. Rietema (1967).
Abrahamsen and Geldart (1980), and Simone and Harriott (1980), using
the collapse technique and certain assumptions, concluded that Emf< ED
< En\band hence Umf < Uo < Umbo
In collapse experiments, the powder is fluidized at a chosen velocity
(say 20 em/s) and the gas suddenly shut off. The bed height is recorded as
a function of time (Fig. 2.5); it falls rapidly in the first few seconds as the
bubbles rise to the surface and then much more slowly as the interstitial
gas flows out and the dense phase collapses. The straight line portion of
the curve is extrapolated back to time zero and it is assumed that the
intercept gives the height Ho which the dense phase would occupy in the

Gas velocity
before cut off=O.027mA;
-

Bubble
escape stage

mb

65

..

E
~

Hindered
...,...- sedimentation
stage

0.64

J:

"t__I J J

60

0.62

H-

1r r

:t-

t__~r

bed height

Dense phase height Ho

55

mf

Solid
consolidation
stage \

50

l{,.f

30

2.54

Hmf

Uo

Umf

~.1

188

dg.568

ILO.066

p~.016

gO.ll8

(p

p~.089
gO.663

Umb

40

iLm

bubbling bed. This height is plotted at various superficial gas velocities on


Fig. 2.6., By further assuming that Darcy's law holds within the dense
phase in bubbling bed, Uo can be calculated. Abrahamsen and Geldart
(1980) give:

Ho

U(cm/s)

(s)

Figure 2.5 Typical collapse curve for group A powder (26


ballotini).

--=

I I

exp (0.09F)

pg)O.118

H?n?43

ILO.371

exp (0.508F)
(pp _ pg)O.663 hO.244

The constants are dimensional and SI units must be used. It should be


noted that these empirical equations indicate that the dense phase 'opens
up' as the mean particle size and particle density decrease and as gas
density (pressure) and viscosity (temperature) increase. Increasing the
fines fraction F also increases the dense phase voidage and gas velocity
relative to conditions at incipient fluidization.

A bed of angular sand of mean sieve size 778 ILm. is fluidized by air. The
particle density is 2,540 kglm3; lL(air) = 18.4 x 10-6 kg/m s; Pg = 1.2 kglm3;
and 24.75 kg of the sand are charged to a bed 0.216 m in diameter. The bed
height at incipient fluidization is 0.447 m. Find:
(a) Emf,
(b) the pressure drop across the bubbling bed,
(c) the incipient fluidization velocity.

24.75
(1T/4)(0.216)2

Emf

1-

_
3
0.447 - 1,511 kglm

1,511
2540

0.405

SINGLE PARTICLES, FlXED AND QUIESCENT BEDS

H
HD
Hmf
M

Step 3. From Eq. (2.23):


O.lM

2.475

----:4 = (1T/4)(0.216)2

!:J.PF =

Answer (b) !:J.PF = 67.5 cm w.g.

Step 4. For angular sand, from Eq. (2.3):


dv

= 1.13 x 778
= 879 x 10-6

Step 5. Archimedes'
Ar

10-6

!:J.p

!:J.PF

Re

number:

Remf

1.2 x 2,540 x 9.81 x (879 x

= -------------(18.4
=

fJ

10-6)3

10-6)2

Sp

59,981

Sy, SB

Step 6. From Eq. (2.26):


.--:-------.
Rem! = (1,135.7 + (0.0408 x 59,981'h - 33.7

V,Vmf

26.15

Vcf

26.15 x 18.4 x
1.2

879

VD

10-6
10-6

Vmb
VB
Vp
VI'

2.14
A
Ar

dpi
dp
dsv
dy
D
F

cross-sectional

NOMENCLATURE
area of bed (= (1T/4) D2)

Archimedes number Pg (pp - pi gdV J-L2


arithmetic mean of adjacent sieve apertures
mean sieve size of a powder
diameter of a sphere having
the same
,
surface/volume ratio as the particle
diameter of a sphere having the same
volume as the particle
diameter of bed
mass fraction of powder less than 45 J-Lm
acceleration

due to gravity

vi

Xi
2

E,Emf,Emb

J-Lmorm

J-L

J-Lmorm

PABS
PB

absolute
pressure
pressure
Reynolds

mean pressure in the bed (= P


drop across a packed bed
drop across a fluidized bed
number

+ !:J.p/2)

Reynolds number at incipient fluidization


velocity (= pgVmtdy/p)
standard deviation
surface area of one particle

Pp

N/m2
N/m2

m~0

m/s

superficial incipient bubbling velocity


volume of bed
volume of one particle

m/s
m3
m3

particle terminal velocity, intercept at E = 1


on log E - log V plot
weight fraction of powder of size dp;
voidages of packed, incipiently fluidized, and
incipiently bubbling beds
fluid viscosity
absolute density of material comprising a powder
bulk density

sphericity of particle

Pg

kg
N/m2
N/m2

surface area of powder per unit volume of


powder and per unit volume of bed, respectively
superficial gas velocity at bed exit and at
incipient fluidization velocity, respectively
velocity at which all particles are
fully supported (Eq. 2.31)
superficial velocity of gas in dense phase in a
bubbling bed

l/J

J-Lm or m

9.81 m/s

mass of powder in a bed


absolute pressure at bed exit

PBT

m
m
m

height of dense phase in a bubbling bed


height of incipient fluidized bed

aerated or most loosely packed bulk density


tapped or maximum bulk density
gas density
particle density
spread of particle size (Eq. 2.9)

PBLP

J-Lmorm

height of packed bed

31

m2/m3
m/s
m/s

kg/m s
kg/m3
kg/m3
kg/m3
kg/m3
kg/m3
kg/m3
J-Lm or m

Abrahamsen, A.R. (1980). M.Sc. Dissertation, University of Bradford.


Abrahamsen, A.R., and Geldart, D. (1980). Powder Technol., 26, 35.
Allen, T. (1981). Particle Size Measurement, 3rd ed., Chapman and Hall, London.
Botterill, 1.S.M., and Teoman, Y. (1980). In Fluidization (Eds 1.R. Grace and 1.M.
Matsen) Plenum Press, p. 93).
Brown, G.G., etal. (1950). Unit Operations, lohn Wiley and Company, New York, p.
214.
Carman P.c. (1937). Trans. Instn. Chem. Engrs (Lond.), 15, 150.
Crowther, M.E., and Whitehead, J.e. (1978). In Fluidization (Eds. J.F. Davidson and
D.L. Keaims), Cambridge University Press, p.65.
Davies, L., and Richardson, 1.F. (1966). Trans. Instn Chem. Engrs, 44, 293.
Ergun S. (1952). Chem. Engng Prog., 48, 89.
Geldart, D., and Abrahamsen, A.R. (1981). Chem. Eng. Prog. Symp. Ser. 77 (205),
160.
Godard, K., and Richardson, J.F. (1968). In Fluidization, Instn. Chern. Engrs Symp.
Ser., p. 126.
,
Guedes de Carvalho, J.R.F. (1981). Chem. Eng. Sci., 36, 1349.
King, D.F., and Harrison, D.L. (1982). Trans. Instn Chem. Engrs, 60, 26.
Knowlton, T.M. (1974). Paper 9b, 67th Ann. meeting of A.I.Ch.E., Washington,
D.C., Dec. 1-5.
_______.
Kozeny, J_(1927). S.B. Akad. Wiss. Wien. Abt. Ila, 136,271.
Partridge, B.A., and Lyall, E. (1969). AERE Rep. M.2152.
Piepers, H.W., Cotaar, E.J.E., Verkooijen, A.H.M., and Rietema, K. (1984). EFCE
Conf. on Role of Particle Interaction in Powder Mechanics, Powder Technol. 37,
55.
Rietema, K.(1967). In Proc. Int. Conf. on Fluidization (Ed. A.A.H. Drinkenburg),
Neth. Univ. Press. Amsterdam, p.154.
Rowe, P.N., MacGillivray, H.J., and Cheesman, D.J. (1979). Trans. Instn. Chem.
Eng, 57, 194.
Simone, S., and Harriott, P. (1980). Powder Technol. 26, 161.
Wen, c.Y., and Yu, Y.H. (1966). A.I.Ch.E.J., 12,610.
Wong, A.C.Y. (1983). Ph.D. Thesis, University of Bradford.

Gas Fluidization Technology


Edited by. D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Characterization of Fluidized Powders

Much of the early experimental work on fluidization related to cracking


catalyst, a finely divided, porous, low density powder having excellent
fluidization properties, Based on the practical experience gained from
operating full-scale catalytic crackers and experimental units for other organic
processes, practitioners of the art formulated rules of thumb concerning the
conditions needed for the successful operation of fluidized beds. As the
technique was applied to processes far removed from oil cracking, different
solids had to be used, but there was a tendency to assume that published
conclusions drawn from using cracking catalyst were also applicable to these
other powders which often had quite different particle sizes and densities.
Experience with catalyst suggested that a powder with a wide size
distribution fluidized more satisfactorily than a powder having a narrow size
range. The term 'more satisfactorily' is used loosely, but workers in the field
understand it to imply smaller wind box pressure fluctuations, less vibration
of the bed, and less tendency to slug, and this is attributed to smaller bubble
sizes promoted by the wide size distrIbution.
However, my own experimental work on bubble sizes in sands having mean
sizes 75 to 470 j.tm (Geldart, 1972) showed no effect due to size distribution,
and it was the study of these and data from the literature which eventually led
to the idea of the powder groups which form the subject of this chapter.
Various attempts (Wilhelm and Kwauk, 1948; Jackson, 1963; Molerus,
1967; Simpson and Rodger, 1961; Verloop and Heertjes, 1970; Oltrogge,
1972) have been made to formulate criteria which can predict whether a
fluid/solid system would fluidize in a 'particulate' or 'aggregative' (bubbling)
manner, and a more detailed discussion is given by Geldart (1973). Most
33

liquid fludized systems are 'particulate' in the sense that as the superficial
velocity is increased the particles move further apart in a more or less uniform
way until they are carried out of the tube; most gas fluidized systems are
'aggregative' in that when gas additional to that required for minimum
fludization is supplied, it passes through the bed as bubbles, leaving the dense
phase at much the same voidage as it is at the minimum fluidization velocity
Urnf However, some liquid systems, like lead shot/water, can behave in a
bubbling mode, and some gas systems, e.g. a fine catalyst in a high pressure
gas, in a 'particulate' or non-bubbling mode; some of the criteria predict
correctly that these extremes of behaviour should occur, but they are not
capable of distinguishing the other differences of behaviour found in gas/solid
systems which are discussed below.

Before discussing the numerical criteria used to discriminate between the


various groups (see Sect~tive
descriptions of their fluidization
behaviour are given. The groups are dealt with in order of increasing particle
size.
3.2.1

Group C

Powders which are in any way cohesive belong in this category. 'Normal'
fluidization of such powders is extremely difficult; the powder lifts as a plug in
small diameter tubes, or channels badly, i.e. the gas passes up interconnected
vertical and inclined cracks extending from the distributor to the bed surface
(Fig. 3.1b). The difficulty arises because the interparticle forces are greater
than those which the fluid can exert on the particle, and these are the result of
very small particle size, generally < 20 J.lom,
strong electrostatic charges, wet,
sticky particle surfaces, soft solids, or particles having a very irregular shape.
The pressure drop across the bed is, on the whole, lower than the theoretical
value (bed weight per unit cross-sectional area) and can be as little as half.
Particle mixing and, consequently, heat transfer between a surface and the
bed is much poorer than with powders of groups A or B. The hardness ofthe
particles has a strong influence, soft materials being more cohesive since they
deform readily and give a larger area for interparticle contacts.
Fluidization can usually be made possible or improved by the use of
mechanical stirrers or vibrators which break up the stable channels, or, in the
case of some powders, usually plastic materials, by the addition of a fumed
silica of sub-micrometre size. Porous particles tend to be non-conductive and
agglomeration may occur due to excessive electrostatic charging. Improvement can generally be effected by humidifying the incoming gas or by making
the equipment wall conducting, e.g. by coating glass with a very thin layer of

tin oxide. However, with non-porous particles humidifying the gas beyond 65
per cent. can increase cohesiveness due to deposition of a liquid film.
If powders in this group do become fluidized or aerated, e.g. by being
transported pneumatically or on an air slide, they may remain aerated for
many minutes after transfer into storage hoppers.

3.2.2

Group A

A considerable amount of research has been devoted to aeratable group A


powders, largely because most commercial fluidized bed catalytic reactors use
them' our understanding of the structure of these powders has been advanced
significantly by the research of Rietema and coworkers (e.g. Rietema, 198~)
and Donsi, Moser, and Massimilla (1975) and their coworkers. Their
extensive researches have shown convincingly that interparticle forces are
present even in powders such as cracking catalyst which exemplifies .t~is
group. Beds of powders in this group expand considerably at veloclt~es
between Umfand the velocity at which bubbling commences, Umb(se~ Fl~.
3.1a), because such powders are slightly cohesive. As the gas velocity IS
increased above Umbthe passage of each bubble disrupts the weak metastable
structure of the expanded dense phase; the bed height becomes smaller
because the dense phase voidage is reduced more quickly with increasing gas
velocity than the bubble hold-up increases. The dense phase eventually
assumes a stable voida~f
an~ Ernb'~nd increasin~ the gas
velocity above about 6 cm/s produces a net Increase In bed expansIOn.
When the gas supply is suddenly cut off, the bed collapses slowly at a rate,
Uc, comparable to the superficial velocity of the ~as in the dense phas~ of the
bubbling bed (0.1 to 0.6 cm/s) (See Fig. 3.2). This makes group A sohds easy
to circulate around fluidized and pneumatic conveying loops; however, the

ease with which they become aerated also makes them liable to flood on
discharge from hoppers (Geldart and Williams, 1985).
Gross circulation of the powder (akin to convection currents in liquids)
occurs even when few bubbles are present, producing rapid mixing and
making the surface resemble a boiling liquid. Bubbles in a two-dimensional
bed appear to split and recoalesce very frequently, resulting in a restricted
bubble size. The bubble size is affected by the mean particle size, the mass
fraction less than 45 JLm, pressure, and temperature. Considerable backmixing of gas in the dense phase occurs, and gas exchange between bubbles
and the dense phase is high, probably due to splitting and recoalescence.
All bubbles rise more rapidly than the interstitial gas velocity, but in freely
bubbling beds the velocity of small bubbles 4 cm) appears to be about 30
to 40 cm/s regardless of bubble size, suggesting that it is the gross circulation
currents which control the rise velocity.
When the superficial gas velocity is sufficiently high and the bed diameter
sufficiently small to cause the formation of slugging conditions, the slugs
produced are axisymmetric; at superficial gas velocities between 0.5 and 1.5
m/s bubble and slug flow breaks down as entrainment becomes enormous and
a transition to transport occurs. Further' increase in velocity and the
recirculation of elutriated solids results in fast fluidization (see Chapter 7).
3.2.3 Group B
Sand typifies powders in this group, which contains most solids in the mean
size'and density ranges:
60 JLm < dp < 500 JLm when

Pp

< dp < 100 JLm when

Pp

250 JLm

l---'-

Time

= 4 g1cm3
= 1 g1cm3

In contrast with group A powders, interparticle forces are negligible and


bubbles start to form in this type of powder at or only slightly above minimum
fluidization velocity. Bed expansion is small and the bed collapses very
rapidly when the gas supply is cut off (see Fig. 3.2).
There is little or no powder circulation in the absence of bubbles and
bubbles burst at the surface of the bed as discrete entities. Most bubbles rise
more quickly than the interstitial gas velocity and bubble size increases with
both bed height and excess gas velocity (U - Urnf); coalescence is the
predominant phenomenon and there is no evidence of a maximum bubble
size. When comparisons are made at equal values of bed height and U - Urnh
bubble sizes are independent of both mean particle size and size distribution.
Backmixing of dense phase gas is relatively low, as is gas exchange between
bubbles and dense phase. When the gas velocity is so high that slugging
commences, the slugs are initially axisymmetric, but with a further increase in

gas velocity an increasing proportion becomes asymmetric, moving up the


bed wall with an enhanced velocity rather than up the tube axis. Transition to
transport occurs between 1 and 3 mfs.
3.2.4 GroupD
Large and/or dense particles belong to this group. All but the largest bubbles
rise more slowly than the interstitial fluidizing gas, so that gas flows into the
base of the bubble and out of the top, providing a mode of gas exchange and
by-passing which is different from that observed with group A or B powders
(see fig. 3.7). The gas velocity in the dense phase is high and solids mixing is
relatively poor; consequently, backmixing of the dense phase gas is small.
Segregation by size is likely when the size distribution is broad, even at high
gas velocities. The flow regime around particles in this group may be
turbulent, that is Pg Umf dpl/L > 1,000, ~ausing some particle attrition with
rapid elutriation of the fines produced. Relatively sticky materials can be
fluidized since the h~momentum
and fewer particle-particle
contacts minimize agglomeration.
Horizontal voids appear close to the distributor: in a narrow column these
may extend across the tube giving bridging and piston-like slugs of solids
which slide up the tube and collapse into wall slugs near the surface; in a large
column these voids often split into bubbles which appear to grow without
coalescence, as through draining gas from the surrounding dense phase. Slug
breakdown and the onset of turbulent fluidization occur above about 3 mfs.
Bubble sizes are similar to those in group B powders at equal values of bed
height and U - Umf If gas is admitted only though a centrally positioned
hole, group D powders can be made to spout even when the bed depth is
appreciable.
A qualitative summary of these properties is given in Table 3.1.

Most obvious

characteristic

"- Typical
Property "solids

1. Bed expansion

Cohesive, difficult
to fluidize

Bubble~fTee range of
fluidization

Starts bubbling

at

Umf

Crushed limestone
coffee beans

FlOUT,

cement
Low when bed
channels. can be
high when fluidized
Initially fast.
exponential

No bubbles.
Channels. and

cracks

Splitting!
recoalcscence
predominate;

maximum size exists;


large wake

4. Solidsmixint

Very low

Moderate

5.

Gas backmixing"

Very low

Moderate

6.

Slug properties

Low
Low

Horizontal voids,
solids slugs.
wall slugs

No. except in very


sballow
Effect on

properties
1 to 7'of:

Mean particle
size within
group

beds

Cohesiveness
increascs as
dp decreases

Particle

Increasing <45 #MJI


fraction improves
properties

size
distributionh

Increasing
pressure,

temperature,
viscosity.

Probably
improves

density of
gas

At equal U- Um,.
b At equal dp

3.3.1

The group AIB Boundo.ry

The most easily observed difference between powders in groups A and B is


whether or not the bed bubbles at or very close to minimum fluidization. If
there is appreciable bed expansion before bubbling commences, then the
powder belongs to group A and is likely to have the other properties
associated with that group.
The equations for the incipient fluidization and bubbling velocities, Eqs
(2.28) and (2.29), are plotted on Fig. 3.3 for particles of two densities
fluidized by air at ambient conditions. The < 45 /Lm fines fraction F is
assumed to be 0.1.

Consider particles of density difference 1,000 kglm3 and mean particle size
100 /Lill.Entering the graph from the left, the minimum fluidization velocity
line is encountered first with a Umf of 0.4 cmls and ,then the minimum
bubbling line at 1 cm/s, giving a value of Umb/Umf =2.5, indicating a bubblefree region and group A powder. By contrast, 100 /Lmalumina (pp = 4,000
kglm3) has a theoretical Umb of 1 cm/s and a Umf of 1.5 cmfs. If will therefore
bubble at minimum fluidization and fall into group B.
Thus for a powder to belong to group A or C:
U
> 1
Umf

(3.1)

Substituting from Eqs (2.28) and (2.29), a powder will be in groups A or C if:
2,300

p~.126

--~-----~.8

fJ-0.523

gO.934

exp (0.716F)

(pp _

pg)O.934

>1

(3.2)

C
(Cohesive)

100

500

dp

1000

mXI0

Figure 3.4 Diagram for classifying powders into groups having broadly similar fluidization
characteristics in air at atmospheric temperature and pressure (after Geldart, 1973)

Figure 3.3 Minimum fluidization velocity and minimum bubbling


velocity for air versus particle size.

<>

He
Ar
<>- Air
Freon 12
2 x C02

If the physical properties of air at ambient conditions are substituted into Eq.
(3.2), Fis put equal to 0.1 and Pg is assumed to be negligible compared with
Pp, then:
pg.934 ~.8

<1

(3.3)

Equation (3.3) is shown in Fig. 3.4 as line XX. Powders with size/density
combinations to the left of the line belong in groups A or C when fluidized by
air at ambient conditions.
The ratio of the incipient bubbling and fluidization velocities - the
velocity ratio - is a useful guide in characterising the fluidization behaviour
of fine powders. From Eq. (3.2) we see that UmJUmf increases as particle size
and density decrease and from Fig. 3.5 that the bed expansion ratio in
bubbling and non-bubbling beds increases with an increase in the velocity
ratio. However, this trend does not continue indefinitely for if the powder

~IE
I-Ie

II

1.5

II

3
Umb
Umf

becomes too fine (typically, less than about 20 ~m) cohesivity assumes
greater importance and the 'quality of fluidization' rapidly deteriorates.
If the physical properties of the powder are maintained constant and no
other effects such as particle sintering come into play, Eq. (3.2) predicts that
UmJUmf
(and hence the fluidization properties) should improve if the
temperature andlor pressure increase, or if the gas composition changes in
such a way that ~O.523 p~.126 increases. Conversely, if the gas is changed from
air to hydrogen a ambient conditions the powder may change from smooth
group A behaviour
group B with its large bubbles or slugs. The reasons for
this are not clear but ay involve gas adsorption on to particles.
The physical proper ies of the gas usually remain constant in a fluidized
process but those of th powder may not. If fines are produced by attrition
and are not lost from the system, F will increase and dp decrease. This will
increase UmJUmf and at first improve the fluidization, but if it continues,
eventually it may change the powder into the excessively cohesive group C,
and result in poor fluidization; this kind of behaviour has been observed
industrially as well as in the laboratory.
Conversely, loss of fines through inefficient cyclones would reduce F and
increase dp; UmblUmf could decline to a value near or less than 1 and move the
powder into group B, causing large bubbles or sluggingand loss of conversion
in a chemical reactor. This has also been observed experimentally.
3.3.2

Group Ale Boundary

Molerus (1982) has attempted to distinguish between groups A and C by


considering the balance between hydrodynamic and cohesive forces. He gives
the criterion that a powder belongs to group C if:

where FH is the adhesion force transmitted in a single contact between two


adjacent particles. Its magnitude depends upon the surface geometry of the
contacting particles, the hardness of the particulate material, and on the type
of forces (van der Waals, capillary, gas adsorption, etc.). Rietema (1984) has
also addressed the problem and produced a qualitative relationship involving
FH, dp, Pp, and ~.
Wong (1983), in a parallel approach, gives the criterion that for a group C
powder:
p~.Q6 d~ ~O.653

FH

At present it is not possible to resolve which of these equations is correct.


Both predict that as FH increases, e.g. by the solids becoming softer, group C
will include larger and larger particles. The area PQ (Fig. 3.4) drawn in by
Geldart (1973) on the basis of fludization properties reported in the literature
implicitly recognizes this, since low density particles (e. g plastics) tend to be
soft and high density materials (e.g. metals) hard. A transition category
designated group AC by Dry, Shingles, and Judd (1983) having properties of
the two others has recently been proposed.
On a strictly empirical basis, a simple test procedure can be used to
ascertain to which group any particular powder belongs. Geldart, Hamby, and
Wong (1984) have shown that standardized measurements of bulk densities
can be used. The bulk densities of the powders are determined by a
standardized technique in the apparatus which is shown in Fig. 3.6. These
features are available on the Nauta-Hosakawa powder tester. Particular care
is taken to ensure that the centre of the stationary chute is in alignment with
the centre of the pre-weighed 100 cm3 cup. The most loosely packed bulk
density, PsLP, is obtained by pouring the powder through a vibrating sieve
and allowing it to fall a fixed height (about 25 cm) into a cylindricalcup. The
amplitude of the vibration is set so that the powder will fill the cup in 20 to 30
s. The excess powder is removed fromthe cup by drawing the flat edge of a
ruler over the top. Care should be taken to ensure the cup is not jarred as this
causes the powder to compact and gives a false reading. The powder and cup
are weighed and an extension piece is then fixed to the cup. The cup is tapped
480 times (with a fall from a height of about 3 cm) whilst extra powder is
added to ensure the powder level does not drop below the surface of the cup.
When tapping ceases the extension is carefully removed, the excess powder
scraped off, and the powder and the cup reweighed. The tapped bulk density,
Psi", can then be calculated.
Strong interparticle forces prevent the particles from rolling over each
other when poured into the cup, giving an open powder structure and low
values of PsLP for cohesive powders. However, this open structure can easily
be disrupted when work is done on the system, so that the powder can be
compacted to produce a closer matrix under mechanical vibration. This
phenomenon gives a high ratio PsT to PsLP in cohesive solids and a low ratio
in free-flowing powders. Thus the ratio PsLP is a reflection of the cohesivity
of a powder. PsT/PsLP is called the Hausner ratio, HR (Grey and Beddow,
1968/69). If the ratio HR > 1.4 the powder is in group C; if 1.25 < HR < 1.4
the powder is in the transition AC group; and if HR < 1.25 the powder is in
group A, B, or D.
3.3.3

Group BID Boundary

There are two criteria which can be used to determine the BID boundary. The
first is based on the mode of gas by-passing, discussed in Chapter 11. In small

particle systems all but the smallest bubbles travel much faster than the
interstitial gas, and gas tends to stay in the bubble' except during coalescence
and splitting (see Fig. 3.7a).

Gas motion
relative to
bubble

U<-b
Emf

Q\

.!;;

~
~o.
Q:J
~
~u
.. 0
a
"0
.0
c:

5Q.l

~.~

00.
0.

c:

d:

en

"

',.
I

.'

...L:..::

-,'

'-'.:: ::- . .,"

.- - ..... ~...
'

eo

::::::

i:.

'

f....
:-...
-0

"

"

. '."

Emf

Umf

"'E

Umf
>--

'".

__

~~.

'"'"

In large particle systems, all but the largest bubbles travel slower than the
interstitial gas and gas short-circuits the bed by entering the bottom of a
bubble and leaving the top. It is clear that when considering groups Band D
we are really looking at a continuum because the form of by-passing which
occurs depends not only on Umf but also on the velocities of bubbles and
hence their sizes. The bubble size depends on the excess gas velocity U-Umf
and the height above the distributor; so within a large, deep fluidized bed
near the distributor there would be small bubbles travelling more slowlythan
Urn/Emf and higher up the bed, large bubbles travelling more quickly.
In spite of this complexity the situation can be simplified considerably by
choosing a bubble size and calculating the density/particle size combinations
of powders in which bubbles smaller than this would rise faster than Umf/Emfl
i.e. in the throughflow mode (Fig. 3.7b).

the particles increase, and as the liquid viscosity decreases (Jackson, 1971).
All these parameters are in the correct qualitative relationship to one another
in Eqs (3.1) and (3.2), which were developed for gas fluidized systems.
Replacing subscript g by L to denote liquid fluidization,
For large particles, 0.5 > Emf> 0.4, say 0.45. Bubbles in commercial
fluidized beds usually have sizes in the range 0.1 to 0.25 m due to limitations
imposed bY~beddepth, gas velocity, or internal tubes. For bubbles smaller
than 0.25 m, Ub < 1.1 mls and a powder will be in group D if:
Umf> 0.5 m/s (approx.)

~or bubbles saller

than 0.1 m,

Ub

(3.7)

< 0.7 m/s and a powder will be in group D

If:
Umf> 0.3 mls (approx.)

(3.8)

The minimum fluidization velocity Umf can be found from Eq. (2.27) and this
has been done for air at ambient conditions to give the size/density
combinations which must satisfy Eqs (3.7) and (3.8), that is lines ZZ and YY
respectively in Fig. 3.4. By and large, the effect of increasing temperature
and/or pressure is to reduce Umf and this would cause YY and ZZ to move to
the right; i.e. the system would become more B-like.
There is another criterion for distinguishing between groups Band D; this
is entirely empirical, being based on experimental measurements of the
spoutability of powders. Whether or not a powder can be spouted depends
not only on the size and density of the particles - nozzle diameter bed
depth, ~nd bed.di~meter a~lplaya part. Baeyens and Geldart (1973)ad~pted
the arbitrary cntenon that If a bed of the powder 0.3 m deep can be spouted it
belongs to group D; this gave the numerical inequality that if:

Umb

--=

Umf

2,300 p?126 p,?523 exp (0.716F)


d~8 go.934 (pp _ pt)O.934

Since there are rarely any particles less than 45 fLm in a liquid fluidized
system F = 0 and the exponential term in the numerator becomes equal to
unity.
In a series of photographs, Harrison, Davidson, and de Kock (1961) show
the appearance of liquid fluidized beds of various solids. Lead shot was
fluidized by various concentrations of glycerol solutions in order to vary the
liquid viscosity, and paraffin was used to fluidize particles having approximately the same size but different densities.
In Table 3.2 the basic data relating to the experiments of Harrison,
Davidson and de Kock (1961) are listed, together with values of Umb/Umf
calculated from Eq. (3.10) and the qualitative appearance of the beds as
judged from the photographs reproduced in Davidson and Harrison (1963). It
can be seen that at values of UmJUmf less than about 2 the beds exhibited
slightly or appreciably disturbed fluidization, which may be construed as
equivalent to bubbling in gas fluidized beds; at higher values of Umb/Umf the
beds appeared to be quiescent. This qualitative agreement is encouraging and
. warrants further investigation. For the present the inequality below may be
used with caution to identify liquid fluidized systems likely to exhibit nonquiescent behaviour at voidages up to 0.75. Thus 'bubbling' will occur if

the powder is spoutable and is therefore in group D. This equation is plotted


as line SS in Fig. 3.4 and gives results similar to those based on the mode of
gas by-passing, King and Harrison (1980) found that the minimum spouting
velocity decreases with increasing pressure which suggests that, in common
with YY and ZZ, SS should move to the right as pressure increases.

3.4

APPLICATION

TO LIQUID FLUIDIZATION

It has long been known that in some liquid fluidized systems instabilities
which resemble bubbles can be seen, for example 770 fLm lead shot in water
(see Davidson and Harrison, 1963). The instabilities develop and grow more
rapidly as the bed expands (i.e. as U/Umf increases), as the density and size of

The concept of four broad groups of powders each of which has distinctive
fluidization characteristics is now widely accepted. Their representation on a
'phase' diagram (Fig. 3.4) when air is the gaseous medium is useful but shows
only part of the picture. As we move away from groups D and B towards
groups A and C, parameters other than particle density and mean size assume
increasing importance, and a multi-dimensional diagram is required. There
remains much to be understood, but two lessons emerge from the ideas
presented above.

Firstly, it is unwise to assume that theories deduced on the basis of data


gathered from a powder in one group are necessarily applicable to a powder
in another. Secondly, although cold modelling is useful and often essential
when studying scale-up, changing the temperature and pressure of the gas
may affect crucially the behaviour of the powder, particularly if it happens to
have a size/density which puts it near the boundary of groups A and C where
interparticle forces play an important role.

A new catalyst to be used in a hydrogenation process has the following


properties:
Skeletal (absolute) density
Particle porosity

Comparison of experimental data on liquid fluidization


(3.10). Data from Harrison, Davidson, and de Kock
(1961).

Weight (%)
of
glycerol

Appearance of
bed of Yoidages
of:

PI

kglm3

10-3

36.3 x
16.6 x 10-3
3.58 X 10-3
X 10-3
1

1,190
1,170
1,110
1,000

75
66
39
0

Particles : lead shot

Pp

UmJUmf

0.6

0.75

(Eq. 3.10)

Q
Q

Q
Q

SB
SB

B
B

6.57
4.36
1.94
0.98

11,320 kglm3, dp

-30
+20

-45
+30

1,500
2,900
7,430
11,320

(Eq. 3.10)

Q
Q

Q
Q

SB

SB
B

19.8
5.9
2.03
1.32

Liquid : PI = 780
JLI =

kglm3

2 X 10-3 Ns/m2

-63
+53

-75
+63

-90
+75

-105 -125 -150


+90 +105 +125

1
particle + lIskeletal
porosity density
1
_
3
0 .4+ 1/3 - 1.364 gm/cm
0.1

UmJUmf

0.75

-53
+45

Step 1. Calculate particle densities and mean sizes:

0.65

Cold model tests using air show the powder to have ideal fluidization
properties.
Conditions in the full-scale plant are such that the gas density and viscosity
will be 1 kg/m3 and 1.3 x 10-5 N s/m2, respectively.
Over a period of several months the conversion efficiency decreases and a
sample of powder from the bed is found to have a particle porosity of 0.1
cm3/gm and only 5 per cent by weight less than 45 p.m.

770 JLm

Appearance of
bed at Yoidages
of:

kg/m3
0.4 cm3/gm

= 3000

3
lf3

= 2.31

gm/cm

From the size distribution table and dp = 1fIx;!dpi) (Eq. 2.8):


Initial mean particle size dp1 = 58 p.m, fines fraction F1 = 0.19
Final mean particle size dp2 = 70 p.m, F2 = 0.05
Step 2. Calculate UmJUmf for cold model, initial, and final reactor conditions,
using Eq. (3.2):
Umb
Cold model U
mf

2,300

= 3.09

1.2126(1.84 X 10-5).523 exp (0.716 x 0.19)


(58 X 10-6).8 [9.81 x (1364 _ 1.2)]934
X

Initial reactor conditions

Final reactor conditions

mb

U
Umf

U
~
Umf

2.51
/-L
/-Ll

1.115

The powde /gas system is clearly in group A under both cold model and initial
reactor co ditions, which would give ideal fluidization behaviour. Under the
final reac or conditions the powder is close to group B with consequent
reduction in bed expansion and catalyst fluidity.

PsT
PsLP

bed voidage and incipient fluidization and


bubbling velocities
viscosity of gas
viscosity of liquid
tapped (maximum) bulk density of powder
most loosely packed (minimum) bulk density
of powder
density of gas
density of liquid
particle or piece density

kg/ms
kg/ms
kg/m3
kg/m3

Step 3. Es 'mate the maximum stable bubble sizes for initial and final reactor
using Eq. (4.13)
(a) Initial conditions. Calculate terminal velocity for particles:

2.7 dp, that is 2.7 x 58 /-Lm = 157 /-Lm and Pp = 1,364 kg/m3
From Eq. (6.12), CDRe2 = 408; from Fig. 6.7, Ret = 9.6; therefore;
lit = 0.795 m/s
From Eq. (4.12): (deq)max =

2 v~

0.127 m

(b) Final conditions. Use the calculation procedure as in (a):

(deq)max = 0.467 m
(c) Comments
The much lower UmJUmf value for the final reaction conditions is a clear
predictor of deteriorating fluidization conditions, and the larger maximum stable bubble size confirms this. Coupled with a lower particle
(internal) porosity it is not surprising that the conversion efficiency
suffered.

mean sieve size of a powder


mass fraction of powder less than 45 /-Lm
adhesive force transmitted in a single contact
between two particles
acceleration of gravity
superficial velocity of gas
rise velocity of an isolated bubble
incipient fluidization and bubbling velocities
terminal velocity of particle

9.81 m/s2
m/s
m/s
rn/s
m/s

Baeyens, J., and Geldart, D. (1973). In Fluidization and Its Applications, Toulouse, p.
263.

Davidson, J.F., and Harrison, D. (1963). Fluidized Particles, Cambridge University


Press.
Donsi, G., Moser, S., and Massimil1a, L. (1975). Chern. Eng. Sci., 30,1533.
Dry, R.J. Shingles, T., and Judd, M.R. (1983). Powder Techno., 34,213.
Geldart, D. (1972). Powder Techno!:, 6,201.
Geldart, D. (1973). Powder Technol., 7, 285.
Geldart, D., Hamby, N., and Wong. A.C.Y. (1984). Powder Technol., 37, 25.
Geldart, D., and Wil1iams,J.C. (1985). Powder Technol., 43, 181.
Grey, R.O., and Beddow, J.K. (1968/69). Powder Techno!., 2, 323.
Harrison, D., Davidson, J.F., and de Kock, J.W. (1961). Trans. Instn Chern. Engrs,
Lond., 37, 328.
Jackson, R. (1963). Trans. Instn Chern. Engrs, 41, 13.
Jackson, R. (1971). In Fluidization (Eds. J.F. Davidson and D. Harrison), Academic
Press.
King, D., and Harrison, D. (1980). Powder Technol., 26, 103.
Moleros, O. (1967). In Proc. Intern Syrnp on Fluidization, Eindhoven, Netherlands
University Press, p. 134.
Molerus, O. (11/82). Powder Techno!., 33, 81.
Oltrogge, R.D. (1972). Ph. D. Thesis, University of Michigan.
Rietema, K. (1984). Powder Technol., 37, 5.
Simpson, H.C., and Rodger, B.W. (1961). Chern. Eng. Sci., 16, 153.
Verloop, J., and Heertjes, P.M. (1970). Chern. Eng. Sci., 25, 825.
Wilhelm, R.H., and Kwauk, M. (1948). Chern. Eng. Prog., 44, 201.
Wong, A.C.Y. (1983). Ph. D. Dissertation, University of Bradford.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Hydrodynamics of Bubbling
Fluidized Beds

The behaviour of most gas fluidized beds is dominated by the rising gas voids,
conveniently termed 'bubbles', which characterize these systems. In analysing
the behaviour of bubbling fluidized beds, it is essential to distinguish between
the bubble phase (or 'lean phase'), i.e. the gas voids containing virtually no
bed particles, and the particulate phase (also known as 'dense phase' or
'emulsion phase') consisting of particles fluidized by interstitial gas. A
bubbling bed can conveniently be defined as a bed in which the bubble phase
is dispersed and the particulate phase is continuous. The flow regimes
observed at high velocity, discussed in Chapter 7, correspond to beds in which
the volume occupied by the lean phase is so high that the particulate phase no
longer forms a continuous medium between discrete bubbles.
Therising bubbles cause motion of the particulate phase which is the main
source of solids mixing in bubbling beds, discussed in Chapter 5. This particle
motion in turn causes the temperature uniformity and high bed/surface heat
transfer coefficients characte'ristic of fluidized beds and reviewed in Chapter
9. Since the gas in a bubble is not in direct contact with the bed particles, it
cannot take part in any reaction between gas and solids. Thus interchange of
gas between bubbles and interstitial gas in the particulate phase can
determine the performance of a fluidized bed reactor. The influence of
bubble characteristics on fluid bed reactors is explained in Chapter 11. Thus
an understanding of the behaviour of the bubble phase in fluidized beds is
essential for understanding the applications of these devices. Section 4.2 of
the present chapter considers the behaviour of individual bubbles and of

53

slugs, i.e. bubbles so large that their shape and rise velocity are determined by
the dimensions of the column containing the fluidized bed rather than by their
own volume.
The gas which fluidizes the bed and forms bubbles must normally be
distributed uniformly over the cross-section of the bed; this is achieved by
introducing it through a gas distributor. Design requirements for gas
distributors are d'
sed in Section 4.3, and the flow patterns observed in the
vicinity of the distributo
re reviewed in Section 4.4. Finally, the behaviour
of continuously slugging an freely bubbling beds is discussed in Sections 4.5
and 4.6, concentrating on prediction of the most important characteristics:
bubble size and flowrate, and bed expansion.
In many respects, bubbles in gas fluidized beds are analogous to gas
bubbles in viscous liquids (see, for example, Davidson, Harrison, and Guedes
de Carvalho, 1977). This analogy has been of immense value in interpreting
the behaviour of bubbling fluidized beds, since bubbles are more readily
observable hi transparent liquidS, and the analogy is used throughout this
chapter. There is, however, an important difference between liquids and
fluidized beds. A gas bubble in a liquid is bounded by a distinct interface;
material can be transferred across this interface by diffusion but not by bulk
flow. On the other hand, the surface of a bubble in a fluidized bed is a
boundary between a particle-lean region and a region of high solids
concentration which is permeable to gas. Therefore bulk flow can occur
between a bubble and the surrounding particulate phase, so that material can
be transferred across the boundary by convection as well as by diffusion.
These processes of interphase transfer are considered in detail in Chapter 11..

where Vb is the bubble volume. A bubble in a viscous liquid shows the form of
Fig. 4.1 if it is sufficiently large for surface tension forces to be negligible
(Clift, Grace, and Weber, 1978), i.e. if
Eo

PI d~q g > 40
(T

where PI and (T are the liquid density and interfacial tension, and the
dimensionless group Eo is known as the Eotvos number. Since fluidized beds
lack a phenomenon equivalent to inteliacial tension, (T is effectively zero.
Thus bubbles in fluidized beds satisfy Eq. (4.2), and the analogy is
reasonable.
The shape of such a bubble can conveniently be described by the 'wake
angle' between the nose and the lower rim, Ow in Fig. 4.1. Grace (1970)
showed that Ow is a function of the bubble Reynolds number:

where Ub is the bubble rise velocity and


shapes of bubbles in liquids are represented
by:
Ow

50

190 exp (- 0.62 Re~.4)

the liquid viscosity. Observed


(Clift, Grace, and Weber, 1978)

JLI

(Eo>

40, Reb > 1.2)

(4.4)

where Ow is expressed in degrees. This relationship is shown in Fig. 4.2, with a


theoretical relationship derived by Davidson et al. (1977):

Reb
4.2.1

= (-2---3-CO-S-o:-+-c-o-s3-Ow-

Shapes of Single Bubbles

Figure 4.1 shows an idealized bubble, the general shape being well known
from photographic studies of bubbles in liquids (see, for example, Clift,
Grace and Weber, 1978) and X-ray studies of bubbles in fluidized beds (see,
for example, Rowe, 1971). The upper surface of the bubble is approximately
spherical, and its radius of curvature will be denoted by r. The base is
typically slightly indented. Since r is not readily determinable, it is usually
more convenient to express the bubble size as its 'volume-equivalent
diameter', i.e. the diameter of the sphere whose volume is equal to that of the
bubble:

which is in qualitative agreement with the empirical result. It is worth noting


that, for Reb < 100, Ow is constant at 50, as predicted by Rippin and
Davidson t1967); this is the shape observed for spherical cap bubbles in low
viscosity liquids such as water. Other parameters of interest are the frontal
diameter of the bubble db and the volume of the circumscribing sphere not
occupied by the bubble Vw' For reasons explained below, Vw is termed the
'wake volume', and the ratio of wake/sphere volume, that is 3Vw/4nr, is
known as the 'wake fraction'
Figure 4.2 shows values of deq/r, dJr, and the
wake fraction, calculated assuming that the wake angle is given by Eq. (4.4)
and the base of the bubble is flat.

tw.

"Note that the wake fraction is sometimes defined as the wake volume Vw divided by the bubble
volume Vb, in which case it is given the symbol ~w' The relationship between the two is ~w =
/.J(1 - /w).

Bubble
volume

Vb

Wake
Angle

8w

gO

Wake
volume

Vw

Quantitative application of these results to fluidized beds is limited by lack


of sufficient information on the effective viscosityof the particulate phase, or
even on how realistic is the assumption that the particulate phase behaves as a
Newtonian medium. However, the effect of viscosity is a convenient starting
point for discussing the effect of particle characteristics on bubble behaviour
in fluidized beds. The essential feature of Fig. 4.2 is that decreasing the
viscosity of the medium surrounding the bubble increases Reb and therefore
causes the bubble to become flatter (lower Ow) and the wake fraction to
increase. Grace (1970) used this effect to infer the effective viscosity of
fluidized beds from bubble shapes reported by Rowe and Partddge (1%5).
Values of Ow ranged from 90 to 134, corresponding to Reb from 2 to 10,
implying effective viscosities from 0.4 to 1.3 N s/m2 (i.e. similar to the
viscosity of engine oil). Apparent viscosities have also been measured by
more conventional techniques (see Schugert, 1971), involving immersing
rotating bodies or surfaces in the bed. Since these techniques disturb the local
structure of the bed, the significance of the results is questionable. However,
the values obtained are generally consistent with those inferred by Grace.
Hetzler and Williams (1969) have given a semiempirical expression for the
effective low-shear viscosityof a fluidized medium. It was derived mainly from
data for liquid/fluidized systems and its general applicability to gas/fluidized
beds is untested, but it seems to explain some of the differences in behaviour
between particle groups. The effective viscosity is predicted to increase with
particle diameter for group A powders, to increase weakly with dp in group B
powders, and to be high but insensitive to dp in group D powder. The
viscosity estimates of Grace (1970) and Schugert (1971) show this general
trend. The results given by Rowe (1971) show the corresponding effect on
bubble shape : on increasing dp from group A to group B, bubbles become

Shape
Parameters

y~-----

Wake Fraction

Reb

Figure 4.2
number.

Bubble parameters as a function of Reynolds

more rounded and the wake fraction decreases, from 0.4 for 50 /-Lm cracking
catalyst to 0.28 to 0.3 in group B ballotini. More recently, Rowe and Yacono
(1976), working with freely bubbling beds of groups A and B reported that
~article size had no detectable effect on average bubble shape.' However, it is
lIkely that the changes of shape resulting from bubble interaction and
coales.cence (see Section 4.6) masked any effect of particle characteristics.
Cranfield and Geldart (1974) report that bubbles in group D beds lack a

58

GAS FLUIDIZATION

TECHNOLOGY

deeply indented base (that is ()w approaches 180); in terms of Fig. 4.2, this
confirms that group D materials display very high effective.viscosity, so that
Reb is low. Although theory is lacking, irregular particles are likely to show
higher effective vi~osity and therefore lower Reb and wake fraction. Results
quoted by Ro~1)971) show exactly this effect: the wake fraction in beds of
irregular particles is typically 0.2 compared to 0.28 to 0.3 in beds of spherical
material. Increasing bubble size in a medium of fixed viscosity increases Reb
and should therefore increase the wake fraction. Rowe and Everett (1972)
observed such an effect, and Rowe and Widmer (1973) showed that the
results were well correlated by:
Vb

w
V = 1- V
=

,ph

,ph

where the braced term is a weak function of Reb, as shown in Fig. 4.2. For
Reb > 100 this term is constant at 0.71, and the resulting expression:

exp (- 0.057db)

where db is the frontal diameter in centimetres. Equation (4.6) was derived


from observations of bubbles with db up to 16 cm, with a range of group B
materials; for db larger than 16 cm, it predicts unrealistically low values of
Vt/V,ph' In the current stage of the technology, all that can be said about
other groups is that the wake fraction is generally larger (up to 0.4) for group
A, smaller for irregular group B particles (typically 0.17 to 0.22), and very
small for group D.
A spherical cap bubble in a liquid carries a closed circulating wake if Reb <
110 (Clift, Grace and Weber, 1978). Since bubbles in fluidized beds fall in this
range, they would be expected to carry closed wakes of particulate phase, and
this has been observed (see Rowe, 1971). To a good approximation, the wake
volume can be taken as the volume of that part of the circumscribing sphere
not occupied by the bubble, Vw (see Figs 4.1 and 4.2). It is important to note
that the decrease in wake fraction on moving from group A through group B
to group D means that the volume of particulate phase transported per unit
bubble volume decreases. This is one cause of the change in particle mixing
characteristics between groups, discussed in Chapter 5.

4.2.2.

below the range for which Eq. (4.7) is strictly valid in liquids. Even so, Eq.
(4.7) has been widely used for fluidized beds, and inaccuracies are no doubt
masked by the erratic velocity variations typically observed (Rowe and
Yacono, 1976). In terms of the volume-equivalent diameter, Eq. (4.7) can be
written:

has been widely used to predict the rise velocity of isolated bubbles in
fluidized beds (Davidson and Harrison, 1963). However, observed rise
velocities are generally smaller than predicted by Eq. (4.9), values of
ut/V(gdeq) from 0.5 to 0.66 being typical for groups A and B (see'Davidson,
Harrison, and Guedes de Carvalho, 1977). From Fig. 4.2, these values
correspond to Reb < 60, broadly consistent with the range inferred from the
bubble shape.
For groups A and B, it is possible to estimate bubble rise velocities from the
shape, as expressed by the wake fraction V wIV'Ph, and this approach is at least
self-consistent. From Eq. (4.8), it follows that:
V2 (

Ub

= -3

)116

1
1- V IV
w

sph

V(gdeq)

For rounded group B particles, Eq. (4.6) can be used to estimate VWlVsph' For
angular group B particles, the wake fraction is typically 0.2, so that
ut/V(gdeq) should be roughly 0.5. The shapes of bubbles in group A particles
suggest that Reb is about 10, so that ub/V(gdeq) should befrom 0.5 to 0.6.
For group D particles, Cranfield and Geldart's (1974) results give:

Rise Velocity of an Isolated Bubble

Davies and Taylor (1950) showed that the rise velocity Ub of a spherical cap
bubble is related to its radius of curvature by:
Ub

2/3 V(gr)

(4.7)

Experimental results show that Eq. (4.7) is reliable for bubbles in liquids if
Reb is greater than about 40; below this value, departures from the spherical
cap shape have a significant effect on Ub(Clift, Grace and Weber, 1978).
Bubbles in fluidized beds typically have Reynolds numbers of order 10or less,

Since db and deq are virtually the same in group D beds, Eqs (4.9) and (4.11)
are equivalent. This is unexpected, since observed bubble shapes suggest that
group D prticles should be treated as low Reb systems (see Section 4.2.1). In
the present state of the technology, all that can be concluded is that treating
the dense phase as a Newtonian medium does not explain all features of
bubble behaviour and that Eqs (4.9) and (4.11) can be used in group D beds.

4.2.3 Bubble Break-Up


Bubbles in viscous liquids and in fluidized beds break up by the process shown
schematically in Fi~. 4.3 (Clift and Grace, 1972). An indentation forms on the
upper surface ofAhe bubble and grows as it is swept around the periphery by
the motio~the
particles relative to the bubble. If the curtain grows
sufficientlY to reach the base of the bubble before being swept away, the
bubble divides. When the surface tension effects are negligible, as in fluidized
beds, the growth rate of such a disturbance decreases with increasing
kinematic viscosity of the medium surrounding the bubble, so that the
maximum stable size a bubble can attain before splitting increases with
increasing kinematic viscosity (Clift, Grace and Weber, 1974). Analysis of
this process leads to realistic predictions of the maximum stable size of
bubbles and drops of conventional fluids (Clift, Grace, and Weber, 1978).
Application to fluidized beds is again limited by lack of sufficient data on
effective viscosities, but it does explain the differences betw~n.powder
groups. In group A particles the effective viscosity and h6frce-'~aximum
bubble size are relatively small, and both increase with particle diameter. For
group B particles the effective viscosity and bubble size are larger and less
dependent on particle size, while for Group D particles the maximum bubble
size is too large to be realized.

that Eq. (4.12) predicts the correct order of magnitude for the maximum
bubble size. Matsen (1973) found that Eq. (4.12) underpredicts the maximum
bubble size, typically by a factor of 4, the difference being larger for very fine
particles. Geldart (1977) suggests that a better fit with experimental data for
group A powders is obtained if the terminal velocity in Eq. (4.12) is
calculated for particles of size:
d;

= 2.7d

(4.13)

where dp is the mean size of the powder (Eq. 2.8). Figure 4.4 shows (deq)max,
for particles of density 2,500 kg/m-', fluidized by air at 1 and 15 bar, and 300
and 1,000 K, calculated from Eq. (4.12) with Vt estimated from the
correlations given by Clift, Grace and Weber (1978) (see Fig. 6.7) and using
Eq. (4.13). For group D and large group B particles, the maximum stable
bubble size is so large that in practice it would not impose a limitation on the
size of bubbles present in a bubbling bed. For both group A and group B
particles, increasing temperature and, especially, increasing pressure reduces
the maximum bubble size, thus tending to make fluidization 'smoother'. This
is consistent with observations (see Chapter 3).
.

;;1000K
//300K

For quantitative predictions of maximum stable size, a method developed


by Harrison et at. (see Davidson and Harrison, 1963) must be applied. It was
proposed that bubbles split when particles are carried up from the base of the
bubble by the circulation of gas within the bubble. The gas was assumed to
circulate with velocity of order Ub, so that a bubble breaks up if Ub exceeds the
terminal velocity of a bed particle. Since the predictions are at best
approximate, Eq. (4.9) may be used to estimate Ub' The resulting expression
for the maximum stable bubble size is:
2v~
(deq)max = .g

(4.12)

where Vt is the particle terminal velocity. Although the physical process on


which Eq. (4.12) is based is now known not to occur, there is some evidence

1mm
10!J.m

100!J.m

1000 !J.m

dp
Figure 4.4 Maximum stable bubble size as a function
of mean size of powder, and gas pressure or temperature (calculated using Eqs 4.12 and 4.13).

4.2.4

Wall Effects:

Slugs

Provided that deq is less than about 0.125 times the bed diameter D, the shape
and rise velocity ot a bubble are unaffected by the walls. For deqID greater
than 0.125, waJlAffects reduce the rise velocity and cause the bubble to be
more rounfte, so that the wake fraction is reduced (Clift, Grace, and Weber,
1978). For deqID between 0.125 and 0.6, the rise velocity can be estimated
roughly (Wallis, 1969) by:
Ub

)
l.13Uboo exp- (dD
eq

(4.14)

where Uboois the rise velocity in the absence of wall effects. For deqlD greater
than 0.6, the bubble velocity becomes completely controlled by the bed
diameter, and the bubble is then termed a 'slug'. Figure 4.5(a) shows
schematically the shape of an ideal slug. The rise velocity of such a slug is
given (see Clift, Grace, and Weber, 1978) by:
Usi = 0.35 Y(gD)

(4.15)

Slugs sometimes adhere to the wall of the tube, as shown schematically in Fig.
4.5(b); in this case the rise velocity is approximately Y2 times the value from

Eq. (4.15).
Figure 4.5(c) shows a rather different type of slug, more accurately termed
a 'plug' , or solids slug, which completely fills the column so that particles rain
through the void rather than moving around it. For groups A and B, this
phenomenon normally occurs only in tubes of small diameter. For coarse,
very angular, or cohesive particles, 'plug' flow occurs in beds of much larger
diameter and is suppressed by roughening the walls (Geldart, Hurt, and
Wadia, 1978).

Clearly slugs can only exist if the diameter of the largest bubble is greater
than q.6D. The question of whether a continuously fluidized bed contains
bubbles or slugs is considered in Section 4.6.

Example 4.1: Calculation of the properties of single bubbles


In the following example, bubble properties are estimated for a fluidized bed
roaster for zinc sulphide ore. Typical conditions for such an operation are
given by Avedesian (1974):
Bed conditions : 1,260 K, 1 bar
Off-gas composition (% molar)
O2
N2

:
:

4.0
77.3

H20: 7.8
S02 : 10.9
The bed is operated at velocities many times minimum fluidization, so that it
will be assumed to be fully mixed with the off-gas composition also
representa~ive of the gas~s within the be~. The bed particles, primarily ZnO,
have denSity 5,300 kglm . The calculatlOns are carried through for 30 p,m
particles (group A) and 100 p,m particles (group B) . ..The latter is typical for
the surface-volume mean diameter of bed particles in such an operation. Two
possible beds are considered (Fig. 4.6):
(a) A production plant of large diameter.
(b) A pilot plant 0.2 m diameter.

Solution
(a) Maximum stable bubble size. The properties
as:

Figure 4.5 (a) Axisymmetric slugs, (b) asymmetric slugs, (c) solids, square-nosed or
'plug' slugs.

Density: 0.32 kg/m3


Viscosity: 5 x 10-5 N s/m2
Particle diameter: dp
30 p,m
Particle type:
Group A
d~ (from Eq. 4.13)
81 p,m
CDRe~
4.72
Ret
0.2
Vt (m/s)
0.42
(deq) from Eq. (4.12)
3.7 cm
Frontal diameter (db)max
4.3 crn

of the above gas are taken

100 p,rn
Group B
270 p,rn
174.6
5
3.18

2.06

rn

2.46 rn

Notes

Equation (4.12) predicts reasonable values for (deq)max for both group A and
group B particles. n practice it is unlikely that the bubbles would reach 2.06
m in diamet . the group B particles because the normal operating bed
depth would be insufficient. Frontal diameter in each case is calculated
assuming that the wake fraction is 0.4; this value is reasonable for the group A
particles, and is probably reasonable also for group B given that the particles
are not smooth and spherical (see Section 4.2.1).
b) Isolated bubble rise velocity. Bubble rise velocity is calculated from Eq.
(4.10). Assuming that the wake fraction, VwIVsph, is constant at 0.4, Eq.
(4.10) gives:
Uboo

= 0.51 V(gdeq)
= 1.61 V deq m/s with deq in m

For a bed of diameter 0.2m, wall effects become significant for deq >
2.5 cm (see Section 4.2.4); they are therefore slightly significant for
bubbles in the 30 /-Lm particles. In a bed of 100 /-Lm particles, wall effects
must be considered. Curve (b) in Fig. 4.6 shows values calculated from
Eq. (4.14) for 2.5 cm < deq < 12 cm and from Eq. (4.15) for deq > 12cm.
Clearly the two equations do not match at deql D = 0.6. The discontinuity
occurs because Eq. (4.14) was developed for bubbles with high Reh, for
which Uboo is given by Eq. (4.9). Bubbles in fluidized beds are more
'rounded' (that is 8w is larger), so that for given deq they rise more slowly,
have smaller dh, and are less sensitive to wall effects. Equation (4.15) is
well tested for fluidized beds, and may therefore be applied with more
confidence. Best estimates are therefore obtained by replacing Eq. (4.14)
by an interpolation of the form shown by the broken curve in Fig. 4.6.

Resulting values, appropriate to a bed of large diameter, are shown as


curve (a) in Fig. 4.6.
4.3.1
Ub

{m/sl

Figure 4.6 Bubble rise velocities calculated


plant, (b) in 0.2 m diameter pilot unit.

in worked Example 4.1: (a) in large production

Types of Gas Distributor

For satisfactory operation of a fluidized bed, the gas must normally be


distributed uniformly across the bed area. In addition, the distributor must
prevent the bed solids from falling through into the 'windbox' and must be
able to support forces due to the pressure drop associated with gas flow
during operation and the weight of the bed solids during shutdown.
For laboratory studies, it is frequently convenient to use a distributor of
some porous material (see Chapter 13), but this is not practical for industrial
fluid beds. One of the simplest types of distributor, also much used in
laboratory studies, is a single perforated plate; i.e. a metal sheet drilled or
punched with a regular array of orifices or slots. Provided that the diameter or
width of the perforations is less than about ten particle diameters, littie
'weeping' of solids occurs. Devices to reduce weeping with this type of
distributor indude the use of two or more staggered perforated plates and a
'sandwich' of coarse particles contained between two perforated plates or
meshes (see Kunii and Levenspiel, 1969). If the fluidizing gas contains dust or
liquid droplets, these particles may deposit in the perforations to block the
distributor. Such deposition is reduced or eliminated by using tapered orifices
or holes as shown in Fig. 4.7(a) (Doganoglu et al., 1978) .
. Figures 4.7(b), (c), and (d) show types of gas distributors which are
particularly suitable for high temperature operation. In each case, the gas
passes up an array of vertical pipes or 'risers', through the metal load-bearing
-plateof the distributor. The plate is covered by a layer of refractory, normally
laid or cast between the risers. The insulation provided by the refractory layer

ensures that the plate is at a temperature close to that of the incoming gases,
although the bed itself may be at a much higher temperature. In the type of
distributor usually known as a 'tuyere', shown in Fig. 4.7(b), the riser has a
single hole whos .is is vertical. Alternatively, the riser may be capped and
the cap .
with radial holes, usually four, sixor eight in number, as shown
in Fig. 4.7(c). With this type of distributor, sometimes termed 'multi-hole
tuyere', the gas enters the bed horizontally. The operating advantage is that
weeping and dumping of solids through the distributor are much reduced if
the radial passages are sufficiently long. A similar advantage is achieved by
the type of distributor shown in Fig. 4.7(d), known as a 'quonset' ('Quonset'
being the American term for what is known in the British Commonwealth as a
'Nissen huf). Here the riser is capped by a horizontal length of pipe cut on a
plane through its axis to form a half-cylinder whose open side is level with the
top of the refractory. The risers may also be capped by bubble caps similar to
the devices used in distillation (see Kunii and Levenspiel, 1969).

If it is desirable to separate the load-bearing and gas distribution


requirements, the gas can be introduced into the bed by a pipe grid, shown
schematically in Fig. 4.7(e). This arrangement is also suitable if more than
one gas stream is to be introduced to the bed without premixing. The pipe
grids are typically horizontal pipes, each with an array of branching smaller
pipes. These small pipes may point upwards, as in Fig. 4.7(e), or be inclined
downwards to reduce passage of bed solids into the pipe grid.
If it is necessary to remove solids continuously from the bed, e.g. ash in
coal processing, it may be convenient to provide a conical base to the bed as
shown in Fig. 4.7(f). Ash is removed from the apex of the cone and fluidizing
gas introduced through pipes in the tapered section. If these pipes are inclined
downwards, as shown in Fig. 4.7(f), solids weeping is reduced.
4.3.2

Design Criteria

From the point of view of gas distribution, the essential requirement is to


design the distributor so that the gases passing through it experience a
sufficiently high pressure drop, Apo. The critical requirement is generally
held to be the ratio of the distributor pressure drop to that across the fluidized
bed, Apo/ApB' Agarwal, Davis, and King, (1962) suggest that Apo should
be 0.1 to 0.3 times APB or 3.4 X 103 N/m2 (0.5 Ib/in2), whichever is the
greater, and this is broadly consistent with the recommendations of Hanway
(1970), Whitehead (1971) and Qureshi and Creasy (1979) point out that
some fluidized beds operate successfully at much lower values of ApdApB'
Apart from the 'ease of fluidization' of the bed solids, a property which is
related to cohesiveness or stickiness but is impossible to define or quantify,
the value of Apo/ApB required depends on the ratio of bed depth H to
diameter D. Qureshi and Creasy (1979) concluded, from a review of
published information, that the distributor pressure drop required for
satisfactory operation was given by:
Apo
--=

APB

0.01

+ 0.2 { 1- exp

(~:)}

so that a pressure drop ratio of 0.21 is required for large diameter beds. If the
bed solids contain tars or other 'sticky' components, it is advisable to increase
this value to at least 0.3.
Since the value of APB is fixed by the weight of solids per unit bed area (see
Chapter 2), selection of an appropriate value for Apo/ApB fixes the required
I::.po, The number of orifices, nozzles, etc., to achieve this pressure drop
must now be selected. We consider first the simplest case of a perforated plate
distributor. If the superficial gas velocity in the windbox is Uw and the

fractional open area of the distributor (i.e. the fraction of the total area open
to gas flow) is fon then the average gas velocity through the orifices is:
Uor

Uw

= -;:lor

It should be noted that Uw differs from the superficial gas velocity in the bed if
the entering gas is at a different temperature from the bed, or if the processes
occurring in the bed cause a change in molar f1owrate, or if the distributor
pressure drop is a significant fraction of the absolute operating pressure.
The pressure drop through the distributor then follows from Bernoulli's
equation:
A

Pgw
Po = -2-

I(c:--)2 Uor

2}

Uw

where Pgw is the density of gas in the windbox and Cd is the orifice discharge
coefficient, which accounts for the fact that the gas does not pass uniformly
through the whole orifice area so that the true average velocity is greater than
Uor' Normally for ~ 1 so that Uor ~ Uw and Eqs (4.17) and (4.18) can be
simplified to give:
for

Uw
Cd

jf.\2Apo
Pgw

Ideally, dor should be as small as is feasible: reducing dor decreases the size of
bubbles formed at the distributor, and therefore improves gas/solid contacting in the bed (see Chapter 11). It is also advisable to check that all
orifices will be equally 'active' in passing gas into the bed, by a criterion
developed by Fakhimi and Harrison (1970) which can be written:
Uw
( -U

)2

>1+

(2)

m/)

2pp g(]-E
------

1--

Lj

(Cdfor)2
--

Pgw

7T

mfw

Umfw

where Pp is the particle density, Emf is the void fraction in the bed at
minimum fluidization, Lj is the length of the jet above the orifice (see Section
4.4.1), and Umfw is the minimum fluidization velocity corresponding to gas at
windbox conditions. If Eq. (4.23) is not satisfied, the orifice diameter should
be reduced to reduce Lj
With distributors of the tuyere type, the form and dimensions of each
tuyere are normally fixed in advance by other considerations. The number of
holes or orifices is then selected to give the required Apo. For an initial
estimate, Eq. (4.18) can be used, rearranged as:
Uort

2Apo

Cd ( --

+ Uw

)112

Pgw

_ j(2APo)

(4.19)

-Cd

--

Pgw

The discharge coefficient depends on the shape of the orifice and is also a
weak function offor (Perry and Chilton, 1973). There is also a possibility that
the gas passing through an orifice into a fluidized bed experiences less
pressure drop than in the absence of particles (see Qureshi and Creasy, 1979).
For square-edged circular orifices with diameter dor much larger than the
plate thickness t, Cd can be taken as 0.6. For t/dor > 0.09, Cd can be estimated
by a correlation given by Qureshi and Creasy:
Cd

= 0.82 (

d:

0.13

where Uort is the first estimate for the mean gas velocity issuing from each
hole into the bed and Cd is estimated for the exit holes; i.e. the upper orifice
for a tuyere (Fig. 4.7b) and the radial holes for a multi-hole tuyere (Fig. 4.7c).
If the exit area per hole is aon the number of holes required per unit bed area
is:

(4.20)

Fixing for by Eq. (4.19) fixes the ratio of centre-to-centre hole spacings to dor
For orifices on equilateral triangle pitch:
s

dor

7T

2forv'3

)tn

0.9523
v'for

while for square pitch:


s
dor

7T

4for

) 1/2

0.8862
v'for

The initial estimate Nl is now improved by estimating the pressure drop


Apo,t experienced by the gas in passing through each hole at rate Qor,} using
standard results for flow through pipe fittings (Perry and Chilton, 1973). Thus
entry into the riser is treated as a sudden contraction, associated with pressure
drop:
Pgw

-2-

(2
2)
UR - Uw

U~

+ Pgw 4 =

Pgw

(3U4~_ U2~)

where UR is the mean gas velocity in the riser, The first term on the left of Eq.
(4.28) results from application of the Bernoulli equation, while the second
represents the mechanical energy loss resulting from formation of a vena
contracta at the shar' contraction.
Pressure drop due t wall friction in the riser is usually negligible. The
upper orifice of a tuye
is treated as a contraction, sharp or rounded
depending on the profile.
ntry to the outlet holes of a multi-hole tuyere is
treated as a 'tee' followed by a contraction, while a 'quonset' is equivalent to a
'tee' with flow entering the branch. The pressure drop ti.PO,I is evaluated as
the sum of the contributions from each change in gas velocity or direction.
The final estimate N for the number of holes per unit area is obtained using
the fact that the pressure drop is proportional to U~r:

N=NIJ(~;~l )

pressure drop across the distributor and bed as 1.54 x 104 N/m2 and estimates
the distributor pressure drop as approximately 0.53 x 104 N/m2. Hence:
Bed pressure drop = 1.54 x 104
= 1.01 X 104
Distributor

0.53
N/m2

pressure drop is the greater of

3.4 x 103 N/m2 and 0.3 x 1.01 x 104


= 3.0 X 103 N/m2
i.e. ti.po = 3.4 X 103 N/m2
(which is only 64 per cent of that estimated by Avedesian for a specific
operating unit).
Calculations are based on gas at the wind box conditions reported by
Avedesian:
Air at 298 K and 1.154 bar
Density Pgw = 1.35 kg/m3
Corresponding superficial velocity Uw

The average gas velocity entering the bed is then

UOr=-N

104

0.159 m/s

Uw

aor

Solution

For complicated tuyere designs it is advisable to determine the pressure drop


in a pilot plant or 'cold model', so that Uor can be selected to give the required
pressure drop under operating conditions.
Normally design calculations of the type summarized here are based on
flowsheet gas rates. It should be noted that ti.po is proportional to U;';
therefore operation below design rates may give insufficient ti.po and hence
poor fluidization. If operation with a range of gas rates is anticipated, the
distributor should be designed to give sufficient pressure drop at the lowest
gas rate foreseen. Designing for 'turndown' in this way will inevitably lead to
higher 6.po and hence higher operating costs at the higher gas velocities. As
a further check, it is also advisable to ensure that ti.po is much larger than the
dynamic pressure of gases entering the windbox, i.e.:
ti.po ~ Pgw Ufn

(4.31)

where Uin is the mean gas velocity in the duct entering the windbox. Normally
this requirement imposes no further restriction on ti.po.

(a) Single-hole tuyeres Figure 4.8(a) shows the cast steel tuyeres used in the
installation described by Avedesian (1974). The pressure drop across each
tuyere is made up of three components:
(i) Contraction loss of entering the 'riser' tube of 22 mm Ld. If the gas
velocity in the 'riser' is UR metres per second then the contraction loss is
estimated from Eq. (4.28) as:
Pgw

3U~ U;')
( -4- - -2-

1.0125 UR

0.01706 N/m

UR

13.44 UR

X (~)

Calculation of tuyere and multi-hole tuyere distributors

This example considers possible distributor designs for the zinc sulphide
roaster introduced in Example 4.1. Since the solids may have a tendency to
'stickiness',
the distributor will be designed following the criterion of
Agarwal, Davis, and King (1962) (see Section 4.3.2) to give ti.po as the
greater of 0.3ti.PB and 3.4 x 103 N/m2 Avedesian (i974) gives the total

(ii) Frictional loss in riser. It may readily be demonstrated


that this
component is negligible.
(iii) Loss on passing through 6 mm hole, Since the length of the 6 mm
opening is several times its diameter, this hole is treated as a sharp
contraction rather than an orifice.

Mean velocity in nozzle


Example 4.2:

Area ratio of contraction

(2~)2

0.0744

which is sufficiently low that the appropriate form of Eq. (4.28) can be
applied (Perry and Chilton, 1973). The contribution to the pressure drop
is then estimated as:

which is clearly t~ominant


Hence:
Total pressure drop

contribution.

= 183.2 U~

- 0.01706 N/m2

Required pressure drop = 3.4 x 1W N/m2


:. U~ = 18.66 (mlS)2
UR = 4.32 mls
Hence, if N holes per square metre are required:
N x

7T

4:" X

(0.022)2 UR

~mm

~mm

superficial velocity = 0.159 m/s

so that N = 96.8 holes/m2. Since there is only one hole per tuyere 968
tuyeres/rn2 are needed.
Area associated with one tuyere = 1.03 x 10-2 m2
i.e. if the tuyeres are arranged on square pitch, the centre to centre spacing is
V(1.03 X 10-2) = 0.102 m. This is precisely the value reported by Avedesian
(1974).
(b) Multi-hole tuyeres Figure 4.8(b) shows a possible alternative design
utilizing a similar 'riser', but with four radial holes each of 3 mm internal
diameter. The total hole area is the same as for the single-hole tuyere in case
(a). The pressure drop components considered in case (a) are still present,
with an additional term arising from the deflection of the gas entering the
holes. Treating the configuration as a 'tee', the extra pressure drop is
estimated (Perry and Chilton, 1973) as Pgw U~12 = 7.95 x 10-2 Uit N/m2.
Hence the total pressure drop is 183.3U~ - 0.01706 N/m2 This is
essentially indistinguishable from the value for case (a), because the pressure
drop is dominated by the velocity in the exit holes which is the same in the two
cases. Thus the spacing for these multi-hole tuyeres should be the same as for
the single-hole tuyeres.
In each case, the gas enters the bed at pressure 1.154 - 0.034 = 1.12bar at
density 1.35 x 1.2111.154 = 1.31 kg/m3. Correcting to these conditions, the
gas velocity entering the bed, Uan is 13.44 UR x 1.154/1.12 = 59.8 mls.

4.4.1

Gas Jets at Distributors

Considerable academic argument has been engendered over whether gas


entering a fluidized bed from an orifice or hole forms a series of bubbles or a
permanent gas jet with low solids concentration. Rowe, MacGillivray, and
Cheeseman (1979) showed that, when gas is introduced through a vertical
hole into a bed held close to minimum fluidization by a background gas flow,
a stream of bubbles forms immediately at the hole. However, when there is
no background gas flow - the case of common practical interest - a gas jet
forms. As in all multi-phase systems, the jet is unsteady, and when the gas
flow through the hole is more than about three times the flow needed to
fluidize the bed, it may break down into a stream of fast-moving bubbles close
to the hole. Whether or not there is a permanent gas void in the region of the
hole, there is ample evidence that the average gas and particle velocities are
very much higher and the average solids concentration is very much lower in
this region than in the rest of the bed. There seems little harm in calling these
regions simply 'jets'. Massimilla and his colleagues have modelled these
regions successfully by a modification of conventional turbulent jet theory
(see, for example, Donsl, Massimilla, and Colantuoni, 1980), so that the

analogy between fluidized beds and conventional liquids holds qualitatively.


In group D particles, rather different phenomena occur (Cranfield and
Geldart, 1974). Voids break away periodically from the distributor, spread
horizontally to form lenticular cavities, and then break up to form a number
of bubbles on a common level. The process is roughly analogous to formation
and break-up of toroidal gas bubbles in liquids (see Clift, Grace and Weber,
1978).
The practical interest in gas jets arises from the fact that they are regions of
very high gas and solids velocity. Therefore solid surfaces in the jet region are
subject to rapid erosion. Distributors such as multi-hole tuyeres which
introduce gas horizontally into the bed must be arranged so that the jets do
not impinge on neighbouring tuyeres or on the wall of the bed. When the gas
enters the bed vertically, the jet penetration determines the lowest level at
which surfaces such as heat exchange tubes should be located, and imposes a
lower limit on operating bed depth. In addition, mass and heat transfer
between gas and particles is rapid in the jet region (Donsi, MassimiIla, and
Colantuoni, 1980). The implications for the performance of fluid bed reactors'
are discussed in Chapter II.
Yang and Keairns (1979) reviewed the correlations available for predicting
the length of the 'jet' established when gas enters a fluidized bed vertically,
concluded that none was sufficiently general, and proposed their own:
Lj
dOT

= 15 {(

Pg
Pp -

)
Pg

U;T }

0.187

gdOT

where Lj is the jet length and Pg and Pp are the densities of the gas and
particles. While Eq. (4.32) appears to be the most reliable correlation for
temperatures
close to ambient, the scant available evidence (Ghadiri and
Clift, 1980) suggests that only the correlation due to Merry (1975) gives
reasonable predictions for elevated temperatures:
Lj
dOT

5.2 (

pgdOT
ppdp

)0.3

11.3 (

U;T
gdOT

)0.2 .- If

where dp is the mean bed particle diameter. Equations (4.32) and (4.33) were
developed for particles in groups A and B, but there is insufficient evidence to
assess their validity for group D. For horizontal jets, another correlation due
to Merry (1971) may be used:
_i
L

dOT

5.25 [ ptlp
ppdOT

UOT
(1 - Emf)

Pgw

ppdpg

f2]0.2

4.5

where Pg and Pgw are the gas densities in the bed and windbox. The
appropriate density for use in Eqs (4.32) and (4.33) is not clear; it is probably
more realistic to use Pgw where this differs from Pg.

4.4.2

Bubble Formation

Davidson and Schiiler (see Davidson and Harrison, 1963) have given an
analysis of bubble formation at a single upward-facing orifice in a fluidized
bed, in which the particulate phase is treated as an incompressible fluid of
zero viscosity. The initial bubble volume Vo is predicted to be a function of
the volumetric gas flow rate through the orifice QOT> but not of the orifice
diameter:

The coefficient in Eq. (4.35) is appropriate for an orifice in a flat plate; for a
protruding tuyere it should be replaced by 1.14. The frequency of bubble
formation follows as:
Qor

tb.o=-V-

The treatment of Davidson and Schiiler assumes that all gas entering the bed
forms bubbles; in fact the bubble flowrate is usually less than QOT> so that Vo is
overestimated. The error becomes particularly significant for Qor less than
about 20 cm3/s; in this case, more complicated semiempirical equations given
by Clift, Grace, and Weber, (1978) should be applied. Equations (4.35) and
(4.36) are consistent with the observation that bubble frequency is only
weakly sensitive to QOT but at high Qor the predicted frequency tends to be
too low. Bubble frequencies around 20 S-l (Rowe, MacGillivray, and
Cheesman, 1979) have been reported for high Qor' Therefore, if Eq. (4.36)
predicts tb.u less than about 10 S-l, it is probably more realistic to assume that
tb.o is 10 S-l and estimate Vo as Qo/tb.o'
Semiempirical correlations to predict the size of bubbles formed at a
distributor containing a number of gas injection points are summarized in
Table 4.1. In these expressions, Ao (U - Umf) is the gas flow in excess of that
required for minimum fluidization passing through each orifice (since Ao =
1IN), and is thus analogous to Qor in Eq. (4.35) (see Section 4.5.2). The form
of the expressions derives directly from Eq. (4.35), which can be written in
terms of the volume - equivalent diameter, deqo, as:

and the result given by Morl and Wen (1975) is essentially Eq. (4.37) in c.g.s.
units. Geldart (1972) and Darton et al. (1977) give slightly larger values, but
the differences are small and there is no clear reason for preferring anyone
correlation. The results all refer to particles in groups A and B; their accuracy
for group D is untested.

Table 4.1 Correlations for bubble size in fludized beds In the followingexpressions,
z indicates distance above distributor, D the bed diameter, and Ao the distributor area
per orifice for multiple orifice or tuyere type distributors. The forms given here for the
correlations of Geldart (1972) and Mori and Wen (1975) require dimensions in
centimetres and velocities in centimetres per second. Any consistent set of units can be
used in the correlations of Darton et al. (1977) and Rowe (1976). The parameter deqm
in the Mori and Wen correlation represents the diameter which would be attained if all
the fluidizing gas above that required for minimum fluidization were to form a train of
bubbles rising on the bed axis. The parameter Za in Rowe's correlation is a
characteristic of the distributor, effectively zero for a porous plate.

Expression

Author
Darton

el

al. (1977)

dcq

for bubble size

= 0.54(U-U

mf)"4(Z

Initial bubble size (at z

[ = 0.411

{A,,{U-Umr)}".4
in centimetre units I

dh

Moti and Wen (1975)

Ii..,

= dcq.m

- (dcq.m - li..,.o)e-O.3zJD

de"."

where dcqm = 0.374 {.".D2(U-Umr)}".4

= 0.347

{A,,(U-Umf)}"4
perforated plates
= 3.76 x 1O-'(U-Umr)2
porous plates

= 5

2d

or

d
Pg or
d

( V~r) -IJ

)0.3

0.2

gdor

Pp p

= 0.095 m
It is typical that the estimate from Merry's correlation is smaller than that
from Yang and Keairns'. To ensure that excessive erosion of surfaces such
as heat exchange tubes is avoided, such surfaces would be located no less
that 0.15 m above the ends o~._
For the second distributor of Example 4.2, the gas enters
the bed horizontally through nozzles of diameter dor = 3 X 10-3 m. From
Merry's correlation, Eq. (4.34):
Lj _- 5.25dor [ -- pgdp

dh." = 1.43 {A.(U-Umf)}O.4g-1I.2


[ = 0.361 {A,,(U-Umr)}"4
in centimetre unitsl

1.43 {A,,(U-Umrllo.4g-0.2
0.027z( U- Umr)lI .

L.

Multi-hole tuyere.

= 0)

<l.,q." = 1.63 {A,,(U-Umf)}U.4g-"2

+ 4v'A,,)""g-"2

(ii) Merry (Eq. 4.33):

ppdor

= 7.8 x 10-3 m
for
for

dh." = 0 for porous plates; no


general results given for
other types of distributor

where Emf has been assumed to be 0.4 and Pg is 0.32 kg/m3 (see
Example 4.1). If this estimate, that the horizontal jets persist for less than
1 cm from the holes, is reliable, then no especial precautions need to be
taken in the orientation of the holes. lithe jet penetration is comparable
to the tuyere spacing, then the holes should be directed so that a jet
cannot reach neighbouring tuyeres and cause excessive erosion.
(b) Initial Bubble Size

(a) Jet penetration

For each of the distributors in Example 4.2, the gas enters the bed with
velocity Vor = 59.8 mls and density Pgw = 1.31 kg/m3. The bed particles,
described in Example 4.1, have density Pp = 5,300 kg/m3 and diameter dp =
10-4 m.
Single-hole tuyere. For the first distributor considered in Example 4.2, the
gas enters the bed vertically through holes of diameter dor = 6 X 10-3 m.
Two estimates for the jet length are then:
(i) Yang and Keairns (Eq. 4.32);
Lj == 15dor

= 0.15 m

I(

rPc.r

Pgw

Pp

Pgw

gdor

0.187

Gas entering the bed reaches the bed temperature rapidly. Avedesian (1974)
reports that the superficial velocity in the bed, corresponding to conditions
for whi~h the distributor has been designed, is V = 0.78 mis, while the
minimum fluidizing velocity Vmf is approximately 0.005 mls.
For the single-hole tuyeres, the bed area per hole is Ao = 1.03 X 10-2 m2
(see'Example 4.2). Hence the excess gas flow per hole is:
Qar

=
=

Ao (V - Vmf)
3

1.03 X 10-2 (0.78 - 0.005)

8.0 x 10- m /s

The Davidson and Schiiler result (Eq. 4.37) then givesthe initial bubble size as
deq,o = 1.38 x (8.0 x 10-3).4 (9.81)-0.2

0.127 m

Since the spaces between tuyeres are filled with castable refractory to the
level of the nozzles, the form appropriate to orifices in a flat plate has been
used.

This result may be compared with the expression in Table 4.1. In c.g.s.
units:
{A (U - Umf) }Oo4 = (S.O X 103)04 = 36.4
o

Hence we obtain the following expressions:


Darton et al:
Geldart:
Mori and Wen:

deq,o

db.o
(deqo

= 15.0 cm = 0.15 m
= 13.1 cm = 0.131 m
= 0.61/3 X 0.131 = 0.110 m)

deq,o =

12.6 cm

0.126 m

Thus the estimates for deq,o are comparable and all are larger than the tuyere
spacing (0.102 m; see Example 4,2). It thus seems unlikely that the tuyeres act
independently. However, in the current state of the art, little further can be
said.
For multi-hole tuyeres, the area and gas flow per injection point is one
quarter ()f the value for the single-hole tuyeres. Assuming that the same
relationships can be used (questionable for horizontal gas entry), we obtain
the following estimates for deq,o:
Davidson and Schuler:

deq,o

= 0.073 m

Darton et al:
Geldart:
Mori and Wen:

deq,o

= 0.086 m
= 0.075 m (deq,o = 0.063 m)
= 0.073 m

db,o
deq,o

(Eq. 4.37)

Again, the values for deq,o are sufficiently large to suggest that the tuyeres are
not generating bubbles independently.

4.5.1

Overall Flow Patterns

When a bed of particles is fluidized at a gas velocity above the minimum


bubbling point, bubbles form continuously and rise through the bed, which is
then said to be 'freely bubbling'. As they rise, bubbles interact and coalesce,
so that the average bubble size increases with distance above the distributor.
Growth by coalescence proceeds until the bubbles approach their maximum
stable size, when splitting becomes frequent. Thereafter, splitting and
recoalescence cause the average bubble size to equilibrate at a value close to
the maximum stable size. In group A particles, for which the maximum
stable diameter is relatively small and is therefore attained close to the
distributor, the average bubble size is therefore typically constant over much

of the depth of the bed. For group B particles, the maximum stable diameter
is larger and typically only attained in the upper levels of deep beds;
generally, the average bubble size then increases steadily with distance from
the distributor. Measurements of bubble size reported by Werther (1975) give
an excellent illustration of this difference between groups A and B. Group D
particles show a behaviour which is again differert (Cranfield and Geldart,
1974): bubbles rise as horizontally associated swaml~; and grow by absorption
of neighbouring bubbles rather than conventional coalescence,
In addition to controlling the vertical profile of bubble size, coalescence
also sets up horizontal non-uniformity in the bubble distribution in beds of
groups A and B. This effect is shown schematically in Fig. 4.9(a), with typical
. coalescence processes shown in Fig. 4.9(b). Bubbles generally coalesce by
overtaking a bubble in front, shown by (i) in Fig. 4.9(b), and may move
sideways into the track of such a bubble before coalescing, shown by (ii) in
Fig. 4.9(b). Thus coalescence can cause lateral motion of bubbles. Bubbles
formed close to the wall of the bed can only move inwards as a result of this
process, since the bubbles surrounding them are only on the side away from
the wall, whereas bubbles nearer to the axis have equal probability of
migrating in any horizontal direction. As a result of this preferential
movement of bubbles away from the wall, an 'active' zone of enhanced
bubble flowrate forms a small distance in from the wall. The higher
concentration of bubbles leads to more frequent coalescence and hence larger
bubbles in this zone. Since the region between the wall and the annular
'active' zone is lean in bubbles, coalescence continues to cause preferential
migration towards the bed axis. Thus the 'active' zone intensifies and moves
closer to the axis with increasing height above the distributor. At heights of
the order of the bed diameter, virtually the whole of the bubble flow occurs
close the ~ axis. This effect was first observed and explained by Grace and
Harrison (1968), but has since been confirmed by many independent studies.
Large beds essentially divide into cells, to form several bubble tracks rather
than a single axial track (Whitehead, 1971). Quantitative prediction of the
development of bubble profiles is possible (Farrokhalaee and Clift, 1980), but
generally requires extensive computer calculations. Qualitative recognition of
this effect helps to interpret the behaviour of freely bubbling beds. An
important consequence of the non-uniformity in bubble flow is gross solids
circulation (see Chapter 5). Generally, solids move up in regions of high
bubble activity and downwards elsewhere; the resulting circulation pattern
(observed, for example, by Masson, 1978) is shown schematically in Fig.
4.9(a). In the upper levels of the bed, particle motion is commonly described
as 'gulf-streaming', up on the axis (or bubble tracks) and downwards near the
walls. At lower levels, the particle motion is down near the axis. Although
coalescence sets up this kind of circulation even if gas enters the bed
uniformly across the distributor, the effect is amplified by the motion of the

dense phase, which reduces bubble formation near the axis and enhances it
near the walls (Werther, 1975; Farrokhalaee and Clift, 1980).
Because of the importance of the walls in initiating the flow pattern shown
in Fig. 4.9(a), the details of the bubble flow pattern depend on the scale of the
equipment as well as on the characteristics of the particles and distributor.
The resulting difficulty in predicting the behaviour of a large bed from tests
on a smaller pilot unit lies behind the problem of scale-up of fluidized bed
reactors discussed in Chapter 11. Moreover, since coalescence effects differ
quantitatively in two and three dimensions, one may not simply assume that a
bubble size distribution measured in a two dimensional bed is the same as that
existing in a three-dimensional bed of the same particles (see Chapter 13).

( a) Overall bubbling pattern


Solids
__
circulation

-..-._

4.5.2

Bubble Flowrate

The 'visible flowrate' in a fluidized bed Qb is defined as the rate at which


bubble volume crosses any level in the bed. For example, the visible bubble
flowrate across section AA' in Fig. 4.9(a) could, in principle, be determined
by noting the volume of each bubble crossing the section in some period of
time and dividing the sum of these volumes by the time of observation. It is
important to note that, because of the bulk 'throughfJ.ow' of gas into and out
of each bubble (see Section 4.1), the flow of gas in the bble phase is greater
than the visible bubble flowrate; the importance of this e ct for chemical
reactors is discussed in Chapter 11.
A first estimate for Qb is given by the 'two-phase theory of fluidization',
proposed by Toomey and Johnstone (1952) and developed by Davidson and
Harrison (1963). As defined by the latter authors, the theory models a
fluidized bed as consisting of:
(a) A particulate phase in which the flow rate is equal to the flowrate for
incipient fluidization, i.e. the voidage .i~essentially con~t~n~at Em!; and
(b) A bubble phase which carries the addItional flow of flUIdIZIngflUId.
This model is then commonly applied by estimating the visible bubble
flowrate as the excess gas flow above that required for minimum fluidization,
i.e.:
Qb

A=

1 t t t t
(i)

(iilO

U - Urnf

where A is the cross-sectional area of the bed.


In practice, Qb is rather smaller than the value given in Eq. (4:38): reported
values for the deficit are typically 10 per cent for group A partIcles, 20 to 30
per cent for group B, and 40 to 50 per cent for group D. The deficit must
result in part from the gas 'throughfloW"in bubbles, but this effect alone d~es
not account satisfactorily for the measured values of Qb (Grace and ClIft,
1974).The gas flow through the particulate phase appears to excee~ the v~l~e
implied by assumption (a) of the two-phase theory, but the extra mterstItIal
flow cannot at present be estimated. Simply replacing Urnf by the minimum
bubbling velocity Urnb does not markedly improve agreement with. exp~riment. Yacono, Rowe and Angelino (1979) have proposed an analysIs WhICh
accounts for both 'throughflow' and increased interstitial flow, but the result
is complex and its general applicability remains to be tested. Equation (4.38)
thus remains the best initial estimate for bubble flowrate, and has been the
basis for much work in fluidization. At present, there is no ready alternative
to using Eq. (4.38) and perhaps reducing the value for Qb obtained by an
empirical factor appropriate to the type of particle being fluidized (see
Chapter 5, Section 5.4, Eq. 5.2 and Fig. 5.7).

4.5.3

Bubble Size and Velocity

Correlations abound in the literature for the mean bubble size across
horizontal sections in freely bubbling beds. Some of the more recent and
more successful are summarized in Table 4.1. That due to Geldart (1972) is
based primarily on data for relatively small beds, and generally gives values
for frontal diameter (db) which are too high for large-scale beds; its value lies
primarily in estimating the bubble diameter formed at the distributor (dbo)'
The remaining three correlations generally give comparable estimates, which
is perhaps not surprising since they are derived by similar semiempirical
arguments and are based on much the same data. The value of deqm in the
correlation of Mori and Wen (1975) represents the maximum bubble size
attainable by coalescence, and is distinct from the maximum stable size
(deq)max discussed in Section 4.2.3. The correlation of Rowe (1976) contains a
parameter Zo characterizing the distributor: essentially, Zo is the height above
a porous distributor at which bubbles would have attained the size formed at
the distributor actually used. Rowe gives some typical values for zo, but no
general results from which it can be estimated. It is therefore necessary to
estimate dbo from, for example, the correlation of Geldart (1972) and then
estimate 2 as:
0

Zo =

4.0 g
(U -

umff

lIt3
f

Davidson and Harrison (1963) suggested that the average velocity of a


bubble in a freely bubbling bed is:

where Ub is the rise velocity of an isolated bubble of the same size (see Section
4.2). Equation (4.41) is developed by analogy with slug flow, and the
applicability to freely bubbling beds is questionabr, (see Section 4.6 and
Turner, 1966). In fact, UA > Ub, primarily as a consequ ce of the increase in
bubble velocity caused by coalescence, and there is some vidence (Farrokhalaee and Clift, 1980) that Eq. (4.41) accounts apprOXI ately for the
average effect of coalescence in a freely bubbling bed. Thus Eq. (4.41) can be
applied to a bed overall, although instantaneous bubble velocities vary widely
about this result and the mean velocity for a given bubble diameter varies
with position in the bed, being highest in the 'active' zones where coalescence
is frequent.
4.5.4 Bed Expansion
Consider a section such as AA' in Fig. 4.9(a), across which the visible bubble
flowrate is Qb and the average bubble velocity is Uk The average fraction of
the bed area occupied by bubbles is then:
Qb

Eb

= --

AUA

Werther (1975) has given a correlation for beds with porous distributors
which could perhaps be adapted to other types of distributors in the same
way.
There is no strong reason for preferring anyone of the correlations by
Darton et al., Mori and Wen, and Rowe; probably the safest approach is to
use all three and average their results. The correlations should be used up to
the heigbt at which either the bubbles attain their maximum stable diameter,
(deq)max, or become large enough to be treated as slugs (deq = O.6D). In the
former case, the bubble size should be assumed constant at higher levels; in
the latter case, the upper levels of the bed should be treated as slugging.
These results were derived for groups A and B. For group D particles,
Cranfield and Geldart (1974) give a correlation for estimating the frontal
diameters of bubbles:

in c.g.s. units, where Umb is the minimum gas velocity at which bubbles form.
Since bubbles do not originate at the distributor in group D particles, Eq.
(4.40) contains no term describing the distributor and should only be used for
h > 5 em.

where A is the cross-sectional area of the bed. The volume of particulate


phase in an elementary height dz at this section in a bed of height H is then:

AHo

= A- foH (1 -

dz

Eb)

=A

1-

Qb)

AUA

dz

(4.43)

from Eq. (4.42). Following assumption (a) of the two-phase theory (see
Section 4.5.2), if the particulate phase remains at the void fraction
corresponding to minimum fluidization, then AHo is equal to the total bed
volume at minimum fluidization, AHmf Equation (4.43) then becomes:

H-Hmf=

Qb

-A

UA

dz

which gives the expansion of the bed above its depth at minimum fluidization.
This general result can be simplified if Qt!A is approximated by the two-phase
theory (Eq. 4.38) and UA is estimated from Eq. (4.41):

H - Hrnf

(U - Urnf) dz

------

(U - Urnf)

(a) Estimation of bubble size


Initial bubble sizes are calculated in Example 4.3(b):
Darton et al: deq.o = 0.150m
Geldart:
dbo = 0.131 m (deq.o = 0.110 m)
Mori and Wen:deq.o = 0.126 m

+ Ub

In general, Ub is a function of z because bubble diameter is a function of z. For


detailed calculations of bed expansion, it is therefore advisable to evaluate
the integral in Eq. (4.45) analytically or numerically, using values of Ub
calculated from the profile of bubble diameters obtained from the correlations in Table 4.1. If Ub varies little or if a single mean bubble diameter and
rise velocity are to be used for the purpose of estimate, then Eq. (4.45)
simplifies to:
H- Hmf
H

For the correlation of Rowe, we characterize the distribut~~'t1~g the Geldart


value for dbO, using Eq. (4.39):
zo

0.1314 X 9.81 )113


0.782
(

0.168m

suggesting that the initial bubble size is equal to that present 0.168 m above a
porous distributor. The correlation of Mori and Wen also requires:
deq.m= 0.374 (7T x 6382
= 590 cm = 5.9 m

X'

77.5)4cm

This is much larger than the maximum stable bubble size, estimated as:
(deq)max= 2.06 m
Equation (4.47), derived by Davidson and Harrison (1963) for the case where
Ub is constant in the bed, is equivalent to a result for slug flow given in Section
4.6. In addition to design calculations of bed expansion, it is also frequently
used to infer mean values of Ub and hence deq or db from measurements of bed
expansion.

in a bed of 100 #Lmparticles (see Example 4.1).


Table 4.2 summarizes resulting values for mean bubble size, which are also
shown in Fig. 4.10. Clearly Geldart's correlation predicts unrealistically rapid
bubble growth, but the other three correlations are in broad agreement.
Reasonable estimates may therefore be obtained by averaging the three
correlations; resulting values are also shown in Table 4.2 and Fig. 4.10. It is
predicted that the maximum stable bubble size is not reached.

Example 4.4: Bubble size distribution and bed expansion

In the following example, bubble properties and mean bed density are
estimated for the fluidized bed roaster considered in Examples 4.1 to 4.3. The
bed in question, as described by Avedesian (1974), has diameter D = 6.38 m
and expanded depth 1 metre. The distributor is of the single-hole tuyere type,
designed in Example 4.2 (a), with one such tuyere per 1.03 x 10-3 m2 of bed
area. The superficial velocity at bed conditions is taken as 0.78 m1s (see
Example 4.3b), so that

(b)

Bubble velocity and bed expansion

Using the values of deq given in column 5 in Table 4.2, the isolated rise
velocity is estimated as in Example 4.1. The average rise velocity at any level
is then calculated from Eq. (4.41), and the average local bubble fraction is
then estimated from Eq. (4.42), in the simplified form derived from the twophase theory:

1. Darton et al.
z
0

2. Geldart

deq

deq

db

3. Mori and
Wen

deq

5. Average

4. Rowe
deq

db

deq
0.129

0.256

0.210

uA

0.578
0.738

1.35

0.573

1.51

0.512
0.474

0.25

0.150
0.220

0.131
0.534

0.110
0.450

0.126
0.193

0.5
0.75
1.0

0.285
0.347
0.406

0.937 0.790
1.340 1.130
1.740 1.470

0.260

0.367

0.310

0.285

0.86

0.326

0.467
0.559

0.393
0.471

0.355
0.422

0.96
1.046

0.391

(O.13)h (0.110)

ub

0.218

1.635
1.734
1.82

11

Eb

0'5
/J,4

(m/s)

1'5

Eb

~~ :::::-....

0.447
0.425

:::,.....
::::::,...

13
a The initial bubble size is taken to apply at the level of the distributor,

although it could be
argued that these values apply at the ends of the distributor jets (see Example 4.3a).
b Values for use in the Rowe correlation deduced from Geldart's correlation.

Resulting values are given in Table 4.2 and Fig. 4.10. The local average
bubble velocity increases, and the average local bubble fraction decreases
with height above the distributor.
Bed expansion is estimated by applying Eq. (4.45) to estimate Hmf In
terms of b, Eq. (4.45) is written:

0'4

0'5

0.573 - 0.212z

(z s 0.4)

0.53 - 0.105z

(z ~ 0.4)

(0.573 - 0.212z) dz

= 0.53H

- 0.052H2

...-';"---J
.--//

= f~.4

/~

/~~.
~::.--

From the values in Table 4.2 and Fig. 4.10, E'b can conveniently be
approximated by a piecewise linear form shown by the broken lines in Fig.
4.10:

H - Hmf

/~

03

H-Hmf=f::E'bdz

4-

deq
(m)

J:

0'1
z(m)

1-0

05

Figure 4.10 Bubble sizes, rise velocities, and volume fractions as a function
of level in bed.

The corresponding

average bubble fraction in the bed is:

4 (0.53 - 0.105z)dz
E'b

+ 0.008

H - Hmf
H

0.486

The value of Hmf corresponds to the equivalent depth of particulate phase at


minimum fluidization conditions. It therefore indicates the weight of particles
per unit cross-sectional area supported by the fluidizing gas, i.e. the bed
pressure drop. Using the above value, the bed pressure drop is estimated as:
Pp (1 -

E'mf)

g Hmf

=
=

5,300 x 0.5 x 9.81 x 0.514


1.34 x 104 N/m2

where the void fraction in the particulate phase, Emf, is assumed to be 0.5.
This estimate is of the right order, but is rather larger than the value of 1.01 x
104 N/m2 estimated by Avedesian (see Example 4.2). Several factors could
contribute to the discrepancy:
(i) The nominal expanded depth of 1.0 m corresponds to the distance from
the distributor to the solids overflow chute. Because of 'splashing' of
particles by erupting bubbles, the mean bed surface will be some
distance below the overflow chute.
(ii) The mean bubble size in the bed is determined from equations based on
data relating to ambient air. If the bubbles are actually smaller, as seems
likely at high temperature, then Ub and UA will be lower, Eb will be
higher, and the bed pressure drop will be lower than the estimated value.
(iii) Equation (4.41), used to estimate bubble rise velocity in a freely
bubbling bed, is known to be an approximation (see Section 4.5.3) and is
not well tested for operation at high gas velocity, as in the present
example.

with c.g.s. units used. The second term on the right is omitted if Hrnf >
60Do.175, and Eq. (4.48) then becomes identical with a result derived for deep
beds by Stewart (see Hovmand and Davidson, 1971).

(a) The maximum stable bubble size, (deq)max, is greater than 0.6D, where D
is the tube diameter.
(b) The gas velocity is sufficiently high.
(c) The bed is sufficiently deep.
Condition (a) depends on the properties of gas and particles (see Section
4.2.3). Conditions (b) and (c) are combined in a criterion for slug flow
developed by Baeyens and Geldart (1974). The minimum gas superficial
velocity for slugging, Urns, is related to the velocity Urnf and bed depth Hrnf at
minimum fluidization by:
Urns

Urnf

+ 1.6

10-3 (60Do.175

Hmf)2

+ 0.07

V(gD)

(4.48)

I
I
I

I
I
I

I
I
I

I
I

I
I

I
_--.J

L_

ttt
Figure 4.11

A-

I
I

I
I

Thus the present example serves to illustrate the limitations as well as the
application of currently available results.

Slugging fluidization can occur in beds of a small diameter and is therefore


likely to be encountered in laboratory and pilot-scale units. Figure 4.11 shows
a schematic section through a slugging bed. Bubbles formed at the distributor
grow by coalescence until they reach the size of slugs (deq > 0.6D). Above
that height, the bed is filled by a succession of slugs, and the bed surface
oscillates widely. Each time a slug breaks through the surface, the level drops
sharply, then rises steadily again until the next slug breaks through.
A fluidized bed will show slug flow if three conditions are met:

-1

I"""-

Bubble and slug growth.

The simple two-phase theory (see Section 4.5.2) generally appears to be a


better approximation for slugging beds than for freely bubbling beds.
Therefore the average 'visible flowrate' of slugs can be estimated as A( U Urnf), and the void fraction in the dense phase can be assumed to be constant
at Emf' Consider now the control volume bounded by AA'B'B in Fig. 4.11,
where the section AA' is deliberately chosen to pass through the dense, or
particulate, phase between slugs. By continuity, the volume flowrate of the
dense phase across AA', Qd, must be equal to the rate at which the fluidizing
gas forms slugs within AA'B'B:

Since the particulate phase moves essentially as a plug (another. respect in


which fluidized particles differ from Newtonian liquids), the particulate phase
velocity across AA' is QdIA, that is (U - Urnf). The slug approaching AA'
moves with its isolated velocity, UsI, relative to these particles so that its
absolute rise velocity is
UsA

(U - Urnf) +

(U - Umf)

Usl

+ 0.35

V(gD)

(4.50)
(4.51)

using the expression for Usl for an ideal slug from Eq. (4.15). Equation (4.50)
has been carried over by analogy to freely bubbling beds (see Davidson and
Harrison, 1963) and is the origin of Eq. (4.41). The derivation of Eq. (4.50)
shows one of the reasons why the theoretical basis of Eq. (4.41) is uncertain:
in a freely bubbling bed, it is not normally possible to define a section like
AA' which passes through no bubbles and across which the particulate phase
velocity is constant.
Matsen, Hovmand, and Davidson (1969) showed that the maximum height
of a slugging bed, Hmax, is related to its depth at minimum fluidization, Hmf,
by:

and the bed depth corresponding to minimum fluidization conditions is taken


as the value calculated in Example 4.4, that is:
Hmf

0.514 m

51.4 cm

(a) Flow regime


At such high fluidizing velocity in a column of diameter less than (deq)max,
slug flow can clearly be anticipated. Applying the criterion developed by
Baeyens and Geldart, Eq. (4.48), slug flow will occur for gas velocities in
excess of:
Ums

=
=
=

0.5 + 1.6 x 10-3 (60 X 20.175


0.5 + 4.0 + 9.3 = 14.~ cmls
0.143 m/s
\

49.6? + 0.07 V(981 x 20)

F.or a very deep bed, the simple f~


of Eq. (4.48) developed by Stewart
gIves:
~
Ums = 0.5 + 0.07 V(981 x 20) = 10.3 cmls

Hmax

Hmf

U - Urnf

Hmf

Us!

which is analogous to Eq. (4.47) for freely bubbling beds. In order to design
the depth of vessel needed for a bed in which slug flow is anticipated, the
overall maximum bed depth is required. Equation (4.52) should therefore be
used with the minimum value for UsI, i.e. the value for an ideal slug (Eq.
4.15), giving
U - Urnf
0.35 V(gD)

If bubbles require a significant distance above the distributor to grow to slugs,


it is best to treat the lowermost region as a bubbling bed, with Eq. (4.55)
applied above the level at which deq reaches 0.6D.
Example 4.5:

Slug flow in a pilot-scale bed

0.78 mls

= 78 cm/s

0.103 mls

Since the operating velocity is so much larger than Um" it is likely that slug
flow will obtain throughout most of the depth of the bed. From Section 4.2.4,
flow may be considered slugging if: .
deq

> 0.6D

= 0.12 m in the present case


The bed depth required to attain this bubble size depends on the distributor
used. For the tuyere distributor considered in Example 4.2(a), the initial
bubble size was estimated in Example 4.4 as 0.129 m, so that slug flow could
be assumed to occur throughout the bed; this seems intuitively correct, since
the required tuyere spacing would give the bed only three nozzles. For a
porous distributor, the initial bubble size would be small, so that a finite bed
depth would be required for the bubbles to attain deq = 0.12 m. The
correlations in Table 4.1 can be used to give a rough estimate for this depth.
Using the correlation of Darton et al., with Ao = 0 for a porous distributor:
deq

In the following example, we consider fluidization of the particles considered


in Examples 4.1 to 4.4 in a column of diameter 0.2 m (that is 20 cm). For
purposes of illustration, the gas velocity is taken as that in Example 4.4:
U

= 0.54 X 0.775.4 x 9.81-0.2


= 0.309zo.8

0.31 m

when deq

X ZO.8

0.12 m

This estimate is likely to be generous, because the constraining effect of the


walls causes more coalescence and hence more rapid bubble growth than in a
large bubbling bed.
.

HYDRODYNAMICS

(b) Bed expansion

Taking slug flow to occur throughout


predict the maximum depth:

Hrnax

= Hrnf
=

the bed, Eq. (4.52) may be applied to

( 1+

U - Urnf)
Us/

0.775 )
0.514 ( 1 + 0.490

where the rise velocity of a single slug, us/, is taken as 0.490 mis, the value
calculated in Example 4.1. From the above:

Hrnax

1.33 m

H
Hd
Hmf
Hmax
Lj
N
Qb
Qor

r
Reb
ReI

s
cross-sectional
A

Ao
Cd

Co
db
db,o
deq

Eo
fb.o
for

fw

area of one hole or orifice

cross-sectional area of bed


area of distributor per orifice
discharge coefficient of orifice
drag coefficient
frontal diameter of bubble
initial frontal diameter of bubble
equivalent volume diameter of bubble
(= (6VJ7T)I/3)
maximum equivalent bubble size attainable by
coalescence
initial equivalent bubble size
diameter of hole or orifice
mean sieve size of particles (= VI (X#pi
average of adjacent sieve apertures
particle size used to calculate VI in Eq. (4.12)
(= 2.7dp)
diameter of fluidizing column
Eotvos number (= PI cPeq glu)
initial bubble frequency
fraction of distributor open to gas (= Naor)
ratio of bubble wake volume to sphere volume
(= 6Vw/7Tdl,)

UA
Ub
Usl
UsA
U
Uin
Urnb, Urnf
Urnfw
Urns
Uor
UR
Uw

OF BUBBLING FLUIDIZED

93

BEDS

acceleration of gravity
height of fluidized bed
height of dense phase in a bubbling bed
height of bed at minimum fluidization velocity
maximum bed height
length of jet
number of orifices per unit cross-sectional area
visible bubble f10wrate
volumetric gas f10wrate per hole
bubble radius
bubble Reynolds number (Eq. 4.3)

9.81 mls
m
m
m
m
m
m-2
m3/s
m3/s
m

particle Reyn.o~ o~mbe'.at tenninal velocity


hole spacmg m a dnlled dIstributor
thickness of dis butor plate
rise velocity of a bubble in a bubbling bed
rise velocity of an isolated bubble
rise velocity of an isolated slug
rise velocity of a slug in a slugging bed
superficial velocity of gas
mean gas velocity in duct entering windbox
minimum bubbling and fluidization velocities,
respectively
minimum fluidization velocity related to windbox
minimum slugging velocity conditions
velocity of gas in hole in distributor
velocity of gas in riser of tuyere
superficial velocity of gas in windbox
particle terminal or free-fall velocity

m
m
m/s
m/s
m/s
m/s
m/s
m/s
mls
mls
m/s
mls
mls
mls

Vsph
Vw
z

bubble volume
initial bubble volume
volume of bubble and wake
volume of wake
distance above distributor

mls
m3
m3
m3
m3
mls

Zo

parameter

[3w

volume of solids in wake as a fraction of bubble


volume

VI

Vb

Vo

in correlation

of Rowe (1976)

94
tlPB, tlPD
Eb
Emf

8w
IL
ILl

Pg
Pgw
PI

Pp
u

GAS FLUIDIZATION

TECHNOLOGY

pressure drop across bed and distributor,


respectively
fraction of bed occupied by bubbles
voidage of bed at minimum fluidization
wake angle
viscosity of gas
viscosity of liquid
gas density
gas density in windbox
liquid density ,
particle density
interfacial tension

N/m

N s/m2
N s/m2
kg/m3
kg/m3
kg/m3
kg/m3
N/m

Doganoglu, Y., Jog, V.,Thambimuthu, K.V., and Clift, R. (1978). Trans. lnstn
Chern. Engrs, 56,239.
Donsi, G., Massimilla, L., and Colantuoni, L. (1980). 3rd Eng. Found. Conf. on
FI';ti~ization(Eds J ..M. Matsen and J.R. Grace), Plenum Press, p.297.
Fakhlml, S., and Hamso~, D. (1970). Cherneca '70, Instn Chern. Engrs, p.29.
Farrokhalaee, T., and ClIft R. (1980). 3rd Eng. Found. Conf. on Fluidization (Eds
J.M. Matsen and J.R. Grace), Plenum Press.
Geldart, D. (1972). Powder Technol., 6,201.
Geldart, D. (1977). 'Gas Fluidization', Short Course. University of Bradford, p.114.
Geldart, D., Hurt, J.M., and Wadia, P.W. (1978). A.I.Ch.E. Svmp. Ser., 74 (176)
60.

.,.

Ghadiri, M., and Clift, R. (1980). Ind. Eng. Chem. Fund., 19,440.
Grace, J.R. (1970). Can. J. Chern. Eng., 48,30.
Grace, J.R., and Clift, R. (1974). Chem. Eng. Sci., 29,327.
Grace, J.R., and Harrison, D. (1968). Instn Chem. Engrs Syrnp. Ser., 30, 105.
Hanway, J.E. (1970). Chern. Eng. Prog. Syrnp. Ser., 66 (105),253.
Hetzler, R., and Williams, M.C. (1969). Ind. Eng. Chern. Fund., 8, 668.
Hovma~d, S., and Davi?son, J.F. (1971). Chapter 5 in Fluidization (Eds J.F.
DaVIdson and D. Hamson), Academic Press.
Kunii, D., and Levenspiel, O. (1969). Fluidization Engineering, John Wiley and Sons
Inc.

Agarwal, J.C., Davis, W.L., and King, D.T. (1962). Chern. Eng. Prog., 58 (Nov.), 85.
Avedesian, M.M. (1974). Paper 5d, 24th Annual Conference of C.S.Ch.E.
Baeyens, J., and Geldart, D. (1974). Chern. Eng. Sci., 29, 255.
Clift, R., and Grace, J.R. (1972). Chern. Eng. Sci., 27,2309.
Clift, R., Grace, J.R., and Weber, M.E. (1974). Ind. Eng. Chern. Fund., 13,45.
Clift, R., Grace, J.R., and Weber, M.E. (1978). Bubbles, Drops and Particles,
Academic Press.
Cranfield, R.R., and Geldart, D. (1974) Chern. Eng. Sci., 29,935.
Darton, R.C., LaNauze, R.D., Davidson, LF., and Harrison, D. (1977). Trans. Instn
Chern. Engrs, 55,274.
Davidson, J.F., and Harrison, D. (1963). Fluidised Particles, Cambridge University
Press.
Davidson, J.F., Harrison, D., Darton, R.C., and LaNauze, R.D. (1977). Chapter 10
in Chemical Reactor Theory, A Review (Eds L. Lapidus and N.R. Amundson),
Prentice Hall.
Davidson, J.F., Harrison, D., and Guedes de Carvalho, J.R.F. (1977). Ann. Rev.
Fluid Mech., 9, 55.
Davies, R.M., and Taylor, G.!. (1950). Proc. Roy. Soc., Ser A, 200,375. ,k

Masson, H. (1978). hern. Eng. Sci., 33, 621.


Matsen, J.M. (1973). . .Ch.E. Syrnp. Ser., 69 (128), 30.
Matsen, J.M., Hovmand, ., and Davidson, J.F. (1969). Chern. Eng. Sci., 24, 1743.
Merry, J.M.D. (1971). Trans. Instn Chem. Engrs, 49, 189.
Merry, J.M.D. (1975). A.I.Ch.E.J., 21,507.
Mori, S., and Wen, C.Y. (1975). A.I.Ch.E.J., 21, 109.
Perry, R.H., ~d Chilton, C.H. (1973). Chernical Engineers' Handbook, 5th ed.,
McGraw-HIli, Kogakusha.
Qureshi, A.E., and Creasy, D.E. (1979). Powder Techno/., 22, 113.
Rippin, D.W.T., and Davidson, J.F. (1967). Chern. Eng. Sci., 22,217.
Rowe, P.N. (1971). Chapter 4 in Fluidization (Eds J.F. Davidson and D. Harrison)
Academic Press.
'
Rowe, P.N., (1976). Chern. Eng. Sci., 31,285.
Rowe, P.N., and Everett, D.J. (1972). Trans. Instn Chern. Engrs, 50,55.
Rowe,P.N., MacGillivray, H.J., and Cheesman, D.J. (1979). Trans. lnstn Chern.
Engrs, 57, 194.
Rowe, P.N., and Pa.rtridge, B.A. (1965). Trans. lnstn Chern. Engrs, 43, T157.
Rowe, P.N., and WIdmer, A.J. (1973). Chern. Eng. Sci., 28,980.
Rowe, P.N., and Yacono, C.X.R. (1976). Chern. Eng. Sci., 31, 1179.
Schiigerl, K. (1971). Chapter 6 in Fluidization (Eds J.F. Davidson and D. Harrison).,
Academic Press.
Toomey, R.D., and Johnstone, H.F. (1952). Chern. Eng. Prog., 48,220.
Turner, J.C.R. (1966). Chern. Eng. Sci., 21,471.
Wallis, G.B. (1969). One-Dirnensional Two-Phase Flow, McGraw-Hili.
Werthe~,~. (1975). Fluidization Technology (Ed. D.L. Keairns), Vol. 1, Hemisphere
PublIshmg Co., p. 215.
Whitehead, A.B. (1971). Chapter 19 in Fluidization (Eds J.F. Davidson and 0
Harrison), Academic Press.
.
Yacono, C.X.R., Rowe, P.N., and Angelino, H. (1979). Chern. Eng. Sci., 34, 789.
Yang, W.-c., and Keaims, D.L. (1979). Ind. Eng. Chern. Fund., 18,317.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Solids Mixing

5.1

1NTRODUITlON

part~miXing is' of importance in the design of both batch and continuous


fluidized beds. For example, the axial and radial transport of solids within the
bed can influence:
(a) Gas-solid contacting
(b) Thermal gradients between a reaction zone and the zone in which heat
transfer surfaces are located
(c) Heat transfer coefficients
(d) The position and number of solids feed and withdrawal points
(e) The presence and extent of dead zones at distributor level.
Problems of mixing in fluid beds fall into four categories:
1.
2.
3.
4.

Particles have uniform size and density.


Particles have uniform size but variable density.
Particles are all of same density but vary in size.
Particles vary in both size and density, e.g. small/heavy - large/light;
small/light - large/heavy.

In category 1 the main problem, generally, is to ensure that the overall solids
transport or circulation flux is high enough to eliminate temperature
gradients; in categories 2,3, and 4 an additional concern is whether the local
composition of the powder - however defined - is everywhere equal to the
overall average. The amount of solids circulation and degree of local mixing
or segregation is primarily determined by the gas velocity, but particle shape,
97

size, density, stickiness, and size distribution all playa part. Bed and column
geometry can also be important.

Although mechanisms of mixing are now better understood, there is still


much to be learnt and experimentation is essential.
5.2.1

Experimental Methods

A selection of the most frequently used methods is listed in Table 13.5 of


Chapter 13. None is entirely satisfactory; they either make use of indirect
measurements (e.g. heat transfer) from which circulation has to be inferred or
involve the use of tracers in batch experiments or in artificial systems such as
two-dimensional beds. There is a great need for a reliable on-line technique
so that mixing can be monitored continuously. 'Freezing' the bed is the most
tedious but simplest technique and is often used. In one version a layer of
tracer is positioned in the bed, the gas flow is turned on for a given time, then
suddenly shut off, and the bed removed a section at a time. In another, a
tracer is released into an already fluidized bed and the bed shut down after a
specified elapsed time.
Where the process demands that solids are fed and removed continuously it
is often important to know the residence time distribution (RID); experimental techniques to determine RID are generally more straightforward and
depend on taking continuous samples of magnetic, chemically active, or
radioactive tracers.
5.2.2

--~

original tracer level

"-

//

Bubble-Induced Particle Mixing and Circulation

It is well known that vertical,mixing in fluidized beds is many times faster than
lateral mixing. Indeed it seems likely that solids move in a net horizontal
direction largely by being (a) carried up to the surface where they are
dispersed sideways by bursting bubbles and (b) carried down to, and across,
the distributor by bubble-free flows of dense phase material.
Much of the early work on mixing mechanisms was done by Rowe and
Partridge (1962, 1965). Layers of coloured and colourless particles were
disturbed by injected bubbles and photographed in two-dimensional beds:
These studies showed that in an idealized case, particles are carried upwards
in the wake of the bubble (defined as the solids occupying the bottom of the
completed sphere) and in the drift (defined as the region behind the
completed sphere) (see Figs 5.1 and 5.2). Particles appear to be pulled into
the wake or drift, carried up the bed for a distance, and then shed. Thus,
particles travel upwards where there are bubbles and, by continuity,
downwards where there are not.

_~,:t
...

.. ~r
..

;. ~ ',!

~L
"

"o.,'t'"

:'~"'\

""-:f":

Mfl~dl ~jltil

l.~

Colourless solid
(b) Coloured solid
(0)

Figure 5.2 Displacement and mixing of solids by a rising bubble


(Rowe and Partridge, 1962).

In freely bubbling beds non-uniform bubble distribution patterns become


established because of the directional probability of bubble coalescence
(Geldart and Kelsey, 1968; Grace and Harrison, 1969;Whitehead, Gartside,
and Dent, 1976). Areas near the wall are almost bubble free. Bubbles which
are initiated close to the wall at the distributor are attracted towards other
bubbles near the centre but can themselves attract only a limited number; the
probability of a bubble moving into the centre of the bed is therefore high,
and of a bubble moving towards the wall, low. As a consequence of there
being fewer bubbles close to the wall, there is a predominantly downward
flow of particles close to the wall. Once established, this return flow of solids
tends to maintain the tendency for bubbles to move inwards from the walls
(Rowe and Everett, 1972). Both these factors are largely responsible for the
characteristic bubble flow profile found in studies of bubble distribution
(Werther and Molerus, 1973; Whitehead, Gartside, and Dent, 1977;
Whitehead and Young, 1967; (see also Fig. 4.7 in Chapter 4).
The apparently random movement of individual particles within a fluidized
bed, as shown in tracks of radioactive tracer particles, led a number of
authors (May, 1959; Lewis, Gililand and Girouard. 1961/62; Kondukov,
1964;'de Groot, 1967;Miyauchi, Kaji, and Saito, 1968)to express mixingdata
in terms of a radial and axial dispersion (diffusion) coefficient obtained by
fitting experimental results to the diffusion differential equation under
specific boundary conditions.
However, due largely to the wide scatter in experimental values and the
lack of correlation in terms of the bubbling phenomenon, this approach has
been abandoned for some time.
Having noticed that most of the earlier work was carried out in the slugging
mode (which is anyway easier to characterize than the freely bubbling mode),
Thiel and others (Thiel and Potter, 1978) fitted the results of th~ir own and
earlier work in terms of a diffusive mixing model for slugging beds. Although
the model relates the diffusion coefficient to actual slugging parameters and
achieves a reasonable agreement, its use is limited since slugging beds are of
little or no industrial importance except in pilot plant work.

5.3

VARIABLES

WHICH INFLUENCE SOUDS CIRCULATION

Whitehead and others (Whitehead, Gartside, and Dent, 1976, 1977;


Whitehead and Dent, 1978) studied the circulation patterns of solids in large
fluidized beds and reported them in a series of papers. The solids used were
largely in Geldart's group B. Their results and those of other workers are
summarized below:
(a) In a bed 1.22 m square by about 1.5 m deep, sand was fluidized at
velocities up to 0.3 mls (see Fig. 5.3). At low velocities bubble-rich

Bubble-lean

regions

00

00

00
00

Figure 5.3 Effect of gas velocity on circulation patterns in a 1.22 x 1.22 m


bed of group B sand (Whitehead et ai" 1977).

areas occurret\, near the corners leaving bubble-free regions close to


the walls and in the centre. Solids were carried upwards with bubbles
and moved downwards elsewhere. At higher velocities the bubble-rich
regions increased in size at the expense of the bubble-free regions; at
still higher velocities the pattern changed entirely to upflow in the
centre and downflow at the walls.
(b) In shallow beds (HID < 0.5), at most velocities the predominant flow
pattern is as shown in Fig. 5A(a) with the central solids downflow
deflecting bubbles towards the walls. In deep beds (HID> 1) the flow
pattern in the distributor region resembles that in a shallow bed but
changes into a central upflow-wall downflow pattern higher up (Fig.
5Ab). The relative size of the regions is a function of gas velocity, with
the 'shallow bed' zone becoming smaller as U - Umf increases.
(c) At equal excess gas velocities, the flow patterns and solids downflow
velocities are similar in different sizes of bed having the same aspect
ratio (HID). This has been explained by Geldart (1980)as follows: The
ratio of bubble size at the surface divided by the bed diameter
(deq,/D) is approximately the same for different bed sizes at equal

(d) The solids flow is not easily changed from its 'natural' circulation
pattern, though gross deliberate 'mal-distribution' at the distributor
can be effective in shallow beds (Fig. 5.4c).
(e) The presence of tubes or baffles modifies the flow patterns shown:
with vertical tubes radial mixing is reduced slightly; with horizontal
tubes, depending on the fraction of the cross-section occupied, vertical
mixing can be inhibited to the point where the bed effectively becomes
staged.
(f) Flow patterns can be time-dependent and cyclic, with changes
occurring over several minutes.
(g) In group A powders bubble sizes and velocities are limited by the
properties of the powder and gas (see Chapter 4, Section 4.2.3). Also,
because the dense phase has a low effective viscosity, circulation
velocities are similar to or higher than the isolated bubble velocities,
and the absolute bubble velocities, UA, turn out to be relatively
independent of bubble size.

'"
I

t
I

~L1J
~

t
I

o t

.~ )1

t:J.

Figure 5.4 Solids circulation patterns in beds having different aspect ratios and distributors at
approximately equal velocities (after Whitehead and Dent, 1982).

values of HID and U - Urnf This is likely to give rise to similar


circulation patterns because solids are transferred laterally from
bubble-rich regions to the downward-moving dense phase largely by
bubbles bursting at the surface and scattering the solids sideways (Fig.
5.5).

t
~

5.4.1

The Parameters

is evident from the foregoing that the mixing rate (J) ought to be defined in
terms of the volumetric flowrate of bubbles through the bed, the amount of
terial each bubble drags up with it, the velocity at which the solids rise, and
the fraction of the bed consisting of bubbles. Thus:

Qb

UA

Eb

visible bubble flowrate


rise velocity of bubbles relative to the wall
the fraction of bed consisting of bubbles

Parameter f3

Figure 5.5 Beds containing the same powder, operated at equal


values of gas velocity and of similar aspect ratio, have similar
circulation patterns.

Rowe and Partridge (1962) showed that the amount of solids travelling in the
the bubble wake Vw, expressed as a fraction of the bubble volume Vh,
decreases with increasing particle size. Baeyens and Geldart (1973) studied
the fraction of the solids in the drift f3d, defined as volume of the drift divided
by bubble volume, and found that this also decreases with increasing particle

size. Their data are shown in Table 5.1 and expressed, somewhat speculatively, as functions of the Archimedes number in Fig. 5.6. For most group B
powders /3d is larger than f3w which averages about 0.35. For group D and

Parameter Qb
The quantity of gas appearing as bubbles should be, according to the twophase theory, equal to (U - Umf)A (see Section 4.5.2). It is in general less
than this, and Eq. (4.38) can be rewritten:

Table 5.1 Wake and drift fractions


(Baeyens and Geldart, 1973)

Powder
Catalyst
Angular sand
Rounded

sand

47
252
470
106
195

Qb

A=

f3w

{3d

0.43

1.00
0.42
0.28
0.70
0.52

0.50
0.25
0.82
0.65

0.26
0.20
0.32
0.30

1.00

where Y decreases with increasing particle size. Baeyens (1981) has


correlated Y speculatively with the Archimedes number in Fig. 5.7.
y

~-

c---

....

0.5
0.4

03

group A solids, f3w may be, respectively, 0.1 and 1. Thus, a bubble of a given
volume produces a much greater degree of particle movement in, for
example, a bed of cracking catalyst than in a bed of coal ash.

Y(U - Urnf)

02
0.1
103

102
Refs.

-----Parameter

Geldart. D. (l96B)

Werlher.J.

(1975)

.., Geldart.

D. (1970)

Werther.J.

and Molerus. D. (1972)

X Geldarl.

D. and

Grace. J.R. and Harrison.

Everell.

D. (1969)

Chavarie. C. (1973)

Baeyens.J.

and

Cranlield,R.R.
Geldart.

(1972)

D. (1981)

D.J. (1970)

Eb

The fraction of the bed consisting of bubbles can be obtained by measuring


the bed expansion so that:

om
1

H - Hrnf

H
Alternatively, it can be predicted approximately from Eqs (4.46) and (5.2):
Parameter

UA

Analysis of movie films of particles in two-dimensional beds (Baeyens and


Geldart, 1973) showed that the particles in the wake travel at a velocity
similar to that of the rising bubble while particles carried up on the drift travel
on average at about 38 per cent. of the bubble velocity, UA, a figure later
confirmed by Ohki and Shirai (1976). Equations (4.41) and (4.9) or (4.10)
may be used to find UA'

Eb

QJA

Y(U - Urnf)

= --;;;:: =

UA

Y(U - Urnf)

U - Urnf

+ Ub

(5.4)

5.4.2 Particle Velocities


Assuming that solids move upwards in the wake and drift of bubbles, and

downwards elsewhere, a mass balance in any horizontal plane cutting the bed
gives:

=(

~~a~:~O:o~~sbed)
x(~~~~~:rd~
move down
particle
velocity
1 -

substituting

Eb -

for
V

Eb

p -

(f3w

+ f3d)

Eb

~:~t~~e~:
.) x (~JI~~rds)
sohdsmove up
velOCIty

velocity range covered. Local values of the particle velocity may be several
times smaller or larger than the average (Nguyen, Whitehead, and Potter,
1978).

(5.5)
5.4.3

vp

UA f3w Eb

+ 0.38

UA f3d Eb

(5.6)

Equations (5.5) and (5.6) can be reformulated


solids circulation rate:

to give an expression for the

from Eq. (5.4) and rearranging:


{

---------

1-

f3w
Eb -

+ 0.38 f3d
}
)
Y (U - Urnf
Eb (f3w + f3d)

( ~~~Iation) =(~~~~ity)
rate
of dense
phase

The term within the first bracket is largely dependent on the particle
properties
and decreases with increasing particle size. Experimentally
measured values of vp (Baeyens and Geldart, 1973) are shown in Fig. 5.8 and
seen to be straight lines; separate measurements of f3 and Y, and values of
Eb calculated
from Eq. (5.4), gave good agreement with the slopes over the

x(

x (~:o~s)
sectIOnal
area
(5.1S)

(em/s)

~~~c~~~
~~ere) x(~e;~rdS)
solIdsmove
solids
upwards
velocity

and the solids flux:

Pp (1 -

As before, substituting for


Vp

Circulation Flux

Emf) (UAf3wEb

+ 0.38

UAf3dEb)

Eb:

Pp (1 -

Emf)

(U - Urnf) Y (f3w

+ 0.38f3d)

(5.10)

This can be compared with the empirical equation of Talmor and Benenati
(1963) obtained in a 0.1 m diameter bed with ion exchange resin particles of
67 to 660 JLm:
J (kg/m2s) = 785 (U - Urnf)e-6,630dp

\
10

The two equations agree reasonably well for group B particles but Eq. (5.11)
increasingly underestimates J for larger group D materials. In Fig 5.9
experimental values from the literature are compared with predicted values
(Eq. 5.10) and reasonable agreement is obtained.
At equal values of excess gas velocity, U - Urnf, the circulation flux
decreases as particle size increases; however, the circulation fluxes in beds of
large particles can be made large by increasing U - Urnf

12

5.4.4

Bed Turnover Time

U - Umf (em Is)


cracking catalyst
47 }Am
fine malachite sand
106 pm
sauthpart sand
195 JoArn
medium malachite sand 252 I'm
caarse malachite sand 470 )-1m

Figure 5.8 Particle velocities at bed wall as a function of excess gas flow
(Baeyens and Geldart, 1973).

(5.11)

The time required to turn the bed over once is:


tT

mass of powder in the bed M


JA

Substituting for M, and for J from Eq. (5.10):

Hrnf
t

(f3w

+ 0.38f3d) Y (U - Urnf)

Whitehead, and Potter (1978) injected CO2 gas near the bed surface and
measured the concentration at 49 positions below the injection point. They
used two sands having Urnf values of 2.5 and 10.7 cm/s and found that the CO2
concentration profiles reflected strongly the downflow patterns of the solids
at various superficial fluidizing velocities (see Section 5.3 (a)). They maintain
that for non-absorbing particles, backmixing commences when the ratio
U/Urnf is about 3, and this reflects the fact that, for a given value of U - Urnf,
both solids mixing and gas backmixing decreases as particle size increases.
However, in view of the wide range of values for Y, f3w, and f3d as particle
size changes, it would be surprising if the critical ratio at which backmixing
commences remained at 3 for all solids.
On the basis of Eq. (5.7) and the criterion that gas backmixing becomes
significant when vp > Urnfi, there are grounds for believing that for group A
and group D solids, the critical values of U/Urnf may be smaller and larger,
respectively.
Gas backmixing is influenced considerably by the degree to which the gas is
absorbed on the particles (Nguyen and Potter, 1974), and Bohle and van
Swaaij (1978) showed that the mean residence of a strongly adsorbing
gas (freon-12) injected into the bed can be a factor of ten longer than that for
a weakly absorbing gas (helium). Thus the critical velocity ratio at which gas
backmixing commences decreases as gas adsorption on the particle increases.

102
J pred
(kg/m2s)

Whitehead

Cranfield

- Geldart

t.

Werther

(1976)

Figure 5.9: Comparison

et al

(1979)

of experimental

(1978)

values of mixing flux with Eq~

For a bed to which solids are continuously added and from which reacted
products are removed (e.g. fluidized bed limestone calcination), the turnover
time should be short compared with the solids residence time: i.e. good
dispersion requires that the feed material should experience as many
circulations as possible before discharge. Little research has been done on this
but a ratio of residence time/turnover time of 5 to 10 seems reasonable.

Many applications of fluidized beds involve the continuous processing of


solids; e.g. the feed may be wet solids and the product dry, or the feed
limestone and the product lime. In a plug flow system all particles have the
same residence time tR, while in a fully mixed system the solids have a wide
range of residence times; this gives a non-uniform solid product and is
inefficient for the high conversion of solids. Consider a quantity of tracer
particles m kg introduced instantaneously at time zero into a bed containing
M kg solids when the overall feed rate is F kg/so If the bed is in plug flow, no
tracer should appear at the solids offtake until the average residence time tR
has elapsed, where:

If the bed is fully mixed, tracer will appear immediately in the offtake at a
concentration Co = m/M and the concentration in the exit, C, will then fall
according to:
This is a function of the downwards velocity of the dense phase, vp' If vp
exceeds the interstitital gas velocity, UrnriEmt. then gas will be carried
downwards with the solids. Working in a 1.22 m square bed, Nguyen,

-=

Co

e -tit

110
"->

.,

__ m kg tracer
-----

Co

1,,->0

M kg solids in bed

"

-'"

F kg/s

'0

c:
c:

.=..

.~

'"

g'"

'0

Iii ..0

~
c:

E'E
'"
c:

A bed may be 'well fluidized' in the sense that all the particles are fully
supported by the gas may still be segregated in the sense that the local
composition does not correspond with the overall average. Defluidization can
be regarded as a special case of segregation in that some or all of the bed,
usually that near the distributor, may be immobile. Segregation is likely to
occur when there is a substantial difference in drag/unit weight between
different particles. Particles having high drag/unit weight migrate to the
surface while those with low drag/unit weight migrate to the distributor.
When fluidized, the upper part of the bed will attain a fairly uniform
composition, while the component that tends to sink forms a concentrated
bottom layer. If the particles differ only in size, the larger particles settle to
the bottom. Nienow and Rowe, who have done a considerable amount of
work on segregation caused by density differences (e.g. Nienow, Rowe, and
Cheung, 1978) have used the words flotsam and jetsam to describe the solids
which occupy, respectively, the top and bottom of the bed. In non-technical
English flotsam means 'floating wreckage' and jetsam, goods thrown out to
lighten a ship and later washed ashore! A quantitative measure of segregation
is required and two parameter are often used:

Mkg solids in bed

solids feed

'E
0

'C
Q)

~
Q)

Ol

~
~
Q)

J - 0 Bed size 0.61x 0.61m


U-Umf

(em Is)

18

Symbol

Weir height

25
""

35

= 15 em

Rate of solids

flow

Mean

residenCe

Mean

particle

time 7 R

size

(a) The coefficient of segregation

8kg/min

2.6 min

XB

= 1.6 mm

Figure S.10 External age distribution of traeer solids in a fluid bed with
constant solids feed - tracer added instantaneously at tflR = 0 (Cranfield,

f= 1 - e-t1tp.

in a completely
~
-

x 100

+ XT

(b) The mixing index M 1 is defined as x/x, where x is the concentration of'the
material under scrutiny at some level in the bed and x is its average
concentration.

(5~

A significant percentage of solids (18.2 per cent.) spends less than 20 p~r cent.
of the average residence time tR in a fully mixed bed. This problem can be
solved in shallow beds ( < 0.15 m) by installing vertical baffles or making the
bed in the form of a long channel (see Chapter 10). Baffles are sometimes
used in deeper beds but become progressively less effective in preventing
short-circuiting of solids as bed depth increases. There is a need for further
work on baffled deep beds.
5.7

-XT)

where XB and XT are the concentration of the material of interest in the


bottom and top halves of the bed, respectively. Clearly Cs can have values
between -100 and + 100 per cent., 0 per cent. being perfect mixing.

1978)

Typical experimental data are shown in Fig. 5.10.


Similarly, the fraction of solids f spending less than time
mixed bed is:

XB

Cs is defined as:

SEGREGATION AND DEFLUIDIZATION

Researchers
often use the words segregation and defluidization interchangeably and this makes for difficulties in interpretation of published data.

5.7.1

Methods of Investigation

A qualitative indication of defluidization can be obtained from the bed


pressure drop-gas velocity curve. Consider a bed which consists of materials
likely to segregate and which is either premixed outside the bed or vigorously
bubbled and then collapsed rapidly. If the gas velocity is gradually increased,
a pressure drop curve OAB (Fig. 5.11) is obtained; the wavy part ofthe curve
AB occurs as the mixture becomes fluidized and then progressively segregates
and defluidizes locally. All the particles are fully supported at Vef, and are
therefore fluidized with no fixed bed region at the distributor. Further
increase in velocity produces bubbles throughout the bed, but they may not
be sufficiently large at the distributor to promote good mixing, i.e. the bed
may be segregated.

5.7.2

Segregation By Density Difference

Segregation is much more likely to occur when the powder contains particles
of different densities than when the size range is very broad; Decamps,
Dumont, and Goossens (1971), Bena et at. (1968), and several other groups
of workers have investigated segregation by density difference. In the papers
of Nienow, Rowe, and coworkers (Rowe, Nienow, and Agbim, 1972;
Nienow, Rowe, and Cheung, 1978) various combinations of heavy and light
particles were fluidized. Rather complex empirical correlations were developed and have proved quite useful:

UTa
UrnfS

(UrnfB)
UrnfS

(d )o.7

1.2+ 0.9 (PH _l)J.l

2.2

Yx

dL

PL

Figure 5.11 Pressure drop curve for wide size distribution powder initially
well mixed (Boland, 1971).

As the gas velocity is decreased, defluidization occurs systematically so that


the pressure drop is being measured across a fluidized bed zone superimposed
on an increasingly deep fixed bed; this gives the curve BCG.
A quantitative
method, the 'freezing' technique, has already been
described in Section 5.2.1 and gives rise to curves of the type shown in Fig.
5.12.
x
I

top

bed

(5.17)

where UrnfB and UmfS are the bigger and smaller mlmmum fluidization
velocities of the two sets of particles; PH and PL are the particle densitities
of the denser and less dense particle; dH and dL are the sizes of denser and less
dense particle; x is the mass fraction of the denser particles in the whole bed;
Hand D are the height and diameter of the bed; UTa is the critical velocity at
which mixing takes over from segregation. In general, UTa> Uet.
Having calculated UTO it is then inserted into Eqs (5.18) and (5.19) to
calculate the mixing index M1:
.

U - UTO
z=----e
U - UmfS

UfU
TO

Thus z and therefore My increase as the gas velocity U increases.


Although developed for relatively small particles at low velocities the
equations have been applied with success to large particle systems and
velocities up to 5 m/s (Geldart et ai., 1981).

t[L:x=x

height

HlDY4

(1 _ e-

I
I

bottom

1.

mass fraction
reference

5.7.3

size

PERFECT MIX

COMPLETE
HORIZONTAL
SEGREGATION

Segregation By Size Difference

Providing the operating gas velocity is larger than Ucf, the velocity of
complete fluidization, there will not be a fixed bed (defluidized) region at the
distributor. However, that is not to say that the bed will be well mixed; it may
be segregated with more coarse at the bottom and more fines at the top.
Boland (1971), working with particles up to about 550 /Lm, developed a
bubbling bed model to predict the value of U - Umf required to avoid

segregation by size. Geldart et al. (1981) studied much larger particles


fluidized at velocities up to 5 mls and concluded that segregation by size
difference appears to be an intrinsic feature of beds of large particles. It
becomes larger as:
(a) The size of the fines decreases.
(b) The mean size of the bed solids increases.
(c) U - Urnf approaches zero.
However, whereas segregation by density difference decreases progressively
with increasing gas velocity, above values of U - Urnf larger than about 2 mls
segregation by size difference becomes relatively constant. The presence of
horizontal tubes appeared to make relatively little difference in the
experiments cited, in which fines were reinjected near the distributor. One
plausible hypothesis is that fines are carried upwards by three mechanisms
and downwards by one. The mechanisms are:
1. Entrainment within the channels or bubbles at velocities which exceed the
terminal velocities of the fines.
2. Bulk movement of fines in the wakeldrift material dragged upwards by
bubbles.
3. Stripping of fines through the voids between particles in the dense phase.
4. The downward movement is achieved by the bulk movement of fines
trapped within the dense phase in the bubble-free regions.
These mechanisms explain in a qualitative way why segregation by size is
relatively insensitive to velocity. Since upward mechanisms 1 and 2 and the
return downward motion of mechanism 4 increase with gas velocity, they tend
to be self-cancelling.
If the size of the coarse is increased, for .agiven superficial gas locityand
fines fraction, the rate of mechanism 3 will increase, partly because
size of
the voids is bigger and also because the interstitial gas velocity, whicfi .s at
least Umf/Ernf, increases. The downward movement of mechanism 4 ill
decrease because U - Urnf which governs mixing has decreased. Qualitative
relationships between the variable were proposed (Geldart et al., 1981):

x~oP = (dp)Q
d
Xf

)b

( U -U U
rnf

rnf

by the softening of the particle with increasing temperature. The result is that
particles may agglomerate to such a size that the gas velocity is insufficient to
support them. This type of defluidization can occur very rapidly (Gluckman,
Yerushalmi, and Squires, 1976). The bed is well-fluidized one moment and
within 5 seconds is dead. As the bed defluidizes and the particles become
loosely stuck together, the fluidizing gas blows a hole through the bed in
order to escape and the pressure drop decreases rapidly.
The tendency to agglomerate depends on (a) the stickiness of the particles,
which is a function of temperature, (b) the available surface area - the
smaller the particles the greater the surface area - and (c) particle
momentum, which is a function of particle size and gas velocity_
In general:
.

Agglomeratmg tendency

(a) x (b)
=

(c)

At the present time it is not possible to predict from first principles the
quantitative relationship between the variables, and experiments are essential
to determine the conditions required for stable operation.
Defluidization can be influenced by the design of the gas distributor and it
is essential to ensure that there are no dead areas at the plate. This might be
taken to imply that a porous plate is best, and it is true that all the solids will
be aerated; however, the small bubbles produced do not have sufficient
energy to provide the vigorous movement required to remix large particles
which may have started to segregate there.
On the other hand, if drilled plates having large holes are used, the large
distance between them (for a given pitch/diameter ratio) allows solids to settle
out. Baeyens (1973) made a study of particle movement caused by bubbles
and suggested that for particle movement over the entire plate:
p <

deq,o

(5.22)

where A is a function of the particle mobility and deq,o is the initial bubble
size produced at the distributor. Geldart (1977) used this approach to derive
Eq. (5.23), which gives the minimum gas velocity required to avoid dead
areas between holes:

(5.20)
(gpf5

Clearly much more work is required in order to develop quantitative


relationships.
5.7.4

Defluidization Through Particle Stickiness/Agglomeration

Defluidization sometimes occurs when interparticle forces become excessive.


This may be brought about by the presence of moisture or binding agents or

K A2.5
where K = 1.73 for holes on square pitch and 2.17 for triangular pitch. For
conservative design, A :=:: 1 for group B solids and 1.5 for group A. Wen,
Krishnan, and Kalyanraman, (1980) used thermistors to detect particle
movement at distributor plates and propose that

30

U - Urnf > - (p - dor)O.716


p

tf.!.205
P

velocities. The whole question of mixing patterns at velocities above 1 mls


also deserves greater attention.

where the units are centimetres per second and centimetres, respectively,
with dp in micrometres. Dead zones are easier to eliminate using tuyeres with
multiple horizontal holes or conical top bubble caps (sometimes called
'Chinese hats').

Mechanical aids or suitable bed internals are sometimes used to promote


mixing andlor to reduce segregation in materials which are naturally cohesive
(group C) or pass through a sticky stage (coal gasification). Alternatively,
they may be used in applications when reaction rate requirements or a desire
to avoid elutriation favour gas velocities which are so low that mixing/heat
transfer would otherwise be unacceptably low.
The use of a stirred bed (Brekken, Lancaster, and Wheelock, 1970)
achieves satisfactory mixing of strongly cohesive powders due to constant
breakdown of any channels formed near the distributor.
Strong particle circulation can be deliberately set up within a bed by
including a draft tube. If excess gas is injected through the tube, the
additional bubble stream will create a pumping action. A .refined method is
described by Decamps, Dumont, and Goossens (1971), who suggest using an
internal airlift to generate more intense vertical mixing and prevent
segregatioon at lower overall gas flowrates, Baxerres, Haewungscharren, and
Gibert (1977) have proposed an alternative to the draft tube, called the
whirling bed.

It is the bubbling phenomenon which causes particle circulation patterns be


established. This circulation is essential where bed homogeneity is requir d
and of particular interest when solid feed streams must be rapidly and totally
dispersed in order to avoid localized accumulations.
Model and empirical equations for mixing, as developed above, enable the
designer to estimate those operating conditions which yield the required
particle circulation. However, there may be other conflicting requirements
such as the degree of conversion and heat transfer. The remaining difficultyin
scale-up of mixing data from small diameter columns to industrial equipment
is a result of the non-equivalent circulation patterns in both sizes. Fluidized
behaviour at equivalent bubble to bed diameter ratios offers a promising
approach to scale-up.
Further work should be directed towards the phenomenon of segregation
by size and density difference, particularly on a larger scale and at higher

Example 5.1

A fluid bed catalytic reactor of 3.5 m diameter is used for a fast, exothermic
reaction releasing 125 kcal/mol of product. Assuming a production of 2,000
kmol/h of product, and 90 per cent. conversion immediately above the
distributor:
(a) What solid flux is required to carry the heat produced to the heat
exchanger surfaces located within the bulk of the bed (average
temperature difference between the bed and surface is 100C)?
(b) What gas flowrate will produce such a flux?
Properties of catalyst:dp : 80 JLm
Cps: 0.2 kcal/kgOC
Urnf: 0.3 cmls
3

Pg : 1 kg/m

Pp

Emf:

JL

1,500 kg/m3
0.5
4 X 10-5 kg/ms

Solution

(a) Production of reaction heat within the distributor zone:


0.9 x 2,000 x 125,000 = 225 x 106 kcallh
Solids flux required:
225 x 106
--= 11,25 X 106 kg/h or 325 kg/m2s
0.2 x 100
(b) The velocity required is calculated using Figs 5.6 and 5.7 in conjunction
with Eq. (5.10) or Eq. (5.11).
With f3w = 0.5, f3d = 1, Y = 1 from Eq. (5.10):
U =: 0.5 m/s

Example 5.2

Limestone of average particle size 2 mm and density 2,700 kg/m3 is fed to a


fluidized calciner with 1.3 m diameter (static bed height 0.8m) at a rate of

3,500 kg/h. The CaC03 concentration in the offtake must be less than 5 per
cent.
(a) What velocity is required to avoid segregation of the limestone feed
within the bed which will be 95 per cent. lime (pp = 1,600 kg/m3)?
(b) Calculate the turnover time of the limestone at the selected velocity.

mixing (M1 = 0.9) the gas velocity should be 4.2 mls. Operation at 1.6 m/s
would give pronounced segregation (M1 = 0.1).
(b) For the limestone, Ar = 37,082. From Figs 5.6 and 5.7 and Eq. (5.13),
operation at U = 4.2 mls gives a turnover time:
-------------

(0.1

+ 0.38 x 0.1)

0.8
x 0.15 x (4.2 - 0.8)

l1.4s

Solution

(a) Assume that the lime produced and the limestone feed have the same
particle size: dcao = 2 mm, PCaO = 1,600 kg/m3. If the offtake contains a
maximum of 5 per cent. of CaC03, the fraction of limestone in the wellmixed bed is less than 0.05 (= x).
For combustion gases at 850C : Pg = 0.28 kg/m3, J.L = 4 X 10-5
kg/ms. Umf can be calculated from the Wen-Yu equation (which slightly
over-estimates at temperature), Eq. (2.27). Therefore:

Calculate UTQ from Eq. (5.17) and use Eqs (5.19) and (5.18) to determine
the mixing index MI for various gas velocities. These are shown on Fig.
5.13. Because conversion takes place continuously, thus gradually
reducing the density of the limestone, there will be relatively little pure
limestone in the bed. However, for conservative design, assume the
residual CaC03 is present as limestone particles. From Fig. 5.13, for good

A rectangular drier (1.5 x 4 m) for polymer powder (dp = 600 p.m, Pp =


1,300 kg/m) is fitted with a multi-hole tuyere distributor (234 tuyeres at
square 0.15 m pitch, each tuyere having four orifices of 6 mm diameter). Its
blower delivers a maximum gas flow of 15,000 N m3/h at 120C.
(a) Which is the largest polymer size that can be treated at 120C in the
equipment without dead zones on the plate?
(b) Is the superficial velocity obtained from the blower sufficient to
guarantee good overall bed mixing for this size of polymer?
Solution

(a) The maximum polymer size can be calculated fromEq. (5.23) taking into
account:
(i) maximum superficial velocity available from the blower (120e) = 1
mls;

(ii) p. = 2 X 10-5 kg/ms and Pg = 0.9 kg/m3;


(iii) from the laminar part of the Wen-Yu equation:
Umf = 3,865 d~ (c.g.s. units);
(iv) with p = 0.15 m, Umax = 1 mis, and A = 1, the value of
1.14 mm (Umf = 0.5 mls).

~.9
08

0.7
0.6
0.5
0.4
0.3
0.2
------

2
U(m/s)

dp,max =

(b) There is no absolute criterion to check overall mixing of this 1.14 mm


powder at 1 m/s. Tentatively, however, one can imagine that optimum
mixing conditions prevail when achieving the maximum heat transfer
rates in the bed.,The velocity required for this condition can be estimated
from either:
0.92 Aro.32 0;;; (U - Umf)max 0;;; 1.25 Aro.36
(5.25)
(Baeyens and Geldart, 1981)
or
Ar
Reopt ~ 18 + 5.22 yeAr)
...

0.1

Example 5.3

(Todes, 1965; see Chapter 9, Fig. 9.4)

SOLIDS

Using these equations with Ar


From Eq. (5.25) : 0.28 ~ (U From Eq. (5.26) : Uopt

UTO

42,512:

Umf)opt ~

vp

0.58 mls

0.75 mls

At U = 1 mls and Umf = 0.5 mis, both equations indicate that mixing
conditions are adequate.

XB,XT
Xf,

A
C
Cs
deqs
deqo

dor
dp
dH, dL
df
D

f
g
H
J

My
p
Qb

t
tT
tR
UA
Ub

U
Ucf
UmfB, UmfS
Umfi

cross-sectional area of fluidized bed


concentration of tracer particles
'coefficient of segregation
equivalent bubble diameter at bed surface
equivalent bubble diameter initiated at distributor
diameter of distributor orifice
average
diameter
particles
average

particle diameter
of, respectively, denser and less dense
in binary mixture
diameter of segregating fines

y
(3w,(3D

m
m
m
m
m

E],
fJH,fJL
1,

Pp

takeover velocity
average downwards particle velocity
volume
average mass fraction of the denser particles in a
binary mixture
mass fraction of material of interest in,
respectively, the bottom and top halves of the bed
mass fraction of fines, respectively, as bed
average or in the top halves of the bed

mls
mls
m3

ratio of volumetric flowrate of bubbles, Qb, and


excess gas flowrate, (U - U mf)A
fraction of solids carried up by a bubble within,
respectively, its wake and drift
fraction of the bed consisting of bubbles
densities of denser and less dense particles in
binary mixture
absolute density of fluidized particles

kglm3
kglm3

m
m

bed diameter
fraction of solids spending less than time t in the
bed
gravitational constant

9.8 m/s2
m
kglm2s

bed height
mixing flux
mixing index
pitch of holes in distributor
visible bubble flowrate

m2

Xf,Top

121

MIXING

time
time required to turn over the bed once
average particle residence time
rise velocity of bubbles relative to the wall
rise velocity of a single isolated bubble
superficial gas velocity
velocity at which all particles are fully supported
(Eq. 2.31)
bigger and smaller minimum fluidization velocities
of two sets of particles
interstitial gas velocity at minimum fluidization

s
s
s
mls
mls
mls
mls
mls
mls

Baeyens, J, (1973). Ph.D. Dissertation, University of Bradford.


Baeyens, J. (1981). 'Gas fluidization', Short Course, I. Chern. E./University of
Bradford.
Baeyens, J., and Geldart, D. (1973). Fluidization et ses Applications, Toulouse, p.
182.
Baeyens, J., and Geldart, D. (1981). J. Powder and Bulk Solids Tech., 4, (4), 1.
Baxerres, J.L., Haewsungscharren, A., and Gibert, H. (1977). Lebensrn, Wiss.
Technol., 10, 191.
Bena, J., Havalda, I., Bafinec, M., and Ilowsky, J. (1968). Colt. Czech. Chern.
Comm., 33,2620.
Brekken, R.A., Lancaster, E.B., and Wheelock, T.D. (1970). Chem. Eng. Prog.
Symp. Ser., 66, (101), 91.
Bohle, W.,and van Swaaj, W.P.M. (1978). In Fluidization (Eds J.F, Davidson and
D.L. Keaims), Cambridge University Press, p. 167.
Boland, D. (1971). Ph.D. Dissertation, University of Bradford.
Chavarie, C., (1973).' Ph.D. Dissertation, McGill Univ.
Cranfield, R.R. (1978). A.I.Ch.E. Symp. Ser., 74 (176), 54.
Decamps, F., Dumont, G., and Goossens, W. (1971). Powder Technol., S, 299.
de Groot, J.H. (1967). Proc. lnterm. Syrnp. Fluidization, Netherlands University
Press, Amsterdam, p. 348.
Everett, D.J., (1970). Ph.D. Dissertation, University of London.
Geldart, D. (1977). 'Gas fluidization', Short Course, I.Chem.E./University of
Bradford.

Geldart, D. (1980. 'Advances in gas fluidization', Short Course at Center for


Professional Advancement, New Jersey.
Geldart, D., Baeyens, J., Pope, D.J., and Vandeweyer, P. (1981). Powder Techno!.,
30, 195.
Geldart, D .. and Kelsey, J.R. (1968). In 'Fluidization', LChem.E. (London) Symp.
Ser., p.1l4.
Gluckman, M.J., Yerushalmi, J., and Squires, A.M. (1976). In Fluidization
Technology (Ed. D.L. Keairns) Vol. II, Hemisphere, p. 395.
Grace, J.R., and Harrison, D. (1969). Chern. Eng. Sci., 24,497.
Kondukov, N.B., (1964). Intern, Chern. Eng., 4,43.
Lewis, W.K., Gilliland, E.R., and Girouard, H. (1961/62). Chern. Eng. Progr. Syrnp.
Ser., 58 (38), 87.
May, W.G. (1954) Chern. Eng. Progr., 55,49.
Miyauchi, T., Kaji, H., and Saito, K. (1963). J. Chern. Eng. (Japan), 1,72.
Nienow, A.W., Rowe, P.N., and Cheung, L.Y.L (1978). Powder Techno!., 20,89.
Nguyen, H.V., Whitehead, A.B., and Potter, O.E. (1978). In Fluidization (Eds J.F.
(Ed. H.M. Hulbert), vol. 133, Am. Chern. Soc., p. 290.
Nguyen, H.V., Whitehead, A.B., and Potter, O.E. (1978). In Fluidization (Eds J.F.
Davidson and D.L. Keairns), Cambridge University Press, p.140.
Okhi, K., and Shirai, T. (1976). Fluidization Technology. (Ed. D.L. Keairns),
Hemisphere, p.95.
Rowe, P.N., and Everett, D.J., (1972). Trans. Instn. Chern. Engrs., 50,42.
Rowe, P.N., Nienow, A.W., and Agbim, A.J. (1972). Trans. Instn. Chern. Engrs, 50,
310.
Rowe, P.N., and Partridge, B.A. (1962). Proc. Syrnp. Interaction between Fluids and
Particles, Instn. Chern. Engrs, p.135.
.
Rowe, P.N., Partridge, B.A., Cheney A.G., Henwood, G.A., and Lyal1,E. Trans.
Inst Chern. Engrs., 50,310.
Talmor, E., and Benenati, R.F. (1963). A.I.Ch.E., 9,536.
Thiel, W.J., and Potter, O.E. (1978). A.I.Ch.E., 24 (4), 561.
Todes, a.M. (1965). Appl. offluid beds in the chemical industry, Part 2. Izd Znanie
Leningrad, pA.
Wen, c.Y., Krishnan, R., and Kalyanaraman, R. (1980). In Fluidization (Eds J.R.
Grace and J.M. Matsen) Plenum Press, p. 405.
Werther, J. (1975). Proceedings of CHISA Conf., Prague.
Werther, 1., and Molerus, 0., (1972). Proceedings of CHISA Conf, Prague.
Werther, J. (1976). In Fluidization Technology Vol. I (Ed. D.L. Keairns), McGr Hill, p. 215.
Werther, 1., and Molerus, O. (1973). Int. J. Multiphase Flow, 1, 103.
Whitehead, A.B., and Auff, A. (1976). Powder Technol., 15, 77.
Whitehead, A.B., and Dent, D.C. (1978). Fluidization (Eds J.F. Davidson and D.L.
Keairns), Cambridge University Press, pA4.
Whitehead, A.B., and Dent, D.C. (1982). A.I.Ch.E., 28, 169.
Whitehead, A.B., Gartside, G., and Dent, D.C. (1976). Powder Techno!., 14,61.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Particle Entrainment and Carryover

In most applications of fluidization the carryover of fine particles in the exit


gases is regarded as a nuisance to be endured in exchange for the advantages
offered by other features of the technique. In a few processes selective
removal of small light solids by elutriation is advantageous, while in the
recirculating (fast) fluid bed, carryover and reinjection of the whole bed
material is essential.
Three processes which typify these different ways of viewing carryover are
catalytic cracking, incineration of sludges from waste-water treatment plants,
and calcination of aluminium trihydrate. In the first, attrition of valuable
catalyst in cyclones and loss of the fines can increase the cost and
controllability of the operation; in the second the solids are a coarse sand
which remains in the bed, and the fine light ash from the burnt sludge is
carried out continuously as an overhead product, thus providing a convenient
method of removal; in the third process, high throughput and the control of
residence time and bed temperature depend on achieving high carryover
rates. In this chapter we shall consider the factors affecting carryover and
discuss ways of predicting the amount and size distribution of solids carried
over.

The terms carryover, entrainment, and elutriation are often used interchangeably to describe the ejection of particles from the surface of a bubbling
fluidized bed and their removal from the unit in the gas stream. Elutriation
123

also describes the technique used to determine the size of subsieve particles or
to prepare narrow cuts from a material having a wide size distribution. This
involves passing air or water at a known velocity upwards through a powder
sample anti! no more is carried out. The material collected is weighed and its
size calculated using Stokes' law.
The terminology and processes occurring in entrainment are best illustrated
through a series of experiments. Imagine a tall vertical column into which
solid particles can be introduced (Fig. 6.1). First drop in a single particle.
Providing it remains more than about 10 particle diameters away from the
wall, its velocity will increase rapidly until it reaches a steady falling rate
which is called its terminal velocity. Ways of calculating this are given later.

If now a few grams of powder are dropped into the tube with no gas flow
the solids may fall as one or more pseudo-particles having effective sizes
larger than those of individual particles, but lower densities. Particles tend to
be shielded by those beneath them and therefore draw nearer to each other.
Very small particles having intrinsically cohesive properties, or particles
which are sticky or damp, also tend to cluster together, and have a higher
terminal velocity than would be expected from their individual sizes.

Small particles
large particles

move upwards
move downwards

Partjcle in this
position mow:s

upwards

Static
gas

Ymax

2Yt

Yavg

Yt

Figure 6.1 Motion of particles falling in a tube


(a) without and (b) with ~p~ard flow ~f gas.

Velocity prOfile
(turbulent flow)

Now pass air up the tube at a superfiCIal velOCItyequal to the partIcle s "
terminal velocity and drop the particle in again. The way it behaves now
depends on several factors. If the gas flow is laminar with respect to the
column there will be the well-known parabolic velocity distribution across the
tube, the maximum velocity at the centre being twice the mean superficial
velocity, and zero gas velocity at the wall. Depending on its radial position the
particle may move upwards or downwards (Fig. 6.1b). If the gas flow is
turbulent not only does the velocity vary across the tube but there are also
random velocity fluctuations with respect to time, and the particle motion is
more unpredictable. Thus, allowing a particle to fall through a stati<::gas and
moving the gas upwards past the particle at the same relative velocity are by
no means identical.

Now pass gas up the tube at .a velocity considerably higher than the
terminal velocity of the mean particle size, and feed in a continuous stream of
particles (Fig. 6.2). They will be picked up '(entrained) by the gas and
transported upwards in dilute phase pneumatic transport. If the size
distribution is so wide that some particles have terminal velocities higher than
the gas velocity then a fractionation (elutriation) of solids occurs with the finer
particles carried upwards and the coarser falling to the bottom of the tube. If
the gas velocity is much higher than the terminal velocity of even the largest
particle and the solids feed rate is gradually increased, the concentration of
solids in the gas increases, together with the pressure drop along the tube,
until a solids/gas loading is reached at which the solids start to fill up the lower
section of.the tube and dilute pha~e flow is no longer possible in the upper
section. The gas/solid suspension then collapses into dense phase or slugging
flow depending on the powder and the tube diameter. The loading at which

300
200

this occurs is known as the saturation carrying capacity of the gas and the
critical .gas velocity is the choking velocity.
Now reduce the gas velocity, allow a fluidized bed to form and cut off the
solids feed. Bubbles bursting at the surface of the bed will project particles
into the freeboard, i.e. the space between the bed surface and the gas offtake
(Fig. 6.3). Depending on their terminal velocities and the gas velocity, solids

100
80
60
40

20
~
"E

:t:

Dilute
phase
transport

e 10
8
6

.- ...
.. . .',
"

..

".",

Disengagement
zone
-TDH (F)

f +

..----

IHold-UP or
density

01
Splash zone
TDH (e)
Bed

are carried up the tube to various heights. The larger particles fall back and
the smallest are carried out of the offtake tube; consequent~, the solids
loading (expressed as kilogram solids per second divided by c~
of gas per second of kilograms of solids per second divided by kilograms of
gas per second; i.e. as kg/m3 or kg/kg) will decline with height. The region
within which the solids loading falls is called the transport disengagement
height )TDH). Some confusion is caused by the application of the term TDH
to two related but different circumstances:
(a) Particles having a terminal velocity larger than the gas ~elocity are flung
upwards and then fall back. These particles are coarses and above the
height they reach are found only fines - particles with terminal velocities
smaller than the gas velocity.
(b) The concentration (or hold-up) of fines decreases with height and, it is
said, eventually reaches a constant value above a certain height.

0-2

0-4

1246810
U-Umb (ftls)

db - bubble dia. at surface


. Figure 6.4 Transport disengagement height as a
function of bubble size and gas velocity (Zenz,
1983) ..

The height which the coarses reach is sometimes called the splash height but
we shall call it TDH(C). The height above which the fines hold-up changes
little or not at all will be called TDH(F). Published correlations for the TDH
are given in Table 6.1.
It is clear that the definition of a particle as coarse or fine will depend upon
its terminal velocity and the superficial gas velocity, so we shall consider next
how best to calculate terminal velocities.
6.3

PARTICLE

TERMINAL

VELOCITY

The terminal (or free-fall) velocity of a particle, VI' figures in most of the
correlations for carryover and is also of considerable importance in
sedimentation, gas and liquid cyclone design, and many other areas of
chemical engineering. It has therefore been a subject of study by numerous
workers, and as there are several excellent comprehensive texts available
(e.g. Clift, Grace, and Weber, 1978; Allen, 1981), the treatment here will be
kept fairly brief.

6.3.1

Spherical Particles

Consider a spherical particle falling freely in an infinite fluid. The forces


acting on it are shown in Fig. 6.5:

TDH (C)
Soroko et ai. (1969):
TDH (C) = 1,200 Hs Re~55Ar-l.I
for 15 < Rep < 300; 19.5 < Ar < 6.5 X 105; Hs < 0.5 m ; dp = 0.7 - 2.5 mm
where Rep = PgUdpl IL ; Hs = settled bed height
. TDH (F)

Gravitational

force - buoyancy force - drag force

accelerating

The drag force can be expressed in terms of a drag coefficient Co (a sort of


friction factor analogous to f in the packed bed equation), the projected area
perpendicular to the flow Ap, and the inertia of the fluid:
F = CoApljz PI v2

Haria et ai. (1980):


TDH(F) = 4.47 d~~s
where deq.s = equivalent volume diameter of bubble at bed surface
Amitin (1968):
TDH(F) = 0.85U1.2 (7.33 - 1.2 IOglOU)

(6.2)

As in flow through pipes and packed beds, the 'friction factor' Co turns out to
be a function of the Reynolds number which in this case is defined as:
Pf v dv
Re=--

(6.3)

JL

For a sphere Ap

7Td~/4 and substituting in Eq. (6.2) gives:

Wen and Chen (1982):

(Eo-. K~';)
/.-.rJ

TDH(F)j = -:-loge
0 01K~
a,
E
p3.5 gO.s
where Ado = 3.07 x 10-9 g 2.5 (U -

force
(6.1)

\ Fournol et ai. (1973):


TDH(F) = 1,000 U2/g

d2

COT
7T

Pf v2

(6.4)

,00

IJ.

eq.s

Equation (6.1) can now be written algebraically:

umff5

and measured values 3.5 < tli > 6.5 m-1: use average

tli,

= 4 m-1

tnIis equation is based on beds smaller than O.6m diameter and it is unwise to use it for large
beds.

7T

"6 (pp

- Pf)g - F

7Tdv Pp

dv

di

When a particle reaches its free-fall or terminal velocity the accelerating force
is zero, v = Vt, and Eq. (6.5) becomes:
7T

= "6 d~ (pp

In the general case, combination

of Eqs (6.4) and (6.6) gives:

4
Co=-

Gravitational
force

- Pf)g

(pp - Pf)dv
2
PfVt

g
(6.7)

By using experimental data from spheres falling in a variety of fluids (Table


6.2), log Co is plotted against log Re to give the so-salled 'standard drag
curve', which has three broad regions (Fig. 6.6):
(a) Laminar region (Re < 0.2). Stokes solved the analytical equations for the
drag on a smooth rigid spherical particle moving on a homogenous
viscous fluid at velocity v:

Table 6.2

Standard

Re

Susbtituting

Vt,ST

from equations in

CoRe2

Co
2,400
1,204
484
244
124
51.5
27.1
14.76
7.03
4.26
2.71
1.57
1.09
0.77
0.555
0.471
0.421
0.387
0.405
0.442
0.474
0.5
0.497
0.376
0.11

0.01
0.02
0.05
0.1
0.2
0.5
1
2
5
10
20
50
100
200
500
1,000
2,000
5,000
10,000
20,000
50,000
100,000
200,000
500,000
106

drag coefficients for spheres (calculated


Clift, Grace, and Weber, 1978)

2.4
4.8
1.21
2.44
4.96
12.87
27.1
59
1.76
4.26
1.08
'3.93
1.09
3.08
1.39
4.71
1.68
9.67
4.05
1.77
1.18
5
1.99
9.4

CoiRe
1

x 10x 10-1

x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x

102
102
103
103
104
104
105
105
106
106
107
108
109
109
1010
1010

2.4
6.02
9.68
2.44
6.2
1.03
27.1
7.38
1.41
4.26
1.35
3.14
1.09
3.85
1.11
4.71
2.1
7.74
4.05
2.21
9.48
5
2.48
7.52

x10S
X

x
x
x
x

X
X
X
X
X
X
X
X
X
X
X
X
X
X
X

104
103
loJ
lOZ
102

10-1
10-1
10-2
10-2
10-3
10-3
10-4
10-4
10-5
10-5
10-5
10-6
10-6
10-6
10-7

this expression for Fin Eq. (6.6) and writin~

gives:

v t,ST

(pp - Pf)g tf;


\
1811-

Spherical/rounded
sand particles (pp = 2,600 kg/m3) smaller than 33
JA-mfalling in air are in the Stokes region; if the expression is used for
larger and/or denser particles then the terminal velocity will be
overestimated. Combining Eqs (6.4) and (6.8) gives:

CD=-

10

Laminllr flow

(Stokesl -~gion
regIOn
I

Tronsition

Turbulent

(Newtonl

,.

region

0,

Ie!

10'

10
Re

10'

10'

10

10'

P"t ~/"

Figure 6.6 Standard drag curve for spheres versus particle Reynolds number.

(b) Turbulent flow (Ret > 1,000). Spherical sand particles larger than 1,500
JA-min air are in the turbulent regime (sometimes called the Newton
region) and CD = 0.43; substitution in Eq. (6.7) and rearrangement leads
to:
Vt,N

=JI:

(pp ~~~dv

(6.11)

Note that the fluid viscosity does not appear in the equation at all and
that Vt is much less dependent on particle size than it is in the laminar
regime (Eq. 6.9). On the other hand, the time and distance taken to
attain Vt.N become considerable
(Table 6.3). The procedure for
calculating the acceleration time will be found in Heywood (1962).
Table 6.3 Time and distance for spheres (pp = 2,600 kglm3) to reach 99
per cent, of their terminal velocities falling in air at ambient conditions

d.

Vt

m/s

24

Time taken
to reach 0.99

Vt

Distance fallen before


reaching 0.99vt
Metres

Ret

which is analogous to the expression for the friction factor for laminar
flow in pipes (j = 16/Re), Particles with terminal Reynolds number> 0.2
reach their terminal velocity very rapidly (see Table 6.3).

10'

30

}Lm

0.07

3 mm 14
3cm 44

0.033
3.5
11.9

1.85 x 10-3
35
453

Particle diameters
61
11,670
15,100

(c) Transitional flow (0.2 < ReI < 1,000). In many applications of fluidized
beds the particles of interest have terminal velocities within the transition
regime and a rapid method of calculation is necessary. It can be seen
from Eq. (6.7) and Fig. 6.5 that VIis a function of Co which is itself a
function of ReI> and ReI contains VI' There are several empirical and
semitheoretical expressions relating CD and Re in the transition region,
the best known being that of Schillar and Nauman (1933). it depends on
successive approximations and is therefore tedious. Moreover, it is not at
all suitable for calculating the size of a particle which has a given
terminal velocity. A more convenient procedure is based on Eq. (6.7).
Multiplying both sides by p d~/ p,2 gives:

vt

Ret = "3 Pt(pp

- Pt)g

d;

=~

'0

,I
,{

1\-\
\

C'

\<2-

\
\

CDRe~ is a dimensionless group (= 4/3 Archimedes number) containing


only the physical properties of the particle/fluid system. The CDversus
ReI data can be recalculated as CD Re~ so that Ret can readily be found
for any value of CoRet, either from a graph of from a polynomial (Fouda
and Capes, 1976).
To find the particle size corresponding to a given terminal velocity, dv
must be eliminated from Eq. (6.7) by dividing by ReI to give:
CD

II

\~

CD

Pp - Pf

p,g

f'\
\

There is as yet no satisfactory way of predicting the terminal velocity of an


irregular shaped particle. We have already encountered (Chapter 2) the
problem of defining the size of non-spherical particles and although sphericity
t{J is a useful parameter, two particles with similar values of t{J can have quite
different shapes (Table 2.1); for example, a regular tetrahedron and a disc
whose diameter is four times its height have t{J values of 0.67 and 0.69,
respectively. A symmetrical particle, such as a sphere, cube, or tetrahedron,
falls vertically in a stable manner at a constant velocity in any orientation,
whereas a disc will move to the side and wobble as it falls. This willbe obvious
to anyone who has watched a leaf fall from a tree of tried to cover a small coin
on the bottom of a bucket of water by dropping a larger coin on to it.

\
'\

'\

\\

(6.13)

Non-Spherical Particles

\
\

ReI
3
PI
vi
Co/Ret is calculated and the plot of Co/Ret versus Ret used to find Ret
and ReI hence dv
Equations (6.12) and (6.13) and the associated graphs, Fig. 6.7, are
valid over the whole range of Reynolds numbers.

6.3.2

"-

"-

1,.0-0----'----'16

,pOO

\
~
\

Nevertheless, in spite of its limitations most corrections to the spherical


particle equations make use of t{J. Pettyjohn and Christiansen (1948)
introduced two correction factors for regular non-spherical shapes with
sphericities between 0.67 and 0.906.
For the laminar regime (Ret < 0.2):
(pp - Pf) g

18p,

t{J

KST

= 0.843 log 0.065

tPv

For the turbuLent regime (Ret > 1,000):


Vt.N

=JI ~
3

(pp - Pf) g dv
KN Pf

Note that Eq. (6.17) implies Co = 0.43 for spherical particles (see Table 6.2).
In the transition regime (0.2 > Ret > 1,000) the method proposed by Becker
(1959) may be used, but it is complicated and has the disadvantage of being
based on large artificial particles. A simple interpolation to obtain a
correction factor, KTR, is sufficiently accurate for most purposes. The
terminal Reynolds number Ret is calculated for a sphere having the same
value of dv as the particle, using CD Re~ and Fig. 6.7, and Vt (sphere) is found.
As estimate is made of '" from Tables 2.1 and 2.2 and KST and KN calculated
using Eqs (6.15) and (6.17). Then:

K=-jK" J( ~:)l(:::=~~
)J(~:)

(618)

The terminal velocity for the non-spherical particle can then be found from:
Vt

6.3.3

KTR

Vt

(sphere)

Effect of Temperature and Pressure on

(6.19)

lit

It can be seen from Eqs (6.9) and (6.11) that increasing the temperature will
cause a reduction in the terminal velocity for particles in the Stokes region
(small particles) and an increase for those falling in the Newton region (large
particles); increasing the pressure has little effect in the Stokes region but
reduces Vt for large particles.

6.4

MECHANISMS

OF ENTRAINMENT AND DISENGAGEMENT,


ANDTDH

There is no general agreement concerning the mechanisms by which solids are


ejected into the freeboard, though bubbles clearly have an important role to
play. While on the one hand postulating that bursting bubbles act like
intermittant jets, Zenz and Weil (1958) also point out that extensive tests
using catalyst failed to establish a relationship between bubble size and TDH.
In a later publication, Zenz (1983) provides a graph showing the variation of
TDH as a function of superficial gas velocity and bubble size at the bed

surface (Fig. 6.4). George and Grace (1978) did experiments which enabled
them to calculate particle ejection velocities; 275 p.m ballotini were found to
be ejected at velocities two to three times the bubble velocity. Leva and Wen
(1971) considered that stratification of fines within a bed was an essential
requirement for entrainment; Geldart et aL. (1981) showed that this ~ay be of
importance in beds of coarse solids containing small amounts of fI~es, but
beds of fine particles which are extremely well mixed also have considerable
entrainment rates. George and Grace (1978) obtained data which they
interpreted as demonstrating that it is the bubble wake which is responsible
for ejecting particles into the freeboard, but observations of the surface of
two-dimensional beds indicate that in general it is the nose of bubble which
ejects the particles; however, ejection from the wake can occur when a
bubble bursts through the base of another which is itself about to break at the
surface.
The picture is complicated even more by three other recent findings;
Morooka, Kawazuishi, and Kato (1980) and Horio et aL. (1980 report that the
solids concentration in the freeboard is non-uniform, with a downwardmoving denser stream of solids at the wall and a dilute upward-moving region
in a central core; Yerushalmi et aI. (1978) deduced that particles form
clusters, or agglomerates, making it possible for fines to remain in the system
even at very high velocities (see Chapter 7); and Geldart et aL. (1979) and
Geldart and Pope (1983) showed that fine particles interact with coarse in the
freeboard, increasing the height to which coarses are lifted and causing the
carryover of particles much larger than would be expected as judged by their
terminal velocities.
It is therefore not surprising that entrainment and carryover cannot be
calculated from first principles, nor that empirical correlations give predictions which differ by factors of over 100.

Many workers have presented their results in terms of an elutriation rate


constant, I(h, where:
Instantaneous rate of
removal of solids of
size dj

== Kih {bed}
area

dt

(XBi

MB)

fraction of bed }
{ consisti?g of size
dpj at time t

Kih

AXBi

Eih

where Eih is the net carryover flux for a component of size dpi when the gas
offtake is at a height h above the bed surface.

For continuous operation, XBi, the concentration of size dpi in the bed, and
MB, the mass of solids in the bed, are constant, and:

K:

Ri =
h A XBi = Eih A
RT =!. Ri = !. K;h A XBi
Eh =!. Eih = !. K;h

or

XBi

Ri

and the total loading presented

in the light of more recent work, we may now picture four (Fig.

(6.22)

Phase 1. A gas stream carrying dispersed solids upwards.

(6.23)

Phase 2. Agglomerates

ejected upwards by the action of bubbles.

(6.23a)

Phase 3. Agglomerates

falling downwards.

Phase 4. An emulsion, more concentrated


the wall.

The solids loading of size fraction dpi in the off-gases:


Pi

freeboard;
6.8):

Eih

than phase 1 moving downwards at

= VA =U

to the cyclone is:


p'

= !. Pi

The concentration of size dpi in the exit gas stream may be calculated from
Eqs (6.22) and (6.23):

For batch operation, XBi changes with time and Eq. (6.21) must be integrated.
Providing MB does not change by more than about 15 per cent., mi.h the mass
of size fraction dpi carried over in time t increases asymptotically as:
mi.!

= XBio

MB /1

- exp

t I
K;h ~:

(6.27)

where XBio is the initial concentration of size fraction dpi i~ the bed.
Similarly, the concentration
of size fraetio~ dpi in the bed declines
exponentially according to:
XBi

= XBio

exp

At )
( -Kih MB

(6.28) ,

As with any batch separation process, such as distillation, it becomes


progressively more difficult to remove a given fraction as its concentration
decreases.

6.6

EFFECT OF SYSTEM AND OPERATING VARIABLES

ON

rib

The rate of carryover cannot be predicted unless K;h is known, and as yet K;h
cannot be calculated from first principles, though several hydrodynamic
models have been proposed.
An approach originally proposed by Lewis, Gilliland, and Lang (1962) and
developed by Kunii and Levenspiel (1969) postulated three phases in the

There is mass transfer between these phases at rates which will depend on gas
velocity, bubble action, particle size and size distribution, and bed diameter.
Although algebraic equations can be written for this model, as yet the
approach has not yielded any useful or realistic predictions.
For the present, the simplified approach of Large, Martini, and Bergougnou (1976) seems more promising. They picture the overall carryover flux

Eih, for a given component of size dpi as consisting of two partial fluxes:

(a) A continuous flux flowing upwards from bed to outlet (this corresponds
to phase 1), E
(b) A flux of agglomerates ejected by bursting bubbles, which decreases
exponentially as a function of freeboard height (this corresponds to the
net algebraic sum of phases 2, 3 and 4).
j",.

Expressed algebraically:
Eih

E j",

+ E io

-a;h

The total solids carryover flux when the gas offtake is at any height h above
the bed surface:
Eh

E",

+ Eo exp (-ah)

Wen and Chen (1982) develop this idea further and propose:
Eh

E",

+ (Eo - E",) exp

(-ah)

carryover rate constant, K~h' increases, though closer examination of some


data (George and Grace, 1978; Geldart, 1981;and Wong, 1983) suggests that
K;h may start to decrease again for fractions less than about 45 Mm,possibly
due to the tendency of cohesive fine particles to agglomerate into clusters.
Particles having

VI

> U - the coarses

Intuition suggests that particles with terminal velocities larger than the
superficial gas velocity should not appear in the exit gas, or that if they do, the
freeboard must be less than the TDH. However, coarses have been found in
the cyclones of a commercial zinc roaster having a 9 m high freeboard
(Avedesian, 1974) and of a fluid bed combustor with a 3.5 m freeboard
(Merrick and Highley, 1972). Geldart and Pope (1983) investigated this
phenomenon in a 0.3 m diameter bed 5 m tall and found that the size and
quantity of coarses carried over depends strongly on the carryover rate of the
fines and can be considerable at fluxes above 2 kg/m2s. The cause is believed
to be collisions and momentum interchange between the fines and coarses; in
effect, the TDH of the coarses, TDH(C), is a function of the f!nes carryover
flux.

On a component basis:
Eih
or K;h

= E + (Eio
= K;oo + (Eo
j",

- Eioo) exp (-ajh)

(6.33)

6.6.2 Concentration of Fines in the Bed

- K;oo) exp (-ajh)

(6.33a)

Equation (6.21) assumes that component carryover rates are directly


proportional to their average concentration in the bed. Setting aside for the
moment the question of segregation within the bed, Figs 6.9 and 6.10 show
this assumption to be justified for both small 2 per cent) and large (> 20
per cent.) concentrations.

If the column is sufficiently tall that the second term on the right-hand side
of Eq. (6.33) can be considered negligible, a measurement at the exit can give
Eioo, and two other measurements lower down can then be used to find
exponent ai and Eio' The ultimate objective is to be able to predict all these
factors from first principles' or, alternatively, from correlations. Unfortunately, many of the published correlations are based on data taken from
columns which were too short, too small in diameter, or both; consequently,
they correlate K;h, not the intrinsic or ejection component rate constants, and
this makes their application to dissimilar systems unreliable. However, we
can make some general comments on the effect which various parameters
have on the carryover of particles.

Particles whose terminal velocities are less than the superficial gas velocity the fines

Particles with vlU < 1 are the most likely to be carried over, and most
researchers agree that as the size of the fines decreases, the component

6.6.3

Type of Solids in the Bed

In powders which have a small mean size, say less than 300 Mm,and a normal
size distribution of particles of equal density, there is little evidence of
segregation within the bed provided it is vigorously bubbling, and the solids
presented to the freeboard by the bubbles have a composition similar to that
of the bed as a whole. If, however, the mean size of the particles is large, say
> 800 Mm, and there is either a very wide or a bimodal distribution, then
there will be vertical segregation, with more fines at the top of the bed even at
high velocities (Geldart et al., 1981). Similarly, if the bed is composed of
solids having different densities, in general the top of the bed will be richer in
the lighter solids. Under these circumstances carryover is likely to be higher
than predicted by correlations based on a well-mixed bed of fine solids.

6.6.4

2'

All correlations include the superficial gas velocity, usually in association with
Vt, as U - Vt, (U - Vt)/Vt, (U - vt)/U, or v/U. Certainly the component solids
carryover flux is very sensitive to gas velocity, but although the dependency is
of the form K~h a if no single value can be assigned to the exponent which
may range from 2 to 4 depending on the systems.

if
o

c:

's::J

Expected

~ 40

Rj

I
U'1
<;f

C(.

line based on
XBj

/.

'0

..

6.6.5 OperatingPressure
As pointed out in Section 6.3.3, terminal velocities are unaffected by gas
density in the Stokes region, slightly in the transition region, and in the
turbulent regime according to V(l/Pg). This means that as the gas pressure is
increased at constant gas velocity, the size of the particles carried over will
increase slightly. More importantly, however, the rate of carryover has been
found experimentally to be directly proportional to absolute pressure (May
and Russel, quoted in Zenz and Weil, 1958).

~.

~ 20
t
o

.~~

Experimental
values

0.2

0.4

OB

0.6

Average concentration

in bed.

x Bi
Figure 6.9 Carryover rate of fines versus average concentration of fines in the bed (from
qeldart et al., 1979).

6.6.6 Bed Diameter


Few data.are available on the effect of seale-up; from a systematic study in
beds from 0.019 to 0.146 m diameter Lewis, Gilliland, and Lang (1962)
concluded that there was little change above 0.1 diameter but that the
entrainment rate increased with decreasing diameter for diameters smaller
than 0.04 m. These conclusions are based largely on data from beds of solids
only 0.1 m deep and may not reflect a general trend, through Ryan (1970)
found a similar effect of bed diameter below 0.15 m diameter. Wohlfarth
(1979), fluidized catalyst in beds 0.46, 1.0 and 4 m in diameter. His results
require careful interpretation as they are presented in dimensionless form,
and conditions in each of the three. beds were not identical. His data show that
at equal velocities and distances above the bed surface, entrainment fluxes
were approximately equal in both the 0.46 and 1 m diameter beds.

~
~
.~

gti 60
E

<1' 50
25mm

U'1

coarse in
bed "'--.;

('\I

6 40
(Xl

....~20
/

'0 10

6.6.7 Internals in the Bed and Freeboard

/ 0

a::

l/

xl

..

Gas Velocity

0.01

0.02

Average concentration

0.03
in bed

XBi

Figure 6.10 Influence of size of coarses in the bed on the


carryover rate of fines (from Geldart et al., 1981).

There are relatively few reports on the effect which baffles, in and above the
fluidized bed, have on entrainment. Geldart (1981)and Colakyan et ai. (1979)
found that the presence of tubes beneath and just above the bed surface had
no discernible effects in beds of coarse solids containing fines. Martini,
Bergougnou, and Baker (1976), working with fine sand, found that louvers
situated just above the bed surface helped to reduce carryover of the larger
fractions. If situated higher in the freeboard they could actually increase
carryover. Similarly, filling the freeboard with tubes would increase the

superficial gas v!;:locity and increase carryover. Cylindrical mesh packings in


the bed reduce carryover (due to smal1er bubbles) but in the freeboard they
increased carryover. (Tweddle, Capes and Osberg, 1970).
The most effective way to reduce carryover is to increase the diameter of
the freeboard thus reducing the gas velocity.

Yagi and Aochi (1955):


K~
d2.
p,(; :
= 0.0015 Re:,6

V:)2

6.7

CORRELATIONS

FOR K;",

In view of the number of variables and the complexity and interrelation of the
mechanisms occurring in and above the bed, it is hardly surprising that the
many published correlations predict widely different rates of carryover when
applied to systems other than those on which they are based. Also there is no
guarantee that the offtake was above TDH(F). A ful1 discussion of the
correlations is beyond the scope of this section, but Fig. 6.11 shows how much
correlations differ and a list is given in Table 6.4. Another problem which
complicates the prediction of carryover rates is that fines may be generated in
the bed due to chemical, hydrodynamic, or mechanical breakdown of the
solids; this makes it difficult to predict the concentration of any particular size
fraction in the bed, and experimental work may be required if carryover is a
crucial parameter. The results should then be compared with published
correlations, and the one which agrees most closely used for subsequent scaleup calculations.

f.-: Merrick

,.. Highley
/ j-CoIokyan

20
10

"'--

"'e

0'

*J

et 01

.' :

et 01

05
02

/:
/ ,,:

f 1.26

4.31

Wen and Hashinger


K;",
Pg(U-vt)

107 ( g
104

lr,

U2

)
)

u2

1.88

when
1.18

when

( g dpi P~

pi Pp

U2
2

g dpi Pp

<3

10-4

>3

10-4

(1960):

1.7 x 10-5 { (U ~ Vt)2}0.S


g pi

( Pp

Re~.725

Pg )

LIS

U~

Vt)U

. Tanaka et al. (1972):

K~",
~U-~

{(U -

4.6 x 10-2

Vt?

g~

l0.5 Re~.3
r

(Pp

Pg )

0.15

Merrick and Highley (1974)


K~", = A
Pg U

130 exp

{_ 10.4 (~)

Umf

0.5 (

K;",
pgU = 23.7 exp
Colakyan et

(V-

5.4ij

at. (1979):

K~", (kglm2s) = 33 ( 1 - ~

K;", = pp(l - Ej)(U - VI)


where
_
{
it. (U Ei
1+
2gD

vi}

Geldort

U2

dpi

0.25}

U - Umf

Wen and Chen (1982):

.vIi

lo

Zenz and WeB (1958):

and Toiloka

)/J.

0.01 Ret

Geldart et ai. (1979) revised (Eq. 6.34):

ond

Zenz and
'Neil
Wen and
'/Hoshi~r

/'
/

et 01

-1/4.7

and it. is obtained from:

U(m/s)

Figure 6.11 Comparison of experimental and


predicted carryover rates (from George and
Grace, 1978).

.!!:....]
Pg

2.5

15.17

{ Pg(U ~ Vt)dpi }

12.3 { Pg(U ~ vt)dpi

-1.5

} -2.5

D2 for Pg(U~

vt)dpi

D for pg(U ~ vI)dpi

2.38

.::;D
2.38

~LT

In the absence of an experimental unit of at least 0.1 m diameter, carryover


must be predicted using correlations.
For the elutriation of particles less than about 100 ~m from well-mixed
beds fluidized at velocities up to about 1.2 mis, the Zenz-Weil (1958)
correlation probably gives the most consistently realistic values of K"
particularly if the whole bed is potentially entrainable (i.e. even the largest
particles have a terminal velocity less than the superficial gas velocity). Such
materials as cracking catalyst, coal char, and other low density solids fall into
this category.
For higher velocities and larger particles, averaging K~"" predicted by the
Zenz- Weil (1958) correlation and the revised Geldart et ai. (1979) correlation
will give values which are not unrealistic:
K~"" = 23.7
-pgU

e-5 .4 vuu

(a) No feed; cyclone 100 per cent. efficient; total recycle of cyclone product
(b) Continuous feed; cyclone efficiency varies depending
recycle of fines

on dpi; partial

(6.34)

For beds of coarse solids containing recycled fines fluidized at up to 5 m/s


(e.g. fines injected into combustors, granulators, or dryers) the carryover rate
constant K;"" of the fines (vtlU < 1) can be predicted from Geldart's (1981)
correlations.
The correlation is also of the form:
K~

~=A
pgU

e-BvJU

31.4;

and for 2.5 mm coarses in the bed:

Linear interpolation of A and B can be made for solids between 1 and 2.5
mm.
High accuracy should not be expected from any of these approaches and in
the present state of knowledge agreement between experiment and prediction
is unlikely to be better than 100 per cent.
6.8

CALCULATION

OF CARRYOVER RATE

Assuming K;h has been predicted from correlations for each size range (e.g.
from Eq. 6.33a using Eo from the last equation in Table 6.1) or determined
experimentally, calculation of the carryover rate requires a knowledge of the
concentration of each size range in the bed. The general case is shown in Fig.
6.12 and many combinations are possible; for example:

Whatever the arrangement, overall and component mass balances must be


done. As an illustration, assume RE = RR = 0 and F and Q are non-zero. A
mass balance on the size fraction dpi gives:
XFiF

An overall mass balance gives:

XpiRT

+ XqiQ

(6.36)

XpiRT = K;h A

RT

= Ei0

XBi

1K;h A

(6.38)
(6.39)

XBi

and in a well mixed bed:

XFi

XB"

(6.40)

KihA +F-RT

This equation cannot be solved directly because, as Eq. (6.39) shows, RT


depends on the value of XBi for each size fraction. In practice a rapidly
converging trial and error loop can be set up, with RT = 0 for the first trial. If
the freeboard height is larger than about 2 m, and the column diameter
exceeds about 0.2 m, then for size fractions having vtlU < 0.5, K;h = K;oo, this
simplifies and shortens the calculation significantly.

There is still a long way to go before carryover rates can be predicted with real
confidence. Cold model experiments in columns at least 0.1 m in diameter can
be of assistance but there is no substitute for making measurements on the
plant. Unfortunately,
due to the different perceptions and objectives of
production and development engineers, these are never easy to make and the
data are frequently incomplete or unsatisfactory.
In addition to the influence of the quantifiable variables discussed in Section
6.6, other parameters often assume importance in a bed operating at high
temperature
and/or pressure: e.g. production of fines by mechanical,
thermal, and chemical attack on the particles at the grid, in the bed, and in
the cyclones; particle stickiness caused by deposition and softening of
impurities on the surface; blocking of cyclone diplegs; leakage around the
distributor which gives mal-distribution of gas. These other effects may occur
in many fluid beds but only become apparent when there is a serious increase
in the rate of solids carryover.

6.10

A powder is fed to a fluidized bed 0.5 m in diameter at the rate of 540 kglh.
The column, which is 3 m tall from distributor to gas offtake, contains 200 kg
of solids having a settled bulk density of 1,450 kg/m3 and a particle density of
2,630 kg/m3. The powder is to be fluidized at 1 m/s at 300C by a gas of density
0.61 kg/m3 and viscosity 2.95 x 10-5 N s/m2. The size distribution of the feed
solids is given in Table 6.5. Estimate the solids loading on the cyclone if (a)
none, (b) all of the elutriated solids are returned to the bed.

The residence time of the size fraction dpi in the bed is:
XBiMB

tR=--1
xFiF
Substituting

from Eqs (6.36) and (6.37) and remembering


MB

= PBmf A
PBmfA

tRi

KihA

EXAMPLE ON ELUTRIATION

that

Hmf

Hmf

+ F-

RT

For very fine particles:


K:h A ~ F - RT

and

K: = K;,

Sieve size
dp, ILffi

-44+60
- 60+85
That is, the residence time is independent
proportion alto the bed depth.

of the concentration

For coarse solids:


tRi

PBmfA

Hmf

and directly

- 85
-100
-120
-150
-200
-350

+ 100
+ 120
+ 150
+
+

200
350

300

215
165
137
115

90
60
(30)

3.3
22.8
16.4
20.3
16.8
13.3
5.2
1.9

XFj

Assumptions
XBi(l)

Particles are sufficiently rounded that dv


dp. The bed is well mixed and the
solids removed from the bottom of the bed have the same composition as the
bed.

Step 1.

Preliminary

e.g. for a 30 J.Lm fraction,

XPi

(b) From Eq. (2.27), Umf

XBi(l)

0.01 m/s.

(c) From data given, static bed depth


equal to Hmf

0.7 m and this is approximately

(d) We need to know the height of the freeboard when the bed is in
operation. Using the same approach illustrated in Example 4.4, the
bubble size at the bed surface deqs is calculated to be 0.5 m, which means
that the bed would be near slugging under ambient conditions. The
average expanded bed depth is calculated (Eq.4.7) to be 1.5m, giving a
freeboard height of about 1.5 m.
(e) Critical particle size corresponding to U = 1 m/s is calculated using Eq.
(6.13) and Fig. 6.7 to be 164 J.Lm, so almost 70 per cent of the solids are
potentially elutriable.

Step 2. Estimate TDH. Since most of the particles are elutriable we are
interested in TDH(F). Values calculated using equations in Table 6.1 are
listed below. Correlations of Horio and Wen and Chen require a knowledge
of bubble size at the bed surface deq,s which in this case has been calculated as
0.5 m.
TDH(F),
Fournol et ai.
Horio et ai.

102

Amitin
Zenz and Weil

6.3

K~co = 7.34 kglm2 s;

F = 540 kg/h = 0.15 kg/s; A = 0.196 m2 Thus:

125 J.Lm.

K~coA + F

= 0.0188 and

calculations:

(a) From Eq. (2.8), dp

0.15 x 0.0188
x 0.196 + 0.15

= 7.34

0.00177

This is done for each size fraction (see Table 6.7, column 3) and so the first
estimate of RT is found. This is then inserted into Eq. (6.4;) and the second
estimate of XT is found. Three trials are sufficient; it can be seen
from the equilibrium values of XBi that compared with the feed, the bed is
deficient in the finest sizes and richer in the coarser.

Table 6.6 Calculation of elutriation rate constants K:=

30

60

90

115

137

0.044

0.19

0.35

0.55

0.74

3.28

1.26

0.59

0.37

0.27

0.19

0.13

K:=, kg/m2s
(Geldart et al., 1979)

11.4

5.18

2.18

0.74

0.27

0.065

0.018

K:= (avg), kg/m2s

7.34

3.22

1.38

0.56

0.27

0.13

0.07

dpi, ~m
VI>

m/s

1(=,

kg/m2s

165

195

1 1.24

(Zenz and Weil, 1958)

3.2
4.3

Clearly the available freeboard height of 1.5 m is below TDH(F) and (a)
particles larger than the critical size (164 J.Lm) will be carried over, (b) the
, carryover rate will be larger than the value based on K~co.

Step 3. Since the powder will be well mixed and dp is less than 300 J.Lm,
, calculate K~oofor each size fraction from Zenz-Weil (1958) and Geldart et at.
(1979) revised, and take the average (Table 6.6).
Step 4. Calculate equilibrium concentration of each size fraction in the bed
using Eq. (6.40). In order to simplify the calculation it, will be assumed that
K~h = K;co. This will enable us to calculate values ofxB; Eioo can then be found
(Eq. 6.30). In the first trial assume RT = 0:

tip,
~m
30
60
90
115
137
165

XFi

XBi

E;",A,
kg/s x 10-3

XBi

Ei",A,
kg/s X 10-3

XBi

Ei",A,
kg/s X 10-3

0.0188
0.0524
0.1328
0.1684
0.203
0.164

0.00177
0.01
0.0473
0.0972
0.15
0.14

2.55
6.31
12.8
10.6
7.94
3.57

0.00183
0.01
0.053
0.117
0.191
0.187

2.63
6.31
14.3
12.8
10.0
4.76

0.00183
0.01
0.054
0.121
0.2
0.197

2.63
6.31
14.6
13.3
10.6
5.0

RT= E,.A

= 43.8 x

10-3

RT

= 50.8

10-3

RT

= 52.4

10-3

Step 5.
(a) From Table 6.7, E", :== 52.4

1 x 0.196

0.354 kg/m2s

K:~

K~h

0.35 kg/m3

mi,t

MB

Step 6. If all the solids from the cyclone are returned, the bed will have the
same composition as the feed and the loading on the cyclone is calculated to
be 0.746 kg/m3

!:i.p

Q
Ri
RT

}
6.12

entrainment flux of solids when offtake is above


TDH(F) (= !. Ej",)
kg/m2s
kg/s
rate at which fresh solids are added to bed
acceleration of gravity
9.81 m/s2
distance above bed surface
height of settled bed
elutriation rate constant for size fraction
dpj when h > TDH(F)
elutriation rate constant for size fraction
dpi when offtake is at height h
kg/m2s
mass of solids size d; carried out of bed in time t kg
mass of solids in bed
kg

g
h
Hs

Therefore total carried over = 0.354 x 0.196 = 0.0694 kg/so Of 0.15 kg/s
fed in, 0.0694 kg/s (46 per cent.) is elutriated and the solids loading to the
external cyclone is:

---

E",

(d) From Eq. (6.32) at h = 1.5 m and taking a = 4 m-1:


Eh = 0.267 + (35.4 - 0.267) e-4 x 1.5

0.0694

entrainment
(= !. Ejo)

10- /0.196 = 0.267 kg/m s.

(b) From the last equation in Table 6.1, the total ejection flux at the bed
surface, Eo, is calculated to be 35.4 kg/m2s, using deqs = 0.5 m (step Id.)

Eo
2

NOMENCLATURE

cross-sectional area of bed at surface


overall and component decay constants
Archimedes number
equivalent volume diameter of a bubble at
bed surlace
mean sieve size of a powder
arithmetic mean of adjacent sieve apertures
volume diameter
diameter of bed
entrainment flux of solids at gas offtake
(= !. Eih)
entrainment flux of size
offtake (= Xpi Eh)
entrainment flux of size
surface (= XBi Eo)
entrainment flux of size
is above TDH(F) (= Xpi

rn2
m-1

Ret
Rep
t, tRi
TDH (C)
TDH(F)
TDH(F)j

m
m or porn
m
m
m
kg/m2s

fraction dpi at gas


kg/m2s
fraction dpi at bed

U
Vt

XBi
XBio
XBit

XEi

XFi

kg/m2s
fraction dpi when offtake
E",)
kg/m2s

Xpi

flux of solids at bed surface

pressure drop across a section of the freeboard


rate at which solids a,re removed from bed
rate of removal of size fraction dpi by gas
rate of removal of solids by gas
terminal velocity Reynolds number Pg Vtdv/po
superficial velocity Reynolds number Pg Udpl po
time, residence time of size dpi
distance above bed surface required for
disengagement of all particles having Vt > U
distance above bed surface beyond which
entrainment rate becomes relatively unchanging
distance above bed surface beyond which
rate for size dpi becomes relatively unchanging
superficial gas velocity
freefall (terminal) velocity of an isolated
particle
equilibriurn concentration of size dpi in bed
initial concentration of size dpj in batch system
concentration of size dp; in batch bed after
time t
concentration of size dpi in solids leaving the
cyclone with the gas
concentration of size dpi in feed entering the
bed
concentration of size dp; in solids entering the
cyclone

kg/m2
kg/s
kg/s
kg/s

m
m
m/s

GAS FLUIDIZATION

TECHNOLOGY

concentration of size dp; in fines returned to the


bed from the cyclone
mass fraction of size fraction dpi in carryover
voidage
voidage in bed at incipient fluidization
gas viscosity
density of fluid
density of gas
solids loading of size fraction dpi in offtake gas
density of particle, including pores

kg/ms
kg/m3
kg/m3
kg solids/m3
kg/m3

Allen, T. (1981). Particle Size Measurement, 3rd ed., Chapman Hall.


Amitin, A.V., Martyushin, I.G., and Gurevich, D.A. (1968). Khirn. Tk. Top. Mas., 3
20.

Avedesian, M.M. (1974). Private commumication.


Becker, H. (1959). Can. J. Chem. eng., 37, 85.
Clift, R., Grace, J.R., and Weber, M.E. (1978). Bubbles, Drops and Particles,
Academic Press.
Colakyan, M., Jovanovic, G., Catipovic, N., and Fitzgerald, T. (1979). AIChE 72nd
Ann. Meeting, Los Angeles.
Fouda, A.E., and Capes, C. (1976). Powder Technol., 13,29l.
Geldart, D. (1981). 'Behaviour of fines in a fluidized bed of coarse solids', Electric
Power Research Institute, Palo Alto, Final Report EPRI CS2094.
Geldart, D., Baeyens, J., Pope D.J., and Van de Wijer, P. (1981). Powder Technol.,
30,195.
Geldart, D., and Pope, D. (1983). Powder Technol., 34,95.
George, S., and Grace, J.R. (1978). A.l.Ch.E. Symp. Ser., 74 (176), 67.
Heywood, H. (1962). Proc. of Symp. on Interaction between Fluids and Particles,Inst.
Chern. Engrs (Land.), p.l.
Horio, M., Taki, A., Hsieh, Y.S., and Muchi, I. (1980). In Fluidization (Eds J.R.
Grace and J.M. Matsen), Plenum Press, p. 509.
Kunii, D., and Levenspiel, O. (1969). Fluidization Engineering, J. Wiley, New York.
Large, J.F., Martini, Y., and Bergougnou, M.A. (1976). Int. Powder and Bulk
Handling Conf., Chicago.
Leva, M., and Wen, C.Y. (1971). In Fluidization (Eds J.F. Davidson and D.
Harrison), Academic Press, London.
Lewis, W.K., Gilliland, E.R., and Lang, P.M. (1962). Chern. Eng. Progr. Symp.
Series, 58 (38), 65.
Martini, Y., Bergougnou, M.A., and Baker, e.G.J. (1976). In Fluidization
Technology (Ed. D.L. Keairns), vol. 2, Hemisphere, Washington.
Merrick, D., and Highley, J. (1972). A.l.Ch.E. Symp. Ser., 70, (137),366.
Morooka, S., Kawazuishi, K., and Kato, Y. (1980). Powder Techn()l., 26, 75.

Pettyjohn, E.A., and Christiansen, E.B., (1948). Chern. Eng. Progr., 44, 157.
Ryan, W.J. (1970). Cherneca 70, Chemical Eng. Conf. Australia, Butterworh and Co.
(Australia) Ltd.
Schillar, L., and Nauman, A. (1933). Z. Ver. Dtsch. Ing. 77, 318.
Tanaka, I., and Shinohara, H. (1978). Int. Chern. Eng., 18, No.2, 276.
Tweddle, T.A., Capes, C.E., and Osberg, G.L. (1970). Ind. Eng. Chern. Proc. Des.
Dev., 9,85.
Wen. C.Y., and Chen, L.H., (1982). A.I.Ch.E.J., 28, 117.
Wen, C.Y., and Hashinger, R.F., (1960). A.l.Ch.E.J., 60,220.
Wohlfarth, W. (1979). Private communication.
Wong, A.C.Y. (1983). Ph. D. Dissertation, University of Bradford.
Yagi, S. and Aochi, T., (1955). Paper presented at Soc. Chern. Engrs. (Japan) Spring
Meeting.
Yerushalmi, J., Cankurt, N., Geldart, D., and Liss, B. (1978). A.I.Ch.E. Syrnp. Ser.,
74 (176), I.
Zenz, F.A. (1983). Chern. Engng., Nov. 28.
Zenz, F.A., and Weil, N.A. (1958). A.l.Ch.E.J., 4, 472.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

High Velocity Fluidized Beds

7.1

INTRODUCTION

Consider a bubbling fluidized bed (Fig. 7.1a). As the velocity of the gas is
slowly raised, the heterogeneous, two-phase character of the bed first peaks,
then gradually gives way to a condition of increasing uniformity culminating
in the turbulent state (Fig. 7.1b) in which large discrete bubbles or voids are
on the whole absent. In the turbulent fluidized bed, there is an upper bed
surface, though it is considerably more diffuse than in a bubbling fluidized
bed because of the greater freeboard activity attending operation at higher
gas velocities,
The turbulent regime extends to the so-called transport velocity. As the

Bubbling
fluidized
bed

Fluidizing

(0 )

gas

Turbulent
fluidized
bed

Fast
fluidized

bed

transport velocity is approached, there is a sharp increase in the rate of


particle carryover, and in the absence of solid recycle, the bed would empty
rapidly. Beyond the transport velocity, particles fed to the bottom of the
column or vessel traverse it in fully entrained transport flow, and the
concentration or density of the resulting suspension depends not only on the
velocity of the gas but also on the flowrate of the ~olids..If the ~olidsa~efed to
the column at a sufficiently high rate - e.g. by cIrculatmg solIds earned over
from the column to its bottom via external cyclones and a standpipe - then it
is possible to maintain in the column a relatively large solids concentration
typical of the fast fluidized bed condition (Fig. 7.10).
.
.
The turbulent and the fast fluidized beds which are the subject of thIS
chapter are referred to as high velocity fluidized beds. As their names suggest,
these regimes lie beyond the bubbling range, yet they presex:ve ~l the
underlying elements associated with the fluidized state (e.g. umformIty of
temperature throughout the bed's volume).
...
Fundamental research into the underlying phenomena of hIgh velOCItyflUId
beds and systematic compilation of design data lag well behind commercial
applications which are developing rapidly. This is due ~n p~t to some real
technical difficulties which stand in the way of research In thISarea. Though
experiments in small equipment may be revealing and useful, all.fl~idization
research requires, at some point, sufficiently large scale, and thISISequally
true for turbulent and fast beds as for bubbling systems. High gas velocities
and large scale fast beds also require a tall structure which may rise beyond 10
m. Nevertheless, the last few years have seen an increase in the number of
investigators attracted to this area.
In this chapter the body of existing data on high velocity fluidizing b~d~is
presented and discussed. Attention is drawn to unknowns and uncertaInties
and, accordingly, to research needs.
7.2

TRANSITION

FROM BUBBliNG

TO TURBULENT FLUIDIZATION

Lanneau (1960) appears to have been the first to recognise the transition.fro~
bubbling to turbulent fluidization and the potential advantages of operatmg m
the turbulent regime. He studied the fluidization characteristics of a fine
powder (see Table 7.1) in a bed 4.6 m deep contained in a column 7.62 cmi.d.
as the velocity of the fluidizing gas (air) was raised to about 1.7 mls. At a
given gas velocity, he recorded tra:es of in~tantaneous point de~sities from
two small-point capacitance probes Inserted m the bed spaced vertically apart
at any desired elevation. Lanneau conducted experiments at two pressure
levels, 1.7 and 5 atmospheres absolute. At both pressures, he noted that at
low gas velocities, up to about 0.3 mis, the tracings from the probes clearly
reflected the pronounced two-phase structure of the slug flow regime. As the
velocity was raised above 0.3 mis, the heterogeneous character of the bed

;;- 250

tlJ

6200

~
~
8.

150

100

50

"C

~
0.3

0.6

0.9

1.2

1.8

Superficial Gas Velocity (m/sJ

Figure 7.2 Lanneau's heterogeneity - the mean deviation of the instantaneous point density from the timeaverage bed density (from Lanneau, 1960).

gradually changed to a condition of increasing homogeneity. Lanneau


suggested that the change was accompanied by the breakdown of slugs into
smaller bubbles containing a higher proportion of solid particles. At gas
velocities beyond about 1 mis, the tracings indicated, in his words, that a
condition of 'almost uniform or "particulate" fluidization' was approached.
To represent the degree of non-uniformity of a fluid bed, Lanneau defined
the parameter 'heterogeneity' as the mean deviation of the instantaneous
point density from the time-averaged bed density (Fig. 7.2). Noting that in his
experiments the heterogeneity peaked at around 0.3 mis, he argued that this
point was the worst from the standpoint of the efficiency of contact between
gas and solid. The more uniform regime that lay beyond a velocity of 0.3 mls
promised a greater degree of interaction between gas and solid, and a higher
efficiency of contact.
Lanneau's paper seems to have gone generally unnoticed. More than a
decade passed before Kehoe and Davidson (1971), working in transparent
equipment, described the transition from bubbling (slugging, in fact) to
turbulent fluidization, and it remained for Massimilla (1975) to provide a
laboratory demonstration of the higher contacting efficiency of the turbulent
fluidized bed (Fig. 7.3).
What Kehoe and Davidson saw was a breakdown of the slugging regime
into a 'state of continuous coalescence - virtually a channelling state with
tongues of fluid darting in zigzag fashion through the bed'. They used the
word 'turbulent' to describe this state. Working with several powders and
utilizing capacitance probes, they reported that the transition to turbulence
occurred at:

158
Table 7.1

159
Transition

from

bubbling

to turbulent

Size
range,
Solids

JAoffi

Catalyst
Catalyst
Catalyst
Ballotini
Ballotini
Ballotini
Catalyst
Catalyst
Catalyst
Catalyst
Catalyst
Catalyst

B
B
B
C
C
C

Summary

fluidization;
Mean
diameter",
p.m

Particle
densit~,
em

Particle
sphericity

30-150
30-150

70
70

2.0
2.0

NA
NA

15-43
15-43
15-43
15-43
15-90
15-90
15-90
40-90
40-90
40-90

22c
22c
22c
22c
26c
26c
26c
55c
55c
55c

1.1
2.2
2.2
2.2
1.1
1.1
1.1
1.1
1.1
1.1

NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
0.4
1.0
1.0
1.0
1.0
1.0

of test conditions

Experimental
methodsb

and results

Pressure
level
(atm)

v"
cmls

Lanneau (1960)
7.5 cm i.d.
C
7.5 cm i.d.
C

1.7
5.0

27
21

Kehoe and Davidson


10 em i.d.
X
10 cm i.d.
X
5 em i.d.
C
6.2 x 0.6 cm
C
10 em i.d.
C,X
6.2 x 0.6 cm
C
10 em Ld.d
C
10 em Ld.d
X
5 em i.d.
C
10 em Ld.d
C

1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0

2.3
10.0
10.0
10.0

Massimilla (1973)
15.6 em i.d.
C

1.0

9.5

The City College studies


15.2 em i.d.
POF
15.2 em Ld.
POF
15.2 em Ld.
POF
5 x 51 em
PT
POF
15.2 em Ld.
5 x 51 em
PT
pdt
15.2 em Ld.
5 x 51 em
PT
PT
5 x 51 cm

1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0

2.3
7.8
10.6
10.6
58.2
58.2
116.0
116.0
95.0

Canada, McLaughlin, and Staub


1-10
PT,C
0.3 x 0.3 m
and
0.61 x 0.61 m
PT,C
0.3 x 0.3 m
1-10
and
0.61 x 0.61 m

Varies
with
pressure
Varies
with
pressure

Apparatus

1.6
3.2
3.2
3.2
2.3
2.3

Uc,
cmls
30
30

Uk
cmIs

Uk/V,

Uncertain
Uncertain

11.0
35.0
40.0
35.0
18.0
17.0
32.0d
44.0
50.0
~ 50.if

5.0
~ 5.0

30-40.0

3.1-4.2

53
61
91
61
122
137
274
150-215
150-215

107.0
61.0
137.0
107.0
274.0
255.0
550.0
Uncertain!
Uncertain

47.1
7.8
12.9
10.9
4.7
4.4
4.7

6.8
11.0
12.5
11.0
7.8
7.4
14.0
4.4

a
b

Unless otherwise

indicated,

Catalyst

NA

50

1.0

Dicalite
FCCHFZ-20
HFZ-20
Alumina
Alumina
Sand
Sand
Glass

~16O
0-130
0-130
0-130
40-200
40-200
80-670
80-670
105-210

33
49
49
49
103
103
268
268
157

1.67
1.07
1.45
1.45
2.46
2.46
2.65
2.65
2.42

Glass

550-750

650

2.48

1.0

Glass

2,500-2,700

2,600

2.92

1.0

0.8
0.8
1.0

FCC
FCC
FCC

0-180
0-180
0-180

60
60
60

0.93
0.93
0.93

NA
NA
NA

Catalyst
Alumina
Ludox

NA
NA
NA

60
95
60

0.94
1.55
1.4

NA
NA
NA

the volume surface mean diameter,

dp is used.

C = capacitance probes; X = x-ray photography; POF = pressure drop fluctuations;


PT = pressure transducers; V = visual.
The mean diameter was computed from 1I(0.5/smallest cut + 0.5l1argest cut).

Thiel and Potter (1977)


V
5.1 em Ld.
V
10.2 em i.d.
21.8 cm Ld.
V
Carotenuto,
PT,C
PT,C
PT,C
d

1.0
1.0
1.0

10.0
10.0
10

NA

0.5-0.65

NA

0.3-0.35

Crescitelli, and Donsi (1974), Crescitelli et ai. (1978)


15.2 em Ld.
1.0
10.0
20.0
1.0
15.2 cm Ld.
36.0
100.0
1.0
15.2 em Ld.
15.0
33.0

The bed depth was 225 cm; in all other experiments


height ranged below 270 cm.
FCC

41
22
22.5

fluid cracking catalyst.

f Owing to wall effects.

conducted

4.1
2.2
0.25

2.0
2.8
2.2
by Kehoe and Davidson

bed
'

where Vt mm
. is the terminal velocity of particles belonging to the smallest cut.
Since then, the transition from bubbling to turbulent fluidization has been
the subject of a number of investigations (Carotenuto, Crescitelli, and Donsi,
1974; Thiel and Potter, 1977; Crescitelli at ai., 1978; at General Electric,
Canada, McLaughlin, and Staub, 1976, and Staub and Canada 1978; and at
the City College of New York, Yerushalmi, Turner, and ~quires, 1976,
Cankurt and Yerushalmi, 1978, Turner, 1978, and Yerushalmt and Cankurt
1978, 1979). The test conditions, experimental methods, and the results of
these investigations are summarised in Table 7.1.

7.2.1

Transition Velocities

The transition from bubbling to turbulent fluidization is gradual and spans a


range of gas velocities which depend on the properties of gas and solids and
also on equipment scale. A consistent delineation of the stages of the
transition has not yet been established. Nor is it certain whether the transition
can be clearly defined. The City College investigators, measuring pressure
500
:

0 0

oSO/DS~D/CALlr - <1200

$Ol/D$~rCC

""-

0-

00
0

"0

'. .........

Hmf (em)
.---110
.---164

----.... .:.:. -- .. -. ....


... ------ ... _1/(1

4
Dimensionless

+ k')

o.

0"

.,

.Po.OOoCS
-0'

.,

0.25

.~.,

'"
~
~
z

~ 0.2

0.20

0.15

0.15

~
ii
...

5IS! 0.10

0.10

t!

0.05

a
(Q)

0.5

1.0

1.5

GAS VElOCI TY U (mls)

200

~~~

1.0

II

2.0

LS

GAS VELOCITY U (mil'

......

1600 _

800
0

SOL./O$" HFZ - 20

SO/.IOS~ No. 85 SANO

1200 ~

700 ~

Much of the work focused on fine powders belonging to Geldart's group A.


The City College studies also encompassed solids belonging to group B, but
the data often showed some scatter owing to strong wall effects in the form of
piston-like movement of solid and some channelling in the two-?imensio~al
bed. Still more severe wall effects precluded altogether expenments With
solids coarser than sand with a mean size of 268 f.Lm (see Table 7.1).
Larger equipment at General Electric afforded tests with coarse solids
belonging to Geldart's group D, uniformly sized glass beads 650 and 2,600
f.Lm in diameter. One of the experimental units, with a square cross-section
measuring 0.3 by 0.3 m, was designed for tests at pressures up to 10 atms.
With the work oriented towards study of fluid bed boilers, extensive tests
were also conducted with tube banks immersed in the beds.
Even a casual inspection of Table 7.1 reveals a lack of coherence and
several apparent inconsistancies; all the work to date does point to several
conclusions.

800>0.14

i!c

0.12

...

~
400

0.10

300

ti

~
g

......

0.04

>

(c)

0.6

0.4

Ei
.. o. 2

0.02
0

~o
0.8

::>

l-

II>

UI

0.06

It:

l!:l
400

. 1.0

0.08

~
j:;

..J

..J

in
l!l
Q

::>

~
g

>800 ~

~
z

500

100

Decreoslng
veloc~y

0.5

(b)

en

Increoslng
veloclly

1I Daa

2.0

300~

6
8
10
Reaction Velocity Constent (11')

Figure 7.3 Ammonoxidation of prol?yl~neto acrylo~itrile


in a turbulent fluid bed (from Masslmdla, 1973). k = k
H IU where k is the reaction velocity constant (S-I). The
uwer dashed curves are predictions based on a slug flow
model (Hovmand and Davidson, 1971).

'r~

.,

'" 0.25

400 \

It!
0.5

1.0

1.5

GAS VELOCITY U (m15)

2.0

(d)

GAS VELOCITY

10

U (m/s)

Figure 7.4 Pressure fluctuations relative to the mean pressure drop across a fluid bed of (a)
Dicalite, (b) fluid cracking catalyst (FCC), (c) HFZ-20, and (d) sand. The upper curve gives
the corresponding mean pressure drop across the bed.

fluctuations, observed that the transition occurs between two velocities: Ue,
the velocity at which the pressure fluctuations peak, and Uk, the velocity at
which the fluctuations, having decayed from the peak value, begin to level
off. Uk marks the end of the transition and the onset of the turbulent regime.
Results for four solids are illustrated in Fig. 7.4. The data were obtained in
a 15.2 m diameter column incorporated in the apparatus shown schematically
in Fig. 7.5. Figure 7.4 gives the relative pressure fluctuations - i.e. the
fluctuations of the pressure drop across the bed divided by the mean pressure
drop - as a function of the superficial gas velocity. Corresponding to each
data point, the figures also display the fluidized density reckoned from the
mean pressure drop across the bed and the bed depth.
These relative pressure fluctuations are akin to Lanneau's heterogeneities,
but apart from noting the approach to an almost uniform fluidization in the

15.2 em 1.0.

BED

34.3em 1.0.
SLOW BED

range of 1 to 1.7 mis, Lanneau does not recognize any clean onset of a
different regime.
Like the City College investigators, Canada, McLaughlin, and Staub (1976)
characterize the transition in terms of the relative pressure fluctuations and
similarly mark the onset of the turbulent regime, Uk, as the velocity at which
the fluctuations begin to level off. With the exception of Thiel and Potter the
remaining investigators cited in Table 7.1 report only a single transition
velocity while stressing the progressive character of the transition. Kehoe and
Davidson note that the transition to turbulent fludization was chosen rather
arbitrarily as 'the velocity above which it was impossible to detect slugs' on
traces from capacitance probes. In a similar fashion, Massimilla, Carotenuto
et ai., and Crescitelli et ai., report the velocity at which the traces from
capacitance probes no longer exhibit any strong and clear periodic element as
the onset of the turbulent regime.
The decay of the relative pressure fluctuations from their peak, on the one
hand, and the gradual loss of a strong periodic component in the traces from
capacitance probes, on the other, reflect essentially the same physical
phenomena, the same progressive changes in the structure of the bed. For this
reason, I have listed in Table 7.1 under Uk all the transition velocities
reported by Kehoe and Davidson, Massimilla, Carotenuto et ai., and
Crescitelli et ai. I have similarly listed the results of Thiel and Potter although,
apart from stating on the basis of visual observations that the transition
velocity marked the breakdown of slugs into smaller bubbles, these authors
are vague on precisely how the transition was determined.
It is apparent marking off Uk from either the relative pressure fluctuations
or the traces from capacitance probes entails a large degree of arbitrary
judgement. Canada, McLaughlin, and Staub, plotting the superficial gas
velocity UIE versus U (Fig. 7.6), noted a clear change of slope near the
transition velocity. A different approach was taken by Kehoe and Davidson
and Crescitelli et ai. Both utilized capacitance probes to deduce the bubble or
lean phase velocity in narrow beds of fine powders undergoing transition from
slugging to turbulent fluidization. The results are shown in Figs 7.7 and 7.8. In
both cases, the experimental results are compared to the familiar expression:
UA

U - Umf

+ O.35(gD)112

(7.2)

arising from the two-phase theory. Crescitelli et ai. also compare the results
with:

Figure 7.5 Schematic of the City College 15.2 cm


fluidization system.

In both cases, the actual lean phase velocities exceed the rise velocity
predicted by the two-phase theory (a point to which I shall return presently),
and there is a sharp change in slope at Uk' However, there is one essential
difference between the results of Kehoe and Davidson, and those of Cresitelli

~
E

o
<100
:J

50
U - Urntlcm/s)

0.5

1.0
Superficial

Figure 7.8 Lean phase velocities in a fluid bed undergoing


transition to turbulent fluidization (from Crescitelli et al.,
1978).Experiments were conducted with Ludox catalyst (see
Table 7.1).

1.5

Gas Velocity,

U(m/s)

Figure 7.6 Correiation of the expansion of a bed of 650 /Lmglass particles


at a pressure of 10 atm (from Canada, McLaughlin, and Staub 1976).

et at. The former show an increase in the lean phase velocity past Uk; the
latter show a decrease. An inconsistency? Or are the two transition velociti~s
somehow dissimilar?

7.2.2

The ratio Uk/Vt

The last column in Table 7.1 shows the ratios of the transition velocity Uk to
the terminal velocity Vt of a particle of size dp The ratios are strongly
dependent on the particle's size and density. The- ratios approach and even
exceed 10 for the finer solids, and decrease sharply with particle size and
density. For the coarse glass beads tested by Canada, McLaughlin, and Staub,
Uk/vt falls below unity and lies in the range of 0.5 to 0.65 for the 650 fJ-m
glass, and 0.3 to 0.35 for coarser 2,600 fJ-msolid. The significance of these
results is discussed later.
7.2.3 Effect of particle size and size distribution
20

10

20

30

40

U - Urn! (cm/s)

Figure 7.7 Lean phase velocities in a fluid bed undergoing transi~ion to


turbulent fluidization (from Kehoe and Davidson, 1971). Expenments
were conducted in a 10 cm column with catalyst B (see Table 7.1).

Although the results summarized in Table 7.1 do not lend themselves to


correlations, they clearly indicate that both Uc and Uk increase with particle
size and density. Little information, however, is available on the effect of
particle size distribution. Kehoe and Davidson added their finest catalyst A to
catalyst C to obtain catalyst B, and noted that as a result the transition

velocity was lowered significantly. This is not surprising since the mean size of
catalyst B is considerably smaller than that of catalyst C. Information on the
effect of size distribution for solids having comparable mean size is lacking.
7.2.4 Effect of bed height
Canada, McLaughlin, and Staub report no apparent effect of bed height over
a range of 25 to 70 cm. Kehoe and Davidson, on the other hand, report that
(for catalyst Band C) a bed 2.55 m in height gave rise to transition velocities
higher than those recorded in beds of the same solids ranging in heights to
only 1.7 m. The shallower beds, they argue, support a greater degree of slug
coalescence and the approach to the 'breakdown' velocity is accordingly
faster.
Lanneau conducted his experiments in a bed 4.6 m deep, and he noted that
the structure of the bed could vary significantly with bed height. When
operating at 5 atm and at a gas velocity of 0.15 mls the pressure traces showed
that the bubbles forming immediately above the distributor were still small.
In the middle of the bed, owing to coalescence, the bubbles had grown large
and the two-phase structure of the bubbling (slugging) regime dominated, but
in the upper section of the bed the bubbles appear to have become smaller
again. The implication that the structure of a fluid bed can vary significantly
with height, especially in tall beds, ought to be taken into account in reactor
modelling.
7.2.5 Effect of pressure
Canada, McLaughlin and Staub report that, at the same gas velocity, the
fluidization grows smoother, as reckoned from the relative pressure fluctuations, as the pressure is raised. For a given solid, the transition velocity Uk
decreases with increasing pressure, but over the range of 1 to 10 atm the
ratio of Uk to the particle's terminal velocity VI remained essentially constant,
0.5 to 0.65 for the 650 Momglass and 0.3 to 0.35 for coarser 2,600 p.m solid.
Lanneau also notes that at a given gas velocity, the fluidization is more
uniform at the higher pressure level. This is reflected in the lower
heterogeneity measured past the peak at 5 as compared to 1.7 atm (Fig. 7.2).
Other workers have also noted the 'smoothing' effect of pressure,
particularly in fluidized beds of group A powders.

drawn between the data from the 15.2em diameter column and that from the
two-dimensional bed. In the latter, the transition generally came into play at
lower gas velocities. The two-dimensional bed is comparable in crosssectional area to the 15.2 cm column, but as it affords growth bubbles larger
that 15.2 cm in that respect it could be regarded as effectively larger.
That bed size may affect the transition in question should not be surprising.
Essentially all the experiments summarized in Table 7.1 were conducted in
equipment of relatively small size which gave rise to slugging behaviour over
the range of gas velocities preceding the transition and overlapping its earlier
stages. Over that range, bed size directly affects the shape, size, and rise
velocity of the slugs, and the fluidized density at a given gas velocity.
Equation (7.2), which is a statement of the two-phase theory for the slugging
regime, is a reminder of that. Since the transition to turbulent fluidization
involves changes in the structure of the bed, it may be expected to show some
dependence on bed size. In beds of sufficient size, where freely bubbling
behaviour prevails, the transition to turbulent fluidization ought to be no
longer influenced by bed size.
The dependence of bed size of the results described here provides a clue to
one of the underlying elements to which the mechanism of the transition must
ultimately be linked - the absolute rise velocity of the bubble (or slug) UA'
All the investigators who have commented upon the transition's mechanism
agree that it entails a breakdown, gradual or otherwise, of the slugs into
smaller bubbles or voids. Clearly, the size the slug attains, its rise velocity,
and its interaction with the emulsion just before the transition must be the
focus of those seeking to understand the forces that drive the transition and
the attendant changes in the bed's structure.
Invoking the two-phase theory, one may gather from Eq. (7.2) that over
the slugging range two different powders in the same bed will give rise to slugs
of comparable size and speed at a given superficial gas velocity. The process
of slug disintegration would accordingly commence earlier for the lighter
and/or smaller powder. This generally accords with the study of Clift, Grace
and Weber, (1974) on the splitting of bubbles.
On the other hand, given two beds of different size holding the same solid,
the transition will commence at a lower gas velocity in the larger because it
supports a larger slug rising at higher velocity. The results of Thiel and Potter
(1977) corroborate this point.
It should be noted that although the results of Kehoe and Davidson do
show a dependence on bed size, the dependence is slight.

7.2.6 Effect of bed size


The results of Kehoe and Davidson, and especially of Thiel and Potter,
indicate that for a given solid the transition velocity Uk decreases with
increasing bed size. The City College results corroborate this if comparison is

7.2.7 Bed expansion


Bed expansion affords more insight into the mechanism of the transition to
turbulent fluidization. If the two-phase theory holds and slug flow persists,

the bed expansion should obey:


H

1+

-=

Hrnf

U - Urnf
1/2
0.35(gD)

(7.4)

Massimilla (1973) reports data which strongly deviate from Eq. (7.4)
beginning with a superficial gas velocity around 15 cm/s (Fig. 7.9).
1.8

I
I

I
I

1.4

I
0
8
o D.
I
I o 0r:t o

1011.

1.2

1
0

Figure 7.9

1.6

'lee

0
0

0
-0

r/rI'.

Hmf
110Cm
164 em

0.4

Bed expansion

data (from Massimilla,

1973).

Massimilla attributes the departure from ideal slug flow behaviour to. slug
velocities exceeding the rise velocity predicted by Eq. (7.1), owing to the
acceleration effect brought about by coalescence of slugs throughout the
column. Above the transition velocity (30 to 40 cm/s), the higher velocity of
the lean phase, he argues, arises from the continuous process of slug
breakdown and the coalescence and rearrangement of the resulting small
bubbles or voids into new slugs. This picture of higher lean phase velocities
accords well with the results of Kehoe and Davidson and Crescitelli et at.
presented earlier (Figs 7.7 and 7.8). An equally plausible explanation is the
expansion of the dense phase which affords higher interstitial flow and
smaller slugs or voids.
7.2.8 Mechanism
All this does not quite add up to a complete picture of the transition to
turbulent fluidization, but several elements of the mechanism do emerge
more clearly. These I attempt to recapitulate briefly below.
The transition to turbulent fluidization involves the gradual breakdown of
large bubbles (or slugs) into smaller bubbles and voids. The process of bubble

splitting is to some extent counterbalanced by coalescence of the resulting


smaller bubbles, but the net effect is a progressive change towards a structure
of greater homogeneity, culminating in the turbulent state where on the
whole large discrete bubbles are absent, and traces for capacitance probes no
longer register a clear periodic element, To be sure, a relatively large void or
slug will occasionally form, travel some distance up, and soon disperse, This
and the recurring process in which the small voids split, coalesce, and
accelerate do lend the phenomena some apparent periodic character. To one
viewing a turbulent fluid bed this impression may be particularly strong when
the bed is narrow and where accordingly the voids tend to segregate towards
the centre, leaving the wall region fairly dense, and when the bed material is
coarse.
Over the transition and well into the turbulent regime, bed expansion falls
below the levels expected from ideal (two-phase) slug behaviour. This is due
at least in part to the high lean phase velocities measured over the same
range. In turn, the high lean phase velocities (which also depart from ideal
slug flow behaviour) appear to arise from the swift acceleration which attends
the recurring coalescence of slugs, in the early stages of the transition, and of
the smaller voids towards the end and past the transition.
The observed bed expansion also arises from the expansion of the dense
phase. This and the active process of splitting and coalescence of voids also
contribute a higher proportion of solids in the lean phase.
The phenomena that underlie the transition from bubbling to turbulent
fluidization remain rather obscure. The key event of course is the breakdown
of the bubbles or slugs, but although this matter has received some attention
(Harrison, Davidson, and de Kock, 1961;Clift and Grace, 1972; Clift, Grace,
and Weber, 1974) no clear and acceptable explanation has been advanced on
why bubbles break. Some evidence was offered earlier showing that the point
or range around which bubbles begin to break may be related, for a given
solid, to some critical value of the absolute rise velocity of the bubble.

The structure of a turbulent fluid bed depends strongly on the transition


velocity Uc' If Uc is small, as in beds of small particles, the sizes of the bubbles
(slugs) attained before the onset of the transition is relatively small, and the
scale of demixing of gas and solid in the resulting turbulent bed is small. In
beds of coarse solids, on the other hand, Uc is relatively high, the size of the
bubbles attained before the transition commences is correspondingly larger,
and the structure of the resulting turbulent bed is more open, marked
by wider channels and by a grosser scale of the demixing patterns of gas
and solid.

These observations are borne out by photographs of the transition in twodimensional beds presented by Kehoe and Davidson (1971).
Consider a fluidized bed of closely sized coarse particles, such as those used
in the experiments of Canada, McLaughlin, and Staub (1976). The turbulent
regime for the 2,600 p,m glass extends for around (0.3 to 0.35)vt up to a
velocity at least equal to Vt. Over this range, since the gas velocity lies below
the terminal velocity of the individual particle, no carryover occurs provided
the freeboard is sufficiently high. In a bed of fine powder, however, the
turbulent regime start at values of Uk/vt much larger than 1, and the ratio
increases as the mean size of the powder decreases (Table 7.1).
The concept that particles combine into loose clusters can be used to
explain the range of the turbulent regime for a fine powder. The mean
effective size and density of these clusters at any gas velocity is such that their
terminal velocities remain greater than the velocity of the gas. This concept
can explain why a bed of fine solid can be maintained at gas velocities that are
10 to 20 times the terminal velocity of its median particles. Entrainment will,
of course, take place as particles are swept from the lean phase near the top of
the bed, and by the erosion of clusters by the surrounding lean phase.
The turbulent regime extends from Uk to the transport velocity UTR. The
approach to the transport velocity is accompanied bya sharp increase in the
rate of carryover. It is difficult, without having made suitable measurements,
to associate the transport velocity with some mean cluster size. By analogy to
the behaviour of beds of closely sized coarse particles, one may surmise that
when the transport velocity is reached, the terminal velocities of a large
proportion of the clusters are smaller than the velocity of the gas. For fine
solids, the transport velocity is typically an order of magnitude greater than
the terminal velocity of the individual median particle. For coarse particles,
the transport velocity lies closer to the terminal velocity of the median
particle.
Though they differ in structure, the turbulent and the bubbling fluid beds
. are similar in several respects. Over the turbulent regime, though particle
carryover rates are relatively high, the fluidized density of the bed remains
quite high and an upper bed level is present, though it is, of course, more
diffuse than in a bubbling fluidized bed. As in the bubbling bed, when solid is
fed to the turbulent bed at a rate matching the saturation carrying capacity,
the concentration of solid above the bed 'decays through the transport
disengaging height and approaches a constant value. If solid is fed to the bed
at a rate above the saturation carrying capacity, the bed level rises until the
new solid rate is matched by solid carried out of the bed. The gas exit now lies
within the transport disengaging height. However, unless the slip velocityi.e. the relative velocity of gas and solid - changes significantly, the fluidized
density of the bed itself remains virtually the same.
Some of the general points made above may be illustrated by experimental
results reported by Yerushalmi and Cankurt (1979)for fluid cracking catalyst

SOllO=FCC

~~30

\
GAS
VELOCITY
U (m/sl.
1.2 \

I'

SOLID RATE

G. (ka Im2_

21.'
0

~13.2

2
4
6
8
HEIGHT ABOVE THE BOTTOM OF 15.2 em BED (m)

(FCC). The experiments were conducted in the 15.2 cm column of the


apparatus shown in Fig. 7.5. Figure 7.10 displays three vertical pressure drop
profiles recorded across the 15.2 diameter column. Each profile corresponds
to a different solid rate. Each data point represents the pressure gradient
between two adjacent pressure taps. The positions of the pressure taps are
indicated along the horizontal axis. At 1.2 mis, the saturation carrying
capacity is about 13 kg/m2s. The corresponding pressure drop profile reflects
the presence of a dense bed (of density around 200 kglm3) at the bottom of
the column and the decay of solid concentration along the transport
disengaging height. The profile also indicates that the solid hold up begins to
level off from about the column's midpoint. As the solid rate is raised beyond
the saturation carrying capacity, the upper bed level rises, and the solid
concentration near the top of the bed increases, until a new equilibrium
condition is reached at which solid leaves the 15.2 cm column at a rate
matching the new feed rate. Figure 7.10 also indicates that the bed density
increases slightly with the solid rate. This arises from the fact that
corresponding to each bed density and solid rate, one may regard the hed as
circulating at a mean solid velocity that is given by the ratio of the solid rate
Gs to the bed density fJB. This mean solid velocity increases with the solid
rate, and accordingly the slip velocity, upon which the bed density ultimately
depends, decreases with the solid rate, giving the somewhat higher bed
densities.
7.4 THE TRANSPORT VELOCITY AND THE FAST BED REGIME
The transport velocity may be regarded as the boundary which divides
vertical gas solid flow regimes into two groups of states. Below it lie the

SOL/ONCC

TIME

,.'--.

(SECONDS) 400

\
0.5

1.0

1.5

SUPERFICIAL GAS VELOCITY

U (m/sl
Figure 7.11 Time required for the pressure drop across a
bed of fluid cracking catalyst (FCC) maintained in a 15.2 cm
column to decay to 2.5 cm of water as a function of the
superficial gas velocity.

bubbling and the turbulent fluid beds where, save for some carryover, the bed
in general experiences no net flow and remains at the bottom of the holding
vessel. These are, to use Lanneau's description, the captive states. Above lie
the transport regimes which encompass a wide range of states from dilute
phase flow to the fast bed condition.
The.transport velocity for the fluid cracking catalyst lies around 1.2 to 1.5
m/s (which is nearly 20 times the terminal velocity of a single particle with a
size equal to the mean diameter of the powder, 49 /Lm). As this velocity is
approached, there is a sharp increase in the rate of particle carryover and in
the absence of solid recycle the bed would empty rapidly. This may be
illustrated by the results, shown in Fig. 7.11, of a simple experiment in which
the flow of recycle solid to the 15.2 cm diameter column was suddenly halted,
and the time it took the bed pressure drop to fall to 2.5 cm of water gauge
(w.g.) was measured at different gas velocities. Because of the continuous
nature of the curve in Fig. 7.11, it could hardly serve to pinpoint the transport
velocity. To do that, other approaches, to be described shortly, proved more
useful.
Beyond the transport velocity, solids traverse the column in fully entrained
flow, and the density or concentration of the resulting suspension depends not
only on the gas velocity but also on the solid rate. If the solid rate is low,
dilute phase flow results (Fig. 7.12). As its name is meant to suggest, solid
concentration is low in the dilute phase flow regime, the particles stream
upwards in relatively straight paths, and slip velocitieslie in the neighborhood
of the fIee-fall velocities of the individual particles. As the solid rate

Figure 7.12 The concentration of a gas/solid suspension as a function of the solid rate.

increases, the suspension becomes progressively denser (Fig. 7.12b), and at a


sufficiently high solid rate, the fast fluidized bed (Yerushalmi, Turner, and
Squires, 1976) is established (Fig. 7.12c). The fast bed condition is marked by
relatively high solid concentrations, aggregation of the particles in clusters
and strands which break apart and reform in rapid succession, extensive
backmixing of solids, and slip velocities that are in order of magnitude greater
than the free-fall velocities of the individual particles. As noted earlier, a way
to create the dense suspension typical of the fast bed condition is simply to
circulate the solids emerging from the top of the column back to its bottom by
means of cyclones and a standpipe (Fig. 7.1c).
Whereas the distinction between dilute phase flow and the fast bed
condition is clear-cut, the suspension density or solid rate at which the fast
bed may be said to come into play (at a given gas velocity) has not been
established. As in the bubbling fluid bed, the extensive solid backmixing in
the fast bed is responsible for the uniformity of temperature throughout the
bed. The study of thermal communication may serve perhaps to delineate the
onset of the fast bed at a given gas velocity. For the cracking catalyst,
extensive backmixing of solid was noted over a range of gas velocities at a
loading around 40 kg/m3
Figure 7.13 presents fluidization data for FCC obtained in the 15.2 cm
column, mostly over the fast bed range. The figure gives the pressure gradient
measured across the middle section of the column as a function of the solid
rate at different gas velocities. Over this range, wall solid friction is relatively

HIGH VELOCITY FLUIDIZED BEDS

175

40

GAS VELOCl'!.V U
2.2

S(/(.IOSFCC

(lIl/s)

4.0

35

30
>

I-

U
0
..J

AP

-zL

(Kt/m31

2.16

25

5.5

>
..J

GAS VELOCITY

j 20
a::
w
I-

U (m/sl

....

>
I-

15

0
..J
W

> 10
a.
::::;

CII

50

100

ISO

200

2SO

SOLID RATE Gs (Kt/nr-SI


Figure 7.13 Pressure gradient versus the solid rate at different gas
velocities. Data were taken with the fluid cracking catalyst (FCC) in
the 15.2 cm column shown in Fig. 7.5 across a section extending from
3.14 to 6.28 m from the bottom of the column.

small (Van Swaaji, Burman, and Van Breugel, 1970;Turner, 1978), and the
pressure gradient can essentially be regarded as the fluidized density p in the
section of the bed in question. Figure 7.13 attests to the high solid
concentrations that can be maintained in the fast bed at gas velocities that are
an order of magnitude greater than those normally employed in a bubbling
fluidized bed of the same solid.
The fluidized density is not uniform along the height of the fast fluidized
bed. The density is normally higher at the bottom, owing in part to
acceleration effects, and decays towards the top. In beds of light powders
(such as the fluid cracking catalyst), the decay is very gradual. In beds of
coarser or heavier solids the decay could be rather sharp.
Slip velocities are high in the fast bed. Those corresponding to the data
shown in Fig. 7.13 and normalized by dividing by Vt for the mean size of FCC
are given in Fig. 7.14. The slip velocity used in Fig. 7.14 is a mean value based
on the mean solid velocity.

100
SOLID RATE

150

200

250

Gs (kg/m'-sl

Figure 7.14 Slipvelocities corresponding to the data shown


in Fig. 7.13.

Slip velocity =
where e is the bed voidage and
defined by:

Vs

Gs

Vs

- -

Vs

is the mean solid velocity;

=p; =

Vs

is turn is

Gs
pp(1 - e)

The large slip ratios can be explained by the aggregation of the solid in the
fast bed into relatively large dense clusters of particles. A large cluster would
naturally have an effective free-fall velocity that is considerably greater than
that of a single particle in isolation. If a cluster is sufficiently large, it cannot

be sustained by the rising gas; it will fall back and will subsequently undergo
disintegration by one mechanism or another - hence the apparent high
degree of solid backmixing in the fast bed.
7.4.1

Effect of bed size

In Fig. 7.15 the data far FCC obtained in the 15.2 cm column (Fig. 7.13) are
compared with similar data obtained in a 7.6 cm diameter column
(Yerushalmi, Turner, and Squires, 1976). There is obviously, a discrepancy
between the two sets of data. At a given solid rate, a higher gas velocity must
be employed in the narrower 7.6 cm bed to register the same pressure drop.
Two factors may contribute to the difference: wall phenomena and wall solid
friction. The results of Van Swaaji, Burman, and Van Breugel (1970) and
Turner (1978) indicate that friction effects, being relatively small, could not
alone be the cause of the observed discrepancy, but wall phenomena could.
Segregation of solid along the wall has been observed in the narrower 7.6 em
column to a degree clearly greater than in the 15.2 cm unit. With the core

Gas velocity, U (m/5I


/1.77

I
I
I
I

somewhat more dilute, gas in the 7.6 cm bed can stream relative to the solid at
higher velocities than in the 15.2 em bed, for the same solid rate and solid
hold up. That is, slip velocities are higher in the 7.6 cm bed. In the extreme
case, if the entire solid segregated along the wall in a dense layer, leaving the
core empty, slip velocities would be even higher. The bed diameter at which
wall phenomena ceases to exert a significant influence, for a given solid, has
not been determined.
In principle the transport velocity should not be dependent on bed size.
However, in laboratory experiments, it might appear to be dependent to the
extent that wall phenomena influence the structure of the bed in narrow
columns and in tests with coarse or heavy particles.

7.4.2

There is essentially -no fundamental information on the effect of pressure on


the phenomena discussed here. Pressure may exert its well-known smoothing
effect, and in one known instance involving highly concentrated suspensions
maintained at velocities around 7 to 10 mis, this has indeed been reported to
be the case (HorsIer, Lacey, and Thompson, 1969). Pressure may also
diminish the degree of aggregation in the fast bed, and thus reduce the extent
of solid backmixing. By stability considerations Grace and Tuot (1979) verify
the concept of cluster formation in the fast bed, and their theory predicts that
pressure should lead to a reduction in clusters size. Similar arguments lead to
the speculation that the transport velocity will decrease with pressure.

7.4.3

I
I
I

50

100

150

200

Solid Rate, Gs (kg/m2 51

Figure 7.15 Comparison between the data for FCC'


obtained in the 15.2 cm column and shown in Fig. 7.13 and
corresponding data for the same solid obtained in a 7.6 em
column.

Effect of pressure

The flooding point

Following the practice of students of vertical gas liquid flow, Yerushalmi and
Cankurt (1979) used the data shown in Fig. 7.13 to prepare plots of the
volumetric gas flux, in cubic metres per square metre, versus the volumetric
solid flux at constant bed densities. The results are shown in Fig. 7.16. Note
that U is in fact the volumetric gas flux and (1 - e)vs is the volumetric solid
flux G/pp; (1 - e) is the volumetric solid concentration.
The curves of Fig. 7.16 are straight lines converging, with one exception, on
the point defined by U = 1.37 and vs(1 - e) = 0.02 m/s. The line
corresponding to p = 32 kgim3 does not converge on this point probably due
to predominance of wall phenomena at this relatively low suspension density.
The intersection is the so-called flooding point (Wallis, 1969), the onset of
concurrent flow of both phases, and U = 1.37 m/s is acordingly the transport
velocity. The flooding point will always have a positive coordinate for
viI - e) because of particle carryover from a turbulent bed. Indeed, Vs (1 ) = 0.02 m/s reflects the fluidized density (about 192 kg/m3) and the

lis" 19.5Kt' S

Figure 7.16 Volumetric flux of gas U versus the solid flux V. (I-E) at constant bed densities.
Drawn from data shown in Fig. 7.13.

saturation carrying capacity (about 20 kg/m~) from a turbulent bed


maintained just below the transport velocity.
Determination of the flooding point thus provides one method for fixing
the transport velocity.

Further experiments to delineate the boundary between the turbulent and the
transport regimes (i.e. the transport velocity) were conducted in the modified
version of the 15.2 cm diameter system shown in Fig. 7.17. The U-tube solid
transfer line, connecting the standpipe ('slow bed') and the 15.2 column, was
replaced by a 60 (from vertical) transfer line also measuring 15.2 cm i.d.
Control of the solid flow in the modified system was afforded by a conical
valve installed at the entrance from the slow bed to the lateral transfer line.
In a typical experiment, air was supplied to the 15.2 cm column through the
ports located at its bottom. At the start of the experiment, the solid control
value was closed; then, with the velocity of the gas set at some desired value,
the solid rate was slowlyraised in small increments. After each increment, the
character of the gas solid flow along the column was observed and the
pressure drops across the various sections of the column were recorded.
Attention was primarily focused on two sections: the bottom section of the
15.2 cm column below the point where solids from the slow bed enter and the
section of the column between the pressure taps numbers 1 and 2 in Fig. 7.17.
We shall concentrate on section 1-2.

UQI < UQz< < UTR<'"

0'

()

ta.P

':
1-2

UQs

."

~L

< Uge

~~<~/~CJ2
\
: .~~ /8 Ua4
/Ugi

/:~ ..
: A':

..

UTR

.',

: \

~~

:p \

~\.QCo

Figure 7.18 Qualitative representation of the pressure gradients measured in


the modified 15.2 cm system between pressure taps 1 and 2.

HIGH VELOCITY FLUIDIZED BEDS

It might be useful to consider the results first in qualitative terms as shown


in Fig. 7.18.'Pressure drop is shown as a function of the solids rate at different
gas velocities both lower and higher than the transport velocity UTR. Consider
an experiment conducted at a gas velocity Ug3; that is at a typical velocity
below the transport velocity. So long as the solids rate remains below the
saturation carrying capacity, solids will traverse the column in dilute phase
flow, and the resulting low pressure drop would correspond to some point
along the curve GA. At point A, the solids rate becomes equal to the
saturation carrying capacity, and a small increase in the solid rate beyond that
produces the collapse of the solid into a fluid bed which forms the bottom of
the column. The bed level will stabilize at some point above or below the
pressure tap numbered 2 in Fig. 7.17. In the latter case, an additional small
increment in the solids rate can bring the bed's upper surface past pressure
tap 2.
The density of the fluid bed at this point corresponds to the pressure
gradient at point A' of Fig. 7.18. The curve A'B shows further increase in the
pressure drop, and hence in the fluidized density, with increasing solid rate.
This comes about since, at a given gas velocity, an increase in solids rate
produces a corresponding increase in the mean solids velocity, and hence a
decrease in the slip velocity upon which the fluidized density depends.
If the relative velocity, Ug3 - vs, between gas and solid is smaller than Uk'
the bed will be bubbling or, in a column of small diameter, slugging, and
pressure drop measurements will show the characteristic fluctuations attending this condition. In this case, the pressure gradients represent mean values.
It should be appreciated that at a gas velocity of Ug3 the solids present
between taps 1 and 2 - ie. at the bottom half of the column - either in the
form of dense fluid bed or a dilute suspension, cannot assume densities
corresponding to pressure gradients between A and A'. As noted above, at a
solid rate around the saturation carrying capacity, it is possible to maintain
the bed level between taps 1 and 2, but pressure drops would not then
represent the bed density.
A slight increase in the solid rate would cause the bed level to rise. A slight
decrease would cause the bed level to fall and before long it would 'empty',
leaving dilute phase flow corresponding to point A in Fig. 7.18. In general, at
all gas velocities below the transport velocity, the solid present in roughly the
bottom half of the column cannot assume stable configuration with densities
corresponding to the area bounded by the vertical axis and the semicircular
dashed curve.
If one repeats the experiment at a gas velocity around the transport velocity
UTR, no apparent discontinuity in the gas solid states will occur. Instead, the
solid concentrations in the section under observation will increase with the
solid rate, and this will be reflected in the pressure gradient across this
section. The same holds for all gas velocities above the transport velocity.

181

400
SOl/OS
350

FCC

'0.64

300.
GAS VELOCITY

,-

(:~t2
(kg

1m3)

U (Ill/,l

.~~~

250

~.-1.3

.'
.,

200

1.5

0
0

0
0

0
0

150

100

50

010203040506010

SOLID RATE

Gs Ikg/1ll2-sl

Figure 7.19 Pressure gradients measured in the modified 15.2


cm system between points 1 and 2. Solid = fluid cracking catalyst
(FCC).

Figure 7.19 gives the pressure gradients measured between taps 1 and 2
using FCC. During the experiments the solid rate suffered continuous
exursions which introduced a scatter in the data. Notwithstanding, the results
shown in Fig. 7.19 support the earlier findings that for FCC, the transport
velocity lies around 1.5 m1s and that the corresponding saturation carrying
capacity GTR is in the region of 20 kg/m2 s.
Figure 7.20 shows the corresponding data for HFZ-20, and indicates that
the transport velocity is about 2.1 mls and that GTR lies around 34 to 39
kg/m2s. Table 7.1 shows that HFZ-20 is similar in size distribution to FCC
but its particle density is about 50 per cent greater.
'
Experiments with fine alumina (dp = 103 p.m, Pp = 2.46 g/cm) reported
elsewhere (Graff and Yerushalmi, 1978) indicate that for this solid the
transport velocity lies around 3.7 to 4 m/s.
The few values of transport velocity presented in this section obviously do
not afford a basis for any correlations, though, like the other transition

182

GAS FLUlDlZA TION TECHNOLOGY

600
0

SOLIOS~HFZ-20

1.0
0

500

1.34

'"
'" '"

400
( :~ }1-2

,GAS
1.65

VELOCITY U Imlsl

2.16

"

300

It.

"

(kg/mil

'I

U~

200

100

./

to rise until it reaches a point where a new equilibrium condition is


established, i.e. until the rate at which solid leaving the pipe becomes equal to
Gs The character of the fluid bed, which has now acquired some net upward
flow, will depend on the relative velocity U - vs' If the relative velocity is
smaller than Uk, the bed will be slugging. If the relative velocity lies beyond
Uk, the bed will be a turbulent fluid bed. In the latter case, a further decrease
in the gas velocity will bring about slugging conditions when the relative
velocity becomes smaller than Uk'
Consider now the case in which the solid rate Gs is greater than GTR In this
case, a continuous decrease in the gas velocity will not produce any collapse
of the suspension. Instead, as may be gathered from Fig. 7.18, the suspension
will grow denser and denser and will gradually assume the condition typical of
the fast fluidized state, with considerable solid refluxing in effect. When the
relative velocity becomes smaller than the transport velocity, the solid will be
transported through the line in the form of a turbulent fluidized bed, and only
when the relative velocity becomes smaller than Uk will choking come into
play in the form of slugging, with its undesirable pressure fluctuations.

80

7.6
Figure 7.20 Pressure gradients measured in the modified 15.2 cm system
between points 1 and 2. Solid = HFZ-20.

velocities, Uc and Uk, the transport velocity UTR with particle size and
density.
The phenomena of choking, described in the pneumatic conveying
literature, is associated closely with the typical experiments described above
(Yang, 1976). To the practising engineer, choking is an undesirable clear-cut
event in which a stable smoothly flowing suspension collapses into a slugging
condition with the attendent large pressure fluctuations (Zenz and Othmer,
1960).

In a typical experiment aimed at determining the choking conditions (i.e.


the choking velocity and the solid concentration at a given solid rate) the solid
rate is set at a given value, and the velocity of the gas is then slowlydecreased
until choking occurs. Reducing the gas velocity at a fixed rate amounts to
moving vertically upwards in Fig. 7.18. If the solid rate Gs is lower than the
saturation carrying capacity GTR (corresponding to the transport velocity),
then as the gas velocity is decreased choking will occur at the velocity for
which Gs is the saturation carrying capacity, and the characteristics of the
suspension just before its collapse are represented by a point on the dashed
curve GAP vertically above Gs
Having collapsed, the solid will form a fluid bed at the bottom of the pipe.
Upon a slight further decrease in the gas velocity, the bed level would begin

REGIMES OF FLUIDIZATION

In Fig. 7.21 the data obtained in the 15.2 em column over the bubbling,
turbulent, and fast bed regime, have been assembled in the form of a
fluidization map depicting the slip velocity versus the solid volumetric
4
SOLID RATE, Gs (Kg/m2

SLIP
VELOCITY ;

19~~
146

SOl/OS~FCC

FAST FLUIDIZED
-

U
--Vs

51

REGIME

(m/sl

9~,TRANSPORT

VELOCITY

~I
I

." TURBULENT REGIME

19.5

,BREAKDOWN OF SLUGS
~.

,!YBBLING

REGIME

"- I-:mf
0.2

1-

0.4

0.6

0.8

1.0

'Chonge in scale

Figure 7.21 Fluidization data for a fluid cracking catalyst (FCC): the slip
velocity versus the solid volumetric concentration. Data obtained in the 15.2 cm
system.

concentration at the bottom of the bed. The slip velocity is used rather than
the superficial gas velocity since it is the former which provides a relevant
measure of the interaction of gas and solid. The figure attempts to highlight
some of the points made above. Over the bubbling and the turbulent regimes,
the relation between the slip velocity and the solid concentration appears
unique. Not so in the fast bed where, corresponding to each solid rate, a
different function connecting these variables exists. If the particles were
uniformly dispersed in the fast bed, then the relation between these functions
would be unique, as it approximately is, for example, in hydraulic transport of
any but very fine solids (Price, Lapidus, and Elgin, 1959). However, the
'structure of a fast bed of fine solids is characterized by the presence of clusters
and streamers of particles. The hydrodynamic behaviour of the fast bed
accordingly reflects the interaction of the gas not so much with individual
particles as with clusters of various sizes. A change in gas velocity or solid
rate, at a given solid concentration, produces a change in the spectrum of
cluster sizes - hence the multiple functions relating the slip velocity to the
solid concentration.
Figure 7.22 is a qualitative fluidization map for fine solids. It is patterned
after Fig. 7.21 but extends further to the dilute phase flow regime. The figure
also displays schematics of equipment typical of the various regimes, each
sized approximately for the same gas-treating capacity.
The difference between dilute phase flow and the fast bed condition was
enumerated in Section 7.4. The distinction between the fast fluidized bed and
the riser transport reactor used in many modem cat cracking plants is less

Transport
riser" reactor
II

~t

:;:~
Turbulent
,', ~;'

-a
vt

.=1
i:S
.

....

--_Urn! tEmf

fluidized

-:"'$: bed

;;:::

Higher
solid
Transport
velocity
Transition to
turbulent
fluidization
1-E

,"

Bubbling
fluidized
bed

clear. As in the fast bed, relatively high solids concentrations (typically


around 160 kg/m3), large slip velocities, solid backmixing, and particle
aggregation are also manifest in the riser reactor (Matsen, 1976). The
difference, which is yet to be fully explored, may lie in the disposition or
structure of the solid. Saxton and Worley (1970) report several radial density
profiles, in the form of contour plots, obtained by radiation attenuation
measurements across a riser reactor. These reveal gross solid segregation.
Some of the contour plots show concentrated zones, most likely representing
solid flowing downward, along the walls, while the core remains relatively
more dilute. Others display large concentrated regions extending into the
core, or a few isolated pockets of high solid concentration occupying positions
away from the wall and surrounded by leaner areas.
If the fast bed is to suit a wider range of physical and chemical process
applications, it should entail a greater degree of homogeneity. Observations
of a two-dimensional bed with powders belonging to Geldart's group A
indicate that a fine scale of gas solid demixing patterns, which extends across
the bed, can be maintained at gas velocities up to 4 to 5 m/s. In contrast, gas
velocities in an FCC riser typically exceed 10 mls beyond some short distance
above the oil feed nozzles.
Observations of two-dimensional beds, however, may be misleading. When
a round 7.6 cm fast bed of HFZ-20 was photographed by X-rays the core was
revealed to be considerably more dilute than the wall region (Gajdos and
Bieri, 1978). This is not altogether surprising. In columns of narrow diameter,
even the bubbling and turbulent fluidized beds give rise to a structure marked
by dense regions along the wall and a leaner core. This will be illustrated in
Section 7.7. The work shows that at lower velocities (around 3 m/s) and at
high solid loadings, the area of the leaner core diminishes and its density
increases.
Figure 7.15 indicates that at the same gas and solid flowrates, segregation
of solid along the walls is more pronounced in the 7.6 diameter unit than in
the 15.2 cm diameter column. Whether this trend extends to columns of
larger diameter remains to be seen since relevant research has yet to be
undertaken. What would be the radial density profile in a fast bed of larger
diameter? Will a relatively thick solid annulus surround a single leaner core?
Or will the profile, apart from some significant wall phenomena, be more
uniform, consisting of a cell-like structure of alternating dense and leaner
regions? What is the effect of gas velocity, solid rate, the properties of solid
and gas, and equipment scale on the structure of a fast bed?
Resolution of these questions must rank high among the research goals that
lie ahead in this area of investigation. Resolution of these questions will place
the riser in better perspective vis-a-vis the fast bed. On the basis of our
laboratory work, and commercial practice to date (Reh, 1971; Yerushalmi
and Cankurt, 1978), it appears that a more efficient contact of solid and gas is

enhanced by the use of fine solids, gas velocities around 3 to 5 mis, and high
solid circulation rates.
7.7

GAS MIXING IN HIGH VELOCITY FLUID BEDS

Cankurt and Yerushalmi (1978) conducted experiments in the 15.2 cm


diameter system, shown in Fig. 7.5, on the extent of gas backmixing in a bed
of a fluid cracking catalyst fluidized at gas velocities spanning the bubbling,
turbulent, and fast bed regimes. In these experiments, a tracer gas, methane,
was continuously injected at a known rate through a tube entering the bed at
any desired elevation. Simultaneously, a stream of sample gas was withdrawn
from locations below the injection point, and its tracer concentration was
analysed and recorded. Tracer could be injected and sampled at any radial
position in the bed. The experiments were conducted at room conditions.
Figure 7.23 gives the radial concentration profile across the bed 5.1 cm
below the injection point at different gas velocities. The tracer was injected at
the centre. Those results lead to the conclusion that the degree of gas
backmixing is high in the bubbling regime (0.21 mls); it diminishes, though it

may still be appreciable, over the turbulent range (0.82, 1.19, and 1.52 m/s)
and, in the absence of a pronounced gas velocity profile, gas is essentially in
plug flow in the fast fluidized bed (1.80 mls and beyond).
This conclusion is reinforced by Figs 7.24 and 7.25, which provide complete
concentration profiles (radial and axial) upstream of the injection point at gas
velocities of 0.21 and 1.52 mls. A striking aspect of these profiles is the
increased concavity of the radial profiles with gas velocity. This suggests a
physical picture of a mixing pattern dominated by downwards movement of
dense phase solid along the wall, and a fast-moving gas flowing upwards
through a leaner core.
120

100

80

~
~

60

....
(,)
(,)

40

20

~7.6

::E J

-5.1

-2.5

2.5

5.1

Distance From Centre (em)

o~

-7.6

-5.1

-2.5

Figure 7.24 Tracer concentration profiles at 0.21 mls.


Tracer injected at centre. Fluidized density = 397 kg/m3.

~i~
T~5.64
2.5

5.1

7.6

Distance From Cent~ (em)

Figure 7.23 Radial concentration profiles measured 5.1 cm


upstream of the injection point at different gas velocities. Tracer
was injected at centre. Fluidized densities in the order of increasing
velocities are 397, 266, 213, 189, 170,64, and 19 kg/m3

Figure 7.26 and 7.27 add further details to this picture. Figure 7.26 shows
radial profiles obtained at three locations below the injection point (i,e,
upstream) when the tracer was injected near the (left) wall of the bed
fluidized at 1.52 mls. Figure 7.27 gives the corresponding axial profiles
measured along the left wall, the right wall, as well as along the centre
of the bed.
-\

UNICANIP
. t\ISLlOTECA
CENTRAl.

~
e....

10

~
u

Figure 7.25 Tracer concentration profiles


at 1.52 mls. Tracer injected at centre.
Auidized density = 189 kg/m3

Figure 7.26 Tracer concentration profiles


at 1.52 mls. Tracer injected near the left
wall.

50

This mixing pattern, as was noted on two earlier occasions, is typical of


beds of small diameter and reflects to a large measure the influence of the
wall. I would expect that influence to become insignificant in beds of large
diameter; the essential feature in which dense phase solids descend against a
rising leaner phase would still prevail, but the random flow of the bubbles,
voids, and channels that make up the lean phase would preclude the sort of
constant profiles that persist on the average in narrow beds.
The curvature of the radial profiles increases with gas velocity up to 1.52
mls. At some velocity beyond that, a sharp change occurs. At 1.80 mis, the
curvature of the profile appears rather small, and at higher gas velocities, the
profile is practically flat. This is illustrated very clearly in Fig. 7.28, in which
the ratio of the concentration near the wall to that at the centre is plotted
versus the superficial gas velocity. The sharp change that is bracketed by 1.52
and 1.80 mls may be laid to a fundamental change in the fluidization
regime - to a change from turbulent to fast fluidization. The transition
velocity (i.e. the transport velocity) is ostensibly higher than the value
determined by other methods simply because the solid itself circulated at a
mean velocity around 0.3 mis, owing, as noted earlier, to the use of tall beds
and the high solid carryover rates which that entailed.

________________________

75

Distance

100

Upstream

125

150

175

(em)

Figure 7.27 Axial concentration profiles at 1.52 mls.


Tracer injected near left wall.

6
4

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
Superficial

Gas Velocity,

U (m/s)

Figure 7.28 Ratio of the tracer concentration near the


wall to that at the centre, measured 5.1 cm upstream of the
injection point. Tracer injected at centre.

1
L

Dense
phose

..,

SOLID "FCC

.-......:....

"'

GAS VELOCITY

.~

Im/S\

.':-.....0.21 .~

g.% \\
1.0

0.82

oCi

1.2

'"

"

\ ,,:
Figure 7.29 Schematic of the countercurrent
two-phase model for the study of gas backmixing.

0.5

1.0

1.5

2.0

DISTANCE UPSTREAM

Theory
The physical picture, sketched above, of the mlXlng patterns over the
bubbling and turbulent regimes invites the application of a simply counter
current model. The model is illustrated in Fig. 7.29, and the governing
equations and their solutions are given below:

de
q -

dx

+ k(C

- c)

=0

2.5

3.0

3.5

Cm)

Figure 7.30 Axial concentration profile at the


centre line at different gas velocities. Tracer
injected at centre.

The relevance of the model is borne out in Fig. 7.30 where semi log plots of
the axial concentration profiles at 0.82, 1.19, and 1.52, and also at 0.21 m/s do
indeed form straight lines. From the intercepts, one may deduce the ratio of
Q to q. Figure 7.31 gives the results. The figure shows the diminishing
percentage of gas backflow with gas velocity over the turbulent range.
100

de
Q dx - k(C - e)

=0

= constant
e = a + /3e-k(r-l)xlq
C = a + f3re-k(r-l)xlq

qe - QC

(7.9)
(7.10)
(7.11)

q
r=-

In the above, q is the volumetric rate of gas flowing upwards and Q is the
corresponding rate for the gas flowing back; a is the tracer concentration far
upsteam.

Figure 7.31 Ratio of backflow to upflow (Q/q) versus the


superficial gas velocity.

Yang, Huang, and Zhao (1983) have recently studied axial and radial
dispersion of helium tracer gas in a fast bed. They found virtually no gas
backmixing in the central part of the riser but some near the walls. Radial
dispersion was found to decrease slightly with increase in gas velocity and to
increase with solid circulation rate. The relationships were expressed in an
empirical corelation:
Er

= 43.4 (

We ~

)
s

1: e )

+ 0.7

High velocity recirculating beds are proving increasingly attractive for certain
applications. Lurgi Chemie and Huttentechnik GmBh of Frankfurt, West
Germany, was the first to appreciate the broad potential of the fast fluidized
bed. Lurgi realized this potential in its successful process, developed jointly
with Vereinigte Aluminum Werke (VAW), for calcining aluminium hydroxide to supply cell-grade alumina (Reh, 1971). The heart of the process is a
high velocity fluid bed furnace (Fig. 7.32), which Lurgi often refers to as the
highly expanded circulating fluid bed furnace.
Calcination is conducted at a temperature between 1,000 and 1,100C.
Heat is provided by direct injection of oil or fuel gas into the lower section of
the furnace. That section is considerably denser owing to the staging of the
combustion. Primary air supplied through a distributor is substoicheiometric
and the combustion is completed with secondary air in the upper portions of
the reactor. The temperature, however, remains remarkably uniform all
along the height of the reactor which, in commercial installations, is typically
around 15 to 20 m. The secondary air is introduced through a few tuyeres.
Reh (1971) reports that the overall gas velocity in the pilot plant which
preceded the first commercial unit was in excess of 3 mls. Given that the mean
size of the alumina particles lies around 50 to 80 /km, one may presume that
below the secondary air ports the bed is maintained in turbulent fluidization.
Above the secondary air ports, the suspension is circulated in the fast bed
mode.
Since its introduction in 1970, the LurgiN AW has proved very successful,
and several of these plants now supply the bulk of alumina in West Germany;
they are also in operation or under construction in other countries in Europe
and Asia. The largest units have a capacity of about 12.5 kg/s and have a
diameter of about 4 m. The Lurgi development suggests that fast beds could
be scaled up rather readily for applications of this type. Lurgi's scale-up was
in two steps: a 12.5 cm i.d. unit was followed by a pilot 90 em diameter and
8,5 m tall unit. Both the pilot and the first commercial plant (which was over 3

Electrostatic
precipitator

Figure 7.32
of alumina.

Flow diagram of the LurgiN A W process for production

m in diameter) achieved design operation within a few weeks of startup.


In the only commercial Fisher-Tropsch plant, in Sasolburg, South Africa,
the synthesis reaction takes place in a circulating fast fluid bed. The plant has
seen successful operation since 1954, and is currently undergoing a large
expansion to a capacity many times its present one. The so-called Synthol
reactor (Fig. 7.33) operates at a pressure around 20 atms. The existing reactor
is about 2.3 m in diameter and about 30 m tall. The gas solid suspension flows
upwards through two heat exchange sections where heat liberated by the
synthesis reaction is removed. The temperature near the bottom of the
reactor, following the mixing of the synthesis gas and the catalyst descending
from the standpipe, is about 315C, and approximately 340C in the settling
hopper (Hoogendoorn, 1973). The gas velocity is about 6 mls in the heat
exchange section and about 2 mls in the reactor. The solid circulation rate is
about 400 kg/m2s. The fine iron catalyst used in the Fischer-Tropsch synthesis
belongs to those cohesive solids which Geldart has classified as group C.
These solids are simply difficult to fluidize at low velocities, either in the
bubbling or even the turbulent regimes. The success of the fast bed Synthol
reactor is due in a large measure to the better solid mixing and the more
efficient gas solid contact brought about by the high gas velocities typical of
the fast bed.
Since the late 1970s, circulating fluid bed boilers have been in commercial
service burning a variety of fuels. A recent survey (Schweiger, 1985)

and chemical reactions and tracer studies are required to elucidate the
interaction of gas and solid.

C
dp

Cooler
groups

Catalyst settling
hopper

Gs
GsTR
Standpipe
----

Cooling
inlet

fluid

Hmf

q
Q
Fresh feed and
recycle

Gas and catalyst


mixture

demonstrates that a major and still growing portion of the market of fluid bed
boilers has been captured by circulating fluid bed boilers. These boilers
typically operate at higher combustion efficiencies and at lower excess air
levels than bubbling-type fluid bed steam generators. They utilize a simpler
and more forgiving fuel-feed system and, when burning high-sulphur fuels,
require less limestone for S02 capture. Also, they are more adaptable to
staged combustion affording NOx control as well as a better load-following
capability.
There are indications that some well-established successful processes
operate in the turbulent regime e.g. acrylonitrile synthesis, roasting of
sulphide ores (Queneau,Bracken,
and Kelly, 1958); others are under
development, e.g. Mobil's methanol-to-gasoline process (Penick, Lee, and
Mazink, 1982). The coming years will see the catalogue of new processes
expanding, but if these processes are to fully reap the potential of high
velocity fluidization, they must be able to draw from a richer reservoir of
fundamental research than that available at present.
There is a need for research at a scale that goes beyond the few inches that
has characterized much of the work to date - for systematic study of solids
spanning a wide range of properties and of the influence of temperature and
pressure. Diagnostic tools that can probe the beds must be brought to bear,

concentration of tracer in the lean phase


concentration of tracer in the dense phase
volume surface mean diameter
column diameter
acceleration of gravity
solid rate per unit cross-sectional area
saturation carrying capacity corresponding to UTR
expanded bed height
bed height at minimum fluidization
volumentric flowrate of gas flowing upwards
volumetric flowrate of gas flowing downwards
ratio of q/Q
superficial gas velocity
absolute rise velocity of bubble, slug, or lean phase gas
velocity at which pressure fluctuations across a bubbling bed
begin to diminish from their peak value
velocity at which turbulent regime commences
minimum fluidization velocity
transport velocity
mean solid velocity
terminal velocity of a single particle
voidage
volume fraction of solid
particle density
bed density

Canada,G.S., McLaughlin,M.H., and Staub,F.W. (1976). A.I.Ch.E. Symp. Ser., 74


(176), 27.

Cankurt, N.T., and Yerushalmi,J. (1978). In Fluidization (Eds J.F. Davidsonand


D.L. Keairns),CambridgeUniversityPress, p. 387.
Carotenuto, L., Crescitelli,S., and Donsi, G. (1974). Ing. Chim. ital., 10, 185.
Clift, R., and Grace,J.R. (1972). Chem. Eng. Sci., 27,2309.
Clift,R., Grace, I.R., and Weber, M.E. (1974). Ind. Eng. Chem. Fund., 13,46.

Crescitelli, S., Donsi, G., Russo, G., and Clift, R. (1978). Chisa Conference, Prague.
Gajdos, U., and BierI, T.W. (1978). Topical report for the US Department of
Energy, Contract No. EX-C-76-01-2449,
Grace, J.R., and Tuot, J. (1979). Trans. Inst. Chem. Engrs., 57,49.
Graff, R.A., and Yerushalmi J. (1978). Seventh quarterly report, prepared for the US
Department of Energy, Contract No. EX-76-S-01-2340.
Harrison, D" Davidson, J.F., and de Kock, j.W. (1961). Trans. Instn. Chem. Engrs.,
39,202.
Hoogendoorn, J.C. (1973). Clean Fuelsfrom Coal., Symposium Proceedings, Institute
of Gas Technology, Chicago, Illinois, p. 353.
Horsier, A.G., Lacey, J.A., and Thompson, B.H. (1969). Chem. Eng. Prog, 65 (10),

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

59.

Kehoe, P.W.K., and Davidson, J.F. (1971). Inst. Chem. Eng. (Lond.) Symp. Ser., 33,
97.

Lanneau, K.P. (1960) Trans. Instn Chem. Engrs, 38, 125.


Massimilla, L. (1973). AI.Ch.E. Symp. Ser., 69 (128), II.
Matsen, J.M. (1976). Fluidization Technology (Ed. D.L. Keairns), Vol. 11,
Hemisphere Publishing Corporation, p. 135.
Price, B.G.L., Lapidus, L., and Elgin, J.e. (1959). A.I.Ch.E.J., 505, 93.
Penick, J.E., Lee, W., and Mazink, J. (1982) Inst. Symp. on Chem. React. Engng,
ISCRE - 7, 1, Boston, Oct. 4.
Queneau, P., Bracken, E.H., and Kelly, D. (1958). J. of Metals, 10,527.
Reh, L. (1971). Chem. Eng. Prog. 67, 58.
Saxton, A.L., and Worley, A.C. (1970). Oil and Gas J., 68 (20), 84.
Schweiger, B. (1985). Power, Feb. 5-I.
Thiel, W.J., and Potter, O.E. (1977). Ind. Eng. Chem. Fund., 16,242.
Turner, D.H.L. (1978), Ph. D. Dissertation, The City College, City University of New
York.
Van Swaaij, W.P.M., Burman, c., and Van Breugel, J.W. (1970). Chem. Eng. Sci.,
25,1818.
Wallis, G.B. (1969). One-Dimensional Two-Phase Flow, McGraw-Hill, Chapter 4.
Yang, G _,Huang, Z., and Zhao, L. (1983). Fluidization IV, Japan, May 29 - June 8,
Eng. Foundation.
.
Yang, W.C. (1976). Proc. of Pneumo Transport, Vol. 3, BHRA Fluid Engineering
(April).
Yerushalmi, J., and Cankurt, N.T. (1978). CHEMTECH, 8, 564.
Yerushalmi, J., and Cankurt, N.T. Geldart, D., and Liss, B. (1978)AI.Ch.E. Symp.
Series, 74 (176), I.
Yerushalmi, J., and Cankurt, N.T., (1979) Powder Technol., 24, 187.
Yerushalmi, J. Cankurt, N.T., Geldart, D., and Liss, B. (1978).AI. Ch.E. Symp. Ser.
74 (1976). Ind. Eng. Chem. Proc. Des. & Dev., 15, 47.
Zenz, F .A., and Othmer, D.F. (1960), Fluidization and Fluid-Particle Systems
Reinhold, New York.

Solid-Gas Separation

The elu,triation of particles is an inevitable consequence of fluidized beds, and


the equipment required to separate the gas and powder provides a significant
contribution to the capital and operation costs of the process. In some
applications the fine powder carried out is valuable and process economics
cannot allow its loss; in others, emission of the particles constitutes a nuisance
or a health hazard; in many applications all of these considerations apply.
Efficient gas cleaning and recovery of the solids form an essential part of
the process requirements for the fluidized bed and are the subject of this
chapter. Because cyclones are the most common device used in fluidized bed
systems they receive the greatest attention, but some of the other techniques
used as alternatives or in addition are also discussed briefly.
Solid-gas separation as a title may be interpreted to mean both 'degassing of
solids (as a direct analogy with solids dewatering in solid-liquid separation)
and gas cleaning, i.e. dedusting of the gas. Only the latter is to be considered
here and the terms solid-gas separation and gas cleaning will be used
interchangeably because the term gas cleaning is quite commonly used, even
in cases when the solids represent the product.
Three phases may be distinguished in any gas-cleaning operation: tranport
of particles onto a surface (separation), collection of separated particles from
the separation surface into discharge hoppers (or particle fixation), and the
disposal of the collected material from the gas-cleaning equipment. The
following account of equipment deals only with the first two phases.

197

8.2

PRINCIPLES

OF PARTICLE SEPARATION
OF EQUIPMENT

AND CLASSIFICATION

In the first phase of gas cleaning, forces are applied to the particles in order to
bring them to a collecting surface; the principles of particle separation are
usually classified according to the nature of the forces involved. These may
be:

(a) External forces due to fields of acceleration which are external to the
gaseous suspension, such as gravity, electrostatic, or magnetic forces, or
(b) Internal forces due to fields of effects which take place within the
suspension itself, e.g. inertial or centrifugal forces, diffusion, coagulation, electrostatic effects of charged particles, thermophoresis, diffusiophoresis, and piezophoresis.
The process of screening, in which particles are classified in relation to their
ability to pass through an aperture in the screen, does not lend itself to the
above classification, but its role in gas cleaning is relatively minor.
Gas-cleaning equipment often combines two or more of the abovementioned principles in one unit; the classification of equipment therefore
does not necessarily follow the same pattern. The most common classification
is into four groups, as follows:
(a) Aeromechanical dry separators in which gravity and/or inertial effects
prevail. This group includes (Svarovsky, 1981b) cyclones, settling
chambers, intertial separators, dual vortex separators, and fan collectors
(or 'mechanical cyclones'). Only the first will be considered in this
chapter.
(b) Aeromechanical wet separators (scrubbers) which make use of diffusional
and inertial effects.
(c) Electrostatic precipitators which depend on electrostatic and gravity
forces.
(d) Filters which use inertial and diffusional effects.

Flowrate-pressure drop relationship; efficiency; economic criteria; suitability for different conditions (the nature of both the dust and the gas),
solids concentration, method of disposal, reliability, etc.
8.3.1

Flowrate-pressure drop relationship

Most gas-cleaning devices have a fixed relationship between static pressure


drop I1p and gas flowrate Q, depending on the configuration of the gas
cleaner. Most frequently, the relationship is expressed in the same way as
with other flow devices as:
I1p

V2 Eu Pg U2

(8.1)

where Eu, the Euler number, is a resistance coefficient (analogous to Co in


Eq. 6.2) which may be a function of the Reynolds number and other
operational variables such as the feed concentration of solids, specific water
consumption with scrubbers, etc.; U is a characteristic velocity calculated
from Q/A, where A is a characteristic area in the separator (e.g. the crosssection of the cylindrical body in cyclones); and Pg is gas density.
One exception in the application of Eq. (8.1) is in air filters, where the
pressure drop is also a function of the amount of dust deposited on the filter.
8.3.2

Efficiency

Efficiency is an important criterion, since it governs the degree of cleaning. It


is best expressed as gravimetric grade efficiency G(d) (see Svarovsky, 1981a,
for details of its evaluation). Comparison of typical grade efficiencies of
aeromechanical dry (D) and wet (W) separators, electrostatic precipitators
(E), and filters (F) is made in Fig. 8.1.

Many gas-cleaning systems combine two or more of the above groups either
by using different equipment in series or by combining these in a single unit.
The above classification of equipment will be followed in the sections on
equipment and a separate section will be devoted to each of the four groups.
8.3

GENERAL CHARACTERISTICS

OF EQUIPMENT
Particle size d (}km)

There are several factors affecting the choice of gas-cleaning equipment for
any particular application. These are:

)0

Figure 8.1 Typical grade efficiency curves of dry separators (D), wet separators (W),
electrostatic precipitators (E), and filters (F).

8.3.3

Economic criteria

Economic criteria consist of the capital and running costs of the dust-arresting
plant. Capital cost is normally expresse~ per thousand cubic metres. of
cleaned gas per hour; it may be further spht mto the cost of the ~on~tr~ctlO.n
material cost of labour erection, design, etc. There are other cntena 10 this
'fi floor
category," such as specific volume of the plant (m3/1,000 ~ 3) , s~ecl.c
area taken (m2/1,OOO m3), etc., the importance of which vanes with different
applications.
.
Running costs include cost of power, mamtenance, water, etc;. ~ower
needed for running the plant consists of the power for pumps, electnclty for
cleaning (electrostatic precipitators), and also the power for blowing the gas
through the plant.
Whenever the total power requirements consist solely of the po~er needed
for passing the given flowrate Q through the separator, the theoretical power
can be calculated from the product of the required pressure drop I:1p and gas
flowrate: Q x I:1p (W). To allow relative comparisons .b.etween differe~t
separators, the theoretical power may be expressed as speclfi~energy per UOit
flowrate, and this is usually in watt-hours per thousand cubiC metres. Thus;
each newton per square metre of pressure drop represents 0.28 Wh/l,OOOm
or in practical engineering units, 1 mm w.g. (column of water) represents 2.73
Wh/1,000 m3 The actual power can only be derived from th~ theoretical
power requirements if the efficiency of the fan and the electnc motor are
known or assumed.

8.3.4

The most important disadvantage is their relatively low efficiency for very
fine particles which leads to their frequent role as a pre-cleaner.
Cyclones are now used in many different fields of technology, but their
most extensive application still remains in gas cleaning, where they are
employed for the separation of relatively coarse dusts.
There are basically two main designs of cyclones available: the reverse-flow
cyclones and the 'uniflow' cyclone., the former being more frequently used.
In both types, tne inlet gas is brought tangentially into a cylindrical section
and a strong vortex is thus created inside the cyclone body. Particles in the
flow are subjected to centrifugal forces which move them radially outwards,
towards the inside cyclone surface on which the solids separate. The two types
of cyclone differ in the direction in which the clean gas leaves the main body,
as is shown in the following.
A typical reverse-flow cyclone (Fig. 8.2) consists of a cylindrical section
joined to a conical section and the clean gas outlet is through a pipe which
extends some distance axially into the cyclone body through the top. The gas
inlet may be tangential, spiral, helical, or axial. Different inlet types lead to
some variations in performance, but the effect is not strong and, in most
cases, the tangential inlet is preferred for its simple construction.

Suitability for different conditions

There are a number of other factors, such as gas temperature and humidity,
the cohesiveness and abrasiveness of the dust, reliability, limits in dust
concentration, etc., which may exert an overiding influence on the final
choice.
8.4

CYCLONES

The advantage of cyclones, and indeed all aeromechanical dry separators,


include:
simple design,
low capital cost,
suitability for higher temperatures,
low energy consumption,
product is dry,
reliability,

1
2
3
4
5

Helical Axial
Spiral
Three other types of gas
inlet available

Gas inlet
Cylindrical port
Conical port
Gas outlet
Top cover

The outer vortex created by the tangential entry is helical, moving


downwards - particles in the flow settle into a dust layer on the wall and this
is pushed down into the apex. Hence, the removal of the dust from the
collection surface is due to the gas flow and not to gravity; gravity has been
found to have little effect on the separation efficiency of cyclones.
The outer vortex reverses its axial direction in the apex and creates the
inner vortex going upward, which carries the gas into the outlet pipe.
As the vortex reaches very far into the apex, it is advisable not to put a
rotary valve or a sliding valve there; it is better to leave the dust outlet clear
and use a discharge hopper underneath the cyclone with the necessary valve
on the hopper outlet. There is evidence of the vortex reaching even into the
hopper itself, but the layer of dust reentrainment is much reduced by using a
discharge hopper.
In the 'uniflow' cyclone, often also called the 'straight-through' cyclone
(Fig. 8.3), the axial direction of the vortex is not reversed but the flow
continues in the same direction and the gas leaves through an outlet pipe
which is at the end opposite to the inlet. While in the reverse-flow cyclone the
separated solids are in powder form, the uniflow cyclone only functions as a
concentrator, i.e. the separated solids leave the unit still suspended in gas
which is usually 5 to 10 per cent of the main flow. A second-stage separator
has to be provided to treat the 'underflow' if the solids are required in powder
form.

-~--f~-*-

....
Particles with small gas
flow (5 to 10 per cent)

It can be shown theoretically (see Eq. 8.6 below) that for a given pressure
drop the cut size of a cyclone is proportional to the square root of the cyclone
diameter; i.e. the smaller the cyclone the higher will be its efficiency. It is
therefore theoretically sound to build multi-cyclone arrangements which use
several smaller units in parallel, with common inlet and exit manifolds. Great
care must be taken in designing the systems to ensure even distribution of
both the gas and the solids between the individual cyclones, because, if this is
not achieved, the resulting blockages (and even backflow in some units)
reduce the overall efficiency and the advantage of using multi-cyclones may
thus be completely lost. Generally, the overall grade efficiency of any multicyclone is never as good as that of the individual cyclone units.
Both reverse-flow and uniflow cyclones are used in multi-cyclone arrange-

Figure 8.4 Schematic diagram of a multicyclone arrangement.

ments, the axial, vane inlet being most commonly employed. Figure 8.4 gives
a schematic diagram of a typical multi-cyclone unit.
Multi-cyclones are generally more elaborate in construction and therefore
more expensive than single cyclones; they are also more liable to abrasion and
blocking of the dust discharge orifice due to the smaller diameter of the
individual units. The latter disadvantage makes multicyclones unsuitable for
cohesive dusts and leads to a lower limit in feed concentration of dust.
8.4.1

Flow characteristics

The static pressure drop measured between the inlet and the gas outlet of a
cyclone is usually proportional to the square of gas flowrate Q: this means
that the resistance coefficient defined as the Euler number, Eu, in Eq. (8.1) is
practically constant for a given cyclone geometry or 'design', independent of
the cyclone body diameter. The characteristic velocity U can be defined for
gas cyclones in various ways but the simplest and most appropriate definition
is based on the cross-section of the cylindrical body of the cyclone, so that:
U-

4Q

1T'D~

where Q is the gas flowrate and Dc is the cyclone inside diameter.


The two other alternatives to the definition of characteristic velocity, the
average inlet or outlet velocities, are not recommended because neither of
them would lead to a sensible comparison of different designs; it can be
argued that the cyclone body diameter is the most important dimension,

determining the manufacturing costs, the space occupied, headroom, etc. As


an example to demonstrate the superiority of the definition of body
characteristic velocity in Eq. (8.2) consider two cyclones, identical in diameter
and in al1 other dimensions except their gas inlet and outlet diameters. One
has a large inlet and smal1outlet while the other has a sman inlet and large gas
outlet: the relative size of the two openings may be such as to result in an
identical pressure drop - flowrate relationship for both cyclones. Using the
body velocity defined in Eq. (8.2), the resistance coefficient Eu in Eq. (8.1)
would be the same for both cyclones; this is to be expected as the cyclones are
of the same size and give identical flowrates for the same pressure drops. If,
however, either the inlet or the gas outlet velocities are used (and some
authors still insist on using those), the resistance coefficients thus obtained
would be very much different for the two cyclones, thus apparently favouring
strongly (and wrongly) one of the two designs depending on which of the two
alternative definitions of U is used.
The resistance coefficient Eu increases a little at high pressure drops and
reduces at high concentrations, but this can usually be compensated for by
plant adjustment, and clean air data are normally taken in design calculations.
8.4.2

Efficiency of separation

The second dimensionless group which characterizes the separation performance of a family of geometrically similar cyclones is the Stokes number Stkso
defined as:

where dso is the cut size (equiprobable size). The cut size and its relationship
to separation efficiency are defined below.
In general, the separation efficiency of gas cyclones depends on particle
size (it is then called the 'grade' efficiency) and increases from zero for ultrafine particles to 100per cent for very coarse particles (see curve D in Fig. 8.1).
The particle size recovered at 50 per cent. efficiencyis referred to 'as the 'cut'
size dso and can be understood as equivalent to the aperture size of an ideal
screen that would give the same separation performance as the cyclone. The
total solids recovery in a particular case then depends on the grade efficiency
(or cut size) which characterizes the cyclone operated under given conditions
and on the size, density, shape, and dispersion of the particles (i.e. the
characte ristics of the feed material).
As the grade efficiency does not rise very steeply with increasing particle
size, some particles in the feed coarser than the cut size will pass through the
cyclone while some particles finer than the cut size will be separated.

As can be seen from Eq. (8.3), the separation efficiency is described there
only by the cut size dso and no regard is given to the steepness of the grade
efficiency curve. If the whole grade efficiency curve is required in design or
performance calculations, it may be generated around the given cut size using
plots or analytical functions of a generalized grade efficiency function
available from the literature or from previously measured data. The
knowledge of the exact form of the grade efficiency is usually not critical in
solid-gas separation applications because only total mass recovery is of
interest, and this is not much affected by the shape of the curve.
Consequently, very little is known about how the shape of the grade
efficiency curve is affected by operating pressure drop, cyclone size or design,
and feed solids concentration.
In powder classification applications, however, including the case of degritting, the shape of the curve determines the amount of the 'misplaced'
material, such as the amount of grit reporting to the gas outlet.
The Stokes number Stkso defined in Eq. (8.3) is usually constant for a given
cyclone design (i.e. a set of geometric proportions relative to the cyclone
diameter Dc), when the cyclone is used to separate granular material at feed
concentrations of less than about 5 g/m3.
Particle size d is best measured as the equivalent Stokes diameter by
sedimentation or elutriation methods. This equivalence is based on the
assumption that, if a spherical particle and an irregularly shaped particle
settle at the same velocity (in gravity or centrifugal fields), they will separate
at the same efficiency. This assumption does not hold for flat or needleshaped particles which assume different orientation in a cyclone than under
gravity or centrifugal settling. Problems are also encountered when the
particles undergo the separation process in an agglomerated state and the
agglomerates are subsequently redispersed into single particles before
particle size analysis.
8.4.3

Reverse-flow cyclone designs

Equations (8.1), (8.2), and (8.3) form the basis of gas cyclone design and
scale-up. There is a whole host of different cyclone designs available today,
which are usually divided into two main groups according to their geometrical
proportions relative to the body diameter: the 'high efficiency' designs and
the 'high rate' designs. Table 8.1 gives the geometry of the two best-known
cyclone designs (refer to Fig. 8.2 for dimensions). Note that only the reverseflow cyclone is considered here because that is the type used most widely in
industrial practice.
The so-called 'high efficiency' cyclones are characterized by relatively small
inlet and gas outlet orifices, and a long body, and give high recoveries. The
'high rate' designs give medium recoveries but offer low resistance to flow so

Table 8.1 Cyclone proportions


diameter
Proportion relative
to diameter
Cyclone
Dc
type
Stairmand,
Stairmand,

H.E."
H. R. b

for two different designs based on cyclone


= 1 (see Fig. 8.2)

ponding values of Eu and Stkso are plotted for several commercial and other
well-known designs. The points are well scattered but a line can be drawn
through them to show a general trend. The line drawn in Fig. 8.5 can be
described by the following approximate equation:

Dc

Eu =
4.0
4.0

2.5
2.5

1.5
1.5

0.375 0.5
0.2
0.575 0.875 0.375

0.5
0.75

0.5
0.75

High efficiency cyclone


b High flowrate cyclone.

that a unit of given size will give much higher air capacity than a high
efficiency design of the same body diameter. The high rate cyclones have
large inlets and gas outlets, and are usually shorter. In order to prevent the
incoming jet of air impinging on the gas outlet pipe, the inlet is spiral (wrapround type) while the high efficiency units can (and often do) have to a simple
tangential entry.
It is interesting to find that, for well-designed cyclones, there is a direct
correlation between Eu and Stkso: high values of the resistance coefficient
usually lead to low values of Stkso (therefore low cut sizes and high
efficiencies), and vice versa. This is shown in Fig. 8.5 where the corres-

. *nnv

xa

t.:Il:!:J "'465-20

VI

Slai,mand

HR.<$>..

R[
--

Stkso

This equation may be used for estimates of cut size of unknown cyclone
designs (of 'reasonable' proportions) from the cyclone flow characteristics
and is intended for guidance only.
Note that the scale-up of cyclones based on Eu and Stkso works well for
near ambient conditions but also predicts the performance reasonably well at
high absolute pressures and high temperatures; this means that there is no
effect of high pressures and temperatures other than that accounted for in the
definitions of Eu and Stkso on gas viscosity and density.
8.4.4

Recommended range of operation

One of the most important characteristics of gas cyclones is the way in which
their efficiency is affected by pressure drop (or flowrate). Correctly designed
and operated cyclones should give pressure drops within a recommended
range, and this for most cyclone designs operated at ambient conditions is
between 2 and 6 inches of water gauge (w.g.) (approximately from 500 to
1,500 Pa). Within this range, the recovery increases with applied static
pressure drop. At higher absolute pressures the limits increase to higher
pressure drops and the equivalence is based on the same inlet velocity.
At pressure drops below the bottom limit, the cyclone represents little
more than a settling chamber, giving low efficiency due to low velocities
within it which may not be capable of generating a stable vortex. Above the
top limit the mass recovery no longer increases with increasing pressure drop
and it may actually decline; it is therefore wasteful to operate cyclones above
this limit, the value of which depends very much on the cyclone design, and
particularly on the geometry at and below the dust outlet orifice (Svarovsky,
1984). This top limit can be as high as 15 in H20 (3,740 Pa).
These recommended limits for pressure drop are demonstrated in Fig. 8.6.
If a given dust is fed into a cyclone and the input dust concentration is kept
constant but the flowrate Q is varied, the total efficiency ET first increases
until a maximum is reached (point B) beyond which it falls off. Theoretically,
it should carryon increasing as indicated by the dashed line in Fig. 8.6 since
higher tangential velocities due to higher flowrates cause stronger centrifugal
fields and should improve separation. In practice, increasing turbulence at
higher flowrates causes reentrainment of particles and, at very high flowrates,
the efficiency falls rapidly. Although, with different dusts, different curves of

The effects on the two most important operating characteristics, flowrate Q


and cut size dso, can be summarized as follows. The flowrate Q (capacity) is
increased by:
(a) Increasing the pressure drop
(b) Increasing the feed solids concentration
(c) Design changes, leading to lower resistance to flow (Larger inlet and gas
outlet orifices, for example)
Operating cut size dso is reduced (i.e. mass recovery is increased) by:

ET = f(Q) would be obtained, the position of the maximum is approximately


constant, and from it a general range of optimum operation can be derived.
Thus, a most desirable range of operation is clearly just before the optimum,
between points A and B, where high separation efficiency is achieved at
reasonable pressure drops tip (or pressure heads H). It should be noted that
as the pressure head H increases with Q (also plotted in Fig. 8.6), operation
beyond the maximum would mean higher energy consumption and hence
higher running costs.
Points A and B are usually found to correspond to H of 40 to 100 m of gas
column, respectively, and this is the recommended operation range. There
are, inevitably, slight variations between values recommended by different
manufacturers, since different cyclone designs have somewhat different
characteristics.
The effect of flowrate Q on grade efficiency G(d) follows the same pattern
as with the total efficiency - the best grade efficiency curve corresponds to
the maximum ET
The increase in separation efficiency with the pressure drop tip, within the
recommended range of tip and for a given cyclone, can be demonstrated
using Eqs (8.1), (8.2), and (8.3). For a given cyclone, gas, and solids, the
equations can be combined to show the dependence of dso on tip as follows:
dso ex: (tip)-lf4

(8.5)

Alternatively, for the same pressure drop and variable cyclone diameter, Eqs
(8.1), (8.2), and (8.3) give another proportionality:
dso ex:

(Dc)lf2

(8.6)

which provides the argument in favour of multi-cyclone units as mentioned


earlier.

(a) Increasing the pressure drop (up to a limit, as discussed above)


(b) Increasing the feed concentration (particle surface properties and gas
humidity play an important role here)
(c) Design changes, usually leading to higher resistance to flow, like reduction in the inlet and outlet orifices: smaller diameter of cyclone body
Dc reduces the cut size
As can be seen from the above points, flowrate and cut size are not
independent and most of the operating and design variables affect them both.
The use of the knowledge of the two dimensionless groups Eu and Stkso for
cyclone scale-up and design is demonstrated in the following example.
Example 8.1

Determine the diameter of a gas cyclone and, if necessary, the number of the
cyclones to be operated in parallel to treat 2.5 m3/s of ambient air (viscosity is
18.25 x 10-6 N s/m2, density 1.2 kg/m3) laden with solids of density 2,600
kg/m3 at a pressure drop of 1200 N/m2 and cut size not more than 6 p.m.
Take the resistance coefficient Eu (related to the superficial velocity in the
cyclone U) equal to 46 and the Stokes number, corresponding to the cut size,
equal to 0.006.
Solution

Cyclone diameter can be calculated from Eqs (8.1) and (8.2):


2

2.5 x 4

Tr.

D=--

1.2 x 46
1,200 x 2

---

This is for the case of one cyclone to treat the whole of the flow. Check on cut
size: from Eq. (8.3):
0.006

x 18 x 18.25 x 10-6 x
2,600 x 4 x 2.5

1T X

0.6953

This is greater than 6 ,um and therefore unsatisfactory. The calculation can
be repeated by using Q12, Q13, etc., until the cut size becomes less than 6
,um. This is reached for Q15, so the solution is that there should be 5 cyclones
operating in parallel, each 0.31 m in diameter.
It should be stressed here, however, that the scale-up procedure shown
above is only approximate because there are effects other than inertial
separation which determine the collection efficiency, such as dust reentrainment and electrostatic effects, which do not necessarily scale up in the same
way.
The feed concentration of solids is known to affect the separation
efficiency. Concentrations higher than about 1 glm3 usually lead to higher
efficiency because of the increased rates of interparticle collisions and the
resulting agglomeration. Most cyclones serving fluidized beds process gas
having concentrations much higher than 1 g/m3 in some cases. It is therefore
fortunate that, if a cyclone is designed on the basis of the scale-up described
here, the estimate of the expected collection efficiency at higher concentrations is usually conservative.
8.4.5

Abrasion and attrition in cyclones; clogging

Abrasion in gas cyclones is an important aspect of cyclone performance and is


affected by the way cyclones are installed and operated as much as by the
material construction and design. Materials of construction are usually steels
of different grades, sometimes lined with rubber, refractory lining, or other
material. Abrasion is worst in two critical zones within the cyclone body: in
the cylindrical part just beyond the inlet opening and in the conical part near
the dust discharge. Abrasion is a serious problem in particular in the power
generation industry and in the production of building materials, although
separation of apparently soft materials like wood chips is also accompanied by
some abrasion.
There has been little published on the effect of operating variables of life
expectancy of cyclones, the most comprehensive study being a Czech
investigation published recently (Storch, 1979), but there is a great need for
more information. This is particularly imperative in applications where
external gas cyclones are operated at high absolute pressures and thus have to
be insured as pressure vessels. The risk of explosion because of thinning of
the cyclone wall due to erosive wear is eliminated in many fluid catalytic

cracking units by housing the cyclones in a common pressure vessel which is


itself not being eroded by the dust.
Erosive wear by a given dust is affected by five most important variables:
design velocity U, direction of particle motion (radius of curvature of its
trajectory R), number of impinging particles, particle size, and time. For th~
same feed material and same cyclone geometry, the abrasion of a cyclone is
proportional to an abrasion index AI:
U"c
AI=R

where U is a characteristic velocity, c is mass concentration of solids, and R is


the radius of the cyclone body. The exponent n is usually between 4 and 5 for
gas cyclones, which indicates a very strong effect of the flow velocity; more
than one cyclone in series, with the first one run at low pressure drop, is a
better alternative than only one stage at a high pressure drop (= high
velocity). The first stage is used as a pre-collector collecting oversize particles
to reduce the concentration and particle size for the second stage, which can
then be run at a higher pressure drop. Reduced concentration and increased
cyclone diameter also reduce abrasion according to Eq. (8.7). The requirement of low velocity is stronger, however, and to achieve a given recovery,
smaller diameter cyclones operated at low pressure drops are actually better
than large diameter units which, for the same recovery, would have to be
operated at higher pressure drops.
The above abrasion index can be used in life expectancy predictions from
tests with the same material on a smaller cyclone, but the actual value of the
exponent would have to be ascertained. Cyclone manufacturers probably
have the best chance to obtain enough data for abrasion studies but they
usually do not publish such information.
Attrition of solids is known to take place on collection in gas cyclones but
little is known about how it is related to particle properties which particle
sizes are most affected, and how attrition can be related to abrasion of the
cyclone walls. Clearly, large particles are more likely to be affected by
attrition so that finer fractions are generated by knocking off corners or
complete breakage of the larger particles.
Attrition is most detectable in recirculating systems such as fluidized beds
where cyclones are used to return the carryover material back to the bed. The
complete inventory of the bed may pass through the cyclones many times per
hour and the effect of attrition is thus magnified and therefore easily
measured.
Even in single-pass systems, however, attrition is sometimes detected and it
demonstrates itself by severe discrepancies in mass balance of different
particle size fractions around the cyclone. It is not uncommon to find that
what is supposed to be the coarse product, i.e. the collected solids, is in fact

just or nearly as fine as the fine product in the gas overflow. In isolated cases,
when the feed solids are very fragile or have a shape which is easily broken
(thin-shelled hollow spheres, scale particles, or such like), the coarse product
may even be finer than the solids remaining in the gas flow and reporting to
overflow. Any meaningful evaluations of grade efficiency curves or any other
assessments of the separation performance of the cyclone are then impossible.
There is clearly a need to investigate the mechanism of attrition with the
view to relating it to the fracture properties of the solids, and to develop a
realistic attrition 'index', similar to that used for abrasion in cyclones (Eq.
8.7). An empirical correlation for abrasion, specific to fluid catalytic cracking
applications, has been proposed by Zenz (1975) and there are several similar
tests used by catalyst manufacturers and users. These are usually based on
attrition in a high velocity jet which enters a small bed of the powder.
Clogging of cyclones is the greatest single source of failures in their
operation. This is usually caused by plugging or overloading of the dust outlet
orifice and as soon as this happens, the cyclone cone quickly fills up with dust,
pressure drop increases, and recovery reduces or ceases altogether. Clogging
may be caused by a variety of reasons such as mechanical defects (bumps on
the cyclone cone, non-circularity of the dust outlet, weld or gasket protrusion,
etc.), changed chemistry, or physical properties of the dust: condensation of
water vapour for example, is a sure cause.
Smaller diameter cyclones are more likely to clog than the large ones; there
is a top limit in feed solids concentration (above which clogging is likely to
occur) for each cyclone and, for given solids, it is roughly proportional to the
cyclone diameter (Svarovsky, 1981b).
8.4.6

Discharge hoppers and diplegs

The design of the dust discharge end is very important for correct functioning
of a gas cyclone. Any inward leakages of air at that end lead to sharp
deterioration of separation efficiency (due to dust reentrainment) while if the
cyclone is under pressure, outward leakages may marginally improve
recovery but lead to pollution and loss of product. It is therefore best to keep
the dust underflow as airtight as possible.
The strong vortex inside a cyclone reaches into the space underneath the
dust outlet and it is most important that no powder surface is allowed to build
within at least one cyclone diameter below the underflow orifice; the greater
the distance the better. In order to reduce the headroom necessary to fulfil
this requirement, a conical vortex breaker is sometimes used just under the
dust discharge orifice which stops the vortex intruding into the discharge
hopper below. This design practice is particularly common in Germany. Some
companies use a 'stepped' cone to counter the effects of reentrainment and
abrasion, and a recent investigation (Svarovsky, 1984)has proved the value of
this design feature.
.

In fluid catalytic cracking beds with internal cyclones, diplegs are used to
return the collected carryover particles into the fluidized bed. These are
vertical pipes of 'chutes' connected directly to the dust discharge orifice of the
cyclone (a stepped cone is usually used in the lower part of the cyclone) or to
the bottom of its discharge hopper (if such a hopper is fitted), extending down
into the fluidized bed below. Returning particles flow down the dipleg in the
lean phase but in the dense phase flow into the lower part before entering the
bed. The level of the dense phase in the dipleg is always higher than the bed
surface and provides a necessary resistance to reduce the upflow of the air
which would reduce cyclone efficiency. Trickle valves are usually fitted to the
bottom end of diplegs in order to provide the seal at start-up, when no
powder is yet returning into the bed; such valves do not allow any return of
the solids into the bed until a sufficient head of solids has accumulated in the
dipleg (see Chapter 12).
As to the designs of diplegs, Zenz (1975) recommends a clear dipleg height
of at least 7 feet (2.1 m) and a minimum inside diameter of 100 mm. The
solids downflow capacity of a dipleg pipe can be calculated from an existing
correlation (Zenz, 1975), and this may be used to size the pipe from the
properties of the solids and the amount of solids returned by the cyclone.
8.4.7

Cyclones in series

The solids recovery of a single cyclone does not continue to rise with applied
pressure drop above a maximum. The recovery can be further increased by
connecting cyclones in series, and this is often done in practice. No more than
two cyclones should be used, however, unless steps are taken so that the cut
size of the subsequent stages is made progressively lower by 'tightening up' on
the design (primary stage of medium or low efficiency design and further
stages of progressively more efficiency design or smaller diameter). The first
stage often has a high dust loading and the operating cut size is thus reduced
by agglomeration; the second stage, if identical in design, may operate at a
greater cut size on account of it receiving much more dilute feed. This does
not make good utilization of the second stage as its recovery is small; it is like
using a fine screen followed by a coarse screen, and design changes in the
second stage should therefore ensure that its cut size is as small or smaller
than that of the first stage.
Apart from the resulting gain in recovery, two-stage systems are also
advantageous for separation of fragile, agglomerated or abrasive dusts in that
the first stage is then designed to operate at low inlet velocity. A large
diameter primary cyclone may be used to collect the grit which would plug or
erode the high efficiency cyclone in the second stage. The two-stage systems
also offer additional reliability in that if the primary cyclone plugs, the
secondary still collects.

The series connection of cyclones is not always necessarily in the direction


of the gas overflow; sometimes it is advantageous to draw off 5 to 15 per cent
of the gas flow through the dust outlet orifice and separate the concentrated
suspension in a small secondary cyclone. Such an arrangement does not
improve the overall recovery of the plant, however, other than through the
beneficial effect it has on the recovery in the first stage.
Several pressurized fluidized bed combustors under test worldwide use a
conventional series connections on overflow, however, with up to four stages
(NCB/CURL 1981;Ernst et al., 1982;Fluidized Combustion Contractors Ltd,
1982). Cyclones are also widely used in series connections with other gascleaning devices such as filters, electrostatic precipitators, or scrubbers,
usually as precollectors, to reduce the load of the high efficiency units that
follow.
8.5

AEROMECHANICAL

WET SEPARATORS (SCRUBBERS)

The separation of particles from gases is aided in scrubbers by the addition of


a liquid, usually water. The presence of water plays an important role in all of
the three phases of particle collection outlined earlier. In the first phase,
particles separate on collecting surfaces, which are either in the form of water
droplets or the wet surfaces of the scrubber structure; in either case, particle
reentrainment is virtually eliminated and this means that scrubbers are
invariably more efficient than the dry separators such as cyclones.
In the second phase, the collected particles are washed off the collecting
surfaces in a film of water and this again is done without any danger of
particle reentrainment into the flow; any reentrained water droplets are
separated in a spray eliminator which is incorporated in most scrubbers.
Finally, in the third phase, the solids are removed from the scrubber in the
form of a slurry, which is more easily handled than dry powders.
All of the above advantages in using water to enhance particle separation
lead to one major disadvantage in that the collected solids may then have to
be separated from the water. Thus what was initially a solid-gas separation
problem becomes a solid-liquid one.
The most important collection mechanisms employed in scrubbers are the
inertial principle, interception, and diffusion; secondary mechanisms include
diffusiophoresis (i,e, mass transfer in the form of condensation leads to a
force on particles causing their deposition), thermophoresis (deposition due
to a temperature gradient), and particle growth due to the condensation of
water on the particles.
.
The grade efficiency function for those types of scrubbers where the inertial
principle is the prevailing separation mechanism was shown by Calvert et al.
(1972) to be of the following general form:
G(d)

1 - exp (-AdB)

(8.8)

where A and B are constants. Packed towers, sieve plate columns, and
Venturi scrubbers (for 2 < Stk < 8) follow the above equation with B = 2,
while for centrifugal scrubbers B is about 0.67. The constant A depends on
the scrubber type, specific water consumption and other variables, and on the
nature of the dust (particularly its affinity to water).
In the other types of scrubbers, where diffusion and other additional
principles contribute aPt'reciably to the separation, more complicated
mathematical models have to be sought (Calvert et al., 1972)

The electrostatic precipitator is an apparatus which cleans gases by using


electrostatic forces to remove the solid or liquid particles suspended in a gas
stream. The dust-laden gas is passed through an intense electrostatic field set
up between electrodes of opposite polarity. The field charges the particles
and generates the separation forces leading to particle collection. The
separation forces are applied directly to the particles, without any need to
accelerate the gas, and this results in a very low power requirement.
Furthermore, the separation efficiency is less dependent on particle size than
with the mechanical, wet or dry, separators; hence electrostatic precipitators
achieve high efficiencies even with extremely fine particles. However, they
are found on few fluidized bed installations and will be discussed no further
here. A thorough discussion can be found in Svarovsky (1981b).
8.7

FILTERS

Gas filtration may be defined as the separation of particles from gases by


passing a gaseous suspension through a porous, permeable medium which
retains the particles. Depending on where in the medium the particle will
separate, two extreme cases can be recognized:
(a) Dust cake filtration, where the particles are deposited in the form of a
cake on the upstream side of a relatively thin filter medium (in singlelayer filters).
(b) Depth filtration, in which particle deposition takes place inside the
medium (in packed filters).
In many practical applications, such as fabric filtration, both cake and depth
filtrations take place, with the latter prevailing in the initial stages of a cyclic
filtration operation and the former prevailing towards the end. In other cases,
such as in sand filters or high inertia fibrous filters, depth filtration alone
takes place throughout the whole cycle. A great majority of the available
texts on gas filtration completely neglect the importance of cake filtration in
industrial fabric filtration and build the necessary theory purely on the basis
of inertial separation and diffusion taking place within the filter medium.

217

SOLID-GAS SEPARATION

The first possible classification of gas filters is according to the dust


concentration or 'loading' they are designed for. The 'primary' gas filters
(mostly bag filters or gravel bed filters) are designed for high dust loading,
usually from 0.2 to about 200 g/m3 because no gas filters can be used above
this limit. The 'secondary' filters are generally employed in lighter dust
loadings of no more than 100 or 200 mg/m3, such as the air-conditioning, and
high efficiency filters.
High dust loadings and high gas volumes can be treated in bag filters. The
filter medium is made from a relatively thin cloth in the shape of a cylinder,
tube, pocket, or bag, operated in a vertical position to facilitate dust removal.
The typical length/diameter ratio of the bags is 20:1, with the diameter from
120 to 150 mm. In exceptional circumstances, the bag length may be up to 10
m.
The filtration mechanism is predominantly that of dust cake filtration, the
cake being formed either on the outside or on the inside of the bag depending
on which way the gas is arranged to flow. In order to be able to treat larger
volumes of air in the smallest possible space, multi-bag arrangements are used
in which many bags are operated in parallel.
The design velocity for a given duty is usually selected from experience with
the same or similar material. The most common range for separation of fine
dusts in bag filters is from 1 to 1.5 crn/s, except for pulse-jet systems which
allow higher velocities up to 3 crn/s. Higher velocities can be used with coarser
dusts. It is not uncommon to find that some installations simply cannot be
operated above a certain limiting value of filtration velocity, when extensive
blocking occurs.
Filtration in bag houses is a cyclic operation and within each cycle the
resistance of the filter changes to the gradual deposition of dust on the filter
medium. A curve which is characteristic for each operating system is the
relationship between pressure drop and the specific mass of the dust collected
on the medium (this is proportional to the volume of gas filtered, at constant
dust loading).
The cleaning cycle in bag filters is triggered either by a timing circuit or
more often by a pressure switch.

8.8 NOMENCLATURE
A
AI
B
c
C
dso

cyclone body length


abrasion index
length of the cone
dust concentration
length of cylindrical part
cut size (equiprobable

size)

m
(see Eq. 8.7)
m
kglm3
m
m

Dc
E
Eu
F
K
L

M
n
R
Q
Stkso
U
t:.p
JL
Pg
Pp

cyclone diameter
dust outlet diameter
Euler number
length of gas outlet pipe
inlet height
inlet width
gas outlet diameter
exponent
radius of cyclone body
gas flowrate
Stokes number for dso
superficial gas velocity based on Dc
static pressure drop
gas viscosity
gas density
particle density

m
m
m
m
m
m
m
m3/s

m/s
N/m2
kg/ms
kg/m3
kg/m3

REFERENCES
Calvert, S., Goldschmid, J., Leith, D., and Mehta, D. (1972). Scrubber Handbook,
Natl Techn. Info. Service, Springfield, Va., PB 213-016.
Ernst, M., Hoke, R.c., Siminski, V.J., McCain J.D., Parker, R., and Dremmel, D.C.
(1982). 'Evaluation of a cyclone dust collector for high temperature, high pressure
particle control', Ind. Eng. Chem. Proc. Dex. & Dev., 21, 158-161.
Fluidized Combustion Contractors Ltd (1982). 'Observation and analysis work
associated with a 1Q0l}-hourtest program in a pressurized fluidized-bed conbustion
facility', EPRI CS-2582, Project 979-3, Topical Report, Sept.
NCB/CURL (1981). 'Fluidized bed combustion, 1000hour test program', Vols 1, 2, 3,
US/DOE/ET-10423-1101, Sept.
Storch, 0., et al. (1979). Industrial Separators for Gas Cleaning, Elsevier, Ansterdam.
Svarovsky, L. (Ed.) (1981a). Solid-Liquid Separation, 2nd ed. Butterworths, London.
Svarovsky, L. (1981b). Solid-Gas Separation, Elsevier, Amsterdam.
Svarovsky, L. (1984). 'Some notes on the use of gas cyclones for classification of
solids', in Proc. First European Symposium on Particle Classification in Gases and
Liquids, Nuremberg, May 9-11, Dechema.
Zenz, F.A. (1975). 'Cyclone separators', Chapter 11 in Manual on Disposal of
Refinery Wastes, Volume on Atmospheric Emissions, API Publication 931, May.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Fluid Bed Heat Transfer

Among the advantageous properties of gas fluidized systems is the extremely


large area of solid surface exposed to the gas. Thus, a cubic metre volume of
100 ~m diameter particles has a surface area greater than 30,000 m2, which is
of a similar magnitude to that of the surface area of the Great Pyramid of
Cheops. Such colossal solid surface area greatly facilitates solid-to-gas heat
and mass transfer operations even though it may not all be accessible to fresh
gas. Additionally, because of the solids mixing generated within the bulk of a
bubbling gas fluidized bed, temperature gradients are reduced to negligible
proportions and the bed possesses a very high effective thermal conductivity
in the vertical direction. High rates of heat transfer are obtainable between
the fluidized solids and an immersed transfer surface. This is also primarily a
consequence of the mixing within the bulk of the bed and of the high thermal
inertia of the solids. The latter property has been used in many industrial
operations, e.g. in the Cat Cracker to transport heat between the regenerator
and reactor, although rarely, as yet, has it been explicitly exploited for its own
sake. Indeed, the advantageous thermal properties of gas fluidized solids as a
high temperature heat storage and transport medium of very low vapour
pressure has still to find application beyond some metallurgical heat
treatment process (Baskakov, 1968; Vier, 1976).
Under the right operating conditions, a gas fluidized bed constitutes a very
satisfactory thermal system. Nevertheless, there have been disappointments
- a consequence of the complexity of bubbling fluidized beds (see
particularly Chapters 4 and 5). Theoretical models descriptive of the heat
transfer mechanism are of limited application because they require knowledge of parameters which are not generally available. Finally, the reader
should be warned against using empirical correlations without first checking
219

back to see that they have been determined


particular problem.

from tests of relevance to the

Because of the very large surface possessed by a mass of small particles,


particle-to-gas heat transfer is only rarely a limiting factor in fluidized bed
heat transfer despite the fact that, based on the total particle surface area,
coefficients are typically only of the order of 6 to 23 W/m2 K. (The exceptions
are such instances as drying, where the need to provide the latent heat of
vapourization can give the bed material a very high effective heat capacity, or
with particles which are reacting and where the reaction is very temperature
sensitive). There is much scatter in the range of the published coefficients
and, indeed, there are many uncertainties involved in their experimental
determination. There is the problem of measuring the relevant gas and solid
temperatures
(Botterill, 1975). Singh and Ferron (1978) have reported
differences between gas and particle temperatures using a 'light-pipe probe'
and thermistor to measure the radiant energy from the particles. Thus, for a
bed of 230 Mm catalyst being cooled from 200C by fluidizing air at ambient
temperature,
differences as large as 70C were detected between gas and
particle temperatures. They suggest that the temperature indicated by an
immersed thermocouple corresponds to a weighted mean. which is 80 per
cent. of the particle temperature and 20 per cent. of that of the gas. Apart
from problems in making the actual measurements, there is also the problem
of interpreting them. Assumptions have to be made as to the nature of the gas
flow through the bed. It was pointed out by Kunii and Levenspiel (1969) that
the most realistic model for this is a plug flow model. Measurements
interpreted on this basis by different investigators are generally consistent
with each other whereas those interpreted on the basis of a model involving
complete gas mixing are widely scattered and give excessively low particle
Nusselt numbers. (The measured temperature gradient at the entry to the bed
is too steep to permit direct distinction between the two models; thus
Richardson and Ayers (1959) noted that the gradient was undetectable after
the first 2 mm or so from the point of entry into beds composed of mean
particle diameters of 200 Mm and less; see also Example 9.1 below.)
Experimental
results, covering particle Reynolds numbers up to 50 and
analysed on a plug flow model (Kunii and Levenspiel. 1969), are correlated
by the equation:

The low limiting Nusselt number obtained can be accounted for by the fact
that not all of the total particle surface area is available for heat transfer with
the gas. Zabrodsky (1963) proposed a model involving 'microbreaks',
pointing out that the gas did not necessarily flow uniformly through the
continuous phase of the bed. Littman and Sliva (1970) showed that,
particularly in packed beds, there is a strong dependence on Reynolds
number because the region near the point of contact between particles is not
fully accessible to the flowing fluid. Reference is made (Botterill, 1975) to two
studies which report that higher coefficients are obtained when the system is
operated at elevated pressure.
For a well-mixed bed of particles of low Biot number (i.e. negligible
internal thermal resistance) operating at a steady temperature, a simple heat
balance relates the change in temperature
of the fluidizing gas to its
penetration into the bed. Thus, on the basis that the gas exchanging heat with
the bed is in piston flow through the continuous phase, then:

for an element of bed dl m deep and of unit cross-sectional area where IlTg is
the change in temperature of the gas flowing through the element of bed and
S6 is the surface area of solids per unit volume of bed. This can be integrated
directly to give:

In

Tg - Tp )
~-~
( Tg.in - Tp

=-

( --hgp S6 ) l
U Pg Cg

6(1 - p)
dp

The distance ln, in which the gas-to-particle temperature


factor:
Tgin - Tp
n=
Tg - Tp

dg

Cg

p~.3 l.f1.3 kg (1 -

5.5 In n

M1.3

difference falls by a

Example 9.1
A bed of 450 }Lm particles is operating at 150C. the temperature and
superficial velocity of the incoming gas are 550C and 0.4 mIs, respectively.
Approximately how far will the incoming gas have penetrated into the bed
before it is cooled to 350C?
Solution
This estimate can be made using Eq. (9.4) and the gas physical properties for
the average temperature over the specified range, i.e. between 550 and
350C. At this temperature of 450C the appropriate physical properties are:
Pg

= 0.68

kg
Cg

=
=

}L

kg/m3
0.04 W/m K
1025 J/kg K
2.8 X 10-5 N s/m2

transfer and a simplification for the situation where that resistance may be
neglected.

Under the usual bubbling fluidized bed operating conditions, the heat
transfer coefficient h between an immersed surface and a gas fluidized bed
can be thought to consist approximately of three additive components
(Botterill, 1975), namely:
(a) The particle convective component hpc, which is dependent upon heat
transfer through particle exchange between the bulk of the bed and the
region adjacent to the transfer surface
(b) The interphase gas convective component hgc, by which heat transfer
between particle and surface is augmented by interphase gas convective
heat transfer and

The gas velocity at 450C for an inlet velocity of 0.4 m/s at 550C will be 0.35
m/s:

550 - 150
_ 150 = 2

= 350

5.55 In 2(2.8 x 10-5)1.3 (450 X 10-6).7 1,025


l - ----------,,....-------- 1.4 mm
2
0.680.3 x 0.353 x 0.04(1 - 0.4)
Note that because Umf is approximately proportional to d~ for laminar flow
conditions. 1 should not be very sensitive to particle diameters for smaller
mean diameters.
However, for particles wjth group D (Geldart 1972)
characteristics, the gas flow conditions will be transitional to turbulent.
Negligible temperature
gradients will be set up within small particles
because the relevant Biot numbers will be small generally, they may be
ignored if the particle Biot number is less than 1/10. Barile, Seth, and Williams
(1970) have studied the transfer of heat between the gas in a rising chain of
bubbles and the continuous fluidized phase. Measured temperature profiles
indicated a rapid approach to bubble/dense phase equilibrium (within the
order of length of one bubble diameter), but this could still be significant
when a highly exothermic reaction is occurring.
McGaw (1976) analysed the problem of gas-to-particle heat transfer in a
cross-flow moving packed bed heat exchanger which is relevant to the design
of particle coolers and circulating bed regenerative systems. He gave both a
rigorous analysis, taking into account internal particle resistance to heat

approximate
range of
significance

40}Lm ~

1 mm

>

800}Lm

and at higher
static pressure

higher temperatures
(> 900 K) and
difference

The particle convective component hpc is particularly dependent on bed


bubbling behaviour (Section 9.3.1) and, at higher fluidizing velocities when
bubbles increasingly shroud the heat transfer surface, bed-to-surface heat
transfer by this mechanism is reduced. Allowance can be made for this (e.g.
Baskakov et al., 1973) by reducing the contribution of the particles convective
component according to the fraction of the time that the transfer surface is
shrouded by bubbles, but this is not generally known.
The situation, however, can be expected to be different when the
operational range is extended into the 'turbulent' and 'fast fluidization'
regimes (see Chapter 7). Much higher gas velocities are then employed. For
particles of millimetre order diameter, the velocity for turbulent fluidization
approaches that of the particle terminal velocity, and bed behaviour is
characterized by a high degree of solids eddy motion superimposed on
cocurrent and countercurrent solid/gas flow (Staub, 1979). For powders of

group A, the operating velocities for fast fluidization exceed the particle
terminal velocity and there is very extensive back mixing in the transporting
solids (Yerushalmi and Cankurt, 1979). Under these conditions, it would
seem profitable, as Staub (1979) suggested, to apply the correlation methods
earlier developed for flowing dense solid/gas suspensions. This is briefly
referred to below (Section 9.4).
9.3.1

Particle convective component hpc

The particle convective component is responsible for the characteristic


marked increase in the bed-to-surface heat transfer coefficient observed as
the particle circulation develops when the fluidized bed begins to bubble (Fig.
9.1). With a continuing increase in the fluidizing gas flowrate, the coefficient

pressure, so It IS the particles which have the capacity to transport heat


through the bulk of the bed. (Not until fluidizing beds of group D material is
there significant heat transfer by gas convection and the interphase gas
convective component hgc becomes truly important.) Thus, particles in the
bulk of the bed exchange heat with the tluidizing gas and, by conduction
through the gas, with each other. Particles usually stay within the bulk of the
bed long enough to come to the same temperature as their neighbours. Then
a 'packet' of particles at the bulk bed temperature is swept into close
proximity with the heat transfer surface under the bubble-induced circulation
patterns occurring within the bed. When they first arrive close to the heat
transfer surface there is a high local temperature difference between the bed
material and the surface; consequently heat transfers rapidly. The longer the
particles reside close to the surface, the more closely the surface and local bed
temperatures approach. Highest mean temperature differences are therefore
to be expected under those bubbling bed conditions which can generate a
rapid exchange of material between the vicinity of the transfer surface and the
bulk of the bed, i.e. with low particle residence time adjacent to the transfer
surface as suggested by the basic Mickley and Fairbanks (1955) model which
can be expressed in the form:
-----------1

V(kmf

Pmf

Cp S)

The terms representing the thermal conductivity and density of the bed, kmf
and Pmf, are for the material at the condition of minimum fluidization
representative of the condition of the 'packet material'; Cp is the heat capacity
of the particles; and S is a 'stirring factor' representative of the frequency of
50

40

r .
""30
0,

:b .!?
., 0;:

20

.:!:
_ +:
u

Figure 9.1 Effect of gas flowrate on bed-to-surface heat


transfer for powders of groups A, B, and D characteristics
when radiant transfer may be neglected.

passes through a maximum. With finer group A powders, which expand


stably after the minimum fluidizing velocity is exceeded before bubbling
begins, there is an additional minor peak in the rising portion of the curve
(Khan, Richardson, and Shakiri, 1978).
It was Mickley and his coworkers who first appreciated the part played by
the particle heat capacity in the process. The volumetric heat capacity of the
particle is of the order of a thousandfold that of the gas at atmospheric

.....
u "

:; 8
0(;
o E

~~
Cl:_
2 3 4 6 810

2 3 4

6 8 10

Ratio of particle-to-gas

2 3 4

6 8 10

thermol conductivity

3 4

6 8 10

kplkg

Figure 9.2 Diagram for estimation of effective continuous phase thermal conductivity
(Baskakov, 1968). Reproduced by permission of the author.

material replacement at the surface. The bed thermal conductivity may be


estimated using the plot given in Fig. 9.2 (following Baskakov, 1968)but the
stirring factor S is not generally known (see later and Eq. 9.13). A limit to this
unsteady state model, however, is reached with larger particles when their
time constant for cooling t exceeds the particle replacement rate, and the heat
transfer process more nearly approximates to a steady state process, i.e.:

t--

1 Pp Cp d~
36

---->
kg

r-1

Decker and Glicksman (1981) then suggest the use of the heat transfer
resistances measured for packed beds in the estimation of rates of heat
transfer.
The chief resistance to heat transfer lies in the low thermal conductivity of
the fluidizing gas, and this is the limiting factor. Particle-particle and
particle-surface contact areas are too small for significant heat transfer to

~'~~
~
i2 800

"'e

~i

Reduci~

i~
!

Hmf

Gro~p A
Some
stobie
bed
ex pons ion

)213

h'

:f.i~1 ,.
Group C

occur through points of solid contact; heat must flow by conduction through
the gas and there is evidence of additional resistance to heat transfer in the
region of increased voidage at the transfer surface (Botterill, 1975). The
strong inverse dependence of coefficient upon particle size over the range 800 to 80 porn (Fig. 9.3) is a consequence of the increase in the percentage
area of transfer surface through which the heat can flow to the particles by
short transfer paths. With very fine powders, however, bed circulation is
inhibited by interparticle forces, and this is responsible for the marked fall in
coefficient as particle size is further reduced and the bed has the fluidization
characteristics of a group C system. Bed-to-surface coefficients increase with
an increase in the gas thermal conductivity upon operating beds at higher
temperatures (see below, Fig. 9.5). It is not until particle size and/or static
operating pressure are sufficiently increased for the fluidizing gas flow
condition through the emulsion phase to enter the transitional or turbulent
flow regime that significant interphase convective transfer occurs through the
gas (hgc, mechanism (b), Section 9.3), and not until then does pressure have
a significant effect on bed-to-surface heat transfer apart from through its
minor influence on overall bed behaviour.
A simple model (Botterill, 1975) suggests that particle packing close to the
transfer surface will be important. The voidage there will be highest with
closely size bed materials and will be affected by the material size
distribution. Thus, bed-to-surface coefficients were observed to fall as
predicted for a reduction in particle packing density at the surface when a
heat transfer element was 'stirred' through a stably expanded bed to give low
effective particle/surface residence times:

Group B

~
~/

(Botterill et at., 1966). With increasing gas flowrates, for either group A or
group B materials, there comes a point where the effect of reducing the
particle residence time caused by the increasing bubble-induced mixing is
countered by the increased bed expansion and the blanketing effect of the
bubble flow across the transfer surface. The resultant maximum in the bed-tosurface coefficient (Fig. 9.1) occurs at velocities relatively closer to that for
minimum fluidization as mean particle size increases. Todes (1965) has given
the approximate correlation for Reop! which is used on Fig. 9.4.
For powders falling within Geldart's group A, the minor peak commented
on above (Fig. 9.1) is a consequence of some initial restricted particle
circulation by diffusive mixing as the bed fluidizes. Additional continuous
phase expansion leads to a reduction in the bed-to-surface coefficient. With a
further increase in gas velocity, the bed reaches the minimum bubbling

.e

Ql
..Q

~ 10

..,'"
g

v v

38o}'om
462}'om
533}'om
x
664}'om
+
775j.4m

887}'om
o 1276}'om
!l 2320 f'm

sand
sand
sand
sand
sand
sand
sand
sand

coal ash

~.p

820

p.rn

occurring in large scale equipment. Bubbles rise up the centre of the bed and
there is the return of solids 'stick/slip' flow down the column wall. The tests
have also most usually only been carried out at comparatively low
temperatures, although the results are presented in generalized correlations.
Van Heerden (1952) tested the then-available correlations with four
hypothetical systems. For one of these systems there was greater than a four
hundred-fold range in the predicted coefficient! The article by Gel'perin and
Einstein (1971), in which they propose a generalized correlation, further
illustrates the scatter in the reported measurements, and casts doubt on the
value of such correlations at all. Nevertheless, Zadbrodsky's (1966) approximate correlation for the maximum bed-to-surface coefficient neglecting
radiation effects is to be recommended for powders of Geldart's group B for
which the particle convective component predominates and with which there
is negligible stable bed expansion - generally within the mean size range of
100 to 800 pm mean diameter:

it;'

a:

Ar

Figure 9.4
Archimedes

Reynolds number, Reop" for maximum bed-to-surface heat transfer as a function of


number Ar and the correlation of Todes (1965) for Reap.'

condition and the continuous phase collapses back to a less expanded


condition. Particle packing is then denser close to the surface and the bed-tosurface coefficient increases under the action of the bubble-generated mixing
to reach the principal peak. Because of the greater degree of continuous
phase expansion obtained with group A materials, however, the coefficient is
less than the small particle size might first have led one to expect.
The design of the distributor and any internals will obviously have a strong
influence on the bubble-induced solids convection patterns established within
the bed (see Chapters 4 and 5; also Kunii and Levenspiel, 1969 ; Davidson
and Harrison, 1971; Botterill, 1975).
Work reported by de Groot (1967) illustrates the pronounced change in
bed behaviour consequent upon changing bubble development patterns
which can result from change in the scale of the equipment. It also illustrates
the great effect that a change in mean particle size or, rather, of fines
concentration of bed material can have through its effect on maximum stable
bubble size. As stressed in the introduction to this chapter, workers are
warned about the limitations of published correlations. They reflect the
particular behaviour of the experimental beds used, and because these have
usually been of small scale in which bubble size is limited by the bed diameter,
definite, simple solids circulation patterns develop unlike those often

This dimensional correlation was determined by following the response of


small spherical calorimetric heat probes which quickly approached the bed
temperature on immersion in the bed. It contains the particle density Pp,
which is related to its heat capacity and the other two key variables: particle
diameter dp and gas thermal conductivity kg. It allows well for the effect of
changing gas conductivity with a change in the operating temperature until,
beyond 600C, radiant heat transfer becomes increasingly significant (see
Example 9.2 and Fig. 9.5; see also Botterill, Teoman, and Yiiregir, 1981). In
Fig. 9.5 the differences between the values obtained with an ordinary oxidised
probe and one of fine gold, and hence of low emissivity, in tests with a bed of
380 pm sand directly indicate the size of the radiant transfer component at
higher temperatures.
Because of the limiting effect of the gas thermal conductivity, the absolute
temperatures of the bed and surface influence the obtainable bed-to-surface
heat transfer coefficient. Thus, higher heat transfer coefficients were
obtained when a hotter calorimetric probe was plunged into a bed of sand at
350Crather than when the probe was colder than the bed before insertion. If
bed-to-surface heat transfer coefficients are measured using a calorimetric
spherical probe of high Biot number so that it takes minutes rather than
seconds after immersion to reach the bed temperature, it is possible to
measure the effect of the changing surface temperature on the transfer
coefficient. Typically, for a bed of 270 pomalumina operating at 500C the
coefficient increased from 410 W/m2 K with the sphere surface at 75Cto 580
W/m2 K when it had nearly heated up to bed temperature (Botterill, Teoman,
and Yiiregir, 1984a).

significant. The value of Remf - 12.5 at Ar - 26,000 marks the approximate


boundary between the dominance of the convective and interphase gas
convective components (BotteriIl, Teoman, and Yiiregir, 1981, 1984b).
Example 9.2
Predict the change
diameter immersed
9.5). The bed is
conductivity of the
and Saxena (1958).

;;:
Ole

...

in bed-to surface heat transfer coefficient for a 10 mm


sphere as a function of operating conditions (data of Fig.
heated by combustion of propane and the thermal
fluidizing gas mixture can be predicted following Mason
The bed material is sand of close size distribution:

500

For air at l70oe, kg - 3.55

10-2 W/m K and using Eg. (9.9):

hmax = 35.8 (2,667)2 (3.55


= 366 W/m2 K

10-2).6 (480

10-6)-0.36

oe 380 /Lm sand


, 480 /Lm sand
530 /Lm sand

Table 9.1 Correspondence between predictions of the Zabrodsky correlation

(Eq. 9.9) and measured values for a bed of 480

Bed
temperature,
C

Excess air,
%

f.l.m

Estimated
gas
conductivity,
W/mK

Figure 9.5 Variation in maximum bed-to-surface heat transfer coefficient with temperature for beds of sand of different particles sizes and
predictions of Zabrodsky's equation (Eq. 9.9) (Botterill, Teoman, and
Yiiregir, 1981).

One can conservatively expect to obtain values of 70 per cent. of the


predicted maximum to a horizontal cooling tube immersed in a hot bed under
good operating conditions when kg is based on the bed temperature. This
reduction is a consequence of two effects: the disturbance caused to the local
fluidization condition by immersing the tube in the bed and the decrease in
gas thermal conductivity adjacent to a surface which is maintained consistently at a lower temperature than that of the bed (see Section 9.4 below).
Absolute pressure has little effect on bed-to-surface heat transfer coefficients except when fluidizing larger denser particles when the interphase gas
convective component of heat transfer (Section 9.3.2 below) becomes

170

360
540
620
800
960

Negligible
proportion
of combustion
products
300
90
70
40
20

sand fluidized by air

Estimated Maximum
coefficient measured
coefficient,
hmaf
W/m K
W/m2 K

3.55 x 10-2

366

370

x 10-2
x 10-2
x 10-2

428
492
512
572
617

430
500
510
610
690

4.6
5.8
6.2
7.45
8.45

10-2
x 10-2

For the case of smaller, less dense powders, Khan, Richardson,


(1978) give the correlation:

and Shakiri

from experiments with group A powders of mean diameters between 40 and


96 p.m. This has not been tested over a range of operating temperatures and
it could be expected that changes in gas physical properties resulting from
changing temperature would affect the fluidization conditions and influence
the bed behaviour, so influencing bed-to-surface heat transfer coefficients
(Botterill, Teoman, and Yiiregir, 1981). It is also a matter of experience that
the behaviour of large reactors can be sensitive to changes in the size
distribution of group A powders consequent upon the effect of changing fines
concentration through their generation by attrition or loss by elutriation.
Bed-to-surface heat transfer coefficients have been observed to vary by as
much as a factor of 2 in the course of operation. This is presumably a
consequence of change in bed behaviour adjacent to vertical heat transfer
surfaces so that the local stability of the bed can change, thereby affecting the
exchange of material between the surface region and bulk of the bed.
Many fundamental models have been proposed to describe the particle
convective heat transfer process (see Botterill, 1975) but these cannot
generally be applied for predictive purposes because the conditions and
thermal properties of the bed directly adjacent to the heat transfer surface are
unknown. An extreme situation is that for a slugging bed - the operational
condition of some high aspect ratio reactors. Donsi et al. (1984) analyse this
situation successfully on the basis of the slug spacing being the characteristic
length of relevance for the heat transfer process and including allowance for
additional solids transfer in the wake of the slug by analogy with the
behaviour of a slug in a liquid system. A simple model containing the
principal features of the fluidized bed condition is the packet replacement
model (Mickley and Fiarbanks, 1955). However, it predicts excessively high
coefficients for very short particle residence times adjacent to the transfer
surface. Accordingly, it has been suggested that perhaps the particles do not
come directly into contact with the surface and that there is an effective gap
between them. The replacement model was therefore modified to incorporate
a resistance to heat transfer in the region of increased voidage close to the
transfer surface. Gel'Perin, Einstein, and Zakovski (1966) give its solution in
the form:
h

w
=-1 [ 1--ln R
pc Ra
2Ra

1+-

2Ra)]
Rw

where Ra is the resistance to heat transfer within the packet phase and Rw is
the additional wall region resistance. Ra may be estimated from development

of Eq. (9.6) as:


T7T
(

kmf Pmf

) 1/2

Cp

T being the packet residence time adjacent to the surface. An approximate


correlation for the prediction of particle residence time adjacent to a surface
T and the time that the surface is shrouded
by bubbles to for vertical
cylindrical surfaces of 15 to 30 mm diameter immersed in beds of mean
particle size between 120 and 650 p.m and fluidized with carbon dioxide and
helium at temperatures
between ambient and 550C has been given by
Baskakov et al. (1973). That for the particle residence time takes the form:

d g

[ U~f(U/~mf-Af

to = 0.33

U
[

mt

(UIU _
d mf

11.1-1

d
D~J

A)2]

11.1-1

pg

where the value of the constant A varied with the bed material and the tube
diameter. However, packet residence time is usually unknown a priori,
although developments in the understanding of mixing processes (Chapter 5)
may eventually permit a realistic estimate of it and of the fraction of time
bubbles shroud the transfer surface. The resistance in the wall region Rw is
independent of time to a first approximation. More recently, this resistance
has been associated with an effective gap between particle and surface
originating from their inherent roughness (Decker and Glicksman, 1981).
However, if the resistance is considered to be generally the region of
increased voidage which extends to about half a particle diameter from the
surface and the heat transfer through this region is solely by steady state
conduction, Rw can be represented by:

where kw is the effective thermal conductivity in the near wall region and can
be estimated following Kunii and Smith (1960).
Very good correspondence
between observation and prediction was
obtained on this basis using Eq. (9.11) to predict the coefficients for flowing
packed bed experiments in which the particle residence time at the heat
transfer surface was known (Denloye and Botterill, 1977). Fortuitously, the
effect of a resistance represented by a 'gas gap' of one-tenth of a particle
diameter was tested in earlier theoretical studies (Botterill, 1975) and found

235
to correspond well in its predictions with observed experimental observations. It corresponds closely with predictions based on the more complicated Kunii and Smith model and is recommended for approximate estimates.
Rw would be greater and more complex in form if conditions had developed
such that there was a return flow of solids descending in 'stick/slip' flow at the
heat transfer surface which was occasionally disturbed by bubbles penetrating
to the surface. The basic forms of this latter model have been developed by
Yoshida, Kunii, and Levenspiel (1969) but, as stressed above, it cannot
usually be applied because the physical properties and condition of the
particulate film adjacent to the transfer surface are not generally known.
However, Bock and Molerus (1980) have demonstrated the basic validity of
this type of model in experiments on heat transfer to vertical tubes in beds up
to 1 m in diameter in which simultaneous measurements of local bubble
behaviour and time-averaged local heat transfer were made. The claims of
general applicability often made for specific mechanistic models, however,
are very questionable. Where bubbling processes are explicity modelled, it is
salutary to test out the implications of that element in the model and see if the
bubble flow predicted is credible (BotteriIl, 1986). In any case, no simple
correlation incorporates the change in mixing consequent upon the transition
from group D to group B type behaviour (see Section 9.3.2 below). It is often
necessary for the models, too, to fall back on limited experimental tests to
evaluate a 'constant' which appears in the model (e.g. Heyde and Klocke,
1979; Martin, 1980).

kmf
Pmf

may be estimated from Fig. 9.2


Pp (1 - E) = 2,600 x 0.6 kglm3.

Ra estimated for

Estimate heat transfer coefficients as a function of packet residence time for


sand of mean particle diameter 200 J.Lm fluidized in air. Assumed physical
properties are Pp = 2,600 kg/m3, kp = 18.9 W/mK, Cp = 800 Jlkg K; Pg =
1.2 kglm3, kg = 0.026 W/mK, Cg = 1,050 J/kg K; bed voidage, E = 0.4.

40 s as Ra

(kplkg

10kg

727)

401T)

= (

and

l/2

4 x 0.26 x 800 x 1,560


9.8 x 10-3 m2 KIW

Rw estImated as 0.1-

hpc for

dp

0.lx200x1O-6

kg

= ------

40 s

89 W/m2K

9.8 x 10-

0.026

7.7

10-4 m2 K/W

7.7 X 10-4
1 - ------In
2 x 9.8 X 10-3

+ 2 x 8.8

107.7 x 10-4
X

)il

R.,
Rw,
(m2 KIW) x 103 (m2 KIW) x 104
40
20
10

Example 9.3

as

5
1
0.5
0.1

9.8
7.0
4.9

3.5
1.6
1.1

0.51

7.7
7.7

7.7
7.7
7.7
7.7
7.7

89
120
163
213
379
418

713

Note, for comparison purposes, that Eq. (9.9) predicts a maximum value of
2
hpc of 414 W/m k, which seems to be a reasonable value for a freely fluidized
system. Following Kunii and Smith (1960), the alternative estimate for Rw
obtained is 7.1 x 10-4 m2 K1W.
h

pc

= -R1 [ 1a

-R In
2Ra

1+ 2Ra)~
-R
w

9.3.2

Interphase Gas Convective Component, hgc

The interphase gas convective component becomes significant when working


with solids of large mean particle diameter (> 800 p.m) and at higher static
operating pressures - Geldart's group D materials (Figs 9.1 and 9.3). The
gas flow condition through the bed is then turbulent or at least in the
transitional flow regime.

A theoretical model of some complexity for this component has been


proposed by Adams and Welty (1979). Experimentally,
Baskakov and
Suprun (1972) estimated its size by analogy from mass transfer measurements
and recommended the following correlations:

Nugc

U
O.0175Aro.46 PrO.33 ( Urn

E~peri.mental cond.iti~ns covered operation up to static pressures of 9 atm


wIth aIr, ~arbon dIoxIde, and argon as the fluidizing gases, with freon at
atmosphenc pressure and with mean particle sizes between 160 and 2,370
~m. The be~ materials included sand, ash, and copper shot and were closely
s~zed.. E~penmental values for the bed-to-surface coefficient components for
aIr flUIdIzed beds of sand operated at atmostpheric pressure and 6 atm as a
function of particles size are shown in Fig. 9.6.

)0.3

for Urnf

1 20C. 1 aIm
2 20C, Gatm
3 GOOC, 1 aIm

< U < Urn

where U is the fluidizing gas velocity and Urn the corresponding value for the
maximum bed-to-surface heat transfer. An alternative empirical correlation
for the interphase gas convective component has been derived by Botterill
and Denloye (1978) as:

i
~100

d1/2
g~

O.86Ar0.39

loJ

< Ar < 2 x

106

is

which has dimensions of m -1/2. The assumption was made that it measured
by the heat transfer coefficient at the point of fluidization. This is not strictly
true because, although no particle convection will occur with the bed in its
quiescent state, there will be conductive heat transfer through the gas
between the particle and surface close to the 'point bf contact' to augment the
interphase gas convective component when the particle thermal time constant
is larger than its residence time close to the surface (Decker and Glicksman,
1981). The relative significance of this conductive contribution compared with
the gas convective component reduces as particle size and operating pressure
increase, so the assumption is most in question for particles < 1 mm mean
diameter fluidized at atmospheric pressure at elevated temperatures. The
corresponding maximum particle convective component (assuming the two
components to be additive) was correlated against the Archimedes number by
the following dimensionless equation:

1,000
Particle diameler (fLm)

Figure.9.6 Effect of particle. size on maximum bed-to-surface heat transfer


coefficI~nt hma at atmospherIC and 6 atm pressure operation. Interphase gas
convectIve compo?ent hgc ta~en as that for quiescent bed; particle convective
component hpc estImated by dIfference (Botterill, Teoman, and Yiiregin, 1981).

Before using the above correlations (Eqs 9.16 to 9.19) it should be noted
that there is some evidence that fines in the size distribution can have a large
effect. Thus, contrary to the predictions of Eqs 9.16 to 9.19, Golan et al.
(1979) came to the conclusion that particle size had a negligible effect on

the bed-to-surface heat transfer with four beds of mean particle diameter
between 800 and 2,700 porn with top fractions ranging from 1,410 to 5,600
pm, but it is significant that the fines distribution (1,410 porn downwards)
was very similar in all four cases. Thus it is possible that the fines present may
have determined what was happening and affected the turbulence in the gas,
so reducing the contribution of the interphase gas convective component.
The Baskakov and Suprun (1972) predictions corresponded well with the
experimental coefficients obtained for quiescent beds of large particles at
ambient temperature and pressure, but increasing discrepancy was observed
as the static operating pressure was increased (Botterill and Denloye, 1978)].
The form of the component will be relatively insensitive to an increase in gas
velocity, and this has been substantiated by Catipovic et ai. (1980). This
is reasonable because there will be little change in the actual gas flow conditions through the continuous phase or adjacent to the transfer surface with
group 0 bubbling beds and, indeed, the maximum bed-to-surface coefficient is relatively insensitive to further increase in fluidizing gas velocity,
though some allowance should be made for direct heat transfer between the
gas flowing through the bubbles and the transfer surface (Catipovic et aI,
1980).
An effect of relative probe/particle heat capacity has been observed
(Botterill, Teoman and Yiiregir, 1984a). When, for example, bed-to-surface
heat transfer coefficients were estimated by following the response of a 10
mm diameter spherical calorimetric test probe suddenly immersed into a hot
bed of 2.3 mm diameter sand particles, the measured coefficients were
approximately 30 per cent. higher than the values predicted using the
correlations given above (Eqs 9.18 and 9.19). However, correspondence was
very close between measured and predicted values when using a calorimetric
exchange surface of much higher heat capacity. The effect becomes
increasingly significant as the ratio of immersed surface-to-particle heat
capacity falls below - 10. This is consistent with the view that, when the heat
capacity of the exchange element becomes comparable with that of the
particles, the element does not extract sufficient heat from the particles to
cool them significantly during their period of close proximity and the timeaveraged particle-to-surface temperature difference is higher, resulting in
higher effective values of the heat transfer coefficients. This effect could be
expected, for example, to increase the heat transfer rate to crushed coal as it
is charged to a fluidized bed combustor above that predicted by the simple
correlation.
Heat transfer coefficients are further enhanced when the object is free
to move (Rios and Gibert, 1984). Polchenock and Tamarin (1983) have
developed the following empirical equation for the prediction of the transfer
coefficient between fluidized bed particles and freely circulating objects:

Numax

0.41ArO.3

D:

d
(

0.2

(pP:)

-0.07
",0.66

where the subscript p refers to the bed particle and 0 to the object being fed to
the bed, and'" is the bed particle shape factor. The correlation was derived
from tests with bed particles of different materials of densities equal to or less
than that of sand and of diameters up to 6.3 mm. They predominantly had
group D characteristics but included some group B materials. The objects
were spherical or cylindrical.

400

52
N

!.
0

."E
C

300
'u
f

.
0

u
.:!.,~

........-

c::

~
'0

~
E::J

200

'x0

-- -~-

1.......

--- ------~
!

Copper probe

o Gold probe

:::E

Figure 9.7 Variation in maximum bed-to-surface heat transfer coefficient with temperature for 1,150!-Lm alumina compared with predictions
of Zabrodsky's equation, (Eq. 9.9) and the Botterill and Denloye
equations (Eqs 9.18 and 9.19).

Figure 9.7 illustrates the close correspondence between the predictions of


the simple correlations (Eqs 9.18 and 9.19) and measured values for heat
transfer between a bed of closely sized 1,148 porn alumina and surface of high

heat capacity until, at about 600C, there is a sharp change in the temperature
dependence of the heat transfer coefficient. The transition is consistent with a
change in the dominant transfer mechanism consequent upon the change in
gas physical properties with increasing temperature; Remf has fallen to -12.5
and the corresponding Archimedes number is about 26,000 (Botterill,
Teoman, and Yiiregir, 1981, 1984b). At higher temperatures, the bed is
behaving with the characteristics of a group B material and the particle
convective mode of heat transfer becomes dominant. (The increasing
difference between coefficients measured to an oxidized copper and a fine
gold surface at the higher operating temperature also illustrates the increasing
contribution of radiant heat transfer). It seems probable that the interstitial
gas flow conditions change at the transition and that there is a change from
'slow' to 'fast' (see Chapter 4) bubbling behaviour, with a consequent
increase in the effectiveness of solids mixing as bed operating temperature is
increased. It is conceivable that bed operating conditions exist under which it
is possible to have effectively slow bubbling behaviour in the lower bed
regions but fast bubbling higher within the bed with consequent implications
for bed-to-surface heat transfer.
9.3.3

Radiative Component hr

For temperatures above which radiant heat transfer is increasingly significant


(generally above 600 C; see the difference between the results for gold arid
copper surfaces, Figs 9.5 and 9.7), local particle packing and the gas
conductivity decrease somewhat in importance (despite the fact that gas
conductivity increases with temperature leading to an increase in the particle
convective component of heat transfer in particular). Baskakov et aI. (1973)
illustrate the increasing fraction of the total heat transfer coefficient
attributable to radiation from tests at temperatures up to 850C (Fig. 9.8). It
can be seen that the effect is also of greater importance with beds of larger
particles. As suggested above (Section 9.3), the three components contributing to the heat transfer process are only approximately additive. It seems
that the radiative process transfers some heat at the expense of the particle
convective mode. Thus, the heat transfer surface receives radiant energy from
the whole particle surface area visible to it and, with larger particles at high
temperatures, relatively less heat flows by conduction through the particle to
pass to the transfer surface through the shortest gas conduction paths
(Botterill, 1975).
In high temperature systems, the bed material is likely to be of a refractory
nature and these materials generally have low emissivities. However, it has
been pointed out by Zabrodsky (1973) that a particulate bed will have a
different overall emissivity to that of its constituent material. Pikashov et ai.
(1969) reported the measured effective emissivities reproduced in Table 9.2,

Figure 9.8 Fractional component of total heat transfer coefficient


attributable to radiation, bed temperature 850C and for three mean
particle sizes (BaskakoY et al. 1973). Reproduced' by permission of the
author.

Table 9.2 Measured effective emissivities (Pikashov et at., 1969), probably underestimated by 10 to 20 per cent (Makhorin, Pikashov, and Kuchin. 1978)
Material of bed

dp
mm

Gas
velocity
UlUmf

Bulk bed Emissivity Effective


temperature, of particle
bed
K
emissivity

Fused MgO and Si02


(irregularly shaped
particles)

1 - 1.15 1.2 - 3.0

770 - 1,470

River sand
(rounded particles)

1-1.5

1.2 - 3.0

770 - 1,370

0.60

0.85

Chamotte
(irregularly shaped
particles)

1-1.5

1.2 - 3.0

720 - 1,370

0.60

0.80

Zirconium dioxide Zr02


(rounded particles)

0.21- 1

1.2 - 4.0

870 - 1,320

0.23

0.59

Alumina Al203
(rounded particles)

1.5-2

1.2 - 3.0

1,070 - 1,720

0.27

0.59

0.95

which Makhorin, Pikashov, and Kuchin (1978) subsequently suggested, were


about 10 to 20 per cent. too low after further refining their measurement
technique. Trace quantities of transition metal oxides will also increase the
emissivity of refractory oxides markedly. The 'effective emissivity' of an
isothermal bed will be higher than that of the actual individual constituent
particles because of the effect of surface reflections. However, even allowing
for variation in emissivity with temperature, which they also measured, in
practical heat transfer situations there also remains the influence of an
immersed surface at a temperature different from that of the bulk of the bed
on the temperature of the bed material directly adjacent to it. (It was the
effect of this reducing the local radiant flux which was responsible for the
lower emissivity values first reported; Table 9.2). For a bed of particles of 250
to 500 ,u.m mean particle diameter operating at 1,100oe, Makhorin,
Pikashov, and Kuchin (1978) reported that the layer of particles directly
adjacent to a surface maintained at lOOoe cooled at an initial rate of between
450 and 860C/s so that their temperature very rapidly fell 50 to 100oe.
Baskakov et al. (1973) have attempted to deal with this problem on the basis
of an 'apparent emissivity' giving estimates according to the difference
between the bed and surface temperatures (Fig. 9.9). It may be anticipated

that the total bed-to-surface heat transfer coefficient h will be of the order of
600 W/m2 K for an immersed tube in a bed operating at 800 e which is of
similar magnitude to the likely steam side coefficient for a unit raising steam
at 3000e and 20 atm pressure. Under such circumstances the tube surface
temperature may approximate to 550 to 600oe, leading to an apparent
emissivity of about 0.6, which seems reasonable.
In contrast to lower temperature systems, maximum heat transfer coefficients with group B bed materials in particular are likely to be less sensitive to
the gas f1owrate. Thus, when the bed is bubbling vigorously while bubbles will
increasingly shroud the heat transfer surface, they will also open up the bed so
that the heat transfer surface can see deeper into it where there has been less
influence of the cooled immersed transfer surface and the bed is behaving
n:or.e like a black body radiator (Botterill, Teoman, and Yiiregir, 1984a).
Sumlarly, the apparent emissivity of the free surface of the bed will be higher.
For rule of thumb estimates, the radiative component can be estimated using
absolute temperatures and an adaptation of the Stefan-Boltzman equation in
the form:
0

hr =

5.673

--------

10-8
Tb

Er
-

cn, ..-_--~)

Ts

where Er is a reduced emissivity to take into account the different emissive


properties of the surface Es and bed Ea, given by:
1
(liEs

+ lIea) -

As indicated above, because of the effect of a cooling surface on the


immediately adjacent particles, the bed temperature there will be lower than
its bulk temperature; allowance can be made for this when using the bulk bed
temperature in Eq. (9.21) by using an apparent emissivity Eapp in place of Er
(Fig. 9.9) following Baskakov et al. (1973). A reasonable value for alumina
for Eapp would be 0.6 to 0.7 and somewhat lower for sand. Er may be as low
as 0.1 with a shiny metal surface. An alternative formula recommended by
Panov et al. (1978) for approximate estimates is:

9.4
Figure 9.9 Variation in apparent emissivity with transfer surface and
bed temperature for a bed of alumina (Baskakov et al., 1973).
Reproduced by permission of the author.

USE OF IMMERSED

TUBES TO INCREASE THE AVAILABLE


TRANSFER SURFACE AREA

HEAT

If the walls of the bed provide insufficient heat transfer surface, the available
transfer area may be increased by the immersion of tubes within the bed.

Indeed, particularly with highly exothermic catalystic reactions employing


group A type materials, the provision of this extra heat transfer surface often
dominates the reactor design.
Insertion of the tube, of course, affects the general bed fluidization
behaviour because of the influence of the inserts on bubble development.
This, in turn, influences the solids circulation patterns and the particle
convective component of heat transfer. However, because of the influence of
the tubes on bed behaviour, beds with such inserts will be less sensitive to the
overall effects of changi:lg scale than will unobstructed beds (see Section 9.3.1
above). Indeed, the use of vertical tubular inserts has been advocated as a
means of controlling bubble growth and hence as a means of reducing the
effect of scale on bed behaviour. The beneficial effect of horizontal tubes in
breaking up slugs has also been reported. However, although the behaviour
of beds containing inserts is less sensitive to change in scale, there is much
variation in their behaviour according to details of their configuration, and
predictions between the published correlations vary by about a factor of 2. The
final choice between horizontal or vertical tube arrangements would seem to
depend largely upon constructional convenience - how best to introduce the
tubes into the bed, support them, and allow for thermal expansion. Thus,
vertical tubes are predominantly used with the deep catalyst beds in high
aspect ratio catalytic reactors and horizontal tubes in low aspect ratio
fluidized bed combustors. Erosion can be severe if jetting bed material
impinges on a tube. There is also danger that tubes exposed within the
freeboard may be prone to erosion damage, although one means of obtaining
turndown. with a steam-raising combustion bed is to arrange to vary the
operating air flowrate in order to control the bed expansion and so vary the
number of horizontal tubes immersed in the bed (Temmink and Meulink,
1983). Results, for example, from a pilot plant facility burning < 25 mm coal
show an approximately linear decrease for the bed-to-tub coefficient from a
value of ~ 330 W/m2 K for an in-bed tube to 150 W/m2 K at a tube height of
300 mm above the expanded bed height.
When horizontal tubes are immersed within a bed, much heat transfer .
surface is exposed to the cross-flow of solids, which is advantageous from the
point of view of heat transfer. Counter to this there is also a tendency for
bubbles to shroud parts of the surface (one would expect this to be the
downward facing surface), and for particles to defluidize and form a stagnant
cap on the top of the tube. There is evidence from tests that this is the general
pattern (see Botterill, 1975; Saxena et aI, 1978), but the region at which the
maximum coefficient occurs moves to higher positions round the tube as the
fluidizing gas velocity is increased (Fig. 9.10). Newby and Keaims (1978)
report measurements where the maximum coefficient even occurs at the top
surface. When tubes are mounted in a staggered array there is a greater

Figure 9.10 Polar diagram showing variation in local heat transfer coefficient round a 35 mm
diameter tube immersed in a bed of 0.37 mm alumina operating at 500"C; excess fluidizing gas
velocities of 0.89 m/s (e) and 0.23 m/s (A), corresponding value of Umf 0.11 m/s (Botterill,.
Teoman, and Yiiregir, 1986a).

tendency for the rising bubbles to displace the stagnant particles from the top
of the tubes periodically. Although very close tube spacing seriously affects
the particle circulation (in the extreme, the bed may even rise above the tubes
so that they become the effective gas distributor if they are so closely spaced),
McLaren and Williams (1969) found no drastic reduction in heat transfer
coefficients as the tube spacing was progressively reduced in low temperature
tests. Their heat transfer results for both in-line and staggered arrays fell on a
single curve if the results were plotted in terms of the narrowest gap between
the tubes. Thus, coefficients fell by about 25 per cent. as the gap reduced
from 282 to 15 mm in tests which fluidized ash of wide size distribution and
mean particle diameters between 400 and 500 p.m but including small
fractions of greater than 1.78 mm diameter. They commented on obtaining
coefficients some 20 per cent. lower when the tube bundle was located close
to the distributor, a region strongly influenced by the functioning of the
distributor and where the bubbles generating particle circulation are still

small compared with those higher within the bed. If gas jetting at the
distributor should increase particle exchange at the transfer surface, this
would, conversely, be expected to increase the particle convective component
of the heat transfer coefficient. However, this would also be likely to be at the
expense of particle attrition and tube erosion damage if the jets actually blow
the particles on to the tubes.
The most extensive series of measurements on heat transfer to a single
immersed tube was carried out by Vreedenberg (1958, 1960), although he
worked predominantly over the rising range of the heat transfer versus the gas
flowrate-curve. He fo~nd it more difficult to correlate the results for a vertical
tube than for a horizontal one, which is to be expected. In the latter instance,
complicated though it is, the prevailing conditions of strong vertical solids
circulation past horizontal tubes is a simpler process than that obtaining at a
vertical surface when local particle residence time at the surface, under the
prevailing bed circulation conditions, can be expected to be a more complex
function of bed operating conditions. The experiments of Piepers, Wiewiorski, and Rietema (1984) with 19 vertical tubes of 2 m length arranged in a
bundle of three concentric rings immersed in a bed of 66 }km mean diameter
catalyst (Umf = 0.0018 m/s) operated at velocities up to 0.4 mls illustrate this
well. Measurements of the local coefficient along a tube mounted at the three
different radial locations were consistent with a predominant circulation of
material up in the central region of the bed, with return flow downwards at
the wall but changing in detail with a change in the operating conditions.
To characterize his material, Vreedenberg used a particle Reynolds group
which included both the particle density and diameter, and he did not,
therefore, include the gas density in it. However, he did not use this group
directly in his correlations but used, instead, the tube diameter in a modified
Reynolds group. For a vertical tube he also included the bed diameter in his
correlations, which is somewhat surprising. For a horizontal tube, Vreedenberg's correlation (1958) takes the form:
hDT)
kg

(~.)

0.3

= 0.66

(U

DT Pp (1 -

Cp}k

E) )

0.44

}kE

for U dp Pp < 2 050


}k

}k2) 0.3

U DT Pp
(
}k

d~ P~

U dp
for --}k

Pp

> 2,550

(9.24)

The form of the correlations for vertical tubular inserts was considerably
more complicated (Vreedenberg, 1960).
Andeen and Glicksman (1976) introduced the factor (1 - E) into the
correlation (Eq. 9.25) so that it should be able to follow the observed
decrease in the bed-to-tube coefficient with increasing gas flowrate beyond
the maximum. Andeen, Glicksman, and Bowman (1978) subsequently
reported that the modified form of correlation would also fit results for
horizontal flattened tubes if the constant is adjusted appropriately, but their
experimental results show considerable scatter within themselves.
The very complete review by Saxena et at. (1978) of work on heat transfer
to immersed tubes at temperatures below which radiant transfer is significant
draws attention to the wide range of results reported. In various circumstances, the different correlations lead to unrealistic predictions with the
Vreendenberg correlations best following the observed temperature dependence of the coefficient (Botterill, Teoman, and Yuregir 1984a, 1984b). Of
seven other correlations also tested, those by Grewal and Saxena (1981) and
Ternovskaya and Korenberg (1971) tended to predict the more extreme
values (Fig. 9.11), but none predicted the' right overall temperature
dependence. As reliable for horizontal tubes, and easier to use, are the simple
correlations recommended in Sections 9.3.1 and 9.3.2 above, namely: for
materials of group B, characteristics to use 70 per cent. of hpc,maxas predicted
by Eq. (9.9) to allow for the effect of the immersed tube when the hot bed is
being cooled (Fig. 9. 11a) and about 75 per cent. of the sum of the components
predicted by Eqs (9.18) and (9.19) for beds of group D characteristics,
immersion of the tube not then affecting the coefficient so much. The
boundary between the two classes of behaviour is at Remf - 12.5 and Ar
-26,000, as outlined above (Botterill, Teoman and Yiiregir, 1984a, 1984b).
This occurs at about 550C for the 1.15 mm alumina bed (Fig. 9.11b). The
detail of the results obtained by Piepers, Wiewiorski, and Rietema (1984) for
heat transfer to vertical tubes would suggest that it is not so easy to estimate
an average bed-to-vertical tube coefficient. Some tests on vertical (Bock and
Molerus, 1980) and horizontal arrays (Catipovic et ai., 1980) showed little
difference between the maximum coefficients obtainable to a tube within an
array and that to a single isolated tube.
Staub and Canada (1978) reported interesting tests of large particle bed
models (dp = 600 and 2,600 }km fluidized at high velocities and pressures up
to 10 atm in the so-called 'turbulent' fluidized bed regime. They reported
results which show a maximum heat transfer coefficient at gas velocities onequarter of the particle terminal velocity, typical values being of about 230
2
W/m
K for 550 }km particles fluidized under atmospheric pressure
conditions; they also found a dependence of the coefficient on the 0.2 to 0.3
power of the gas density depending on the particle size and tube geometry.

,,

t:l

a;

a;

"- . \.
"- "-.
"-\

"

'\.

\
\
\

\
\

c:

.~

'\

oJ>
0."0

" ,,

,,

'"

'\

"

,
,
\

::J

a.

0
Cl

" ae'..."

\
0

'\

!!I

~'t:

<Ii
<7

"
U

ai

~.

0
~

c:

"
u

"0

'..." !..~
::J

0
0

'"

<;
a.

:e

!!

1 +(

"0

to

150
dp

6)

10

0.73

(0.42Pp

(1 -

Eav)

t.zZ.4

)J

0.45

pgU

'"

"0

This compares with powers of 0.39 and 0.15 for the gas and particle
convective components, as indicated by the correlations of Eqs (9.18) and
(9.19), but one can hardly expect the correlations for ordinary bubbling bed
flow to be applicable. Staub (1979) has developed a credible heat transfer
model based on the correlations established for flowing dense solids/gas
suspensions. This uses gas and solid superficial velocities pertinent to the
upflow and downflow situations as a prediction within his flow model. The
simplified form of his general relation predicts the ratio of Nusselt numbers
(based on the tube diameter) for the conditions with the bed present Nu and
without particles Nugas in the form:

C\l

E
E

()I

0
0

0
0

0
0
<t

C\l

'"

2w/M) IUa!OlH800 Ja/sUOJI loa4 aq",-ol-paq

wnw!xDW

(>t 2w/Ml lUI~!Ueo~JI'SUDJJ JOl4 Iqnl-Ol-plq

~
WnW!XDW

for 20 p.m < dp < 1,000 p.m; for 1,000 p.m < dp < 3,000 IJ-ma value of dp =
10-3 m should be used in Eq. (9.26). t.Zm in the expression is a mixing length
which is the average distance between tube centres in this model. The average
bed hold-up (1 - Eav) is taken from bed expansion data or the correlation
data given by Staub and Canada, (1978). Wood, Kuwata, and Staub (1980)
have extended this model to heat transfer to tubes in the splash zone of a
fluidized bed of large particles.
Because of the reported range of behaviour, it is only prudent to carry out
tests on a model of as large a scale as possible. Preferably, when concerned
with a multi-unit system, this should be a 'unitcell' of the proposed system,
i.e. incorporating all the features of one full-scale complete unit of the
proposed design. This would be the area of bed fed from one fuel distribution
point, say, in the design of a combustor. However, because bed behaviour
can change markedly with operating temperature (Botterill and Teoman,
1980; Botterill, Teoman, and Yiiregir, 1981), the results of cold tests should
be interpreted with caution.
For a bed of given cross-sectional area, the depth of bed required in order
to immerse adequately the tubes providing the necessary heat transfer surface
area will be dependent on the chosen tube diameter and pitch. The available
surface area of tube within a given volume of bed increases both as the tube
diameter is increased and as the pitch is decreased. Higher fluidizing gas
pumping power cost in order to overcome the pressure drop across the bed is
the penalty to be paid for deeper beds. Thus, for example, the choice can be
important in the design of atmospheric pressure fluidized bed combustors
for steam-raising plant. At higher superficial velocities, a larger immersed

surface area may be necessary in order to be able to hold the required


combustion bed temperature, and the cost of the bed depth through increased
pumping power charges must be balanced against the increase containment
costs of a bed of greater cross-sectional area (Elliott, Healey, and Roberts,
1971). Staub and Canada (1978) point out that finned tubes, if sufficient~y
effective in increasing available heat transfer surface area, can be of value m
reducing the necessary bed height.
In applications where it is required to remove more or less heat f.r~ma bed
operating as a heat source, the generally high bed-to-surface coeffl~le~tm~y
well pose problems when the heat demand is small. As noted earlier I~ thiS
section, one way to overcome this and provide adequate turn down IS by
arranging more or less heat transfer surface to be immersed within the bed by
varying the fluidizing gas flowrate and therefore varying the degree of overall
bed expansion. This may not be operationally satisfactory for the maintenance of efficient fuel combustion, and there is again danger of increased
attrition damage to the particles and erosion damage to horizontal tubes
exposed in the free board while the bed is being operated in a less expanded
state, although the effect of the immersed tubes is to reduce the violence of
bubble eruptions at the bed surface. Alternatively, it is possible to regulate
the rate of heat removal through tubes permanently immersed within a
section of the bed in smaller units by varying the degree of fluidization
adjacent to the tubes. Thus, if there is a section of distributor with a separate
gas supply for that region, the degree of fluidity ~f ~e~ around the tu~es can
be regulated, so controlling the heat transport wIthm It. Bec~use honzo~tal
fluidbed 'conductivities' are only a few per cent. of those m the vertical
direction, it is also necessary to ensure that adequate bed material circulation
will occur from the bulk of the bed to the heat transfer section. A design to do
this is a feature of the combustor described by Virr (1976).

9.5

FINNED TUBE TRANSFER UNITS

The use of finned tube heat transfer elements has been advocated for
applications where the bed-side coefficien~.may be limiting. Generally, in
deep bed applications, bed-to-surface coeffICIentsbased on the total expo~ed
surface area are considerably reduced. Thus, Staub and Canada (1978) gIve
values showing something like a 30 per cent. reduction after allowing a
correction for the fin efficiency, but this obviously depends on the geometry
of the system (see also Saxena et ai., 1978). Not only .will p~rticle ~cce~s
within the fins depend on the relative size of particle and fm spacmg but It wIll
also depend on the relative location of neighbouring tubes. Nevertheless,
rough calculations indicate that a tube with 2.1 fins per centimetre length and

with a fin structure that will not exceed a maximum fin tip temperature of
500C in a 900C combustion bed, offers a threefold improvement in heat
transfer capability compared with a bare tube bank (Staub and Canada,
1978)..
Shallow fluidized beds have an advantage through their low pressure drop,
and the most efficient gas/solid contacting also occurs in the region of the
distributor, where bubbles are growing rapidly by coalescence (see Chapter
4). They also afford the basis for very efficient waste heat recovery units
(Virr, 1976). In this latter application, finned tubes are immersed in shallow
fluidized beds and the regions between the fins behave virtually as separate,
dilute fluidized beds. Bubbles remain restricted in size by the closeness of the
fin spacing, which may be as small as 3 to 5 mm or 15 to 20 times the particle
diameter. In operation, bed expansion as high as 400 per cent. is usual while
still obtaining high bed-to-surface coefficients, as good as those to a bare
tube, because of the very short particle residence times obtaining at the heat
transfer surface; particle velocities are possibly an order of magnitude higher
than in a deep bed and this more than compensates for the greatly reduced
bed density. Based on the outside area of the tube, obtainable coefficients are
in the range of 1 to 4 kW/m2 K. Because the particles in the bed present a very
high surface area to the gas, very good heat recovery can be achieved from a
hot fluidizing gas. Typically a unit can be operated with an overall pressure
drop as low as 50 mm water gauge from the suction developed by an exhaust
fan which is used to suck the hot flue gases through the bed. For space
heating, 150 kW of hot water has been recovered using a small unit to cool
flue gases from 600 to 180C. Another unit is producing 3.5 tonne/h of steam
from the waste gases from a 1,200 h.p. diesel engine in a 32,000 tonne dead
weight oil tanker.

Reference was made in the introduction to the present limited applications of


fluidized solids as a basic heat transfer medium. these have mainly been in
metallurgical heat treatment processes (Baskakov, 1968; and Virr, 1976). In
one application requiring a reducing atmosphere, for example, a rich fuel/air
mixture was introduced through the distributor. Higher within the bed, above
the zone where the specimens for treament were placed, additional air was
injected to complete the combustion of the fuel and provide the required heat
release to maintain the overall bed temperature. Natural bed circulation
caused the hot solids to flow downwards through the bed from the principal
combustion zone to the heat treatment zone, replacing those carried upwards
towards the upper combustion region by rising bubbles.
Work described by Newby and Keairns (1978) formed part of a design
study for the development of a fluidized bed stream-raising heat exchanger in

a liquid metal cooled nuclear reactor system. In this scheme, heat would be
transferred between the lower horizontal tube array through which the liquid
metal circulated to an upper bank of steam-raising tubes by the fluidized bed.
Thus they thought to use the fluidized solids as an inert heat transfer medium,
so obviating the danger of direct sodium/steam contact in the event of tube
failure.
There is much potential for the use of beds of particles in heat storage
associated with generation and recovery systems. There is a possible
advantage in separating the heat generation from steam-raising sections of the
boiler plant by interposing a large store between the combustion bed and the
steam generating plant (Botterill and Elliott, 1964). For an electricity
generating power station, the combustion equipment could then be rated at
the average load over 24 hours, whereas only the heat exchangers and
electricity generating equipment would have to be rated for the peak
capacity. A cube of side 40 m would provide adequate store for a 500 MW
unit, and this is no bigger but much cheaper than a larger boiler. In the
defluidized state heat losses from the store would be low. The system offers
improved reliability against combustor failure through the period of grace
afforded by the energy in store. By controlling the rate at which material is
fluidized and withdrawn from the store, the turndown problem occurring
when tubes are immersed directly within the combustion bed (referred to in
Section 9.4 above) is overcome; such a system posseses very quick startup
capabilities, limited only by the rate at which the turbine can accept load.
Many high temperature processes produce dirty exhaust gases at varying
rates. In the steel industry, for example, the BOS unit operates intermittently, producing exhaust gases at 1,600C, and the order of energy to be
disposed of is typically about 8 GJ/t of steel production. Currently, this is
wasted and is expensive to dispose of. It could be transferred to a store using a
falling cl<JUdheat exchanger and recovered from there using fluidbed
techniques for transport and the heat transfer operation. A solids flow rate of
3 kgls is readily capable of transporting energy at a rate of 2 MW. Basic
studies on the feasibility of this and solar energy storage utilizing fluidized bed
tehcniques are outlined by Bergougnou et at. (1981).

A more detailed review of work on fluidized bed heat transfer up to 1984has


been undertaken by Botterill (1986). Gas-to-particle heat transfer is not
usually a limiting factor in fluidized bed operations because of the large
surface area presented by particulate solids. The correlation (Eq. 9.1) may be
used to estimate the heat transfer coefficient, although precise measurements, as outlined in Section 9.2 above, are difficult to make. This remains a
problem area.

The three basic mechanisms involved in bubbling bed-to-surface heat


transfer have been identified as outlined in Section 9.3 above. To a first
approximation they are additive. It is the capacity of the circulating particles
to transport heat which is the significant feature of the particle convective
component hpc (Section 9.3.1). In this mechanism, gas conductivity is the
limiting factor and the level of operating temperature is therefore important
because of its influence on it. Particle packing and residence time close to the
transfer surface are usually the controlling factors; the former is dependent on
particle size distribution and bed fluidizing conditions, and the latter is also
much influenced by the general design of the bed. For particles of Geldart's
group B between about 100 and 800 JLm, Zabrodsky's dimensional
correlation (Eq. 9.9) is to be recommended. For group A materials where
there can be significant bed expansion there is the correlation (Eq. 9.10), but
this has only been tested at temperatures close to ambient.
The interphase gas convective component hgc (Section 9.3.2) becomes
significant when gas flow conditions through the continuous phase cease to be
laminar. Gas heat capacity and hence static operating pressure are important.
A correlation for its estimation is given in Eq. (9.18) and, for the
corresponding maximum particle convective component, in Eq. (9.19).
Transition from group D to group B type behaviour occurs as Remffalls below
-12.15 with Ar -26,000 as operating temperatures increase (Fig. 9.7).
Progress has been made in modelling radiant heat transfer (Lindsay 1983),
Rule of thumb estimates can be made using an apparent emissivity to allow
for the influence of the immersed heat transfer element on the bed directly
adjacent to it (Section 9.9.3). Thus, although the bed emissivity and that of its
surface will be higher than for the individual particles forming the bed (e.g.
see Table 9.2), an apparent emissivity for the bed adjacent to the transfer
surface between 0.6 and 0.7 is more reasonable in many instances.
The range of coefficients to be expected if radiative transfer can be
neglected is indicated in Fig. 9.3. Knowing the necessary conditions to obtain
good bed-to-surface heat trasnfer, the challenge remains to engineer the
system so that it can be obtained. As indicated in Section 9.4, the available
correlations for the prediction of heat transfer to arrays of tubes predict a
wide range of values. For ease of application and general reliability of
estimate, the simpler correlations as outlined in Section 9.3 are to be
preferred. Thus a conservative estimate for the bed-to-horizontal-tube
coefficient for beds of group B type materials would be about 0.7 x hpc.max,as
predicted by the Zabrodsky equation (Eq. 9.9), and for the group D type
materials, about 75 per cent. of the sum of the predictions for the correlations
of Eqs (9.18) and (9.19). Different correlations will apply for the condition of
turbulent fluidization (Eq. 9.26). It should be appreciated that most of the
correlations for heat transfer to tubes have been developed from lower
temperature experiments only. 'Unitcell' tests need to be carried out and

255

FLUID BED HEAT TRANSFER

particularly so because the effects of lateral mixing may become very


important when there is a large number of tubes immersed within the bed.
Operations with 'large' local heat sources should be tested if these are likely
to occur in practice. Cold tests need to be interpreted with care because of the
pronounced changes in bed behaviour that can occur with a change in
temperature of operation (Botterill and Teoman, 1980; Botterill, Teoman,
and Yuregir, 1981, 1984a, 1984b). The conditions of fluidization can be used
to regulate heat removal from a bed, but this must also be determined
according to the particular operational characteristics of the bed. As indicated
in Sections 9.5 and 9.6, the potential of fluidized solids in basic heat transfer
operations is only just beginning to be exploited.

Do
DT
to
g
h
hgc
hgp
hpc
hr
H
Hmf
ke

kg
kmf
kp

kw
Ra
Rw

constant in Eqs (9.13) and (9.14)


heat capacity of gas
heat capacity of particle
, diameter of particle
diameter of immersed or freely circulating object
diameter of immersed tube
fraction of time surface is shrouded by bubbles
acceleration due to gravity
heat transfer coefficient
interphase gas convective component of
bed to surface heat trasnfer coefficient
gas-to-particle heat trasnfer coefficient
particle convective component of bed-to-surface
heat trasnfer coefficient
radiative component of bed-to-surface heat
transfer coefficient
bed height
bed height at minimum fluidization
effective thermal conductivity of bed
thermal conductivity of gas
thermal conductivity of bed at minimum
fluidization
thermal conductivity of particle
effective thermal conductivity in near wall region
resistance to heat transfer within packet phase
resistance to heat transfer at wall

Tg
Tp
Ts
LlZm
IE
lEapp

9.8 NOMENCLATURE
A
Cg
Cp
dp

S
Sa
V
Vmf
t
Ta

lEa
Jlkg K
Jlkg K
m
m
m

lEeff
IEp
IEr
s
j.L

Pg
m/s2
W/m2 K
W/m2 K
W/m2 K
W/m2 K
W/m2 K
m
m
W/mK
W/mK

Pmf
Pp
u
'T

t/J

Ar
Nugc
Nugp
Pr
Rep
Remf

stirring factor (Eq. 9.6)


surface area of solids/unit volume of bed
gas velocity
gas velocity at minimum fluidization
cooling time constant
temperature of bed
temperature of gas
temperature of particle
temperature of surface
mixing length (Eq. 9.26)
bed voidage
apparent emissivity
emissivity of bed
effective emissivity
particle emissivity
reduced emissivity (Eq. 9.22)
emissivity of surface
gas viscosity
gas density
density of bed at minimum fluidization
particle density
Stefan-Boltzman constant
packet residence time
particle shape factor
Archimedes number, gd~pg(pp _ Pg)1j.L2
interphase gas convective Nusselt number, hgcdplkg
gas-to-particle Nusselt number, hgp dplkg

m/s
m/s
s

K
K
K
K
m

N s/m2
kg/m3
kg/m2
kg/m3
W/m2 K4
s

gas Prandtl number, Cgj.L/ kg


particle Reynolds number, dpPgVI j.L
Reynolds number for minimum fluidization
condition, dppgVmfl j.L

9.9 REFERENCES
W/mK
W/mK
W/mK
m2KIW
m2KIW

Adams, R.L, and Welty, I.R. (1979). A.I.Ch.E. J., 25,395.


Andeen, B.R., and Glicksman, L.R. (1976). ASME-AIChE Heat Transfer Conference, Paper 76-HT-67.
Andeen, B.R., Glicksman, L.R., and Bowman, R. (1978). In Fluidization - Proc of
the Second Engineering Foundation Conference, (Eds I.F. Davidson and D.L.
Keairns) Cambridge University Press, p. 345.

Barile, R.G., Seth, H.K., and Williams, K.A (1970). Chem. Eng. J., 1,236.
Baskakov, A.P. (1968). High Speed Non-oxidising Heating and Heat Treatment of
Metals in a Fluidized Bed, Metallurgy, USSR.
Baskakov, A.P., Berg, B.V., Virr, O.K., Phillippovsky, N.F., Kirakosyan, V.A.,
Goldobin, J.M., and Suprun, V.M. Powder Technol., 8, 273.
Baskakov, AP., and Suprun, V.M. (1972). Int. Chem. Eng., 12, 119.
Bergougnou, M.A., Botterill, J.S.M., Howard, R.J., Newey, D.C., and Teoman, Y.
(1981). Third International Conference on Future Energy Concepts, London, p. 61.
Bock, RI., and Molerus, O. (1980). Proceedings of the 1980 Fluidization Conference
(Eds J. R. Grace and J.M. Matsen), Plenum Publishing Corporation, New York, p.
217.
Botterill, I.S.M. (1975). Fluid Bed heat Transfer, Academic Press, London.
Botterill, J.S.M. (1986). "Advances in Fluidized Bed Heat Transfer" in Transport
Phenomena in Fluidising Systems. (Eds. L.K. Doraiswamy and B.D. Kulkarni)
Wiley EasternlWiley Halsted, in press.
Botterill, I.S.M., Brundrett, G.W., Cain, G .L., and Elliott, D.E. (1966). Chem. Eng.
Prog. Symp. Ser, 62, (62), 1.
Botterill J.S.M., and Denloye, A.O.O. (1978). A.I.Ch.E. Symp. Ser., 74 (176),194).
Botterill, I.S.M., and Elliott, D.E. 1964). Engineering, 198, 146 (July 31).
Botterill, J.S.M., and Teoman, Y. (1980). Proceedings of the 1980 Fluidization
Conference (Eds J.R. Grace and J.M. Matsen), Plenum Publishing Corporation,
New York, p. 93.
Botterill, I.S.M., Teoman, Y., and Yiiregir, K.R. (1981). A.l.Ch.E. Symp. Ser., 77
(208), 330.
Botterill, I.S.M., Teoman, Y.,and Yiiregir, K.R., (1984a). Powder Technol., 39,177.
Botterill, J.S.M., Teoman, Y., amd Yuregir, K.R., (1984b). XVI ICHMT Symposium, Dubrovnik, 5-7 Sept.
Catipovic, N.M., Jovanovic, G.N., Fitzgerald, T.L, and Levenspiel, O. (1980).
Proceedings of the 1980 Fluidization Conference (Eds J.R. Grace and J. Matsen)
Plenum Publishing Corporation, New York, p. 225.
Davidson, J.F., and Harrison, D. (1971). Fluidization, Adademic Press, London.
Decker, N., and Glicksman, L.R. (1981). AI.Ch.E. Symp. Ser, 77 (208), 341.
de Groot, J.H. (1967). In Proceedings of the International Symposium on Fluidization,
Ed. A.AH. Drinkenburg Netherlands University Press, Amsterdam, p. 348.
Denloye, A.O.O., and Botterill, J.S.M. (1977). Chem. Eng. Sci., 32,461.
Donsi, G., Lancia, A., Massimilla, L.,and Volpicelli, G. (1984). In Fluidization (Eds
D. Kunii and R. Toei), Engineering Foundation Conference, p. 347.
Elliott, D.E., Healey, E.M., and Roberts, AG. (1971). Conference of the Institute of
Fuel and L'Institut Francais des Combustibles et de L'Energie, Paris.
Geldart, D. (1972). Powder Technol., 6, '85.
Gel'Perin, N.I., and Einstein, V.G. (1971~'In Fluidization (Eds J.F. Davidson and D.
Harrison), Academic Press, London, p. 471.
Gel'Perin, N.I., Einstein, YG., and Zakovski, A.V. (1966). Khim Prom., 6, 418.
Golan, L.P., Cherrington, D.C., Diener, R., Scarborough, C.E., and Wiener, S.C.
(1979). Chem. Eng. Prog., 75, 62 (July).
Grewal, N.S., and Saxena, S.c. (1981). Ind. Eng. Chem. Proc. Des. & Dev., ZO, 108.
Heyde, M., and Klocke, H-J. (1979). Chem. Ing. Technik, 51,318.
Khan, A.R., Richardson, J.F., and Shakiri, K.J., (1978). In FluidizationProceedings of the Second Engineering Foundation Conference (Eds I.F. Davidson
and D.L. Keairns), Cambric!ge University Press, p. 345.
Kunii, D., and Levenspiel, O. (1969). Fluidization Engineering, Wiley.

Kunii, D., and Smith, J.M. (1960). A.J.Ch.E. J., 6, 71.


Lindsay, J. (1983). Radiative Heat Trasnfer in Fluidized Beds, Ph.D. Thesis,
University of Cambridge.
Littman, E., and Sliva, D.E. (1970). Heat Transfer 1979, Paper CE.1.4, Fourth
InternatIOnal Heat Transfer Conference, Versailles, Elsevier Publishing Corp.
McGaw, D.R. (1976). Powder Technol., 13,231.
McLaren, J., and Williams, D.F. (1969). J. Inst. Fuel., 42,303.
Makhorin, K.E., Pikashov, V.S., and Kuchin, G.P. (1978). In FluidizationProceedmgs of the Second Engineering Conference (Eds J .F. Davidson and D L
K~airns), Cambridge University Press, p. 39.
. .
Martm, R (1980). Chem. Eng. Technik., 52, 199.
M~son, E.A., and Saxena, S.c. (1958). The Physics of Fluids, 1,361.
MIckley, H.S., and Fairbanks, D .F. (1955). A.J. Ch. E. J.. 1, 374.
NeWby,.R.A., and Keairns, D.L. (1978). In Fluidization - Proceedings of the Second
Engl1leermg FoundatIOn Conference (Eds J.F. Davidson and D.L. Keairns)
Cambridge University Press, p. 320.
'
Panov, O.M., Baskakov, A.P., Goldobin, Yu.M., Phillippovsky, N.F., and Mazur
Yu.S., (1978). Inzh-Fiz. Zh., 36,409.
'
Piepers, RW.: Wiewi.orski.'P., and Rietma, K. (1984). In Fluidization (Eds D. Kunii
. and R. Toet), Engmeenng Foundation Conferences, p. 339.
Plkashov, V.S., Zabrodsky, S.S., Makhorin, K.E., and Il'Chenko, AI. (1969). Bull.
B.S.S.R., Acad. Sci., Physical Energetics Ser., 100, No.2.
Polchenok, G.I., and Tamarin, A.I., (1983). 1nzh-Fiz.Zh., 45,427.
~chardson, J.F., an~ Ayers, P. (1959). Trans. Instn. Chem. Engrs, 37,314.
Rios, G.M:, and GIbert, H. (1984). Fluidization (Eds D. Kunii and R. Toei)
FoundatIOn Conferences, p. 363.
Saxena, S.C., Grewa~, N.S., Gabor, J.D., Zabrodsky, S.S., and Galershtein, D.M.
(1978)1dvances
m Heat Transfer (Eds T.F. Irvine and J.P. Hartnett), Vol. 14,
AcademIC Press, New York, p. 147.
Singh, AN., and Ferron, J.R. (1978). Chem. Eng. J., 15,169.
Staub, F.W. (1979). J. Heat Transfer, 101, 391.
Staub, ~.W ..' and Canad~, G.S. (1978). In Fluidization - Proceedings of the Second
Engme~rlng F?und~tlOn Conference (Eds J.F. Davidson and D.L. Keairns)
Cambndge Umverslty Press, p. 339.
'
. Temmink, H.M.G., and Meulink, J. (1983). 'Operating experience with TNO 2m x
1m atmospheric fluid bed boiler facility', 3rd European Coal Utilization Conference, 11-13 Oct., Amsterdam.
Ternovskaya, A.N., and Korenberg, Yu.G. (1971). Pyrite Klining in a Fluidized Bed
Izd Khiminya, Moscow.
'
Todes, O.M. (1965). Applications of Fluidized Beds in Chemical Industry, Part II, Izd.
Zanie, Leningrad, pp. 4-27.
..
.
v~n Heerden, C. (1952). J. Appl. Chem., 2, Supplement Issue No.1, S7.
Vm, M.J. (1976~.In Procee~ings of the Fourth International Conference on Fluidized
Bed CombustIOn, The MItre Corporation, p. 631.
Vreedenberg, H.A (1958). Chem. Eng. Sci., 9, 52.
Vreedenberg, H.A. (1960). Chem. Eng. Sci., 11,274.
Wood,. ~.~., Kuwata, M., and Staub, F.W. (1980). Proceedings of the 1980
FLUidization Conference (Eds J.R. Grace and J. Matsen), Plenum Publishing
Corporation, New York, p. 235.
Yerushalmi, J., and Cankurt, N.T. (1979). Powder Technol., 24, 187.
Yoshida, K., Kunii, D., and Levenspiel, O. (1969). Int. J. Heat and Mass Trans+er 12
529.
J',
,

Zabrodsky, S.S. (1963). Int. J. Heat and Mass Transfer, 6,23.


Zabrodsky, S.S. (1966). Hydrodynamics and Heat Transfer in Fluidized Beds, The
MIT Press.
Zabrodsky, S.S. (1973). Int. J. Heat and Mass Transfer, 15, 241.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 by John Wiley & Sons Ltd.

Fluid Bed Drying

Fluidizing with hot air is an attractive means for drying many moist powders
and granular products. The technique has been used industrially since 1948,
and today it enjoys widespread popularity for drying crushed minerals, sand,
polymers, fertilizers, pharmaceuticals, crystalline materials, and many other
industrial products.
The main reasons for this popularity are as follows:
(a) Efficient gas solids contacting leads to compact units and relatively low
capital cost combined with relatively high thermal efficiency.
(b) The handling of the particles is quite gentle compared to some other
types of dryer. This is important with fragile crystals.
(c) The lack of moving parts, other than feeding and discharge mechanisms,
keeps reliability high and maintenance costs low.
The main limit on the applicability of fluid bed dryers is that the material
being dried must be fluidizable. Some potential feedstocks are too wet to
fluidize satisfactorily. This is usually due to an excessive amount of surface
moisture on the particles, causing them to agglomerate. This problem may be
overcome by flashing off the surface moisture in a pneumatic conveying dryer
preceding the fluid bed dryer. Another limitation is encountered if the
product has a very wide size distribution, so that at an air velocity high
enough to fluidize the large particles there is an unacceptable loss of small
particles from the bed. To some extent this limitation can be overcome
nowadays by the use of a vibrating fluid bed (see below).

The potential user of fluid bed drying has a wide variety of equipment to
choose from. The main categories are described briefly below. Within each
category there are many variations offered by equipment vendors.
10.2.1

Batch dryers

The first distinction to be made is between batch and continuous fluid dryers.
Batch dryers are normally used when the production scale is small and several
different products have to be made on the same production line. Figure 10.1

Dried
product

shows a typical batch fluid bed dryer. The wet feed is loaded into the cabinet
and clamped to the filter sock module. The cabinet doors are then closed and
the blower started. An adjustable damper controls the degree of air
recirculation. The circulating air may be heated by a steam tube battery or by
gas firing. Batch fluid bed dryers are particularly popular in the pharmaceutical and dyestuffs industries.
10.2.2

Continuous 'well-mixed' dryers

The first continuous fluid bed dryer was the 'well-mixed' type, illustrated in
Fig. 10.2, which was introduced in the United States in 1948. It is usually of
circular cross-section, and takes its name from the fact that the particle
residence time distribution approaches the perfect mixing law:
E(t)

~exp

(-

t~ )

-Fuel
Air

In this expression, E(t)dt is the fraction of particles with residence times


between t + dt, and tR is the mean residence time. Because of the near-perfect
mixing the bed has a nearly uniform composition and temperature equal to
the composition and temperature of the outlet product stream. Hence, the
moist feed falls into a bed of almost dry particles. For this reason this type of
continuous fluid bed dryer can handle wetter feedstocks than can other types
to be described later.
The main disadvantage of the 'well-mixed' type of fluid bed dryer is
the wide particle residence time distribution, leading to a wide range of
moisture content in the product particles. The average moisture content
?f the product may be acceptable, but 40 per cent. of the particles stay
In the dryer for less than half the average residence time and 10 per cent.
for less than one-tenth of the average (see Chapter 5, Section 5.6). Hence,
some particles will emerge quite wet. For some products, particularly
many polymers, this is unacceptable. Furthermore, the wide distribution
makes it difficult to achieve a very low average product moisture
content.
Despite this drawback, the 'well-mixed' bed is still the most popular type of
continuous fluid bed dryer in North America. In Western Europe it has been
superseded in many applications by the 'plug flow' bed or by the vibrated
fluid bed.

10.2.3

Continuous 'plug flow' dryers

These are beds of shallow depth (typically 0.1 m) in which the particles flow
along a channel whose length/width ratio is much greater than unity (typically
in the range 4:1 to 30:1). The objective is a more or less close approach to
plug flow of the particles, so that they all emerge with approximately the same
moisture content. For length/width ratios up to about 10 the particle flow
channel may be straight, but for higher ratios a reversing path or spiral path is
more practicable as shown in Fig. 10.3. Straight channel designs are often
provided with baffles normal to the direction of particle flow in an attempt to
improve the approach to plug flow.

-air

Exhaust
to
cyclone
Adjustable
weir

In addition to reducing the spread of product particle moisture contents, a


plug flow bed will normally require a smaller bed volume than a well-mixed
bed to achieve the same average product moisture content. However, the
simple plug flow bed has some disadvantages. Firstly, the moist feed falls into
an area where the particles are still comparatively wet. Consequently, there
may be fluidization difficulties at the feed end with some feeds which could be
handled quite satisfactorily in a well-mixed bed. Secondly, the hot air passing
through the bed towards the discharge end does comparatively little drying
and therefore does not give up much of its heat. Hence, the thermal efficiency
of a simple plug flow bed is lower than that of a well-mixed bed. Thirdly, the
temperature of the particles rises as they flow along the bed and towards the
discharge end it approaches the inlet air temperature. With a heat-sensitive
product this limits the inlet air temperature which can be used, thereby
reducing the thermal efficiency still further. Variations on the simple plug
flow bed have been developed to overcome these difficulties. The most
important of these variations are the vibrated fluid bed and the multi-stage
bed.
10.2.4

I Dried
product

Plan view

Feed--BL....

This is simply a straight channel plug flow bed with a vibrating distributor, as
shown in Fig. 10.4. Alternatively, it may be viewed as a vibrating fluid bed

~G_

Product

(bl

Plan view

Vibrated fluid bed dryers

Flexible
Z:=couplings

Feedl'

_______

~I

Product

F = feed
P= product

Figure 10.3 Continuous 'plug flow' fluid bed dryer: (a) straight
path; (b) reversing path; (c) spiral path.

conveyor which uses hot air as the fluidizing medium. Compared to the
simple plug flow bed, it has the advantage that any agglomerates arising at the
feed will be kept moving by the vibrations of the distributor, hopefully until
they have dried sufficiently to break up.
Furthermore,
feeds with a wide size distribution can be processed
successfully in this type of bed. The air velocity can be set low enough to
avoid excessive elutration of the smaller particles while the largest particles
are kept moving by the vibration of the distributor.

Finally, these beds are often used to dry feeds consisting entirely of large
particles with a minimum fluidization velocity of the order of 1 m/s. With a
static distributor an air velocity considerably in excess of the minimum would
be needed to ensure adequate fluidization and conveying, but this would be
far more air than is required to satisfy mass and heat balance considerations.
If the distributor is vibrated the air velocity can safely be kept in the vicinity of
the minimum fluidization velocity, with consequent savings in capital and
operating costs.
Present technology imposes two limitations on vibrated fluid bed dryers.
Firstly, the materials of construction available for vibration mountings limit
air inlet temperature to about 400C. Secondly, if the bed is to vibrate at its
natural frequency, which is the simplest and most economical way of
operating, bed lengths are limited to about 8 m. Therefore, vibrated fluid bed
dryers are limited to intermediate temperatures and intermediate throughputs. Finally, they have a higher maintenance requirement than static fluid
beds.
10.2.5
Multi-stage

Multi-stage beds

beds are often used for the following purposes:

(a) To accomplish drying and cooling in the same vessel, which is usually a
straight channel plug flow unit with a divided chamber below the
distributor (see Fig. lO.Sa.)
Feed

1:::~~t~~?

Exhaust
-air

~tt'---==-.~-===;...

=...------1~ld~

~
Product

IWell-mixed
lsectian

I Plug flow
: section

Figure 10.5 Multi-stage fluid bed dryers: (a) dryer plus cooler;
(b) plug flow fluid bed following well-mixed fluid bed (plan view).

(b) To combine the ability of a 'well-mixed' bed to handle wet feeds with the
ability of a 'plug flow' bed to achieve a comparatively uniform product
moisture content, through having a 'well-mixed' section followed by a
'plug flow' section (see Fig. 1O.5b)
Purposes (a) and (b) can be combined in a three-stage unit.
In case (b) the temperature of the air supplied to the 'well-mixed' stage can
be higher than that supplied to the 'plug flow' stage because the higher
evaporation rate in the first stage will help to keep the bed temperature down.
A higher air velocity may also be used in the first stage to help in rapid
dispersion of the wet feed.
10.2.6

Fluid bed dryers with internal heating

When a finely divided, heat-sensitive powder is being dried in a fluidized bed


there are limitations on both the velocity and the temperature of the inlet air.
Consequently, the rate of heat input from the air per unit distributor area may
be quite low, and if the air is the only heat source the distributor area required
to perform the drying duty may be large. It can be reduced by supplying part
of the heat through steam tubes or heated baffles immersed in the bed. For
example, if half the heat is supplied in this way the distributor area needed is
halved. This can give substantial savings in capital and operating costs.
Heat transfer coefficients from immersed heating surfaces to the bed
particles increase with decreasing particle size (see Chapter 9). The converse
is true for gas-to-particle heat transfer, when account is taken of the enforced
reduction in air velocity with decreasing particle size. Hence, the use of
internal heating surfaces becomes more attractive the smaller the particle
size.
When internal heating surfaces are used, the bed depth is usually
determined by the need to keep the heat transfer surfaces immersed, rather
than by consideration of drying kinetics.
10.2.7

Fluid bed granulation

Fluidized beds can also be used for making dry powder from a feed which is a
slurry or solution. The feed is sprayed on to the bed, usually with a pneumatic
atomizer to give a very fine atomization. The particles in the bed are
continually growing, by one or both of two mechanisms. A drop may strike a
particle and form a thin layer of liquid on the particle surface, whereupon the
layer immediately dries and the particle grows by one layer. This mechanism
gives a hard, dense particle structured like an onion. Alternatively, the
particle may strike another particle before the layer has dried, in which case
the liquid may act as a binder and hold the particles together. This gives a

product conslstmg of agglomerates of finer particles. Which mechanism


predominates depends very much on the materials and operating conditions,
and must be ascertained by experiment.
In batch granulation the bed initially contains fine seed particles. The
required amount of feed is then added at an appropriate rate. After feed
addition is complete, there may be a further period during which the grown
particles are thoroughly dried, followed finally by a cooling period when the
fluidizing medium is changed from hot to cold air. Batch fluid bed granulation
is very popular in the pharmaceutical industry.
In continuous fluid bed granulation, a stream of particles is continually
withdrawn from the bed and classified into fines, size, and oversize. The fines
and crushed oversize are returned to the bed for further growth, while
product of the required size is taken off. Figure 10.6 shows the principal
options.

Top
spraying
Wet
~R
feed~
Submerged
feed

'-Surface
---""""solids
OR
-Bottom

Batch drying curves

If a batch of moist powder is dried in a fluidized bed with periodic withdrawal


of samples for determination of moisture content, the resulting curve of
powder moisture content X (defined as the weight of H20 divided by the
weight of dry solid) versus time t will probably look something like the curve
of Fig. 1O.7(a). This is conventionalIy divided into two portions, the first

w ----------(df'-

Separated solids
ar dust
Direct~OR
to fluid bed

10.3.1

..,.........

Solids treatment
(size reduction and/
or classification

Product

removal

-~I

x-x.
Figure 10.7 Batch drying curves.

The main potential problem with fluid bed granulation is an operational


one. If the ratio of liquid to solid in the bed becomes too high the particles will
rapidly form large agglomerates the size of golf balls and fluidization will be
lost. The plant must then be shut down and the bed dug out. The operator
gets only a few minutes warning of the onset of this 'wet quenching' process.

called the constant rate period and the second called the falling rate period.
The moisture content at the transition between the two periods is called the
critical moisture content Xcr' If drying is continued for long enough, X will
approach the equilibrium moisture content Xe. For a given material this is a
function of relative humidity and temperature. At any point on the curve the
amount of removable moisture remaining, X - Xc, is called the free moisture
content. It should be noted that in reality Xcr rarely appears as a sharply

defined point on the experimental drying rate curve. There is usually some
curvature at the transition from constant rate to falling rate. Xcr is probably
best defined as the point at which the forward extrapolation of the constant
rate line intersects the backwards extrapolation of the falling rate curve.
The rate of drying - dXldt can be determined at any point by
differentiating the X versus t curve. A graph of - dXldt versus the free
moisture content X - Xe is an alternative way of representing the drying
characteristics of a material (see Fig. 1O.7b).
As a rough approximation, the constant rate period may be regarded as
corresponding to the removal of surface moisture from the particles, while the
falling rate period corresponds to the removal of internal moisture. Since
most materials will not fluidize satisfactorily if there is substantial surface
moisture, the constant rate period in fluidized bed drying may be so short as
to be unobservable, except under very mild drying conditions. On the other
hand, a non-porous material such as sand may show virtually no falling rate
period.
10.3.2

Constant rate period

In the constant rate period the surface of the particle is wet enough for the
layer of air adjacent to the surface to be saturated. Hence, the drying rate is
determined by the rate at which vaporized moisture can be transported across
the boundary layer surrounding a particle. During this period the temperature of the particle surface remains constant at the wet bulb temperature
Twb of the air. If Pwb is the partial pressure of vapour at the wet bulb
temperature, P is the partial pressure in the bulk air stream, and Kp is the
mass transfer coefficient based on partial pressure, the rate of moisture
removal Nc per unit particle surface area (called the drying flux) in the
constant rate period is given by:
Nc = Kp (Pwb - p)

(10.2)

Since the particle temperature does not rise during this period, all the heat
transferred across the boundary layer from gas to particle must be used for
evaporation. Hence, an alternative formulation for Eq. (10.2) is:
hgp
NC=T(T-

Twb)

where hgp is the gas-to-particle heat transfer coefficient, A is the latent heat
of vaporization and T is the bulk gas temperature.
For the air-water system the wet bulb temperature of the air is almost equal
to the adiabatic saturation temperature, which is readily obtained from
psychrometric charts. For organic solvents in air or an inert gas, TWb can be
calculated if the psychrometric ratio of the solvent-gas system is known (see,

for example, Keey, 1978). Hence, if either Kp or hgp is known, Nc can be


calculated for any drying conditions.
There is a plethora of data on gas-to-particle heat transfer coefficients in
fluidized beds of small particles of diameter less about 1 mm. This subject is
covered in Chapter 9. One of the most frequently used correlations is that
proposed by Kothari and discussed in Kunii and Levenspiel (1969):
Nu

0.03Re~3

(10.4)

This suggests that the heat transfer coefficient, and hence the drying rate in
the constant rate period, should be proportional to the gas velocity U raised
to the power 1.3. This was confirmed by Mostafa (1977) in experiments on the
drying of small particles of silica gel, molecular sieve, and vermiculite in a
fluidized bed.
Many fluid bed dryers, particularly those with vibrating distributors,
operate on particles of diameter greater than 1 mm. Many less data are
available for this size range, and what there are suggest a much lower
dependence of drying rate on gas velocity. No generally applicable correlation is yet available. Zabeschek (1977) found that with a fluidized bed of
aluminium silicate particles the particle-to-gas mass transfer coefficient was
roughly proportional to If'.5 when the particle diameter was 2.76 mm, and
was almost independent of U when the particle diameter was 4.30 mm.
Subramanian, Martin, and Schliinder (1977) also concluded, from an analysis
of other investigators' results, that transfer rates are effectively independent
of gas velocity when the particle diameter is much greater than 1 mm.
Equations (10.2), (10.2), and (10.4) have a number of other practical uses.
They enable one to predict the effects of bulk gas temperature, bulk gas
humidity, and small particle diameter on the drying rate in the constant rate
period.
10.3.3

Falling rate period

In the falling rate period the rate of moisture migration to the surface of a
particle is insufficient to keep the layer of air adjacent to the particle surface
saturated. Hence, the drying rate is no longer determined solely by conditions
in the boundary layer. It also depends on the pore structure of the material
and on the mechanism of moisture migration. There may, in fact, be several
simultaneous mechanisms. These include capillary action, vapour diffusion,
diffusion along internal surfaces, and, in the case of cellular materials,
diffusion across cell walls. The balance between these mechanisms may
change as drying proceeds. For example, capillary motion may predominate
during the early part of the falling rate period when the pores are relatively
full, while vapour diffusion may dominate towards the end when only small
pockets of moisture remain in the solid structure. In general, the falling rate

curve cannot be predicted a priori and must be determined by experiment.


However, having determined the falling rate curve experimentally at one
set of operating conditions, it is often possible to predict approximately how it
will change when the operating conditions are changed. This is useful in that it
enables a range of process options to be explored with only a small amount of
experimental data. To do this we make use of the concept of the characteristic
drying curve, first proposed by van Meel (1958), elaborated by Keey (1978),
and shown by experiment to be valid for several widely different materials
(see, for example, Zabeschek, 1977, and Gummel and Schliinder, 1977).
Essentially, the characteristic drying curve concept is based on the
assumption that at any point on a batch drying curve the drying flux N is given

t Increasing
I

No

I
I
I

,'--Locus
I

of {X-x"Jc,

I
/

I
I

by:
N

Ncf

X - Xe)
( X _ X
cr

Nc V

Nc is the drying flux which would be observed at the prevailing external


conditions if the material was wet enough to be in the constant rate period of
drying. Above the critical moisture content V takes the value unity; below
the critical moisture content it is assumed to be a function only of the free
moisture content remaining in the material, expressed in dimensionless form
by using the critical free moisture content as a reference quantity. Nc can be
estimated theoretically by the methods outlined above, while the form of V
can be estimated by fitting Eq. (10.5) to a single measured batch drying curve
(see Fig. 10.8); V is assumed to be unaffected by changes in external
conditions. For analytical purposes Eq. (10.5) is sometimes simplified by
assuming that:
V=

X-X
Xcr

N=Nc

Xe

(X-X)
X

Figure 10.8 Normalization of batch drying curves:


(a) batch drying curves at different external conditions;
(b) normalized batch drying curve.

cr

-;

For many materials this may be an adequate approximation if only a rough


answer is required; it is equivalent to representing the curve on Fig. 10.7(b)
by a straight line joining the points (Ne, Xcr) and (0, 0). Greater accuracy canbe achieved by approximating the curve either by a series of straight line
portions or by an empirical power function.
The concept cannot be expected to hold exactly, since the critical moisture
j::ontent tends to increase slowly as drying conditions are made more severe.
However, it has been shown to hold approximate Iy for materials as diverse as
molecular sieve, paper, textiles, ion exchange resin, iron ore, gelatin, and

crystalline agglomerates. Its virtue is that it enables batch drying curves to be


transformed from one set of operating conditions to another without any
fundamental knowledge of the structure of the material or of the mechanism
of moisture movement within it. It is unlikely to give acceptable accuracy
when applied to conditions giving an order of magnitude difference in the
drying rate, because then the change in critical moisture content will be too
great.
10.3.4

Effect of bed depth on drying rate

The available experimental evidence indicates that. for materials which dry
relatively quickly, - dXldt is almost inversely proportional to bed depth
(Mostafa, 1977; Reay and Allen, 1982). Therefore, the drying rate - WB
dXldt, where WB is the dry bed weight, is effectively independent of bed

depth, and increasing the bed depth does not increase the drying rate. It
appears that most of the drying takes place in a shallow layer just above the
distributor, and increasing the bed depth merely reduces the frequency with
which an individual particle enters this layer. If we assume that the gas flow
can be divided into an interstitial flow and a bubble flow with most of the heat
and mass transfer occurring in the former and not much gas exchange
occurring between them, then a plausible explanation is that the mass transfer
rate is so rapid that the interstitial gas becomes almost saturated a short
distance above the distributor (see Fig. 10.9a).
Bed depth
I
I
I

- ---------

..- -Soturation

the bed surface (see Fig. 1O.9b). If relatively quick drying materials have an
appreciable falling rate period they also tend towards this type of behaviour
when the drying rate has declined to a very low value.

10.4.1
Let F

= solid

Mass and heat balances

mass flowrate (dry basis)

= gas mass flowrate (dry basis)

= solid

moisture content (kg liquid/kg dry solid)

= gas moisture content

(kg vapour/kg dry gas)

Subscripts:

F(Xi
Let HF

Bed depth
I
I

Xo)

= G (o - j)

enthalpy of 1 kg of dry solid

+ associated liquid

Hg

== enthalpy of 1 kg of dry gas + associated vapour

Qw

= rate
= rate

Qh

- - - - - - - - - - - ~----

of heat loss from walls


of heat input from heater immersed in bed

Saturation

I
I
I
I
t
I
I
I

I
I

I
I
I
I
I

I
t

A heat balance over the dryer yields:


FHFj

+ GHgj +

Qh

On the other hand, with a material such as wheat which dries very slowly,
- dX/dt is almost independent of bed depth (Reay and Allen, 1982).
Therefore, the drying rate - WB dX/dtis directly proportional to bed depth. It
may be surmised that in this case the intersitial gas is still far from saturated at

+ GHgo +

(10.7)

Qw

Let Cs, Ch Cg and Cv be specific heats of solid, liquid,gas and vapour


respectively. Let TF be the temperature of the solid and associated liquid, and
tg the temperature of the gas and associated vapour. Let A. be the latent
heat of vaporization at
(2,500 kJ/kg for water). Then, referring to an
enthalpy base of
we have:

aoe,

Figure 10.9 Postulated increase of vapour


partial pressure with height above the distributor in the interstitial gas flow: (a) fast drying;
(b) slow drying.

= FHFo

HFj
HFo
Hgj
Hgo

aoe

= (Cs + XjC1)
=
=

TFj
(Cs + XOC1) TFo
(Cg + jCv) Tgi + jA
(Cg + oCv) Tgo + o A

Substitution in Eq. (10.7) and rearrangement

(10.8)
(10.9)
(10.10)
(10.11)
yields:

G (Cg + jCv) Tgj - (Cg + oCv) Tgo - A(o - i)


= F (Cs + XoC1) TFo - (Cs + jCD TFj + Qw

+ Qh

If a psychrometric chart is available for the gas-vapour system in question, Hgi


and Hgo would normally be read off directly from it. Equation (10.12) would
still be used for computerized calculations, however, and is essential for many
organic vapour-inert gas systems for which no psychrometric chart is
available.
10.4.2

Calculation of required mean particle residence time tR

The starting point of this calculation is always a batch drying curve measured
in laboratory equipment at a suitable bed depth z and gas velocity U. The
latter is chosen to be high enough to ensure good fluidization but low enough
to avoid excessive elutriation. The measured batch drying curve can be
transformed to other bed depths and gas velocities by means of the rules
presented earlier.
For simulation of a plug flow fluid bed dryer the batch drying curve should
be measured with the air inlet temperature Tg; constant at the desired value.
This will usually be dictated by material temperature limitations, with an
appropriate safety margin. If the deviation from plug flow is small, the
temperature and moisture content histories of the material as it flows along
the bed will correspond to its temperature and moisture content histories on
the batch drying curve, so the required mean particle residence time tm can be
read directly from the appropriate curve.
The calculation is more complicated for a bed with a well-mixed particle
residence time distribution. This case has been analysed in detail by Vanecek,
Markvart, and Drbohlav (1966). In this case, in order to simulate the passage
of particles through the continuous bed the batch drying curve should refer to
a constant bed temperature rather than a constant air inlet temperature. This
means that during the course of the laboratory experiment the air inlet
temperature must be progressively reduced as the drying rate declines in
order to prevent the bed temperature rising. A good control system is
needed.
In the continuous operation let Xl be the moisture content of a particle
after residence time t in the bed, and let Xo be the average moisture content
of the product. If E(t) is the residence time distribution function of the
product we have:
Xo

l;t

E(t) dt

(10.13)

and using Eq. (10.1) for the residence time distribution in a perfectly mixed
vessel we obtain:

In this expression the function Xl is given by the batch drying curve measured
at constant bed temperature.
In the general case, Eq. (10.14) must be integrated numerically with an
arbitrarily chosen value of tR, and by trial and error the value of tR found
which gives the desired Xo' This is not difficult with a computer or a
programmable calculator.
In certain special cases an analytical solution may be possible. For example,
suppose the batch drying curve can be approximated by a linear falling rate
curve with no constant rate period, i.e.:
dX
dt

--=

k(X-

X)

Integrating between (Xi, 0) and (Xl, t) we obtain:


Xt-Xe
X _ X

Xl

= exp (-kt)
e

Xe

(Xi - Xe) exp (-kt)

Substituting for Xl in Eq. (10.14) and integrating yields:


Xj-Xe

Xo

10.4.3

= Xe + 1 + k tR

Calculation of the bed dimensions

Let the bed area be A, the bed depth z, and the bed density
measured in the laboratory for a given gas velocity U). Then:
Bed hold-up

= FtR =

PB

(easily

PBzA

and

Since tR is calculated using a batch drying curve corresponding to a selectpd


bed depth z, A can be calculated from Eq. (10.16).
The calculation of tR, and hence of A, may be repeated for other bed
depths using the batch drying curve transformation rules. However, if the

material is one which dries relatively quickly so that - dX/dt is approximately


inversely proportional
to bed depth, then A is likely to be sensibly
independent of bed depth. On the other hand, if the material dries very
slowly so that - dX/dt is almost independent of bed depth, A is likely to be
inversely proportional to bed depth.
Having calculated A, the corresponding gas mass flowrate G at the selected
gas velocity can be calculated from:
G

PgUA

(0.20
)
0.04 - 1

800 s

From Eq. (10.16):

_ 1.39 x 800 _
2
A - 500 x 0.200 - 11.12 m

From Eq. (10.17):

= 0.70

x 11.12

7.78 kg/s

From the mass balance Eq. (10.6):

(10.17)

The gas exit humidity Yo can now be calculated from the mass balance Eq.
(10.6). If reasonable values are assumed for Qh and Qw, the gas exit
temperature Tgo can then be estimated from the heat balance Eq. (10.12).
The psychrometric chart should then be used to ascertain how far Tgo is above
the dew point of the exit gas. A difference of at least 10C is usually
considered necessary to avoid the risk of condensation in the cyclones. If they
are too close, additional bed area may be required to increase G and thereby
reduce Ygo'
If desired, the entire calculation procedure can be repeated with batch
drying curves corresponding to different gas velocities until an optimal design
is achieved.

G (Xi - Xo)

1.39
7.78 (0.20 - 0.04)

Yo

+ Yi

= 0.034 kg H20/kg

+ 0.005
dry air

Reference to a psychrometric chart shows that air of this humidity has a dew
point of 34C. Therefore, since the air leaves the bed at 80C there is no risk
of condensation.
TFi
Therefore,

20C,

Cs

1.0 kJ/ktC,

C1

4.2 kJ/kgOC

from Eq. (10.8):


HFi

Example 10.1:

= 0.005

tR

(1.0

+ 0.20

x 4.2) 20

36.8 kJ/kg

Design of a continuous perfectly mixed fluid bed dryer


and from Eq. (10.9):

Let 6,000 kg/h of a wet porous solid entering at 20C be dried from a water
content of 20 per cent. by weight (dry basis) to an average water content of 4
per cent. by weight. The bed depth is to be 200 mm, the bed temperature
80C, and the superficial mass velocity pgU of the inlet air 0.70 kg/m2s.
Under these conditions the bed density is 500 kg/m3 The humidity of the inlet
air is 0.005 kg H20/kg dry air. The specific heat of the dry solid is 1.0 kJ/kg C
and the specific heat of liquid water is 4.2 kJ/kg C. There is no internal heater
in the bed. Heat loss from the walls may be taken as 5 per cent. of the heat
content of the inlet air. At the selected operating conditions the constant bed
temperature batch drying curve for this material consists entirely of a linear
falling rate curve obeying the equation - dX/dt = 0.0005X S-l. Calculate the
required mean residence time, bed area, mass flowrate of air, and air inlet
temperature. Check that the exhaust air is far enough above its dew point to
avoid condensation problems in the cyclone, and estimate the thermal
efficiency of the dryer.
Solution

HFo = (1.0

+ 0.04 x 4.2) 80 = 69.6 kJ/kg

From a psychrometric chart, air at 80C with a humidity of 0.034 has an


enthalpy Hgo = 148 kJ/kg.
Qh
Rearranging

0,

Qw = 0.05 Hgi

the heat balance Eq. (10.7) yields:


Hg;

0.05 GHgi
G

Hgi

= G (HFo

I~
I~:~:
(HFo

HFi)

+ Hgo

(69.6 - 36.8)

Xi

= 0.20,

Xo

6,000
1.20 x 3,600

0.04,

1.39 kg/s dry solid

Xe

+ Hgo

HFi)

+ 148

0.~5

} 0.~5

Reference to a psychrometric chart shows that air with an enthalpy of 162


kJ/kg and a humidity of 0.005 has a temperature of 147. This is the required
air inlet temperature.
Heat content of the inlet air

Hgj

7.78 x 162

The dimensionless group B = DptR/ L 2, known as the axial dispersion


number, is a measure of the degree of particle backmixing along the length of
the bed, and hence of the degree of deviation from plug flow. For perfect plug
flow B = 0 and for perfect mixing B = 00.
From experiments on a variety of materials in beds of depth ranging up to
10 cm, Reay (1978) derived the following correlation for particle diffusivity:

= 1,260 kJ/s

For water at 80C, the latent heat of vaporization

= F(X

=
=

Xo)

507 kJ/s

PARTICLE RESIDENCE TIME DISTRIBUTION


'PLUG FLOW' BEDS

IN

Perfect plug flow of particles is an idealization which is never achieved in a


real 'plug flow' fluid bed dryer. The extent of the deviation from the ideal will
depend mainly on the geometry of the bed, the mean residence time of the
particles, the gas velocity, and the size and density of the particles. In. many
applications it is important to control this deviation within specified limits.
Particle mixing in a horizontal plane, which is what determines the particle
residence time distribution (RTD) in shallow beds, has been shown to obey
Fick's second law of diffusion for bed depths up to 10 cm (Reay, 1978):

ac

ax2

(10.18)

where c is the concentration of labelled particles monitored at a distance x


along the bed, t is time, and Dp is the diffusivity of the particles. If the
particles have a net flow velocity Vs along the bed an additional term must be
added to the right-hand side of Eg. (10.18), which becomes:

iflc
Pax

-=D-'--v2

ac
s~

The net flow velocity Vs = LltR, where L is the bed length. In terms of the
dimensionless variables e = t/tR and ~ = x/L, Eg. (10.19) becomes:

ac

DptR

371

iflc

ac

ae-=2 a~2 -ag-

10-4 (U - Umf)

u~i

m2/s

(10.21)

(a) The RTD can be represented


by function Eo = f (B, 8), where
Eode is the fraction of particles which have dimensionless residence
times between e and e + de.
(b)

J:

Eo de
than 8.

fraction of particles with dimensionless residence times less

(c) For small deviations from plug flow, i.e. for small values of B, the
solution to Eq. (10.20) yields the symmetrical curve:

iflc

at = D

ac
at

p'

where the gas velocity U and the minimum fluidization velocity Umf are in
metres per second. The term (U - Umf) in the numerator is clearly a measure
of the extent of bubbling in the bed, but the denominator term is completely
empirical.
Having established that particle diffusion in a horizontal direction in a
shallow fluidized bed obeys the same basic equation as molecular diffusion,
we ca~ make use of the body of theory developed for deviation from plug flow
of flUids through process vessels. A good account of the latter is given in
Chapter 9 of the book by Levenspiel (1972). For our present purpose the
most important concepts and results are the following:

2,300 x 1.39 (0.20 - 0.04)


507
=-x 100 =400/<
1,260
0

10.5

= 2,300 kJ/s. Therefore:

Eo

2 Y( 1TB)

exp

{(1-8)2
4B

The mean is at e = 1 and the dimensionless standard


RTD about the mean is given by:
Uo

!!....

(10.22)
deviation of the

Y(2B)

tm
This approximation is reasonable up to B = 0.1, which covers most
practical cases of true 'plug flow' fluid bed dryers.
(d) An alternative way of representing deviations from plug flow is to
consider the bed as consisting of a number n of perfectly mixed beds in
series. For perfect plug flow n = 00. With this model the dimensionless
standard deviation Uo = V(Un). Hence, comparing the two models we
have n = 1I(2B).

From the above analysis the designer can select the maximum value of B
which can be tolerated if the deviation from plug flow is to be kept within the
specified limits. Dp can be estimated from Eg. (10.21) and the required tR
estimated from the batch drying curve. Hence, the minimum acceptable bed
length L can be calculated. The process design calculations outlined earlier
yield the required bed area A, so the required bed width W can be found. In
the case of a bed with reversing or a spiral flow channel, Wand L should be
interpreted as the width and total length of the particle flow channel.
It should be emphasized that the validity of the above analysis has only
been confirmed for bed depths up to 10 cm. There is some evidence in the
literature that Dp may be an order of magnitude larger in much deeper beds
(Mori and Nakamura, 1965; Highley and Merrick, 1971). this is thought to be
due to the establishment of regular particle circulation cells in deeper beds, in
contrast to the more random particle motion in shallow beds. This large
increase in Dp with deeper beds suggests the use of the shallowest practicable
bed depth when the objective is a close approach to plug flow.
Calculation of the residence time distribution in a
'plug flow' fluid bed dryer
Bed length L

=5m

Mean particle residence time tR

= 1,000 s

Gas velocity U

= 0.4

Minimum fluidiization velocity Urnf

m/s

bed area

Et
E9

residence time distribution (RTD) function


dimensionless RTD function
function of parameter in parenthesis
solids feed rate (dry basis)
gas mass flowrate (dry basis)
gas-to-particle heat transfer coefficient
enthalpy of inlet solids
enthalpy of outlet solids
enthalpy of inlet gas
enthalpy of outlet gas
proportionality constant
thermal conductivity of gas
mass transfer coeffcient
dryer length

N
Nc
Nu

number of perfectly mixed stages


drying rate
drying rate in constant rate period
Nusselt number, hgp dplkg

partial pressure of vapour

fO
F

G
hgp
HFi
HFO
Hgi
Hgo
k

0.1 m/s

kg
kp

Dp

DptR

-4

3.7

=0=

2.4

10

0.4 - O.
0 1/3
.1
= 2.4 x 10-4 m2/s

10-4
25

1
=-

2B

1,000

52

0.0096

axial dispersion number, DptRI L 2


concentration of labelled particles
specific heat of gas
specific heat of liquid
specific heat .of solids
specific heat of vapour
particle diameter
particle diffusivity

kJ/kg K
kJ/kg K
kJ/kg K
kJ/kg K
m

m2/s

kg/s
i kg/s

W/m2K
kJ/kg
kJ/kg
kJ/kg
kJ/kg
W/mK
kg/m2s (N/m2)
m

partial pressure of vapour at the wet bulb


temperature
Prandtl number. Cg pJkg
rate of heat supply from immersed heater
rate of heat loss from dryer
Reynolds number, U dpfJgI f.L

kJ/s
kJ/s

bulk temperature of gas


solids inlet temperature
solids outlet temperature
gas inlet temperature
gas outlet temperature
wet bulb temperature
time
mean residence time
mls
gas velocity
mls
minimum fluidization velocity
mls
net flow velocity of particles
Bed width
weight of solids in bed (dry basis)
distance along bed
kg H20/kg solids
moisture content (dry basis)
kg Ho/kg solids
critical moisture content
kg H20/kg solids
equilibrium moisture content
kg H20/kg solids
inlet moisture content
kg H20/kg solids
outlet moisture content
mean outlet moisture content from a
kg H20/kh solids
well-mixed bed
kf H20/kg solids
moisture content at time t
kg H20/kg dry air
gas inlet humidity
kg H20/kg dry air
gas outlet humidity
bed height
dimensionless time, t/tR
kJ/kg
latent heat of vaporization
Ns/m2
viscosity of gas
defined in Eq. (10.5)
dimensionless distance, x/L
bulk density of bed
density of gas
standard deviation of RID function
standard deviation of dimensionless RTD function-

Gummel, P., and Schliinder, E.U. (1977). Verfahrenstechnik, II (12), 743.


Highley, J., and Merrick, D. (1971). Chem. Eng. Prog. Symp. Ser., 67 (116), 219.
Keey, R.B. (1978). Introduction to Industrial Drying Operations, Pergamon Press.
Kunii, D., and Levenspiel, O. (1969). Fluidization Engineering, John Wiley and Sons,
Chapter 7.
Levenspiel, O. (1972). Chemical Reaction Engineering, 2nd ed., John Wiley and Sons,
Chapter 9.
Mori, Y., and Nakamura, K. (1965). Kagaku Kogaku, 29, 868.
Mostafa, 1. (1977). 'Studies of fluidized bed drying' Ph.D. Thesis, Imperial College of
Science and Technology, University of London.
Reay, D. (1978). Proceedings of the First International Symposium of Drying, McGill
University, Montreal, Canada, published by Science Press, p. 136.
Reay, D. and Allen, R.W.K. (1982). Proceedings of the Third International Drying
Symposium, Birmingham, 2, p. 130-140.
Subramanian, D., Martin, H., and Schliinder, E.U. (1977). Verfahrenstechnik, II
(12), 3.
van Meel, D.A. (1958). Chem. Eng. Sci., 9,36.
Vanecek, V., Markvart, M., and Drbohlav, R. (1966). Fluidised Bed Drying,
Leonard Hill.
Zabeschek, G. (1977). Ph.D. Dissertaion (in German), Universitat Karlsruhe,
Germany.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Fluid Beds as Chemical Reactors

The earliest applications of fluidization were for the purpose of carrying out
chemical reactions (see Chapter 1). Since that time there have been many
successful chemical processes involving fluid bed reactors. The principal
applications are listed in Table 11.1 under three headings:
Table 11.1

Some applications of fluidjzed--bedsas chemical reactors

Catalytic gas phase reactions


Hydrocarbon cracking
Catalytic reforming
Phthalic anhydride manufacture
Acrylonitrile production
Aniline production
Synthesis of high-density and low density polyethylene
Fischer- Tropsch synthesis
Chlorination or bromination of hydrocarbons
Oxidation of S02 to 503
Methanol to gasoline process
Non-catalytic gas phase reactions
Hydrogenation of ethylene
Thermal cracking, e.g. to give ethylene or for fluid coking
Gas-solid reactions
Roasting of sulphide and sulphate ores
Calcination of limestone, phosphates etc.
Incineration of waste liquors and solids refuse
Combustion of coal, coke, peat, biomass
Gasification of coal, peat, wood wastes
Pyrolysis of coal, silane
Catalyst regeneration
Fluorination of V02 pellets
Chlorination of rutile, ilmenite
Hydrogen reduction of ores

(a) Catalytic gas phase reactions. These are cases in which the solid particles
serve as catalysts, undergoing no chemical changes and limited physical
changes during the course of the reaction. The catalyst materials are
manufactured or selected to combine favourable activity and selectivity
characteristics with suitable physical properties - particle size and
density, resistance to attrition, etc. The most important application
historically is the catalytic cracking of petroleum.
(b) Non -catalytic gas phase rections. In this case the particles are inert. Their
purpose is to provide a uniform heat transfer medium or nuclei upon
which the product material can be deposited (as in fluid coking) or both.
Applications in this category are relatively rare.
(c) Gas-solid reactions. Reactions of this type involve the fluidized particles
themselves. The useful product may be the solid, a gas, thermal energy,
or some combination of these. In reactions of this type there is generally
less control over the properties of the solids than can be achieved in
catalytic reactions. Moreover, the solids undergo changes in their
chemical composition and their physical characteristics during the course
of the reaction. The most important applications of fluidization for gassolid reactions are in roasting of ores, calcination reactions, regeneration
of catalysts, and processing of coal and biomass.
The principal advantages of fluidized beds which make them attractive for
many chemical processes relative to fixed beds and other types of gas-solid
reactors are:
(a) Temperature uniformity. 'Hot spots' are avoided.
(b) Favourable rates of heat transfer to immersed tubes or to the walls of the
column.
(c) Ease of solids handling. Solids can be continuously added and removed
from the system. In catalytic reactors the deactivated solids may be fed to
a parallel regenerator bed and then recycled to the main reactor.
(d) Scale of operation. Successful operations have been achieved with
columns as small as 0.05 m and as large as 30 m in diameter.
(e) Turndown. The gas flowrate can be varied over a wide range.
(f) Pressure drop. The pressure drop across a fluidized bed of solids is less
than for the same bed at the same superficial gas velocity under fixed bed
conditions, especially for fine particles.
On the other hand, fluidized beds have certain limitations and disadvantages
which must be clearly recognized when considering what type of reactor is
best suited to a given process:
(a) Backmixing of both solids and gas is generally substantial, resulting in
lower conversions than with most competing types of reactor.

(b) By-passing of gas via bubbles or jets causes gas-solids contacting to be


unfavourable. This leads to a further lowering of conversions, and may
also contribute to poor reactor selectivity.
(c) Entrainment of solids may lead to loss of expensive materials or product
as well as to pollution of the atmosphere and the need for pollution
control equipment.
(d) Attrition, erosion. and agglomeration may cause serious operational
problems.
(e) Scale-up and design of fluidized bed reactors is an uncertain undertaking
because of the complexity of fluidized beds and limitations in the extent
to which their behaviour can be predicted and modelled.
(f) There are some limitations on the size of particles that can be treated.
Generally speaking, gas fluidized beds are not well suited for particles of
mean size less than about 30 /Lm or greater than about 3 mm.
Some of these limitations can be lessened by modifications in design. For
example, particles too small to fluidize properly can often be treated by
stirring or vibrating the bed. Backmixing of solids can be greatly reduced by
using stages in series. By-passing of gas can be lessened by means of properly
designed baffles. However, all of these modifications involve considerable
extra cost and operational complexity which may make the fluid bed process
uncompetitive.
In the remainder of this chapter, we consider methods of modelling fluid
bed reactors for purposes of design, scale-up, simulation, and control.
Considerable effort has gone into devising models, but the results have been
disappointing on the whole. Nevertheless, the models help to give a good
understanding of the phenomena involved_and.of the factors and variables
which influence the performance of fluidized bed reactors. While it is
impossible to predict the conversion or selectivity of fluidized beds with
complete certainty. the models do provide a rational basis for scaling-up
fluidized beds and for simulating existing reactors. Because the contacting
between solids and gas is quite different in three different regions - the
region just above the gas distributor (which we will call the grid region), the
bed proper, and the freeboard region above the bed surface - these regions
are considered separately.

Fluidized beds are far too complex hydrodynamically to be used for deriving
reaction kinetics. Instead, kinetic parameters should be obtained in reactors
where the hydrodynamics can be described with confidence. Fixed beds are
well characterized by dispersed plug flow models (see Levenspiel, 1972;Wen
and Fan, 1975), but temperature gradients may be severe. Consequently, it
may be better to use a spinning basket reactor (TajbI, Simons, and Carberry,

1966) or other suitable laboratory reactors (e.g. Carberry, 1964) to derive


reaction kinetics.
Kinetic data should be obtained for a range of temperatures spanning the
range in which operation is to take place. It is also essential that kinetic data
be obtained under much higher conversion conditions than are anticipated for
the fluidized bed as a whole. This is- because of the two-phase nature of
fluidized beds: Most reaction occurs in a dense phase region where the gas is
highly converted, the final composition being achieved by mixing of this gas
with poorly reacted dilute phase gas.
Catalytic reactions occurring on the interior surface of porous catalyst
particles involve the following sequence of physical and chemical steps:
(a) Mass transfer of reactants
catalyst particle.

from the bulk gas to the outside of the

(b) Diffusion of reactant into the catalyst pores.


(c) Chemisorption

of reactant species

(d) Chemical reactions on the surface


(e) Desorption
(f)

of product molecules

Diffusion of products back to the exterior surface of the particle

which gas volume changes due to reaction can be neglected. Allowance for
volume changes accompanying reaction is complicated due to the fact that
most reaction takes place in the dense phase of fluidized bed reactors, while
most of the added volumetric flow probably finds its way into the bubble
phase.
11.3

BUBBliNG

REGIME: GAS MOTION IN AND AROUND BUBBLES

Bubbles in fluidized beds are responsible for by-passing of gas and for limiting
gas-solid contacting. While there are important similarities between bubbles
in fluidized beds and bubbles in liquids, the gas motion in the two cases is
fundamentally different. In gas-liquid systems, the gas must circulate entirely
within the bubble. In fluidized systems, on the other hand, the bubble
boundary is permeable, and gas circulates through the bubble, entering at the
bottom and leaving through the upper surface. This factor, commonly
referred to as 'throughflow',
has important consequences for gas-solid
contacting and for interphase (bubble to dense phase) mass transfer in
aggregative fluidized beds.
Consider a bubble rising at steady velocity Ub' Whether or not elements of
gas leaving the top of the bubble recirculate through the dense phase and
reenter the bottom of the same bubble depends on the ratio

(g) Mass transfer of products from the exterior surface into the bulk
Catalyst particles used in most fluidized bed processes are quite small,
typically 50 to 120 J.tm in mean diameter, small enough for external mass
transfer and internal diffusion resistances to be minor. Temperature gradients
inside the catalyst particles can also usually be neglected. For many purposes
it is permissible to use a single apparent or overall volumetric rate constant
kn, defined as:
'A

-1 dNA
V dt
p

bubble velocity
a

remote interstitial velocity

Ub Emf

U mf

For a > 1 recirculation does occur. The region surrounding the bubble in
which gas recirculates (see Fig. 11.1) is called the 'cloud'. According to the

= knC;'

where n is the order of reaction and C A is the local bulk concentration of


gaseous reactant A. We will assume rate expressions of this type for catalytic
reactions throughout this chapter. The rate constant kn is based on unit
particle volume. Where diffusional resistances are significant, the kn thus
defined includes allowance for the effectiveness of the catalyst pores. If the
kinetics follow more complex rate expressions (e.g. Smith, 1970), these
expressions may be readily incorporated within the reactor models described
below, but solution will then generally require numerical integration.
In addition to the order of reaction and rate constant as a function of
temperature,
laboratory kinetic tests should yield the stoichiometry of the
desired reaction and indications of side reactions or other difficulties. The
heat of reaction and equilibrium conversion can be calculated from
thermodynamic
considerations. In this chapter we consider only cases in

"'ass transfer at
bubble / cloud boundary

Figure 11.1 Shape of cloud surrounding a two-dimensional gas bubble in a fluidized


bed, as observed by Rowe, Partridge and Lyall (1964) for a = 2.5. Resistances to
mass transfer may be accounted for at the bubble/cloud boundary and/or at the
cloud/emulsion interface.

theory of Davidson and Harrison (1963), the cloud for a spherical bubble also
has a spherical boundary with the same centre and with volume:*
Vc = Vb

( 3)
a-I

1T

d~q

2 (a _

1)

(11.3)

An alternative analysis (Murray, 1965) leads to a non-spherical cloud which is


smaller and which has its centroid above the centre of the (assumed) spherical
bubble. An empirical equation (Partridge and Rowe, 1966) which has been
fitted to the results of this analysis gives
Vc

Vb (~)
a-I

0.195 1T d~q
a-I

Example 11.1: Cloud volumes


Estimate the ratio of cloud volume to bubble volume for a bubble of volume
equivalent diameter 0.10 m rising through an air-fluidized bed of:
(a) Cracking catalyst particles: dp = 80 ,urn, Umf = 0.005 mis,
(b) Silica sand: dp = 250 ,urn, Umf = 0.056 mis, Emf = 0.45

Emf

0.50

= 450 ,urn, Umf = 0.16 m/s. mf = 0.45


= 700 ,urn, Umf = 0.32 mis, Emf = 0.48
= 1.2 mm, Umf = 0.54 mis, Emf = 0.48

(c) Silica sand: dp


dp

(e) Dolomite:

dp

Solution
The rise velocity of the bubble relative to the dense phase in each case is
(from Eq. 4.9):
Uh

0.71 (gdeq)l/2

0.71 (9.8

0.10)1/2 mls

Case

Umf, m/S

Emf

a
b

0.005
0.056
0.16
0.32
0.54

0.50
0.45
0.45
0.48
0.48

c
d
e

a
70
5.6
2.0
1.05
0.62

V.jVb from Eq. (11.3)

V.jVb from Eq. (11.4)

0.043
0.017
0.65
0.25
1.2
3.1
54
21
No clouds since a < 1

(11.4)

Experimental results (Rowe, Partridge, and Lyall, 1964; Anwer and Pyle,
1974) indicate closer agreement with the Murray analysis, but both theories
have been used as the bases for reactor models, as we will see below. For a
review of the fluid mechanics of fluidization and analyses of gas motion
around bubbles, see Jackson (1971).
As shown by Eqs (11.2) to (11.4), the cloud volume is strongly related to a
and hence to particle size. Typical cloud volumes are estimated in the
following sample problem.

(d) Dolomite:

The values of Umf and Emf together with values of a calculated from Eq.
(11.2) and V c!Vb from Eqs (11.3) and (11.4) appear below:

0.70 rols

'The terms cloud and cloud volume refer in this book only to the outer dense phase mantle
(particle and interstitial gas) surrounding the bubble in which gas recirculates. Occasionally
(e.g. Partridge and Rowe, 1966) the terms are used to include also the gas contained in the
bubble.

Clouds form the basis for a number of the reactor models to be discussed
below. For a < 1, gas emerging from the upper surface of bubbles does not
recirculate. Instead the gas short circuits through the bubbles on-its way
through the bed. Hence cloud models are not applicable to large particle
systems (often called 'slow bubble' systems) where a < 1. A qualitative
explanation of why a plays such a role may be helpful.
All of the theories suggest that the velocity of the throughflow gas at the
bubble roof is proportional
to the interstitial gas velocity, Umf/Emf'
(Davidson's theory gives a proportionality constant which is three times
greater than Murray's, and this accounts for the larger predicted clouds.) The
gas emerging into the dense phase from the bubble surface encounters
particles which are moving tangentially around the bubble surface with
velocities of order Ub and decreasing with increasing radial distance from the
bubble surface. This situation is shown in Fig. 11.2. If the interstitial velocity
Umrlemf is much less than Ub (i.e. for large-a), the particles near the bubble
drag the gas around the bubble surface in a tight trajectory and a thin cloud is
formed. As Umf/emf increases (i.e. as a decreases), the gas elements
penetrate further into the dense phase before the drag caused by the solid
particles is able to reverse the direction. A larger cloud therefore results. For
a sufficiently large Umf/Emf where a < 1, the gas is able to penetrate so far
that it never recirculates.
Pyle and Rose (1965) predicted gas streamlines inside rising bubbles. The
most notable feature of their predictions was that there were regions at all
values of a in which gas would circulate entirely within the bubble, without
ever making contact with particles. Total by-passing of this gas would
therefore occur. In practice, molecular diffusion, bubble dilations, and
raining through of solids are expected to cause rapid mixing of the gas within
each bubble. Hence almost all subsequent workers have assumed perfect
mixing within each bubble unit, and we will limit our attention to models of
this type.

/nerlia force. 60s elemerts


cross the bubble bollldO!'y
with a velocity of order Umf.

6asmotion
inside006'

I+

Drag force due to parfides


moving tangential to I:t.JbbIe
boundary. The velocity of these

particles is of order ub

Figure 11.2 Forces acting on an element of gas emerging from a bubble into the
dense phase. The frame of reference is moving with the bubble.

The performance of fluidized bed reactors is greatly influenced by the rate of


. mass transfer between the dilute phase (bubble phase or bubble plus cloud
phase) and the remaining dense phase. As shown in Fig. 11.1, two resistances
to mass transfer may be considered for cases in which cloud formation occurs:
(a) Resistance at bubble boundary. Mass transfer at the outer surface of the
bubble occurs due to two parallel mechanisms: convective flow (also
called throughflow) and diffusion.
(b) Resistance at cloud boundary. Mass transfer at the cloud boundary
occurs due to molecular diffusion, due to gas adsorption on (or capture
by) particles moving through the cloud region, and due to shedding of
elements of the cloud during wake shedding, bubble shape dilations, or
as a result of coalescence.
Most of the chemical reactor models assume that one of these resistances is
much larger than the other, and this allows the clouds to be lumped with
either the dense phase or bubbles so that two-phase models, rather than

three-phase models, can be adopted. An exception is the model of Kunii and


Levenspiel (1969) which allows for resistances in series at the bubble and
cloud boundaries. If the resistance at the cloud boundary is ignored, then no
distinction is made between gas elements in the cloud and those in the dense
phase at large at the same level, and the models consider a bubble'phase and a
dense phase comprising all (or essentially all) particles and interstitial gas. If,
on the other hand, the resistance to mass transfer at the bubble boundary is
ignored, the clouds are lumped with the bubbles as one phase. The remainder
of the bed, often called the emulsion, then comprises the other phase.
Interphase mass transfer rates have been measured or deduced in a number
of studies. Almost all of the work has been for single isolated bubbles. If due
allowance is made for end effects and for the tendency of bubbles to grow
when injected into a bed fluidized at superficial velocities somewhat in excess
of Urnf, Sit and Grace (1978) have shown that the transfer rate can be
expressed as the sum of two terms. For isolated three-dimensional bubbles
the interphase mass transfer coefficient is then given by:

The first term represents the throughflow component, as given by the analysis
of Murray (1965). (In a number of reactor models the corresponding result
from Davidson's analysis (Davidson and Harrison, 1963) is adopted, leading
to the first term being larger by a factor of 3.) The second term in Eq. (11.5)
describes diffusion according to .the-penetration theory. The bulk flow or
convective term is dominant for large particles, but the diffusion term
becomes more important for small particles, as shown in the following
worked example.
Example 11.2: Interphase mass transfer rates
Determine the two components of the interphase mass transfer coefficient
and the total coefficient for the transfer of hydrogen in air at 20Cfor the five
cases of Example 11.1. The molecular diffusivity is 6.9 x 10-5 mls.
Solution
From Example 11.1, deq = 0.10 m and Ub = 0.70 m/s. We are given qj) = 6.9
x 10-5 m2/s. Inserting these values into Eq. (11.5) with the appropriate
values of Urnf and Emf from Example 11.1, we obtain the following results:

294

GAS FLUIDIZATION

TECHNOLOGY

Case

Particles

Convection
term, mm/s

Diffusion
kq,
term, mm/s mm/s

a
b
c
d
e

80 /Lm catalyst
250 /Lm sand
450 /Lm sand
700 /Lffidolomite
1.2 mm dolomite

1.3

17.5

14.0
40.0
80.0
135.0

16.6
16.6
17.2
17.2

18.8
30.6
56.6
97.2
152

Bulk

flow/kq
0.07
0.46
0.71
0.82
0.89

In addition to the influence of decreasing particle size, the diffusion term also
becomes more important as the molecular diffusivity increases and as the
bubble size decreases.
The fact that Eq. (11.5) gives a good representation of transfer rates from
the bubble phase implies that the principal resistance to mass transfer resides
at the bubble boundary, not at the cloud boundary. Theories for mass transfer
at the cloud boundary based on diffusional transfer alone (Partridge and
Rowe, 1966; Kunii and Levenspiel, 1969) predict much lower mass transfer
coefficients than are measured or inferred in practice. This implies that
neglected mechnisms of transfer at the cloud surface (in particular due to
adsorption by particles moving through the cloud region, due to unsteady
bubble motion, or due to shedding of cloud/or wake elements) are sufficient
to cause rapid dispersion of gas in the dense phase once gas elements are
transferred from a bubble. See Walker (1975), Chavarie and Grace (1976),
Lignola, Donsi and MassimiIIa (1983), and Grace (1984) for further
discussion and evidence on this point.
For bubbles undergoing interaction and coalescence, there is evidence of
enhanced interphase mass transfer (Toei et ai., 1969; Pereira, 1977; Sit and
Grace, 1978, 1981) compared with isolated bubbles. Indications are that the
throughflow term in Eq. (11.5) is augmented by 20 to 30 per cent. for a freely
bubbling bed when averaged over time, while the diffusion term is unaffected
(Sit and Grace, 1981). For freely bubbling beds the following relationship
should be used (Sit and Grace, 1981):
kq

Umf
=-+
3

(4 flJ _

Emf !lb.)

1T

Early workers tended to treat fluidized beds as if the gas and solids were
intimately mixed without segregation into dilute and dense phases. This led to
simple 'single-phase models' in which it was assumed that reactor performance could be determined from the residence time distribution of gas
elements. In practice, however, reactor performance is likely to be affected
less by the residence time distribution than by the 'contact time distribution',
i.e. by the distribution of times that gas elements spend in the vicinity of, and
in effective contact with, solid particles. Gas elements in the bubbles are in
poor contact with solids, while those in the dense phase are in intimate
contact. Hence single-phase models are unable to account for the true flow
and mixing behaviour of aggregative fluidized beds, and we must therefore
look at more complex models in which two or more phases are considered.
Single-phase models do, however, provide limits for the two-phase models
considered below. If we consider a first-order reaction (n = 1), the limiting
results are as follows:

deq

The enhancement of transfer is therefore more important for large particles,


where the bulk flow term is dominant, than for fine particles, where the
diffusion term plays the major role. Equation (11.6) is used in reactor
modelling below.
Interphase mass transfer rates can also be inferred from residence time
distribution data, if an appropriate unsteady state two-phase model is used.
This possibility is covered in Section 11.9 below.

--

=exp (-k1)

(11.7)

1
1 + ki

(11.8)

CAin
CAout

_Perfectly mixed reactor

--=-CAin

where k1 is a dimensionless rate constant defined by:


k1

k1 x total particle volume


.,
volumetnc flUIdflowrate

One or other of these results (depenaIng on the assumed dense phase gas
mixing pattern) will be approached as limiting cases for the simpler two-phase
models considered below, as mass transfer of reactants between phases
becomes very rapid or as the proportion of gas passing through the dilute
phase approaches zero.
11.6

112

CAout

Plug flow reactor

TWO- AND THREEPHASE REACTOR MODELS:


BASIC ASSUMPTIONS

In modelling any chemical reactor system, a primary objective must be to


represent accurately the key physical or hydrodynamic features of the system
before inserting chemical kinetics and other chemical parameters. At the
same time, it is important that models be sufficiently straightforward that
their application does not demand excessive computation. The many models
of fluid bed reactors that have been proposed and described differ in
complexity and in their view of what constitute the principal physical or
hydrodynamic features of the fluidized bed reactor.

The models considered in this section are based upon parallel onedimensional flow in two or three phases, as illustrated in Fig. 11.3. The word

Figure 11.3 Schematic diagrams showing phases chosen in alternative


fluid bed reactor models: B = bubbles; C = clouds; D = dense phase; E
= emulsion. (a) Two-phase models with all solids assigned to the dense
phase; (b) two-phase models with clouds included with bubbles;
(c) three-phase models.

'phase' in this context refers to a region which may include both gas and solid
particles. The phases are distinguished from one another in terms of the
volume fraction of solids, by physical appearance, and/or through their flow
characteristics. In one group of models, shown in Fig. l1.3(a), the dilute
phase is either free of particles or assumed to contain only widely dispersed
solids; in these models no distinction is made between the cloud and other
interstitial gas elements. In a second group of models, Fig. l1.3(b), the cloud
gas and solids (with or without bubble wakes) are lumped with the bubbles to
give a .combined bubble/cloud dilute phase. Gas within the cloud is then
assumed to have the same composition as gas within the bubble surrounded
by the cloud. In a third group the cloud is treated separately from both
bubbles and the remaining dense phase (often called the 'emulsion'), so that
there are three phases. This third group of models is illustrated schematically
in Fig. 11.3(c).
Having decided upon the general physical nature of the phases the model
builder must then decide on the distribution of gas flow between the phases,
on the extent of interphase mass transfer, and on the degree of axial
dispersion in each phase. Since each of these decisions is controversial, the
number of alternative assumptions which can be made is large, and the
number of combinations of assumptions even larger. Table 11.2 outlines the
assumptions upon which models have been based in the critical areas of phase
composition, flow distribution, axial dispersion and interphase mass transfer.
Table 11.3 lists some of the more popular and more recent models, and shows

A.

Nature of dilute phase:


1. Bubble phase completely free of particles.
2. Bubbles containing some widely dispersed solids.
3. Bubble-eloud phase, i.e. clouds are included.

B.

Division of gas between phases:


1. Governed by two-phase theory of fluidization (Chapter 4).
2. All gas carried by bubbles.
3. Some downflow of gas in the dense phase permitted.
4. Other or fitted parameter.

C.

Axial dispersion in dilute phase:


1. Plug flow.
2. Dispersed plug flow.

D.

Axial dispersion in dense phase:


1. Plug flow.
2. Dispersed plug flow.
3. Stagnant.
4. Well-mixe,dtanks in series.
5. Perfect mixing.
6. Downflow.
7. Bubble-induced turbulent fluctuations.

E.

Mass transfer between phases:


1. Obtained from independent gas mixing or mass transfer studies.
2. Fitted parameter for case under study.
3. Empirical correlation from pr-eviousor pilot plant data.
4. Bubble to dense phase transfer obtained from experimental or theoretical
single bubble studies.
6. Enhancement due to bubble interaction included.

F.

Cloud size (if applicable):


1. As given by Davidson theory.
2. Murray or modified Murray.
3. Wake not specificaIIyincluded or assumed negligible.
4. Wake added to cloud.
5. Recognized but assumed negligible.

G. Bubble size:
1. Not specifically included.
2. Single representative size for entire bed.
3. Allowed to increase with height.
4. Obtained from separate measurements, correlation, or estimated.
5. Kept as fitting parameter.

Assumptions or approaches embodied in some of the principal two-phase


and three-phase reactor models; letters and numbers in the table refer to Table 11.2

Table II.3

Assumptions
A

Authors
(a) Two-phase models:
Shen and Johnstone (1955)
Lewis, Gilliland, and Glass (1959)
May (1959)
Van Deemter 1961
Orcutt, Davidson, and Pigford (1962)
Partridge and Rowe (1966)
Mireur'and Bischoff (1967)
Kato and Wen (1969)
Bywater (1978)
Darton (1979)
Werther (1980)
Grace (19B4)
(b) Three
Kunii and
Fryer and
Fan, Fan,

phase models:
Levenspiel (1969)
Potter (1972a)
and Miyanami (1977)

1
2
1
1
1
3
1
3
1
1
1
2

B
1
2
1
4
1
1
1
2
4
1
1
2

2,1 1 or 2
3
1
2
1

1 1 or 5 3 NA
1 1 or 5 3 NA
1 NA
2
1
1 NA
2
1
1 1 or 5 4 NA
2,4
5
1
1
2
1,3 NA
1
1,3
4
1
1
5
5
1
7
4,5
5
5
1
3 NA
1
1
4,6 NA
1
3
1
1
2

3
6
2

4,5
4,5
4,5

Chavarie and Grace (1975). Reactant concentration profiles in both the


bubbles and dense phase were shown tq give a much more discriminating test
of the models than has been achieved with overall conversion determinations,
A qualitative comparison of the key features of the primary models tested
appears in Table 11.4, Generally speaking, all of the models tested were

G
1
1
1
1
2,5
4
1
3,4
2,4
3,4
3,4
2,4

1,4 - 2,4
4,5 2,4
2,3 3,4

the assumptions which have been adopted in these models, using the
categories given in Table 11.2. This list of models is far from exhaustive. An
c;:xpandedlisting of pre-1970 models appears in the article by Grace (1971).
Except for assumptions regarding axial dispersion in the dilute phase, where
all models except one assume plug flow, we find that there is no unanimity in
the assumptions adopted by different authors. Even in cases where the same
number appears in a given column in Table 11.3, competing models may
employ different approaches, For example, there are a number of expressions
which may be used to predict interphase mass transfer rates, as discussed in
Section 11.4
Not surprisingly, models whose assumptions differ as radically as those
listed in Table 11.3 can give very different predictions of chemical conversions
and selectivities. For simulation of existing reactors, operated over limited
ranges of variables, it may be possible to use a number of different models,
especially if fitted constants are introduced. On the other hand, for reactor
scale-up or for prediction of reactor performance from first principles, it is
extremely important to choose a suitable model from the many alternatives.
Experimental testing of a number of competing models was carried out by

Table 11.4 Comparison of key assumptions underlying models investigated by


Chavarie and Grace (1975)

Model or

Interphase mass
transfer

Solids assigned
bubble phase

10

Orcutt, Davidson, and Pigford


Orcutt, Davidson, and Pigford
Rowe (1964)

High
High
None

None
None
Davidson cloud

Partridge and Rowe (I 966)

Low

Murray cloud

Kato and Wen (1969)


Kunii and Levenspiel (1969)

Intennediate
Two resistances in
series (one low.
one high)

Davidson cloud
(later include wake)
Little or none

Gas flow through


bubble phase

Axial dispersion
in dense phase

Two-phase theory
Two-pbase theory
(U-Um,)A
+ gas
carried in clouds
(U-Um,)A
+ gas
carried in clouds
All

None
Perfect mixing
None

All

None
Well-mixed units
in series
None

deficient in certain respects, although some gave better representations than


others, From a statistical point of view none of the models was acceptable, a
conclusion which was also reached by Shaw, Hoffman, and Reilly (1974)
using a very different experimental technique in which selectivities, rather
than concentration profiles, were the basis for model evaluation.
In the next two sectio!1S---lWo
of the two-phase models are presented in
greater detail. The first of these, one of the Orcutt, Davidson, and Pigford
(1962) models, is presented because it is simple, widely used, and illustrates
many of the features of the two-phase models. In Section 11.8 another model,
the two-phase bubbling bed model, is presented. This is nearly as simple to
apply as the Orcutt model and appears to give a better representation of the
performance of bubbling fluidized bed reactors for most practical conditions.
The models treated in this chapter all assume negligible heat and mass
transfer resistances between gas and particles within the dense phase. The
entire bed is also treated as if it were isothermal. If these simplifying
assumptions are not adopted, energy balances must be written in addition to
mole balances, and separate accounting must be made of the particles. This
complicates the models greatly and is beyond the scope of this book. For a
discussion of non-isothermal reactor models and of the significance of heat
and mass transfer resistance within the dense phase, see Bukur, Caram, and
Amundson (1977).

In Sections 11.7 and 11.8 we consider only gas phase solid catalysed
reactions. Application of these models to gas-solid heterogeneous reactions is
treated in Section 11.12 below.
11.7

Substitution of this expression into Eq. (11.13) gives an equation for CAd
which can be solved for certain specific values of the reaction order n. The
final exit concentration of A is then obtained from a mole balance at z = H,
i.e.:

ORCUTT MODEL WITH PERFECT MIXING IN DENSE PHASE

The key assumptions of this model are that no solids are associated with the
bubble phase while gas is assumed to be in plug flow in the bubble phase and
perfectly mixed in the dense phase. We generalize the model somewhat by
letting the fraction of gas which flows through the bubble phase at any height
be {3,and by letting the interphase mass transfer coefficient (volumetric rate
of transfer per unit bubble surface area) be kq. In the model proposed by
Orcutt, Davidson, and Pigford (1962), reproduced by Davidson and Harrison
(1963), the following expressions for {3and kq were assumed:
{3
kq

= 0.75

Umf

U-Umf

= --

U
0.975go.25

(11.10)

q; 0.5

dO.25

(11.11)

Table 11.5 Outlet {;oncentrations derived from generalization of the model proposed
by Orcutt, Davidson. and Pigford (1962) for some simple cases

O-order
irreversible
(rA = ko)
Half-order
irreversible
(rA = ko.5 <:15)

CAou<

~-C

1+

CAin

eq

Other values of the flow distribution parameter {3and of the interphase mass
transfer coefficient kq can be used instead, based on more recent experimental evidence (see Chapter 4 and Section 11.4, respectively).
With the assumption that there are no particles in the bubble phase, the
change in the molar flux of species A at any height in the bubble phase must
be accounted for by interphase transfer, Le.:
{3U dCAb = kq (CAd -

CAb)

(11.12)

ObEbdz

where ab is the interfacial bubble area per unit bubble volume. A mole
balance over the entire dense phase yields:
(l-{3)U(CAin-CAd)

Jff

kq(CAb-CAd)abEbdz=(I-Eb)(1-Emf)HknCA2

(11.13)
where kn is the nth-order rate constant as defined in Eq. (11.1). Here mass
transfer resistance within the dense phase is ignored so that the gas
composition is treated as being the same in the particles and local interstitial
gas. This assumption is valid unless the particles are large or the reaction very
rapid (Bukur, Caram, and Amundson, 1977). The boundary condition on
CAb is CAb = CAin at z = O. CAdis a constant throughout the column for this
model in view of the assumption of perfect gas mixing in the dense phase.
Integration of Eq. (11.12) yields:
CAb = CAd + (CAin-CAd) exp

-kqabfhz

{3U

kb

= 1-

CAin

(k' )2
0.5

1 - f3e-X

First-order
irreversible:

4(1-I3-X)2

1+----

(kfJ.5l

+ I3kie-x

1 - ,Be-x

A-+B

I-V

2(l-,Be-X')

ki

(rA = ktCA)
Second-order
irreversible
(rA = k2Ci
(1-,Be-X?

First-order
reversible:
A~B,
(rA = kt(CA-CBIKe))
Consecutive firstorder: A -+ B -+ C;
no B in feed
kACA;
= kBCB

rA =
rB

1-,Be-x

CAout
---

ki - (ki?/{ki

Ke(I-l3e-X

= as for fiITst-ord'er Irreversl"ble case ab ove

CAin

CBOUl
kA
-C-.
= l+k' _(.le-X
Am
B ~

(l_,Be-X)2
l-(.lo-X+k'
~

order O. 0.5, 1, and 2, for a reversible first-order reaction, and for a


consecutive first-order reaction. All of these reactions are assumed to be
carried out under isothermal conditions with negligible volume change
occurring due to reaction. All of the dimensionless outlet concentrations are
given as functions of {3and two other dimensionless groups, a dimensionless
rate constant:

k'

k H '1n

)CAin<,,-J)

mf\

"

Emf

and a dimensionless interphase mass transfer group:


kqabEJ/

=---

(3U

X can also be thought of as the number of mass transfer units or the number

of times a bubble is flushed out during passage through the bed.


The dimensionless outlet concentration is shown in Fig. 11.4 for a first-order

"",

"

"

"

,,
,
\

11.8

\ '\

Single-phase fiug flow reacfr)(\ \


'--\\

""'\

\
\

Single-phase perfectly mixed ~\

001

"

\',

,,

"

...

o
~

absence of particles in the bubble phase, gas must be transferred to the dense
phase before reaction can take place. The expressions given in Table 11.5
could be used to construct curves for other values of {3or for reactions of other
orders. Generally speaking, the predictions become more sensitive to the
dimensionless hydrodynamic parameters {3and X with increasing order of
reaction. An example showing the use of this model to estimate chemical
conversions is given below in Example 11.3.
Although this model is useful in discussing the effects of different
parameters and simple enough to lead to a number of analytic results for
conversions and selectivities, as shown in Table 11.5, results are generally
disappointing. The predicted gas concentration profiles are not observed in
practice (Chavarie and Grace, 1975; Fryer and Potter, 1976). Moreover, the
lack of any solids associated with the bubble phase appears to lead to
underestimation of conversion for fast reactions. We turn therefore to a more
realistic two-phase model.

1
K; - k,Hmfrt-cmfJ
U

...

-.,

'- ...._-

-----

Figure 11.4 Outlet concentration for a first-order chemical reaction according 10 the
Orcutt model for f3 = 0.75 and several different values of the interphase mass transfer
group X.

reaction (n = 1) and for a particular value of {3,{3= 0.75, i.e. with 75 per
cent. of the gas passing through the bubble phase. The limiting results for
single-phase plug flow and perfectly mixed reactors, Eqs (11.7) and (11.8),
,respectively, are also shown in the figure as broken lines. As X ~ 00 or as {3
~ 0, the conversion, 1-CAout/CAin, predicted by the two-phase theory
approaches Eq. (11.8), i.e. that of the single-phase perfectly mixed reactor.
For this particular two-phase model, conversions are always less favourable
for finite X than for the perfectly mixed case. For slow reactions (k1 < 0.1),
the outlet concentration is controlled by chemical kinetics and virtually
independent of the hydrodynamics. However, for intermediate and fast
reactions, interphase transfer plays a very significant role, since, in the

TWO-PHASE BUBBliNG

BED REACTOR MODEL

The two-phase bubbling bed model (Grace, 1984) can be regarded as a


simplification of the three-phase bubbling bed model of Kunii and Levenspiel
(1969) and of the countercurrent backrnixing model of Fryer and Potter
(1972a). Under many circumstances it is possible to lump the cloud region
with one of the other phases to give a simpler two-phase representation
(Chavarie and Grace, 1975). The assumption of no net vertical gas flow in the
dense phase introduced by Kunii and Levenspiel is adopted. The accuracy of
the model is hardly affected for most practical purposes by these simplifications, providing that the parameters of the model are chosen carefully.
Since the net vertical flow of gas is assumed to be entirely through the bubble
phase, reflecting the tendency for downflowing solids to drag interstitial gas
dt>wnwardswith them, the (3 parameter of Section 11.7 is set equal to 1 and
ddes not appear. On the other hand, we require two new parameters, <!>b and
rPct, to describe the fraction of the bed volume occupied by solids which are
part of the bubble phase and dense phase, respectively. Values of <!>b and rPd
are discussed below.
Consider two parallel phases as shown in Fig. 11.3(a). The reaction is
assumed to be isothermal and to involve negligible volume change. A mole
balance on component A in the dilute or bubble phase yields:

where <!>b is the fraction of the bed volume occupied by bubble phase solids.
For the dense phase, a mole balance on component A gives:

kqahEb(CAd-CAb)

+ kn4>dQ3 =

(11.19)

where 4>dis the fraction of the bed volume occupied by solids assigned to the
dense phase.
These equations can be integrated in a straightforward manner with the
boundary conditions CAb = CAd = CAin at z = 0 for several simple cases. The
results for these cases are given in Table 11.6. For cases where the kinetics or
Table 11.6

Zero-order
irreversible
(rA = ko)
First-order
irreversible:
A --+ B (rA

Outlet concentrations derived from the two-phase bubbling bed reactor


model for some simple cases

CAout

ra

kACA;

= kBCB

1 - ~ (4)0 + rPd)

13 = 1; i.e.:

kqab~H
X=--U
where kq should be obtained from Eq. (11.6), while:
6
ab = =deq
The mean bubble
growth correlations
advocated by Fryer
from relationships
dimensionless rate

(11.21)

diameter d"q can be obtained from one of the bubble


outlined in Chapter 4 , evaluated at height 0.4H, as
and Potter (1972b). Similarly, Eband H can be estimated
given in Chapter 4. For an nth-order reaction, the
constant group is:

CAin

= ktCA)

Consecutive firstorder: A --+ B --+ C;


no B in feed;
rA =

group is the same as given by Eq. (11.17), but with

CAout
CAin

= exp { -k~{X(4)o
X

+ rP~) + k~4>orPd
+ k1 rPd

CAout

-C-- = exp (-FA)


Ain

Caou,
CAir

_ G{exp(-FA) -exp (-Fa)}


Fa - FA

where FA

k:"'{X( 4>0 + rPd) + k:'" 4>orPd}


X + k:'" rPd

= -~--'----"---'-'-~;""::':"-

k~{X(4)o + rPd) + k~ rPl/>d}


X + k~ rPd

boundary conditions are more complex, numerical integration is required,


and is generally straightforward.
The results for first-order irreversible
reaction in Table 11.6 can also be obtained by simplification of the Kunii and
Levenspiel (1969) expression for the case where the mass transfer resistance
between the cloud and emulsion.is negligible relative to that at the bubble
interface. In the limiting cases where the dimensionless interphase mass
transfer group, X, approaches infinity, the results for this model approach
those for a single-phase plug flow reactor, e.g. Eq. (11.7) for a first-order
reaction.
The dimensionless groups which determine chemical conversions and
selectivities from this model are similar to, but not identical to, those which
appeared in the model considered in the previous section. The mass transfer

Let us now consider the variables <!>band 4>d, giving the fraction of bed
volume occupied by solids associated with the bubble and dense phase,
respectively. Kunii and Levenspiel (1969) defined a variable I'b = <!>bleb,
which is the volume fraction of bubbles occupied by dispersed or raining
solids, and cited experimental evidence suggesting that 0.001 ::;:;'Yb ::;:; 0.01.
Hence the lower and upper limits on <!>b
may be taken as:
O.OOleb ~ <!>b~ O.01Eb

(11.23)

The value of o/d can be approximated by including all of the solids in the
clouds, wakes, and emulsion in the dense phase and assuming the dense
phase voidage to be emf' Hence, we may write:

The actual conversion should be bounded between that predicted using the
upper and lower bounds for <!>b
given in Eq. (11.23). For slow reactions the
solids assigned to the bubble phase have little influence and the two limits are
indistinguishable. However, for fast reactions, <!>bplays a significant role.
The limits proposed for <!>b
are conservative in that all cloud and wake solids
are assigned to the dense phase, rather than to the bubble phase. In view of
this fact, the model can also be used for the so-called 'slow bubble' or
cloudless bubble case (a < 1).
The dimensionless outlet concentration for an irreversible first-order
reaction, as predicted by the model for 4>d= 0.4, is plotted in Fig. 11.5. Two

-------- --"

"

X-O1

"(,

,,

,,

gas velocity is 0.50 mls. The molecular diffusivity of the reacting species is 3.3
X 10-5 m2/s. Use both the Orcutt model and the two-phase bubbling bed
model to estimate the conversion if: (a) no baffles are present so that
unrestrained bubble growth occurs and (b) baffles are present limiting the
bubble size to 0.15 m.
Solution

We are given D = 1.30 m, H = 2.60 m, Urnf = 0.020 mis,


12.8 S-I, U = 0.50 mis, r!IJ = 3.3 X 10-5 m2/s.
Hence U - Urnf = 0.48 m/s.

Emf

0.48,k1 =

Case (a) No baffles present

Figure 11.5 Outlet concentration for a first-order chemical reaction


according to the two-phase bubbling bed reactor model with cPd = 0.4,
cPt, = 0.0 and 0.005, and different values of the interphase mass transfer
group X.

values of .<!>b, <!>b = 0 and <!>b = 0.005, are chosen to demonstrate the
influence of this variable. Note that <!>b has little effect for slow reactions or
for cases where interphase mass transfer is rapid, but that even a small
fraction of solids in the bubble phase can lead to a significant increase in
conversion for fast reactions and slow mass transfer (high ki and low X). As in
Fig. 11.4, all curves in Fig. 11.5 converge where the reaction becomes
kinetically rate controlled, i.e at small values of the dimensionless rate
constant. Conversion is again seen to depend strongly on interphase mass
transfer for intermediate and fast reactions. In general, the conversion
becomes somewhat more sensitive to the parameters X, cPt" and <Pd as the
order of reaction increases.
The following sample problem illustrates the use of the two-phase bubbling
bed model and of the Orcutt model outlined in the previous section.

The mean bubble diameter, d"q, can be obtained from the Mori and Wen
correlation (see Chapter 4), evaluated at a height of O.4H = 1.04 m above the
distributor. Assuming bubbles form at the distributor with an initial diameter
of 10 mm, we calculate d.,q = 0.300 m. The corresponding value of UA, the
absolute rise velocity of the bubbles, is found using Eq. (4.41):
UA

0.71 v(gdeq) + (U-Urnf)

1.70 mls

Hence, from Eq. (4.42):


U-Urnf
=--=-

0.48

1.70

UA

=0.282

=r6 = 20.0m

-1

eq

f3

U-U f
U rn

0.96

Example 11.3: Chemical conversion for catalytic gas phase reaction


A first-order irreversible gas phase chemical reaction is to be carried out in a
fluidized bed of diameter 1.30 m. The expanded bed height is 2.60 m. The
catalyst particles have a superficial velocity and voidage at minimum
fluidization of 0.020 mls and 0.48, respectively, and an activity (based on
particle volume) at the bed temperature of 12.8 S-I. The operating superficial

From Eq. (11.11) the interphase mass transfer coefficient is:


0.25

kq

= 0.75 X 0.020 + 0.975 x


=

0.0284 mls

O~~~O
]

(3.3

10-5).5 mls

Hence, from Eq. (11.17), the dimensionless mass transfer group is:
0.0284

= --------

x 20.0 x 0.282 x 2.60

CAin

0.96 x 0.50

-66.6

CAout

--=exp

0.867

+ 0.373) + 66.6 x 0.00028 x 0.373}}


0.509 + 66.6 x 0.373

{0.509(0.00028

= 0.596
Similarly, for 4>r, = 0.00282, we obtain:

e-X
The dimensionless

0.420

Aout
C
=
CAin

kinetic rate constant is obtained from Eq. (11.16):

12.8 x 1.87 x (1-0.48)


0.50

0.503

24.9
Case (b) With baffles present

We may now evaluate the outlet concentration


Table 11.5; Le.:
CAout

--

CAin

The estimated

of the reacting species from

1 + 0.96 x 0.420 (24.9-1)


=-------=
1 - 0.96 x 0.420 + 24.9

conversion is then simply

0.417

(l-CAoutICAin)

or 58 p~r cent.

Two-phase bubbling bed model:


The interphase

mass transfer coefficient is now calculated from Eq. (11.6):

_ 0.020
kq 3

{4

= 0.0174

10-5
7T

x 0.48 x 1.70 } 0.5

x 0.300

mls

.mls

0.0174
X

x 3.3

x 20.0 x 0.282 x 2.60

0.50

ki

12.8 x 2.60
0.50

66.6

The fraction of the bed occupied by bubble phase solids is expected to lie in
the range O.OOlEb to 0.01b; therefore:
0.00028 ~ 4>r, ~ 0.00282
while the fraction of the bed occupied by dense phase solids is:
<f>d

(1-0.282)

x (1-0.48)

0.373

The conversion
may now be calculated using the first-order
relationship in Table 11.6. For 4>r, = 0.00028, we calculate:

reaction

We are
before,
For the
and k1

given deq = 0.15 m. Hence we may proceed with the calculations as


yielding UA = 1.342 mis, Eb = 0.358, ab = 40.0 m-I, Hmf = 1.67 m.
Orcutt model, {3 = 0.96 as before, while kq = 0.0309 mis, X = 2.40,
= 22.2 are new calculated values. We proceed as before, giving
CAout/CAin = 0.123. Hence the new estimated conversion is 88 per cent. For
the two-phase bubbling bed model, k~ = 66.6 as before, while we now obtain
kq = 0.0201 mis, X = 1.50, 0.00036 ~ 4>r, ~ 0.00358, and <f>d = 0.334. We
calculate limiting values of CAoutlCAin as before, giving new lower and upper
bounds for the conversion of 76 to 81 per cent.
Note the marked increase in conversion which can occur when baffles are
introduced to keep bubbles relatively small, under conditions where
hydrodynamic conditions are rate controlling. Note also that there can be
significant differences between the predictions of alternative reactor models.
The primary reason for the difference in the present case is the considerably
larger interphase mass transfer group in the case of the Orcutt model.
The two-phase reactor models can be used to show the effect of reactor
scale on conversion. As a process is scaled up from the laboratory scale to the
industrial scale, it is a common experience that the conversion tends to
decrease. This decrease in conversion is related to the fact that bubbles tend
to grow larger in columns of larger scale (see Chapter 4). The larger bubbles,
in turn, lead to: shorter gas residence time, since Ub a d;qo.5; decreased
interphase mass transfer (X oc d;q1.5 to -1.75). An additional factor, in many
cases, is that it is difficult to achieve as uniform a distribution of incoming gas
at the grid for columns of larger scale. Separate models should be used for
reactors which are small enough for slug flow to occur, (see Section 11.14). It
is then possible to model the influence of reactor scale on conversion over the
range of conditions in which experimental data are available. As illustrated in
the worked example above, introduction of internal baffles or tubes can be
used to keep bubbles small, thereby counteracting the adverse effect of
increased reactor scale.

The expressions in Table 11.6 for consecutive first-order reactions of the


type A -'> B -'> C can be used to predict the selectivity to an intermediate
product B. Similar expressions are given in Table 11.5 for the Orcutt model.
An important feature of fluidized bed reactors is that the selectivity of an
intermediate for a given conversion of A is always less favourable for
consecutive reactions than in a single-phase isothermal reactor. This is due to
the fact that the reaction in a fluidized bed takes place primarily in the dense
phase where the local conversion is high, whereas the flow is predominantly
in the bubble phase where conversion is relatively low. This gives more
opportunity for C to form in the dense phase than in the case where the same
amount of A is converted in a homogeneous single-phase reactor, e.g. a
packed bed.

conditions can be best described by extending the two-phase models discussed


in the previous section to incorporate unsteady terms. May (1959) used a twophase model in which plug flow is assumed in the bubble phase and dispersed
plug flow is assumed in the dense phase* with crossflow (i.e. interphase mass
transfer) between the two phases. Based on data available in the literature
Mireur and Bischoff (1967) have correlated the dense phase axial dispersio~
coefficient and crossflow in terms of HID and UlUrnf, respectively.
For a transient case in which chemical reaction is ignored, the Orcutt model
wi~h perfect mixing in the dense phase, discussed in Section 11.7, may be
wntten:

11.9 BUBBUNG REGIME: AXIAL AND RADIAL GAS MIXING


For packed beds, flow in the axial direction generally does not deviate greatly
from plug flow, and deviations can be accounted for in terms of the dispersed
plug flow model. This model assumes that the mixing steps are both small in
scale and random in nature.
In gas fluidized beds, on the other hand, most of the axial mixing is due to
phenomena associated with bubbles. Some causes of the spread in gas
residence time distribution from plug flow are as follows:
(a) In general, bubble gas moves at a different velocity than interstitial gas.
At a > 1 elements associated wit4 bubbles and their clouds tend to
spend less time jn the bed than gas elements in the dense phase. For a <
1, they spend longer.
(b) Different bubbles rise at different velocities due to having different sizes,
different coalescence histories, etc.
(c) Net solids circulation patterns become established, causing gas to move
upwards in some regions more than in others. At UlUrnf ~ 1, gas in the
dense phase can actually have a downwards absolute velocity (Latham,
Hamilton, and Potter, 1968).
Although a number of workers have inferred overall axial dispersion from gas
residence time distribution data, the dispersed plug flow model is inappropriate in this case, since the bubbles which induce mixing are neither small in
scale nor random in nature. Hence, overall axial dispersion coefficients have
no physical significance, and their correlation in terms of Peclet numbers
should be avoided.
In general, measured residence time distributions tend to follow neither
plug flow nor perfect mixing. (Plug flow may, however, be approached for
very tall slugging beds or for a multi-stage unit.) The gas flow under transient

Here we have made allowance for Ep, the internal porosity of the particles,
and have assumed rapid equalization of concentration between the outer
surface and inner volume of particle, usually a good approximation for the
small particles used in fluidized bed processes. Transfer functions have been
derived for the case where f3 = 1 by Heimlich and Gruet (1966) and
Barnstone and Harriott (1967), and good matching between residence time
distribution data and the model were demonstrated for reasonable values of
the dimensionless interphase transfer parameter X for beds of 0.076 and 1.50
m diameter. This technique has been shown to be a possible means of
estimating interphase mass transfer rates by Fontaine and Harriott (1972),
who cycled the mput concentration of different tracer gases.
In terms of the two phase bubbling bed model (Section 11.8), the unsteady
state equations in the absence of chemical reaction are:

"Note that our criticism of the dispersed plug flow model above applied to the case where it is
~sed for .thebed ~s a whole, no~to cases where it is used to describe an individual phase. There
IS some Justification for expectmg small random mixing steps in the dense phase.

312

GAS FLUIDIZATION

TECHNOLOGY

This model has been used in conjunction with a step input change by Yoshida
and Kunii (1968). Solution requires numerical integration.
In principle, either of these models can b~ used for proc~ss cont~ol of
fluidized bed chemical reactors when the reactIOnterms are remserted m the
above equations from Sections 11.7 and 11.8. In practice, simpler control
models are usually adopted for this purpose. .
.
.,
Although the two-phase models provide a good basIs for treatI~g ax~al
dispersion in gas fluidized beds, there are no proven methods for dealIng wI.th
radial dispersion. Experimental measurements are scarce. When radIal
dispersion coefficients have been reported, they tend to be abo~t an.orde~ of
magnitude smaller than values which ~ave .b~en.ascribe~ to axIal dI~persIOn
coefficients. The principal cause of radIal mIXmgm bUb~lIngsystems ISlate~al
motion of bubbles assocated with coalescence of non-alIgned bubbles. SolIds
circulation patterns may also playa significant role. .
..
Both axial and radial mixing can be altered by addmg baffles to flUIdIZed
beds. In general, any baffle which impedes bubble an? solids,motion in a
given direction or which tends to make bUbb~esmore umform will reduce t~e
mixing in that direction. For example, vertIcal tubes tend to red~ce r~~Ial
dispersion, while horizontal screens or perfora~ed plates reduce ~Ial mIXm~
and cause gas residence times to be more umform. For a multI-stage umt
(stages in series), the overall gas flow pattern can be made to approach plug
flow conditions.

The models considered in the previous sections give reasonable descriptions


of the bubbling region of fluidized beds operated in the bubbling bed regime.
However, they do not give a good description of the hydrodynamics and
contacting in the entry region just above the gas distributor or grid.
The first separate model for the grid region was proposed by Behie and
Kehoe (1973). The model is shown schematically in Fig. 11.6(a). Gas entering
the column through upward facing horizontal orifices is assumed to pass
initially through a solids-free jet region (designated J). At the top of the jets,
the gas is distributed partly to the bubble phase (denoted by B) and partly to
the dense phase (designated D), according to the two-phase theory of
fluidization (see Chapter 4). Interphase mass transfer is assumed to occur
both between the jets and the dense phase and between the bubbles and
dense phase, with transfer being significantly more rapid in the former case.
Perfect mixing of gas was assumed to apply to the entire dense phase. In an
alternative formulation (Wen, 1979) (see Fig. 11.6b) the dense phase was
divided into two compartments, both well mixed, one extending up to the top
of the jets and the other corresponding to the bubbling zone.
The above model was modified by Grace and deLasa (1978) to allow for
solids entrained in the jets and to consider the case where gas in the
corresponding dense phase region is stagnant, rather than perfectly mixed.
The model is represented schematically in Fig. 11.6(c). Either of the models
considered in Section 11.7 or 11.8 may be used for the bubbling region higher
in the bed. Considering only the lower part of the bed corresponding to the
jet region, we may write separate mass balances for the two regions or phases.
For a first-order gas phase reaction under steady state isothermal conditions,
with one-dimensional flow assumed in each phase:

Dense phase: kqjaj(CAd

{oj

(bJ

(cJ

Figure 11.6 Schematic diagrams showing grid models: (a) Behie


and Kehoe (1973); (b) Wen (1979); (c) Grace and deLasa (1978).
J '= jet region; B '= bubble phase; D ~ dense phase; m = well
mixed; p '= plug flow; s '= stagnant regIOn.

CAj)

+ kI(l-j)

(l-mf)CAd

=0

(11.30)

where aj is the jet interfacial area per unit bed volume, j is the fraction of
the cross-sectional bed area occupied by jets, and 'Yj is the volume fraction of
the jets occupied by catalyst particles. The final terms must be modified for
non-first-order reactions. Integration of Eqs. (11.29) and (11.30) wit~ the
boundary conditions CAj = CAin at z = 0 yields the following concentration at
the level z = Hj, corresponding to the top of the jet or grid region:
CAHj = CAin exp

Xjki
X

+ ki -

k{ j'Yj
(l-j)(l-mf)

(11.31)

Example 11.4: Chemical reaction in the grid region

k1(1-Ej)(1-Emf)Hj
U
are analogous to X and kj or k~ in the bubbling bed models above.
Equation (11.31) can be used to estimate the concentration of the reacting
species A at the top of the grid region; the bubbling bed models outlined
earlier in this chapter can then be used to predict reaction rates in the
remainder of the bed. There are, however, several difficulties with these
models. Rowe, MacGillivray, and Cheesman (1979) studied the entry region
above single orifices for beds of fine and intermediate size particles with the
aid of X-ray photography and showed that the region directly above the
orifice is characterized by periodically forming elongated voids rather than
steady jets. This implies that models like that represented by Eqs (11.29) to
(11.31) can only represent the grid region behaviour in a time mean sense.
However, this difficulty is also present, and probably to a greater degree, for
the bubbling bed models.
A more serious problem is the estimation of parameters to use in the
model. There is a lack of data upon which to base most of these variables. Of
the various correlations for jet penetration, Hj (see Chapter 4), that of Merry
(1975) appears to have the soundest basis. To a first approximation, the
average jet diameter can be taken as V2 to 2 x dor where dor is the orifice
diameter. Hence Ejmay be estimated to lie in the range 2 to 4for and Qjin the
range 4 v2fo/dor to 8fo/don where for is the fractional free (orifice) area in
the grid plate. The volume fraction of solids entrained in the jets, 'Yj, is
expected to lie in the range 0.001 to 0.006 (Merry, 1976); except for fast
reactions, these values are generally too low to have an appreciable effect on
conversion. Of greater important is the jet to dense phase interphase mass
transfer coefficient kqj. Behie and Kehoe (1973) based some sample calculations on mass transfer data obtained in a 0.61 diameter column of 60 f.Lm
cracking catalyst (Behie, Bergougnou, and Baker, 1976). The values of kqj
were of order 1 mis, such a high value that the consequences for reactor
modelling were considerable at least for reasonably rapid kinetic rate constants
(Behie and Kehoe, 1973; Grace and deLasa, 1978). However, values of kqj are
very difficult to measure, and other workers (deMichaele, Elia, and Massimilla
1976; Sit, 1981) have inferred or measured transfer coefficients which are lower
by an order of magnitude. The following illustrative problems demonstrates
the importance of the interphase transfer step.

Estimate the conversion in the grid region of a fluidized bed reactor assuming:
(a) kqj = 1.0 mls and (b) kqj = 0.10 mls for a bed of 100 f.Lm catalyst particles of
density 2,000 kg/m3. The superficial gas velocity is 1.0 mls and the gas density is
1.0 kg/m3. The grid is a multi-orifice plate containing 12.7 mm diameter holes
on a square lattice with a hole-to-hole spacing of 0.102 m. The kinetic rate
constant is 12.8 S-l. Assume that Emf= 0.48.
Solution
The fractional free area in the grid is given by:
7T

for =

X (12.7?
X 1022

= 0.01218

The gas velocity through the orifices is therefore:


Uor

U
1.0
= for = 0.01218 mls = 82.1 mls

The height of the grid region can be estimated


correlation of Merry (1975); i.e.:

Hj

= 5.2dor

(pg~or)
Pp

0.3

/1.3 ( U~r)

0.2

1/

gdor

1.0 x 0.0127)
= 5.2 x 0.0127 x ( 2000
x 0.0001

0.3

from the jet penetration

13
.

82.12
)0.2
9.8 x 0.0127 -1

=0.303 m
Ej :::::(2 to 4) for; take Ej = 3for = 0.0365
Qj :::::(4 V2 to 8)fo/dor; take Qj = 4 v'3 forldor = 6.64 m-1
'Yj =

o.(Xn

to 0.006; take 'Yj= 0.0035

From Eq. (11.33):


j }2.8 x (1-0.0365)
k 1 --------------(a) For kqj

x (1-0.48)
1.0

x 0.303

1.0 m/s:

From Eq. (11.32):

Xj =

1.0 x 6.64 x 0.303


1.0
= 2.01

1.943

Therefore, from Eq. (11.31):


CAHj
CAin

exp

-2.01 x 1.943 1.943 x 0.0365 x 0.00351


(2.01 + 1.943) - (1-0.0365) x (1-0.48)

(b) Similarly, for kqj


Xj = 0.201

and

roasting of ores, calcination reactions, and hydrogen reduction of ores. The


equation for such a reaction may be written:

0.37

0.10 m/s:
AHj

0.83

CAin

from Eqs (11.32) and (11.31). The final result is insensitive to the choice
of 'Yj and not strongly affected by the choice of aj and Ej for the
conditions under consideration. However, there is a very strong influence
of the interphase mass transfer coefficient (the estimated conversion in
the grid region varying from 63 to 17 per cent. for a tenfold variation in
kqj). Note that these conversions can be appreciable with respect to those
in a much deeper bubbling bed (cf. Example 11.3).
The type of mass transfer envisaged in the model outlined above is probably
dominant only for relatively fine particles (for example dp < 200 /Lm). For
coarse particles (e.g. > 1 mm), the grid region, at least for upward facing
horizontal orifices, may resemble a series of parallel spouted beds (Mathur
and Epstein, 1974). The primary mass transfer is then via the bulk outflow of
gas, as predicted by Mamuro and Hattori (1968). Models and experimental
results for a catalytic first-order reaction in isothermal spouted beds are given
by Piccinini, Grace, and Mathur (1979) and Rovero et al. (1983).
Considerable work remains to be performed before reaction in the grid
region can be predicted with confidence. Even for the simplest grid geometry
of a multi-orifice plate, there is a need for definitive interphase mass transfer
results and reliable parameter estimation methods, and there is little to go on
for intermediate sizeparticles. For more complex grid geometries, that
reaction can be rapid in the grid region and that minor changes in grid
character can have significant effects (Cooke et al., 1968; Walker 1970;
Hovmand, Freedman, and Davidson, 1971; Chavarie and Grace, 1975;Bauer
and Werther, 1981).

In previous sections we have treated only gas-phase reactions catalysed by


solid particles. While many applications of fluidised beds are for reactions of
this type, there are also many applications in which both the gas and solid
take part in the reaction (see Table 11.1). Gas-solid reactions of this type
include coal conversion processes (combustion, gasification, and pyrolysis),

where b is a stoichiometric constant.


The physical properties of a solid particle undergoing reaction tend to
change with time. In some cases the particle may shrink during the course of
reaction, e.g a coal particle undergoing combustion evolves volatile matter
and then is converted into gases (CO, CO2, and H20) and ash. In other cases
one of the products of reaction is itself a porous solid, or the reaction leaves
behind a porous structure, e.g. in the calcination of limestone or in the
burning of coke deposits from porous cracking catalyst particles. Alternatively, the reaction may produce a solid product of larger volume than the
original, thereby tending to block the pores, as in the sulphation of calcined
limestone sorbent particles during coal combustion. It is clear that the rate
expressions in these different cases may differ significantly.
Kinetics of gas-solid reactions are considered in detail in a number of texts
and references (e.g. Smith, 1970; Levenspiel, 1972; Carberry, 1976). In the
general case, the following steps occur in series during reaction of a solid
particle:
(a) Mass transfer of gaseous reactant component from the bulk gas to the
outside of the solid particle
(b) Diffusion of gaseous reactant into the pores of the particle
(c) Reaction of the gaseous reactant with the solid surface at some distance
inside the particle
(d) Diffusion of gaseous products back to the outer surface of the solid
particle
(e) Mass transfer of gaseous products from the exterior surface into the bulk
gas
Not all of these steps are present in all cases. Steps (a) and (e) tend to
encounter less resistance than steps (b) and (d), respectively. However, the
relative resistance to steps (b), (c), and (d) depends on many factors including
temperature, particle size, pore structure, and the nature of the reaction,
reactants, and products. The most common reaction model for single particles
and one that is the best simple representation for most gas-solid reactions
(Wen, 1968)is the shrinking core model (also called the unreacted core model
or shell progressive model). Reaction is visualized as occurring at a sharp
front, separating unreacted material from product; this front begins at the
outside of the particle and advances inwards as the reaction proceeds. This
model adopts a pseudo steady state assumption which relies on the reaction
front migrating inwards at a rate much lower than the speed at which gas can

diffuse to the front. In most cases one of steps (b), (c), or (d) may be assumed
to be rate controlling, i.e. to provide the bottleneck which determines the
overall rate of reaction. When the chemical reaction step controls, it can be
shown (e.g. Levenspiel, 1972) that the time required for complete conversion
of a spherical particle under isothermal conditions exposed at time 0 to a gas
concentration C:'" is:
tcr

PBdp

=.

(11.35)

2bkls CAMa

Here the reaction is assumed to be first order, kls being the reaction rate
constant based on the unit surface area of unreacted core; i.e.:
rA

=-

1
4~e

dNA

--;:it = k1sCA

where rue is the radius of the unreacted core.


When the reaction is fast enough that the resistance to chemical reaction
and diffusion are of similar magnitude, it is possible (Kunii and Levenspiel,
1969) to retain Eq. (11.35), but with k1s replaced by an effective rate constant
kef. If the diffusion is through a layer of porous solid or ash material, then:
1
kef

l/k

1s

+ dpl12qj)s

(11.37)

where !:i's is the effective diffusivity of the gas through the porous solid. For
particles which shrink as the reaction proceeds so that the reaction front is
always at the exterior surface, then:
kef

1
l/k1s

+ l/kd

where kd is the mass transfer coefficient between the particle outer surface
and the bulk gas surrounding the particle.

Heterogeneous gas-solid reactions tend to be more complex than catalytic


gas-phase reactions in the following ways:
(a) The particles usually undergo significant physical changes (e.g. shrinkage
growth, density changes, shape changes) during the course of the
reaction.
(b) The kinetics of the reaction may be complex, as outlined in the previous
section. The controlling mechanism may change as the reaction proceeds
or be different for large and small particles.

(c) Energy balances are often required in addition to material balances,


especially where the reactor finds its own temperature level (as in catalyst
regenerators), where heat transfer is the rate-limiting step (as in
calcination of limestone), or where reacting particles are at a different
temperature from inactive particles (as in coal combustion).
(d) Since the solids (in addition to the gas) are involved in the reaction,
residence time distributions and mass balances are required for the
particles as well as for the gas phase components. The mass balance for
the solids must be related to that of the gas through the stoichometry of
the reaction or reactions. To account for the behaviour of different size
fractions (feed and exit streams, growth or shrinkage, elutriation,
attrition, and mixing), solids population balances are often written.
Many of the reactor models which have been applied for gas-solid reactions
have been devised for a specific reaction. Table 11.7 lists a number of these
models from the literature together with the principal assumptions. Note that
a number of areas where assumptions must be made are common with the
models for gas phase solid catalysed reactions (cf. Table 11.2). Because of the
additional areas where assumptions must be made, the number of possible
models is even larger than before. Many of the models proposed have never
been subjected to direct experimental testing.
In order to illustrate the procedure to be followed in predicting the
conversion for a continuous gas-solid reaction, we will assume that the
reaction follows the shrinking core model and that the reactor is isothermal.
For the time being, we further adopt a hydrodynamic model which parallels
that given in Section 11.7 for a gas phase reaction, i.e. we assume no particles
in the bubble phase, plug flow of gas in the bubble phase, and perfect mixing
of gas in the dense phase. In addition, the solids are assumed to be perfectly
mixed in a single-stage system. The procedure then involves a sequence of
steps:
1. Given the bed diameter, particle and gas properties, bed depth and gas
properties, work out the required hydrodynamic features. In particular,
evaluate the bubble diameter, bubble velocity, and fraction of bed volume
occupied by bubbles (deq, Ub, a~d Eb) at a representative height, generally
z = OAB. Hence find ab = 6ldeq, the bubble area per unit volume, and
Hmf. Note that the particles will generally be nearly completely reacted so
that the particle properties should be based on the solid product, not on
the feed material.
2. Calculate the interphase mass transfer coefficient kq from Eq. (11.6) or
Eq. (11.11).
3. Calculate the dimensionless groups f3 and C from Eqs (11.10) and (11.17),

SC,CR

3:8,C,

General

As in their catalytic model

SC

2:8 and D

and

Penetration + leakage of
circulating gas

Coal combustion. monodispcrsc and


wide dl' di!itribution: some allowance
for volatiles

As Davidson

Coal combustion
Roasting of ZnS: allow for particles in
bubbles; compare with experimenfs

and Harrison

(1963)

Carbon combustion; consider multi


plicity of steady states
General model: bubbles grow with height

I1Jd,;'l

~~~~a~~~~:~~
~a~~:a":ta(~71)
As Kunii and levenspiel

b
C

CR

Coal combustion and SO~ - limestone


n.: bubbles grow with height
Coal combustion and sulphlltion of
additives: bubbles gmw with heigbt;
compare with pilot plant data
Coal combustion: bubbles grow with
height: compare with data
Coal combustion including devolatilizution Ilnd S removal; bubbles grow
with height; compare experimental
dala
Coal burning; general comparison with
literature data
Catalysl regenerator; separate allowance
for grid region; some experimental
data
Shallow bed: general model applied to
zinc roasting; sleudy and unstcady
solutions

(1969)

Dispersion
model
(lateral)

2:B and 0

= continuous reaction modeli Diff . control by diffusion through. ash layer; SC = shrinking core model; 0 = order. rxs = reactions.

B _ bubble phase; C ,. cloud phase; D - dense phase; E


emulsion; I =- inlentitiaJ gas; J
jet phase; P - panides.
C in S :;; wellmixed compartments
in series; PF:;; plug tklw; PM
perfect mixing; PB
population balance; A = attrition
aUowance; St
stagnant; NS
not specified.

various

As Davidson and Harrison (1963)

11Id"

II

consider

cases

General model; tcst for ZnS rousting:


hubbies grow with height; allow fnr
solids in hubbies; single. consecutive.
and paralic I reactions
Coal combustion; few active particles
among inscns
Coal dcvolalization
and comhustion;
few active paTliclcs among incrts

I lid,.

As Hovmand, Freedman.
Davidson (1911)

1st or Oth
order with
respect to
solid

model;

limiting

and E

allowance:

= entrainment

\)

II

~
0Il

If
~~I

....

:..: I
+\:i,)

rol
'-""

Example 11.5: Zinc Roaster


Operating conditions for a zinc roaster operated by Canadian Electrolytic
Zinc Limited at VaIIeyfield, Quebec, have been reported by Avedesian (1974)
to be as follows:
Bed diameter: 6.38 m
Operating pressure: 101 kN/m2
Operating temperature: 1,233 K
Air superficial velocity: 0.78 rn/s
Minimum fluidizing velocity: 4.8 mm/s
Bed weight: - 30,000 kg (estimated from the bed pressure drop quoted by
Avedesian as 10 kN/m2)
Solids feed rate: 2.48 kg/s (dry basis)
Certain hydrodynamic features of this roaster have already been considered
in Examples 4.3 and 4.4. For the illustrative purposes of this worked example,
we assume uniform solids of diameter 60 /Lm and a mean bubble size of 0.1
m. The reaction is known to proceed according to the shrinking core model.
According to Fukunada et al. (1976), the surface chemical rate constant is
given by an Arrhenius equation:
kls

= 2.96

13

x 10. exp

-3.14 x 1<r }

RT

0.21 x 1.013 x 105


8.314 x 1,233

0.21P

CAin = ~

Now proceed through th,e steps given above:


1.

UA

= 0.71

ab

=--

x 0.10) + 0.78-0.0048

\/(9.8

0.10

m/s

1.479 m/s

60 m-1

0.78-0.0048
1.479

0.524 (from Eq. 4.42); in practice we expect

never to exceed about 0.40. Therefore,

assume

Eb =

0.40.

We expect the solids in the bed to be highly reacted. Hence their density
will be:
Ps

4,100 x 0.0814
3
0.0975
kg/m

rn/s

The effective diffusivity of O2 and S02 through the solid ash may be taken as
9.0 X 10-6 m2/s. The density of the concentrate is 4,100 kg/m3. Estimate the
conversion of both gas and solids assuming that the bed operates entirely in
the bubble regime with a mean bubble diameter of 0.10 m.

Hmf

30,000
3,420 x ('lT14) X 6.382 x 0.5 m

0.548
1-0.40

0.548 m

0.915 m

Product solids f10wrate from the bed F1

2.48 x 0.0814
0.0975
kg/s

Solution
We assume Emf = 0.50 and estimate a molecular diffusivity of oxygen in air
at 1,233 K of 2.5 x 10-4 m2/s. The reaction may be written in the form of Eq.
(11.34) as:
02(g)

+ 3 ZnS(s) - 3 ZnO(s) + 3S02(g)

so that b = 2/3. The molecular weights of ZnS and ZnO are 0.0975 and 0.0814
kg/mol, respectively.

Mean solids residence time

30,000
2.07

14,500 s

At 1,233 K:
k1s

= 2.96

13

x 10

exp

-3.14 x 1<r)
8.314 x 1,233

rn/s

kq

= 0.3

x 0.0048

4 x 2.5 x 10-4 x 0.50 x 1.479)


'IT x 0.10

1/2

rn/s - 0.050 rn/s

{3

1 - 0.0048
0.78

(11.30); 4Jt,and 4Jdare estimated as outlined in Section 11.8. For step 4, the
gas conversion XA is now found as a function of k~ from Table 11.6. The
mean concentration of gas to which the solids are exposed CA, to be
calculated in step 5, is found by integration. For a first-order irreversible
reaction, it can be shown that:

0.994

From Eq. (11.17)


X

0.050 x 60 x 0.40 x 0.915


0.994 x 0.78
0.78

1.415

C~

4Jt,C + 4JdCAd
CAinXA
= ----4Jt,+ 4Jd
(4Jt,+ 4Jd)k ~

= -----~.Ab

X A = 1 - exp

0.758 + ki

_ 2.075 x 0.758
3
CAd- 0.758 + ki mol/m
ef

1/1.47 + 60 x 10-6/(12 x 9 x 10-6)


4,100 x 0.OO60

tcr

2x

2f3

x 0.809
tcr

XB

k~ {X (4Jt,+ 4Jd~ + k~ 4Jt,4Jd}


X + k 1 4Jd

= 0.809 mls

2.34
=--

O.0975CAd CAd
ter

= 1 - 4 x 14,500 = 1 - 58,000

The model developed above and the example are for the case of uniform
solids of unchanging particle size. In most fluidized bed processes, particles
fed to the bed have a wide distribution of sizes, and fines are elutriated from
the bed. To take these factors into account for particles of unchanging size,
we make the following changes in the procedures given above for calculating
the solids conversion:
1. The feed particle size distribution is represented by N size fractions of
representive size dpi and mass fraction Xi (generally determined by sieve
analysis).
2. Calculate the time for complete reaction of each size fraction (ter)i from
Eq. (11.35) or another appropriate kinetic model with dp replaced by dpi'
3. Determine the mean residence time of particles in this size range from the
equation:
M

7. From Eq. (11.42):


2.48
XA

All other steps are identical to those outlined above.

0.758k1

= 0.0975 2/3

= munder+ (1-1/i)

XB

x 0.78 x 7Ti4 x 6.382 x 2.075

0.737XB

After trial and error, ki = 26; CAd = 0.0588 mollm3; tcr = 39.8s; XA =
0.737; XB = 0.9993,
Hence the solids are more than 99.9 per cent. converted while the oxygen
is nearly 74 per cent. converted according to the model.
To replace the Orcutt model by the two-phase bubbling bed reactor model
(outlined in Section 11.8 for a gas phase reaction), the steps are essentially the
same. However, in step 3, {3is set equal to unity; X is calculated from Eq.

K~A

where M is the mass of solids in the bed, munderis the underflow or


overflow mass withdrawal rate of product; K~ is the elutriation rate
constant, and 1/iis the efficiency of any cyclones used to recycle solids to
the bed. Both K~ and 1/ican be estimated from standard correlations.
4. The mean conversion of particle of size dpi can be calculated from:
XBi

1 -"

"

(Ie,)'

(1 - XBi)E(dpj, t)dt

(11.46)

where XBi> the conversion of particles of size dpi for a residence time t, is
obtained from standard kinetic models and the exit age distribution is
generally assumed to correspond to perfect mixing of each particle size
fraction so that:

E(dpi,t)

exp{ -(tIT;)}

= -----

T;

Hence, for example, the shrinking core model with chemical reaction
control XBi is given by Eq. (11.41) with T and tcr replaced by and (tcr);,
respectively.
5. Finally, to calculate the overall conversion, we must sum over the different
size fractions fed to the bed:
Tj

It is possible to extend the above procedure to cover more complex cases, e.g.
multi-stage beds, beds in series, or other cases where the solids residence time
distribution does not correspond to perfect mixing. For particles which shrink
or grow as a result of chemical reaction, attrition, or agglomeration,
population balances are required to account for the different size fractions. A
number of these cases have been considered by Kunii and Levenspiel (1969)
and in the models listed in Table 11.7.
As demonstrated by the models outlined above, modelling of the
disappearence of gaseous reactants follows the same general lines as twophase modelling of catalytic gas phase reactions. However, it is common for
the gaseous reagents (e.g. oxygen in the case of coal combustion) to be
present well in excess of the stoichiometric requirements. In this case, the
residence time distribution of the solids, including losses by elutriation, plays
a critical role in determining the overall conversion of solids. Where high
conversions of solids are required, it is often desirable to place two or more
units in series and to provide better solids recovery than is possible with a
single cyclone.

The freeboard region can playa significant role in determining the overall
performance of fluidized bed reactors. For example, nitric oxide levels in coal
combustors change rapidly in the freeboard region (Gibbs, Perlira, and Beer,
1975), while the temperature level, coke conversion, and afterburning in
cracking catalyst regenerators depend strongly on the freeboard (deLasa and
Grace, 1979).
The contribution of the freeboard arises from the following considerations:
(a) Additional contacting of gas and solids can take place in the freeboard
region. Although the hydrodynamics of the region are not well
understood, contacting appears to be favourable, with little or no
tendency of the gas to separate into two phases (as in the bubbling bed)
and near plug flow for the gas. This effective contacting and lack of

backmixing, coupled with the fact that entrained particles tend to be finer
than those in the bed, make up to some extent for the low hold-up of
solids in the region above the dense bed surface.
(b) Whereas temperatures tend to be highly uniform in the dense bed,
significant temperature gradients are possible in the freeboard region.
This may lead to increased reaction rates, unwanted side reactions, or
recycling of additional heat to the bed itself as solids return from cyclones
or by gravity.
(c) Where heat transfer surfaces extend into the freeboard region, additional
heat transfer occurs. Heat transfer coefficients are nearly as favourable
near the bed surface as for immersed tubes, but the coefficients fall off
rapidly with increasing height. Radiation losses from particles in the
freeboard can also be substantial.
The contribution of the freeboard region will depend on a number of factors:
(a) The extent of reaction which has already occurred in the dense bed. (It is
evident that if a reactant has already been used up before reaching the
bed surface, no further reaction involving consumption of that reactant
can take place in the freebaord.)
(b) The height and geometry of the freeboard region, including expanded
areas and fixed baffles or tubes, if any.
(c) The fines content of the powder.
(d) The efficiency of the cyclone or other solid recovery system in returning
entrained fines to the bed where they can be reentrained.
(e) The operating conditions, principally the superficial gas velocity,
pressure, and temperature.
To a lesser extent, reaction in the freeboard may also depend on the depth of
the dense bed and the presence of immersed tubes through their influence on
the bursting pattern of voids at the bed surface and hence on the ejection of
solids into the freeboard.
The freeboard itself may be divided into different regions, as shown in Fig.
11.7:
(a) Splash zone: this is immediately above the dense bed surface.
(b) Disengagement zone: This extends up to transport disengaging height
level; it may be taken to include the splash zone since there is no distinct
boundary between the splash zone and the disengagement zone.
(c) Region beyond TDH. In regions (a) and (b), particles travel both
upwards and downwards, amid periodic bursts (or splashes) of particles.
In zone (a), on the other hand, particle motion is almost solely upwards

first order gas phase solid catalysed reaction, their model leads to the
following expression:
CAz
CAH

/-gf3W(QB/A)

--=exp

(I-Emf) (z-H)
+ dpl6kd)

V(V-lit) (1Ik[

where f3w is the wake volume/bubble volume for bubbles reaching the bed
surface. The volumetric flux of bubbles, QB/A, can be estimated as detailed in
Chapter 4, while g, the fraction of wake particles ejected into the freeboard
as bubbles burst at the bed surface, varies with deq, but is typically of order
0.4 (George and Grace, 1978). The mass transfer coefficient kd can be
obtained for steady flow of gas past the particle at a relative velocity of VI as
detailed by Clift, Grace, and Weber (1978). Sample calculations performed
by Yates and Rowe (1977) show that the freeboard region can be more
effective than an equal height of dense bed.
Example 11.6

Figure 11.7 Freeboard region of fluidized bed and its zones.


Other regions are also indicated. Typical time-mean Yoidages
are shown for each region.

and the distribution of particles is much more uniform, although


streamers or clusters can often be identified. For relatively large particles
without fines, essentially all of the ejected particles return to the dense
bed surface, and region (c) may be neglected. Region (c) is also absent if
the column is successfully designed to have the freeboard height equal to
the TDH.
It is clear from the discussion above that the behaviour in the freeboard
region is complex and depends on many different factors. The general
inadequacy of correlations for entrainment and TOH (see Chapter 6)
complicates the treatment of the freeboard region still further.
Yates and Rowe (1977) adopted a simple model in which gas and solids
were assumed to remain isothermal. Radial gradients were ignored. The slip
velocity between solids and gas was assumed to be V-Vt throughout the
entire freeboard. Two extreme cases, plug flow and perfect mixing, were
assumed for the gas mixing. For the more realistic case of plug flow and for a

Narrow size fraction 60 ILm spherical catalyst particles are being used in a
fluidized bed reactor. The particle density is 2,650 kg/m3 and the catalyst
activity is 10.0 s-\ the reaction being first order. The fluidizing gas has the
properties of air at 20C and 101kPa. The molecular diffusivity of the reacting
gas in air is 2.0 x 10-5 m2/s. Estimate the fraction of the gas leaving the
expanded bed surface which is converted in the freeboard region if the total
height of this region is 4.0 m. The superficial gas velocity is 1.0 mls. Assume
isothermal behaviour.
.
Solution

From Clift, Grace, and Weber (1978), Vt = 0.26 mls and kd = 0.73 mls.
Estimates for Vmfand Emf are 0.004 m/s and 0.50, respectively. From George
and Grace (1978), we estimate g = 0.4. A reasonable estimate for f3w is
0.33. The visible bubble flowrate may, to a first approximation, be estimated
from the two-phase theory so that:
QB
11
=
Hence, Eq. (11.49) yields:
CAz
CAH = exp

1.0 - 0.004 = 0.996 m/s

-0.4 x 0.33 x 0.996 x 0.5 x 4.0


1.0 x (1.0 - 0.26) x {1/10.0 + 0.000060/(6 x 0.73)}

= 0.029

Hence, the model indicates that only about 3 per cent. of the gas which is
unconverted leaving the surface of the dense bed is still unconverted at the

top of the freeboard region. This illustrates the prominent influence which the
freeboard can exert.
Generally speaking, it is necessary to allow for temperature gradients in the
freeboard region by writing a differential energy balance. For plug flow of
gas, this may be written:
dT
CgPg dz

hP(T-Tw)
VA

dXA

+ !i.HrCAin & =

where the particles and gas in the freeboard are assumed to have the same
temperature
T locally (usually a reasonable approximation), while h is the
freeboard suspension-to-wall heat transfer coefficient, P the column perimeter, Tw the column wall temperature, and !i.Hr the heat of reaction. The
middle term can be neglected for well-insulated systems. Allowance may be
made for radiation losses or cooling tubes in the freeboard by incorporating
extra terms.
Miyauchi and Furusaki (1974) used an equation like Eq. (11.50) to show
the effect of the freeboard on selectivity. The freeboard could be beneficial or
harmful to the yield of a desired product, depending on the mechanism of
reaction and whether the reaction is exothermic or endothermic. For
example, for reactions of the type A _ B _ C where B is the desired product,
the freeboard region is helpful for endothermic reactions, but not for
exothermic reactions.
DeLasa and Grace (1979) treated the freeboard region of a gas-solid
heterogeneous
reaction, the burning of coke deposits from deactivated
cracking catalyst particles in a regenerator. Separate allowance was made for
upward and downward moving solids. Material balances were written for
both the solid reactant (coke) and the gaseous reactant (oxygen); energy
balances were used in addition. The hydrodynamics of particle ejection and
trajectories in the freeboard were taken from George and Grace (1978).
Numerical integration was required to solve the equations for different
monodisperse
particle sizes and conditions matching those in an actual
industrial regenerator. The freeboard was shown to have a profound effect for
shallow beds, especially for particles whose terminal velocities are close to the
superficial gas velocity. The model has been extended to describe non-steady
state conditions by deLasa and Errazu (1980).
Chen and Wen (1982) modified the above approach by assuming axial
dispersion of gas in the freeboard region and by assuming an exponential
decay of solids concentration above the bed surface. The model was shown to
be in favourable agreement with experimental results obtained in a 1.8 m
square fluidized combustor. Fee et ai. (1982) considered the part of the
freeboard in which particles are ejected and fall back to the bed surface as a
continuously stirred tank reactor. Significant sulphur capture was shown to
occur in the freeboard of a fluidized bed combustor.

11.14

REACTORS OPERATING IN THE SLUG FLOW REGIME

Two phase models similar to those described earlier for the bubbling regime
have also been written for fluidized beds operating in the slug flow regime.
The best known slug flow reactor models are those of Hovmand et al (1971)
and Raghuraman and Potter (1978). Yates and Gregoire (1980) found that
the latter gave better agreement with experimental conversions obtained in a
0.10 m column. It is essential to allow for favourable gas-solid contacting in
the region below the level where slug flow is reached (typically z=2D) in
applying slug flow reactor models.

11.15

REACTORS OPERATING IN THE TURBULENT FLUIDIZATION

REm~

In industrial practice, many fluidized bed reactors operate in the turbulent


fluidi~ation. regime. This is especially the case for solid catalysed gas phase
reactIons SInce most catalysts fall within the group A category where
superficial gas velocities of the order of 0.3 m/s are reportedly (Avidan, 1982)
sufficient to initiate this regime in reactors of commercial size. In smaller
sc~le reactors, higher gas velocities are needed for this regime. Perhaps for
thIS reason, there appear to be almost no data available in the open literature
showing reaction rates for fluidized beds operated in the turbulent regime.
The appearance of turbulent fluidized beds (see Chapter 7), with tongues
of gas and solids darting backwards and forwards, suggest that this regime
should give rise to effective gas-solid contacting. Van Swaaij (1978) and Wen
(1979) suggested that beds operated in this regime might be modelled
according to a single-phase plug flow model (Section 11.5). On the other
hand, Avidan (1982) suggests that appreciable axial gas mixing occurs and
that an axially dispersed single phase model should be appropriate. Since
many of the experimental results correlated by Van Deemter (1980) were for
reactors, large and small, operated within this regime, his summary of results
can provide estimates for the effective axial dispersion coefficients needed for
this simple model. Experimental data are required to verify that there is
negligible interphase mass transfer resistance.

11.16

REACTORS OPERATING IN THE FAST FLUIDIZATION

REGIME

There has been considerable interest in recent years in circulating beds and in
riser reactors for both gas-solid reactions and solid catalysed reactions.
Operation within the fast fluidization regime (see Chapter 7) confers a
number of advantages including effective gas-solid contacting, increased gas
throughput,
ability to treat cohesive and low density materials (e.g.
peat,wood waste, caking coals), good radial mixing, improved turndown,

temperature
uniformity throughout the entire reactor volume, and (for
co~b.usti()n reactions) reduced NOx emissions. because of secondary air
addltlons. Erosion, attrition and particulate emissions can all be controlled
within reasonable limits by good design.
Wen (1979) suggested tentatively that a single-phase plug flow model
(sectio~ 11.5) mi~ht be appropriate for this regime in view of the effective
gas-solId contactmg. On the other hand, there is strong evidence of phase
segregation into either a Core of dilute phase moving upwards at high velocity
surrounded by a thin annular region of predominantly downward dense
part~cle motion (Bieri, Gajdos, and McIver 1980) or of the occurrence of
partIcle clusters (Yerushalmi, Cankurt, Geldart, and Liss, 1978). In either
case, there may well be a need for a two-phase model. The two-region model
of Van D~emter (1961), which allows for a dilute upflow and a dense region,
may provide a good basis for such models. However, additional work on the
hydrodynamics of the fast fluidization regime and reaction studies on beds
operated within this regime are required to allow the parameters needed for
the ~odel to be specified. Energy balances can be incorporated when
r~qUlre?, as for other regimes. For modelling of gas-solid reactions in
cIrculating beds, it would appear to be appropriate to assume that the solids
are perfectly mixed.

C/'
flis
E(dpj, t)
for

FA ,Fa
Fin

g
G
h
H
Hj
Hmf

k{
kls

column diameter
molecular diffusivity of component in gas stream
effective diffusivity of gaseous component through porous solid
exit age distribution function
fractional free area in the grid
functions defined in Table 11.6
solids mass feedrate
acceleration of gravity
function defined in Table 11.6
bed-to-wall or suspension-to-wall heat transfer coefficient
bed height
height of jet penetration
bed height at minimum fluidization
dimensionless first-order rate constant defined in Eq. (11.33)
first-order reaction rate constant based on unit surface area as
defined by Eq. (11.36)
first-order rate constants for disappearance
in Table 11.5

11.16
ab
aj

A
b
CA,Ca

C~
CAb,Cb
CAd, Cd
CAN
CAin, Cin

CAj
CAout

NOMENCLATURE

dimensionless rate constants, kAHmf (l-Emf)/U


and kBHmf
(l-Emf)/U
dimensionless rate constants, kAH/U and kBH/U
gas-to-particle mass transfer coefficient
effective kinetic rate constant given by Eq. (11.37) or
Eq. (11.38)

interfacial bubble area per unit bubble volume


interfacial jet area per unit bed volume
cross-sectional area of column
stoichometric
concentration

constant defined in Eq. (11.34)


of A and B

gas phase concentration to which a reacting particle is exposed


concentration of A in the bubble phase
concentration of A in the dense phase
average concentration of A at z=H
inlet concentration of A

Cg

concentration of A in the jet phase


outlet concentration of A
specific heat of gas

deq
dor
dp

diameter of equivalent volume sphere


diameter of grid orifice
particle diameter

kn
k~, k~

kq
kqj

Ke
K*
MB
n

bubble diameter

of A and B, defined

NA
p
QB

nth-order kinetic rate constant


dimensionless kinetic rate constants (Eqs. 11.16 and 11.22)
bubble to dense phase mass transfer coefficient
jet to dense phase mass transfer coefficient
equilibrium constant defined in Table 11.5
elutriation rate constant
molecular weight of B
reaction order
number of moles of reactant A
column perimeter
volumetric flowrate of void units
time rate of disappearance of reactant A or B
radius of unreacted core of reacting particle

334

8.314 J/mol K

gas constant

time
time for complete reaction of a solid particle
absolute temperature
wall temperature

U
Umf

true velocity of bubbles relative to wall in a freely


bubbling bed
rise velocity of single isolated bubble
gas velocity through grid orifice
superficial gas velocity
minimum fluidization velocity
terminal settling
bubble volume
cloud volume
particle volume
mass fraction of
number of mass
number of mass

velocity of a particle

particles
transfer units (Eq. 11.7 or Eq. 11.20)
transfer units in grid region (Eq. 11.32)

vertical position coordinate measured upwards from


the grid
ratio of bubble velocity to interstitial gas velocity (Eq. 11.2)
ratio of gas flow via bubble phase to total gas flow
wake volume per unit bubble volume
volume fraction of bubbles occupied by dispersed particles
(= 4>t,!Eb)
'Yj

/;J.Hr
Eb
Ej
Emf

Ep

T/

PB

volume fraction of jets occupied by dispersed particles


heat of reaction
fraction of bed volume occupied by bubbles
fraction of grid entry region occupied by jets
bed voidage at minimum fluidization
internal particle porosity
cyclone separation efficiency
density of solid reactant B
solids mean residence time
fraction of bed volume occupied by solids associated with the
bubble phase
fraction of bed volume occupied by solids associated with the
dense phase

gas phase conversion (= 1-CAou/CAin)


solid phase conversion

Anwer, J., and, Pyle, D.L. (1974). 'Gas motion around bubbles in fluidized beds. La
Fluidization et ses Applications, Soc. Chim. Ind .. Paris. pp. 240-253.
Avedesian, M.M. (1974). 'Roasting zinc concentrate in fluidized beds - practice and
principles', 24th CSChE Meeting, Ottawa.
Avedesian, M.M., and Davidson, J.F. (1973). 'Combustion of carbon particles in a
fluidized bed', Trans. lnstn Chem. Engrs, .51, 121-131.
Avidan, A. (1982). 'Turbulent fluid bed reactors using fine powder catalyst,
AIChE-CIESC Mtd., Beijing, Sept.
Bamstone, L.A., and Harriott, P., (1967), Frequency response of gas mixing in a
fluidized bed reactor, AIChE J. 13,465-475.
Baron, R.E., Hodges, J.L., and Sarofim, AF. (1978). Mathematical model for
predicting efficiency of fluidized bed steam generators', AI.Ch.E. Symp. Ser., 74
(176), 120-125.
Basu, P., Broughton, J., and Elliott, D.E. (1975). 'Combustion of single coal particles
in fluidized beds', inst. Fuel Symp. Ser., A3 (1), 1-10.
Bauer, E., and Werther, J. (1981). 'Scale-up of fluid bed reactors with respect to size
and gas distributor design - measurements and model calculations', Proc. 2nd
World Congress of Chem. Engng. 3,69-72.
Becker, H.A., Beer, J.M., and Gibbs, B.M. (1975). 'A model for fluidized bed
combustion of coal', Inst. Fuel Symp. Ser., Al(I).
Behie, L.A., Bergougnou, M.A., and Baker, C.G.J. (1976). 'Mass transfer from a
grid jet in a large gas-fluidized bed', in Fluidization Technology (Ed. D.L.
Keairns), Vol. 1, Hemisphere, Washington, pp. 261-278.
Behie, L.A, and Kehoe, P. (1973). 'The grid region in a fluidized bed reactor',
A.I.Ch.E. J., 19, 1070-1072.
Bieri, T.W. Gajdos, L.J., McIver, A.E., and McGovern, J.J. (1980). DOE Rep. FE2449-11.
Bukur, D., Caram, H.S., and Amundson, N.R. (1977). 'Some model studies of
fluidized bed reactors', in Chemical Reactor Theory: A Review, (Ed. L. Lapidus
and N.R. Admundson), Prentice-Hall, Englewood Cliffs, NJ, pp. 686-757.
Bywater. R.J. (1978). 'Fluidized bed catalytic reactor according to a statistical fluid
mechanic model', A.I.Ch.E. Symp. Ser., 74 (176),126-133.
Carberry, J.J. (1964). 'Designing laboratory catalytic reactors', Ind. Engng. Chem., S6
(11), 39-46.
Carberry, J.J. (1976). Chemical and Catalytic Reaction Engineering, McGraw-Hill,
New York.
Chang, C.C., Rong, S.X., and Fan, L.T. (1982). 'Modelling of shallow fluidized bed
reactors', Can. J. Chem. Eng., 60, 781-790.
Chavarie, C., and Grace, J.R. (1975). 'Performance analysis of a fluidized bed
reactor' ,Ind. Eng. Chern. Fund., 14,75-91.

Chavarie, c., and Grace, J.R. (1976). 'Interphase mass transfer in a gas-fluidized
bed', Chem. Eng. Sci., 31, 741-749.
Chen, L.H., and Wen, C.Y. (1982). 'Model of solid gas reaction phenomena in the
fluidized bed freeboard'. A.l.Ch.E.J., 28, 1019-1027.
Chen, T.P., and Saxena, S.c. (1977). 'Mathematical modelling of coal combustion in
fluidized beds with sulphur emission control by limestone and dolomite', Fuel, 56,
401-413.
Clift, R., Grace, J.R., and Weber, M.E. (1978). Bubbles, Drops and Particles,
Academic Press, New York.
Cooke, MJ., Harris, W., Highley, J., and Williams, D.F. (1968). 'Kinetics of oxygen
consumption in fluidized bed carbonizers', Inst, Chem. Engrs Symp. Ser., 30,
21-27.
.
Darton, R.c. (1979). 'A bubble growth theory of fluidized bed reactors', Trans. lnstn
Chern. Engrs, 57, 134-138.
Davidson, J.F., and Harrison, D. (1963). Fluidized Particles, Cambridge University
Press, Cambridge.
deLasa, H.I., and Errazu, A. (1980). In Fluidization, Eds. J.R. Grace and J.M.
Matsen, Plenum, New York, pp. 563-570.
deLasa, H.I., and Grace, J.R. (1979). 'The influence of the freeboard region in a
fluidized bed catalytic cracking regenerator', Al.Ch.. J., 25, 984-991.
deMichaele, G., Elia, A., and Massimilla, L. (1976). 'The interaction between jets
and fluidized beds', Ing. Chirn. Ita!., 12, 155-162.
Errazu, A.F., deLasa, H.I., and Sarti, F. (1979). 'A fluidized bed catalytic cracking
regenerator model: grid effects', Can. J. Chern. Eng., 57, 191-197.
Fan, L.T., Fan, L.S., and Miyanami, K. (1977). 'Reactant dynamics of catalytic
fluidized bed reactors characterized by a transient axial dispersion model with
varying physical quantities', Proc. Pachec Conference, pp. 1379-1388.
Fee, D.C., Myles, K.M., Marroquin, G., and Fan L-S., (1982). 'An analytical model
for freeboard and in-bed limestone sulfation in fluidized-bed coal combustors',
Proc. 7th Int. Fluidized Bed Combustion Conference, 1121-1126.
Fontaine, R.W., and Harriott, P. (1972). 'The effect of molecular diffusivityon mixing
in fluidized beds', Chem. Eng. Sci., 27,2189-2197.
Fryer, C., and Potter, O.E. (1972a). 'Countercurrent backmixing model for fluidized
bed catalytic reactors; applicability of simplified solutions', Ind. Eng. Chem.
Fund., 11,338-344.
Fryer, C., and Potter, O.E. (1972b). 'Bubble size variation in two-phase models of
fluidized bed reactors', Powder Techno!., 6,317-322.
Fryer, C., and Potter, O.E. (1976), 'Experimental investigation of models for
fluidized bed catalytic reactors', A.l.Ch.E. J. 22,38-47.
Fukunda, Y., Monta, T., Asaki, Z., and Kondo, Y. (1976). 'Oxidation of zinc
sulphide in a fluidized bed', Meta!. Trans., 7B, 307-314.
George, S.E., and Grace J.R. (1978). 'Entrainment of particles from aggregative
fluidized beds, A.I.Ch.E. Symp. Ser., 74 (176), 67-74.
Gibbs, B.M. (1975). A mechanistic model for predicting the performance of a
fluidized bed coal combustor, Inst.Fuel Symp. Ser., (1) A5.
Gibbs, B.M., Pereira, F.J., and Beer, J.M. (1975). 'Coal combustion and NO
formation in an experimental fluidized bed', Inst. Fuel Syrnp. Ser., (1) D6.
Gordon, A.L., and Amundson, N.R. (1976). 'Modeling of fluidized bed reactors:
combustion of carbon particles', Chern. Eng. Sci., 31, 1163-1178.
Grace, J.R. (1971). 'An evaluation of models for fluidized bed reactors', Al.Ch.E.
Symp. Ser., 67 (116), 159-167.

Grace, J.R. (1984). 'Generalized models for isothermal fluidized bed reactors', in
Recent Advances in the Engineering Analysis of Chernically Reacting Systems (Ed.
L.K. Doraiswamy), Wiley, New Delhi.
Grace, J.R., and deLasa, H.I. (1978). Reaction near the grid in fluidized beds
A.I.Ch.E. J., 24,364-366.
Heimlich, B.N., and Gruet, I.C. (1966). 'Transfer function analysis of fluidized bed
residence time distribution data', Chern. Eng. Prog. Symp. Ser., 62 (67), 28-34.
Horio, M., and Wen, c.Y. (1976). 'Analysis of fluidized-bed combustion of coal with
limestone injection', in Fluidization Technology (Ed. D.L. Keaims), Vol. 2,
Hemisphere, Washington, pp. 289-320.
Horio, M., and Wen, C.Y. (1978). 'Simulation of fluidized bed combustors:
Combustion efficiency and temperature profile', A.I. Ch.E. Syrnp. Ser., 74 (176),
101-11I.
Hovmand, S., and Davidson, J .F. (1971). 'Pilot plant and laboratory scale fluidized
reactors at high gas velocities; the relevance of slug flow', Chapter 5 in Fluidization
(Eds J.F. Davidson and D. Harrison), Academic Press, London, pp. 193-259.
Hovmand, S., Freedman, W., and Davidson, J.F. (1971). 'Chemical reaction in a
pilot-scale fluidized bed', Trans. Instn Chern. Engrs, 49, 149-162.
Jackson, R. (1971). 'Fluid mechanical theory', Chapter 3 in Fluidization (Eds. J.F.
Davidson and D. Harrison, Academic Press, London.
Kato, K., and Wen, C.Y. (1969). 'Bubble assemblage model for fluidized bed catalytic
reactors', Chern. Eng. Sci., 24, 1351-1369.
Kunii, D., and Levenspiel, O. (1969). Fluidization Engineering, Wiley, New York.
Latham, R., Hamilton, c., and Potter, O.E. (1968). 'Backmixing and chemical
reaction in fluidized beds', Brit. Chem. Eng., 13, 66fHJ71.
Levenspiel, O. (1972). ChemicalReaction Engineering, 2nd ed., Wiley, New York.
Lewis, W.K., Gilliland, E.R., and Glass, W. (1959). 'Solid catalysed reaction in a
fluidized bed', A.I.Ch.E. J., 5,419-426.
Lignola. P.G., Donsi, G., and Massimilla, L. (1983). 'Mass spectrometric measurements of gas composition profiles associated with bubbles in a two-dimensional
bed', A.I.Ch.E. Symp. Ser., 79 (222),19-26.
Mamuro, T., and Hattori, H. (1968). Flow pattern of fluid in spouted beds, J. Chem.
Engng Japan, 1, 1-5.
Mathur, K.B., and Epstein, N. (1974). Spouted Beds, Academic Press, New York.
May, W.G. (1959). 'Fluidized-bed reactor studies', Chem. Engng. Prog., 55 (12),
49-56.
Merry, J.M.D. (1975). 'Penetration of vertical gas jets into fluidized beds', A.I.Ch.E.
J., 21, 507-510.
.Merry, J.M.D. (1976). 'Fluid and particle entrainment into vertical jets in fluidized
beds', AI.Ch.E. J., 22,315-323.
Mireur, J.f., and Bischoff, K.B. (1967). 'Mixing and contacting models for fluidized
beds', A.I.Ch.E.J., 13,839-845.
Miyauchi, T., and Furusaki, S. (1974). 'Relative contribution of variables affecting the
reaction in fluid bed contactors', AI.Ch.E. J., 20, 1087-1096.
Mori, S., and Wen, C.Y. (1976). 'Simulation of fluidized bed reactor performance by
modified bubble assemblage model', in Fluidization Technology (Ed. D.L.
Keaims), Vol. 1, Hemisphere, Washington, pp. 179-203.
Murray, J.D. (1%5). 'On the mathematics of fluidization', J. Fluid Mech., 21,.
465-493, and 22, 57-80.
Orcutt, J.C., Davidson, J.F., and Pigford, R.L. (1962). 'Reaction time distributions in
fluidized catalytic reactors', Chern. Eng. Prog, Syrnp. Ser., 58 (38),1-15.

Partridge, B.A., and Rowe, P.N. (1966). 'Chemical reaction in a bubbling gas
fluidized bed', Trans. Instn Chern. Engrs, 44, 335-348.
Pereira, LA.F. (1977). Ph.D. Thesis, University of Edinburgh.
Piccinini, N., Grace, J.R., and Mathur, K.B. (1979). 'Vapour-phase chemical reaction
in spouted beds: verification of theory'. Chem. Eng. Sci., 34, 1257-1263.
Pyle, D.L., and Rose, P.L. (1965). 'Chemical reaction in bubbling fluidized beds',
Chem. Eng. Sci., 20,25-31.
Raghuraman, J., and Potter, a.E. (1978), Countercurrent backmixing model for
slugging fluidized bed reactors, A.l. Ch.E. J., 24, 698-704.
Rajan, R., Krishnan, R., and Wen, C.Y. (1978). 'Simultation of fluidized bed
combustors: coal devolatilization and sulphur oxides retention', A. I. Ch.E. Symp.
Ser., 74 (176), 112-119.
Rovero, G., Piccinini, N., Grace, J.R., Epstein, N., and Brereton, C.M.H. (1983).
'Gas phase solid-catalysed chemical reaction in spouted beds', Chem. Eng. Sci., 38,
557-566.
Rowe, P.N. (1964). 'Gas-solid reaction in a fluidized bed', Chem. Eng. Prog., 60 (3),
75-80.
Rowe, P.N., MacGillivray, H.J., and Cheesman, D.J. (1979). 'Gas discharge from an
orifice into a gas fluidized bed', Trans. Instn Chem. Engrs, 57, 194-199.
Rowe, P.N., Partridge, B.A., and Lyall, E. (1964). 'Cloud formation around bubbles
in fluidized beds', Chem. Eng. Sci., 19,973-985.
Shaw, I., Hoffman, T.W., and Reilly, P.M. (1974). 'An experimental evaluation of
two-phase models describing catalytic fluidized bed reactors', A.I. Ch.E. Syrnp.
Ser., 70 (141), 41-52.
.
Shen, C.Y., and Johnstone, H.F. (1955). 'Gas-solid contact in fluidized beds',
A.I.Ch.E. J., 1,349-354.
Sit, S.P. (1981). Grid region and coalescence zone gas exchange in fluidized beds',
Ph.D. Dissertation, McGill University.
'Sit, S.P., and Grace, J.R. (1978). 'Interphase mass transfer in an aggregative fluidized
bed', Chem. Eng. Sci., 33, 1115-1122.
Sit, S.P., and Grace, J.R. 1981. 'Effect of bubble interaction on interphase mass
transfer in gas fluidized beds', Chem. Eng. Sci., 36, 327-335.
Smith, J.M. (1970. Chemical Engineering Kinetics, 2nd ed., McGraw-Hill, New York.
Tajbl, D.G., Simons, J.B., and Carberry, J.J. (1966). 'Heterogeneous catalysis in a
continuous stirred tank reactor', Ind. Eng. Chem. Fund., S, 171-175.
Toei, R., Matsuno, R., Nishitani, K., Hayashi, H., and Imamoto, T. (1969). 'Gas
interchange between bubble phase and continuous phase in gas-solid fluidized beds
at coalescence', Kagaku Kogaku, 33, 668-674.
Van Deemter, J.J. (1961). 'Mixing and contacting in gas-solid fluidized beds', Chem.
Eng. Sci., 13, 143-154.
Van Deemter, J.J. (1980). 'Mixing patterns in large-scalefluidized bed, in Fluidization
(Eds LR. Grace and J.M. Matsen), Plenum Press, New York, pp. 69-89.
Van Swaaij, W.P.M. (1978). 'The design of gas-solids fluid bed and related reactors',
A.C.S. Symp. Ser (Eds. D. Luss and V.W. Weekman), 72,193-222.
Walker, B.V. (1970). 'Gas-solid contacting in bubbling fluidized bed', Ph.D.
Dissertation, Cambridge University.
Walker, B .V. (1975). 'The effective rate of gas exchange in a bubbling fluidized bed',
Trans. lnstn Chem. Engrs, 53, 225-266.
Wen, C.Y. (1968). 'Noncatalytic heterogeneous solid fluid reaction models', Ind Eng.
Chern., 60,34-54.

W~n, C.Y. (1979). 'Chemical reaction in fluidized beds', in Proc. N.S.F. Workshop on
Fluidization and Fluid-Particle Systems (Ed. H. Littman), Rensselaer Polytechnic
Inst., pp. 317-387.
.
Wen, c.Y., and Fan, L.T. (1975). Models for Flow System~ and Chemical Reactors,
Marcel Dekker, New York.
Werther, J. (1980), 'Mathematical modeling of fluidized bed reactors', Int. Chern.
Eng., 20, 529-541.
Yates, J.G., and Gregoire, J.Y. (1980), 'An experimental test of slugging-bed reactor
models', In Fluidization, eds. J.R. Grace and J.M. Matsen, Plenum, New York,
pp. 581-588.
Yates, J.G., and Rowe, P.N. (1977). 'A model for chemical reaction in the freeboard
region above a fluidized bed', Trans. Instn Chern. Engrs, 55, 137-142.
Yerushalmi, J., Cankurt, N.T., Geldart, D., and Liss, R., (1978). 'Flow regimes in
vertical gas-solid contact systems', A.I.Ch.E. Syrnp. Ser., 74 (176),1-13.
Yoshida, K., and Kunii, D. (1968). 'Stimulus and response of gas concentration in
bubbling fluidized beds', J. Chem. Eng. Japan, 1, 11-16.
Yoshida, K., and Wen, c.y. (1970). 'Noncatalytic solid-gas reaction in a fluidized bed
reactor', Chem. Eng. Sci., 25, 1395-1404.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Solids Transfer in Fluidized Systems

In chemical processes employing the continuous circulation or transfer of


particulate solids, the success or failure of the venture as a whole is often
dependent upon the successful operation of the solids transfer system. The
large fluidized catalytic cracking (FCC) units used in the petroleum industry
and the coal conversion plants to produce synthetic fuels from coal are just
two examples of processes where this is true.
Because of the importance of the solids transfer system, its successful
design is critical to a process. Unfortunately, the design of such systems can
be extremely complex. Gas-solids mixtures (whether pneumatically conveyed
or moved in gravity flow in standpipes) can be transported in several different
regimes - each with its own inherent quirks and peculiarities. The problem
facing the designer of such a system is to couple these various regimes to
achieve a smooth, stable transfer of solids over a wide range of operating
conditions. This is a difficult task at best but it is made even more difficult
because the understanding of many aspects of two-phase gas-solids flow is
sadly inadequate. Thus, lacking a sound scientific basis for design, in practice
many transport systems are often still designed on the b~sis of operating
experience or rules of thumb.
Pneumatic transport of solids can be classified into four different regimes:
horizontal dilute phase flow, vertical dilute phase flow, horizontal dense
phase flow, and vertical dense phase flow. The boundary between dilute and
dense phase conveying is not clear-cut. The parameter of solids/gas loading
(kilograms of solid per kilogram of gas) in the conveying line has often been
used to distinguish between dilute and dense phase flow. Solids/gas loadings
of 0.01 to 15 kg of solid per kilogram of gas were used to denote dilute phase
341

flow, while dense phase flow was characterized by solids/gas loadings of 15 to


over 200 kg of solids per kilogram of gas. However, these are just very rough,
rule-of-thumb guidelines, and there can be much overlap. An analysis of each
type of conveying is given in the following section.
It is the objective of this chapter to present information which will increase
the probability of a successful design of solids transfer systems. In only one
chapter there is not space enough for exhaustive theory. Therefore, the
emphasis will be placed primarily on practical design.

In the design of industrial vertical dilute phase pneumatic conveying systems,


the main consideration is generally that of choosing the correct velocity at
which to transport the solids. Too Iowa velocity will result in unacceptable,
unstable slugging flow; too high a velocity will result in excessive gas
requirements and high pressure drops. The general relationship between
velocity and pressure drop per unit length, AP/ L, in a dilute phase vertical
riser is shown schematically in Fig. 12.1.

FRICTIONAL
RESISTANCE
PREDOMINATES

SLPERFICIAL VELOCITY, U
Figure 12.1 Phase diagram for dilute phase vertical pneumatic conveying.

Line AB is the pressure drop-velocity relationship for gas alone passing


through a dilute phase pneumatic conveying line. Curve CDE is for a solids
mass flux G1, while curve FG is for a solids mass flux Gz which is greater than
G1. At C, the gas velocity is very high and the conveying line is very dilute. As
the conveying velocity is decreased from C to D, the gas and solids both rise
more slowly. The solids inventory in the lift line rises, thus increasing the
static head. However, the frictional resistance portion of the total pressure
drop predominates in this region; thus, as the velocity decreases, so does the
pressure drop.
In the region D to E, the decreasing velocity results in a rapid increase in
solids inventory in the line. The static head of solids now predominates over
the frictional resistance and the pressure drop rises. Near E, the bulk density
of the mixture becomes too great for the gas velocity to support and the
mixture begins to slug, or choke. The superficial gas velocity at E is termed
the choking velocity Uch, which is the minimum velocity which can be used to
smoothly transport solids in vertical dilute phase conveying lines, and is
obviously an important design parameter. The curve FG defines the pressure
drop-velocity relationship for a higher mass flux, Gz. Feeding the conveying
line at the solids rate Gz will cause the conveying line to choke at a higher gas
velocity. Thus, choking can be reached by decreasing the gas velocity at a
constant solids flowrate or increasing the solids flowrate at a constant gas
velocity.
Although operation near choking will result in the minimum gas requirement for pneumatic solids transfer, the choking region is a very unstable one.
A small decrease in the gas velocity near choking causes the average pressure
drop in the lift line to rise rapidly. This is accompanied by large fluctuations in
the pressure drop as the line starts to choke. In large diameter conveying lines
the slugging of the solids can also cause excessive vibration which can be
structurally damaging if allowed to persist for any length of time.
A good operating point for vertical dilute phase conveying lines is at, or
slightly to the right of, D. At this point, the AP/L versus velocity curve is
relatively flat, and small perturbations in the system will not cause large
changes in the conveying line pressure drop. Lift gas requirements at D are
still low and yet far enough away from choking to be 'safe'. Perhaps most
importantly, this is the point of minimum lift-line pressure drop.
If centrifugal fans are used to supply the gas for the riser, there is another
reason why operation in the region D to E is unsatisfactory. Centrifugal fans
are characterized by reduced capacity as outlet pressure is increased. Leung,
Wiles, and Nicklin (1971a) have noted that if a slight system perturbation
caused an increase in solids flowrate, the pressure drop in the lift line would
increase. This would raise the pressure at the outlet of the fan and would
. result in a decrease in the output of the fan, producing a lower velocity in the
lift line. This would shift the system operating point on the AP/L versus

velocity curve towards E, producing a further increase in the system pressure


drop. This cycle would continue until choking conditions were reached,
resulting in unstable flow in the pipe. This destructive cycle does not occur if
the fan operates at velocities to the right of D.
Zenz and Othmer (1960) are generally considered to be the first to describe
choking. They considered it to be the degeneration of dilute gas-solids flow
into dense, slugging flow. Matsen (1981) and Yang (1982) have recently
suggested that choking occurs because of the formation of clusters or particle
aggregates. Indeed, Grace and Tuot (1979), through instability analysis,
showed that a uniform gas-solid suspension is unstable and that there is a
tendency for clusters to form (see Chapter 7).
However, not every gas-solids mixture will choke. It is possible for solids to
undergo a transition from a dilute phase directly to a denser non-choking,
fluidized bed type of transport. Yousfi and Gau (1974) were the first to
observe such a transition, and developed a criterion to precit when choking
,would occur; i.e.:
v2
g~

> 140

for choking to occur

(12.1)

Yang (1976a) extended this concept further and developed the following
criterion for choking:

vf

Fr= -< 0.35


gD > 0.35

no choking
choking occurs

This criterion (which is based on the maximum stable bubble size theory of
Davidson and Harrison, 1963) predicts that light and small materials are less
likely to choke, and takes into account the diameter of the line being
employed.
A third criterion for choking was proposed by Smith (1978):
Vt

f!"'-I n (1 - e)
05

(gD) .

>0.41

Leung (1980) considered these three choking criteria and concluded that
Yang's criterion was most useful since it contained the important parameter
of pipe size and was most consistent with the experimental data.
The concept that not all materials will choke is also supported by the work
of Canning and Thompson (1980). They found that large diameter particles
invariably formed slugs in vertical (and horizontal) pneumatic conveying,
while finer solids did not tend to slug.
The 'fast fluidized bed' (see Chapter 7) researched extensively by workers
at the City College of New York (Yerushalmi, Turner, and Squires, 1976)

appears to be a pneumatic conveying line operating in the relatively dense


phase region near choking. This has been proposed by both Yang (1982) and
Matsen (1981). Indeed, Gajdos and BierI (1978), using probe measurements
and X-ray exposures, reported that fast fluidization was phenomenologically
no different than classical dilute phase conveying. Both regimes were found
to consist of high velocity dilute core, surround by a low velocity dense annulus.
This annular core flow in a riser has also been reported by Batholomew and
Casagrande (1957), Van Zoonen (1961), and Saxton and Worley (1970).

12.2.1 Particle Velocity


The particle velocity in vertical dilute phase conveying lines is an important
parameter because it determines the residence time of the solids in the line.
The solids velocity in the conveying line differs from the gas velocity in the
line by the slip velocity:
v,

= IVg

v'lip

(12.4)

The slip velocity is often approximated by the terminal velocity of the


particles, and if this assumption is made, Eq. (12.4) becomes:

v, = IVg

vtl

(12.5)

Measurements by Hinkle (1953) using high speed photography and coarse


particles, showed that this relationship applied within 20 per cent. Capes
and Nakamura (1973) trapped a wide variety of solids in a 7.6 cm diameter
riser to determine particle velocities and concluded: (a) that slip velocity was
often greater than the terminal velocity and (b) that the deviation between
V,lip and Vt was greater for particles with high terminal velocities. This was
attributed to particle-to-wall friction and particle recirculation in the lift line.
However, their data also indicate that for particles with terminal velocities
below 7.6 mis, Eq. (12.5) can be used as a reasonable approximation.
Matsen (1981) has proposed that the slip velocity is basically only a function
of the voidage of the dilute, flowing suspension. He argues that as the voidage
decreases (due to increased solids loading, for example) the slip velocity
increases. He attributes this to cluster formation, but a solids annular flow
model would also explain this voidage-slip velocity relationship.
12.2.2 Choking Velocity Correlations
In spite of the extensive literature on all aspects of pneumatic conveying; the
most reliable way to determine design parameters for a particular pneumatic
conveying system is still by experiment. This generally requires the
construction and operation of a test rig, but this is not always possible because

347

SOLIDS TRANSFER IN FLUIDIZED SYSTEMS

of insufficient time and/or funds. There are, however, numerous correlations


for predicting conveying design parameters - so many, in fact, that a
designer could easily feel like the proverbial mosquito in a nudist colony he just would not know where to begin! Comparisons of many of these
correlations have been published, and although somewhat limited, the
comparisons provide a reference point for discussion.
The choking velocity defines the lower limit of the gas velocity for a dilute
phase vertical pneumatic conveying system. A comparative study of several
choking velocity correlations was carried out by the Institute of Gas
Technology for the US Department of Energy (Institute of Gas Technology,
1978). This report has an extensive review of the available choking velocity
correlations. The correlations of Zenz (1964), Rose and Duckworth (1969),
Leung, Wiles, and Nicklin (1971a), Yousfi and Gau (1974), Knowlton and
Bachovchin (1976), Yang (1975), and Punwani, Modi and Tarman (1976)
were evaluated using both low pressure and high pressure data. The
correlation of Punwani, Modi, and Tarman (1976) was recommended for use
in predicting choking velocities. This correlation is basically the Yang (1975)
correlation modified to take into account the considerable effect of gas
density on choking velocity shown by Knowlton and Bachovchin (1976). The
substantial dependence of Uch on gas density is illustrated in Fig. 12.2. The
Punwani correlation is shown below:
2gD

(E-4.7 -

1)

ch

(Vch -

GAS DENS ITY ,


0

10-3

30

50

40

20

EXPERIMENTAL
Ul

+'

!:>

"c+-

16

U
0

Ib/s-ft2

30 (146.5)
20 (97.7)

Ul

LI NES REPRESENT
CORRELATION

14

U
0

-'

-'

4 W

>

12

C!J

C!J
Z
.....

>t.....

>

",

o 40 (195.3)

>l.....

PTS,

(kg/s-m2)

18

Z
.....
~

10

J:

J:

U
8

2
6
0

= 8.72

20

10

kg/m3

0.5

1.0

1.5

2.0

GAS DENS !TV,

p~.77

Vt)

2.5
Ib/H3

3.0

3.5

The effect of gas density on choking velocity (Knowlton and


Bachovchin, 1976).
Figure 12.2

Gs

-=
Pp

(Vch

VI) (1 -

Ech)

A worked example of how to calculate the choking velocity using the Punwani
correlation is shown below.

Uch
=--

Vch

Ech

To calculate the actual gas velocity at choking, Eqs (12.6) and (12.7) must be
solved simultaneously for Ech and Vch' This correlation predicted the -data
tested within 25 per cent.
Yang (1982) has recently modified his earlier choking velocity correlation
to include the effect of gas density. The modified Yang choking velocity
correlation is also shown below:
. 2gD

1) _

(E~4.7 2

( Vch

VI )

Gs

Pp

6.81 x 10

(Vch

(.!!1..).2.2

Vt) (1 -

Pp
Ech)

Example 12.1: Choking velocity calculation -

Punwani correlation

Calculate the choking velocity for coal particles with an average particle size
of 300 ~m being conveyed at a rate of 1,816 kg/h using nitrogen in a 75 mm
diameter pipe at a system temperature of 20C and a system pressure of 17 bar
gauge.

= 21.6 kg/m3
= 1,200 kglm3
~ = 1.84 X 10-5

Pg
Pp
D
G

=75 mm

114 kg/m2 s

N s/m2

Solving the Punwani correlation for the choking velocity consists of solving
Eqs. (12.6) and (12.7) simultaneously for Vch and Ech'
In order to calculate the choking velocity a value must be obtained for the
terminal velocity VI> the method outlined in Chapter 6, Section 6.3, may be
used; alternatively, since the flow regime is transitional, Eq. (12.11) is
appropriate. Assume that the particle size given in the example is the volume
diameter dv.
0.153go.71d~.14 (pp _ pg)O.71
VI

p~.29

/LO.43

For the conditions given:

and presented in the Coal Conversion Systems Technical Data Book (Institute
of Gas Technology, 1978). Several of these correlations (Hinkle, 1953; Currin
and Gorin, 1968; Konno and Saito, 1969; Rose and Duckworth, 1969;
Richards and Wiersma, 1973; Knowlton and Bachovchin, 1976; Leung, 1976;
and Yang, 1976) were compared in the Data Book using the low pressure data
of Curran and Gorin (1968) and the high pressure data of Knowlton and
Bachovchin (1976). The correlation of Yang was found to predict the low
pressure data better than the other correlations. However, overall, the
modified Konno and Saito correlation was found to predict the data
somewhat better than the Yang correlation and, although much less
sophisticated than the Yang correlation, is much simpler to use. The modified
Konno and Saito correlation is presented below:

_ U2Pg
6.PT 2

(1)

Gs

Vs

=-

2fgPgU2L

D
(3)

0.057Upg8Lg
v'(gD)
(4)

(6)
(12.12)

0.095 m/s
Vs

(12.6) and (12.7) then become:

2 x 9.81 x 0.076 (E.;4.7 - 1)


(Vch _ 0.5)2
(Vch

Combining

sVs

(2)

Pp

Equations

= 8.72

10-

x (21.6)77

0.5) (1 - Ech) = 0.095

the equations we have:


(E~h4.7- 1) (1 - ECh)2 = 5.69

10-4

Using the trial and error method:


Ech = 0.9528

0.095

= vg

Gs

VI

and 8 =-UPg

and where the numbered terms have the following significance:


(1)
(2)
(3)
(4)
(5)
(6)

pressure
presure
pressure
pressure
pressure
pressure

drop due to gas acceleration,


drop due to particle acceleration,
drop due to gas-to-pipe friction,
drop related to solid-to-pipe friction,
drop due to the static head of the solids,
drop due to the static head of the gas.

If the gas and the solids are already accelerated in the lift line, then the first
two terms should be omitted from the calculation of the pressure drop.
A worked example of how to calculate the pressure drop in vertical dilute
phase pneumatic conveying using the Konno and Saito correlation is given
below.

1-Ech

Uch

= 2.51 mls
= Vch Ech =
12.2.3

2.51 x 0.9528

2.39 m/s

Vertical Dilute Phase Pressure Drop Corre/Qtion

Many correlations are also available for use in determining the pressure drop
in vertical pneumatic conveying lines. Twenty such correlations were found

Example 12.2: Pressure drop in vertical dilute phase conveyingKonno and Saito method
Determine the pressure drop in a dilutephase pneumatic conveying line in
which coal of an average particle size of 300 /Lm is being conveyed at a
velocity of 12 mls with nitrogen in a 75 mm diameter conveying line at a rate
of 1,861 kg/h. The temperature of the line is 21C and is at a pressure of 17 bar
guage. The line is 15 m long.

..
:
...
@
".::.:.

Solution

".a

Assume the gas is already accelerated when the particles enter the line. Since
the gas is already accelerated then, from Eg. (12.12):

APT

2fgPgU2L
Gsvs +
D

0.057UpgfJLg
GsLg
V(gD)
+~
+ PgLg

~
z

(12.13)

W
..J

f-

....
Z

= (12 - 0.5)
=

::l

m/s

11.5 m/s

,6p

Q.
Q.

Gs

= --

UPg

114

= ---12 x 21.6

0.44 kg solid/kg gas


I

2fg x 21.6 X (12)2 x 15

0.057 x 12 x 21.6 x 0.44 x 15 x 9.81

0.075

V(9.81 x 0.075)

= 114 x 11.5+-------+-------------

AP
T

o
~
a
w

::l

en
en

w
~

Q.

114 x 15 x 9.81
11.5
+ 21.6 x 15 x 9.81

To calculate the Reynolds number to determine the friction factor we have:


Re

DUpg
-fJ--

0.075 x 12 x 21.6
1.84 X 10-5

= 1.06 x 10

and

fg = 0.004
Therefore:
6..PT

= 1,311 + 4,976 + 1,115 + 1,459 + 3,178 = 12,039 N/m2

Since 10 ,000 N/m2=

1.45 Ib/in2:
APT

1.74lb/in2

As in vertical dilute phase pneumatic conveying, the problem confronting the


designer of a horizontal dilute phase transfer line is selecting a suitable gas
velocity. The general relationship between superficial velocity and API L for
a horizontal dilute phase transfer line is sho~n in Fig. 12.3. In many ways this
relationship is similar to that for a vertical dilute phase conveying line. Line
AB represents the curve obtained for gas alone travelling through the pipe,

curve CDEF for a mass flux G1, and curve GH for a mass flux G2 greater
than G1 At C, the gas velocity in the horizontal line is sufficiently high to
carry solids at a rate G1 through the pipe in a very dilute suspension. If the gas
velocity is reduced while continuing to feed at the constant mass flux G1, the
frictional resistance, and thus the pressure drop in the line, will decrease. The
solids will move more slowly and their concentration in the pipe will increase.
At D, the gas velocity is insufficient to keep the solids in suspension, and they
begin to settle to the bottom of the pipe. The gas velocity at D is termed the
saltation velocity, Usa1t> and is a strong function of solids loading.
If the gas velocity is reduced further to E, the pressure drop will rise rapidly
as the solids continue to deposit on the bottom of the pipe constricting the
space available for gas flow. As the gas velocity is reduced between E and F,
the depth of the particle layer and the pressure gradient both increase. In this
regime, some particles may move slowly in dense phase flow along the bottom
of the pipe, while, simultaneously, others travel in suspension in the gas in the
upper part of the pipe.
The saltation velocity sets a lower limit on velocity for any horizontal
dilute phase pneumatic transfer ~s-tem. It is desirable, however, to operate at as low a velocity and pressure drop as possible and still remain far
enough from saltation to be 'safe' if a system upset should occur. Therefore,

the saltation velocity needs to be determined to set a lower bound on this


operating point.
12.3.1

CONVEYING MEDIUM, AIR


SPHERICAL PARTICLES
---- ANGULAR PARTICLES

Saltation Velocity Correlation

There have been many efforts to correlate the saltation velocity. Two
comparisons of many of the published saltation velocity correlations have
been made (Jones and Leung, 1978;and Arastoopour et al., 1979). The Jones
and Leung comparison was made using eight different correlations (Thomas,
1962; Doig and Roper, 1963; Zenz, 1964; Rose and Duckworth, 1969; Rizk,
1973; Matsumoto etal., 1974a, 1975b; Mewing, 1976). The comparisons were
made using air-solids data only, and at low pressure. Jones and Leung
recommended that the Thomas correlation be used for design.
The Zenz correlation predicts that as particle size is decreased, the saltation
velocity at first decreases, then passes through a minimum, and then increases
again. This latter effect is explained as being due to small particles becoming
'trapped' in the laminar sublayer and needing a higher velocity to dislodge the
particle. However, for small particles (less than about 80 JLm in size), the
Zenz correlation at times predicts unbelievably high values of the saltation
velocity. It is best applied to particles larger than this size.
A worked example of how to calculate the saltation velocity using the Zenz
correlation is shown below.

Figure 12.4 Single-particle saltation velocities as a function of particle size in a 63.5 mm


diameter lucite tube (Zenz, 1964).
113

= /

(the units of ware in metres per second)

and
Example 12.3:

Calculation of saltation velocity-method

of Zem;

Calculate the saltation velocity of -10+ 100 US mesh sand being conveyed at
a rate of 1,362 kg/h through a 150 mm diameter tube with nitrogen at 21C
and 0.34 bar gauge.
3

2644 kg/m
Pg = 1.61 kg/m3
JL = 1.84 X 10-5 N s/m2
D = 150mm
Gs = 21.4 kg/m2 s
dpi (10 mesh particle) = 2,000 /-Lm /.
dpi (100 mesh particle) = 150/-Lm
Pp

Solution

The Zenz correlation utilizes a 'graphical procedure in conjunction with


various calculations to determine the saltation velocity Usa1t' First a
calculation of the single-particle saltation velocity, Uso, and a factor called Sa
must be made. In Fig. 12.4, Usolw is plotted versus dp/!:J. where:

!:J. =

1/3

12/

(the units of!:J. are in metres)

Substituting values:
w

!:J.

= 0.626 m/s

1.82

10-5 m

18.2 /-Lm

For the largest particle:


2,000
18.2

= 109.9

For the smallest particle:


dp)
( !:J.

=
s

150 = 8.24
18.2

Then from Fig. 12.4 for angular particles:

(Usc!w))
(Uso/w)s

(dplll)l)
(dplll)s
~..

SA

r~

I.

where S ~ is the slope of the line joining the coordinates


smallest particles in Fig. 12.4: Therefore:

2- =

(109.9)
8.24

3.3

for the largest and

Technical Data Book (Institute of Gas Technology, 1978) (Culgan, 1952;


Hinkle, 1953; Engineering Eqipment User's Association, 1963; Curran and
Gorin, 1968; Rose and Duckworth, 1969; and Yang, 1976a). The correlations
of Yang and Hinkle were found to predict the data far better than the other
correlations. The accuracy of both correlations was about the same for the
data tested. The Hinkle correlation, however, is much simpler than the Yang
correlation and is presented below:

SA

IlPT

S~ =0.29
According to Zenz,if SA is negative or less than 0.05, it should be taken as
0.05. However, it is not in this case and;

w(7)

(Uso)s

w(3.3)

4.38 mls

0.626 x 3.3

2.06 m/s

Notice that (Uso). is larger than (Uso)s' However, (U so)s can som~tiJJle pe
larger than (Uso)I' This can occur because the small particles can be ~in
the laminar boundary layer. Always use the largest Uso to calculate Usa1t,
however.
Figure 12.4 was determined from experiments in a 63.5 mm diameter tube.
The single-particle saltation velocities will now have to be corrected for the
larger sized 150mqt diameter 'tube. U.so !las been shown to be proportional to
DO.4 Therefore:
.
(Uso)150

(Uso)63.5

(150 \0.4
~

63.5)'

= 4.38 X (2.36)4 = 6.17 mls

[(Uso)sh50

The saltation velocity at loaded conditions is related to the single-particle


saltation velocity by:

.. = 0.213
Pp
21.4
2,644

(SA)1.5 (Usalt-Uso)
Uso

0.213 x (0.29)1.5 (Usa1t-6.17)


6.17

12.3.2

Horizontal Dilute Phase Pressure Drop Correlations

Several pressure drop correlations developed for horizontal dilute phase


pneumatic conveying were also compared in the Coal Conversion Systems

cJ2Pg

G
svs

2/gpgU2 L
D

where:

= U (1

Vs

(Uso)1

- 0.0638d~3pg.5)

Is = 2-!i

Co

8 Pp

!l.-

(U -

dp

II )2
s

Vs

An illustration of how to calculate horizontal dilute phase pressure drop using


the Hinkle correlation is given in Example 12.4

EX~Plel~

Pressure drop in horizontal dilute phase conveying Hinkle co"elotion

Calculate the pressure drop in a horizontal dilute phase pneumatic conveying


line in which coal with an average particle size of 300 }Lm and density 1,200
kg_m3 is being conveyed at a velocity of 18 m/s with nitrogen in a 75 mm
diameter conveying line at a rate of 1,816 kg/h. The temperature of the line is
21C and the line is at a pressure of 1.36 bar gauge. The line is 24 m long.
Assume the gas is already accelerated when the solids enter the line.
Since the gas is already accelerated when the solids are added to the line:
IlP
T

Pg

7f

= 2.83

=
Vs =
}L

=
=

+Gv +
ss

2-1'

U2L
D

'1gPg

kglm3

1.84 X 10-5 N s/m2


3 pg.5)
U (1 - 0.0638

4.

18 {I - 0.0638 (300 x 10-6).3 (1,200)0.5}


14.5 m/s

_ 1.816
4
_
2
Gs - 3,600 '17'(0
..075)2 - 114 kg/m s

Solution

At high Reynolds numbers, Co in Eq. (12.19) is a constant at 0.43 and:


_ 3 x 2.83 x 0.43 x 0.075 (18 - 14.5)2
fs - 8 x 1,200 x 300 x 10-6
14.5
fs

= 0.0056

From Eq. (12.20):


182 x 24) (
0.0056 14.5
114 )
1 +-- --0.075
fg
18 18 x 2.83
0.075 x 18 x 2.83 _
x 5
1.84 X 10-5
- 2.08 10

tlP = 114 x 14.5


DUpg

= -p.- =
fg

2fg(2.83

+ -------

= 0.004

fsvs
0.0056 x 14.5
-=----=
fgU
0.004 x 18

1.13

Zenz (1964) recommends that iffsvifgU > 1 then assume [.viU fg


calculation of the pressure drop; Le.:
tlP = 1,653 + 4,694 = 6,347 N/m2 = 0.92 Ib/in2

= 1 in the

The various correlations described above were found to be the 'best'


correlations for predicting Usalt, Uch, the horizontal dilute phase pressure
drop, and the vertical dilute phase pressure drop, based on the data used.
However, for the following reasons, care must be taken in applying these
correlations and automatically assuming that they will accurately predict the
design parameters for a particular system:

Therefore, when applying these correlations, bear in mind that the state of
the art (as applied to pneumatic conveying correlations) still leaves much to
be desired.

Bends complicate the design of pneumatic dilute phase transfer systems, and
when designing a conveying system it is best to use as few bends as possible.
Bends increase the pressure drop in a line, and also are the points of most
serious erosion and particle attrition.
Solids normally in suspension in straight, horizontal, or vertical pipes tend
to salt out at bends due to the centrifugal force encountered while travelling
around the bend. Because of this separation, the particles slow down and are
then reentrained and reaccelerated after they pass through the bend, resulting
in the higher pressure drops associated with bends.
There is a greater tendency for particles to salt out in a horizontal pipe
which is preceded by a downflowing vertical-to horizontal bend than in any
other configuration (Patterson, 1959). If this type of bend is present in a
system, it is possible for solids to remain on the bottom of the pipe for very
long distances following the bend, before they redisperse (Fig. 12.5b). The
particles in a horizontal pipe following an upflowing vertical-to-horizontal
bend are not as likely to salt out since they can easily redisperse into the gas
stream due to gravity (Fig. 12.5a). Therefore, it is recommended that

J.

A
(a)

VERTICAL (Up) TO
HORIZONTAL ELBOW

(a) The correlations were evaluated fot a very limited amount of data.
(b) There is disagreement between studies as to which correlation is 'best'.
Note that the Jones and Leung (1978) analysis resulted in a different
'best' correlation for USal1 than did the Arastoopour et al. (1979) study.
(c) The 'best' correlations can generally be expected to predict parameters
no more accurately than 30 to 40 per cent.
(d) If the correlations are to be applied to high temperature and/or high
pressure systems, their accuracy may be much worse than 30 to 40 per
cent. since nearly all of the correlations were developed from data
obtained at ambient conditions.

(bl VERTICAL <Down) TO


HORIZONTAL ELBOW

..
.
.
....... .
'

_:

C
(e)

BU NDED TEE HOR I ZONT AL


TO VERTICAL (Up)
J

downflowing vertical-to-horizontal bends be avoided if at all possible in dilute


phase pneumatic conveying systems.
As noted above, bends increase the pressure drop in a conveying system. In
addition to the pressure drop due to the bend itself, because the bend slows
down the solids there is also a pressure drop due to subsequent particle
reacceleration by the gas. The length of straight pipe needed before the solids
reattain their steady state velocity can be considerable (Scott, Richard, and
Mooij, 1976).
In the past, designers of dilute phase pneumatic conveying systems
intuitively thought that gradually sloped, long radius elbows would reduce the
erosion and increase bend service life relative to 90 elbows. Zehz (1960),
however, recommended that blinded tees (Fig. 12.5c) be used in place of
elbows in pneumatic conveying systems. The theory behind the use of the
blinded tee is that a 'cushion' of stagnant particles collects in the blinded or
unused branch of the tee, and the conveyed particles then impinge upon the
stagnant particles in the tee rather than on the metal surface, as in a long
radius or short radius elbow.
Bodner (1982) determined the service life and pressure drop of various
bend configurations. He found that the service life of the blinded tee
configuration was far better than any other configuration tested and that it
gave a service life 15 times greater than that of radius bends or elbows. This
was due to the 'cushioning accumulation of particles' in the blinded branch of
the tee which he observed in glass bend models. Bodner also reported that
pressure drops and solid attrition rates for the blinded tee were approximately
the same as those observed for radius bends.
Marcus, Hilbert, and Klinzing (1984) reported that long radius bends have
higher pressure losses than short radius bends due to the extra length of the
bend.
Schuchart (1968), Morikawa et al. (1978), Tsuji (1980), lung (1958), and
Weidner (1955) also investigated pressure drops through bends in dilute
phase conveying systems. However, in spite of this work, at present there is
no reliable method of predicting accurate bend pressure drops other than by
experiment for the actual conditions expected. In industrial practice, bend
pressure drop is often approximated by assuming that it is equivalent to
approximately 7.5 m of vertical section pressure drop. In the absence of any
reliable correlation to predict bend pressure drop, this crude method is
probably as reliable and as conservative as any.
The formulation of an accurate method of determining bend pressure drop
is one area which needs much work.
12.5

well dispersed, the inclined pipe pressure drop was found to be less than the
vertical dilute phase pressure drop. At lower gas velocities the solids formed
nodes on the bottom of the pipe. A further reduction in the conveying
velocity resulted in slugging or choking in the pipe. This choking condition
occurred at a higher velocity than observed in vertical pipes. No useful
correlations exist to predict choking velocities and pressure drops in inclined
pipes.
12.6

COMBINED SYSTEMS

Many times a dilute phase pneumatic conveying system will consist of a


combination of horizontal and vertical transfer lines with various bends in the
line. This system is more complex than the simple vertical or horizontal lines
considered earlier. If no slanted lines or downward flowing vertical-tohorizontal bends are used in the system, the lower velocity limit for the
system will be the saltation velocity, since this velocity is always greater than
or equal to the choking velocity for a particular gas-solids mixture (Zenz,
1960). The determination of the final design ofthe transfer system depends
upon the pressure drop balance in the total circulation loop.
12.7

EFFECT OF PIPE DIAMETER

It is important to know the effect of pipe diameter on dilute phase conveying


design parameters (that is Ueb, Usa1t, and t1P) when scaling up in size.
Surprisingly, there is very little information available in the literature on the
effect of pipe diameter on these parameters, even though most test rigs
operate in the 5 to 7.5 cm diameter range and commercial conveying lines are
in the 30 to 122 cm diameter range.
Analysis of the data of Patterson (1959) on his work with pulverized coal
shows that the pressure drop in his dilute phase conveying line was greater for
30.5 em diameter pipe than for 20.3 cm diameter pipe.
Woebcke and Cofield (1978) found that the dilute phase pressure drop in
horizontal conveying increased with increasing pipe diameter. They also
reported that the 'minimum velocity' (saltation velocity) in their tests
increased with increasing pipe diameter over the range of 3.2 to 15.2 cm
diameter pipes.
The Zenz (12964)correlation for saltation velocity also predicts an increase
in Usa1t with pipe diameter to the 0.4 power, i.e.:
Usa1t

ex:

iJO.4

DILUTE PHASE CONVEYING IN INCUNED PIPES

The only data available on the dilute phase flow of solids through inclined
pipes are due to Zenz (1960). At very high velocities where the solids were

although this dependency was based on work with rather small tubes, 3.2 to
6.4 cm in diameter.

The Yang (1976a), Punwani, Modi, and Tarman (1976), and modified
Yang (1982) correlations for the choking velocity also predict an increase in
Uch with pipe diameter. This dependency is approximately pipe diameter to
the 0.25 power, i.e.:
Uch

DO.25

In summary, theory and sketchy literature results both predict an increase in


Uch and Usa1t with pipe diameter. This increase is generally proportional to
the 0.25 to 0.4 power of the pipe diameter. Even sketchier literature results
indicate that dilute phase pneumatic conveying pressure drop increases with
pipe diameter. More systematic studies are definitely needed in this area.

12.8

DENSE PHASE PNEUMATIC CONVEYING

Dilute phase conveying is by far the most common type of conveying. It can
also be the most economical. Sandy, Daubert, and Jones (1970) calculated
that the minimum power requirements for dilute phase vertical pneumatic
conveying were approximately an order of magnitude less than the power
requirements for dense phase vertical pneumatic conveying of an air-fused
alumina system. The large power requirements needed for vertical dense
phase transport were a result of the high pressure drops needed to move
solids upwards in dense phase flow.
However, it is not always the case that dilute phase pneumatic conveying
will be less costly than dense phase pneumatic conveying as it depends
strongly on the particular gas-solids system being used. In general, the tradeoff which must be made is between the higher pressure drop per unit length
(I1P/L) associated with dense phase pneumatic conveying versus the higher
gas flows needed for dilute phase pneumatic transport.
In spite of the high pressure drop per unit length needed, dense phase
conveying offers certain advantages over dilute phase conveying which
may tip the scales in its favour for a particular process. Since the solids
velocities are much lower in dense phase conveying than in dilute phase
conveying, solids attrition and pipe erosion are considerably reduced. This
feature is particularly desirable when expensive. or fragile solids are to be
transported, and it is desired to minimize particle breakup and elutriation
losses. Dense phase conveying also decreases the demand on a particle
collection system.
Gas requirements for dense phase pneumatic conveying (as noted above)
are much lower than for dilute phase pneumatic conveying. Sandy, Daubert,
and Jones (1970) found that vertical dilute phase gas requirements for their
air/fused alumina system were over 20 times those for vertical dense phase
conveying. In instances where the conveying gas is extremely expensive, or

where it is desired to minimize the amount of diluting transfer gas fed to a


reactor, dense phase transport is an alternative choice.
In this chapter, dense phase conveying will denote all solids conveying in
which the solids are not completely suspended in the gas stream. Solids can be
conveyed in dense phase both vertically and horizontally. In vertical dense
phase conveying the entire line is filled with solids in either a packed bed or
fluidized bed slugging mode. In horizontal dense phase conveying the line
mayor may not be completely filled with solids.
12.8.1

Packed Bed Vertical Dense Phase Conveying

In vertical dense phase packed bed conveying, the solids are transported
through a tube by forcing transporting gas upwards through the interstices of
the particles. When the frictional forces between the gas and the particles are
sufficient to overcome the weight of the particles plus the frictional forces of
the solids sliding on the tube walls, the solids will start to move upwards. This
occurs at a pressure drop per unit length, I1P/L, approximately equal to that
at minimum fluidization for that particular material. This I1P/L value is also
approximately equal to the bulk density of the solids.
In order to prevent fluidization of the particles in the pipe or tube the outlet
of the conduit must be restricted in some manner. This can be accomplished
by placing a valve, orifice, or an expanded section of a non-fluidized bed of
solids at the outlet of the pipe. This effectively prevents the solids from
expanding and becoming fluidized.
Because of the restriction at the outlet of the pipe, the relative velocity
between gas and solids may now be greater than the minimum fluidization
velocity, and the solids will move upwards as a packed bed. This operating
mode is shown as lines GH and 1] in the phase diagram of Fig. 12.6, where the
mass flux G1 is less than G2 The familiar fluidization and dilute phase vertical
conveying I1P/L versus velocity curves (ODE and ABC, respectively) are
also shown in this figure for reference. Curve OD is the packed bed I1P/L
versus velocity curve. At D the I1P/L is sufficient to cause the bed to expand
and normally become fluidized. However, if the bed is restricted (e.g. by a
screen small enough to let only gas pass through it) the I1P/L continues to
increase with velocity at zero solids flow. Thus, ODF is the I1P/L versus
velocity curve for G = 0 in vertical dense phase transport.
Berg (1954a, 1954b, 1954c, 1954d) received several patents relating to the
vertical transport of solids in a dense phase packed bed regime in order to
reduce particle attrition and loss of expensive catalyst material. Because of
bubble formation at high solids flowrates and excessive pressure drops in the
transport line, both due to gas expansion, Berg had to taper or increase the
diameter of the transport line every few feet. Alternatively, he also found that
removing some of the gas would also prevent excessive pressure drops in the

the gas to support the solids. The Leung, Sandy et al. and Wen and Galli
correlations are extremely sensitive to voidage and particle size. Their
accuracy is strongly dependent
upon the correct prediction of these
properties. The Wen and Galli correlation is shown below:
150(1 -

ef /L(Ule - vs)
(I/Jdvf~
(1)

1.75(1-e)Pg(Ule

- vs)z

!/Jdve
(2)

(l-e)(pp

- Pg)

2pgU
D

fg

where:
(1) and (2) are the pressure drop due to gas-solids friction,
(3) is the pressure drop for the gas to support the solids,
(4) is the gas-wall friction loss.

SUPERFICIAL VELOCITY, U

Figure 12.6 .Phase diagram for dense phase vertical pneumatic


conveying.

line. Sandy, Daubert, and Jones (1970) were also forced to use increasingly
large flow areas to keep 6.PIL from becoming excessively high in their dense
phase vertical transport line.
Extreme care must be exercised in placing an expanded section of pipe at
the exit of the line in dense phase vertical transport (Knowlton and Hirsan,
1980) so that its area is not too large. A large diameter section requires a gas
velocity large enough to produce a 6.PIL equal to the bulk density of the
material before solids will flow. This amount of gas also has to pass through
the smaller diameter section of pipe preceding the expanded section. If the
ratio of the large diameter pipe to the small diameter pipe is too large, the
6.PIL in the smaller diameter section can become exorbitantly high.
Several correlations have been developed to predict the pressure drop in
dense phase vertical packed bed transport (Berg, 1954a, 1954b, 1954c, 1954d;
Sandy, Daubert, and Jones, 1970; Leung, Wiles and Nicklin, 1971b; and Wen
and Galli, 1971). The Berg correlation is relatively complex. The Leung
correlation is the simplest and is merely the Ergun equation applied to a
moving packed bed. The Sandy and the Wen and Galli correlations a~so
employ the Ergun equation to predict the pressure drop due to gas-sohds
friction, but also include other terms, primarily the pressure drop needed for

12.8.2

Fluidized Bed DenSe Phase Vertical Conveying

If the solids in a dense phase vertical transport line are not restricted as in
dense phase packed bed flow, they can be transported in dense phase in
fluidized bed flow. However, because of gas expansion, this flow generally
reverts to bubbling or slug flow unless the transport line is tapered.
This type of flow is characterized by lines LM and NP on the phase diagram
of Fig. 12.6, where NP represents a solids mass flux, Gz, greater than Gr.
Although these curves are informative, they are not as useful as they are in
dilute phase conveying. In dilute phase conveying, the gas velocity can be
reduced while maintaining a constant mass flux. The solids are simply
transferred at the same flow rate, but in a denser state. This is usually not the
case in dense phase vertical transport with fluidized (or packed bed) flow.
This is because the transport gas rate also generally determines the solids flow
rate. When the superficial transport gas velocity is changed, the solids
flowrate also changes. Thus, a knowledge of how the solids flowrate varies
with gas velocity is also needed for the design of dense phase pneumatic
conveying systems. A typical solids flowrate versus velocity curve for an
unrestricted vertical dense phase transfer line operating in slug flow is shown
in Fig. 12.7, along with a curve showing the variation in pressure drop in the
transport line as a function of solids flowrate (Fig. 12.8). The curves were
obtained in conjunction with a non-mechanical J-valve study by Knowlton
and Hirsan (1980).

GAS FLUIDIZATION

364

TECHNOLOGY

J-VALVE AERATION, actual


1.5

1.0

0.5

2.0

3.5

3.0

2.5

12.8.3

m3/h

In horizontal dense phase conveying, it is not necessary that the solids


completely fill the pipe cross-section. In fact, it is usually the case that they
will not. At any velocity less than that necessary to keep the solids in
suspension (i.e. less than the saltation velocity) the solids are considered to be
in dense phase flow.
The saltation velocity is the velocity at D in Fig. 12.3 for a mass flux, Ct.
The portion of curve CDEF described by DEF in Fig. 12.3 is considered to be
in dense phase flow. As the velocity is reduced from D to E, a particle layer
appears on the bottom of the pipe. From E to F the depth of the particle layer
increases and the pressure drop in the pipe builds up.
Wen and Simons (1959) studied horizontal dense phase gas solids flow in
0.5,0.75. and 1 in diameter glass pipe over a wide range of solids/gas loadings
(80 to 800 kg of solids per kilogram of gas). They visually observed that the
solids (glass beads and coal ranging from approximately 70 to 750 ,urn) went
through four separate flow regimes, depending on the solids/gas loading or
the gas velocity in the pipe.
At high gas velocities the solids were observed to be in dilute phase
suspension (Fig. 12.9a). At lower gas velocities, the solids settled in the pipe

4.0

...c.
32

"-.Jl
0
0
0

J-VALVE

DIAM:

J-VALVE
(61 em)

LENGTH:

28

MATERIAL:

24

-<
>I
')

14
24

"-.:i.

SIDERITE

12

SYST. PRESS.'
(55 kPo)

AERATION GAS'

NITROGEN

psig

W
>
..J

-<
>I
to

')

I
C!l

20

I
C!l

:J
a
a::

:J
a
a::

..c.
01

In.

(300 fLm)

,
W
>
..J

I-

16

W
I-

I-

-<
a::

3:
..J
lI..

I-

...:

12

IX
3:

a
..J

en
0
...

U-

Dense Phase Horizontal Conveying

...J

rn

a
,.,.

..J

en

0
rn

0.4

0.8

1.2

1.6

2.0

2.4

J-VALVE AERATION, actual ft3/min


Figure 12.7 Dense phase vertical fluidized solids flow rates as a function
of motive gas (Knowlton and Hirsan, 1980).

SOLIDS FLOW RATE THROUGH DENSE-PHASE LIFT, kg!h


3

o
w~

~':J~ ~

WE

~o

1.3

1.2

.0..
lJ..<I

....
..J

...,

:J'(1)0

DENSE-PHASE LIFT DIAM. 3 In. (7.6 em)


MATERIAL: SIDERITE (300 I'm)
SYST. PRESS.' 8 pslg (55 kPo)
AERATION GAS. NITROGEN

,
o..~

t.4

o LOWER DENSE-PHASE

LIFT
LIFT

A UPPER DENSE-PHASE

1. 1

~~
0.. ,
..J
...
'lJ..0..
:J<I

WE
(1)0

WO

(1)0

~~
o..w

"':lJ.. 1.0

iE~
IW
~~

10..

W
0.13

----;;--

WO

15

~Ic:.

a 8

A
I

~~
00
5

10

11

SOLIDS FLOW RATE THROUGH DENSE-PHASE LIFT, 1000


Figure 12.8 Pressure
(Knowlton and Hirsan,

drop in dense
1980).

phase

vertical

fluidized

12

13

Ib/h

flow versus

solids

Figure 12.9 Solids flow regimes


horizontal pneumatic conveying.

ln

dense

phase

and dune formation occurred (Fig. 12.9b). Decreasing the gas velocity even
more (or, equivalently, increasing the solids/gas loading) caused more solids
to settle to the bottom of the pipe and the solids were observed to travel in
small dunes or ripples across the top of a thick, relatively stationary, salted
layer (Fig. 12.9c). A further increase in solids loading eventually caused a
complete blockage of the pipe.
In some cases, Wen and Simons observed that intermittent slug flow of
solids (Fig. 12.9d) occurred instead of dune formation (Fig. 12.9b). This was
accompanied by unstable, wildly fluctuating pressure drops across the line.
Because of the uncertainties associated with horizontal dense phase
conveying (HDPC) and the high probability of plugging the pipe, most
designers have shied away from horizontal dense phase conveying. They have
opted instead to use the dilute phase conveying for its known reliability. In
fact, HDPC has long been considered somewhat of a 'black art' by many
people. This attitude has grown because there has been no systematic
procedure with which to predict what system to use with a particular material.
Some materials could be conveyed relatively easily in the HDPC mode, while
others would plug the line and the system would fail.
A paper by Canning and Thompson (1980) appears to offer the first
systematic procedure which can be used to determine what horizontal dense
phase system to use with a particular type of material. This work, along with
the earlier work of Flatt (1976), has helped to significantly advance the state
of the art of HDPC.
There are three basic types of HDPC systems (Fig. 12.10). The simplest is
the blow tank, or simple-pipe system., depicted in Fig. 12.1O(a). With this
system, pressurizing and motive air is usually supplied to the top of a tank and
the solids and secondary motive air are added to the conveying line through a
valve below the tank.
The second type of system is a pulsed phase system in which gas is added to
the conveying line in pulses (Fig. 12.10b). This type of system is used to
artificially form solid plugs of selected length for those solids which do not
naturally form slugs.
The third type of system is the by-pass type (Fig. 12.lOc). With this system,
some of the conveying air is by passed around the solids so that the plug
length can be limited and plugging will not occur.
Canning and Thompson found that the system which was best for a
particular solid depended on whether or not the particles to be conveyed
formed slugs, and, if they did what type of slugs were produced. They
also found that a particular solid behaved similarly in both horizontal and
vertical flow. They therefore analysed slug formation using choking-velocity
theory.
Yang (1976a) argued that when the velocity of a gas slug in vertical
conveying was greater than the terminal velocity of the particles, then no

SIMPLE-PIPE

SYSTEM (Blow Tank)

MOTIVE AIR

PULSED SYSTEM
MOTIVE AIR

(PULSED-FLOW

AIR KNIFE

BYPASS SYSTEM
MOTIVE AIR

Figure 12.10 A schematic representation of three primary dense phase feed


systems.

stable slugs were possible. In this case the particles could be fast fluidized
without slugging or choking. Therefore, for unstable or asymmetric slugs, the
terminal velocity must be less than the slug rise velocity, or:
Vt

< 0.35(gD)o.5

Similarly, for stable slugs or choking to exist:


Vt

> 0.35(gD)o.5

For fine particles, the terminal velocity can be written:


Vt

(pp - pg)d~g
= -----

18p.

for Ret

< 0.2

For higher Reynolds numbers other expressions


Combining Eqs (12.22) and (12.24) we have:

for

Vt

should be used.

< 0.35(gD)o.5

(pp - pg) gd~


18jL

Canning and Thompson reported that with the aid of the fluidization
diagram it was possible to choose a suitable HDPC system for a particular
powder based only on the fluidization classifiaction of the powder. The
system best suited for a particular powder type is described below.
Group C Solids

Thus, we now have a relationship between dy and (pp - pg) for various pipe
diameters. This relationship may now be plotted on a Geldart (1973)
fluidization diagram, as shown in Fig. 12.11, to distinguish between the
various slugging regions.

D
STRONG, STABLE
AXISYMMETRIC
SLUGS

UNSTABLE
ASYMMETRIC
SLUGS (Dunes)

NO
~

SLUGGING

~-:.
..~
.........

......

...

""

""

Group A Solids
These materials are the best candidates for HDPC. They retain air and
remain fluidized, and can be generally conveyed at low velocities and at high
loadings with little risk of pipe blockages. All three types of systems shown in
Fig. 12.10 can be used with these materials. For long (greater than 200 m
distances) it may be advisable to use a by-pass system.
Group B Solids
These materals are the worst type to use in HDPC. At high loadings they tend
to cause pipe vibrations and blockage. The by-pass system can be used to
convey these materials at higher loadings and at lower velocities than a dilute
phase system.

...

"

These are materials which are generally very difficult to convey in HDPC
because of their cohesiveness.

'1"
Group D Solids
100

1,000

These materials can be conveyed easily if they have a narrow particle size
distribution. All three system types may be used to convey these particles.

MEAN PARTICLE SIZE (jLm)

Figure 12.11 The Geldart fluidization diagram applied to dense phase pneumatic
conveying (Canning and Thompson, 1980).

The overriding concern in HDPC is to avoid blocking the pipe. Blockages


are most likely to occur with materials that form weak, asymmetric, unstable
slugs. These materials are found in the group B region of the Geldart
fluidization diagram. With these materials, the gas slugs move faster than the
interstitial gas and the material conveys in dune flow. When the dunes grow
to fill the pipe, severe vibration and eventual blockage of the pipe may result.
In the group D region of the diagram, the gas slugs move more slowly than
the interstitial gas and stable slugs can form which completely fill the pipe
diameter.

12.8.4

Extrusion Flow

Zenz and Rowe (1976) studied a group of materials which can be made
to flow through pipes in extrusion-type flow. In this type of flow, the
solids occupy the entire pipe cross-section and flow much like plastic
being extruded through a round die. Zenz and Rowe found that only
certain solids would flow in this manner. These solids were generally of
low bulk density and very small particle size. The solids also had to exhibit
a property called 'bulk deform ability' , characterized in this study by a test
wherein a steel rod was placed on top of the solids. If the rod fell through
the solids without aerating them, the solids could be conveyed in extrusion
flow.

12.8.5

Air-Assisted Gravity Conveyor

Solids can be transferred against a pressure if gas flows upwards relative to


the downwar.d fl~wing solids, thus generating the required 'sealing' pressure
drop. The dlrectlo~ of the actual gas flow relative to the standpipe wall,
however, can be either up or down and still have the relative gas-solids
velocity Vr directed upwards.

Another type of system which can transfonn solids in HDPC is the airassisted gravity conveyor. With air-assisted gravity conveying, the solids are
fluidized with air distributed along the length of the conveying pipe or
channel (Woodcock and Mason, 1976). The solids then flow by gravity along
the bottom of the channel. This type of conveying uses a very small amount of
energy. Although solids can be made to move against gravity with this type of
conveying, it is generally used only with 'downhill' conveying.

The standpipe has been in use for over 40 years. It was 'invented' by a
research team from Jersey Standard in the 1940s (Campbell, Martin, and
Tyson, 1948) working on feeding cracking catalyst into fluidized beds during
the Second World War.
The purpose of a standpipe is to transfer solids from a region of low
pressure to a region of higher pressure. This is indicated in Fig. 12.12 where
solids are being transferred from a fluidized bed at pressure PI to another
fluidized bed operating at pressure P., which is higher than Pl.

CASE I
GAS FLOWING
UPWARD RELATIVE
TO PIPE WALL

CASE II
GAS FLOWING
DOWNWARD RELATIVE
TO PIPE WALL

I
I
I
I

Vg I

I
I

Vi

9I
I
I

UPPER
FLUIDIZED

BED

+
DIRECTION

OF

PRESSURE DROP
Figure 12.13 Schematic representation of gas flow down
and gas flow up relative to the standpipe wall.

LOWER

This can be shown with the aid of Fig. 12.13 and the definition
relative velocity. Mathematically:

FLUIDIZED

Vr

BED

= Ivs

vgl

where:
Vr

= relative gas solids velocity (also called

Vs

= solid

velocity

vg = interstitial gas velocity (UJe)

VSlip)

of the

In solids transfer systems, most people tend to visualize what is occurring in a


particular pipe by mentally travelling along with the solids. Therefore, in this
chapter the positive reference direction in determining Vr will always be the
direction the solids are flowing. For vertical pneumatic conveying this positive
reference direction is generally upwards. For standpipe flow, the positive
reference direction is downwards.
In Fig. 12.13, solids are being transferred downwards in a standpipe from
pressure PI to a higher pressure Pz. Solid velocities in this figure are denoted
by the length of the thick-lined arrows, gas velocities by the length of the
dashed arrows, and the relative velocity by the length of the thin-lined
arrowS.
In case I, solids are flowing downwards and gas is flowing upwards relative
to the standpipe wall. The relative velocity is directed upwards and is equal to
the sum of the solids velocity and the gas velocity, i.e. (for downwards being
the positive direction):
Vr

Vs

(-Vg)

Vs

+ vg

(12.27)

For case II, solids are flowing down the standpipe and gas is also flowing
down the standpipe relative to the standpipe wall, but at a velocity less than
that of the solids. For this case, the relative velocity is also directed upward
and is equal to the difference between the solids velocity and the gas velocity,
i.e.:

For this case, if one were riding down the standpipe with the solids, the gas
would appear to be moving upwards.
Gas flowing upwards relative to the solids causes a frictional pressure drop
to be generated. The relationship between pressure drop per unit length
(I1PIL) and relative velocity is determined by the fluidization curve,
schematically shown in Fig. 12.14. This curve is usually generated in a
fluidization column when the solids are not flowing. However, the relationship also applies for moving solids in standpipe flow.
As the relative velocity through the bed of solids is increased from zero,
I1PIL increases linearly with Vp This is the packed bed region. At some Vn
the I1P generated is equal to the weight of the solids per unit area and the
particles become fluidized. The relative velocity at this point is called the
interstitial minimum fluidized velocity, IJrnf, or Urnr!Ernf' The I1PI L at IJrnf is
designated as I1PIL)rnf and is often referred to as the fluidized bed 'density'
since I1PIL has the units of density.
Increases in Vr above Vrnf do not lead to further increases in I1PIL. For
Geldart group B materials, any increase in gas flow in excess of that required
at Vrnf goes into the formation of bubbles. Therefore, as Vr increase, API L
stays relatively constant and then begins to drop as the bubble volume in the

FLU ID I ZED BED

REGION

f-

(Bubbling)

C!l

~
--J

~(~:L---------a:
w
Q.

~
a:
a
w
a:
::>
UJ
UJ

a:
Q.

RELATIVE VELOCITY, Vr
Figure 12.14 Typical-fluidization curve for Geldart group B solids.

bed increases. The I1PIL versus Vr curve shown in Fig. 12.14 is for group B
solids.
The fluidization curve for Geldart fro up A materials is slightly different
from that for group B solids. As Vr is increased above Vrnffor these materials,
the solids expand without bubbles forming for a certain velocity range, and
I1PIL drops over this velocity range. The velocity where bubbles begin to
form is called the minimum bubbling velocity, Vrnb, to differentiate it from
Vrnf' A typical fluidization curve for group A solids is shown in Fig. 12.15.
Standpipes generally operate in one of three basic flow regimes - packed
bed flow and fluidized bed flow, and streaming flow:
1. Packed bed flow. In packed bed flow, Vr is less than Vmf i.e. (Urnf/Ernf) and
the void age in the standpipe is essentially constant. As Vr is increased,
I1PIL increases in packed bed flow. Often when a standpipe is operating
in the moving packed bed flow regime a flow condition is reached which
causes the standpipe to vibrate and a loud 'chattering' of the solids occurs.
This flow mode is not present at low solid velocities and also decreases in
severity as the relative gas/velocity approaches Vrnf' This flow regime
should be avoided but, unfortunately, no method exists to predict when it
will occur. Little work has been conducted on this problem, and this is also
an area which needs much work.

PACKEDBED
REGION

FLUIDIZED
BUBBLING
REGION

FLUIDIZED
NON-BUBBLING
REGION

I-

C!l
Z
W
...J

(a) All gas in excess of Vrnf travels through the bed in the form of bubbles.
(b) Gas travels through the emulsion at an interstitial velocity equal to
Vrnf

!::
(l>p)
Z ::>

These four types of bubbling fluidized bed flow are discussed individually
below. In the discussion, it is assumed that the two-phase theory of
fluidization holds (see Chapter 4). this theory holds that:

L mf

Type 1 flow. For this type of bubbling fluidized bed flow, both the gas
flowing through the emulsion and the bubbles flow upwards relative to the
standpipe wall, as indicated in Fig. 12.16(a). This type of bubbling
fluidized flow occurs in a standpipe when Vs is less than Vrnf. For fluidized
bed bubbling type 1 flow:
Vs < Vrnf
(12.29)

----------

0::

I
I
I
I
I
I
I
I
I
I
I
I
I
I Vmb
I
I

a.
a.
o
0::
o
W

0::

::>
(J)
(J)

0::

a.

TYPE 2

!cb
o
o

o
2. Fluidized bed flow. In fluidized bed solids flow,

is equal to, or greater


than, Vmf. The voidage in the standpipe can change along the length of the
standpipe and t:J.PIL does not increase with Vr There are two kinds of
fluidized bed flow:
Vr

(a) bubbling fluidized bed flow and


(b) non-bubbling fluidized bed flow.
When a group B solid is operating in fluidized bed flow it always operates
in the bubbling fluidized bed mode since bubbles are formed at all relative
velocities above Vmf. However, for group A solids, there is an operating
'window' corresponding to a relative or slip velocity between Vrnf and Vrnb,
where the solids are fluidized but no bubbles are formed in the standpipe.
A standpipe operating with a relative velocity above Vrnb and with group A
solids operates in the bubbling fluidized bed flow.
In the bubbling fluidized bed mode of standpipe operation there are
four different types of regimes which can be distinguished:
Type
Type
Type
Type

1.
2.
3.
4.

Emulsion
Emulsion
Emulsion
Emulsion

gas flow up, bubble flow up, net gas flow up


gas flow down, bubble flow up, net gas flow up
gas flow down, bubble flow up, net gas flow down
gas flow down, bubble flow down, net gas flow down

;~
0
0
0

TYPE 3

ieb

0
0

0
0

0
0

0
0

TYPE 4

EMULSION GAS UP

EMULSION GAS DOWN

EMUlSION GAS DOWN

EMULSION GAS DOWN

BUBBLE GAS UP

BUBBlE GAS UP

BUBBLE GAS UP

BUBBLE GAS DOWN

NET GAS UP

NET GAS UP

NET GAS DOWN

NET GAS DOWN

(a)

(b)

(el

(d)

Figure 12.16 Schematic representation of four types of bubbling fluidized bed standpipe
flow.

Type 2 flow. In this regime, schematically depicted in Fig. 12.16(b), Vs is


greater than Vrnf so that emulsion gas will flow down the standpipe relative
to the wall. In order for the bubbles in the standpipe to rise, Vs must also be
less than the bubble rise velocity, Ub. The volumetric rate of upward
bubble flow (Qb) is greater than the volumetric rate of downward-moving
dense phase gas flow (Qd) for this flow regime, making the net gas flow
upwards. Therefore, for bubbling fluidized bed type 2 flow:

throughout the entire standpipe length is not desired because the full dense
phase 'head' of the solids is not developed.
Type 3 flow.
This type of bubbling fluidized bed flow is similar to type 2
flow in that Vs is greater than Vmf but less than lb, as shown in Fig.
12.16(c). However, for this type of flow, Qb is less than Qd, resulting in a
net gas flow downwards. Therefore, for type 3 bubbling fluidized bed flow:

Type 4 flow.
In this case (Fig. 12.16d), the solids velocity is so great that
it exceeds lb' The bubbles cannot rise in the standpipe and are carried
down the standpipe relative to the standpipe wall at a velocity equal to
Vs-lb' The net gas flow is obviously downward and so for type 4 bubbling
fluidized bed flow:

Bubbles, especially large bubbles, are undesirable in a standpipe. If a


standpipe is operating in the bubbling fluidized bed mode such that the
solids velocity, v s, is less than the bubble rise velocity lb, then bubbles will
rise and grow by coalescence. The bubbles rising against the downward
flowing solids hinder or limit the solids flowrate (Knowlton and Hirsan,
1978; Eleftheriades and Judd, 1978). The larger the bubbles, the greater
the hindrance.
When the solids velocity in the standpipe is greater than the bubble rise
velocity, the bubbles will travel down the standpipe relative to the
standpipe wall. It is also possible in this case (Zenz, 1984) for bubbles to
coalesce and hinder solids flow through the standpipe.
Bubbles also reduce the 6.Pt L or 'density' of the solids in the standpipe.
Thus, a standpipe operating in a strongly bubbling regime will require a
longer length to seal the same differential pressure than for a standpipe in
which the same solids are at slightly above minimum fluidizing conditions.
In light of the above discussion, it is evident that for optimum fluidized
bed standpipe operation:
(a)

in a standpipe operating with group B solids should be maintained


just slightly greater than Vmf;
(b) Vr in a standpipe operating with group A solids should be maintained
between Vmf and Vmb'
Vr

3. Streaming flow. Standpipes can also operate in a streaming, dilute phase


flow characterized by high voidages. In general, this type of flow

Other subclassifications of these basic flow regimes have been noted and
mathematically defined by Leung and coworkers (Leung, Wiles, and Nicklin,
1971a, 1971b; Leung and Wilson, 1973; Leung, 1976; Leung and Wiles,
1976; Leung and Jones, 1978a, 1978b, 1978c; Leung, 1980), who have studied
standpipe flow extensively over the past ten years. Their efforts have
significantly advanced the understanding of solids flow in standpipes. Their
detailed analyses of standpipe operation have demonstrated that multiple
steady state standpipe solids flow regime combinations are possible. More
recently, Jones, Teo, and Leung (1980) have analysed the stability of
standpipe operation. Any serious, in-depth study of standpipe operation
should include a review of their work.
Basically, the three flow regimes described above are the ones most
frequently existing in a standpipe. Various combinations of these flow
regimes can also exist. Fluidized bed flow can exist above packed bed flow
and streaming flow is extremely common above fluidized bed flow and even
packed bed flow.
12.9.1

Overflow and Underflow Standpipes

There are two types of standpipe configurations which can be employed the overflow standpipe (Fig. 12.17a) and the underflow standpipe (Fig.
12.17b). The overflow standpipe is so named because solids overflow from
the top of the fluidized bed into the standpipe. In the underflow standpipe,
the solids are introduced into the standpipe from the underside, or bottom, of
the bed or hopper.
In any gas-solids flow system a pressure drop 'loop' can be defined such
that the sum of the pressure drops around the loop will be zero. In any such
loop in which a standpipe is a component, the standpipe is usually the
dependent part of the loop. This means that the pressure drop across the
standpipe will automatically adjust to balance the pressure drop produced by
the other pressure drop components in the loop. How the standpipe pressure
drop adjustment is made is different for overflow compared to underflow
standpipes.
Consider the overflow standpipe system shown in Fig. 12.18(a). Solids are
being transferred from the upper fluidized bed to the lower fluidized bed
against the pressure differential P 2 - P l' This pressure differential is composed of the pressure drop in the lower fluidized bed from the standpipe exit
to the top of the bed (6.PLB), the pressure drop across the grid of the upper
fluidized bed (6.Pgrid), and the pressure drop across the upper fluidized bed
(6.PUB), i.e.:

PRESSURE PROFILE
IN COLUMN
PRESSURE PROFILE
IN STANDPIPE

li~./..

.,.
..

:::.:
..:. :.:.:.:-T
.
.: : : APLB
......
.... . ...' .l..

.'.~'
,"

\"t)~';llf~~t,i

::':P2'::

II

',_

l>Psp

PRESSURE

( a) OVERFLOW

STANDPIPE
Figure 12.17 Overflow and underflow standpipe configurations.

This pressure drop must be balanced by the pressure drop in the overflow
standpipe. If the standpipe is operating in the fluidized bed mode at a I1PIL
equal to (I1PI L)mf, the solids height in the standpipe Hsp will adjust so that
the standpipe pressure drop I1Psp will equal the product of (I1PIL)mf and
Hsp, i.e.:
I1Psp = Pz-PI

= ( ~:-

)mf Hsp

t.Psp

(12.36)

The pressure diagram for this system is also shown in Fig. 12.18(a).
If the gas flowrate through the two beds is increased, I1Pgrid will increase
while the pressure drops across the two beds will remain essentially constant,
so that P2-PI increases to P2-PI. The pressure drop across the standpipe
will then also adjust to P2- Pl' This occurs because the height of solids in the
standpipe will increase from Hsp to H~p so that:
I1Psp = P2-PI

(~:

)mfH~p

(12.37)

This is illustrated in Fig. 12.18(b) and is also' reflected in the pressure


diagram. If the increase in pressure drop is ever such that Hsp must increase to

------<"""

PRESSURE

a value greater than the standpipe height available, the standpipe will not
operate.
In Fig. 12.19(a), solids are being transferred through an underflow
standpipe from the upper fluidized bed to the lower fluidized bed against the
differential pressure, PZ-P1.The differential pressure, PZ-PI, consists of the
pressure drop across the grid of the upper fluidized bed ~Pgrid' However,
there is also a pressure drop across the solids flow control valve ~Pv for this
standpipe configuration. Therefore, the standpipe pressure drop ~Psp must
equal the sum of I1Pv and I1Pgrid, i.e.:

PRESSURE
-

PRESS~E

----

PRESSURE PROFILE

PROFILE IN COLUMN
IN STANDPIPE

(b)

AP)
L II
-------

AP)

L
---

I
I

IIV),.I

standpipe. Therefore, t:..PI L in the standpipe must increase to balance the


pressure drop loop. This will occur (if the standpipe had been operating in the
packed bed mode before the increase in t:..Pgrid) by a simple increase in Vr in
the standpipe. This can be visualized with the aid of Fig. 12.19(c).
For Case I, the pressure drop loop in the bed was satisfied by having the
standpipe operate at point I on the t:..PIL versus Vr curve, as shown in Fig.
12.19(c). When the pressure drop in the system increased, the Vr in the
standpipe automatically adjusted to generate a higher t:..PIL, (t:..PILhh to
balance the higher system pressure drop.
If the pressure drop across the grid would have increased to a value such
that the product of (t:..PIL)mf and the standpipe height, Hsp were less than the
sum of t:..Pv and t:..Pgrid, then the standpipe would not seal and, therefore,
would not operate.
From the discussion above, it is evident that the pressire drop in an
overflow standpipe adjusts because of a solids level change in the standpipe.
the pressure drop in the packed bed underflow standpipe adjusts by a I1PIL
change due to a changing Vr.
In addition to the way the two standpipes adjust to balance pressure drops
in the loop, there are other distinguishing characteristics of overflow and
underflow standpipes.

II

I
I
I
I

I V ,.) II
I

Overflow Standpipes
(a) It is necessary to immerse the exit end of an overflow standpipe in a
fluidized bed. If not immersed, the standpipe will empty of solids, a
pressure seal will not form, and solids will not flow through the
standpipe.
(b) The overflow standpipe operates 'automatically'. This means that the
solids which enter the top of the overflow standpipe are discharged at the
same rate into the fluidized bed without the need of a valve, i.e.
'automatically' .

Thus, for this underflow standpipe case, the standpipe must generate a
pressure drop greater than PZ-P1. This is shown as case I in the pressure
diagram of Fig. 12.19(b).
If the gas flowrate through the fluidized beds is increased, t:..Pgrid will
increase. If it is assumed that t:..Pv remains constant, then t:..Psp must increase
to balance the pressure drop loop. This is shown as case II in the pressure
diagram of Fig. 12.19(b). Unlike the previous case discussed for the overflow
standpipe, the solids level cannot rise to increase the pressure drop in the

(c) The overflow standpipe usually operates with solids in the fluidized
regime. However, it is not necessary that it do so. The overflow
standpipe can also operate in the packed bed mode.
(d) The overflow standpipe generally operates only partially full of solids.
This is because an overflow standpipe is designed so that it will have a
height 25 to 30 per cent. greater than needed to develop the design
pressure drop.
(e) Since the solids enter the overflow standpipe from the top of the bed,
large solids, or agglomerates, may collect at the bottom of the feed
fluidized bed since they cannot be discharged.

Underflow Standpipes

(a) Solids enter the underflow standpipe from the bottom of a fluidized or
packed bed. Therefore, it is necessary to have a valve at the exit of the
underflow standpipe to maintain solids in the bed. This valve must
control the solids flowrate and the standpipe usually does not operate
'automatically' .
(b) The underflow standpipe operates completely full of solids.
(c) The underflow standpipe usually operates in the packed bed mode with
group B solids. With group A solids, either the fluidized bed or the
packed bed mode can be used.
(d) The underflow standpipe does not need to be buried in a fluidized bed to
operate. The valve at the bottom of the standpipe enables the overflow
standpipe to also discharge the solids above the bed, if desired.
(e) The underflow standpipe will discharge agglomerates which collect near
the bottom of the bed if they are not larger than the standpipe diameter.

AERATION
TO STANDPIPE

AERATION
TO STANDPIPE

One of the most common types of standpipes is the underflow standpipe


operating with group A solids in the fluidized bed mode. This type of
standpipe is used extensively in the petroleum industry to feed cracking
catalyst from fluidized beds into dilute phase pneumatic conveying lines (Fig.
12.20).
In this type of underflow standpipe, the standpipe pressure drop will not
adjust to balance changes in system pressure because the standpipe is
operating in the fluidized bed mode. This standpipe is designed so that the
standpipe length is great enough to generate more 'head' than needed. The
excess pressure is then <burned up' across the slide valve used to regulate the
solids flow. The slide valve is also used to adjust the pressure drop in the
system loop to balance pressure drop changes in the other pressure drop
components.
The solids velocity Vs in the standpipe carrying cracking catalyst is generally
of the order of 1 to 2 mis, while Vrnf is approximately 1 cmls. This means that
gas is also flowing down the standpipe at a velocity vg of about 1 to 2 m/s.

the stan?pipe wall. The cracking catalyst friction loss in an 8 in. (20 em)
diameter standpipe, as reported by Matsen (1976), is shown in Fig. 12.21. In
this figure, the standpipe friction loss is presented as a function of the true
density of the catalyst-gas mixture in the standpipe (usually measured with
radiation in commercial standpipes). When the true density of the mixture in
the standpipe increases above that at minimum fluidization, the friction loss
due to solids sliding on the standpipe wall increases sharply.

12.9.2 Friction Loss

12.9.3 Standpipe Aeration

For a standpipe operating in a fluidized mode, the pressure buildup in the


standpipe is nearly entirely due to the hydrostatic 'head' of solids. The
pressure loss due to solids/wall friction is basically negligible in the fluidized
bed mode. This is not the case for packed bed flow in standpipes.
When the relative velocity in a standpipe is less than Vrnf, the solids-wall
friction can be significant. Basically, the percentage of the solids weight not
supported by the gas will be supported by the friction of the solids sliding on

The purpose of adding aeration gas to a standpipe is generally to maintain the


solids in the standpipe in a fluidized state near the minimum fluidization
conditi?~. !1:is results in t~e highest standpipe density, or IlP/L, possible,
thus mlllImIzlllgthe standpIpe length and also the solids/wall friction loss.
There are two primary instances where aeration is added to standpipes to
maintain fluidized flow:

.'

if"

.~

DILUTE-PHASE
CONVEYI NG LINE "

DILUTE-PHASE
CONVEYING LINE

Figure 12.20 Typical solids flow loop in a catalytic cracker with underflow standpipe.

(a) To prevent defluidization due to gas compression in long standpipes,

300
C')

"-OJ
.:t

, 200

00

t-

OO
0
-I
Z
0
....

:r:
<.?
W
:r:
100

tO

.....

~
lJ...

400

500

600

700

800

900

TRUE CATALYST-GAS MIXTURE DENSITY,

1000

kg/m3
Fig,ure 12.22 Standpipe pressure profiles as a function of height in an underflow standpipe in
which VOIdage decreases wIth height.

(b) To prevent
standpipes.

defluidization

due to high solid flow rates

m overflow

Defluidization Due to Gas Compression


This is by far the most common reason for adding aeration to standpipes.
Consider the underflow, fluidized group A cracking catalyst standpipe system
shown in Fig. 12.20, where gas and solids are both flowing downwards
relative to the standpipe wall. The length of this type of standpipe can be
considerable (30 to 45 m) in commercial units. In such a long standpipe at low
pressures, the percentage change in gas density can be significant from the top
of the standpipe to the bottom. Typically, the absolute pressure may vary
from 1.36 to 3.4 bar, giving an increase in gas density of 250 per cent.
Kojabashian
(1958) showed that the solids voidage in the standpipe
decreases along the standpipe as the pressure increases. This is because the
gas volume decreases or 'shrinks' and the solids move closer together, thus
causing the voidage e to decrease. The voidage can change so much because
of the pressure change that the flow regime in the lower part of the standpipe
can change from fluidized bed flow to packed bed flow. Kojabashian (1958)
and Leung (1976) have shown that several different pressure profile
variations can exist, as shown in Fig. 12.22.

If the standpipe is fluidized over its entire length and if the fluidized density
is constant throughout the entire length of the standpipe, then the standpipe
pressure profile will take the form shown in Fig. 12.22(a).
As noted above, since e decreases as the gas density increases down the
standpipe, the solids flow regime can be in fluidized bed flow at the top of the
standpipe and then revert to packed bed flow at the bottom of the standpipe.
The effect that decreasing e has on Vr can be seen by expressing the relative
velocity equation:

v
r

Gs
Pp (1-e)A

--

Gg

Pg A

(12.40)

If e decreases percentage-wise faster than Pg increases, then Vs decreases


and vg increases - causing Vr to decrease also. If Vr decreases to a value less
than Vmf at the bottom of the standpipe, but gas is still flowing upwards
relative to the solids, then the pressure profile will take the form shown in
Fig. 12.22(b). the pressure buildup in the standpipes for this case is less than
that shown in Fig. 12.22(a) by the difference, l:1P, indicated in the figure.

As noted by Kojabashian (1958), in some instances E can decrease so


much that vg becomes greater than Vs and Vr becomes negative in Eq. (12.28)
Pressure reversal then occurs (i.e. the gas faster than the solids), as shown in
Fig. 12.22(c).
Adding aeration to the standpipe can keep the voidage from decreasing at
the bottom of the standpipe and prevent the transition to packed bed flow, as
indicated in Fig. 12.22(b) and (c). It is generally more effective to add
aeration gas at several points along the standpipe to 'replace' the gas volume
lost because of the pressure change. Matsen (1973) and Zenz (1984) have
both developed equations to calculate the amount of aeration which must be
added to this type of standpipe to prevent defluidization. Dries (1980) has
also discussed this problem in detail.
For bubbling fluidized bed flow, the effect of adding aeration is produced in
different directions, depending on the type of flow in the standpipe. For
bubbling fluidized bed flow types 1, 2, and 3 (as described above), the effect
of adding aeration to the standpipe is produced above the aeration injection
point (Fig. 12.23a). This is because Vs > Ub and the bubbles added to the
standpipe via the aeration will rise upwards relative to the standpipe wall for
these three flow types.

BUBBLING
FLUIDIZED-BED
FLOW TYPES
L

2, AND3

BUBBLING
FLUIDIZED-BED
FLOW TYPE
4

For type 4 bubbling fluidized bed solids flow where Vs < Ub, bubbles added
to the standpipe by aeration are carried down the standpipe with the solids,
and the effect of the aeration is produced below the aeration injection point
(fig. 12.23b).
It must be emphasized here that the minimum amount of aeration
necessary to maintain fluidization in the standpipe is the optimum amount of
aeration to use. Too much aeration gas added to the standpipe will cause
large bubbles to form which will hinder and decrease the flow of solids down
the standpipe.
Defluidization

Due to Solid Flow Rate Increases

Sauer, Chan, and Knowlton (1984) have recently shown that overflow
standpipes operating in a bubbling fluidized bed mode can defluidize if the
solids flowrate is increased above a limiting value. Defluidization in this case
is not caused by solids/wall friction but is thought to occur because of the
force produced by the dilute phase streaming solids falling on the top of the
dense phase solids in the standpipe. This force may 'compress' the fluidized
column of solids, decrease E, and cause defluidization. Sauer, Chan and
Knowlton (1984) showed that adding aeration to the bottom of the standpipe
at a single location could prevent defluidization. This is because the solids
velocities in the standpipes they were operating were always less than the
bubble rise velocity (bubbling fluidized bed flow types 1, 2, and 3). If Vs had
exceeded Ub in their standpipes (bubbling fluidized bed flow type 4), aeration
would have had to be added at the top of the standpipe.
12.9.4

Figure 12.23 Direction of aeration effectiveness


for v, < Ub and v, > Ub'

Angled Standpipes

Angled (or inclined) standpipes are used in many cases in industry to transfer
solids between two points which are separated horizontally, as well as
vertically. Angled standpipes would seem to be a very convenient way to
transfer solids on paper; operationally, angled standpipes are not recommended.
Sauer, Chan, and Knowlton (1984) studied angled standpipes and a hybrid
combination of an angled standpipe and a straight standpipe operating in
fluidized bed bubbling flow. They used transparent lines to observe the flow
of solids in the angled sections in these standpipes. Invariably, they found that
the gas and solids separated, the gas channelling via bubbles along the upper
half of the pipe while the solids flowed along the bottom portion of the
standpipe. This is illustrated in Fig. 12.24.
Sauer, Chan and Knowlton (1984) also found that the pressure buildup in
angled standpipes was much lower than in straight standpipes due to
gas-solids separation. Because of this separation, the gas does not completely

SOLID FLOW RATE, THOUSANDS


0.7

OF kg/h

5
I

0.10

W
0..
t-<

0..

I(f)

0.05

~
-.J

"'0..
<J

-60
ANGLED ST ANDP I PE
-----STRAIGHT
STANDPIPE
D:

2.5

in.

MATERIAL:

o
~CREASING
VOIDAGE AND
GAS VELOCITY

(6.4

em)

SAND (275 fLm)

SOLID FLOW RATE, THOUSANDS

10

12

OF

Ib/h

fluidize the solids and produce the full hydrostatic head of solids. This is
illustrated in Fig. 12.25, which compares the aPIL produced by a straight
overflow standpipe and an angled overflow standpipe as a function of the
solids flowrate through them. As can be seen, the straight or vertical
standpipe produces a much higher standpipe density.
If at all possible, underflow standpipes to be operated in the fluidized
bubbling bed mode should be designed to be straight. The section of overflow
standpipes operating in dense phase flow should also be designed to be
straight. The section of an overflow standpipe operating in dilute phase
streaming flow can be angled with no adverse effects, however. This is
indicated in Fig. 12.26.
12.9.5 Stripping in Standpipes
During the transfer of solids from one fluidized bed to another via a
standpipe, a substantial amount of gas can be transferred with the solids in the

NO

YES

Figure 12.26 Correct and incorrect way to utilize


angles in an overflow standpipe.

interstices. For safety reasons, or to prevent contamination or dilution of


product gas, it is frequently desired to prevent gas being transferred via a
standpipe from coming into contact with the gas in the receiving bed (Fig.
12.27). To accomplish this, a neutral gas is usually added to the standpipe to
replace, or 'strip', the normal downcomer gas in the void spaces between the
solids.

Moving Packed Bed Stripping


Knowlton et al. (1981) have published the only systematic study of stripping in
moving packed bed standpipes. They found that a standpipe that operated in
a moving packed bed regime with non-porous solids and that had stripping
gas added to it could be analysed using a relative gas-solids velocity concept.
They found that two cases must be considered: (a) gas travelling down with
the solids and (b) gas travelling up the standpipe.
Gas Flow Down
For the standpipe

shown in Fig. 12.28(a),

CASEI
GAS TRAVELLING
DOWNWARD

the gas is initially travelling


CASE

II

GAS TRAVELLING
UPWARD

.t

I I
I I
I I

I
I
I

p: ,
I

I I

I I I
I
I I

I
I

I I

I I
II

I
I

I
I

'012345

'012~45
STRIPPING
GAS
INJECTION
(N2)

STRIPPING
GAS (N2)

I
I
I

:
I
I

I
I
I
I
I

I I
I
I

, t

I
I
I

I
I
I

I
I
I

I
I

If the solids being transferred are porous, then a substantial amount of gas
may be trapped in the pores and transferred along with the interstitial gas.
During the process of stripping, the gas trapped in the pores may also diffuse
into the bulk of the interstitial gas and increase the stripping requirements
significantly relative to non-porous solids stripping requirements.
Stripping gas can be added to standpipes operating in both the moving
packed bed and fluidized bed modes.

, t I
: t
I

Figure 12.27 Schematic of stripping in a standpipe between two fluidized beds.

I I I :

v.

:
012345

V.

I
I

012345

Figure 12.28 Relative gas-solid velocities in a stripping standpipe: (a) gas flowing downwards
and (b) gas flowing upwards.

downwards relative to the standpipe wall. The dotted arrows designated '0'
for case I in Fig. 12.28(a) indicate the magnitude and direction of the gas
above and below the stripping gas injection point (SGIP) when no stripping
gas is being added to the standpipe. When the stripping gas is added to the

standpipe, the pressure drop in the standpipe above the SGIP rises, whereas
the pressure drop in the standpipe below the SGIP falls, as shown in Fig.
12.29. However, the total standpipe pressure drop (Fig. 12.29) remains
constant.

This trend continues until no gas is carried down the standpipe with the solids
(point '3' in Fig. 12.28a). At this point, vg = 0 above the SGIP. If more
stripping gas is added, stripping gas will now flow up the standpipe above the
SGIP, as well as down the standpipe section below the SGIP. At this point,
the gas from the upper fluidized bed is completely stripped; i.e. it cannot
travel down the standpipe.
Gas Flow Up

q,
:r
'080
E
u

a:060
a:
c
IJJ

a:

5len
IJJ

a:
a.

40

...o

--0

.!:

a:
o

is 20 !i=~

Material -20+40 meshOttawa sand


Solidsflow rate 6300 Ib/h (2857.7 kg/h)
<n
Stripper diameter 4 in (10.2em)
~ 10 I- Stripper length 11.25ft (3.43m)
a.
0 Totalstandpipe t,p
/1 t,P below Nz injection point
o t,p above Nz injection paint
00
0.1
0.2
NITROGENSTRIPPINGGAS FLOW RATE, ACFM

i:5

0.2
0.4
0.6
NITROGENSTRIPPING GAS FLOW RATE, actual m3/min x 10-2

In Fig. 12.28(b), the gas in the standpipe is initially moving upwards. When
stripping gas is added to this standpipe, the standpipe pressure drop above the
SGIP rises and the pressure drop across the standpipe below the SGIP falls.
Thus, the Vr in the standpipe above the SGIP rises and the Vr in the standpipe
below the SGIP decreases. Therefore, all the stripping gas that is initially
added goes up the standpipe above the SGIP, diluting the gas already moving
upwards.
In the standpipe section below the SGIP the pressure drop falls; therefore,
the amount of gas moving up the pipe in this section has to decrease. As more
stripping gas is added, this trend continues until no gas moves up the pipe
(point '3' in Fig. 12.28b) and the standpipe is effectively stripped.
Thus, in a packed bed of non-porous particles, the interstitial gas can be
stripped with another gas regardless of whether the interstitial gas was
initially moving upwards or downwards. The actual stripping section is below
the SGIP in the former case and above it in the latter.
Stripping Gas Injection Point Location

Because pressure drop is proportional to the relative velocity, Vr increases


in the standpipe section above the SGIP and decreases in the standpipe
section below the SGIP when stripping gas is added to the standpipe. Arrows
designated '1' in Fig. 12.28(b) represent the gas velocities in the standpipe
above and below the SGIP after a small amount of stripping gas has been
added to the standpipe. The amount of gas being carried down the standpipe
above the SGIP decreases at condition '1' relative to condition '0' in order to
increase Vn and hence the pressure drop in this section as stripping gas is
added. However, the amount of gas flowing down the standpipe section
below the SGIP increases to decrease Vr in that section because of the
decrease in pressure drop upon the addition of stripping gas.
As more stripping gas is added to the system:
(a) The standpipe pressure drop above the SGIP increases.
(b) Vr increases in the standpipe above the SGIP.
(c) The amount of gas being carried down the standpipe in the section above
the SGIP decreases even more.

Changing the location of the SGIP in a standpipe affects the amount of gas
needed to strip. It does so because the pressure drop distribution in the
standpipe changes.
Moving the SGIP can, therefore, reduce the amount of gas needed to strip
in a standpipe. For gas travelling down the standpipe, moving the SGIP closer
to the top of the standpipe reduces the amount of stripping gas needed.
For gas travelling up the standpipe, moving the SGIP towards the bottom
of the standpipe reduces the stripping gas requirement.
In actual practice, stripping efficiency may fall off if the SGIP is located too
close to the ends of the standpipe. This is because there may be a minimum
standpipe length necessary for complete stripping because of mixing,
diffusion, or other factors.
Stripping in Fluidized Standpipes
If Vs in a standpipe is greater than Vmf then the interstitial gas in the standpipe
cannot be stripped before the solids fluidize. Knowlton et al. (1981) described

the results of several tests in which the standpipe tiP/Labove


the SGIP
became high enough so that Vr exceeded Vmf and thl-solids began to fluidize in
that section. Two cases must again be considered: gas travelling down with
the solids and gas travelling up the standpipe.
Emulsion

Gas Down

When stripping gas is added to a fluidized standpipe, the pressure drops


above and below the SGIP remain essentially constant. Any excess stripping
gas added to the standpipe will, therefore, travel up the standpipe as bubbles.
In Types 2 and 3 bubbling fluidized flow, before the stripping gas is added,
interstitial gas is still travelling down the standpipe in the emulsion phase with
the solids.
For a very brief period standpipe gas is travelling downwards in the
emulsion phase, while stripping gas is travelling upwards in the bubble phase.
The stripping gas in the bubble phase and the standpipe gas in the emulsion
phase then interchange over the height of the stripper. The result of this
situation is that there is a gradient in the stripper wherein the standpipe gas
concentration is lower at the bottom than at the top. This condition exists
until the stripping gas travelling up the standpipe effectively 'absorbs' all of
the standpipe gas and the standpipe is 'stripped'. For Type 4 bubbling bed
flow, complete stripping cannot be achieved for the same reasons noted for
Type 1 flow. No matter how much stripping gas is added to the standpipe, the
interstitial standpipe gas flow downwards above the SGIP will continue to
flow down the standpipe. Adding stripping gas results only in dilution of this
gas.
Emulsion

At this condition, solids would be


theoretically no stripping gas would be
will not occur because solids flowrates
certain degree of mixing and diffusion
optimum stripping a standpipe should
possible. Therefore, vg = 0 should be
section design.

falling through stagnant gas, and


needed. Realistically, however, this
do not stay constant, and there is a
in a 'real' standpipe. However, for
be operated as close to vg = a as
the criterion for optimum stripping

A non-mechanical
valve is a device which uses only aeration gas in
conjunction with its geometrical shape to control the flowrate of particulate
solids through it. Non-mechanical
valves offer several advantages over
traditional mechanical valves:
(a) They have no moving mechanical parts which are subject to wear and/or
seizure.
(b) Constructed
sive.

of ordinary

pipe and fittings, they are extremely

Gas Up

When the standpipe section above the SGIP becomes fluidized for gas flowing
in Type 1 bubbling fluidized flow, complete stripping cannot be achieved. If
more stripping gas is added to the fluidized standpipe for this case, the
standpipe gas will be diluted by the bubbles of stripping gas. However, no
matter how much stripping gas is added, the initial interstitial standpipe gas
flowing upwards below the SGIP will continue to flow up the standpipe and
complete stripping will not occur. Thus, for the case of gas flowing initially up
the standpipe operating in the fluidized bed mode, the interstitial gas cannot
be stripped, only diluted.
Stripping Section Design
The minimum amount of stripping gas needed to strip interstitial gas in a
moving packed bed standpipe would be when vg = a initially in the standpipe.

Figure 12.30 L-valve feeding into dilute phase


and dense phase systems.

inexpen-

(c) They can be fabricated 'inchouse', which avoids the long delivery times
often associated with the purchase or replacement of mechanical valves.
(d) They can be used to control or feed solids into either a dense phase or
dilute phase environment,
such as a fluidized bed or a pneumatic
conveying line (Fig. 12.30).
The three most commonly used types of non-mechanical valves are the Lvalve, J-valve, and the reverse seal. These three devices are shown
schematically in Fig. 12.31. The only difference between these devices are

their shape, the direction in which they discharge solids, and the distance
between the control aeration injection point and the solids discharge point.
All three devices operate on the same principle.
The most common of the non-mechanical valves is the L-valve (so named
because it is shaped like the letter 'L') because its shape is easiest to construct
and it is slightly more efficient than either the J-valve or the reverse seal.
Since the principle of operation is the same for all non-mechanical valves,
non-mechanical
valve operation will be presented primarily through a
discussion of the characteristics of the L-valve.
12.10.1

Principles of Operation and Design

When aeration gas is added to a non-mechanical valve, solids do not begin to


flow immediately. There is a certain threshold amount of aeration which must
be added before solids will begin to flow (Fig. 12.32). However, after the

W
I-

<

0::

:3
o
-.J
i.l..
(f)

>~

d
(f)

THRESHOLD
AERATION

AERATION RATE,

m3/min

Figure 12.32 Typical solids flowrate versues aeration rate


curve for a non-mechanical valve.

Figure 12.31 The three most common types of non-mechanical valves:


L-valve, J-valve, and reverse seal.

threshold aeration is reached, additional aeration gas added to the valve


causes the solids flowrate to increase and decreasing the amount of aeration
to the valve causes the solids flowrate to decrease. In general, there is no
hysteresis in the aeration versus solids flowrate curve for a non-mechanical
valve.
Solids flow through a non-mechanical valve because of drag forces on the
particles produced by the aerating gas. When aeration gas is added to a nonmechanical valve, gas flows downwards through the particles and through the
constricting bend. This relative gas-solids flow produces a frictional drag

force on the particles in the direction of flow. When this drag force exceeds
the force needed to overcome the resistance to solids flow in the constricting
bend, the solids begin to flow through the valve.
The solids flow profile through a non-mechanical valve changes with the
solids flowrate through the valve. This is illustrated in Fig. 12.33 for the case
of the L-valve.

~I)I
LOW

t
\)
MEDIUM

J}

t
J\

)\
HIGH
~

STAGNANT

SOLIDS REGION

Figure 12.33 Solids flow pattern through an Lvalve


as a function of solids flowrate.

At low solids flowrates, the solids flow in a narrow zone near the inside of
the constricting elbow. Most of the solids in the elbow are stagnant. Similarly,
only a shallow upper layer of solids is moving in the horizontal section and the
solids appear to be moving in 'ripples' near the top of the pipe.
At the medium solids flowrate shown in Fig. 12.33, the stagnant region of
solids is much reduced. Also, a larger portion of the solids are now moving in
the horizontal section of the L-valve.
At high so Ids flowrates the entire mass of solids in the horizontal section of
the L-valve is in motion.
Non-mechanical
valves work best with Geldart B materials. These
are granular materials in the area shown on the Geldart fluidization diagram

of Fig. 3.4. Group C materials (cohesive materials such as flour) do not work
well in L-valves. In general, group C materials do not flow well in any type of
pipe and are even difficult to fluidize.
Group A materials (materials which have a substantial deaeration time)
also generally do not work well in L-valves. These materials retain air in their
interstices and remain fluidized for a significant period of time and thus often
flow uncontrollably through the L-valve like water.
Group D solids are very large solids with particle diameters generally 1/4 in
(6,350 Mm) or greater. These materials can be made to flow through an Lvalve, but the L-valve has to have a large diameter and much aeration has to
be used. The large Group D solids also work best in an L-valve if they are
mixed in with smaller material. The smaller particles fill up the void spaces
between the larger particles and decrease the voidage. This increases the drag
on the entire mass of solids when aeration is added and makes the solids
easier to move through the constricting bend.
From the above discussion it is seen that L-valves are best utilized with
granular material in the size range between 100 and about 8,000 Mm.
Knowlton and Hirsan (1978) and Knowlton, Hirsan, and Leung (1978)
have shown that the operation of a non-mechanical valve is dependent upon
system pressure drop and geometry. In general, a non-mechanical valve wilf
operate with an underflow standpipe operating in moving packed bed flow
above it. The standpipe in turn is usually attached to a hopper.
Consider the L-valve solids feed system shown in Fig. 12.34 with its
associated pressure drops. The aeration injection location is the high pressure
point in the system; the low pressure common point is at the top of the feed
hopper. The pressure drop balance is such that the sum of the L-valve,
fluidized bed, and piping pressure drops must equal the standpipe and hopper
pressure drops, Le.:
~Pstandpipe

~Phopper

= ~PL.valve

~Pbed

~Ppiping

(12.44)

The hopper pressure drop is usually negligible so that Eq. (12.41) may be
written as:

The standpipe is the dependent part of the pressure drop loop in that its
pressure drop will adjust to exactly balance the pressure drop produced by the
sum of the components on the independent side of the loop. However, there
.!s_~_Il1aximumpressure dr?!, p~r uni~ length (~P{ L) !pat ~h.e movIng'pac~~
('
. bed staIl.<!p.ip~.c~nde~~.19p. ThIS m~xImum value ISthe flUIdIzed bed pressure ~
drop'per unit length, (~P/L)mf,'Jor ~~@il?,
The independent pressure drop can be increased by increasing the bed
pressure drop (increasing the bed height), the piping pressure drop

independent side of the loop at the desired solids flowrate.


length of standpipe necessary Lmin is:

The minimum

L . = .:lPindependent
mln
(.:lPIL)mf
The actual length of standpipe selected for the L-valve design should be
greater than Lmin to allow for the possibility of future increases in the solids
flow requirements and to act as a safety factor. Typical standpipe lengths
might be 1.5 to 2 times Lmin.
Knowlton and Hirsan (1978, 1980) and Knowlton (1979) have also shown
that the total aeration QT, which causes solids to flow through a non
mechanical valve, is not just the external aeration added to the valve. It is the
sum of the external aeration gas QA and the standpipe gas Qdc which
determines the solids flowrate through the valve. Thus, for gas travelling
down the standpipe with the solids:

~PBED

QT = QA
~

6PL-VALVE

~PSTANDPIPE

Qdc

For gas travelling up the standpipe:

+ 6PHOPPER = 6PBED + 6PPlPING + 6f\.-VALVE

QT = QA - Qdc

This difference is shown schematically

in Fig. 12.35.

Que

(increasing the gas flowrate), and the L-valve pressure drop (increasing the
solids flowrate). For a constant gas velocity and bed height, as the solids
flowrate into the bed is increased, the independent half of the pressure drop
loop increases. The standpipe pressure drop adjusts to balance this increase.
If the solids flowrate continues to increases, the standpipe pressure drop will
also keep increasing to balance the pressure drop on the independent side of
the loop. This will occur until the .:lPIL in the downcomer reaches (.:lPIL)mf'
Because of its lesser capacity to absorb pressure drop, a short standpipe will
reach its maximum .:lPIL at a lower solids flowrate than a longer standpipe.
Thus, the maximum solids flowrate through an L-valve depends on the length
of the standpipe above it.
The pressure drop in a standpipe is generated by the relative velocity Vr
between the gas and solids in the standpipe. When Vr reaches the value
necessary for minimum fluidization of the solids, a transition from a packed
bed flow to fluidized bed flow occurs. Any further increase in Vr results in the
formation of bubbles in the standpipe. These bubbles hinder the flow of solids
through the standpipe and thus decrease the solids flowrate.
(To
determine the minimum standpipe length necessary for an L-valve
V\\
\feeding into a fluidized bed, it is:necessary to know the pressure drop on the
~)

Ooc

t
OA

QA

~or

Or;

OA + QDe

~Qr

Or ; QA - QDe

Figure 12.35 Effective aeration flows in an L- valve for gas flowing up and gas flowing down
the standpipe.

To determine the diameter of an L-valve it is necessary to select the linear


solids velocity desired in the standpipe. Nearly all L-valves usually operate
over a linear particles velocity range of 0 to 0.3 mls in the standpipe.
Although an L-valve may be designed for any linear solids velocity, a typical

value near 0.15 mls is generally selected to allow for substantial increases or
decreases in the solids flowrate, if necessary.
Unfortunately, at this time there are no published correlations to predict Lvalve pressure drop and aeration requirements. The best way to determine
these parameters is to estimate them from basic data from a small L-valve test
unit or to extrapolate them from the data given by Knowlton and Hirsan
(1978).

4/6~ ~o spuDsn04~ '31'1l:l M01.:l SOIlOS


to

"-

0,..;

~
c

Example 12.5: L-valve design

The minImum standpipe length and the L-valve diameter for an L-valve
feeding limestone into a fluidized bed from a hopper at a rate of 56,820 kg/h is
to be determined. The configuration for this feed system is the same as that
shown in Fig. 12.34 and:

v
ci

'"

z
(,J

q2
-

!i

a:
z

~
a:
ILl
~

In 'Ot

'"

<DO

oj::
~
a:
"-ILl
c;j~

ILl

>

..

c:

W
0
..
>
.2 ffiO

=.-

(!)~

In

on ci

U(Qvrt)

<D

<co
"'I'"

1-%

'ci" '"0
%

1>0<1

'Ot

i'!

c;j

W= Ph v,;A
0

0
0

'"

Q
4/ql

to

~o spuDsno41 '31'1l:l M01.:l

Ul

'Ot

SOIlOS

OJ

OJ

:>(;j

~
<Il

::l

I
..J

"-

OJ

0;
"-

0;::

I",

.~
d

.g
"eo...

III
II.

"
E

o~

1
v
ci

<Il

c:

.2

<Il

c;j

c:

ti

Ul..J

....

!!i

Ow"
% ":r'"
ct VI
VI ct

ffi ~
tt::O<C
ffi

To determine the L-valve diameter, first select a linear limestone velocity, vs,
in the standpipe. Choose Vs = 0.15 mls. Then:

- 0'

~ ~~.s ~ ~ ~

:.J %
~ 2

= 1,520 x 0.15A

0a:

<!l

56,820
3600
,

"!~

Ijw

ci

CD CI::.--

429 cm

.9

w c:e

00

0.2 + 0.136 + 0.136


0.0011

--

-:JfI'

a ci0'"

~
~

(ilP/L)mf

c:

:'j

E
"
0
OJ

ILl

Lmin

,9

ilPindependent

~
c:

Limestone bulk density = 1,520 kg/m2


ilPL-valve = 0.2 bar
ilPbed
= 0.136 bar
ilPpiping
= 0.136 bar
Limestone fluidized bed density = 0.0011 bar/em

Q,

'E
,.,..

Given:

,9
<;;i
u
.2

<Il

"0

:.::l
0
CI'l
'Q

'"
M

..

~
=lOll
~

1rD2
A = -4-

56,820
3,600 x 1520 x 0.15

= 0.069

4
D2 = (0. ;69) - = 0.088
1r

0.296 m

0.97 ft

30 cm

12.10.2

12.10.3 L-valve Horizontal Length

In order for an L-valve to operate properly, the L-valve horizontal section


length must be kept between a minimum and maximum length. The minimum
length must be greater than the horizontal length through which the solids
flow due to their angle of repose, as shown in Fig. 12.37. Mathematically, the
horizontal section must be greater than L where:

Aeration Tap Location

It is desirable to add aeration to a non mechanical valve as low in the


standpipe as possible. This will result in the maximum standpipe length and
minimum non mechanical valve pressure drop, both factors which result in
increasing the maximum solids flowrate through the valve. However, if the
aeration is added too low to a non mechanical valve, gas by-passing results
and solids flow control is insensitive and not effective.
Knowlton and Hirsan (1978) studied the effect of aeration tap location on
the operation of a 7.6 cm diameter L-valve. They found that aeration was
most effective if it was added a distance of 15.2 em (a length/diameter (LID)
ratio of 2) or more above the centerline of the horizontal section of the Lvalve. Aeration added at the centerline of the horizontal section or at the
bottom of the centerline of the standpipe was found to be ineffective. The
aeration tap locations tested are shown in Fig. 12.36, together with the solids
flow rate versus aeration rate curves for each aeration tap location tested.
As shown in Fig. 12.36, the solids flowrate was increased from zero to its
maximum value by increasing the aeration rate to the L-valve. For taps Band
1 (located at the bottom of the standpipe and flush with the L-valve
centerline, respectively), the solids flowrate increased to a maximum value of
approximately 910 kg/h and then decreased when more aeration was added to
the taps. For taps 2, 3, and 5, the solids flow rate increased with increasing
aeration up to a value of approximately 8,180 kg/h.
The poor performance of the bottom two taps was due to gas by-passing
from the aeration tap to the horizontal section of the L-valve and not being
efficiently utilized to drag the solids through the constricting bend. The
aeration added to the upper aeration taps was fully used to drag the solids
through the bend. Therefore, all of these solids flowrate versus aeration
curves for the upper taps were basically the same.
In summary, L-valve aeration should be added above the constricting bend.
To assure good operation it should be added at an LID of approximately 2
above the centerline of the horizontal section of the L-valve.

V
L=-tan ex:
For design purposes the minimum horizontal length should actually be about
1.5L.
If the L-valve is designed properly, solids flow through it in micro pulses at a
relatively high frequency, and for all practical purposes the solids flow is
steady. For very long L-valve horizontal lengths, the solids flow becomes
intermittent (surging, stopping, then surging again). In these 'long' L-valves,
a dune builds up to such a size that it blocks the pipe and flow stops. Gas
pressure builds up behind the dune and finally becomes so great that the dune
collapses and the solids surge through the L-valve. Another dune then builds
up after the gas pressure is released. This pattern then repeats itself. This
cycle does not occur if the L-valve horizontal section length is kept below that
equal to an LID of about 12.

Knowlton and Hirsan (1978) have determined the effect of varying


geometrical and particle parameters on the operation of the L-valve, which
are summarized below.
L-valve aeration requirements increase with:
(a) Increasing vertical section diameter
(b) Increasing particle size
(c) Increasing particle density

QA

1.\\\

L-valve pressure drop increases with


(0)

(a) Incresing aeration (i.e. increasing solids flowrate)


(b) Increasing particle density
(c) Decreasing horizontal section diameter

W3

W2 >W\

No equations presently exist with which to determine aeration requirements.


Thus, L-valve design must be determined from test rig data or from
extrapolation from the curves of Knowlton and Hirsan (1978).

< WI

H2

12.10.4

Automatic L-Valve

An L-valve can also operate 'automatically' at the bottom of an overflow


standpipe, such as a cyclone dipleg. Automatic operation means that if the
solids flowrate through the standpipe is suddenly changed, the L-valve will
automatically adjust to accomodate the changed flowrate.
Consider the L-valve shown in Fig. 12.38(a) in which solids are flowing
through the L-valve at the rate WI while aeration is being added to the Lvalve at the rate QA' The solids above the aeration point are flowing in the
fluidized bed mode and are at an equilibrium height HI' The solids below the
aeration point are flowing in the moving packed bed mode. The fraction of
aeration gas needed to move the solids at rate WI is flowing around the elbow
with the solids, while the remaining portion of the aeration gas is flowing up
the standpipe.
,]'1 L
h
If the solids flowrate decreases from WI to.W3(Fig. 12.38if, the soiids level
in the standpipe will initially rise because solids are being fed to the L-valve
faster than they are being discharged. However, the increased height of solids
in the standpipe offers an increased resistance to gas flow and so a larger
fraction of the aeration gas QA will now flow around the elbow increasing the
solids flowrate through the L-valve. If enough aeration gas is being added to
the L-valve, the system will reach equilibrium at a point where a larger
fraction of the aeration gas is flowing around the L-valve constriction causing
solids to flow at rate Wz. The height in the standpipe above the L-valve will
reach equilibrium at height Hz greater than HI. and the system is in balance.
If the solids flowrate decreases from WI to W3 (Fig. 12.38c), the solids level

QA

L:

\\\\

~'1\
Ie)

(b)

Figure 12.38 Automatic L-valve operation.

in the standpipe above the L-valve will fall, decreasing the resistance to gas
flow, and a larger fraction of the constant aeration flow QA will pass up the
standpipe. The reduced flow of gas through the L-valve will result in a lower
solids flowrate. Balance is reached where solids are flowing through the Lvalve at rate W3 and the solids in the standpipe are at equilibrium height, H3
Very little work has been done with this type of L-valve system. Findlay
(1981) has shown that this type of L-valve operation is possible, but much
work will have to be done before the solids flowrate range and aeration
limitation over which the L-valve will operate automatically can be estimated
accurately.
12.11

SOLIDS FLOWRATE

ENHANCEMENT

FROM HOPPERS

Ginestra, Rangachari, and Jackson (1980) and Chen, Rangachari, and


Jackson (1984) have developed one-dimensional equations to predict the
relationship between pressure drop, soilds flowrate and control orifice

diameter in a vertical standpipe operating below a conical mass t10w hopper.


They found experimentally that several different flow regimes are possible
and that the flow regimes could be predicted with their mathematical model.
Knowlton (1980) studied the effect of increasing the length of an
unconstrained
standpipe attached to a feed hopper on the solids gravity
f10wrate from the hopper. He found that the solids f10wrate from the hopper
increased as the standpipe length was increased and that the solids f10wrate
increase was more pronounced for small particle sizes than for large particle
sizes (Fig. 12.39).

PRESSURE,

em

H20
o

SAND SIZE: 87 p.m


PIPE DIAMETER: 1.5in(3.8cmJ

f-

0.2

(!J

W
..J

PIPE LENGTH,

z
<

60

f-

..0

50

a 523 fLm

.r.L
.....

173 fLm

0'
0 40
0

lD

o 87 fLm

20

PIPE DIAMETER,

....J

lL

-<:
0::

en

....
0

Cl
W

0.6

(em)

....
..J
N

47 (119)

<

6 135 (343)

L:
~
0

0.8

0 168 (427)

274 (696)

-80

-60

....
d
en

-40

PRESSURE,

0
10 ....J
lL

20

en

Cl
....J

in

X
W
f-

30

-<:

~
~
0

a0

1.5 in.
(3.8 em)

X
W
f-

CI)

PIPE LENGTH,

SAND SIZE

.r.L
.....

W
a..
a..

0.4

Figure 12.40
pipe system.

-20

20

40

in

Pressure profiles in a hopper-unconstrained

stand-

Cl

10

en

0
30

60

80

120

PIPE LENGTH,
Figure 12.39
unconstrained

Gravity solids flowrate


standpipe.

from a hopper

in

as a function of standpipe

length in an

For unrestricted standpipe flow, the solids flow through the pipe at a rate
faster than their discharge from the hopper, and a dilute phase streaming flow
results. The resulting profile in the hopper-standpipe
system with an
unconstrained standipe is shown in Fig. 12.40, for the case where the top of
the hopper and the bottom of the standpipe are both at atmospheric pressure.
When the standpipe is operating in dilute phase flow, a constant pressure
gradient over most of the standpipe length is developed. This produces a low
pressure region at the top of the standpipe immediately below the hopper
outlet. The longer the standpipe, the lower the pressure at the top of the
standpipe.

The decrease in pressure at the top of the standpipe causes the pressure
drop across the hopper outlet to increase. The gas flow through the particles
in the hopper then has to increase, which increases the solids f10wrate through
the hopper outlet.
Chen, Rangachari, and Jackson (1984), Judd and Dixon (1978), and Judd
and Rowe (1978) have also reported similar profiles in unrestricted
standpipes. Chen, Rangachari, and Jackson (1984) have also predicted the
flow profile and the increase in solids flowrate using their one-dimensional
mathematical model.

area
particle size
volumetric particle diameter
tube or pipe diameter
gas-wall friction factor

410

is
Fr
g

Og
Os
Hsp
L

solids-wall friction factor


Froude number, U2/gD
gravitational constant
gas mass flux
solids mass flux
height of solids in standpipe
length
minimum L-valve standpipe length
exponent in the Richardson-Zaki
equation,
UWt =

P
ti.P
ti.PT
ti.PIL
(ti.PIL)mf

QT

Re
Rep
Sa
Ub

U
Uch
Usalt
Uso
Vch

Vg
Vmb

9.81 m/s2
kglm2 s
kglm2s
m
m
m

F!'

pressure
pressure drop
total pressure drop
pressure drop per unit length
pressure drop per unit length at minimum
fluidization
amount of aeration gas added to a non
mechanical valve
amount of gas flowing in the bl.lbble phase
amount of gas flowing in the dense phase
amount of gas flowing down or up a standpipe
relative to the sandpipe wall
amount of gas flowing around the L-valve bend
pipe Reynolds number, pgDUIp.
particle Reynolds number, pgdpUl p.
exponent indicative of the 'powderiness' of a solid
bubble rise velocity
superficial gas velocity
choking velocity, superficial
superficial saltation velocity
single particle saltation velocity
interstitial choking velocity, Uch/E
interstitial gas velocity, UIE
interstitial minimum bubbling velocity, Umbl E

N/m2
N/m2
N/m2

N/m3

rnIs
rnIs
rnIs

m/s
m/s
m/s
m/s
rnIs

m/s

Vr

interstitial minimum fluidization velocity, U mil E


relative gas solids velocity

Vs

solids velocity

m/s

Vmf

rnIs

slip velocity between solids and gas


particle terminal velocity
solids flow rate
angle of repose
pseudo-length used in the Zenz saltation
correlation
voidage
voidage at choking
void age at minimum fluidization
loading factor used in Konno and Saito correlationgas viscosity
solids bulk density
gas density
particle density
sphericity of a solid particle
pseudo-velocity used in the Zenz saltation
correlation

m/s
m/s
kgls

N s/m2
kg/m3
kg/m3
kg/m3

m/s

Arastoopour, H., Modi, M.V., Punwani, D.V., and Talwalkar, A.T. (1979). Paper
presented at International Powder and Bulk Solids Conference/Exhibition,
Chicago.
Batholomew, R.N., and Casagrande, R.M. (1957). Ind. Eng. Chem., 49, 428--431.
Berg, C.H.O. (assigned to Union Oil Company of California) (1954a).
Berg, C.H.O. (assigned to Union Oil Company of California) (1954b). US Patent
2,684,870.
Berg, C.H.O. (assigned to Union Oil Company of California) (1954c) U.S. Patent
2,684,872.
Berg, C.H.O. (assigned to Union Oil Company of California) (1954d). US Patent
2,684,873.
Bodner, S. (1982). Proceedings of Pneumatech I, International Conference on
Pneumatic Conveying Technology, Powder Advisory Centre, London.
Campbell, D.L., Martin, H.Z., and Tyson, C.W. (1948). US Patent 2,451,803.
Canning, D.A., and Thompson, A.I. (1980). Paper presented at the Annual Meeting
of the American Institute of Mechanical Engineers, San Francisco.
Capes, C.E., and Nakamura, L. (1973). Can. J. Chem. Eng., 51, 31-38.
Chen, Y.M., Rangachari, S., and Jackson, R. (1984). To be presented in Ind. Eng.
Chem. Fund.
Culgan, J.M. (1962). Ph. D. Thesis, Georgia Institute of Technology.
Curran, G.P., and Gorin, E. (1968). Prepared for Office of Coal Research, Contract
No. 14-01-0001-415, Interim Report No.3, Book 6, Washington, DC.
Davidson, J.F., and Harrison, D. (1963). Fluidized Particles, Cambridge University
Press.

Institute of Gas Technology (1978). Coal Conversion Systems Technical Data Book,
Prepared for the US Department of Energy by the Institute of Gas Technology
under Contract No. EX-76-C-Q1-2286.
Doig, 1.D., and Roper, G.H. (1963). Australian Chern. Eng., 1, 9-19.
Dries, H.W.A. (1980). In Fluidization Eds J.R. Grace and J.M. Matsen, Plenum
Press, New York.
Eleftheriades, C.M., and Judd, M.R. (1978). Powder Technol., 21,217227.
Engineering Equipment Users' Association (1963). EEUA Handbook, No. 15,
Constable, London.
Findlay, J. (1981). Personal communication.
.
Flatt, W. (1976). Paper 43, Proceedings of Pneurnotransport 3, BHRA FlUid
Engineering, Bath.
.
Gajdos, L.J., and Bierl, T.W. (1978). Topical report prepared by CarnegIe-Mellon
University.
Geldart, D., (1973). Powder Technol, 7,285.
Ginestra, J.e., Rangachari, 1., and Jackson, R. (1980). Powder Technol., 27, 69.
Grace, J.R., and Tuot, J. (1979). Trans. Instn. Chern. Engrs., 57, 47.
Hinkle, B.L. (1953). Ph.D. Thesis, Georgia Institute of Technology.
Jones, P.L, and Leung, L.S. (1978). Ind. Eng. Chern. Proc. Des. & Dev., 17, 4.
Jones, P.L, Teo, C.S., and Leung, L.S. (1980). In Fluidization (Eds J.R. Grace and
J.M. Matsen), p. 469.
Judd, M.R., and Dixon, P.D. (1978). A.I.Ch.E. Syrnp. Ser., 74 (176), 38.
Judd, M.R., and Rowe, P.N. (1978). In Fluidization (Eds J.F. Davidson and D.L.
Keairns), Cambridge .University Press, p.110.
Jung, R. (1958) Forschung Ing. Wes., 24 (2), 50.
.
Knowlton, T.M. (1977). Institute of Gas Technology Project IU-4-8 Annual Report
for 1977, Sponsored by American Gas Association.
Knowlton, T.M. (1979). Proceedings of Powder and Bulk Solids Conference/
Exhibition, Philadelphia, Pennsylvania.
Knowlton, T.M. (1980). Proceedings of the 6th International Conference on FluidizedBed Combustion, Atlanta, Georgia.
Knowlton, T.M., Aquino, C., Hirsan, 1., and Sishtla, C. (1981). A.I.Ch.E. Syrnp.
Ser., 77 (205), 189.
Knowlton, T.M., and Bachovchin, D.M. (1976). In Fluidization Technology (Ed.
D.L. Keairns), Hemisphere Publications Corporation.
Knowlton, T.M., and Hirsan, 1. (1978). Hydrocarbon Processing, 57,149-156.
Knowlton, T.M., and Hirsan, 1. (1980). Paper E3, Proceedings of Pneurnotransport 5,
BHRA Fluid Engineering, London, p. 257.
Knowlton, T.M., Hirsan, 1., and Leung, L.S. (1978). In Fluidization (Eds J.F.
Davidson, D.L Keairns), Cambridge University Press.
Kojabashian, C. (1958). Ph.D. Thesis, Massachusetts Institute of Technology,
Cambridge, Massachusetts.
Konno, H., and Saito, S.J. (1969). Chern. Eng. Japan, 2,211-17.
.
Leung, L.S., and Jones, P.J. (1976). In Fluidization Technology (Ed. D.L. Kemrns),
Vol. II, Hemisphere Publishing Corporation, Washington, De..
.
.
Leung, L.S. (1980). Proceedings of Pneurnotransport, 5, BHRA FlUIdEngmeenng,
London.
Leung, L.S., and Jones, P.J. (1978a). Paper Dl, Proceedings of Pneurnotransport 4
Conference, BHRA Fluid Engineering.
Leung, L.S., and Jones, P.J. (1978b). Powder Tech., 20,145--160.
Leung, L.S., and Jones, P.J. (1978c). In Fluidization (Eds. J.F. Davidson and D.L.
Keairns), Cambridge University Press.

Leung, L.S. and Wiles, R.J. (1976). Paper C4-47-58 , Presented at Pneumotransport
3 Conference, University of Bath.
Leung, L.S., Wiles, R.J., and Nicklin, D.J. (1971a). Ind. Eng. Chern. Proc. Des. &
Dev., 2, (10), 183-189.
Leung, L.S., Wiles, R.J., and Nicklin, D.J. (1971b). Paper presented at Pneumotransport 1 Conference, Churchill College, Cambridge.
Leung, L.S., and Wilson, L.A. (1973). Powder Techno!., 7, 343.
Marcus, R.D., Hilbert, J.D., Jr., and Klinzing, G.E. (1984). Pipelines, 4, 103-112.
Matsen, J.M., (1973). Powder Technol., 7, 93.
Matsen, J.M. (1976). In Fluidization Technology (Ed. D.L. Keairns), Vol II,
Hemisphere Publishing Corporation, Washington, DC, p. 135.
Matsen, J.M. (1981). Proceedings of the 73rd Annual AIChE Meeting, New Orleans.
Matsumoto, S., Hara, M., Saito. S., and Maeda, S. (1974a). J. Chern. Eng. Japan. 7
(6), 425-30,
Matsumoto, S., Harada, S., Saito, S., and Maeda, S. (1974b). 1. Chern. Eng. Japan. 8,
331.
Mewing, S.F. (1976). B.Eng. Thesis, University of Queensland.
Morikawa, Y., Tsuji, Y., Matsuik, J., and Hani, Y. (1978). Int. J. Multiphase Flow, 4,
575.
Patterson, R.e. (1959). ASME J. Engineering for Power, 1959, 43-54.
Punwani, D.V., Modi, M.V., and Tarman, P.B. (1976). Paper presented at the
International Powder Bulk Solids Handling and Processing Conference, Chicago.
Richards, P.C., and Wiersma, S. (1973). Paper A1-1-15, Presented at Pneumotransport 2 Conference, Guildford.
Rizk, f. (1973). Doktor-Ingenieurs Dissertation, Technische Hochschule Karlsruhe.
Rose, H.E., and Duckworth, R.A. (1969). Engineer, 227,392-396,430-433,478-483.
Sandy, e.W., Daubert, T.E., and Jones, J.H. (1970). Chern. Eng. Prog. Syrnp. Ser.,
66, 105, 133-142.
Sauer, R.A., Chan, 1.H., and Knowlton, T.M. (1984). A.I.Ch.E. Syrnp. Ser., 80
(234), 1-23.
Saxton, A.L., and Worley, A.C. (1970). Oil and Gas. J .. 1971,92-99.
Schuchart, P. (1968). 1. Chern. E. V.D.1.N.Y.G. Symposium on the Engineering of
Gas-Solids Reaction, Brighton.
Scott, A.M., Richard, P.D., and Mooij, A. (1976). Paper DI0, presented at
Pneumotransport 3 Conference, Bath.
Smith, T.N. (1978). Chemical Eng. Sci., 33, 745.
Thomas, D.G. (1962). A. U. Ch.E.J., 8 (3), 363-378.
Tsuji, Y. (1980). Paper H2, Proceedings of the 5th International Conference on the
Pneumatic Transport of Solids in Pipes, BHRA Fluid Engineering, Cranfield,
Bedford, pp. 363-388.
Van Zoonen, D. (1961). Proceedings of the Symposium on the Interaction between
Fluids and Particles, Inst. Chemical Engineers, London, p.64.
Weidner, G. (1955). Forschung Ing. Wes., 21 (5), 145.
Wen, C.Y., and Galli, A.F. (1971). In Fluidization (Eds J.F. Davidson and D.
Harrison), Academic Press, London.
Wen, e.Y. and Simons, H.P. (1959). A.I.Ch.E.J., 5 (2), 263-267.
Woebcke, H.N., and Cofield, W.W. (1978). Coal Processing Technol., 3, 79.
Woodcock, e.R., and Mason, J.S. (1976). Paper El, Pneumotransport 3 Conference,
Bath.
Yang, W.C. (1975). A.I.Ch.E.J.,
21, 1013-1015.
Yang, W.e. (1976a). Paper presented at Pneumotransport 3 Conference, Bath.

Yang, w.e. (1976b). Paper presented at the International Powder and Bulk Solids
Handling and Processing Conference/Exhibition, Chicago.
Yang, W.e. (1982). Paper presented at the 74th A.I.Ch.E. Annual Meeting, Los
Angeles, California.
Yerushalmi, J., Turner, D.H., and Squires, A.M. (1976). Ind. Eng. Chern. Proc. Dev.
& Dec., 15, 47-53.
Yousfi, Y., and Gau, G. (1974). Chern. Eng. Sci., 29, 1939-1946.
Zenz, F.A. (1964). Ind. Chern. Fund., 3, 65-75.
Zenz, F.A. (1984). Paper presented at Katalistiks Symposium, Vienna, Austria.
Zenz, F.A., and Othmer, D.F. (1960). Fluidization and Fluid-Particle Systems,
Reinhold, New York.
Zenz, F.A., and Rowe, P.N. (1976). In Fluidization Technology (Ed. D.L. Keairns),
Vol. II, Hemisphere Publishing Corporation, Washington, DC, p.151.

Gas Fluidization Technology


Edited by D. Geldart
Copyright 1986 John Wiley & Sons Ltd.

Instrumentation and
Experimental Techniques

The behaviour of fluidized beds is influenced by many factors - particle


properties, gas properties, column geometry, and operating conditions.
Two powders which appear to be similar may behave quite differently in the
same apparatus while the same powder may fluidize in a different manner in
another piece of equipment.
In developing successful fluidization processes, it is essential to minimize
scale-up problems by forethought at the experimental stage. While there is no
general route to success, there are certain guidelines which should be
followed to perform consistent and reliable experiments. We attempt to
provide some useful guidelines and to provide typical or useful referencei in
this chapter. While we believe that the guidelines will be useful under most
circumstances, we caution the reader that each fluidized bed process has
unique elements and special problems which may not be adequately covered
by the general treatment given here.

In order to design a fluid bed system, the engineer should be fully familiar
with the characteristics of the powder to be handled. As a first step, one
should check the Geldart classification of the powder (see Chapter 3). As a
minimum requirement,
hydrodynamic
equivalence between laboratory
powders and process materials requires that they be in the same classification.
The extrapolation of results for material in one powder group to another is
not advised.
Particle size and density are, however, not the only important parameters.

415

Variations in particle shape, moisture content, entrained air, porosity,


surface configuration, etc., make it impossible to design a standard system to
suit all requirements. Some factors which influence design are listed in Table
13.1. For each specific powder, extensive experimentation is preferred to
extrapolation, if time and money permit.

Powder parameter

Main design influence

Particle size and density

Aerated specific weight


Size distribution as given by
sieve analyses; density
distribution
Particle shape
Aeration or deaeration
characteristics
Temperature limitations
Moisture content and/or
sensitivity
Resistance to attrition

Agglomeration or cohesive
properties

Gas flow requirements, key


elements in many correlations,
powder group
Hopper volumes, feeder capacities
Elutriation, dedusting equipment,
seals, segregation, feeding method
Factor only if extreme (e.g. flaky
or very angular)
Requirements for deaeration, hopper
design, standpipe flow
Heat transfer surface, insulation
needs
Feeder design, gas velocity,
residence time in bed.
Carryover, distributor design,
cyclones
Materials of construction, placement
of heat transfer surfaces, distributor design
Segregation, use of stirrers, means
of withdrawing product, gas fiowrate,
avoidance of internal surfaces.

13.2.1 Components
The components of the gas supply system are shown schematically in Fig.
13.1. The blower or compressor must be capable of supplyingsufficientgas to
the column to give a superficial velocity of at least 1 to 2 mls and preferably 3
m/s. For studies of circulating beds and fast fluidization, velocities of 5 to 12
mls should be attainable. In the latter case, the air should be split into a
primary and secondary stream whose relative flows are in the range of

.1

.1

..
1

1
1
1

approximately 5:1 to 1:3, and minor amounts of air will also be required for
aeration of recirculating solids. The primary pressure losses in the system are
those across the bed of particles, the distributor, and any valves. Allowance
should be made for losses across cyclones, filters, and flow measuring devices
as well as pipes, fittings, and noise-abatement devices.
Pressure pulsations arising from the blower can alter fluidization behaviour. For many purposes, elements giving pressure drops upstream of the
bed provide sufficient damping of these pulsations. A reservoir may be used
to provide further damping, as shown in Fig. 13.1. A filter should be installed
to remove oil and water droplets. A surge tank and filter are not required if
the blower or compressor is replaced by one or more pressurized gas
cylinders. However, these gas bottles are only practical for small columns and
limited running times.
Shown next on Fig. 13.1 is a loop which permits the incoming gas to be
humidified or dehumidified to a desired level. In cases where humidity
control is not necessary (e.g. for large dry particles), it may be possible to do
without part or all of this loop. The pressure regulator shown in the diagram
is set at the calibration pressure of the rotameters. To minimize fluctuations
in float level due to pressure oscillations in the fluidized bed, the calibration
level should be as high as possible, consistent with the capability of the blower
or compressor. The rotameter valves should be downstream as shown in Fig.
13.1, so that the rotameter calibration can be used directly without having to
correct for pressure.

13.2.2

'Two-dimensional' columns are of rectangular cross-section, the width being


considerably greater than the thickness. The fluidized particles are contained
in the gap between two flat transparent faces, separated by a distance which is
usually in the range 10 to 25 mm. A schematic diagram of a typical set-up
appears in Fig. 13.2.

Leak testing

There are two ways to evaluate gas leaking under pilot plant conditions. The
pressure drop method involves completely sealing the unit which is then
pressurized. After turning off the gas, the decrease in pressure is noted as a
function of time. Alternatively, in the pressure hold-up method, the rate of
gas addition to maintain the unit at the required pressure is recorded.
To locate points of leakage, the soap bubble method is usually employed.
Special gas tracers can also be used and localized with special detectors. To
test equipment components for leakage of process gases, other methods are
available such as the use of special emission detectors.
Aside from the fluidization column, process valves account for most
leakage; safety relief valves and sample points should be carefully checked.
Leakage from piping and flanges is generally minor. The remaining sources of
leakage are the blower or compressor, dedusting equipment, etc. Leak
tightness becomes more critical as the size and density of the powder
decrease, since a small leak has a much greater proportional effect for
powders of low Umf

Dimensions and uses of a number of two-dimensional columns which have


been reported in the literature are given in Table 13.2. There is no standard
column size, with reported thicknesses as small as 5 mm and as large as 63 mm
and heights up to 7.3 m. Table 13.2 shows that a wide variety of experiments
have been carried out in two-dimensional columns.
The two-dimensional column has proved to be especially useful:
(a) For studies of bubble properties. Bubbles span the bed thickness and
hence are readily viewed with the aid of back lighting.
(b) As an educational tool in helping those unfamiliar with fluidized beds to
visualize and understand basic phenomena.
(c) As a means of qualitatively
given powders.

viewing the fluidization

characteristics

of

While two-dimensional columns are useful for qualitative purposes, there are
important quantitative
differences between two- and three-dimensional
fluidized beds. These arise from quantitative differences in rise velocities of

isolated bubbles, different bubble coalescence properties, differences in


bubble shapes and wake characteristics, increased jet stability in two-'
dimensional columns, different mechanisms of solids ejection into the
freeboard, and reduced solids mixing. Rowe and Everett (1972) used X-rays
to study the influence of the thickness of a rectangular column of 0.30 m
width. The thickness was increased in stages from 14 mm (a two-dimensional
column) to 0.30 m (a fully three-dimensional column). For the range of
conditions studied, there was a significant change in bubble properties for
thicknesses less than about 0.10 m, with thickness having little discernible
influence once this value was exceeded, The minimum column thickness
(Lyall, 1969) is 30 mean particle diameters for spherical particles, but this
ratio should be somewhat greater when the particles are angular, sticky, or
have a broad size distribution.
Some factors to be considered in designing and building two-dimensional
columns are as follows:

..,.

'C

""
~:O~~og
(""",
x -

B("I"'l

ON
CN"""

~"'(/)~~~tr)~ _8

\(")("'I~~lr}

'"
_z~8
x

C"'!V"l

XV1XV")V1X

OO-X

xZZXXXXX~X

-N:x:("'-i:r""l

XXI"'"!

"1"XXI.Q("")V')("')("'jX_

q.-;XX~""';lI!~

"":NV'"lO"':"";"';""':_-n

("Itn-

..r.X"':XX.o

XXNNX("'.lXX
....
X--

Xv: XIJ"lV: X

X X X X X LI1X

~t.r.OO~
..c(""">""'J'"M("""';OC

Alr)~~~g~~

(a) Materials of construction. The front and rear faces should be transparent
and are generally of plastic or armour glass. Plastic materials are cheaper
and easier to work with, but are more subject to scratching, discoloration, and electrostatic effects. The spacers separating the front and rear
plates may be of any material, but it is best that the entire inside surface
of the column be of uniform roughness and electrical properties. Special
construction techniques are required to make two-dimensional columns
suitable for studying fluidization at elevated pressures (Subzwari, Clift,
and Pyle, 1978; Varadi and Grace, 1978; Kawabata et al., 1981).
(b) Distributor and windbox. The column should be supported in such a way
that the windbox-distributor
assembly can be removed in order to
replace the distributor, repair leaks, or make other modifications. The
most popular types of distributor are perforated plates, porous plates,
and supported filter paper or cloth. The pressure drop across the
distributor should be designed to be approximately 10 to 50 per cent. of
that across the bed of particles. In order to improve the uniformity of gas
flow across the distributor, the wind box maybe partitioned, with the flow
to each section of the distributor measured and controlled separately. It
is essential to provide an effective seal around the outside of the
distributor plate to prevent gas from short-circuiting up the walls. A
pressure relief valve may be installed in the windbox to prevent the
pressure from exceeding its design maximum. Alternatively, a V-tube
manometer can be used for this purpose, with the manometer fluid
discharged if the pressure in the windbox is excessive.
(c) Alignment.

The column must be perfectly vertical

(d) Reinforcement. The front and rear faces tend to bulge, due to hydrostatic
pressure forces inside the fluidized bed. Consequently, it is usually

(e)

(f)

(g)

(h)

(i)

necessary to install horizontal metal supports at intervals to keep the


faces parallel. Spacing of these supports depends on the thickness and
stiffness of the walls, the width and depth of the bed, and the tolerance
level of distortion.
Electrostatic 'effects. Because of their lar~ surface area, twodimensional columns are more subject to electr'6static charging effects
than normal fluidized beds. In some cases, this rl"ay cause a layer of
particles to build up on both faces, totally obscu~g the phenomena
under study. General techniques for reducing electrostatic effects are
discussed in Section 13.3.7 below. A technique which is often useful for
two-dimensional columns is to cover all internal faces of the column
(including the spacers) with a layer of Cellotape or Scotchtape.
Ports. Provision should be made for emptying the bed, usually by
installing a port of inside diameter at least 20 mean particle diameters just
above the distributor. The layer of solids which cannot be ~racted by
opening this port and fluidizing the bed can be removed by tipping the
entire column or by vacuum cleaning. Pressure taps should be installed
just above the distributor and at regular intervals (typically 0.1 to 0.2 m)
above the grid. The taps are usually drilled through one of the spacers in
order not to interfere with viewing. If ports are required for bubble
injection, gas sampling, etc., these are best installed before the column is
erected. Injection ports should be at least 40 mm above the distributor.
Tubes should not protrude beyond the inner surface of the column.
Opening the column. The column should be constructed in such a way
that one of the faces can be removed for modification, adding or
replacing surface coatings, cleaning, etc. The column must be leak tested
carefully after each opening.
Photography. Sufficient space should be left behind the column for
placing floodlights to give even illumination. Tracing paper attached to
the outside back surface helps to give an evenly bright backdrop. There
should be a clear viewing area in front, extending 8 m or more if possible
to allow photography with negligible distortion.
Vibration. Gentle vibration or tapping is sometimes required to initiate
fluidization or to prevent bridging effects. Bridging, if not prevented,
may cause dead zones or lead to instability of injected bubbles.
13.3.2

Figure 13.3 .Schematic diagram


cylindrical column.

of semi-

full cylindrical column. In this case, front illumination is required for bubble
viewing and photography, since bubbles generally do not span the entire bed
thickness. *
While this geometry is very popular and generally reliable in spouted bed
studies (Mathur and Epstein, 1974; Whiting and Geldart, 1980; Geldart et
al., 1981), its use for fluidized beds has been restricted. Hatate, King, Migita,
and Ikari (1985) describe its use for bubble visualization. Gabor (1972)
employed this geometry in some heat transfer work. Levy et al. (1983) studied
particle ejection mechanisms at the bed surface in a half-column. Yang et al.
(1984) have studied jet penetration and bubble properties in a large (3 m
diameter) semicircular vessel with a conical grid. However, Rowe, MacGillivray, and Cheesman (1979) found that flat surfaces alter the behaviour of
grid jets as compared with the case where the jet is centred in a fully
cylindrical vessel. Hence, there is some doubt about how accurately
semicylindrical columns represent the behaviour in fully cylindrical vessels.

Semicylindrical (or Half-) Columns

A column geometry which is intermediate between a two-dimensional corumn


and a cylindrical three-dimensional geometry consists of a cylindrical vessel
which has been sliced in half, with a flat plate installed along the diametral
plane. A diagram showing this geometry appears in Fig. 13.3. The idea is that
one should view, through the flat face, what happens across the diameter of a

13.3.3

Laboratory and Pilot Scale Three-Dimensional Columns

Most laboratory and pilot scale fluidized beds are cylindrical. Square and
rectangular cross-sections are also quite common, especially when horizontal
'Bubbles sometimes move away from the face, thus negating the purpose of the unit. It is
therefore useful to slightly tilt the unit so that bubbles 'run up' the face.

tubes are required. The full cylindrical column has the virtues of simplicity
and structural strength. Standard steel, glass, and perspex tubes are readily
suited to construction of small-scale columns.
Although early work in the field of fluidization was often performed in
smaller diameter columns, it now seems to be generally accepted that there is
a minimum diameter of about 0.10 m below which results tend to be
misleading. Smaller columns almost always give rise to slugging, while
bridging and electrostatic effects may also be severe. In no case should the
column diameter be less than 100 times the mean particle diameter. The
upper extreme of laboratory column diameter is about 0.6 m. This is also a
useful size for pilot plant units, large enough to allow useful data to be
obtained and to aIlow a man to work inside, and yet small enough to avoid
excessive fabrication, materials handling, and blower costs. The influence of
bed diameter on bubble size and the onset of slug flow are treated in
Chapter 4.
The usual scale-up or scale-down procedure for fluidized beds is to keep the
height fixed while varying the diameter (and hence the aspect ratio) (Grace,
1974). The distributor of the smaller column should be identical (i.e. a
portion of that from the full-scale unit rather than a scaled-down replica).
Transparent
sections or windows should be included if possible. The
windbox-distributor
assembly should be removable, and safety valves should
be provided in case of excessive pressures. A port should be provided just
above the distributor for emptying the solids, and pressure taps should be
installed at 0.1 to 0.2 m intervals. The height of fluidization columns varies
widely, from approximately 1 m for the smallest laboratory units to 20 m or
more for pilot plants and full-scale vessels. Entrainment may be reduced by
providing a tapered, expanding freeboard. The angle of taper must be greater
than the angle of repose so that solids do not lodge on the sloping section.
Space restrictions and the need for accessibility usually dictate that cyclones
in small-scale units be external rather than internal. The superficial velocity
range over which the cyclone operates effectively can be extended by
installing a hinged flapper at the inlet to the cyclone to vary the inlet width
and gas velocity. SmaIl diameter multiclones are available commercially.
Collected solids should be returned to the bed just above the distributor, with
the standpipe to be not less than 50 mm in diameter. Circulating bed columns
should be at least 6 m tall, with the solids returned to the main column by a
solids seal, e.g. by an L-valve, J-valve, or (slow) fluidization section.
13.3.4

Wall Roughness

Surface roughness may affect fluidization behaviour, especially in the slug


flow regime (Ormiston, Mitchell, and Davidson, 1965; Geldart, Hurt, and
Wadia, 1978) or for immersed tubes (Yates et al., 1984). Roughening
probably leads to a condition approaching 'no slip' at the boundary, whereas

particles can slide or roll freely along smooth walls. As a result, roughened
walls can promote arching or bridging of particles.
Since wall effects are unlikely to be significant in large columns, laboratoryscale columns should be smooth walled. Offsets, protruding tubes, recesses,
etc., which could act as wall roughness elements, should be minimized.
13.3.5

High Pressure Columns

There has been considerable interest in recent years in fluidized beds


operating at elevated pressures because of improvements in the homogeneity of fluidization, increased gas throughputs, and chemical equilibrium
considerations.
'High pressure fluidization' generally refers to pressures
greater than 5 bar, with the range 10 to 80 bar receiving most attention.
Details of columns used in some high pressure studies are given in Table 13.3.
We have not included details of pressurized fluidized bed combustion (PFBC)
pilot plants which are summarized in a recent report (Grace, Lim and Evans,
1983).
Visual observation of beds operating under high pressure has been
achieved by:
(a) Thick-walled or toughened transport materials (de Carvalho, 1976;
Subzwari, Clift, and Pyle, 1978; Varadi and Grace, 1978)
(b) X-ray photography

(de Carvalho, 1976; Rowe et ai, 1984)

(c) Enclosing a thin-walled transparent column in a pressurized steel vessel


equipped with sight glasses (Knowlton, 1977; Chan and Knowlton, 1984;
Chitester et al., 1984)
Gas may be supplied either via a compressor or from pressurized cylinders.
Special care is required in venting the exit air to atmosphere to prevent
excessive noise and because particle entrainment increases with increasing
pressure.
All instruments must be capable of withstanding the design
pressure. For example, measuring the pressure drop across the bed may
require construction of special manometers because of the high absolute
pressures. In addition, special precautions may be required to prevent
backflow of solids into the windbox during pressurization of the column.
Safety standards must be rigidly adhered to when designing and operating
high pressure equipment.
13.3.6

13.3.7

Elimination of Electrostatic Effects

The following conclusions can be drawn from the literature (Loeb, 1958;
Harper; Richardson and McLeman, 1960; Soo, Ihring, and El Kouh, 1960;
Torobin and Ga.uvin, 1961; Boland and Geldart, 1971/72):
(a) Particulate systems exhibit charging as a result of particle-particle
and
particle-wall rubbing. The latter becomes more important for units of
large wall area (e.g. two-dimensional beds) and fine particles. Both
positive and negative charged particles are present and there is some
evidence that the air stream may hold a balance of charge.
(b) During fluidization, strong electrostatic forces are likely to be set up in
beds of dielectric particles (see Lewis, Gilliland, and Bauer, 1949; Miller
and Logwinuk, 1951; Davies and Robinson, 1960; Ciborowski and
Wlodarski, 1962; and Katz and Sears, 1969). The charges may alter the
fluidization characteristics leading to increased pressure drops, decreased
bed expansion, and a general trend towards aggregative behaviour or
even channelling conditions.
To eliminate static electrification

effects, differenttechniques

can be used:

(a) The use of humid air (60 to 70 per cent. relative humidity) has been
shown to increase the conductivities of insulators. Generated charges in
the fluidized bed are thus neutralized or conducted to earth.
(b) Insulating surfaces can be treated
Gabor, 1967) or with a layer of
dimensional columns. Electrically
metallic coatings are now available

with conductive films (Forest, 1953;


tape, as discussed above for twoconducting plastics and transparent
commercially.

(c) Air may be ionized by electrical discharge or by radioactive sources


(Henry, 1953; Loeb, 1958). Anti-static sprays available commercially
may also be added periodically or continuously to the fluidizing gas
upstream from the distributor.
13.4

GAS DISTRIBUTORS FOR LABORATORY-AND


PILOTSCALE COLUMNS

At one time it was common practice to use porous plates (e.g. sintered
bronze, sintered plastic, or woven stainless steel) as distributors in laboratoryscale fluidized beds. This practice has declined for the following reasons:

Columns Operating Under Vacuum

There has been little work on fluidization under vacuum. Zabrodsky,


Antonishin, and Parnas (1976) report some Russian work on heat transfer
under vacuum. Germain and Claude I(1976) describe a set-up which was used
to study fluidization at pressures less than 30 torr.

(a) Commercially available sintered plates tend to have variations of


thickness or permeability which lead to dead regions and regions of high
flow. Improvements can sometimes be made on an ad hoc basis e.g. by
adding a coating layer of glue to reduce flow through overly active
regions, but this is time-consuming and uncertain.

(b) These materials tend to be expensive, fragile, and difficult to machine.


Pores may also become blocked by entrained dust or oil.
(c) Results are unlikely to be representative of large-scale fluidized beds
since porous plates give rise to a large number of very small bubbles while
distributors used in commercial operations (see Chapter 4) tend to give
jets and much larger initial bubble sizes.
Drilled or punched plates are often employed in laboratory-scale columns.
The pressure drop across a perforated plate distributor, t.PD, can be
calculated from the equation:

. VA

CdnorAor

j2t.P
--

Pg
where Cd is a discharge coefficient which can usually be assigned the value of
0.6 (Kunii and Levenspiel, 1969). The ratio of t.po to the pressure drop
across the bed should be in the range of 0.1 to 0.5, with the higher value
advisable for materials which are difficult to fluidize, shallow beds
(HID < 0.5), and applications which are sensitive to bubble size such as
certain chemical reactions. For most' purposes, a value of 0.3 is quite
sufficient. An equation for calculating the ratio of distributorlbed pressure
drop has been proposed by Sathiyamoorthy and Rao (1981). The holes should
be uniformly distributed across the plate. The size of hole is governed by a
compromise between competing factors:
(a) Small orifices, especially those smaller than about 1 mm in diameter, are
difficult and expensive to drill or punch.
(b) Better performance is usually given by a large number of small holes than
a small number of large holes having the same free cross-sectional area.
The usual hole diameter adopted in experimental fluidized beds is in the
range 1 to 6 mm.
Fakhimi and Harrison (1970) proposed a criterion for keeping all grid holes
unblocked and in operation. In small-scale units, hole blockage and backflow
of solids can be prevented by stretching a piece of cloth or wire mesh over the
perforated plate. The cloth or screen is sometimes put on the underside of the
plate. This prevents backflow of solids, but is not as effective in preventing
hole blockage.
Other types of distributor which may be useful in the laboratoy include:
(a) Filterpaper, cloth, or screens sandwiched between commercially available punched plates, the plates typically having a free cross-sectional area
of about 30 per cent.
(b) A shallow packed bed of heavy particles (e.g. lead shot) supported on a
screen or on a plate of high free cross-sectional area. For the advantages
of this type of distributor, see Hengl, Hiquily, and Couderc (1977).

Use of a low pressure, drop screen as the sole gas distributor is not
recommended. Similarly, single-orifice plates should not be used except for
spouted beds.
When a small-scale pilot-scale unit is to be the basis for scale-up, the
distributor should be a full-size portion of that in a larger unit, not a scaleddown version. For design of grids for large-scale fluidized beds, see Chapter 4
and the recent paper by Geldart and Baeyens (1985).

13.5.1

Photographic Techniques

Cine photography has been one of the chief techniques which has been used to
investigate bubble behaviour and particle motion in fluidized beds. Four
different set-ups have been employed:
(a) Two-dimensional columns, generally viewed with the aid of back lighting
as shown in Fig. 13.2
(b) X-ray or y-ray photography of three-dimensional
and Carpenter, 1971; Rowe, 1971)

columns (e.g. Orcutt

(c) Three-dimensional columns photographed from the side through the wall
or semi-cylindrical columns (see Section 13.3.32) photographed using
front lighting
(d) Three-dimensional columns photographed from above with the aid of top
lighting
Each of these has its limitations. Two-dimensional
fluidized beds give
quantitative differences from three-dimensional beds, as discussed in Section
13.3.1. For X-ray photography, the maximum bed thickness is of order 0.2 m,
and gas flowrates must be modest to allow identification of individual
bubbles. Normal photography of three-dimensional beds only yields information about the behaviour at the wall, which is generally not representative
of behaviour in the interior. Photography from above requires low gas
flow rates in order that the bed surface should not be obscured by particles
ejected into the freeboard. Despite these limitations, each of these methods
has yielded some useful data within the constraints mentioned (Lyall, 1969).
For bubbling fluidized beds, bubble and maximum particle velocities are of
order 1 m/s. To obtain resolution within 1 mm therefore requires a shutter
speed of about 10-3 s or better: Most cine photography of bubbling fluidized
beds has been at speeds of 24 to 80 frames/s, with frame exposure times
ranging from 0.02 to 0.002 s. Hence, there is some inherent ambiguity in
defining the position of individual particles or bubble boundaries. Sharper
resolution can be obtained with high speed photography.
High speed
photography is also required to view high velocity beds where superficial
velocities exceed 3 m/s.

.J'>.
VJ

Table 13.4

Reference
Gerald

Example

Type
Piezoelectric

(1951)

crystal

Hot wire

Morse and Ballou (1951)


Bakker and Heertjes (1958)
Yasui and Johanson (1958)

Capacitance
Capacitance
Light transmission

Dotson

Capacitance

Lanneau
Fukuda,

Capacitance

(1960)
Asaki, and Kondo (1967)

Capacitance

Kunii, Yashida, and Hiraki (1967)


Whitehead and Young (1967)

Capacitance
Light transmission

Park et al. (1969)

Electroresistivity

Behie et al. (1970)

Pitot tube

Cranfield

Inductance

(1972)

Geldart and Kelsey (1972)


Pattureaux, Vergnes, and Mihe
( 1973)
Werther and MoJerus (1973)
Werther (1974)

Capacitance
Piezoelectric

Calderbank,
(1976)

Resistivity

Pereira,

and Burgess

in fluidization

research

Details

Osberg (1951)

(1959)

of probes

Capacitance

"--------

Diaphragm stretched over ring of


5 cm diameter
Wire supported on mica sandwiched
between steel supports
Three parallel plates
Three parallel plates (see Fig. l3.4a)
Two probes, each with 0.3 mm diameter
lamp and 4 mm prism
Coaxial cell, 5 cm diameter x 7.6 cm
long with central rod and six
grounded outer rods
Stainless steel strips 19 mm long
and 7.6 cm apart
1.5 mm diameter brass tips separated
by 24 mm
Not specified
196 probt;:s, each with 6 mm spacing
between source and photodiode
1 mm of two tips exposed, 9.5 mm
apart
10 mm o.d. ellipsoidal nosed tube
with backflushing
Signal coil encapsulated in epoxy
resin; forms cylinder 2 mm diameter x
4.5 mm long
Two plates 5 mm apart
Sensor 4 mm diameter connected to
piezoelectric element
Needle 3 mm long and 0.4 mm diameter
protrudes from 1.1 mm diameter tube
(see Fig. l3.4b)
Five prongs, three-dimensional
(see Fig. l3.4d)

Property

measured

Particle impacts
Bed level
Uniformity of fluidization
Local void fraction
Bubble properties

(J)

'T1

Uniformity

of bed density

S
0

N
Local voidage fluctuations
Bed uniformity, bubble
properties
Bubble frequency
Bubble properties
Bubble frequency, volume
fraction, pierced length
Jet momentum

-l

0:z
-l

m
()

:r:
:z
0
r

-<

Bubble properties

Bubble properties
Particle collision frequency
and amplitude
Bubble properties
Bubble properties,
transfer

mass

._---~---

Okhi and Shirai (1976)


Rietema and Hoebink (1976)

Light transmission
Capacitance and
suction

Dutta and Wen (1979)

Optical

Ishida, Hishiwaki,

Light reflectance
transmission

Fibre optics (see Fig. 13.4c)


Capacitance probes as Werther
combined with suction capillary
alongside
Live central rod with six longer
earthed outer rods
1 mm diameter bead
Two light-emitting diodephotodiode pairs with 5 mm gap,
one pair 20 mm above the other
LED and photocell separated by
3 mm or more
Fibre optics bundles connected
to signal processing set-up

Mayhofcr

Capacitance

Light reflectance

Array of fibre optics sensors

Thermocouples

Measures changes and differences


in temperature
Commercial cell made by Bosch
mounted on tube
15 m W laser, five illuminating fibres
of diameter 0.13 mm in 0.85 mm tube
and correia tor
Four sensors, one ahead of the others
Deformable 2.3 mm o.d. element in
steel shell; strain gauge
Three 0.5 mm diameter tubes in water
cooled 1.6 mm sheath
0.8 mm diameter aligned fibres with
screen cylinder in between
2 Thermistors 2 mm apart below
10 W heater

and Neuzil (1977)

Vines (1977)
Masson and Jottrand

Okhi, Ishida,

Thermistor
Light transmission

(1978)

and Shirai (1980)

and Sbirai (1980)

Lau and Whalley (1981)


Ljungstrom

and Lindqvist

Patrose and Caram (1982)

Zirconia electrochemical cell


Optical

Glass and Mojtahedi (1983)


Raso, Tirabasso, and Donsi

Fibre optic
Impact

Flemmer

Pneumatic: jet
impingement
UV light absorption

(1984)

Ishida and Hatano


Glicksman

(1982)

(1984)

and McAndrews

(1985)

Thermistor

Local particle velocity


Bubble properties coupled
with sampling from bubble
phase
Bed uniformity near grid
Regimes near distributor
Bubble properties

Z
(J)

Bubble properties
Velocity and direction of
particle flow; diameter
and velocity of bubbles
Gives image showing jets and
bubbles
Detluidization of
agglomerating solids
Local instantaneous oxygen
concentration
Local particle velocities

-l
:>J

c:
I:::
m
:z

-l

::J

:z

:z
tJ
tl1

."
m

;>:I

:
tl1

Bubble properties
Velocities and concentrations
of particles in jet region
Void detection, bubble
properties
Concentration
of tracer gas
Downward

velocity of particles

:z

-l

-l
tl1
()

:r:
:z
0
c:
tl1
(J)

When It IS required to follow the motion of individual particles, front


illumination is required, and it is helpful to have a small fraction (typically 1
to 2 per cent.) of the particles coloured to aid identification. It is possible, for
example, to apply a surface coating of paint to spherical particles by rolling
them down an incline to which a thin layer of fresh paint has been applied.
Coloured glass beads are also available commercially. It is important that the
properties of the marked particles be representative of those of the entire
assembly of particles. Streak photography and use of successive flashes have
not found much application in fluidization research.
Frame-by-frame
analysis of bubble properties presents some special
problems, especially when the flow due to translation of bubbles is measured
(Grace and Clift, 1974). These problems arise from transfer of gas between
separate bubbles through the permeable intervening dense phase. Bubble
areas maybe measured by means of a planimeter, by weighing paper cutouts
traced from the projected images, or by visually comparing the areas with a
set of standard circles.
High speed television recorders are now becoming available. The advantages of instant playback are considerable, and although expensive at present,
video systems are likely to displace cine photography quite rapidly.
13.5.2

Probes in Fluidized Beds

Because fluidized beds are opaque and behaviour near the walls is often not
representative
of that in the interior, many workers have inserted probes to
obtain local property measurements. A partial listing of these probes appears
in Table 13.4, and some sketches showing probe geometries are given in
Figure 13.3. For earlier reviews of probes in fluidized beds see Dutta and
Wen (1979) and Fitzgerald (1979).
The following characteristics are required of immersed probes (Werther
and Molerus, 1973):
(a) Minimal disturbance of the bed
(b) Measurement

of local properties

(c) Rapid reponse to transients


(d) Mechanical
(e) Mobility,
(f)

strength
i.e. it should be possible to relocate the probe

Compatibility

with the solids and gas

Almost all of the probes employed in fluidized studies have been unique
'homemade'
devices, rather than items which can be purchased from
suppliers. Many of the probes have distinguished between times when a small
measuring volume is devoid of particles and when particles are present. The
physical property used to discriminate between dense phase and dilute phase

.~

QJ

L :;
0'0

:.:

III

00.
-:;::
_"5

-,E

;~
.-

has most often been capacitance, light transmission, or electical resistivity.


Recording of the probe output gives a direct indication of the uniformity of
fluidization and the degree of bUbbling. Data processing techniques, such as
cross-correlation of signals from nearby probes, can give detailed information
on bubble properties. For discussions of treatment of probe signals, see
Werther and Molerus (1973), Werther (1974), Park, Lee and Capes (1974),
Calderbank, Pereira, and Burgess (1976) and Gunn and AI-Doori (1985).
Probes inevitably cause some local disturbance in fluidized beds. It is
therefore important to minimize interference while measurements are being
taken. Horizontal cylinders in fluidized beds lead to some bubble generation
at or near the equator, to a film of gas on the underside, and to a stagnant cap
of particles on the top surface (Glass and Harrison, 1964). In addition, when
bubbles impinge directly on obstacles of this nature, bubble splitting may be
induced (Cloete, 1967; Hager and Thomson, 1973; Lockwood, 1977) and the
object is subject to transient forces (Hosny and Grace, 1984). Bubbles can
envelop and adhere to vertical tubes, and this causes elongation and increased
velocities of rise (Grace and Harrison, 1968; Rowe and Everett, 1972; Hager
and Thomson, 1973; Rowe and Masson, 1980; Yates etal., 1984). Adherence
may also be caused if the tube is inclined at a small angle to the vertical (Rowe
and Everett, 1972; Hager and Thomson, 1973). If a tube is allowed to vibrate

nQ
II

II

L.!..,

""

probe than large ones. Probes near the wall of the column, horizontal
supports, and multi-prong probes were shown to cause severe interference.
These observations can be used as a guide to designing probes for fluidized
beds. Generally speaking, vertical supports are preferable to horizontal
supports. The probe itself should be as small and have as few elements as
possible. The supporting tube should be suspended from above where
possible and at least 30 particle diameters in from the wall of the column. The
support should be sufficiently rigid to withstand the buffetting action of the
bed with minimal vibration. Geldart and Kelsey (1972) showed the
importance of support geometry with respect to whether or not bubble
splitting is induced. Their photographs showed that capacitance detector
plates, mounted directly on the bottom of an 8 mm diameter suport tube at
right angles to the tube, caused bubble splitting and some adherence to the
tube. Considerably less interference occurred if the plates were offset from
the tube by means of a short, smaller diameter tube inclined at an angle of 45
to the horizontal, as shown in Fig. 13.5(c). Conversely, Rowe and Masson
(1981) found a disturbing lack of agreement when comparing bubble heights
and velocities measured by X-rays and measured by a light probe. On the
other hand, Lord et al. (1982) checked their optical probe data against video
records of a two-dimensional bed and found that the probe could be used to
provide accurate data. Gunn and AI-Doori (1985) found that proper signal
processing and noise discrimination were of critical importance in obtaining
accurate data from probes. Caution is clearly required in the use of probe
data.
When a probe consists of two parts separated by a gap, e.g. a light source
and photodetector or two plates forming a capacitor, the gap width should be
sufficient to allow particles to circulate freely through the gap. For rounded
particles of narrow size distribution, this requires a gap width of at least 20 to
30 mean particle diameters. Somewhat larger gap widths are required for
angular particles or for powders having broad size distributions.
13.5.3

Figure 13.5 Sequences showing interaction of probes and rising


gas bubbles, from Geldart and Kelsey (1972). Reprinted with
permission.

transversely,
bubbles can be formed along its entire length (Rowe and
Everett, 1972). X-ray photographs (Rowe and Masson, 1980, 1981) have
shown that probes can cause bubbles to elongate, to accelerate along the
support, to deflect sideways, to flatten prior to reaching the probe, and to
split. Small bubbles were found to be more subject to disturbance by the

Capacitance Measurements in Fluidized Beds

Many of the probes used in fluidization research have used capacitance


measurements to record fluctuations in local bed porosity (see Table 13.4). In
addition, a number of workers have avoided probe interference effects by
affixing capacitance plates to opposite walls of experimental columns. For
two-dimensional beds, resolution has been good enough to allow profiles of
local voidage around bubbles (Lockett and Harrison, 1967) and tubes
(Fakhimi and Harrison, 1980) to be determined in the manner. For threedimensional columns, a capacitance signal across the entire column diameter
or width allows determination
of the arrival of large bubbles, thereby
facilitating the measurements of slug or bubble velocities (Angelino, Charzat,

and Williams, 1964; Ormiston, Mitchell, and Davidson, 1965). Capacitance


elements have also been mounted on vertical and horizontal tubes (Ozkaynak
and Chen, 1978; Chandran and Chen, 1982; Fitzgerald, Catipovic, and
Jovanovic. 1981) to allow measurements of bubble properties and local
voidage which can be related to heat transfer models.
The principle employed in using capacitance variations is simple. The
capacitance of solid media differs markedly from that of solids-free gas. The
capacitance of a region therefore depends on the concentration of solid
particles in that region. Hence, local capacitance variations can be used to
provide information on local voidage variations, emulsion density, and on the
arrival or departure of bubbles. The probe can be water cooled to allow
measurements
in high temperature
beds (Almstedt and Olsson, 1982),
although this requires an increase in probe size. A disadvantage of the
capacitance technique is that the measuring volume is never precisely
defined. Suitable electrical circuits for monitoring capacitance variations have
been presented in a number of the cited references. Werther and Molerus
(1973) describe a method of discriminating between signals due to arrival of
bubbles and those corresponding to dense phase voidage fluctuations. In their
case, electronic counting was found to be better than auto-correlation of
signals from nearby probes for measuring bubble velocities. Gunn and AlDoori (1985) chose a threshold acceptance voltage on the basis of twodimensional bed cine films and rejected apparent bubbles with velocities
outside a realistic range of 0.05 to 2 m/s.

13.5.4

Other Non-Interfering Diagnostic Techniques

Limited information on the hydrodynamics of fluidized beds can also be


obtained by other techniques which are non-interfering. These techniques are
also applicable at high bed temperatures and pressures:
(a) By measuring the expanded bed height and height fluctuations visually,
some information can be inferred on bubble sizes and whether or not
slugging is occurring (King and Harrison, 1980).
(b) Gamma-ray attenuation can be used to measure bed level, density
profiles and phase hydrodynamics (Batholomew and Casagrande, 1957;
Baumgarten and Pigford, 1960; Saxton and Worley, 1970; Schuurmans,
1980; Weimer, Guyere and Clough, 1985). X-ray absorption can be used
in a similar manner (Bierl et af. 1980).
(c) Microwave scattering has been used to provide information on bubble
sizes and velocities (Eastlund, Jellison, and Granatstein, 1982).
(d) Pressure fluctuations
sed further below.

provide information on hydrodynamics,

as discus-

Quantity

Method/principle

measured

High speed photography


in two-dimensional bed

Rowe et at. (1965)


Gabor (1967)
Baeyens and Geldart (1973)
Loew. Schmutter. and Resnick (1979)
Tanimoto et ai. (1980)

Wake, drift. particle


velocities

Addition of tagged tracer


particles

Geldart and Cranfield


(1972)
Whitehead, Gartside, and Dent (1976)
Bieri et ai. (1980)

Solids circulation
and velocity

X-ray photography
Drag force or impact of
particles upon piezoelectric
crystal
Rotation of small impeller

Nienow and Cheesman

Segregation

Heat transfer using


thermistor beads

(1980)

Local velocity

Botterill and Bessant (1976)


Marsheck and Gomezplata (1965)
Vines (1977)
Wen etai. (1978)

Local velocity

Tracking of radioactive
particles

Guinn (1956)
Kondukov (1964)
Littman (1964)
Van Velzen el ai. (1974)
Masson, Jottrand and Dang Tran
(1978)
Lin, Chen, and Chao (1985)

Tracking of radiopill
tracer

Handley and Perry (1965)


Merry and Davidson (1973)
Rao and Venkateswarku (1973)

Transit time of heated


particles
Transit time of particles
viewed by optical fibre
light detectors

Kawanishi and Yamazaki (1984)


Valenzuela and Glicksman (1984a)

Thermistor probe
Response curves using
magnetic tracer

Schulz-Walz (1974)
Okhi, Walawender, and Fan (1977)
Okhi etai. (1980)
Patrose and Caram (1982)
Glicksman and McAndrews (1985)
Woollard and Potter (1968)

Chemically active tracer

Kamiya (1955)
Yagi and Kunii (1961)

Radioactive

tracer

Fluorescent

tracer

Guinn (1956)
Qin and Liu (1982)
Morooka, Kayo, and Kato

Ferromagnetic tracer
and inductance

patterns

Heertjes, Verloop. and Willems


(1970171)
Donsi and Massimilla (1972)

Avidan (1980)

Local velocity

Local velocity
RTD C(I) curve

RTD C(t) curve


Pathlines
RTD
RTD. solids
circulation

(e) The rate of collapse of the bed surface after interrupting the gas flow to
the bed can be used to estimate the voidage of the dense phase (Rietema,
1967; Geldart and Abrahamsen, 1980; and Chapter 2).
(f) Sound emission from the column may allow an experienced operator to
identify slugging conditions or stick-slip flow.
(g) Heat flux probes or temperature-sensitive
hot spots and refractory damage.

13.6.1

Typical references

Stimulus-response

Gilliland and Mason (1952)


Yoshida and Kunii (1968)
Atimtay and Cakaloz (1978)

Diffusion-type
(dispersion)

Overcashier, Todd, and Olney (1959)


Baems, Fetting, and Schugerl (1963)
Kobayashi and Arai (1964)
Schugerl (1967)
Van Deemter (1980)
Yang, Huang and Zhao (1984)
May (1959);
Kobayashi and Arai (1964)
Kwauk, Wang, Li
Chen, and Shen (1986)

paint can give indications of

Overall Approaches

Solids mixing can be examined with a view to determination of axial or radial


dispersion coefficients, gross solids circulation rates or turnover times,
bubble-induced particle motion, or residence time distribution of solids in the
system. Experimental
methods for determining particle velocities and
residence time distribution (RID) are summarized in Table 13.5.
Due to the limitations of visual observation and the tedious nature of
tracer-suction methods, methods have been developed for direct measure. ment of local particle velocities inside the fluidized bed. Local velocities are
then detected by a small obstacle which detects the drag force exerted on it by
a flow of particles, by measurement of the rate of heat transfer, by following
the path of a tracer, or by statistical correlation of the optically observed
movement of particles. The RTD is determined from response curves using
magnetic, chemically active, or radioactive tracers.
The various techniques of Table 13.5 have advantages and disadvantages.
For a method to be successful, the particle flow should be obstructed as little
as possible, and detection should be independent of particle mass and overall
bed characteristics such as voidage. Tracer and optical transit methods seem
most appropriate,
although the latter requires the use of electronic signal
processing. The tracking method developed recently by Lin, Chen, and Chao
(1985) is also very powerful, but requires multiple detectors and sophisticated
and extensive data storage and processing facilities.
13.6.2

Technique

Two-phase or
two-region
models

Cankurt and Yerushalmi


(1978)
Whitehead et al. (1980)

Only overall response is


obtained and gas contact
time history (bubble gas
versus emulsion gas) is
ignored
Models assume small and
random mixing steps.
Approach is better for
radial than for axial mixing.
Most common approach
It is virtually impossible to
evaluate separately gas
flow and dispersion in
phases and interphase
exchange. Some parameters are often assumed
with one or more parameter then chosen to fit the
experimental data
Tracer is looked for
upstream of injection level.
Little information is
obtained on the mechanism
of mixing or for systems
with limited axial
dispersion

pressure (imposed and hydrostatic). Sampling is stopped by retracting the


inner pipe to cover the sampling port. Similar simple devices have been
described by other workers (e.g. Ljungstrom and Lindqvist, 1982). Because
of overall circulation patterns in fluidized beds and dead regions near the
grid, it is best to provide a sampling system which can sample solids from
more than a single locality. While fluidized beds are generally well mixed, it is
best to avoid the regions immediately above the distributor and below the bed
surface if representative samples are to be withdrawn from a single level. This
advice applies also to sampling from defluidized (slumped) beds. Solids
captured in the cyclones can conveniently be sampled by means of lock
hoppers or Y-junctions installed between the cyclone and return line.

Solids Sampling Techniques

Ernst, Loughnane, and Bertrand (1979) describe a simple device for sampling
in-bed solids during operation of the Exxon coal combustion 'miniplant'. A
closed stainless steel pipe was made to slide within an open-ended close
tolerance pipe by means of pneumatic pistons. Particles enter when the inner
pipe is slid outwards exposing a small hole near the closed end. These
particles are pushed into a storage vessel outside the combustor by the

13.7.1

Overall Approaches

The flow pattern of gas in a fluidized bed has been studied both by steady!
unsteady state tracer techniques and by calculations based on reactor
performances using reactor models. Results have been expressed by means of

stimulus-response curves, diffusion-type models, and a variety of more or less


sophisticated two-region models. Limitations of these approaches and typical
references appear in Table 13.6.
13.7.2

Gas Sampling Techniques

Gas sampling is often required to measure local gas concentrations in


chemical reaction, mass transfer, and gas mixing studies. A requirement for
sampling is that the sampling rate be sufficiently small, that there is negligible
interference with the surrounding bed, and that a layer of solids, held by
suction, does not build up on the sample tube. These conditions are considered
to be satisfied if the sampling tube is much smaller than the bed dimensions
and if the mean gas velocity in the tube is Umf or less. The end of the sampling
tube must be covered with a fine mesh screen or porous tip to prevent inflow
of particles. Samples may be withdrawn continuously or in pulses. Tubes may
be inserted from the side of the bed or from above.For continuous sampling,
it is important that pressure fluctuations (associated with the passage of
bubbles, for example) be small relative to the overall pressure driving force in
order to obtain a constant flowrate and a proper time-mean concentration.
In testing of reactor models, it is desirable to obtain separate measurements
of concentration in the bubble and dense phases. For a two-dimensional
column or in the wall region of a three-dimensional bed, this can be achieved
by manually actuating a solenoid valve (Chavarie and Grace, 1975) or by
synchronization of concentration traces with cine films (Lignola, Donsi, and
Massimilla, 1983). For the interior of a three-dimensional column, more
sophisticated electronics are required by which opening of a solenoid is
controlled by the output from an optical probe or other sensor (see Table
13.4) capable of distinguishing between the two phases, e.g. the combined
capacitance suction probe described by Rietema and Hoebink (1976).
To obtain accurate concentration measurements, deposition, chemical
reactions and adsorption in the sampling line must be minimized. This may
require special materials, heat tracing of the sample line, or quenching of the
gases. Use of a carrier gas is sometimes useful. Tubes should be as short as
possible. Enlargements, fitting, etc., should be avoided in order to minimize
axial dispersion along the sample lines. Since radial gradients are appreciable
in many fluidized beds, care must be exercised when a concentration is to be
measured which is supposed to represent an average for a given level or an
average outlet value.
In-bed concentrations of oxygen at high temperature can also be measured
by zirconia electrochemical cells like those used for measured emissionsfrom
automobiles (e.g. Ljungstrom and Lindqvist, 1982). However, the signals are
difficult to interpret and subject to significant error in the presence of local
homogeneous combustion of gaseous components like CO and methane (Lim
et aI. 1986).

13.8

CHEMICAL REACTION STUDIES IN FLUIDIZED BEDS


OF LABORATORY AND PILOT PLANT SCALE

Experimental fluidized bed reactors range from simple tubes fitted with a
porous distributor to highly sophisticated pieces of equipment requiring
specialized design. Key requirements in planning a reactor are: (a) to make
sufficient provision for instrumentation that useful data can be obtained; (b)
to foresee possible problems and modifications and make the reactor flexible
enough to allow these to be solved or accommodated. It is always more
difficult, for example, to add sampling ports after a reactor is operating than
to incorporate additional ports at the design stage, especially for pressurized
reactors which are subject to special safety codes and inspections. Table 13.7
lists some common problems that arise in fluidized bed reactors and suggests
what can be done at the design stage or during operation to solve these
problems.
Since their hydrodynamics are not well understood (see Chapter 11),
fluidized beds are usually not in themselves suitable for obtaining kinetic
information. If heats of reaction are sufficiently small that reaction under
isothermal conditions can be achieved, packed beds can be used to generate
the required kinetic data. For larger heats of reaction, spinning basket
reactors are often useful. The kinetic data obtained must cover not only the
full range of temperature to be encountered in the fluidized bed reactor but .
also the full range of concentrations. Because conversion in the dense phase
tends to be much higher than that in the bubble phase, this means obtaining
kinetic data under high conversion conditions if reactor models are to be
tested.
Methods for measuring gas and solids concentrations by sampling and by
in-bed probes are outlined in earlier sections of this chapter. Many reactions
carried out in fluidized beds are highly exothermic or involve substances
which are potentially explosive. Monitors and alarms are commonly required
for in-bed temperature, pressure, flowrates, and concentrations of key
components to keep the reactor within safe limits. The integrity of refractory
should be checked at regular intervals, and the reactor shell should be
investigated for hot spots during operation. Pressure relief valves or rupture
discs should be provided and checked regularly. The refractory should be
designed to withstand sudden reactor depressurization.

13.9

MEASUREMENT OF ENTRAINMENT AND OF FREEBOARD


AND CIRCULATING BED HYDRODYNAMICS

Measurements of entrainment and related phenomena in fluidized beds pose


special problems and are subject to large errors. Many workers who have
studied entrainment (e.g. Zenz and Weil, 1958)collected solids captured by a

Table 13.7

Some common problem areas which should be considered when designing or


operating experimental fluidized bed reactors

Problem
Increase gas velocity; ensure uniform distributor; avoid
closely spaced internals; return solids from cyclone near grid
Install screens over holes; ensure high velocity gas entry;
provide for removable or accessible grid
Design removable grid; ensure high distributor pressure
drop; install screens to avoid hole blockage
Provide vibration pads; install surge tank between blower and
bed; vary gas flowrate; introduce baffles to break up slugs;
furnish viewing ports to observe bed
Avoid jet impingement; consider antiabrasion pads; decrease
gas velocity
Avoid jet impingement; decrease gas velocity
Attrition
Humidify incoming gas; use antistatic agents; connect entire
Electrostatic charges
column to ground
Alter temperature; dry incoming gas; increase gas velocity;
avoid dead regions; improve atomization of liquid feed (if
any); stir bed contents or add ultrasonic vibration; add inert
solids
Avoid horizontal and shallow surfaces; dry incoming gases;
alter temperature and/or gas velocity
Prevent dead regions near grids; ensure that adjacent internal
Hot spots, overheating,
surfaces are at least 30 particle diameters apart; increase gas
danger of explosion
velocity; increase or provide more uniform cooling
Change total transfer surface area or driving force; preheat
Extinguishing of
feed gas or solids
reaction
Change operating conditions (e.g. temperature, pressure,
Low conversion or
unwanted side reactions feed composition, gas velocity); use improved catalyst;
provide more uniform distributor; introduce baffles;
introduce quenching in freeboard
Add cyclone in parallel or series; reduce attrition (as
detailed); lower gas velocity or pressure; introduce baffles
just above or below bed surface; furnish viewing ports;
provide expanding freeboard
Test for leaks; operate under negative pressure; enclose
Escape of noxious
column
fumes or dust
Increase dipleg diameter; heat trace dipleg; avoid bends and
Blockage of dipleg
inclined portions containing dense phase flow
Avoid erosion and vibration (as detailed); choose tube
materials with corrosion and erosion in mind, taking account
of local concentration fluctuations; allow for thermal
expansion; increase tube support
Avoid bends; increase inlet size: dilute feed with inert
particles; aerate feed; consider alternate feed method
(e.g. screw feeding, pneumatic feeding)
Avoid attrition, solid deposits, and hot spots (as detailed);
Failure to reach
preheat feed; improve heat transfer and insulation; provide
steady state
more reliable feed system for solids and liquid (if any)
Solids segregation

single cyclone over a period of time. These particles were then weighed and
sieved. This technique has a number of disadvantages:
(a) It assumes perfect collection in the cyclone.
(b) Since fines are not returned to the bed during the period of collection,
bed properties change with time.
(c) No information is given on the hold-up of solids in the freeboard.
Merrick and Highley (1972)accounted for essentially all of the particles carried
over by adding a second cyclone and sampling the stack gases. Blyakher and
Pavlov (1966) allowed entrained particles to be carried over and used the
weight change to determine the total entrainment. Changes in bed composition
with time may be minimized by ensuring that the mass of solids collected is
much less than the total mass of solids in the bed or that the time of sampling
is less than the particle cycle time.
Techniques used to determine the hold-up of solids in the freeboard include
static pressure profiles (e.g. Nazemi, Bergougnou, and Baker, 1974),
attenuation of a laser beam (Horio et ai., 1980), and butterfly or disc valves
which trap the suspended solids (e.g. Geldart and Pope, 1978). The static
pressure method has the disadvantage that frictional losses and particle
deceleration, in addition to the hold-up of solids and gas, contribute to the
measured pressure drop, making it difficult to determine the hold-up
accurately by this method. The trapping valves are only useful if they can be
closed very quickly and if they do not disturb the flow in the freeboard when
not in service. Static pressure drops are also commonly used to measure
suspension densities in circulating beds. A single porous butterfly valve in.the
return vessel has been used by a number of investigators beginning with
Yerushalmi, Turner, and Squires (1976) to determine solids circulation rates
in circulating bed systems.
To determine the transport disengaging height experimentally, the flux of
solids should be measured. Since the flux tends to vary with radial position as
well as with height, both radial and axial profiles are required. Fournol,
Bergougnou, and Baker (1873) employed an isokinetic suction probe, where
the suction velocity was set equal to the superficial gas velocity to obtain local
flux measurements. A difficulty with isokinetic sampling is that bubble arrival
at the bed surface is irregular, causing particle ejection in bursts, rather than
in a uniform manner. Bubble eruption also causes the gas flow to be highly
turbulent, especially in the lower part of the freeboard (Levy and Lockwood,
1980, 1983). Because bubble bursts cause release of both solids and gas,
isokinetic sampling requires suction velocities in excess of the superficial gas
velocity (Pemberton, 1982). George and Grace (1978) designed a special
trapping device to capture the particles ejected by individual bubbles.
Soroko, Mikhalev, and Mukhlenov (1969) used photography to determine
the transport disengaging height.

Recent workers have realized the importance of turbulence and of


separately measuring gas and particle velocities in the freeboard. Hot wire
anemometry has been used to measure 'gas velocities (Horio et ai., 1980;
Pemberton, 1982); Horio et ai. used the optical fibre prove of Okhi and Shirai
(1976) to measure particle velocities. Laser Doppler anemometry has been
employed by Levy and Lockwood (1980, 1983) to measure both gas and
particle velocities in the freeboard, as well as to indicate the size of the
particles in the measuring volume. These techniques are difficult to apply in
fluidization research because of the presence of a broad spectrum of particle
sizes and the relatively high hold-up of solids.

Time-averaged pressure measurements are useful in determining the minimum fluidization velocity and expanded bed level and in estimating the holdup of. solids in the unit. Pressure taps should be positioned at regular
intervals, typically 0.1 to 0.2 m, up the side of the vessel. One tap is usually
installed in the windbox below the grid. There should always be a tap just
above the grid (Sutherland, 1964). Although the pressure drop across the grid
is an indicator of the gas flowrate, it is advisable to measure the gas flowrate
by another method (e.g. rotameter or upstream 'orifice plate) since the
distributor holes may become partially blocked. The pressure drop across the
distributor plate can be used to indicate whether hole blockage or changes in
grid performance have occurred. Whitehead and Dent (1978) have shown
that pressure profiles along the distributor plate can be used to deduce
information regarding solids circulation patterns there.
For fully fluidized materials, the pressure drop between two taps is very
nearly equal to the weight of solids (and gas) held up between the two levels
divided by the bed cross-sectional area. Lower measured pressure drops
usually indicate chanelling or partial defluidization, some of the weight of the
solids being supported by interparticle contact. When absolute pressure i~
plotted against height, there is a sharp break in slope at the surface of the
expanded bed at low gas velocities, and this gives a convenient means of
estimating bed depth. Some deviation from a linear profile for the dense bed
may occur at the lowest pressure tap, and this has been a~cribed (Rowe and
Partridge, 1965) to radial gas flow associated with bubble formation. Changes
in slope may also indicate solids segregation in the bed (Chiba, Chiba,
Nienow and Kobagoshi, 1979).
The pressure measurements described above are generally obtained with
simple manometers or Bourdon-type pressure gauges. A gauze or fine screen
should be installed flush with the wall of the vessel to prevent particles from
entering the pressure taps. For sticky solids, it may be necessary to purge the
manometer lines continuously in order to prevent build-up of a deposit over

the pressure taps. The linear velocity in the purge line should be 1 to 2 m/s. In
some cases, it is more convenient to use a thin tube covered by a screen
lowered from the top of the column to measure the pressure at different
levels, instead of taps on the wall. For time-mean pressure measurements, it
is often useful to damp out pressure fluctuations by having the pressure lines
expanded to a tube of larger bore containing fixed particles (e.g. sand), in
effect providing a capacitance.
Aside from the information which can be derived from time mean pressure
drops, further information may be obtained from pressure fluctuations if one
or more taps are connected to a rapid response (e.g. diaphragm type)
pressure transducer and high speed recorder. To minimize damping of the
signal, the transducer and connecting tubes should have small volumes.
Absence of pressure fluctuations may indicate channelling or defluidization.
Pressure fluctuations of irregular amplitude and frequency usually indicate
normal bubbling, with the mean amplitude becoming larger with increasing
bubble size. Large amplitude oscillations of a fairly regular nature normally
denote slugging. High frequency, low amplitude fluctuations tend to be
associated with the turbulent regime of fluidization (Yerushalmi et ai., 1978).
Hence, the output from a pressure transducer can be useful as an indicator of
the regime of fluidization and of any shifts in bed behaviour or regime
(Broadhurst and Becker, 1976; Ho et ai., 1983; Svoboda et ai. 1984). Within
the bubbling and slugging regimes, statistical analysis and correlation of
pressure fluctuations have been used to infer bubble and slug properties
(Sitnai, 1982; Zhang, Walsh and Beer, 1982; Fan, Ho, -and Walawender,
1983). Probes can also be prepared which measure local pressure fluctuations,
but interpretation of the signals is difficult (Lord et al., 1982).

,13.11.1

Gas-to-solids heat transfer and temperature measurements

In both steady and unsteady state experiments, the gas-to-solids heat transfer
coefficient is deduced from a heat balance over a differential section of height
dl:

He~t into the)


( sectIOnby gas

._ (hea~ out of )
sectIOnby gas

= (heat tr~nsferred)
to particles

(13.2)

Solving this differential equation with specific boundary conditions yields the
heat transfer coefficient hgp: It is generally assumed that the solids

temperature Tp is uniform throughout the bed. This approach also assumes


that gas elements associated with the bubble or dense phase are at the same
temperature at a given height, i.e. the interphase heat transfer resistance is
ignored. Aside from this last assumption, a principal reason for the
experimental scatter in reported values of hgp is the uncertain temperature of
gas and solids, as measured by thermocouples embedded in the fluidized
system. Bare and suction thermocouples have been used. The differences in
thermocouple readings have been discussed by various authors. From
thermocouple response and use of 'shielded' thermocouples, Baeyens and
Goossens (1973) demonstrated that a bare thermocouple reads the solids
temperature, although emulsion gas and solids are at the same temperature in
a well-mixed bed. Using suction thermocouples (Walton, Olson, and
Levenspiel, 1995; Chang and Wen, 1966), a different reading is obtained
depending on the pressure drop of the suction probe. A high pressure drop
probe yields a value close to Tg, whereas a low pressure drop probe measures
a temperature close to the bubble temperature since bubble gas is sampled
preferentially.
It is therefore advisable to determine solids and gas temperature using bare
microthermocouples and high pressure drop suction thermocouples, respectively. Errors may be caused by radiation or by a build up of solids on the
thermocouple. Solids temperatures can be read by other .techniques such as
thermistors, miniature microelectric devices (Barker, 1967), or heat-sensitive
rings (Yates and Walker, 1978). Optical pyrometry can be used at high bed
temperatures.
13.11.2

has been used including horizontal and vertical tubes, finned and flattened
tubes, vertical surfaces, and spheres. Some investigators have measured
spatially and time-average heat transfer coefficients, while others have
measured time-average local values and still others have devised means of

Bed-to-Surface Heat Transfer

Coefficients for wall-to-bed heat transfer are commonly calculated from a


heat balance using the area S of the transferring surface and the difference
between the wall temperature Twand bulk bed temperature Tb, both
measured by means of thermocouples. Hence:

Q
hbw

S(Tw-Tb)

where Q is the heat transfer rate.


A typical experimental set-up is illustrated in Fig. 13.6. The experiment
involves a heater (or cooler) immersed in (or on the containing wall of) the
column. Electrical heating can be used instead of circulating a heat transfer
fluid inside the heat transfer surface. Axially and radially moveable
thermocouples are used to measure bed temperatures; thermocouples are
also fixed to the heater or cooler surface. Axial heat losses at the ends of the
heat transfer element should be reduced, e.g by guard heaters (Baeyens and
Goossens, 1973). As shown in Table 13.8, a variety of heat transfer surfaces

Figure 13.6 Typical experimental set-up for studying bed-to-surface heat transfer:
1 compressor,2 valve and water trap, 3 pressure gauge, 4 rotameters, 5 thermometer,
6 distributor, 7 fluidization column, 8 expanded freeboard, 9 filters, 10 manometer,
11 constant head tank, 12 water rotameter, 13, 14 thermometers, 15 thermocouples,
16 multi-chaneI recorder, 17 heater, 18 cooling jacket, 19 supports for heater, 20
voltage regulators.

Table 13.8 Examples

of typical and unusual experimental


equipment
bed-to-surface
heat transfer coefficients

used to measure

References

Transfer surface

Details

Aspects studied

Fitzgerald, Catipovic,
and Jovanovic
(1981)

51 mm diameter tube
with five platinum
film heating
elements around
periphery

Resistance heating of
elements with
simultaneous
capacitance and
local pressure
measurement

Instantaneous local
heat transfer coefficients. pressure.
and capacitance for
horizontal tube

George and Grace


(1982)

25 mm o.d. horizontal
tubes in arrays

Bed electrically
heated; tube cooled
by silicone oil

Freeboard transfer,
different configurations. and tube
spacing

Golan, Lalonde, and


Weiner (1982)

100 mm o.d. horizontal tube bundle

Bcd heated by propane burning; water


cooling

Local, average
convective and
radiative coefficients measured

Temperature profiles
and fluctuations

Ozkaynak, Chen and


Frankenfield (1984)

Radiometer probe of
diameter 73 mm

Total and radiative


heat transfer

Effect of fins, tube


spacing, and particle size

Surface mounted
vertically in beds at
temperatures to
800C

Piepers, Wiewiorski,
and Rietema (1984)

65 mm vertical tube
bundle

One of 19 tubes with


10 electrically
heated copper
elements

Effect of radial position and height as


well as superficial
velocity

References

Transfer surface

Details

Aspects studied

Mickley and Trilling


(1949)

Vertical 12 mm
diameter calrod
heater; Externally
heated bed also

Single tube on axis of


bed, electrical
heating

Temperature distributions, effects of


particle diameter,
bed height, gasvelocity

0.13 mm diameter
horizontal wire
45 mmlong
Vertical 6.4 mm tube,
610mm long

Electrical heating,
bridge circuit

Effect of gas and


particle properties

Six 100 mm long


sections, electrical
heating
Single tube; fixed and
moveable thermocouples
16 mm tubes with and
without fins, electrical heating

Influence of height,
gas velocity, bubble
frequency

Jacob and Osberg


(1957)
Mickley, Fairbanks
and Hawthorn
(1961)
Baeyens and Goossens
(1973)
Bartel and Genetti
(1973)

10 mm diameter
vertical central
heater
Seven horizontal bare
or finned tubes

Tuot and Clift (1973)

10 mm square vertical
platinum film

Guard heater;
Wheatstone bridge

Local instantaneous
coefficients in twodimensional column

Andeen, Glicksman,
and Bowman (1978)

Flattened horizontal
tubes of height
32 mm and width
13 mm

Array oftubes, electrically heated

Effect of superficial
velocity bed depth,
and tube shape

Canada and
McLaughlin (1978)
Hoebink and Rietema
(1978)

32 mm o.d. horizontal
tubes
30 copper blocks,
each 20 x 30 mm
arranged in 3 x 10
vertical array
Vertical cylinder of
diameter 18 mm
and length 100 mm

5 and 10 row banks,


staggered array

Effects of pressure
and particle size

Blocks separated by
insulation and kept
at same temperature
Wooden rod wrapped
with copper wire

Effect of heater length


and vertical
position

40x20x1Omm
copper block

Electrically heated

Effect of pressure

30 mm diameter,
277 mm long horizonta! tube,
isolated longitudinalS mm wide strip

Strip heated electrically and insulated


from heated
remainder of tube

Angular and vertical


dependence of local
heat transfer coefficient

9.5 mm diameter
sphericalorimetric
probe

Embedded thermocouple; different


surfaces to allow
radiation to be
evaluated

Effect of particle size


and type, bed
temperature;
contribution of
radiation

Borodulya, Ganzha,
and Podberezsky
(1980)
Xavier et al. (1980)
Abubakar

et ai. (1980)

Botterill, Teoman,
and Yuregir (1981)

Effect of pressure and


particle diame ter

obtaining instantaneous local values. The listing in Table 13.8 is far from
complete. Entries have been chosen to illustrate the wide variety of
experimental equipment and to give references where detailed information is provided. Experimental attempts to measure radiation heat transfer in
fluidized beds have been reviewed by Ozkaynak, Chen, and Frankenfield
(1984).
The methods available for measuring surface temperature and their errors
and difficulties are discussed by Baker, Ryder, and Baker (1961) and Walker
and Rapier (1965). Thermocouples mounted onto a surface should not
perturb the bed or alter the surface roughness. Errors for brazed thermocouples are discussed by Walker and Tapier (1965).
Care should be taken if a direct electrical current is passed through the heat
transfer element since pick-up voltages can be of the same magnitude as the
thermocouple output. To minimize this effect, electrically insulating cements
with high thermal conductivity should be used, e.g. cements based on copper
oxide. Care should also be taken to minimize and estimate heat losses in
. order to obtain accurate surface-to-bed heat transfer coefficients.

Table 13.9

Summary of common instrumentation needs for fluidized


bed equipment
Position

Property
Gas flowrate
Solids feedrate
Solids efflux

Upstream
Inlet
Outlet

Bed depth

Bed surface

Pressure profile

Bed

Pressure drop

Grid

Pressure fluctuations
Temperature
Temperature

Bed

Temperature

Freeboard

Gas composition

Upstream

Gas composition
Gas composition
Gas composition

Bed
.Freeboard
Outlet

Upstream
Bed

Solids composition
Solids composition

Bed
Inlet, outlet

Heat flux

Tubes, jacket

Provision or comments
Rotameters, orifice plate
Weigh belt, load cells on hopper
Lock hopper, load cells, filter on
cyclone gas outlet
Sight glass, pressure profile, sensor
lowered from above, y-rays
Taps at regular intervals connected
to manometers
Taps just above and below distributor
plate
Tap above distributor connected to
transducer and high speed recorder
Thermocouple in feed line or windbox
Thermocouple in dense bed region
(one usually sufficient)
Series of thermocouples at regular
intervals
Sampling from feed stream or windbox to
gas chromatograph or other analysers
Sampling tubes
Sampling tubes
Sampling well above bed, e.g. near
cycloneinlet
Grab sampler or exit port(s)
Procedure for collecting representative
samples
Thermocouples at inlet and outlet

The most common instrumentation needs for fluidized beds are summarized
in Table 13.9. Additional instrumentation may be required for particular
applications.

cross-sectional area of column


orifice area
gas concentration response as function of time to pulse stimulus
discharge coefficient

t:J.PD
Pg

specific heat capacity of gas


bed diameter
expanded bed height
bed to wall heat transfer coefficient
gas to particle heat transfer coefficient
number of orifices in the distributor
heat transfer rate
surface area of heat transfer surface
particle surface area per unit bed volume
bulk bed temperature
gas temperature
particle temperature
wall temperature
superficial gas velocity
vertical height coordinate
pressure drop across the distributor
gas density

Abubakar, M.Y., Bergougnou, M.A., Tarasuk, J.D., and Sullivan, J.L. (1980).'Local
heat transfer coefficients around a horizontal tube in a shallow fluidized bed', J.
Powder & Bulk Solids Technol., 4, 11-18.
Almstedt, A.E., and Olsson, E. (1982). 'Measurements of bubble behaviour in a
pressurized fluidized bed burning coal using capacitance probes', Proc. 7th Intern.
Fluidized Bed Combustion Conf., pp. 89-98.
Andeen, B.R., Glicksman, L.R., and Bowman, R. (1978). 'Heat transfer from
flattened horizontal tubes', in Fluidization (Eds J.F. Davidson and D.L. Keairns),
Cambridge University Press, pp. 345-350.
Angelino, H., Charzat, c., and Williams, R. (1964). 'Evolution of gas bubbles in
liquid and fluidized systems', Chem. Eng. Sci., 19,289-304.
Anwer, J., and Pyle, D.L. (1973). 'Gas motion around bubbles in fluidized beds', in
La Fluidisation et ses Applications, Cepadues Editions, Toulouse, France, pp.
240-253.
Atimtay, A., and Cakaloz, T. (1978). 'An investigation of gas mixing in a fluidized
bed', Powder Technol., 20,1-7.
Avidan, A.A. (1980). 'Bed expansion and solid mixing in high velocity fluidized beds',
Ph.D. Thesis, City University of New York.
Avidan, A.A., and Yerushalmi, J. (1982). 'Bed expansion in high velocity
fluidization', Powder Technol., 32, 223-232.
Baerns, M., Fetting, F., and Schugerl, K. (1963). 'Radiale und axiale Gasvermischung
in Wirbelschicten', Chem lng. Tech., 35, 609.
Baeyens, J., and Geldart, D. (1973). 'Particle mixing in a gas fluidized bed', in La
Fluidisation et ses Applications, Cepadues Editions, Toulouse, France, pp.
182-195.

Baeyens, I., and Goossens, W.R.A. (1973). 'Some aspects of heat transfer between a
vertical wall and a gas fluidized bed', Powder Techno/., 8, 91-96.
Baker, H.D., Ryder, E.A., and Baker, N.H. (1961). Temperature Measurements in
Engineering, 2nd ed., Wiley, New York.
. .
,
Bakker, P.J., and Heertjes, P.M. (1958). 'Porosity measurements in a flUidizedbed,
Brit. Chem. Eng., 3, 240-246.
Barker, J.J. (1967). 'Microelectronic device for measuring heat transfer coefficients in
fluidized beds', Ind. Eng. Chem. Fund., 6, 139-142.
.
Bartel, W.J., and Genetti, W.E. (1973). 'Heat transfer from a hOrIzontal bundle of
bare and finned tubes in an air fluidized bed', A.I.Ch.E. Symp. Ser., 29, (128),
85-93.
Bartholomew, R.N., and Cassagrande, R.M. (1957). 'Measuring solids concentration
in fluidized systems by gamma-ray absorption', Ind. Eng. Shem., 49. 428-431.
Baumgarten, P.K., and Pigford, R.L. (1960). 'Density fluctuations in fluidized beds',
A.I. Ch.E.J., 6, 115-123.
Behie, L.A., Bergougnou, M.A., Baker, C.G.J., and Bulani, W. (1970). 'Jet
momentum dissipation at a grid of a large gas fluidized bed', Can. J. Chem. Eng.,
48, 158-161.
Bieri, T.W., Gajdos, L.J., McIver, A.E., and McGover~, J.J. (1980). 'Studies in
support of recirculating bed reactors for the processmg of coal', DOE Rept ..
FE-2449-11.
Blyakher, I.G., and Pavlov, V.M. (1966). 'Reducing dust ~nt,rainment from conical
fluidized bed equipment through the use of support gnds , Int. Chem. Eng., 6,
47-50.
Boland, D., and Geldart, D. (1971/72). 'Electrostatic charging in gas fluidized beds',
Powder Technol., 5,289-297.
Borodulya, V.A., Ganzha, V.G., and Podberezsky, A.I. (1980). 'Heat transfer in a
fluidized bed at high pressure' in Fluidization (Eds J.R. Grace and J.M. Matsen),
Plenum Press, New York, pp. 201-207.
Botterill, J.S.M., and Bessant, D.J. (1976). 'The flow prope~es of fluidized solids',
Powder Technol., 14, 131-137.
Botterill, J.S.M., and Desai, M. (1972). 'Limiting factor in gas fluidized bed heat
transfer', Powder Technol., 6, 231-238.
Botterill. J.S.M., Teoman, Y., and Yuregir, K.R (1981). 'Temperature effects on the
heat transfer behaviour of gas fluidized beds', A.I. Ch.E. Symp. Ser., 77 (208),
330-340.
Broadhurst, T.E., and Becker, H.A. (1976). 'Measurement and spectral analysis of
pressure fluctuations ~n slugging be~s', in Fluidization Technology (Ed. D.L.
Keairns), Vol. I, Hemisphere, Wasdhmgton, DC, pp. 6~5.
.
Calderbank, P.H., Pereira, J., and Burgess, J.M. (1976). The phySIcal and mass
transfer properties of bubbles in fluidized ~eds of electrically ~onducting par:ticles',
in Fluidization Technology (Ed. D.L. Kealrns), Vol. I, HemIsphere, Washmgton,
DC, pp. 115-167.
.
.,.
Canada, G.S., and McLaughlin, M.H. (1978). 'Large particle flUIdizatIOnand heat
transfer at high pressures', A.I.Ch.E. Symp. Ser., 7~ .(179)2!-37.
.
..
Cankurt, N.T., and Yerushalrni, J. (1978). 'Gas backmIxlOgm hIgh velocity flUIdIzed
beds', in Fluidization (Eds J.F. Davidson and D.L. Keaims), Cambridge
University Press, Cambridge, pp. 387-393.
Chan I.H. and Knowlton, T.M. (1984). 'The effect of pressure on entrainment from
bubbling gas-fluidized beds', in Fluidization (Eds D. Kunii and R. Toei),
Engineering Foundation, New York, pp. 283-290.

Chandran, R, and Chen, J.e. (1982). 'Bed-surface contact dynamics for horizontal
tubes in fluidized beds', A.I.Ch.E.J., 28,907-914.
Chang, T.M., and Wen, e.y. (1966). 'Fluid-to-particle heat transfer in air-fluidized
beds', Chem. Eng. Prog. Symp. Ser., 62, (67),111-117.
Chavarie, e., and Grace, J.R. (1975). 'Performance analysis of a fluidized bed
reactor', Ind. Eng. Chem. Fund., 14,75-91.
Chiba, S., Chiba, T., Nienow, A.W., and Kobayashi, H. (1979) 'The minimum
fluidization velocity, bed expansion and pressure-drop profile of binary particle
mixtures', Powder Technol., 22,255-269.
Chitester, D.C., Kornosky, RM., Fan, L-S., and Danko, J.P. (1984). Characteristics
of fluidization at high pressure', Chem. Eng. Sci., 39, 253-261.
Ciborowski, J., and Wlodarski, A. (1962). 'On electrostatic effects in fluidized beds',
Chem. Eng. Sci., 17,23-32.
Cloete, F.L.D. (1967). Discussion, Proceedings of International Symposium on
Fluidization (Eds A.A.H. Drinkenburg), Netherlands University Press, Amster
dam, pp. 305-307.
Cranfield, R.R. (1972). 'A probe for bubble detection and measurement in large
particle fluidized beds', Chem. Eng. Sci., 27,239-245.
Crowther, M.E., and Whitehead, J.e. (1978) 'Fluidization of fine particles at elevated
pressures', in Fluidization (Eds J.F. Davidson and D.L. Keairns), Cambridge
University Press, Cambridge, pp. 65-70.
Davies, G., and Robinson, D.B. (1960). 'A study of aggregative fluidization', Can. J.
Chem. Eng., 38,175-183.
de Carvalho, J.F.RG. (1976). 'Fluidization under pressure', Ph.D. Thesis, Cambridge
University.
de Carvalho, J.R.F.G., King, D.F., and Harrison, D. (1978). 'Fluidization of fine
powders under pressure', in Fluidization (Eds J.F. Davidson and D.L. Keairns),
Cambridge University Press, pp. 59-64.
Donsi, G., and Massimilla, L. (1972). 'Bubble-free expansion of gas fluidized beds of
fine particles', A.I. Ch.E.J., 19, 1104-1110.
Dotson. J.M. (1959). 'Factors affecting density transients in a fluidized bed',
A.I.Ch.E.J., 5, 169-174.
Dutta, S., and Wen, C.Y., (1979). 'Simple probe for fluidized bed measurements',
Can. J. Chem. Eng., 57,115-119.
Eastlund, B.J., Jellison, G., and Granatstein, V. (1982). 'Microwave diagnostics of
bubbles in fluidized bed combustors', Proc. 7th Intern. Fluidized Bed Combustion
Conf., Vol. 1, pp. 99-110.
Ellis, J.E., Partridge, B.A., and Lloyd, D.I. (1968). 'Comparison of predicted with
experimental butadiene yields for the oxydehydrogenation of butenes in a gas
fluidized bed', Instn. Chem. Engrs. Symp. Ser., 30, 43-52.
Ernst, M., Loughnane, M.D., and Bertrand, R.R. (1979). 'Instrument methods in
fluid bed pilot plant, coal processing technology', A.I.Ch.E., 5, 166-170.
Fakhimi, S., and Harrison, D. (1970). 'Multi-orifice distributors in fluidized beds: a
guide to design', in Chemeca Proceedings, Butterworths, Australia, pp. 29-46.
Fakhimi, S., and Harrison, D. (1980). 'The void fraction near a horizontal tube in a
fluidized bed', Trans. Instn Chem. Engrs, 58, 125-131.
Fan, L.T., Ho, T-C., and Walawender, W.P. (1983). 'Measurements of the rise
velocity of bubbles, slugs and pressure waves in a gas-solid fluidized bed using
pressure fluctuation signals', A.I.Ch.E.J., 29,33-39.
.Fitzgerald, T.J. (1979). 'Review of instrumentation for fluidized beds' , in Proc. N. S.F.
Workshop, Rensselaer Polytechnic Inst., pp. 292-309.

Fitzgerald, T.J., Catipovic, N.M., and Jovanovic. G.N. (1981). 'Instrumented


cylinder for studying heat transfer to immersed tubes in fluidized beds', Ind. Eng.
Chern. Fund., 20,82-88.
Flemmer, R.L.e. (1984). 'A pneumatic probe to detect gas bubbles in fluidized beds',
Ind. Eng. Chem. Fund., 23, 113-119.
Forrest, 1.S. (1953). 'Methods of increasing the electrical conductivity of surfaces',
Brit. J. Appl. Phys., 4, Suppl.2, S37-39.
Fournol, A.B., Bergougnou, M.A., and Baker, e.G.J. (1973). 'Solids entrainment in
a large gas fluidized bed', Can. J. Chem. Eng., 51, 401-404.
Fukuda, M., Asaki, Z., and Kondo, Y. (1967). 'On the nonuniformity in fluidized
beds' , Mem. Fac. Eng., Kyoto Univ., 29, 287-305.
Gabor, J.D. (1967). 'Wall effects on fluidized particle movement in a two-dimensional
column', in Proceedings of International Symposium on Fluidization (Ed. A.A.H.
Driilkenburg), Netherlands University Press, Amsterdam, pp. 230-240.
Gabor, J.D. (1972). 'Wall to bed heat transfer in fluidized beds', A.I.Ch.E.J., 18,
249-250.
Garcia, A., Grace, I.R., and Clift, R. (1973). 'Behaviour of gas bubbles in fluidized
beds', A. I. Ch.E.J., 19, 369-370.
Gbordzoe, E.A.M., Bulani, W., and Bergougnou, M.A. (1981). 'Hydrodynamic
study of floating contactors (bubble breakers) in a fluidized bed', A. I. Ch.E. Symp.
Ser., 77 (205), 1-7.
Geldart, D., and Abrahamsen, A.R. (1980). 'The effect of fines on the behaviour of
gas fluidized beds of small particles', in Fluidization (Eds J. R. Grace and J .M.
Matsen), Plenum Press, New York, pp. 453-460.
Geldart, D., and Baeyens, J. (1985). 'The design of distributors for gas fluidized beds',
Powder Techno!., 42, 67-78.
Geldart, D., and Cranfield, R.R. (1972). 'The gas fluidization of large particles',
Chern. Eng. J., 3, 211-231.
Geldart, D., Hemsworth, A., Sundavadra, R., and Whiting, K.J. (1981). 'A
comparison of spouting and jetting in round and half-round fluidized beds', Can. J.
Chern. Eng., 59, 638-639.
Geldart, D., Hurt, J.M., and Wadia, P.H. (1978. 'Slugging in beds of large.
particles', A.I.Ch.E. Symp. Ser., 74 (176), 60-66.
Geldart, D., and Kelsey, J.R. (1972). 'The use of capacitance probes in fluidized
beds', Powder Technol., 6, 45-60.
Geldart, D., and Pope, D.J. (1978). 'Elutriation from a multisize fluidized bed,
A.I.Ch.E. Meeting, Miami Beach.
George, S.E., and Grace, J.R. (1978). 'Entrainment of. particles from aggregative
fluidized beds', A.I.Ch.E. Symp. Ser., 74 (176).67-74.
George, S.E., and Grace, I.R. (1982). 'Heat transfer to horizontal tubes in the
freeboard region of a gas fluidized bed', A.I.Ch.E.J., 28,759-765.
Gerald, C.F. (1951). 'Measuring uniformity of fluidization', Chem. Eng. Prog., 47,
483-484.
Germain, B., and Claudel, B. (1976). 'Fluidization at mean pressure less than 30 torr,
Powder Techno!., 13, 115-121.
Gilliland, E.R., and Mason, E.A. (1952). 'Gas mixingin beds of fluidized solids, Ind.
Eng. Chem., 44, 218-224.
Glass, D.H., and Harrison, D. (1964). 'Flow pattern near a solid obstacle in a
fluidized bed', Chem. Eng. Sci., 19,1001-1002.
Glass, D.H., and Mojtahedi, W. (1983). 'Measurement of fluidized bed bubbling
properties using a fibre-optic light probe', Chem. Eng. Res. & Des., 61, 37-44.

Glicksman, L.R., and McAndrews, G., (1985). 'The effect of bed width on the
hydrodynamics of large particle fluidized beds', Powder Technol., 42, 159-167.
Godard, K., and Richardson, J.F. (1968). 'The behaviour of bubble-free fluidized
beds', Instn Chem. Engrs Symp. Ser., 30, 126-135.
Golan, L.P., Lalonde, G.V., and Weiner, S.e. (1982). 'High temperature heat
transfer studies in a tube filled bed', Proc. 6th Intern. Fluidized Bed Combustion
Conf., 3, 1173-1184.
Grace, J.R. (1974). 'Fluidization and its application to coal treatment and allied
processes', A.I.Ch.E. Symp. Ser., 70 (141), 21-26.
Grace, I.R., and Clift, R. (1974). 'On the two-phase theory of fluidization', Chem.
Eng. Sci., 29,327-334.
Grace, J.R., and Harrison, D. (1968). 'The effect of internal baffles in fluidized beds'
a guide to design', lnstn Chem. Engrs Symp. Ser., 27. 93-100.
.
Grace, J.R., and Harrison, D. (1969). 'The behaviour of freely bubbling fluidized
beds', Chem. Eng. Sci., 24, 497-500.
Grace, J.R., Lim, e.J., and Evans, R.L. (1983). 'Preparatory Report of pressurized
fluidized bed combustion', Report to Government of Canada under Contract
ISU82--Q0251.
Guinn, V.P. (1956). Nucleonics., 14 (May), 69.
Gunn, D.J. and AI-Doori, H.H. (1985). 'The measurement of bubble flows in
fluidized beds by electrical probe', Int. J. Multiphase Flow, 11, 535-551.
Hager, W.R., and Thomson, W.J. (1973). 'Bubble behaviour around immersed tubes
in a fluidized bed', A.I. Ch.E. Symp. Ser., 69 (128), 68-77.
Handley, M.F., and Perry, M.G. (1965). Rheol. Acta, 4,225.
Harper, W.R. (1960). Soc. Chem. Ind. Monograph, 14, 115.
Harrison, D., and Leung, L.S. (1962). 'The coalescence of bubbles in fluidized beds'
in Proc. Symp. Interactions Fluids Particles, Instn Chern. Engrs, London, pp:
127-134.
Hatate, Y., King, D.F., Migita, M., and Ikari, A. (1985). 'Behaviour of bubble in a
semi-cylindrical gas-solid fluidized bed', J. Chem. Eng. Japan., 18,99-104.
Heertjes, P.M., Verloop, J., and Willems, R. (1970/71). 'The measurment of local
mass flow rates and particle velocities in fluid-solids flow, Powder Technol., 4,
38-40.
Hengl, G., Hiquily, N., and Couderc, J.P. (1977). 'A new distributor for gas
fluidization', Powder Technol., 18,277-278.
Henry, P.S.H. (1953). 'Electrostatic eliminators in the textile industry', Brit. J. Appl.
Phys., 4, Supp!. 2, S78-83.
Hiraki, I., Yoshida, K., and Kunii, D. (1966). 'Behaviour of bubbles in a twodimensional fluidized bed', Chem. Eng. Japan, 4, 139-146.
Ho T-e., Yutani, N., Fan, L.T., and Walawender, W.P. (1983). 'The onset of
slugging in gas-solid fluidized beds with large particles', A. I. Ch. E.J. , 35, 249-257.
Hoebink, J.H.B.J., and Rietema, K. (1978). 'Wall-to-bed heat transfer in a fluidized
bed', in Fluidization (Eds J.F. Davidson and D.L. Keairns), Cambridge University
Press, pp. 327-332.
Horio, M., Taki, A., Hsieh, Y.S., and Muchi, I. (1980). 'Elutriation and particle
transport through the freeboard of a gas-solid fluidized bed', in Fluidization (Eds
J.R. Grace and J.M. Matsen), Plenum Press, New York, pp. 509-518.
HorsIer, A.G., and Thompson, B.H. (1968). 'Fluidization in the development of
gas-making processes', Instn Chem. Engrs Symp. Ser., 30, 58-66.
Hosny, N., and Grace, J.R. (1984). 'Forces 011 a tube immersed within a fluidized
bed', in Fluidization (Eds D. Kunii and R. Toei), Engineering Foundation, New
York, pp. 111-120.

Ishida, M., and Hatano, H. (1984). 'The behaviour of gas and solid particles in a
gas-solid fluidized bed detected by optica: fiber probes', in Fluidization (Eds D.
Kunii and R. Toei), Engineering Foundation, New York.
Ishida, M., Nishiwaki, A., and Shirai, T. (1980). 'Movement of solid particles around
bubbles in a three-dimensional fluidized bed at high temperature', in Fluidization
(Eds J .R. Grace and J.M. Matsen), Plenum Press, New York, pp. 357-364.
Jacob, A., and Osberg, G.L. (1957). 'Effect of gas thermal conductivity on local heat
transfer in a fluidized bed', Can. J. Chem. Eng., 35, 5-9.
Jin, Y., 'Yu, L.LZ., and Wang, Z. (1980). 'A study of particle movement in a gas
fluidized bed', in Fluidization (Eds J.R. Grace and J.M. Matsen), Plenum Press,
New 'York, pp. 365-372.
Karniya, Y. (1955). Chem. Eng. Japan, 19, 412.
Katosonov, LV., Menshov, V.N., Zuev, A.A., and Anokhin, V.N. (1974). 'Density
of fluidized beds of fine grained materials and its variation under pressure', Zh.
. Prikl. Khim., 47, 1861-1866.
Katz, H.H., and Sears, J.T. (1969). 'Electric field phenomena in fluidized and fixed
beds', Can. J. Chem. Eng., 47,50-53.
Kavlick, V.J., and Lee, R.S. (1970). 'High pressure electrothermal fluidized bed
gasification of coal char', Chem. Eng. Prog. Symp. Ser., 66 (105), 145-151.
Kawabata, J., Yumiyama, M., Tazaki, Y., Honma, S., Chiba, T., Sumiya, T., and
Endo, K. (1981). 'Characteristics of gas fluidized beds under pressure', J. Chem.
Eng. Japan, 14,85-89.
Kawanishi, K., and Yamazaki, M. (1984). 'Structure of emulsion phase in fluidized
bed', in Fluidization (Eds D. Kunll and R. Toei), Engineering Foundation, New
York, pp. 37-44.
..
King, D.F., and Harrison, D. (1980). 'The bubble phase in high-pressure fluidized
beds', in Fluidization (Eds J.R. Grace and J.M. Matsen), Plenum Press, New
York, pp. 101-108.
Knowlton, T.M. (1977). 'High pressure fluidization characteristics of several
particulate solids, primarily coal and coal-derived materials', A.I.Ch.E. Symp.
Ser., 73 (161), 22-28.
Knowlton, T.M., and Hirsan, J. (1980). 'The effect of pressure on jet penetration in
semi-cylindrical gas fluidized beds', in Fluidization (Eds J.R. Grace and J.M.
Matsen), Plenum Press, New York, pp. 315-324.
Kobayashi, H., and Arai, F. (1964). 4th Syrup. Chern. React. Eng., Soc. Chern.
Engrs, Osaka.
Kondukov, N.B., Kornilaev, A.N., Skachko, LM., Akhromenkov, A.A., and
Kruglov, A.S. (1964). 'An investigation of the parameters of moving particles in a
fluidized bed by a radioisotopic method'. Int. Chem. Eng., 4,43-47.
Kunll, D., and Levenspiel, O. (1969). Fluidization Engineering, Wiley, New York.
Kunll,- D., Yoshida, K., and Hiraki, J. (1967). 'The behaviour of freely bubbling
fluidized beds', In Proc. Intern. Symp. on Fluidization (Ed. A.A.H. Drinkenburg),
Netherlands University Press, Amsterdam, pp. 243--254.
Kwauk, M., Wang. N., Li, Y., Chen, B., and Shen, Z. (1986) 'Fast Fluidization at
ICM' , In Circulating Fluidized Bed Technology (Ed. P. Basu), Pergamon Press,
Toronto, pp. 33--62.
Lanneau, K.P. (1960). 'Gas-solid contacting in fluidized beds', Trans. Instn Chem.
Engrs, 38, 125-137.
Lau, LT., and Whalley, R.J.P. (1981). 'A differential thermal probe for anticipation
of defluidization of caking coals', Fuel Proc. Technol., 4, 101-115.
Lee, B.S., Pyrcioch, E.J., and Schora, F.e. (1970). 'Hydrogasification of coal in highpressure fluidized beds', Chem. Eng. Prog. Symp. Ser., 66 (105), 152-156.

Levy, E.K., Caram, H.S., Dille, J.e., and Edelstein, S. (1983). 'Mechanisms for
solids ejection from gas fluidized beds', A.I.Ch.E.J., 29, 383-388.
Levy, Y., and Lockwood, F.e. (1980). 'Two phase flow measurements in the
freeboard of a fluidized bed using laser Doppler anemometry', Israel J. Technol.,
18, 146--151.
Levy, Y., and Lockwood, F.e. (1983). 'Laser Doppler measurements of flow in
freeboard of a fluidized bed', A.I. Ch. E.J., 29, 889-895.
Lewis, W.K., Gilliland, E.R. and Bauer, W.e. (1949). 'Characteristics of fluidized
particles', Ind. Eng. Chem., 41, 1104-1117.
Lignola, R.G., Donsi, G., and Massimilla, L. (1983). 'Mass spectrometric measurements of gas composition profiles associated with bubbles in a two-dimensional
bed', A.I.Ch.E. Symp. Ser., 79 (222),19-26.
Lim, e.J., Ko, G.H., and Grace, J.R. (1986) 'Oxygen probe measurements and their
accuracy in coal combustors'. To be published.
Lin, J.S., Chen, M.M., and Chao, B.T. (1985). 'A novel radioactive tracking facility
for measurement of solids motion in gas fluidized beds', A.I. Ch. E.J., 31, 465-473.
Littman, H. (1964). 'Solids mixing in straight and tapered fluidized beds'.
A./.Ch.E.J.,
10,924-929.
Littman, H., and Homolka, G.A.J. (1973). 'The pressure field around a twodimensional gas bubble in a fluidized bed', Chem. En~. Sci., 28,2231-2243.
Ljungstrom, E., and Lindqvist, O. (1982). 'Measurements of in-bed gas and solids
compositions in a combustor operating at pressures up to 20 bar' , Proc. 7th Intern.
Fluidized Bed Combustion Conf., 1,465-472.
Lockett, M.J., and Harrison, D. (1967). 'The distribution of voidage fraction near
bubbles rising in gas-fluidized beds', Proceedings of International Symposium on
Fluidization
(Ed. A.A.H. Drinkenburg), Netherlands University Press,
Amsterdam, 257-267.
Lockwood, D.N. (1977). 'Effects of heat exchanger tube spacing and arrangement on
the quality of fluidization', Proc. Pachec Conf., Denver, pp. 1177-1181.
Loeb, L.B. (1958). Static Electrification, Springverlag, Berlin.
Loew, 0., Schmutter, B., and Resnick, W. (1979) 'Particle and bubble behaviour and
velocities in a large-particle fluidized bed with immersed obstacles',' Powder
Technol., 22,45-57.
Lord, W.K., McAndrews, G., Sakagami, M., Valenzuela,I.A., and Glicksman, L.R.
(1982). 'Measurement of bubble properties in fluidized beds', Proc. 7th Intern.
Fluidized Bed Combustion Conf., 1,76-88.
Lyall, E. (1969). 'The photography of bubbling fluidized beds', Brit. Chem. Eng., 14,
501~506.
Marsheck, R.M., and Gomezplata, A. (1965). 'Particle flow pattern in a fluidized
bed', A.I.Ch.E.J.,
11,167-173.
Masson, H., and Jottrand, R. (1978). 'Measurement of local bubble properties in a
fluidized bed', in Fluidization (Eds J.F. Davidson and D.L. Keairns), Cambridge
University Press, pp. 1~.
Masson, H., Jottrand, R., and Dang Tran, K. (1978). Intern. Congo on Mixing in the
Chemical Industries, Mons.
Mathur, K.B., and Epstein, N. (1974). Spouted Beds, Academic Press, New York.
May, W.G. (1959). 'Fluidized bed reactor studies', Chem. Eng. Prog., 55 (12), 49-56.
May, W.G., and Russell, F.D. (1954). 'High pressure fluidization', Paper given at
North New Jersey A.e.S. meeting.
Mayhofer, R., and Neuzil, L. (1977). 'Influence of grid on fluidized bed inhomogeneity', Chem. Eng. J., 14, 167-173.

Merrick, D., and Highley, 1. (1972). 'Particle size reduction and elutriatiQn in a
fluidized bed processes', A.l.Ch.E. Syrnp. Ser., 70 (137), 36?,-378..
.
Merry, J.M.D., and Davidson, J.F. (1973). 'Gulf stream CIrculatIOnm shallow
fluidized beds' Trans. Instn Chern. Engrs, 51, 361-368.
Mickley, H.S., F~irbanks, D.F., and Hawthorn,. R.D: (19~1~.'The relation betwee~
the transfer coefficient and thermal fluctuations m flUIdized bed heat transfer,
Chern. Eng. Prog. Syrnp. Ser., 57,51--60.
. .
. .
Mickley, H.S. and Trilling, C.A. (1949). 'Heat transfer charactenstIcs of flUidized
beds' Ind. Eng. Chern., 41,1135-1147.
.
.
Miller, e.O., and Logwinuk, A.K. (1951). 'Fluidization studies of solid particles, Ind.
Eng. Chern., 43,1220-1226.
.
Mogan, J.P., Taylor, R.W., and Booth, F.L. ~1971):'!he va.lueof the exponent m th,e
Richardson and Zaki equation for fine solids flUIdizedWIthgases under pressure ,
Powder Techno!., 4, 286-289.
.
.
.
Morooka, S., Kago, T., and Kato, Y. (1984). 'Flow pattern of..solid particles .m
freeboard of fluidized bed', in Fluidization (Eds D. Kunll and R. Toe1),
Engineering Foundation, New York, pp. 291-298.
Morse, R.D., and Ballou, C.O. (1951). 'The uniformity of fluidization, its
measurement and use' Chern. Eng. Prog., 47, 199-211.
Nazemi, A., Bergougnou: M.A., and Baker, C.G.J. (1974). 'Dilute phase holdup in a
large gas fluidized bed', A.l.Ch.E. Syrnp. Ser., 70 (14), 98-107.
Nguyen, T.H., and Grace, J.R. (1978). "Forces on objects immersed in fluidized
beds', Powder Technol., 19,255-264.
."
.
Nguyen, X.T., and Leung, L.S. (~972). 'A note on bubble formatIOnat an onfice m a
fluidized bed', Chern. Eng. Scz., 27, 1748-1750.
'
..
Nienow, A.W., and Cheesman, D.J. (1980). 'The effect of shape on t~e ml~lt.Ig~nd
segregation of large particles in a gas-fluidized bed of small ones, m FlUldlzatlon
(Eds J.R. Grace and J.M. Matsen), Plenum Press, ~ew Yo:k, pp. 373-~80.
Ohki, K., Ishida, M., and Shirai, !. (19~0). 'Th~ ?ehaVlO~r?f Jets.a~d ~artlcles near
the distributor grid in a three-dImenslOnal flUIdizedbed, m FlUidizatIOn (Eds J.R.
Grace and J.M. Matsen), Plenum Press, New York, ~~. 421-42~. .
..,
Ohki, K., and Shirai, T. (1976). 'Particle velo~ity in flUldIZ~dbeds, m FLUIdizatIOn
Technology (Ed. D.L. Keairns), Vol. I, HemIsphere, W~shmgton, DC, pp. 95-110.
Okhi, K., Walawender, W.P., and Fan, L.T. (1977). 'The measurement of local
velocity of solid particles', Powder Techno!., 18,171-178.
..
Orcutt, J.C., and Carpenter, B.H. (1971). 'Bubble coalescence and the slmulat~onof
mass transport and chemical reaction in gas fluidized beds', Chern. Eng. SCI., 26,
1046-1064.
.
Ormiston, R.M., Mitchell, F.R.G., and Davidson, J.F. (1965). 'The velOCityof slugs
in fluidized beds' Trans. Instn Chern. Engrs, 43, 209-216.
Osberg, G.L. (1951). 'Locating fluidized solids bed level in a reactor', Ind. Eng.
Chern., 43, 1871-1873.
Overcashier, R.D., Todd, D.B., and Olney, R.B. (1959). 'Some effects of baffles on a
fluidized system', A.I.Ch.E.J., 5,54-60.
.
.
Ozkaynak, T.F., and Chen, J.e. (1978). 'Averag.e res~d~ncetIm:s of emulslOn and
void phases at the surface of heat transfer tubes m flUidizedbeds, A.I.Ch.E. Syrnp.
Ser., 74 (174), 334-343.
,
.
Ozkaynak, T.F., Chen, J.e., and Fran~~nfield, T.R .. (1984~.~n expenme~tal
investigation of radiation (Eds D. Kunu and R. ToeI), Engmeenng Foundation,
New York, pp. 371-378.
Park, W.H., Kang, W.L., Capes, C.E., and Osberg, G.L. (1969). 'The properties of

bubbles in fluidized beds of conducting particles as measured by an electroresistivity probe', Chern. Eng. Sci., 24,851-865.
Park, W.H., Lee, N.G., and Capes, C.E. (1974). 'Wall effects in point probe
measurements of radial bubble distributions', Chern. Eng. Sci., 29, 339-344.
Partridge, B.A., Lyall, E., and Crooks, H.E. (1969). 'Particle slip surfaces in bubbling
gas fluidized beds', Powder Technol., 2, 301-305.
Patrose, B., and Caram, H.S. (1982). 'Optical fiber probe transit anemometer for
particle velocity measurements in fluidized beds', A.I.Ch.E.J., 28, 604-609.
Patureaux, T., Vergnes, F., and Mihe, J.P. (1973). 'Construction of piezoelectric
transducer for statistical study on impact of particles in fluidized beds', Powder
Technol., 8, 101-105.
Pemberton, S.T. (1982). 'Entrainment from fluidized beds', Ph.D. Thesis, Cambridge
University.
Piepers, H.W., Wiewiorski, P., and Rietema, K. (1984). 'Heat transfer on a vertical
tube bundle immersed in a 0.70 m fluidized bed', in Fluidization (Eds D. Kunii and
R. Toei), Engineering Foundation, New York, pp. 339-346.
Pyle, D.L., and Harrison, D. (1967). 'The rising velocity of bubbles in twodimensiorial fluidized beds', Chern. Eng. Sci., 22, 531-535.
Qin, S., and Liu, G. (1982). 'Application of optical fibres to measurement and display
of fluidized systems', in Fluidization Science and Technology (Eds M. Kwauk and
D. Kunii), Science Press, Beijing, pp. 258-266.
Rao, V.L., and Venkateswarku, D. (1973). 'Determination of velocities and flow
patterns of particles in mass flow hoppers', Powder Techno!., 7, 263-265.
Raso, G., Tirabasso, G., and Donsi, G. (1983). 'An impact probe for local analysis of
gas-solid flows', Powder Technol., 34, 151-159.
Richardson, J.F., and McLeman, M. (1960). 'Pneumatic conveying: solids velocities
and pressure gradients in a one-inch horizontal pipe', Trans. Instn Chern. Engrs,
38, 257-266.
Rietema, K. (1967). 'Application of mechanical stress theory of fluidization', in
Proceedings of International Syrnposium on Fluidization (Ed. A.A.H. Drinkenburg), Netherlands University Press, Amsterdam, pp. 154-166.
Rietema, K., and Hoebink, J. (1976). 'Mass transfer from single rising bubbles to the
dense phase in three-dimensional fluidized beds', in Fluidization Technology (Ed.
D.L. Keairns) Vol. I, Hemisphere, Washington, DC, pp. 279-288.
Rowe, P.N. (1971). 'Experimental properties of bubbles', Chapter 4 in Fluidization
(Eds J.F. Davidson and D. Harrison), Academic Press, London, pp. 121-191.
Rowe, P.N., and Everett, D.J. (1972). 'Fluidized bed bubbles viewed by X-rays',
Trans. Instn Chern. Engrs, 50,42--60.
Rowe, P.N., Foscolo, P.U., Hoffman, A.C., and Yates, J.G. (1984). 'X-ray
observation of gas fluidized beds under pressure', in Fluidization (Eds D. Kunii and
R. Toei), Engineering Foundation, New York, pp. 53-60.
Rowe, P.N., MacGillivray, H.J., and Cheesman, D.J. (1979). 'Gas discharge from an
orifice into a gas fluidized bed', Trans. Instn Chern. Engrs., 57,194-199.
Rowe, P.N., and Masson, H. (1980). 'Fluidized bed bubbles observed simultaneously
by probe and X-rays', Chern. Eng. Sci., 35, 1443-1447.
Rowe, P.N., and Masson, H. (1981). 'Interactions of bubbles with probes in gas
fluidized beds', Trans. Instn Chern. Engrs, 59, 177-185.
Rowe, P.N., and Partridge, B.A. (1962). 'Particle movement caused by bubbles in a
fluidized bed', in Proc. Syrnp. Interactions Fluids Particles, Instn Chern. Engrs,
London, pp. 135-142.
Rowe, P.N. and Partridge, B.A. (1965). 'Aggregative fluidization', Chern. Eng. Sci.,
I

20,985.
Rowe, P.N., Partridge, B.A., Cheney, A.G., Henwood, G.A., and Lyall E. (1965).
The mechamsms of solids mixing in fluidized beds', Trans. lnstn. Chem Engrs 43
271-286.
' ,
Rowe, P.N., Partridge, B.A. and Lyall, G. (1964). 'Cloud formation around bubbles
in fluidized beds', Chem. Eng. Sci., 19,973-985.
Sathiyamoorthy, D., and Rao, C.S. (1981). 'The choice of distributor to bed pressure
drop in gas fluidized beds'. Powder Technol., 30, 139-143.
Saxton, A.L., and Worley, A.C. (1970). 'Modern catalytic cracking design', Oil &
Gas J., 68 (20). 82-99.
Schugerl, K. (1967). 'Experimental comparison of mixingprocesses in two- and threephase fluidized beds' in Proceedings of International Symposium on Fluidization
Netherlands University Press, Amsterdam, pp. 782-794.
'
Schulz-Walz, A. (1974). 'Messing von Schuttgutgeschwindigkeiten am Beispiel der
Schwingsiebzentrifuge, Chem. Ing. Techn., 46, 259.
Schuurmans, H.J.A. (1980). 'Measurements in a commercial catalytic cracking unit'.
Ind. Eng. Chem. proc. Des & Dev., 19, 267-27l.
Sit, S.P., and Grace, J.R. (1978). 'Interphase mass transfer in an aggregative fluidized
bed', Chem. Eng. Sci., 33, 1115-1122.
Sitnai, D .. (19~2). 'Utilization of the pressure differential records from gas fluidized
beds with mternals for bubble parameters determination' Chem. Eng. Sci 37
1059-1066.
.
,.,
,
Soko!o.' V.E., .Mikhalev, M.F., and Mukhlenov, I.P. (1969). 'Calculation of the
mimmum height of the space above the bed in fluidized bed contact equipment' .
Int. Chem. Eng., 9, 280-281.
'
Soo, S:L:, Ihring, H:K., and El Kouh, A.G. (1960). 'Experimental determination of
statistical properties of two-phase turbulent motion', Trans. ASME J. Basic Eng.
82, 609-621.
'
,
Staub, F ..W., and Canada, G.S ..(197~): 'Effect of tube bank and gas density on flow
behaViour and heat transfer m flUidiZedbeds'. In Fluidization (Eds J.F. Davidson
and J:?L. Keairn~), Cambridge Univerisity Press, Cambridge, pp. 339-344.
Subzwan, M.P., Chft, R., and Pyle, D.L. (1978). 'Bubbling behaviour of fluidized
beds at. elevate~ pre.ssures', in Fluidization (Eds J.F. Davidson and D.L. Keairns),
Cambndge Umverslty Press, Cambridge, pp. 50-54.
Sutherland, J.P. (1964). 'The measurement of pressure drop across a gas fluidized
bed', Chern. Eng. Sci., 19,839-841.
SV~boda, K., Cerm.ak, ~., Hartman, M., Drahos, J., and Selucky, K. (1984).
Influence of particle Size on the pressure fluctuations and slugging in a fluidized
bed', A.I.Ch.E.J., 30,513-517.
Tanimoto, ~., S:hiba, S., .C~iba, T., and Kobayashi, H. (1980). 'Mechanism of solid
segregatIon m gas flUidiZed beds', in Fluidization (Eds J.R. Grace and J.M.
Matsen), Plenum, New York, pp. 381-388.
Torobin, L.B., and Gauvin. W.H. (1961). 'Fundamental aspects of solids-gas flow'
Can. J. Chern. Eng., 39, 113-120.
'
Tuot, J., and Clift, R. (1973). 'Heat transfer around single bubbles in a twodimensional fluidized bed', A.I.Ch.E. Symp. Ser., 69, (128),78-84.
Vale~z.uel~, J.A., and Gli~ksman, L.R. (1984a). 'An experimental study of solids
mlXlng m a freely bubblmg two-dimensional fluidized bed' Powder Technol 38
63-72.
,.,
,
Valenzuela: J.A:, and Gli~k~man, L.R. (1984b). 'Residence time and dispersion of
fine particles m a 2-d flUidizedbed of large particles', in Fluidization (Eds D. Kunii
and R. Toei), Engineering Foundation, New York, pp. 161-168.

VanI?e.em~er, J.J. (1980). 'Mixing patterns in large-scale fluidized beds'. in


FLUIdizatIOn(Eds J.R. Grace and J.M. Matsen), Plenum Press. New York.
pp.69-89.
Van Velzen, D., Flamm, H.J., Langenkamp, H., and Castle. A. (1974). 'Motion of
solids in spouted beds', Can. J. Chem. Eng., 52, 156--16l.
Varadi, T., and Grace, J.R. (1978). 'High pressure fluidization in a two-dimensional
gas flu.idized ?ed',. in Fluidization (Eds J.F. Davidson and D:L. Keairns),
Cambndge Umversity Press, Cambridge, pp. 55-58.
Vines, S.N. (1977). 'Gas-solid regimes in the distributor region of a fluidized bed',
Ph.D. Thesis, University of Virginia.
Wace, P.F., and Burnett, S.J. (1961). 'Flow pattern in gas-fluidized beds'. Trans.
Instn Chem. Engrs, 39, 168-174.
Walker, V., and Rapier, A.c. (1965). UKAEA Report. JRG1026 (W).
Walton, J.S., Olson, R.L., and Levenspiel, O. (1952). Gas-solid film coefficients of
heat transfer in fluidized coal beds'. Ind. Eng. Chem., 44, 1474-1480.
Weim~r,.A.W., ~'Yure, D.C., and Clough, D.E. (1985). 'Application of a gammaradiation denSity gauge for determining hydrodynamic properties of fluidized
beds', Powder Technol., 44,179-194.
Wen, C.Y., Krishnan, R., Khosravi, R., and Dutta, S. (1978). 'Dead zone heights
near the grid of fluidized beds', in Fluidization (Eds J.F. Davidson and D.L.
Keairns), Cambridge University Press, Cambridge, pp. 32-37.
Werther, J. (1974). 'Bubbles in gas fluidized beds', Trans. Instn Chem. Engrs, 52,
149-169.
Werther, J., and Molerus, O. (1973). 'The local structure of gas fluidized beds', Int. J.
Multiphase Flow, 1, 103-138.
Whitehead, A.B., and Dent, D.C. (1978). 'Some effects of distributor slope and
auxiliary gas injection on the performance of gas-solid fluidized beds', in
Fluidization (Eds J.F. Davidson and D.L. Keairns), Cambridge University Press,
Cambridge, pp. 44-49.
Whitehead, A.B., Gartside, G., and Dent, D.C. (1976). 'Fluidization studies in large
gas-solid systems', Powder Techno/., 14,61-70.
Whitehead, A.B., Potter, O.E., Nguyen, H.V., and Dent, D.C. (198Q). 'Gas
backmixing in 0.61 m and 1.22 m square fluidized beds', in Fluidization (Eds J.R.
Grace and J.M. Matsen), Plenum Press, New York, pp. 333-340.
Whitehead, A.B., and Young, A.D. (1967), 'Fluidization performance in large scale
equipment', in Proceedings of International Symposium on Fluidization (Ed.
A.A.H. Drinkenburg), Netherlands University Press, Amsterdam, pp. 294-302.
Whiting, K.J., and Geldart, D. (1980). 'A comparison of cylindrical and semicylindrical spouted beds of coarse particles', Chem. Eng. Sci., 35,1499-1501.
Woollard, I., and Potter, O.E. (1968). 'Solids mixing in fluidized beds', A.I. Ch.E.J.,
14, 338-391.
Xavier, A.M., King, D.F., Davidson, J.F., and Harrison, D. (1980). 'Surface-bed
heat transfer in a fluidized bed at high pressure', in Fluidization (Eds J .R. Grace
and J.M. Matsen), Plenum Press, New York, pp. 209-216.
Yagi, S., and Kunii, D. (1961). 'Fluidized solids reactors with continuous solids feed,
Chem. Eng. Sci., 16,364-371.
Yang, G., Huang, Z., and Zhao, L. (1984). 'Radial gas dispersion in a fast fluidized
bed', in Fluidization (Eds D. Kunii and R. Toei), Engineering Foundation, New
York, pp. 145-152.
Yang, W-C., Revay, D., Anderson, R.G., Chelen, E.J., Keairns, D.L., and Cicero,
D.C. (1984). 'Fluidization phenomena in a large-scale cold-flow model', in
Fluidization (Eds D. Kunii and R. Toei), Engineering Foundation, New York, pp.

Yates, J.G., Cheesman, D.J., Mashingaidze, T.A., Howe,.C., and Jefferis, G. (1984).
'The effect of vertical rods on bubbles in gas fluidized beds', in Fluidization (Eds D.
Kunii and R. Toei), Engineering Foundation, New York, pp. 103-110.
Yates, J.G., and Walker, P.R. (1978). 'Particle temperatures in a fluidized bed
combustor', in Fluidization (Eds J.F. Davidson and D.L. Keaims), Cambridge
University Press, Cambridge, pp. 241-245.
Yerushalmi, 1., Cankurt, N.T., Geldart, D., and Liss, B. (1978). 'Flow regimes in
vertical gas-solid contact systems', A.I.Ch.E. Syrnp. Ser., 74 (176),1-13.
Yerushalmi, J., Turner, D.H., and Squires, A.M. (1976). 'The fast fluidized bed', Ind.
Eng. Chern. Proc. Des. & Dev., 15, 47-53.
Yoshida, K., and Kunii, D. (1968). 'Stimulus and response of gas concentration in
bubbling fluidized beds', J. Chern. Eng. Japan, 1,11-16.
Yoshida, K., Fujii, S., and Kunii, D. (1976). 'Characteristics of fluidized beds at high
temperatures',
in Fluidization Technology (Ed. D.L. Keairns), Vol. 1, Hemisphere, Washington, DC, pp. 43-48.
Zabrodsky, S.S., Antonishin, N.V., and Pamas, A.L. (1976). 'On fluidized bed-tosurface heat transfer', Can. J. Chern. Eng., 54,52-58.
Zenz, F.A., and Weil, N.A. (1958). 'A theoretical-empirical
approach to the
mechanisms of particle entrainment from fluidized beds', A.!. Ch.E.J., 4, 472--479.
Zhang, M.C., Walsh, P.M., and Beer, J.M. (1982). 'Determination of bubble size
distributions from pressure fluctuations in a fluidized bed combustor', 7th Intern.
Fluidized Bed Combustion Conf., 1, 75-75J.

Index

Abrasion in cyclones, 210, 211


Abrasion index, 211
Acrylonitrile, 2, 5, 194, 285
Adhesion, 42
Aeration, see Standpipes, and Collapse
Rate
Aero-mechanical separators
dry, 198
wet, 198, 214, 215
Aggregative fluidization, see Bubbling
fluidization
Agglomerates, see Clusters
Agglomeration, 24, 38, 115, 261, 326
Applications, 4-7,192-195,285,316,331
Archimedes number, 3, 24, 105
Attrition, 1, 38, 42, 123, 246, 250, 326,
332,358
in cyclones, 211
Axial dispersion of gas, 296-297
Backmixing
of gas, 37,39, 108-109,286
of solids, 173, 286
Baffles, 103, 110,262,264,307-309,312,
327
Bends
pneumatic conveying, 357, 358
Boilers, see Combustors
Bridging, 38
Bubblelbubbles, 1,54
cloud, 289-292, 297, 310
coalescence, 37, 45, 78, 79, 100,294
drift fraction, 98, 103, 104
effect of temperature and pressure on,
61
effect of wall on, 65
formation at distributor, 75-78
frequency, 75

maximum stable size, 37, 50,60,61,


78,79,85,
88, 228
shape, 54, 56
size, 22, 37, 45, 46, 58, 59, 77, 297, 319
correlations for, 76, 82
splitting, 37, 60, 167-169
velocity, 58, 59, 85, 319
wake fraction, 55, 56, 57, 58, 98, 103,
104, 135, 329
Bubble point, see Velocity, minimum
bubbling
Bubbling fluidization, 34, 53-95, 219
Bypassing, see Gas bypassing
Calcination, 108, 117-118, 123, 192,318,
321
Capacitance measurements, 157, 163,
169, 430--436
Carman-Kozeny equation, 20
Carryover, see Entrainment
Catalysts, 5, 290
cracking, 17,24,33,36,39,57,
123,
141, 172-174, 181, 184,219, 246,
329
Catalytic cracking, 4, 33, 36, 123, 184,
219,285, 286, 341
Catalytic reactions, 36, 117
gas phase, 285, 287-289, 306
Channels/channelling,
25, 34, 160
Characterization of fluidized powders,
33-51
see also Powders, effect of properties
on fluidization
Chemical reaction
engineering, 8
studies, 441--443
Chemical reactor models, see Reactor
models

Chemical reactors, 11,295


conve~ion in, 26,42, 298-309
Choking, 126, 343-348
Circulating fluid beds, see Fast
fluidization, and Solids circulation
Clusters, 135, 170, 175, 184, 344
Cohesive solids/powders, 1,21,34, 35,
36, 39,42, 43, 62, 116, 125
Cold models, 6, 48, 421-440
Collapse rate of bed, 25, 27, 36, 39, 438
Combustion/combustor, 5, 6, 7,139,193,
194, 214, 250, 285, 316
turn down, 252
Conversions, see Chemical reactors
Cut size, 202, 207
Cyclones, 2, 200-215
clogging, 147,210,212,442
cut size, 202, 207
design, 209-210
efficiency, 42, 202, 204-209
flow characteristics, 203-204
high efficiency, 205, 206
high rate, 206
loss of fines from, 42
multi-, 202, 203
pressure drop in, 207
reverse flow, 201, 205
uniflow, 202
De-aeration, see Collapse rate
Defluidization, 111, 114, 115, 387
Dense phase, 53, 81
velocity in, 26, 28
voidage of, 27, 81, 305
Dense phase flow/conveying, see
Pneumatic transport
Density
absolute, skeletal, 17,49
aerated,22
bulk, 17,43
definition of, 14
of fludized bed, 174, 180
particle or piece, 3, 49, 229
Dilute phase flow/conveying,
see Pneumatic transport
Distribution of gas between phases,
81,297
Distributors,
dead zones, 115-116
effect of solids mixing, 102
general, 2, 25, 54, 65, 228, 421,
427-429

perforated plate, 65
pipe grid, 67
porous plate, 65, 82, 91, 427
pressure drop across, 67-73
tuyere type, 66,69,70,72,73,75,77,
78
weeping/dumping, 65-66
Drift, see Bubble, drift fraction
Dryers; fluid bed,S, 119, 259-283
batch, 260
continuous plug flow, 262
continuous well-mixed, 26(}-261
process design of, 273-278
vibrated, 263
Drying,
constant rate, 268
falling rate, 269
Electrostatic precipitators, 198, 215
Electrostatics, 34, 198, 422, 424, 427
Elutriation, see Entrainment
Elutriation rate constant, 135-144
correlations for, 142
effect of baffles, 141
effect of bed diameter, 141
effect of gas velocity, 141
effect of pressure, 141
effect of size of particles, 138
Emissivity of particles, 240, 241, 242, 243
Emulsion, see Dense phase
Energy balance, 299,319, 330
Entrainment, (see also Elutriation rate
constant), 1,37, 123-150
measurement of, 441, 443-444
mechanisms, 134
Ergun equation, 13, 20, 362
Erosion, 1, 74, 77, 244, 246, 250, 358
Euler Number, 199,205,206
Excess gas velocity, 45,81,107
Expansion,
ofbed,27,32,36,37,4O,41,83-86,
105,227,244
of dense phase, 26, 28, 36
of sluggingbed, 90-92, 167-168
of turbulent bed, 167-168, 249
Extrusion flow, 369
Fast fluidization, 3, 37, 155-195,344-345
applications of, 192-195,331-332
heat transfer in, 223
FCC, see Catalysts
Filte~, 215, 216, 447

bag, 216
Fines fraction, 28, 42
effect on conversion in freeboard, 327
effect on entrainment, 138, 139
effect on heat transfer, 228, 237
Finned tubes, see Tubes, finned
Fischer-Tropsch synthesis,S, 193, 194,
285
Fixed beds, (see also Packed beds) 11,
18-21
Fluid bed dryer, see Dryer
Fluidization velocities, see Velocities
Fluidized bed combusto~, see
Combustor
Freeboard, 126,445
reaction in, 326-330
Gas adsorption, 25, 42, 288, 292
Gas bypassing, 38, 45-46, 81, 287,
289-292
Gas cleaning, 197-217
Gas interchange/exchange, 38, 53,
292-300
Gas sampling, 440, 441
Galileo number, see Archimedes number
Gas-solid contacting, 97, 289, 331
Gas-solid reactions, 285, 316-326
Gasification, 4, 8, 285, 316
Gas metering, 416-418
Gas mixing, 186-191, 310
Gas supply, 416-418
Geldart fluidization diagram, 3, 33-51,
368
Grade efficiency, 199, 202, 204
Granulators/granulation, 2, 267, 266
Grid, see Distributor
Hardness of particles, 34, 42, 43
Hausner ratio, 43
Heat sensitive powders, 263, 265
Heat storage in fluid bed, 252
Heat transfer, 1, 34, 53, 96, 220
bed-to-surface,223-234
effect of pressure on, 230, 237
maximum coefficient, 238, 239
measurement of, 446-449
Heat transfer correlations
bed-to-tube, 229-250
interphase gas convective component
223, 224-235
particle-to-gas,22(}-222
radiative component, 223, 229,

24{}-:243
turbulent fluidized bed, 249
Holdup
measurement of 36,441,443-444
Hoppers, 35,37,212,407-409
Humidity, 34, 35, 418, 427, 442
Hydrodynamics
of bubbles,S, 53-92, 302, 429
Incipient fluidization velocity, see
Velocity, minimum fluidization
Inclined pipes,
in pneumatic conveying, 358-359
Instrumentation, 415-462
Interparticle forces, 36, 43, 48, 227
Interphase mass transfer, see Mass
transfer
Interstitial gas velocity, see Velocity,
inte~titial gas
Jets/jetting, 73, 74, 76, 77, 313, 314
Kinetics of drying, 267-272
see also Reaction kinetics /
Lateral mixing, see Solids mixing (radial)
Leaks, 418
Liquid fluidization, 26, 34, 46-47
Mass transfer,
interphase 74, 292-294, 296, 304, 306,
311,316
Mechanical stirring, 34
Methanol-to-gasoline process, 6, 194
Minimum bubbling velocity, see
Velocity, minimum bubbling
Minimum fluidization velocity, see
Velocity, minimum fluidization
Mixing, see Gas mixing, Solids mixing
Moving beds, 4, 20, 391
see also Standpipes
Multi-stage beds, 264, 326
Non-bubbling fluidization, see Quiescent
fluidization
Non-catalytic gas-phase reactions, 285
Non-mechanical valves, L- and J-valves,
363,395-407
aeration in, 396-398, 400

Packed bed, 11, 18-21


flow, 361, 373
pressure drop in, 19
Particle ejection, 126, 135,327
Particle density, see Density
Particle growth, 318
Particle separation, see Gas cleaning
Particle shape, 11,
see also Sphericity
effect on heat transfer, 239
Particle size,
mean diameter, 3, 14
median diameter, 14
sieve diameter, 11, 16
surface diameter, 12
surface volume diameter, 11
volume diameter, 11,13
Particle size distribution, 11, 14
narrow, 33
wide, 33, 259
Particle velocity, 105-107
in conveying, 345
Particulate fluidization, see Quiescent
fluidization
Photographic techniques, 419, 423., 429,
432
Pilot scale plant, 88, 300, 424, 429
slug flow in, 90
Pneumatic transport, 35, 36, 341-370
dense phase, 360-370
dilute phase, 125, 341-360
Polyethylene, 5
Pore volume, 17
Porous solids, 17,34, 317, 390
Powders,
effect of properties on fluidization, 3,
33-50,416
narrow size range, 22, 23
Pressure,
.
effect on Umf and Umb, 24, 25
effect on maximum stable bubble size,

61
effect on entraiment, 141
effect on fast fluidization, 177
effect on heat transfer, 230, 237
effect on turbulent fluidization, 157
gradient, 19,21
Pressure drop,
in compressible flow, 20
in conveying, 342-344, 348-351,
354-356
in fluidized bed, 1,22

in gas cleaning, 199


in moving beds, 20-21, 361-363, 374
in packed beds, 19-20
Pressure fluctuations, 33, 162, 163, 182,
418,436, 445
Pressure measurement, 444-445
Probes, 432-437
see also Capacitance measurements
Quality of fluidization,
Quiescent fluidization,

42
11, 25, 26, 33, 34,

40
Quiescent bed, 25, 238
Reaction kinetics, 287
effect of particle size, 317-319
gas/solid reactions, 31~326
Reactor models,
single phase, 295, 332
two phase, 295, 297, 298, 324
three phase, 295, 297, 298
Reactors,
fluidized bed, 53, 74, 285-335
Reaction,
in the freeboard, 32~330
in the grid region, 313-316
Regimes,
bubbling, 33, 34
of fluidization, 4, 33-51, 183-185
non-bubbling, 33, 34
Relative velocity between gas and
particles, 371-376, 391-395
Residence time, 108
of particles, 146,233,276,277
Residence time distribution, 98, 109,
260-261,278-280,295,438
Roasting,
of sulphide ores, 5, 194
of zinc ore, 63, 84, 139,322
Saltation velocity, 351-354, 365
Sampling, see Gas sampling, and Solids
sampling
Saturation carrying capacity, 126, 170,
171, 180, 181
Scale-up, 1,5,6,48,80,210,423,424
of reactors, 287, 309
Scrubber, see Aero-mechanicaI
separators
Segregation, 26, 38, 97,110-114
by density difference, 113
by size difference, 113,

p4

radial, in fast fluidization, 185


Selectivity, 286, 287, 304, 310
Semi-cylindrical columns, 422-423
Shell chlorine process, 6
Shrinkage of particles, 317-319
Shrinking core model, 317-318, 326
Sintering of particles, 24, 42
Slip velocity, 21,174, 175,345
Slugs, slugging beds, 23, 33, 37, 39, 42,
54,62,88-92,125,163-165,365
reactions in, 331
Softening of particles, 24
Softness of particles, 43
Solid-gas separation, 197-217
Solids circulation, 78-80, 10{}-109, 310
and heat transfer, 225, 244
Solids feeding (see also Non-mechanical
valves)
blow tank, 367
Solids mixing, 1,34,37,38,39,53,79,
97-122,437-439
axial, 97, 98
mechanisms, 98
mixing index, 111-113
radial, 97
velocity, 101 (see also Particle
velocity)
Solids sampling, 438
Solids transfer, 341-414
Sphericity, 13, 18, 132
definition of, 12
Splash zone, 126, 329
Spouted beds/spouting, 4, 23, 38, 39, 46,
316
Standpipes, 2, 26, 370-395
aeration of, 383-387
friction loss in, 382
inclined, 387-388
pressure drop in, 377-382
voidage in, 26
Sticky particles,
(see also Cohesive solids) 34, 38, 97,
114-115, 116, 125, 147
Stirrers/stirred beds, 34, 116
Stokes law, 124, 130
Stokes number, 204-206
Stripping,
in standpipes, 388-395
Temperature gradients, 97, 219, 327, 330
Terminal velocity, see Velocity, terminal
Thermal conductivity

of fluidized bed, 219, 226


of gas, 227, 229
Thermocouples, 220, 446-450
Throughflow, see Gas bypassing
Tracers, 98, 100, 109, 18~19O
Transport disengagement height (TDH)
12~128, 134, 139, 142, 148,327
correlations for, 128
measurement of, 443
Transport reactors, 4, 184
Transport velocity, 155, 170-172,
178-183
Tubes,
horizontal, 160,244-245
vertical, 244, 247
Tubes as heat transfer surfaces
effect of mixing, 103
erosion damage to, 244, 246, 250
finned, 250, 251
shallow bed heat exchangers, 250, 251
variation in coefficient around tube,
245
see also Baffles
Turbulent fluidization, 38, 155, 169-171
effect of bed height, 166 I
effect of bed size, 166
effect of particle size, 165
effect of pressure, 157, 166
heat transfer in, 249
reaction in, 331
Two-dimensional beds, 37,185,419-422
bubbles in, 37, 80
Two-phase bubbling bed reactor model
303-309,324
Two-phase theory, 81, 84, 89, 105, 167
Velocity, velocities
excess gas, 45, 107
interstitial gas, 38, 289
minimum bubbling, 24-25, 81, 227
minimum fluidization, 21-24, 29, 46
of bubbles, (see Bubbles-velocity)
of choking, 126,343-348
of complete fluidization, 22, 26, III
of solids, 101
saltation, 351-354, 365
terminal, 60,124,126,170,344,348,
367-368
of spherical particles, 127-132
of non-spherical particles, 132-134
see also..Relative velocity
Vertical mixing, see Solids mixing, axial

Vibrated fluid bed dryer, 259, 263-264,

269
Vibration, 33, 34, 43, 422, 442
Viscosity
effect on fluidization, 24, 42
of fluidized beds, 56, 58, 60, 103
Visible bubble flow rate, 103
Voidage
definition, 17
in fluidized beds, 26-27
of dense phase, 27
prediction of, 18

variation with particle and wall


roughness, 10
variation with particle shape, 18
variation with particle size, 19
variation with size distribution, 19
see also Expansion
Wake, see Bubbles, wake
Wall effects, 62, 65, 160, 424
Windbox, 65, 57, 421

You might also like