Professional Documents
Culture Documents
Handbook of Fluidization and Fluid-Particle Systems
Handbook of Fluidization and Fluid-Particle Systems
FLUIDIZfiTION and
FLUID-PARTICLE SYSTEMS
edited by
Wen-Ching Yang
Siemens Westinghouse Power Corporation
Pittsburgh, Pennsylvania, U.S.A.
CHEMICAL INDUSTRIES
Consulting Editor
HEINZ HEINEMANN
Berkeley, California
1.
2.
3.
4.
5.
6
7
8
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19
20
21
22
23
24
25
26.
27
28
29
30
31.
Fluid Catalytic Cracking with Zeolite Catalysts, Paul B Venuto and E. Thomas Habib, Jr.
Ethylene: Keystone to the Petrochemical Industry, Ludwig Kniel, Olaf Winter, and Karl Stork
The Chemistry and Technology of Petroleum, James G Speight
The Desulfurization of Heavy Oils and Residua, James G. Speight
Catalysis of Organic Reactions, edited by William R. Moser
Acetylene-Based Chemicals from Coal and Other Natural Resources, Robert J. Tedeschi
Chemically Resistant Masonry, Walter Lee Sheppard, Jr.
Compressors and Expanders: Selection and Application for the Process Industry, Heinz P. Bloch, Joseph A.
Cameron, Frank M. Danowski, Jr, Ralph James, Jr, Judson S. Sweanngen, and Marilyn E. Weightman
Metering Pumps. Selection and Application, James P. Poynton
Hydrocarbons from Methanol, Clarence D Chang
Form Flotation: Theory and Applications, Ann N. Clarke and David J. Wilson
The Chemistry and Technology of Coal, James G. Speight
Pneumatic and Hydraulic Conveying of Solids, O. A Williams
Catalyst Manufacture: Laboratory and Commercial Preparations, Alvm B. Stiles
Characterization of Heterogeneous Catalysts, edited by Francis Delannay
BASIC Programs for Chemical Engineering Design, James H. Weber
Catalyst Poisoning, L. Louis Hegedus and Robert W McCabe
Catalysis of Organic Reactions, edited by John R Kosak
Adsorption Technology: A Step-by-Step Approach to Process Evaluation and Application, edited by Frank L.
Slejko
Deactivation and Poisoning of Catalysts, edited by Jacques Oudar and Henry Wise
Catalysis and Surface Science: Developments in Chemicals from Methanol, Hydrotreating of Hydrocarbons,
Catalyst Preparation, Monomers and Polymers, Photocatalysis and Photovoltaics, edited by Heinz Heinemann
and Gabor A Somorjai
Catalysis of Organic Reactions, edited by Robert L. Augustine
Modern Control Techniques for the Processing Industries, T. H. Tsai, J W. Lane, and C. S. Lin
Temperature-Programmed Reduction for Solid Materials Characterization, Alan Jones and Brian McNichol
Catalytic Cracking: Catalysts, Chemistry, and Kinetics, Bohdan W Wojciechowski and Avelino Corma
Chemical Reaction and Reactor Engineering, edited by J. J. Carberry and A. Varma
Filtration: Principles and Practices: Second Edition, edited by Michael J Matteson and Clyde Orr
Corrosion Mechanisms, edited by Florian Mansfeld
Catalysis and Surface Properties of Liquid Metals and Alloys, Yoshisada Ogino
Catalyst Deactivation, edited by Eugene E Petersen and Alexis T. Bell
Hydrogen Effects in Catalysis: Fundamentals and Practical Applications, edited by Zoltan Paal and P G. Menon
32. Flow Management for Engineers and Scientists, Nicholas P. Cheremisinoff and Paul N. Cheremisinoff
33. Catalysis of Organic Reactions, edited by Paul N. Rylander, Harold Greenfield, and Robert L. Augustine
34. Powder and Bulk Solids Handling Processes: Instrumentation and Control, Koichi linoya, Hiroaki Masuda, and
Kinnosuke Watanabe
35. Reverse Osmosis Technology: Applications for High-Purity-Water Production, edited by Bipin S. Parekh
36. Shape Selective Catalysis in Industrial Applications, N. Y. Chen, William E. Garwood, and Frank G. Dwyer
37. Alpha Olefins Applications Handbook, edited by George R. Lappin and Joseph L. Sauer
38 Process Modeling and Control in Chemical Industries, edited by Kaddour Najim
39. Clathrate Hydrates of Natural Gases, E. Dendy Sloan, Jr
40 Catalysis of Organic Reactions, edited by Dale W Blackburn
41. Fuel Science and Technology Handbook, edited by James G Speight
42. Octane-Enhancing Zeolitic FCC Catalysts, Julius Scherzer
43. Oxygen in Catalysis, Adam Bielanski and Jerzy Haber
44. The Chemistry and Technology of Petroleum: Second Edition, Revised and Expanded, James G. Speight
45 Industrial Drying Equipment: Selection and Application, C. M. van't Land
46 Novel Production Methods for Ethylene, Light Hydrocarbons, and Aromatics, edited by Lyle F. Albright, Billy L.
Crynes, and Siegfried Nowak
47 Catalysis of Organic Reactions, edited by William E. Pascoe
48 Synthetic Lubricants and High-Performance Functional Fluids, edited by Ronald L. Shubkin
49 Acetic Acid and Its Derivatives, edited by Victor H Agreda and Joseph R. Zoeller
50 Properties and Applications of Perovskite-Type Oxides, edited by L G Tejuca and J. L. G. Fierro
51 Computer-Aided Design of Catalysts, edited by E Robert Becker and Carmo J. Pereira
52. Models for Thermodynamic and Phase Equilibria Calculations, edited by Stanley I Sandier
53 Catalysis of Organic Reactions, edited by John R Kosak and Thomas A Johnson
54 Composition and Analysis of Heavy Petroleum Fractions, Klaus H. Altgelt and Mieczyslaw M. Boduszynski
55. NMR Techniques in Catalysis, edited by Alexis T. Bell and Alexander Pines
56. Upgrading Petroleum Residues and Heavy Oils, Murray R. Gray
57. Methanol Production and Use, edited by Wu-Hsun Cheng and Harold H. Kung
58. Catalytic Hydroprocessing of Petroleum and Distillates, edited by Michael C. Oballah and Stuart S. Shih
59. The Chemistry and Technology of Coal: Second Edition, Revised and Expanded, James G Speight
60. Lubricant Base Oil and Wax Processing, Avilino Sequeira, Jr
61. Catalytic Naphtha Reforming- Science and Technology, edited by George J. Antos, Abdullah M. Aitani, and
Jose M. Parera
62. Catalysis of Organic Reactions, edited by Mike G. Scares and Michael L. Prunier
63. Catalyst Manufacture, Alvin B. Stiles and Theodore A Koch
64. Handbook ofGrignard Reagents, edited by Gary S. Silverman and Philip E. Rakita
65. Shape Selective Catalysis in Industrial Applications: Second Edition, Revised and Expanded, N. Y. Chen,
William E. Garwood, and Francis G. Dwyer
66. Hydrocracking Science and Technology, Julius Scherzer and A. J. Gruia
67 Hydrotreating Technology for Pollution Control- Catalysts, Catalysis, and Processes, edited by Mario L.
Occelli and Russell Chianelli
68. Catalysis of Organic Reactions, edited by Russell E. Malz, Jr.
69. Synthesis of Porous Materials: Zeolites, Clays, and Nanostructures, edited by Mario L. Occelli and Henri
Kessler
70. Methane and Its Derivatives, Sunggyu Lee
71. Structured Catalysts and Reactors, edited by Andrzej Cybulski and Jacob A. Moulijn
72. Industrial Gases in Petrochemical Processing, Harold Gunardson
73. Clathrate Hydrates of Natural Gases: Second Edition, Revised and Expanded, E Dendy Sloan, Jr
74. Fluid Cracking Catalysts, edited by Mario L. Occelli and Paul O'Connor
75. Catalysis of Organic Reactions, edited by Frank E. Herkes
76. The Chemistry and Technology of Petroleum: Third Edition, Revised and Expanded, James G. Speight
77 Synthetic Lubricants and High-Performance Functional Fluids- Second Edition, Revised and Expanded,
Leslie R. Rudnick and Ronald L. Shubkin
78 7776 Desulfurization of Heavy Oils and Residua, Second Edition, Revised and Expanded, James G. Speight
79. Reaction Kinetics and Reactor Design: Second Edition, Revised and Expanded, John B. Butt
80. Regulatory Chemicals Handbook, Jennifer M. Spero, Bella Devito, and Louis Theodore
81 Applied Parameter Estimation for Chemical Engineers, Peter Englezos and Nicolas Kalogerakis
82 Catalysis of Organic Reactions, edited by Michael E Ford
83 The Chemical Process Industries Infrastructure Function and Economics, James R Couper, O Thomas
Beasley, and W Roy Penney
84 Transport Phenomena Fundamentals, Joel L Plawsky
85 Petroleum Refining Processes, James G Speight and Baki Ozum
86 Health, Safety, and Accident Management in the Chemical Process Industries, Ann Mane Flynn and Louis
Theodore
87 Plantwide Dynamic Simulators in Chemical Processing and Control, William L Luyben
88 Chemicial Reactor Design, Peter Harriott
89 Catalysis of Organic Reactions, edited by Dennis G Morrell
90 Lubricant Additives Chemistry and Applications, edited by Leslie R Rudnick
91 Handbook of Fluidization and Fluid-Particle Systems, edited by Wen-Chmg Yang
92 Conservation Equations and Modeling of Chemical and Biochemical Processes, Said S E H Elnashaie and
Parag Garhyan
93 Batch Fermentation Modeling, Monitoring, and Control, All Qmar, Gulnur Birol, Satish J Parulekar, and
Cenk Undey
94 Industrial Solvents Handbook, Second Edition, Nicholas P Cheremisinoff
Preface
Every chemical engineer, whether a student or practicing, has looked up technical information in Perrys Chemical
Engineering Handbook. Its compilation was one of the most important contributions to the chemical engineering
education and profession. After more than six decades, it remains one of the elds most useful general-purpose
reference books. It was in this spirit of serving the profession that I undertook the task of compiling the Handbook
of Fluidization and Fluid-Particle Systems. Through future revisions and additions, I sincerely hope that this handbook will become an archivable reference volume for every practitioner in this eld, spanning the boundary of
various disciplines. Fluidization and uid-particle system engineering is being applied in industries as diverse as
basic and specialty chemicals, mineral processing, coal and biomass gasication and combustion for power generation, environmental technologies, resource recovery, FCC petroleum rening, pharmaceuticals, biotechnology,
cement, ceramics, and other solids handling and processing industries. The rst focused handbook ever published in
this extended eld, it collects all relevant and important information in a single volume. Both fundamentals and
applications are emphasized. Furthermore, all authors are internationally recognized practitioners in the area of
uidization and uid-particle systems.
This handbook contains 28 chapters and is authored by 34 internationally recognized experts from seven
countries; half of them are professors. Particle characterization and dynamicsimportant in all aspects of particle
production, manufacturing, handling, processing, and applicationsare discussed in Chapter 1. Chapter 2 presents
the ow through xed beds and summarizes packing characteristics of spherical and nonspherical particles, pressure-drop correlations for ow through xed beds, and heat and mass transfer. Bubbling uidized beds are presented in detail in Chapter 3, which covers all important aspects including jetting phenomena and particle
segregation, topics not addressed extensively in other books on uidization. Other important design considerations
are treated in separate chapters: elutriation and entrainment in Chapter 4, eect of temperature and pressure in
Chapter 5, gas distributor and plenum design in Chapter 6, eect of internal tubes and baes in Chapter 7, attrition
in Chapter 8, and modeling in Chapter 9. Heat transfer (Chapter 10) and mass transfer (Chapter 11) are also
treated. The approaches for designing and scaling up uidized bed reactors are elucidated in Chapter 12, General
Approaches to Reactor Design, and Chapter 13, Fluidized Bed Scaleup.
Important industrial applications for uidized bed reactors are also discussed, including uid catalytic cracking
(Chapter 14), gasiers and combustors (Chapter 15), chemical production and processing (Chapter 16), coating and
granulation (Chapter 17), and uidized bed drying (Chapter 18).
The important variation of bubbling uidized bedsthe circulation uidized bedsare discussed in detail in
Chapter 19. Chapter 20 summarizes other nonconventional uidized beds, including spouted beds, recirculating
uidized beds with a draft tube, jetting uidized beds, and rotating uidized beds. The solids handling, transport
and circulating devices are described in Chapter 21, Standpipe and Nonmechanical Valves, and Chapter 22,
Cyclone Separators. Pneumatic transport is covered in Chapters 23 and 24. Instrumentation and measurement
requirements are reviewed in Chapter 25.
The last three chapters examine the uidized beds and uid-particle systems involving liquid.
This handbook took more than four years to complete. Along the way, content was altered, format was changed,
and chapters were revised to t the page limitation. The nal product is indeed one to be proud of by all who
participated. A monumental endeavor such as this could not have been possible without the cooperation and
dedication of all the authors, especially those who were asked to revise their chapters, sometimes several times. I
am truly indebted to them all for taking the time out of their busy schedule and for their cooperation, dedication,
and conscientious eort. The sta of the publisher, Marcel Dekker, Inc., also deserves credit for their patience and
tenacity in shepherding the project to its eventual completion. Finally, I thank my family, especially my wife, Rae,
for their continuous encouragement.
Wen-Ching Yang
Contents
Preface
Contributors
1
Attrition
Joachim Werther and Jens Reppenhagen
Modeling
Thomas C. Ho
10
Heat Transfer
John C. Chen
11
Mass Transfer
Thomas C. Ho
12
13
14
15
16
17
18
19
20
21
22
Cyclone Separators
T. M. Knowlton
23
24
25
26
LiquidSolids Fluidization
Norman Epstein
27
28
LiquidSolids Separation
Shiao-Hung Chiang, Daxin He, and Yuru Feng
Contributors
Hsiaotao Bi Department of Chemical and Biological Engineering, University of British Columbia, Vancouver,
British Columbia, Canada
John C. Chen
Ye-Mon Chen
Shiao-Hung Chiang Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh,
Pennsylvania, U.S.A.
Sakamon Devahastin Department of Food Engineering, King Mongkuts University of Technology Thonburi,
Bangkok, Thailand
Norman Epstein Department of Chemical and Biological Engineering, University of British Columbia, Vancouver,
British Columbia, Canada
Liang-Shih Fan
Department of Chemical Engineering, The Ohio State University, Columbus, Ohio, U.S.A.
Yuru Feng Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh,
Pennsylvania, U.S.A.
Leon R. Glicksman Departments of Architecture and Mechanical Engineering, Massachusetts Institute of
Technology, Cambridge, Massachusetts, U.S.A.
Mohammad Golriz
John R. Grace Department of Chemical and Biological Engineering, University of British Columbia, Vancouver,
British Columbia, Canada
Ernst-Ulrich Hartge
Masayuki Horio
Japan
Behzad Jazayeri
Department of Chemical Engineering, The Ohio State University, Columbus, Ohio, U.S.A.
Peijun Jiang
Yong Jin
S. B. Reddy Karri
George E. Klinzing
T. M. Knowlton
Rafal P. Kobylecki
Tokyo, Japan.
Paul R. Mort
Arun S. Mujumdar
Richard A. Newby Science & Technology Center, Siemens Westinghouse Power Corporation, Pittsburgh,
Pennsylvania, U.S.A.
Jens Reppenhageny Technical University Hamburg-Harburg, Hamburg, Germany
Gabriel I. Tardos Department of Chemical Engineering, The City College of the City University of New York,
New York, U.S.A.
Mayumi Tsukada
Tokyo, Japan
Yao Wang
Fei Wei
Joachim Werther
Guoqiang Yang
Wen-Ching Yang Science & Technology Center, Siemens Westinghouse Power Corporation, Pittsburgh,
Pennsylvania, U.S.A.
J. G. Yates
Current aliation: Energy Engineering Department, Czestochowa Technical University, Czestochowa, Poland.
y Current aliation: BMH Claudius Peters, Buxtehude, Germany.
1
Particle Characterization and Dynamics
Wen-Ching Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
1.1.1
Surface Diameter
The surface diameter, ds , is dened as the diameter of a
sphere having the same surface area of the particle.
Mathematically it can be shown to be
1=2
Sp
where
Sp surface area of
ds
p
the particle 2
SurfaceVolume Diameter
The surfacevolume diameter, dsv , also known as the
Sauter diameter, is dened as the diameter of a sphere
having the same external-surface-area-to-volume ratio
as the particle. This can be expressed as
dsv
6Vp dv3
2
Sp
ds
Feret Diameter
The Feret diameter, dF , is a statistical diameter representing the mean value of the distances between pairs
of parallel tangents to a projected outline of the particle, as shown in Fig. 1. The Feret diameter is usually
used in particle characterization employing the optical
imaging technique to be discussed later.
Sieve Diameter
The sieve diameter, dA , is dened as the width of the
minimum square aperture in the sieve screen through
which the particle will pass. The sieve diameter will be
discussed in more detail when particle size analysis by
sieving is discussed later in the chapter.
Stokes Diameter
The Stokes diameter, dst , is the free-falling diameter of
the particle in the Stokes law region and can be calculated from
s
18mUt
dst
rp rf g
where
terminal
velocity
Ut
4
of the
particle
Free-Falling Diameter
The free-falling diameter, df , is the diameter of a sphere
having the same density and the same free-falling velocity (or terminal velocity) as the particle in a uid of
same density and viscosity. In the Stokes law region,
the free-falling diameter is the Stokes diameter dened
earlier.
Martin Diameter
The Martin Diameter, dM , is also a statistical diameter
dened as the mean chord length of the projected outline of the particle, which appropriately bisects the area
of the projected prole, as shown in Fig. 1. Like the
Feret diameter, the Martin diameter is also employed
most often during particle characterization using the
optical imaging technique to be discussed later.
Only four of the above particle size denitions are of
general interest for applications in packed beds and
uidized beds. They are the sieve diameter, dA , the
volume diameter, dv , the surface diameter, ds , and
the surfacevolume diameter, dsv . The most relevant
diameter for application in a uidized bed is the surfacevolume diameter, dsv . For applications in catalytic reactors with dierent isometric catalyst shapes,
Rase (1990) suggested that we use the equivalent diameters summarized in Table 1.
Drag Diameter
The drag diameter, dD , is dened as the diameter of a
sphere having the same resistance to motion as the
particle in a uid of the same density and viscosity
and moving at the same velocity.
Perimeter Diameter
The perimeter diameter, dc , is the diameter of a circle
having the same perimeter as the projected outline of
the particle.
Figure 1 Illustration for projected area diameter, Feret diameter, and Martin diameter.
Sphere
dp diameter of a sphere
Cylinder, dy 6 ly
dp
6dy
4 2dy =ly
dp 1:5do di
Mixed sizes
dp P
1
xi =dpi
dp dv ; dv dA
1.1.2
Table 2a
Relative Proportions
f dsv =dv
1:1:2
1:2:2
1:1:4
1:4:4
1:2:4
Height diameter
Height 2 diameter
Height 4 diameter
Height 12 diameter
Height 14 diameter
1:1:1
1:1:2
1:2:2
1:1:4
1:4:4
1:2:4
0.93
0.92
0.78
0.70
0.79
0.87
0.83
0.73
0.83
0.69
0.81
0.77
0.77
0.68
0.64
0.68
0.67
0.83
Shape
Spheroid
Ellipsoid
Cylinder
Rectangular parallelpiped
Rectangular tetrahedron
Regular octahedron
Sphericity
0.86
0.66
0.80.9
0.73
0.630.75
0.700.90
0.28
0.58
0.84
0.65
0.700.90
0.89
0.75
0.85
0
B
fop @
Volume of particle
C
A
Volume of the smallest circumscribing
sphere
11=2
7
6op @
b
t
l
e2
b
e1
Flatness ratio
10
Elongation ratio
11
12
Vp
da3
13
p
where da 4Ap =p.
The projected area diameter, da , is the diameter of
the sphere with the same projected area as the particle.
A number of methods have been suggested for obtaining an estimate for da . Even if da is available, the
Heywood shape factor can only be evaluated if Vp is
known. For naturally occurring particles or if a distribution of particle sizes or shapes is present, Vp may not
be readily available. Automatic techniques for charac-
ke
p
e1 e2
14
0.56
0.51
0.47
0.38
0.26
0.23
0.16
0.13
0.16
0.524
0.696
0.328
0.785
0.547
0.262
0.370
1.1.3
Mp
mass of the particle
PARTICLE CHARACTERIZATION
TECHNIQUES
2.1
2.1.1
mended. Detailed discussion of sieve analysis techniques can be found in standard textbook such as Allen
(1975) and Kaye (1981).
Sieve Analysis
2.1.2
Imaging Technique
Direct measurement of particle dimensions is also possible from enlarged photographic or electronic images
of microscopes. There are three types of microscopes
commonly employed, i.e., the optical microscope, the
scanning electron microscope (SEM), and the transmission electron microscope (TEM). The optical
microscope is employed for particles from 1 mm to
about 150 mm. Both SEM and TEM make use of electron beams and can be used for particles from 5 mm
down to as small as 0:01 mm. They are especially useful
for revealing the surface morphology of extremely
small particles.
Particles to be imaged in an optical microscope are
usually dispersed in a drop of viscous uid on a glass
slide. Their images are then visually compared with a
set of standard circles, geometric shapes, or linear grids
to derive their actual sizes and shapes. The Martin and
Feret diameters are also often used to characterize a
particle. The Martin diameter is dened as the magnitude of the chord that divides the image into two equal
areas with respect to a xed direction (see Fig. 1). The
Feret diameter of a particle image is dened as the
length of the image as projected with respect to a
specic reference direction (see Fig. 1). Both Martin
and Feret diameters are intended to be statistical diameters, i.e., after characterization of many images.
Thus any slight departure from the true randomness
in the eld of view of particle images can produce bias
in the particle size characterization. Modern instruments couple television cameras interfaced with computers and sophisticated software can speed up the
imaging analysis considerably.
Martins statistical diameter is also referred to as the
mean linear intercept or mean chord and has been
shown to relate to the specic surface in the following
manner (Herdan, 1960):
Martin diameter
4
rSv
16
8
10
12
14
16
18
20
25
30
35
40
45
50
60
70
80
100
120
140
170
200
230
270
325
400
635
Standard
opening, mm
Tyler
mesh No.
British
mesh No.
Standard
opening, mm
German
DIN No.
Standard
opening, mm
5.66
4.76
4.00
3.36
2.83
2.38
2.00
1.68
1.41
1.19
1.00
0.84
0.71
0.59
0.50
0.42
0.35
0.297
0.250
0.210
0.177
0.149
0.125
0.105
0.088
0.074
0.062
0.053
0.044
0.038
0.020
312
4
5
6
7
8
9
10
12
14
16
20
24
28
32
35
42
48
60
65
80
100
115
150
170
200
250
270
325
400
5
6
7
8
10
12
14
16
18
22
25
30
36
44
52
60
72
85
100
120
150
170
200
240
300
3.353
2.812
2.411
2.057
1.676
1.405
1.204
1.003
0.853
0.699
0.599
0.500
0.422
0.353
0.295
0.251
0.211
0.178
0.152
0.124
0.104
0.089
0.076
0.066
0.053
212
3
4
10
11
12
14
16
20
24
30
40
50
60
70
80
100
6.000
3.000
2.400
2.000
1.500
1.200
1.020
0.750
0.600
0.540
0.490
0.430
0.385
0.300
0.250
0.200
0.150
0.120
0.102
0.088
0.075
0.060
diameter is usually fairly constant for the same material. For Portland cement, it is about 1.2; for ground
quartz and ground glass, approximately 1.3 (Herdan,
1960).
The microscopic measurement technique is most
suitable for particles relatively uniform in size and
granular in shape, because a large number of particles,
Figure 3 Graphic representation of sieve size analysislogarithmic-scale graph expressing cumulative % oversize or undersize.
size in the gravity sedimentation technique. The measured speeds are then converted to Stokes diameters by
applying the Stokes equation [Eq. (4)], assuming that
the particles are all spherical in shape. Since the irregularly shaped particles fall with dierent orientations
in the vertical direction and thus dierent settling velocities, similar irregularly shaped particles can have a
range of Stoke diameters. Disturbances caused by the
presence of other particles, the concentration eect,
can also be important. Recommendations for the
upper limit particle concentration to be used during
the sedimentation analysis varies from 0.01 to 3.0%
(Kaye, 1981). An upper limit of 0.05% was recommended by Kaye, and in certain practical cases up to
0.2% by volume is permissible. The probability of
forming clusters increases with particle suspension
concentration. The clusters tend to fall at a higher
speed and thus introduce measurement error.
The hindering eect of the containing wall on the
falling speed of the particles cannot be ignored either.
For a spherical particle, the eect can be expressed by
the Landenburg equation as
dp
V1 Vm 1 2:4
D
17
18
19
Characterization by Elutriation
As in sedimentation, the concentration of the particles in the elutriators aects the results of measurement. Also the velocity in the elutriators tends to be
parabolic rather than uniform and thus introduces
errors in measurement. Centrifugal force can also be
applied like that in the sedimentation to enhance the
performance. Dierent designs, including horizontal
elutriators, are available.
2.1.5
Figure 5
Laser-diraction spectrometry.
Sensing technique
DR10
Electrical sensing zone
Laser diraction and scattering
X-ray sedimentation
Photosedimentation
Light attenuation
DR50
Electrical sensing zone
Laser diraction and scattering
X-ray sedimentation
Photosedimentation
Light attenuation
DR90
Electrical sensing zone
Laser diraction and scattering
X-ray sedimentation
Photosedimentation
Light attenuation
Aluminum
oxide
Barium
titanate
Boron
nitride
Silicon
nitride
whisker
1.00
0.61
0.82
0.82
1.00
1.00
0.77
1.10
0.95
1.15
1.00
0.94
0.58
0.96
1.00
0.61
0.69
0.75
0.96
1.00
0.97
0.84
0.78
1.33
1.00
1.11
1.05
0.96
1.20
1.00
1.32
0.74
1.11
1.00
1.41
0.86
1.05
1.04
1.00
1.15
0.90
1.01
1.20
1.00
1.33
1.14
1.17
1.46
1.00
1.56
1.53
1.78
1.00
2.32
1.01
2.43
0.98
2.3
2.3.2
2.3.1
20
Attrition Index
Table 6 Applicable Particle Size Ranges for Various Particle Sizing Methods
Particle sizing methods
Sieving
Dry
Wet
Microscopic examination
Optical
Electronic
Zone sensing
Resistivity
Optical
Elutriation
Laminar ow
Cyclone
Gravity sedimentation
Pipette and hydrometer
Photoextinction
X-ray
Centrifugal sedimentation
Mass accumulation
Photoextinction
X-ray
Centrifugal classication
Gas permeability
Gas adsorption
Cascade impaction
Measured dimension
> 10
2500
Sieve diameter
1.0100
0.01500
0.61200
1.0800
Volume
375
850
Stokes diameter
1100
0.5100
0.1130
Stokes diameter
0.525
0.05100
0.15
0.550
0.140
0.00550
0.0530
Stokes diameter
Abrasiveness Index
Erosiveness Index
3.1.1
CD
F
CD
1=2rf Ur2 Ap
21
1
F CD rf Ur2 Ap
2
22
24
Rep
for
23
or
3.1.2
CD f bRep c
for
24
CD 0:44
for
25
Fs 3pmdp Ur
26
27
33
or
m
CD 24
dp Ur rf
24
Rep
28
The linear relation between the Fs and Ur is an important property of the Stokes law. It enables considerable
simplications in many problems.
3.3
29
4 d p r p rf g
CD
3
rf Ur2
30
or
More recently, Proudman and Pearson (1957) provided the key to a more straightforward calculation
of further correction terms. The results for the drag
force on the sphere in accordance with Stokes,
Oseens, and Proudman and Pearsons calculations
are given in Table 7. A very large and important collection of solutions of such problems can be found in
the book by Happel and Brenner (1965).
3.6 Empirical Drag Coefcient Expression
The preferred correlations to be used in dierent
ranges of particle Reynolds numbers were recommended by Clift et al. (1978) and are summarized in
Table 8. In 1986, Turton and Levenspiel proposed a
single correlation applicable for the complete range of
Reynolds numbers and considerably simplied the calculation of the single-particle drag coecient.
The Turton and Levenspiel (1986) equation is
or
Ur Ut
dp2
rp r f g
18m
CD
31
Wall Effect
i
24 h
1 0:173Re0:657
p
Rep
0:413
1 16:300Rep1:09
34
Fs 3pdp mUr
n
o
3
F Fs 1 Rep O Re2p
16
Proudman
and Pearson (1957)
n
o
Rep
3
9
F Fs 1 Rep
O Re2p
Re2p ln
2
16
160
Table 8
CD
log10
CD
20 Rep 260
log10
CD
3
24
16 Rep
CD Rep
1 0:881 0:82w 0:05w2
24
i
24 h
1 0:1315Re0:820:05w
p
Rep
CD Rep
1 0:7133 0:6305w
24
i
24 h
1 0:1935Re0:6305
p
Rep
CD 29:78 5:3w
CD 0:1w 0:49
CD 0:19
8 104
Rep
CD
i
24 h
1 8:1716e4:0655f Rep0:09640:5565f
Rep
73:69 e5:0748f Rep
Rep 5:378e6:2122f
35
0:4607Rep
24
3:3643Re0:3471
p
Rep 2682:5
Rep
36
3.7
CD l
5:96 5:51f
for
2
103 < Ret < 2
105 ;
0:67 < f < 1
39
for
CD 0:95
42
for
Ar > 1:18
106 dp2
43
dp being in meters.
4
3.8
dp
"
#1=3
1=3
rf rp r f g
3
1=3
2
dp
Ar
Re
C
p
4 D
m2
45
"
U U
#1=3
r2f
m rp rf g
1=3
Rep
4 Rep
1=3
3 CD
Ar
5
X
49
an X n
n0
46
4.2
acceptance. The tables, however, are not very convenient to use, especially for applications using a computer. The tables are now tted by Fouda and Capes
(1976) in the form of
48
where
and
X log10
Ut
Q
Ut
Q
or log10 Pdt
Fouda and Capes (1976) proposed polynomial equations tted to the Heywood (1962) tables for calculating large numbers of terminal velocities. The Heywood
tables were recognized as being a simple and accurate
method for calculating both the terminal velocity and
the equivalent particle diameter, and they enjoyed wide
51
53
50
log10 Ut =Q
log10 Pdt
log10 Pdt
log10 Ut =Q
Polynomial coecients
a0
a1
a2
a3
a4
a5
a0
a1
a2
a3
a4
a5
1.37323
2:06962
0:453219
0:334612 101
0:745901 102
0:249580 102
0:785724
0:684342
0:168457
0:103834
0:20901 101
0:57664 102
54
and
logRet Y; 1 0:77481 0:56032 log Y
0:024246log Y2 0:0038056
log Y3
55
#
"
4 g rp rf m
Y
3
Ut3 r2f
57
These equations are good for 0:01 < Ret < 16000
and 0:67 < f < 1. The equation tted the experimental
terminal velocities of limestone and lime particles to
better than 20%
4.3
58
59
Ut ip
60
61
where
4dp2 g 1 e rp rf
CD Rep
3mUt p
62
and
4dp2 g rp rf
24
CD Reip
3mUt ip
R
b p
k
63
i
24 h
1 0:0853Rep1:0930:105w
Rep
65
As far as the wall eect is concerned, the experiments show that, at low Reynolds numbers, the wall
eect for a porous sphere is of the same order of magnitude as that for an impermeable sphere. At high
Reynolds numbers, however, the wall eect become
smaller and less signicant for a porous sphere.
CHARACTERIZATION OF MULTIPARTICLE
SYSTEMS
In most applications, the systems and processes contain large amounts of particles with size distribution;
each size may also possess a distinguished shape. To
describe properly these systems and processes for
design and analysis, they need to be adequately
characterized to reect their physical and chemical
potentials. In the following sections, dierent average
particle diameter denitions are introduced along with
statistical descriptions of particle size distribution.
Depending on applications, one denition may be
more suitable than others. Thus care must be exercised
to select the proper characterization for each process.
5.1
Arithmetic Mean
The arithmetic mean is dened as the sum of all diameters divided by the total number of particles
P
ni dpi
i
P
66
dav
ni
i
vP
u n d2
u
i pi
u
ds t iP
ni
i
5.1.3
Volume Mean
5.1.4
VolumeSurface Mean
5.1.5
3
ni dpi
2
ni dpi
1
Px
i
dpi
i
69
Weight Mean
where
3
ni dpi
vi P 3
ni dpi
71
Surface Mean
67
5.1.6
Length Mean
P
i
dl P
2
ni dpi
ni dpi
72
5.1.7
Geometric Mean
Harmonic Mean
dav log1 log dg 1:151 log2 sg
d1=2 dg
dm log1 log dg 2:303 log2 sg
76
77
78
Figure 10
z dc
i
z P
dc
i
z lndpi ;
83
84
85
86
87
88
89
90
91
92
93
94
2
2
2
80
5.3
95
loglog
100
constantn logSa
Wf
96
97
when the same powder is pulled on top of a at surface. Normal powders have angles of repose of around
35 . The angles of repose for commonly occurring
powders are summarized in Table 10.
The angle made by the free surface of the powder
still remaining in the test device with the at bottom of
the test device, the angle a in Fig. 12, is called the angle
of internal friction of the powder. The angle of internal
friction is also an intrinsic property of the powder and
can be considered as a shearing plane of the powder. In
order for the powder to ow, the angle has to be higher
than the angle of internal friction of the powder.
Normal powders have angles of internal friction of
around 70 . The angles of internal friction for commonly occurring powders are also listed in Table 10.
Both the angle of repose and the angle of internal
friction can be modied and reduced by ow aids
such as aeration, vibration, and the addition of other
powders.
Material
FCC catalyst
Sand
Resin
Wet ashes
Wheat
Oats
Angle of repose
degree
Angle of
internal friction
degree
32
36
29
50
21
79
64
82
55
NOMENCLATURE
Ap
Ar
as
b
C
CD
da
dA
dav
dc
dD
df
dF
dg
=
=
=
=
=
=
=
=
=
=
=
=
dgc
dgm
dh
di
dl
dst
dsv
dt
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
dv
dv
dvs
dw
dy
D
DR10
=
=
=
=
=
=
=
dM
dm
do
dp
dpi
dp
ds
ds
DR50 =
DR90 =
d1=2
e1
e2
f
F
Fs
g
=
=
=
=
=
=
=
dp3 rf rp rf g
Archimedes number, Ar
m2
exterior surface area/volume of catalyst pellet
breadth
concentration of particle suspension
drag coecient
projected-area diameter
sieve diameter
arithmetic mean particle diameter
perimeter diameter
drag diameter
free-falling diameter
Feret diameter
geometric mean particle diameter or geometric
median particle diameter
geometric median diameter by count
geometric median diameter by weight
harmonic mean particle diameter
inside diameter of a ring
length mean particle diameter
Martin diameter
mode particle diameter
outside diameter of a ring
equivalent particle diameter
particle diameter of size i
dimensionless particle diameter
surface diameter
surface mean particle diameter
Stokes diameter
surfacevolume diameter
particle diameter of a sphere having the same
terminal velocity as the particle
volume diameter
volume mean particle diameter
volumesurface mean particle diameter
weight mean particle diameter
diameter of a cylinder
sedimentation column diameter
particle diameter ratio at the cumulative mass of
10%
particle diameter ratio at the cumulative mass of
50%
particle diameter ratio at the cumulative mass of
90%
median particle diameter
atness ratio
elongation ratio
frequency of observation
force exerting on a particle
force exerting on a sphere
gravitational acceleration
Ic
Io
k
k
K
kc
ke
l
ly
L
Lw
M
ni
R
Rep
Ret
Sa
Sa m
Sp
Sv
t
T
U
Ur
Ut
Ut ip
Ut p
Vm
Vp
V1
Wf
Ws
W200
xi
Xc
Xo
m
r
rf
rp
f
fop
fw
6
6op
a
=
=
=
=
b
=
=
b
b
l
=
=
=
=
s
sg
=
=
=
working sphericity
circularity
operational circularity
an empirical constant depending on equipment
and particle characteristics
angle of internal friction
an empirical constant depending on equipment,
particle, and suspending uid characteristics
normalized sphere radius
angle of repose
density ratio
ratio of resistance experienced by a porous
sphere to an equivalent impermeable sphere
number, surface, or weight
standard deviation
geometric standard deviation
REFERENCES
S Akapo, TM Khong , CP Simpson , B Toussaint , JG Yates.
A novel method for determining the particle density of
porous aeratable powders. Powder Technol 58:237242,
1989.
T Allen. Particle Size Measurement. London: Chapman and
Hall, 1975.
B Aschenbrenner. A new method of expressing particle
sphericity. J Sediment Petrol 26:1531, 1956.
LG Austin. Conversion factors to convert particle size distributions measured by one method to those measured by
another method. Part Part Syst Charact 15:108111, 1998.
Babcock & Wilcox. Steam. 40th ed. Barberton: The Babcock
& Wilcox Company, 1992, pp 12-812-9.
HA Becker. The eects of shape and Reynolds number on
drag in the motion of a freely oriented body in an innite
uid. Can J Chem Eng 37:8591, 1959.
CF Bohren, RD Human. Absorption and Scattering of
Light by Small Particles. New York: Wiley-Interscience,
1986.
MP Bonin, M Queiroz. Local particle velocity, size, and concentration measurements in an industrial-sale, pulverized
coal-red boiler. Comb and Flame 85:121133, 1989.
B Buczek, D Geldart. Determination of the density of porous
particles using very ne dense powders. Powder Technol
45:173176, 1986.
EB Christiansen, DH Barker. The eect of shape and density
on the free settling of particles at high Reynolds numbers.
AIChE J 11:145151, 1965.
R Clift, JR Grace, ME Weber. Bubbles, Drops, and
Particles. New York: Academic Press, 1978.
RP Davuluri, TM Knowlton. Development of a standardized
attrition test procedure. In: LS Fan, TM Knowlton, eds.
Fluidization IX. New York: Engineering Foundation,
1998, pp 333340.
N Epstein, G Neale. On the sedimentation of a swarm of
permeable spheres. Chem Eng Sci 29:18411842, 1974.
2
Flow Through Fixed Beds
Wen-Ching Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.
PACKING CHARACTERISTICS OF
MONOSIZED SPHERES
For convenience, the regular packing can be considered to be constructed from regular layers and rows.
The two basic layers are the square layer with a 90
Figure 2 Six possible arrangements of monosize spheresone cubic (Figure 2a), two orthorhombic (Figures 2b and 2c), one
tetragonalspheroidal (Figure 2d), two rhombohedral (Figure 2e).
Copyright 2003 by Taylor & Francis Group LLC
Points of
contact
Porosity, %
Surface of spheres
per cm3
Cross-sectional
pore area per cm2
47.64
1.57/R
0.2146
8
10
39.54
30.19
1.81/R
2.10/R
0.2146
0.093
12
25.95
2.22/R
Cubic
Orthohombic
(two orientations)
Tetragonal-Spheroidal
Rhombohedral
(two orientations)
1.4
61 e
pdp3
Table 2 Correspondence of
Coordination Number and
Voidage
n 22:47 39:39e
Coordination Number
3
4
5
6
7
8
9
10
11
12
Voidage
0.7766
0.6599
0.5969
0.4764
0.4388
0.3955
0.3866
0.3019
0.2817
0.2595
61 e
dp
1 es
1 Xb aXb 1 es
ds
db
where
The packing of binary mixtures of spheres is the simplest system and thus has received a lot of studies. The
packing of binary mixtures of spheres depends on their
diameter ratio and the percentage of large particles in
the mixtures. Following McGearys (1967) development, the minimum fraction of the coarse particles
for closest packing is 73%, and the closest packing
voidage is 0.14.
The dependence of maximum observed binary
mechanical packing of spheres on the ratio of diameters between the large and small spheres shows
that when the size ratio, or diameter ratio, decreases,
the changes in voidage increase. With a diameter ratio
larger than about 7 to 1, the change in voidage is only
very slight. This ratio is very close to the critical
diameter ratio of 6.5 for the smaller sphere to pass
through the triangular opening formed by three larger
spheres in close packing (see Fig. 7).
Abe et al. (1979) carried out theoretical analysis of a
packed bed of binary mixture and proposed equations
to estimate voidage at dierent degree of mixing of
binary mixtures of spherical and nonspherical particles. If cohesive forces are absent, the binary mixture
will yield a packed bed of minimum voidage at Xmin , as
shown in Fig. 3. In the region Xb < Xmin , large particles are distributed randomly and evenly in the packed
bed. This state is called complete mixing. In the region
Xb > Xmin , there are two dierent states possible. In
the rst state, the small particles are distributed in
a 1 fc
fc 1
for
ds
0:5
db
fc 1:2
for
ds
0:25
db
fc 1:4
for
ds
0:125
db
1 es Xb X b
2
Xb 1 X b 1 1 es a
X b 1 es
Xb 1 X b 1 1 es a
10
where
X b
a
a 1 aa1 es
11
and
a
1 eb
1 1:5ds =db 3
12
For Xmin Xb 1;
eav 1
1 eb
Xb 1 bs bc 3
13
where
s
1
1
bc 3
Xb
and
bs
Figure 3 Packed bed of minimum voidage for binary particle mixtures.
n
ds
db
14
15
1
1:355r1:566
p
G
for
rp 0:824
17
and
1
1
G
for
rp 0:824
18
where
rp
ds
db
19
20
Vb Vs
V b Vs 1
21
1 eb
1 eb es
22
16
The empirical constant G is independent of the composition of the mixtures but depends on the size ratio
of the particles. Yu et al. (1993) gave the following
values for G:
V V b Xb 2
V Vb Xb
2G
Vs
Vs
2
V X b V s Xs
V Xb Vs Xs
1
Vb 1
Vb 1
Figure 5 Eect of changing particle size ratio on the packing of binary particles. (Adapted from Finkers and Homann, 1998).
3
1
eb 1 r p
24
rstr
1 es
The structural ratio is then used to calculate the
empirical constant G for use in Eq. (16).
G rkstr 1 ek
25
b
A value of 0:63 was suggested for k in Eq. (25). The
proposed approach is good for both spherical and
nonspherical particles. For binary particles with size
distribution in each fraction, the approach did not
fare as well. By changing the value k to 0:345, the
proposed approach gave excellent results for data by
Sohn and Moreland (1968). It was suggested that k
relates to the particle size distribution in each fraction
to provide an even more general equation of particle
packing in a packed bed.
Packings of ternary and quaternary mixtures of
solid particles, considerably more complex systems,
Figure 7
ing.
26
6:464
dS 2= 3 1
27
1:366
dM 0:732
28
db
1
4:444
dM 0:225
30
dc
Dc
31
db
1
2:415
dM 0:414
rings, and Berl saddles. Coulson (1949) studied packings of cubes, cylinders, and plates and found that the
results depend on the height of fall of the particles. His
results are shown in Table 3a. For random packed
beds of uniform-sized particles, Brown (1966)
suggested that the packing porosity depends on the
sphericity of the particles and can be related as
shown in Fig. 8 and Table 3b based on experimental
ndings. For small grains observable under the microscope, the sphericity can be obtained following Wadell
(1935) and the average values obtained from the following equation after observing a sample of particles.
Id
Ic
e0;cylinder
e
Ic Id
Ic Id 0;disk
36
Normal charge
Dense packed
Tablets
Extrudates
Short
Long
Spheres
Uniform
Mixed sizes
Irregular
0.36
0.31
1 0:43 dp =D
0.40
0.46
0.33
0.40
1 0:46 dp =D
1 0:46 dp =D
0.40
0.36
0.42 (average)
0.36
0.32
0.42 (average)
1 0:42 dp =D
1 0:3 dp =D
0.85
0.80
0.75
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
0.80
0.75
0.70
0.67
0.63
0.59
0.55
0.51
0.48
0.45
0.42
0.40
0.38
0.36
0.34
0.32
4.1
HR 1:478
100:136f
39
Figure 8 Relation between packing voidage and sphericity. (Adapted from Zou and Yu, 1996).
Particle shape, particle size, particle size distribution, particle coecient of restitution, particle
surface properties (friction)
Container shape, container size, container surface
properties (friction)
Deposition method, deposition intensity, velocity of
particle deposition
Vibratory compaction, pressure compaction
dp
dp
e 0:6
40
ew 1 0:6
D
D
dp
dp
0:6
41
ec e 0:31 e 1 0:6
D
D
where e is the average voidage experimentally measured or the ratio of void volume to the total volume
of the packed bed. When the ratio of particle diameter
to the bed diameter is less than 0.02, the correction for
wall eect becomes negligible.
Figure 9 is a plot obtained by Benenati and
Brosilow (1962) in a bed of spheres showing the radial
oscillations of voidage away from the containing
cylindrical wall. The observation was conrmed by
Propster and Szekely (1977). Experiments performed
by Goodling et al. (1983) also indicated that for
uniformly sized spherical particles, the oscillations in
voidage can be up to 5 particle diameters from the
wall. The oscillations in voidage are down to 2 to 3
diameters from the wall, for a mixture of two spherical
sizes, and down to 1 particle diameter from the wall for
a mixture of three particle sizes.
Propster and Szekely (1977) also found that a
marked local minimum in voidage existed in the inter-
Figure 9 Schematic radial oscillations of voidage away from the containing cylindrical wall. (Adapted from Propster and
Szekely, 1977.)
have correlations for xed bed bulk voidage for commonly used catalysts such as spheres, cylinders, and
rings. The following correlations are those proposed
by Dixon (1988). A more comprehensive review was
conducted by Haughey and Beveridge (1969).
For spherical particles
2
dp
dp
dp
0:5 42
0:412
e 0:4 0:05
D
D
D
dp
dp
e 0:528 2:464
0:5
0:5
0:536
D
D
43
3
0:5
dp
dp
1
e 1 0:667
2
D
D
dp
0:536
D
44
2
dpv
dpv
e 0:36 0:10
0:7
D
D
dpv
0:6
D
45
e 0:677 9
dpv
0:625
D
2
0:6
dpv
0:7
D
46
e 1 0:763
dpv
D
2
dpv
0:7
D
47
Darcys Law
The theory of laminar ow through homogeneous porous media is based on a classical experiment originally
performed by Darcy in 1856. Darcys experiment is
described in Fig. 10. The total volume of the uid
percolating through the xed bed, Q, can be expressed
in terms of the height of the bed and the bed area as
Q
KAh2 h1
h
49
P2 rf gh2 z2
51
dimensions M1 L3 T1 , has to be elucidated. The dependence of K 0 on the porous medium and on the liquid
has to be separated for practical applications. Nutting
(1930) proposed that
k
K0 s
m
52
rf rf P
54
@rf
div rf !
g
@t
55
ReB
Udp rf
m1 e
57
P dp
e3
fB
L rf U 2 g 1 e
58
61
edp
e
S 61 e
62
or
f /
1 e3n
1
3
e
Re2n
p
67
1 e2
e3 Rep
68
1 e
e3
69
Because of the large dependence on the voidage, a 30to-70-fold increase in f is predicted as the voidage
changes from 0.3 to 0.7 in the turbulent and viscous
ranges, respectively. The expression found above for f
at low Reynolds numbers has been experimentally
proved to be correct. At high Reynolds numbers,
however, the expression seems to predict a somewhat
stronger dependence of the friction factor on the
voidage than is actually found experimentally. For a
single particle, voidage becomes 1 by denition, and
the Blakes Reynolds number approaches innity.
This is not surprising, since the Blake analogy based
on a capillary ow analogy breaks down in this
range, because such an analogy has no physical meaning.
7.3
61 e
S
dp
63
64
gdp
L
65
or
1 e3n
e3
66
Udp rf
mem
70
7.4
for
71
3
L
ge3 dp2
ge Vp =Sp 2
72
73
Vp
51 e2 mUL
The pipe ow analogy breaks down beyond the creeping ow range. Thus beyond the creeping ow range,
purely empirical models must be employed, the only
remaining link with the pipe ow is the structure of the
dimensionless groups.
7.5
this type is by Ergun (1952) employing Blakes denition of drag coecient and the Reynolds number, as
shown in Eqs. (57) and (58).
P gdp f
e3
1 e
75
0:875
2
L 2rf U 1 e
fRep
74
75
with
B
ln H0 ln H1
t
H0 > H1 > L
76
7.6
fp
rf U 2 L
1 ee4:8
dp
79
80
Equation (79) well represents the steady-state expansion characteristics of a turbulent regime in uidized
beds.
7.7
P dp e3
1501 e
1:75
Rep
rf U 2 L1 e
81
P dp e4:8
17:3
0:336
rf U 2 L1 e Rep
82
Equation (82) gives a signicantly better representation for data at higher Reynolds numbers, where
Ergun equation consistently overestimates the
observed friction factor. Wentz and Thodos (1963)
also proposed a general equation for friction factor
for packed and distended beds of spheres:
fw
0:351
1:2
Re
0:05
83
84
85
CD f
es 0:803e1 e
8f e2
;
31 e 1 K1 e1=3
f
P gdp
L 2rf U 2
87
and
"
4:8
CD f 0:63
Ref
#2
88
where K is a constant
The correlation allows the prediction of pressure
drop or drag in single-particle and multiparticle system
with a single curve. The correlation provides good
agreement with data in the creeping ow and intermediate regimes. Deviation is observed in the turbulent
regime. For the data in the distended bed, the correlation also provides a good t in the highly turbulent
range. The range of applicability may be extended by
the application of average particle size denitions,
shape factors, and wall eect correlations.
7.10
89
k
2e3
95
edp
61 e
96
90
94
edp
61 eM
97
and
M 1
4dp
6D1 e
98
91
92
93
Figure 11
Radial gas velocity proles in a 50.8 mm and 101.6 mm diameter bed. (Adapted from Schwarz and Smith, 1953.)
qp hs ap Ts T
99
accuracy to 25%
100
q hi Ai T Tw
101
102
dp
0:3
D
dp
0:2
D
103
9.1.3
q hw Ai TR Tw
104
dp
0:3
D
105
9.1.4
dp
0:2
D
106
hw dp e
0:27
kg 1 e
9.2
107
500 <
dp Urf
< 6000
m1 e
dp
< 0:15
D
108
110
rf Cpg Dp U
kg
111
Pep
Ped
rf Cpg dp U
kg
114
n 0:14Ped
113
where
Ped dp
1 e2
7:83
Lp
e
For Ped < 10, Eq. (115) implies that just one particle
layer is enough to heat the percolating gas to the surface temperature of the particles.
9.3
116
at
y0
and
0 < z < Lh
117
T To
at
y1
118
T To
at
z0
119
@z Cpg G @r2
r @r
!
120
at
r Ry
and
0 < z < Lh
@T
@2 T1 1 @2 T1
rf U1 Cpg 1 ke1
@z
r @r2
@r2
121
T To
at
r1
122
T To
at
z0
123
hav
s
4 ke Cpg G
p Lh
124
hav
s
4 ke Cpg G 1 ke
p Lh
2 Ry
125
126
2
1=kg 1=ks
ks
k
ln s 1
ks kg
kg
0:0935kg
127
Botterill and Denloye (1978) conducted an even more
detailed analysis of the heat transfer between an
immersed cylindrical heater and the packed bed by
dividing the heat transfer into two regions, a region
of higher voidage within half a particle diameter
from the immersed cylindrical surface, and the region
of constant voidage outside the wall region. The radial
transfer of heat for the wall region (region 1) is
governed by
and
!
at
dp
R t < r < Rt
2
z>0
128
@z
r @r2
@r2
Heat transfer in region 1 was assumed to take place by
three dierent mechanisms, neglecting the heat transfer
through radiation. The three mechanisms are (1) heat
transfer by turbulent diusion in the radial direction,
ka , (2) heat transfer by molecular conduction in the
uid boundary layer, kb , and (3) heat transfer
through the thin lm near particle contact points,
kc . Mechanisms (1) and (2) operate in series while
mechanism (3) operates in parallel. The resulting
eective bed conductivity can thus be expressed as
ke1 kc
1
kg
kg kg =ka kg =kb
130
ka
0:045Re1 Pr
kg
131
where
Re1
d p U 1 rf
m
and
Pr
Cpg m
kg
132
kc
1 ew
ew
kg
2fw 2=3kg =ks
133
where
(
)
1
K 1=K 2
1
fw
4 ln K K 1=K
3K
ew 1
1 e 0:7293 0:5139Y
1Y
where Y dp =2R.
ks
134
kg
135
"
for
2 #
dp
1 19:4
D
145
137
ke2 Keo
0:125Re2 Pr
kg
kg
for
NOMENCLATURE
Ai
ap
CD
CD1
CD f
Cj
Cjs
=
=
=
=
=
=
=
Cpg
D
D
D
db
dc
=
=
=
=
=
=
Dc
dM
dp
dpe
dpv
=
=
=
ds
dve
dyi
dyo
f
fB
fe
fp
=
=
=
=
=
=
=
=
fw
g
!
g
G
G
hav
=
=
=
=
=
hi
hs
hw
h1
h1
=
=
=
=
138
Equations (137) and (138) need to be solved numerically. The agreement with experimental data was
reported to be within 25%. The prediction of
Gabors model was consistently higher than the
experimental data. The theoretical predictions show
an almost linear dependence of the heat transfer coefcient on the particle Reynolds number. Because of
its simplicity, Gabors model may be used as a rst
approximation.
9.4
m
2 1:1Sc1=3 Re0:6
rf qj
with accuracy to 25%.
The eective radial diusivity is
@Cj
Njr qr
@r z;r
eqr
1 0:38
Udp m Re
140
141
142
for
Re > 400
143
144
h2
h2
Ic
Id
k
k
K
=
=
=
=
=
=
=
K
K0
ks
=
=
=
kb
kc
Kd
ke
ke1
ke2
koe
=
=
=
=
kg
ks
ks
ksj
Ks
=
=
=
=
=
L
Lh
Lp
m
mp
n
n
Np
Nu
NuD
Ped
Pep
P
P1
P2
P
Pe
q
!
q
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Q
qp
r
rp
=
=
=
=
rstr
R
Rt
Rh
Re
ReB
Rep
=
=
=
=
=
=
Repe
Ref
Re1
Re2
Ry
S
=
=
=
=
=
Sc
Sp
T
To
TR
Ts
Tw
Tw
T1
T2
U
U1
U2
V
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Vb
Vp
Vs
Xb
Xmin
=
=
=
=
=
Xs
y
z
z
z1
z2
qj
b
m
rf
rp
f
=
=
=
=
=
=
=
=
=
=
=
=
fc
radius of sphere
radius of the heater
hydraulic radius
Reynolds number, DUrf =m
Reynolds number proposed by Blake, Eq. (57)
Reynolds number based on particle diameter,
dp Urf =m
Reynolds number proposed by Brownell and
Kats, and dened in Eq. (70)
Reynolds number for multiparticle systems
Reynolds number in region 1
Reynolds number in region 2
radius of cylindrical heater
specic particle surface area; total particle
surface area per unit volume of bed
Schmidt number, m=qj rf
particle surface
temperature
initial gas temperature
temperature at inside radius of the vessel
temperature of a single particle
temperature at the heater wall
temperature at wall
temperature in region 1
temperature in region 2
supercial uid velocity
supercial uid velocity in region 1
supercial uid velocity in region 2
specic volume of solid material,
V 1=1 e
volume of big particles
particle volume
volume of small particles
volume fraction of big particles
volume fraction of large particles where the
voidage of the binary mixture is the minimum
volume fraction of small particles
horizontal coordinate
vertical coordinate
axial distance along the heat transfer surface
vertical coordinate at point 1
vertical coordinate at point 2
diusivity
layer spacing
uid viscosity
uid density
particle density
Wadells sphericity, the ratio of the surface
areas of a volume equivalent sphere and the
actual particle
an equivalent sphericity or aspect ratio dened
as the ratio of the maximum length to the
diameter of the circle having the same
projected area normal to the maximum length
fd
e
eav
eb
ec
ehc
emin
eo
eo;cylinder
eo;disk
eo1;cylinder
eo1;disk
eod;cylinder
eod;disk
es
es
esc
ew
e
REFERENCES
E Abe, H Hirosue, A Yokota. Pressure drop through a
packed bed of binary mixture. J Chem Eng Japan
12:302, 1979.
AR Balakrishnan, Pei DCT. Heat transfer in gassolid
packed bed systems, a critical review. Ind Eng Chem
Process Des Dev 18:3040, 1979.
JJ Barker. Heat transfer in packed beds. Ind Eng Chem
57(4):4351, April 1965.
E Barnea, RL Mednick. A generalized approach to the uid
dynamics of particulate systems part III: general correlation for the pressure drop through xed beds of spherical
particles. Chem Eng J 15:215227, 1978.
RF Benenati, CB Brosilow. Void fraction distribution in beds
of spheres. AIChE J 8:359361, 1962.
FC Blake. The resistance of packing to uid ow. Trans Am
Inst Chem Eng 14:415, 1922.
JSM Botterill, AOO Denloye. A theoretical model of heat
transfer to a packed or quiescent uidized bed. Chem Eng
Sci 33:509515, 1978.
GG Brown. Unit Operations. New York: John Wiley, 1966.
LE Brownell, DL Katz. Flow of uids through porous media
I. Single homogeneous uids. Chem Eng Prog 43:537
548, October 1947.
SP Burke, WB Plummer. Gas ow through packed columns.
Ind Eng Chem 20:11961200, 1928.
PC Carman. Fluid ow through granular beds. Trans Inst
Chem Eng 15:150, 1937
3
Bubbling Fluidized Beds
Wen-Ching Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
Group B
Group C
Group D
For
Umb
1
Umf
8 104 gdp2 rp rf
m
1
!
dp
X xi
Umf
1
2
dpi
Umb Kdp
UB
8 104 rp rf gdp2
memf
Umf
rp rf gdp
dominant inuence of cohesion forces, and its demarcation can be expressed by the equation
10
rp rf dp3 g
K1 102
FH
2.3
10
rp rf dp g
15
2
rf Umf
are good also for gases other than air and for temperature and pressure other than ambient.
11
Group
Group
Group
Group
C boundary
A/C boundary
A/B boundary
B/D boundary
Ar
Ar
Ar
Ar
0.97
9.8
88.5
176,900
(12)
(13)
(14)
(15)
Figure 3 Bed collapsing stages during bed collapsing characterization of powders. (Adapted from Kwauk, 1992.)
16
where U1
fB 1 fw ee UB ee 1 fB Fw fB Ue
1 fB 1 ee
H He
and fB 0
H0
2.5.2
18
During the hindered settling stage, the bed surface settles at a constant velocity, and the settled layer at the
bottom, layer D in Fig. 3, also increases in height. The
change in bed height for both layers can be expressed
as
H 2 H e U2 t
H1
HD
U t
He H1 2
He H 1
1 exp K3
t
He
U 2 He
1 ee
He
He H1 K3 ee ec He H1
19
20
2.5.3
dH
K3 HD H1
dt
21
K3 HD H1
dt
dt
ys
tc
dp rp rf H1
24
22
3
or
H3 HD Hc H1 expK3 t tc H1
23
Not all powders exhibit all three stages during bed
collapsing experiments. Only Geldarts Group A
powders exhibit all three stages. For Group B and D
powders, the rst bubble escaping stage and the second
hindered settling stage are practically instantaneous.
The transition between the hindered settling stage
and the solids consolidation stage, the point tc ; Hc
in Fig. 3, is called by Yang et al. (1985) the critical
point. The better to qualify the powders through the
bed collapsing tests, Yang et al. (1985) also dened a
dimensionless subsidence time of a powder, ys .
Copyright 2003 by Taylor & Francis Group LLC
2
Umf
< 0:13
gdp
Reference
Wilhelm and Kwauk (1948)
DB max
2Ut2
< 30; DB max
dp
g
Rowe (1962)
3
Umf
p f Hmf
< 100
gmD
p f q3
gdp < C1 depending on e and emf
m
" 3
#m
gdp rf rp rf
rmf 0:5
< C2
2
rf
m
Fluidization regime
0 U < Umf
Umf U < Umb
Fixed bed
Particulate regime
Bubbling regime
Ums U < Uc
Slug ow regime
Uc U < Uk
Uk U < Utr
Transition to turbulent
uidization
Turbulent regime
U > Utr
Fast uidization
U Utr
Pneumatic conveying
xed bed to the bubbling regime is the minimum uidization velocity for Group B and Group D powders
while for Group A powders, it is the minimum bubbling velocity. The homogeneous expansion, also
called particulate uidization, occurs only in Group
A powders for gassolids systems. The particulate
uidization is especially important for liquid uidized
beds because where most of the occurrence is observed.
Liquidsolids uidization will be discussed in Chapter
26, LiquidSolids Fluidization.
3.2
3.3
d p rf U c
0:936Ar0:472
m
28
dp rf Uk
1:46Ar0:472
m
for Canada et al. (1978) data
29
dp rf Uk
1:41Ar0:56
m
for Yerushalmi et al. (1978) data
30
Rek
Rek
The extensive literature data based on absolute pressure uctuation and bed expansion measurements up
to 1989 were correlated by Cai et al. (1989) to be
Rec
dp rf Uc
0:57Ar0:46
m
31
0:941
d p rf U k
Ut
16:31Ar0:136 p
m
gD
for Ar 125
Rek
Rek
d p rf U k
U 0:0015
2:274Ar0:419 pt
m
gD
for Ar 125
Retr
35
36
dp rf Utr
2:28Ar0:419
m
37
d p rf U c
1:24Ar0:45
m
for
Retr
d p rf U k
1:95Ar0:453
m
dp rf Utr
1:41Ar0:483
m
32
33
Since the Uk and Utr are very similar, they suggest that
the following equation alone may be used to calculate
both Uk and Utr for simplication:
Rek;tr
dp rf Uk dp rf Utr
1:45Ar0:484
m
m
34
38
Perales et al. (1990) also suggested equations to calculate Uk and Utr as follows.
Rek
dDB dDB dh
UB
dt
dh dt
40
dDB
1
dh
41
or
3.4
for
42
The phenomenon of uidization can best be characterized by a P=L versus U plot such as the one shown in
Fig. 6. Below a characteristic gas velocity known as the
minimum uidization velocity, a packed bed of solid
particles remains xed, though a pressure drop across
the bed can be measured. At the minimum uidization
velocity, all the particles are essentially supported by
the gas stream. The pressure drop through the bed is
then equal to the bed weight divided by the crosssectional area of the bed, P W=A. Further
increases in gas velocity will usually not cause further
increases in pressure drop. In actual practice, however,
pressure drop at minimum uidization velocity is actually less then W=A because a small percentage of the
bed particles is supported by the wall owing to the less
43
L
cos2 g
dp2 e3
and assuming
72
180;
cos2 g
we have
P
mU
459 2mf
L
dp
45
46
gdp2 rp rf
m
Figure 6 Pressure drop vs. uidization velocity plot for determination of minimum uidization velocity.
44
47
Remf
gdp2 rp rf
m
dp Umf rf
m
Remf
48
F 3pmUdp
and
68:5 3pmUmf dp
rp rf pdp3 g
6
50
q
C12 C2 Ar C1
There are many sets of values suggested for the constants C1 and C2 based on dierent databases used for
correlation. A simplied set is shown in the table.
Reference
Wen and Yu (1966a)
Richardson (1971)
Saxena and Vogel (1977)
Babu et al. (1978)
Grace (1982)
Chitester et al. (1984)
Umf
0:00081rp
m
rf gdp2
51
Another widely employed approach is by simplication of the Ergun equation for the packed bed. By
combining the original Ergun equation as shown in
Eq. (52) and the pressure drop equation at minimum
uidization as shown in Eq. (46)
P gfdp e3
1 e
0:875
75
L 2rf U 2 1 e
fRep
52
1 emf
1
Remf 1:75 3 Re2mf
2 3
femf
f emf
53
and
1 emf
11
f2 e3mf
54
C1
C2
33.7
25.7
25.3
25.25
27.2
28.7
0.0408
0.0365
0.0571
0.0651
0.0408
0.0494
or
55
Umf 0:00075
or
rp rf gdp2
m
56
For Ar>107 ,
p
Remf 0:202 Ar or
s
rp rf gdp
Umf 0:202
rf
57
1
d sv X x
i
dpi
60
s 0:5
61
Ut
90
Umf
63
s 0:3
64
The two-phase theory of uidization was rst proposed by Toomey and Johnstone (1952). The model
assumed that the aggregative uidization consists of
two phases, i.e., the particulate (or emulsion) phase
and the bubble phase. The ow rate through the emulsion phase is equal to the ow rate for minimum uidization, and the voidage is essentially constant at emf .
Any ow in excess of that required for minimum uidization appears as bubbles in the separate bubble
phase. Mathematically, the two-phase theory can be
expressed as
GB
U Umf
A
65
58
with
K4 > 1:0
66
67
1 e 3
dB
dB
H0
( " #
)
Wds
dS
dS 2
H H0
1 1 e 3
dB
dB
H0
70
where H0 and H represent the initial and the expanded
bed height. Assuming the particles to be perfect
spheres, the optimum two-component system with
the maximum surface area has been computed to be
the one in which the diameter of the nes is 22.5% that
of the larger particles, and in which the smaller particles constitute 25% of the mixture by weight. A study
by Geldart (1972) employing the surfacevolume particle diameter seemed to conrm this as well. Some of
the data by Matheson et al. (1949) and replotted by
Geldart are presented in Fig. 8.
The importance of nes content in the uidized bed
reactors was also investigated by Grace and Sun (1990)
from a dierent aspect. They found that nes spent a
much longer time in the bubble phase and considerably
enhanced the concentration inside the bubbles due to
the through ow of uidizing gas through the bubbles.
This enhanced particlesgas contact and improved the
reactor performance.
De UB rf
23e0:004y
m
De UB rp 1 emf
Ree
73
UB
Hmf 1 VTmf Hmax VTmax
BED EXPANSION
8.1
Lmf
1 e
74
Hmax
UA
76
77
Hmf
UB
78
79
8.1.3
75
8.1.2
80
b 1:917
U Umf 0:8
g0:4
and
p
B 4 Ao
82
YU Umf
1
Hmf
Hmf
d g
83
where d 0:027U Umf
and c 0:915 U
Umf 0:4 for porous distributor plates. Y is a correction
for deviation from the two-phase theory. Quantitative
values of Y are shown in Fig. 26.
For a distributor with N holes per cm2 , c can be
calculated by
1:43 U Umf 0:4
c 0:2
84
N
g
0:94
The bubbles in a uidized bed have two basic properties. In general, they rise at a nite velocity, and they
usually grow in size owing primarily to static pressure
or to coalescence. It has been found that there is a
Bubbles in Liquids
89
87
p
G
0:35 gD
A
88
The factor 1.2 stems from the fact that the peak velocity at the middle of the tube is about 1.2 times the
average velocity, owing to the nonuniform velocity
UB
H0
U
H H0
90
91
93
94
0
@x
@y
95
@P
@x
and
Uy V y K
@P
@y
96
Figure 10
0
@x
@y
0
@x2 @y2
98
97
R
RB
3
UB 2Umf =emf
UB Umf =emf
100
The uidizing uid in this case moves downward relative to the bubble motion. The uid ows past the
ctitious sphere of radius R with velocity UB
Umf =emf at r 1. Inside the sphere of penetration
of radius R, the uid leaves the roof of the buble and
recirculates back to the base of the bubble as shown in
Fig. 11. The radius of the penetration for a 3-D bed can
be calculated from Eq. (101).
2
31=3
UB
2
6Umf =emf
7
7
RB
R6
101
4 UB
5
1
Umf =emf
In this case, the bubbles in the uidized bed are accompanied by a cloud while rising through the bed. For
fast bubbles where the bubble velocity is large, or for
uidized beds of ne powders when the minimum
uidization velocity is small, the cloud is usually very
jRj
0:8jRj
21=3
103
VB 1:138
104
Bubble formation in a uidized bed was found experimentally to be very similar to that in an inviscid liquid.
At a very low gas ow rate, the frequency and size of
the bubbles formed are primarily governed by a balance between the surface tension of the liquid and the
buoyancy force of the bubble. The inertia of the liquid
moved by the rising bubbles becomes more improtant
than the surface tension at higher gas rates. It is in this
G1:2
g0:6
106
At high gas ow rates where the bubble sizes are independent of the bed viscosity, the inviscid liquid theory
can predict the bubble sizes satisfactorily. At low gas
ow rates where the viscosity eects are quite pronounced, the inviscid liquid theory underestimates
the bubble sizes. In this case, the following equation
by Davidson and Schuler (1960) should be used.
VB 1:378
G1:2
g0:6
107
G
fn
fn bubble frequency
108
Vs 16VB0:42
109
The bubbles in uidized beds grow in size due primarily to three factors:
1. The eective hydrostatic pressure descreases
toward the top of the uidized bed.
2. Bubbles coalesce in the vertical direction with
the trailing bubble catching up the leading
bubble, and
3. Bubbles coalesce in the horizontal direction
with the neighboring bubbles.
The eect of the hydrostatic pressure is usually small,
and the bubbles grow in size owing largely to coalescence. There are a number of bubble coalescence
models in the literature. A few of the more wellknown ones are discussed here.
Geldart (1972) found that the uidization behavior
of Group B powders was independent both of the
mean particle size and of particle size distribution. In
particular, the mean bubble size was found to depend
only on the type of the distributor, the distance above
the distributor plate, and the excess gas velocity above
that required at the minimum uidization condition,
U Umf . Mathematically, it can be expressed as
DB DBo KH n U Umf m
110
113
114
117
Rowe (1976) suggested the following equation to estimate the bubble size in a uidized bed:
DB
118
12RB
If the bubbles are hemispheres, the volume of each
individual bubble can be calculated by the equation
! "
#
2R3B
pD3e
120
3
6
Figure 13
U Umf Ac
p
g
2=5
; Ac catchment area
121
122
and
h lDc
123
124
n1
X
0:62g0:25 l
5=4
D
25n=12
U Umf 0:5 eo n0
125
1:85g0:25 l
D5=4 D5=4
eo
U Umf 0:5 en
126
ment in the bed is also taken into account in developing the bubble coalescence model, the model may not
accurately predict the actual phenomena occurring
in industrial uidized beds. Bubble growth in large
uidized beds was also studied by Werther (1967a).
9.5.1
130
131
Figure 15 Patterns of bubble coalescence and solids circulation in large industrial uidized beds. (Adapted from Whitehead,
1967.)
10
SLUGGING BEDS
eS
pD3 =8
1
2
pD =43D 6
135
136
137
10.2
ating gas velocity and the bed depth. The so-called wall
slugs, which appear as half of a round-nosed slug, can
also be classied as a variation of the type A slug.
10.1
Slugging Criteria
for slugging
132
133
134
Hmf
0:35 gD
138
D
UB
where NT is the number of bed diameters between the
rear of a leading slug and the nose of a trailing one.
Combining Eqs. (138) and (139) gives the minimum or
collapsed bed height of a slugging bed:
Hmin
N D U Umf
140
1 1 T
Hmf
Hmf
UB
10.3
Figure 17
fS
142
LS
NT D
This means that the slug frequency is independent of
the operating velocity. The only unknown in Eq. (142)
is the spacing between the slugs, which has to be
determined experimentally. Values of NT from 2 up
to 8 have been observed experimentally (Kehoe and
Davidson, 1973b).
Hovmand and Davidson (1968) derived an equation
for calculating the slug length by assuming an interslug
spacing to be two times the diameter of the bed.
0:5
LS
L
0:495 S
1 B 0:061 1:94B 0
D
D
143
The equation was later modied by Kehoe and
Davidson (1973b) to read as
0:5
LS
L
1 B 0:061
0:495 S
144
D
D
TS 0:061 B 0
where
U Umf
> 0:2
UB
145
D in cm
146
D 2:51D0:2
0:13D0:47
D in cm
147
Wall Slugs
1
1
Hmf
UB1
Hmf
UB2
Hmf
D2
D2
Hmf
UB2
9
8
>
>
>
>
=
<
h
1
149
1
>
U Umf >
Hmf
>
>
;
:
1 1
UB1
For Case II where Hmf < h,
Hmax
h
U2 Umf
U2 Umf
1
Hmf
Hmf U1 Umf
UB2
h
1
1
Hmf 1 U1 Umf =UB1
150
151
For Case II where Hmf > h,
Hmax
h U2 Umf
1
Hmf
Hmf
UB1
h
U Umf
1
2
Hmf
UB2
152
11
Figure 18
region (jet penetration depth) may characterize the performance of the reactor and sometimes dictate the
reactor design. Because of the sizes of solids to be
fed continously into the beds, the feeding nozzles and
thus the jets are usually larger than 15 mm.
There is another category of jets in uidized beds
that tend to be substantially larger in size. For example, a large central jet is applied in a jetting uidized
bed, similar to a spouted bed, to induce solids mixing
and circulation, to facilitate heat and mass transfer
from one region of a uidized bed to an adjacent
region, and to promote granulation and agglomeration. This type of jetting uidized beds were used for
solids mixing and drying, coal gasication, and powder granulation. The study of large jets, substantially
larger than 15 mm in diameter, is not as common
(Yang, 1998b). The jetting uidized beds are discussed in Chapter 20, Other Nonconventional
Fluidized Beds.
An ambiguity in studying the jetting phenomenon is
the lack of consensus on what constitutes a jet (Rowe
et al., 1979; Yang and Keairns, 1979). The gas issuing
from an orice might be in the form of bubbles, a
pulsating jet (a periodic jet), or a permanent amelike
dor
25:4
dp
154
or
Um
d
3:84 o
Uj
x
156
or
Um
d
6:30 o
Uj
x
157
For small particles, like the cracking catalyst, experiments by Behie et al. (1970, 1975, 1976) indicated that
the potential core, where the properties of the jet are
essentially similar to that at the jet nozzle, is within
about half of the nozzle diameter. Donsi et al. (1980)
found that the potential core length depended on the
particle characteristics and nozzle size in a very com-
Umf p
Umf atm
161
References
rf Uj
LB
26:6
p
do
rp gdp
!0:67
rf Uj
Lmax
p
19:3
do
rp gdp
U
Ucf
!0:83
Lj
Lmin Lmax
2
0:24
U
Ucf
0:54
"
!
!#0:472
Uj2
Lmax
Ucf atm
rf
7:65
do
Ucf p rp rf gdo
"
!
!#0:835
Uj2
Lmax
Ucf atm
rf
C
do
Ucf p rp rf gdo
U Ucf
for
U > Ucf
0:0144
Lj
Lmax
1:3 0:5 logrf Uj2 ;
do
Lmin Lmax
2
0:35
Uj
0:919dp
Lmax
!0:3 2
!0:2 3
Uj2
41:3
15
gdo
!0:47
rp dp 0:585 rf do Uj 0:654 Uj2
Lj
814:2
do
r f do
m
gdo
Lj
15:0
do
"
rf
rp rf
Uj2
Lmax
1:3
do
gdp
!0:38
Uj2
gdo
!#0:187
rf dp Uj
m
0:13
Shakhova (1968)
Zenz (1968)
Merry (1975)
rf
rp
!0:56
do 0:25
dp
Table 3 Continued
Jet penetration correlations
References
0:5 q 1:5; for a single jet q 0:5; for multiple jets q 1:5
!
Lj
rf Uj
p
7:8
do
rp gdp
Shakhova (1968)
Lmax
1:48 0:5 logrf Uj2 ; rf in lb/ft3 and Uj in ft/s
do
"
#0:4
!0:2
Lj
rf Uj2
dp 0:2
rf
4:5 5:25
do
1 "rp gdp
rp
do
Zenz (1968)
0:044
Figure 19
Merry (1971)
"
#k8
Umf atm
Umf p
162
where
k1 0:30 to 0:83
k3 0 to 1:25
k2 0:3 to 0:83
k4 0 to 0:5
k5 0:4 to 0:944
k7 0:2 to 0:472
k6 0 to 0:13
k8 0 to 0:472
!
167
Figure 20
Schlitchting type and Tollmien type similarities for gas velocity proles in jets in uidized beds.
G 0:53VB fn
168
and
G 0:37VB fn
169
The inlet jet velocity for a single-phase (gas) jet is calculated based on the cross-sectional area of the jet
nozzle and the total volumetric ow rate of the jet. If
the jet also carries solids, as in the case of feeding solids
pneumatically into a uidized bed, the inlet jet velocity
should be calculated from the equation
rf Uj 2s
M g Ug Ms Us
At
170
ow rate. The solid particle velocity, Us , can be calculated assuming the gas/solid slip velocity to be the
terminal velocity of a single particle of the average size.
For a concentric jet with the inner jet carrying solid
particles, the inlet jet velocity can be calculated based
on the equation
rf Uj 2c
11.5
Mg i Ug i Ms i Us i Mg o Ug o
At
171
11.5.1
11.6
1 ej ej ez p2 e1 edp
Vj
16r
1 ez 1 ez
173
jets and found that the jets behave like two isolated jets
only at low nozzle velocities. At high nozzle velocities,
the jets interact, and the jet penetration depth becomes
constant. The jet interaction is divided into three different regions, as shown in Fig. 22.
Yang (unpublished data) also has studied the jet
interaction for two and three jets in a two-dimensional
beds. His observation of the jet interaction, shown in
Fig. 23, is very much similar to that reported by Wu
and Whiting. Yang also developed two correlations for
the stagnant emulsion regions between the adjacent jets
based on the pitch and jet diameter ratio. The height of
the stagnant emulsion regions between the jets is inversely proportional to the interaction between the jets.
The higher the jet velocity, the more intense the jet
inaction, which in turn results in lower height of the
stagnant emulsion regions between the jets, as evident
in Fig. 23. The proposed equations are
Uj
Yh
1:1660 0:02876
for two jets 174
Pj
Ut
Uj
Yh
0:4445 0:01220
Pj
Ut
Particle mixing and segregation phenomena are important in industrial uidized beds, where particles of wide
size distribution or particles of dierent densitites are
usually handled. Studies have indicated that the uidized bed reactors can be operated in dierent modes
either to promote particle mixing or to enhance the
Schematic of jet interaction of multiple jets in uidized beds. (Adapted from Wu and Whiting, 1988.)
176
177
do
rp rf gdo
12
For multiple-jet systems such as that above a gas distributor plate, Wu and Whiting (1988) suggested to
make use of the jet half angle discussed earlier to derive
an equation for jet penetration depth. Their proposed
equation is
Figure 22
rp dp 0:236 Pj do
Lm 8:79
rf d o
2
Figure 23 Schematic of jet interaction of multiple jets in uidized beds. (Adapted from Yang, unpublished data.)
Figure 24
U Umf dp rf 0:21
Dsr
0:46
U Umf Hmf
m
0:24
rp rf 0:43
hmf
dp
rf
179
3
"B
Umf DB
16 1 "B
"mf
178
Figure 25
12.1.2
where Y is a correction for deviation from the twophase theory, and the factor 0.38 for drift ux is
from the experimental evidence that the particles carried up in the drift travel on average at about 38% of
the bubble velocity. The value of wake and drift fraction, bw and bd , can be obtained from Fig. 25, while
the correction for deviation from the two-phase theory,
Y, can be found from Fig. 26, also due to Geldart
(1986). Values of bw , bd , and Y for common materials
are summarized in Table 4.
Dependence of wake and drift fraction on particle characteristics. (Adapted from Geldart, 1986.)
180
Figure 26
Correction for deviation from the two-phase theory. (Adapted from Geldart, 1986.)
1
1 U Umf
fB
181
182
and
o 7:2U Umf exp4:1U Umf
Size
mm
bw
bd
47
252
470
106
195
0.43
0.26
0.20
0.32
0.30
1.00
0.42
0.28
0.70
0.52
1.00
0.50
0.25
0.82
0.65
183
184
Hmf
YU Umf bw 0:38bd
185
W
Fo
186
For a completely mixed stirred tank reactor, the residence time distribution can be expressed following
ideal backmix reactor model.
1 t=tR
Rt
e
187
tR
Rt dt is the fraction of solids staying in the bed for the
time period between t and t dt. Yagi and Kunii
(1961) found experimentally that Eq. (187) represented
the particle residence time in a uidized bed quite well.
The fraction of solids spending less than time t can
then be calculated from
f 1 et=tR
188
dence time distribution of systems with various combinations of series and parallel reactors, the reader can
consult Levenspiel (1979).
12.2
DUbf rf
18:12 0:0192Ar0:5 18:1
m
190
DUtf rf
2
0:5
Retf
24:0 0:0546Ar 24:0
m
191
Rebf
B S
B 6 S
Case II
IIa
IIb
B > S
B < S
Case IV
IVa
IVb
B > B S
B < B S
Case III
IIIa
IIIb
12.2.3
dB =dS 10
Case I
Ia
Ib
Ro
dp F dp J
dp
R
193
R 1 xF rF dp F xF rJ dp J
Ro 1 xF rF xF rJ
194
197
198
Umf F
1 Umf F =Umf J xF Umf F =Umf J
200
0:437
dJ rF
m 0:17
d F rJ
202
d p Umf rf
0:0054Ar0:78
m
204
where
3
d p rf rp rf g
Ar
m2
rp
n
X
rpi
arithmetic mean
205
206
i1
and
d p 1=
n
X
xi
d
i1 pi
harmonic mean
207
bed is completely mixed and homogeneous, corresponding to the Umf M shown in Fig. 29.
12.2.5
XJ u
XJ
208
XF u
XF
209
Figure 30
210
1 SX F
1 S
S
1 XF
XJ
1
1 eZ
211
U UTo
eU=UTo
U Umf F
212
Umf F
Umf F
0:7
0:9rR 11:1 dER
2:2xJ 0:5 H 1:4
213
The rR in Eq. (213) is the density ratio; the dER , the
size ratio; and the H , the reduced bed height. They are
dened as
r
214
rR J
rF
dER
f J dJ
fF dF
215
H
H 1 exp
D
216
signicantly larger than those measured experimentally. Equations (211) and (213) are only recommended
for binary mixtures with volumetric jetsam concentration less than about 50%, and Eq. (213) is also only
good for systems with a particle size ratio less than 3.
Equation (213) cannot be applied to a bed with a high
aspect ratio where slugging occurs.
The equations of Nienow et al. were modied later
by Rice and Brainnovich (1986) and Peeler and Huang
(1989). A more recent study on segregation by size
dierence was conducted by Wu and Baeyens (1998).
They found that the excess gas ow rate required to
prevent segregation in a uidized bed with a wide size
distribution of powder can be calculated from the mixing index expression shown in the equation
0:75
G
217
M 1 0:0067dR 1:33 B
A
For dR 2, good mixing M 0:9 can be achieved
when the visible bubble ow GB =A is larger than
about 0.094 m/s. Since temperature has limited eect
on the visible bubble ow rate, it is expected that temperature has very little eect on the particle separation
due to dierence in size alone.
Simple quantitative equation can only be formulated for binary systems of near spherical particles.
For binary systems of granular particles, Rowe et al.
(1972) suggested the following dependence for segregation tendency.
Segregation tendency
2:5 0:2
r
dB
/ U Umf F J
rF
dS
218
12.2.9
minimum uidization velocity of the large component to promote the particle mixing. The velocity
was then decreased to a preselected value between
Umf F and Umf J . The bed was slumped after a predetermined duration of operation, and the bed was
removed layer by layer for analysis. The same procedure was repeated with longer operation duration until
the steady state concentration prole was reached. It
was found that a clean interface could usually be
obtained between the otsam-rich and the jetsam-rich
beds when the particle size ratio of the binary components was larger than 3. For systems with size ratio less
than 3, a transition region with changing composition
existed between the two segregated beds.
A more comprehensive investigation was carried
out by Yang and Keairns (1982b) employing acrylic
particles (rp 1110 kg/m3 ) as the otsam and dolomite particles rp 2610 kg/m3 ) as the jetsam. Both
types of particles have a relatively wide particle size
distribution. Four mixtures of 20, 40, 60, and 80 weight
percent of dolomite were studied at velocities ranging
from the minimum uidization velocity of the acrylic
particles to slightly higher than that of the dolomite
particles. The experiments were performed in a specially constructed bed with a main gas line leading to
the bed and a gas bypass line, both controlled by individual but electrically interlocked solenoid valves. A
known concentration of acrylicdolomite mixture
was rst passed through a Rie sampler at least four
times to mix the mixture, and the mixture was then
placed in the uidized bed. Air was turned on to a
desirable reading on the rotameter with the solenoid
valve in the bypass line open and that in the main line
closed. The needle valve in the bypass line was then
adjusted to give a pressure drop in the line equivalent
to that expected from the uidized bed. An electrical
switch interlocking both solenoid valves was then
turned on to open the solenoid valve in the main line
and to shut the solenoid valve in the bypass simultaneously. After a predetermined time, the switch was
again turned o to reverse the ow. The uidizedbed content was then vacuumed o layer by layer
and the particle concentrations analyzed by screening.
The procedure was then repeated for a dierent
separation time duration, for a dierent separation
gas velocity, or for a dierent particle mixture.
The transient particle concentration proles for a
20 w/o and 60 w/o dolomite starting mixture are presented in Figs. 32 and 33. They are shown for 3, 5, 10,
and 20 seconds of operation. When the jetsam concentration is low (e.g., 20 w/o dolomite mixture), both the
top and bottom layers have relatively uniform concentration proles throughout the transient time period at
dierent operating velocities (see Fig. 32). For the
60 w/o dolomite mixture, however, the same is true
only at lower operating velocities (Figs. 33a and
33b). At higher operating velocities, especially those
higher than the minimum uidization velocity of the
dolomite (Fig. 33d), concentration gradients start to
develop in both layers with a fuzzy transition between
them. At equilibrium, the upper uidized-bed layer
and the bottom packed-bed layer usually have uniform
particle concentrations if the operating uidizing velocities are lower than the minimum uidization velocity
of the jetsam. There is also a small transition zone
between the two layers. The jetsam in the bottom
packed bed increases with an increase in operating
velocity, while the upper uidized bed almost consists
of pure otsam.
Visual observation of the experiments in this study
indicated that the particle separation mechanisms were
dierent at high and low uidization velocities with the
minimum uidization velocity of dolomite, Umf J , as
the demarcation. At an operating velocity lower than
Umf J , there was a distinct packed bed at the bottom
in equilibrium with a uidized bed at the top with a
short transition zone in between. Mixing between these
two beds appeared to be minimum. When the operating velocity was increased to higher than Umf J , the
whole bed appeared to be uidized, although a fuzzy
particle interface was usually discernible. Both particle
mixing and particle separation were apparently occurring along the bed height. At equilibrium the particle
concentration proles in both layers were essentially
uniform. This was approximately true even when the
operating velocity was higher than Umf J . During the
transient time period, this was still true except for the
case with high jetsam concentration and high operating
velocity (Fig. 33d). It appears, then, that considerable
simplication in the mathematical model may be
possible, at least for the highly segregating system of
acrylicdolomite.
A simple mathematical model was developed by
Yang and Keairns (1982b) by assuming that the particle segregation is a uidized bed could be simulated by
two perfectly mixed uidized beds in series with particle interchange between them. The short transition
region observed experimentally was ignored in the
model. The particle exchange was accomplished by
bubble wakes from the bottom to the top uidized
bed and by bulk solids ow in the reverse direction.
The resulting equation is
Figure 32
ln
Transient particle separation proles for acrylicdolomite systems20 w/o dolomite concentration.
V1 1 CF W VJ1
o
V1 1 CF W VJ1
U Umf2
A
fw
1 "w
t
V1
where
219
U Umf2
A
fw
1 "w
V1
221
There is certainly some experimental evidence indicating that the assumption of CF W 1:0 is reasonable
at least for highly segregating systems (Yang and
Keairns, 1982b).
Equation (220) was used to t the experimental data
with excellent results. The success of the correlation
does suggest possible physical signicance of the para-
Figure 33 Transient particle separation proles for acrylicdolomite systems60 w/o dolomite concentration.
A gas uidized bed is a complex reactor even if it contains only particles of similar size, shape and, density.
The physical phenomena occurring in a gas uidized
bed depend not only on the particle characteristics, the
operating pressure, and the temperature that will
change the properties of the gas uidizing medium
but also on the physical size of the bed. New ndings
and surprises are still continuously being reported.
When the bed material consists of particles of dierent
sizes and densities, the mixing and segregation phenomena are much less understood. The equilibrium
particle concentration prole established inside a gas
uidized bed at steady state is actually a dynamic equilibrium between the competing processes of mixing
and segregation. Several mathematical models have
been proposed to evaluate both the transient and the
equilibrium particle concentration proles. However,
they are all restricted to systems of binary mixtures
only.
The models proposed so far can be broadly classied into two categories. The rst group of models is
constructed on the basis of mechanistic phenomena
observed in a bubbling gas uidized bed. Thus they
are usually similar in concept and only dierent in
details. They all divide the bed into the wake (or bubble) phase and the bulk (or emulsion) phase. The
assumptions made for the particle exchange between
the two phases and the excursion of certain phenomena distinguish each individual model. The models in
this group are due to Gibilaro and Rowe (1974),
Burgess et al. (1977), and Yoshida et al. (1980). The
D
DB
DBo
DB max
De
Dsr
=
=
=
=
=
=
12.4
e
F
=
=
fF ; fJ
FH
fn
fNF
Fo
Fr
(Fr)c
=
=
=
=
=
fS
fVF
fw
=
=
=
g
G
GB
Gj
h
H
H1 ; H2
=
=
=
=
=
=
=
H3 ; HD
Hc
He
Hmax
Hmf
Hmin
Ho
H1
I
J
K; K1 ; K2 ;
K3 ; K4
L
Lj
Lm
=
=
=
=
=
=
=
=
Lmf
LS
ls
M
= constants
= bed depth; or bed width
= jet penetration depth
= jet penetration depth for multiple-jet
systems
= bed depth at minimum uidization
= slug length
= orice separation distance or pitch
= the mixing index
Mg
Mg i
Mg o
Ms
Ms i
n
No
=
=
=
=
=
=
=
NT
P
Pj
=
=
P
q
=
=
r1=2
RB
Rcf
=
=
Rebf ; Remf , =
Retf
(Re)p
(Re)t
Rt
S
t
tb
=
=
=
tc
ti
tR
tT
U
UA
UB
Ub
Ubf
=
=
=
=
=
=
=
xF ; xJ
XF ; XJ
X F; X J
XF u ; XJ u
y
Y
Yh
z
rB ; rS
rB S
rf
rF ; rJ
rp
rp
"
"1
"b
"B
"B
"c
"e
"j
"mf
"s
"w
"z
cf
cp
f
fB
fF ; fJ
ab
b
bd
bw
l
s
m
y
ys
REFERENCES
Abramovich GN. The Theory of Turbulent Jets. Cambridge,
MA: MIT Press, 1963.
Anagbo PE. Derivation of jet cone angle from bubble theory.
Chem Eng Sci 35:14941495, 1980.
Babu SP, Shah B, Talwalkar A. Fluidization correlations for
coal gasication materialsminimum uidization velocity and uidized bed expansion ratio. AIChE Symp Ser
74(176):176186, 1978.
Baeyens J, Geldart D. Fluidization et ses Applications.
Toulouse, 1973, p 182.
Baeyens J, Geldart D. An investigation into slugging uidized beds. Chem Eng Sci 29:255265, 1974.
Baskakov AP, Malykh GA, Shihko II. Separation of materials in equipment with a uidized bed and with continuous
charging and discharging. Int Chem Eng 15(2):286289,
1975.
Basov VA, Markhevka VI, Melik-Akhnazanov TKh,
Orochke DI. Investigation of the structure of a nonuniform uidized bed. Intern Chem Eng 9:263266, 1969.
Beeckmans JM, Minh T. Separation of mixed granular solids
using the uidized counter-current cascade principle. Can
J Chem Eng 55:493496, 1977.
Beeckmans JM, Stahl B. Mixing and segregation kinetics in a
strongly segregated gas-uidized bed. Powder Technol
53:3138, 1987.
Beeckmans JM, Inculet II, Dumas G. Enhancement in segregation of a mixed powder in a uidized bed in the presence of an electrostatic eld. Powder Technol 24:267
269, 1979.
Behie LA, Bergougnou MA, Baker CGJ, Bulani W. Jet
momentum dissipation at a grid of a large gas uidized
bed. Can J Chem Eng 48:158161, 1970.
Behie LA, Bergougnou MA, Baker CGJ. Heat transfer from
a grid jet in a large uidized bed. Can J Chem Eng 53:25
30, 1975.
Behie LA, Bergougnou MA, Baker CGJ. Mass transfer from
a grid jet in a large gas uidized bed. In: Fluidization
Technology. (Keairns DL, ed.). New York: McGrawHill, 1976, Vol. 1, pp 261278.
Bi HT, Fan LS. Existence of turbulent regime in gas-solid
uidization. AIChE J 38:297301, 1992.
Bi HT, Grace JR. Flow regime diagrams for gas-solid uidization and upward transport. Int J Multiphase Flow
21:12291236, 1995a.
Bi HT, Grace JR. Eect of measurement method on velocities used to demarcate the onset of turbulent uidization.
Chem Eng J 57:261271, 1995b.
Bi HT, Grace JR, Zhu JX. Regime transitions aecting gassolids suspensions and uidized beds. Chem Eng Res Des
73:154161, 1995.
Blake TR, Wen CY, Ku CA. The correlation of jet penetration measurements in uidized beds using nondimensional
hydrodynamic parameters. AIChE Symp Series
80(234):4251, 1984.
Grace JR, Lim CT. Permanent jet formation in beds of particulate solids. Can J Chem Eng 65:160162, 1987.
Grace JR, Sun G. Fines concentration in voids in uidized
beds. Powder Technol 62:203205, 1990.
Habermann A, Winter, F. Hofbauer H, Gogolek PEG.
Residence time distribution of particles in uidized bed
reactors for metallurgical processes. In; Fluidization
(Fan LS, Knowlton TM, eds.). New York: Engineering
Foundation, 1998, pp 117124.
Harrison D, Leung LS. Bubble formation at an orice in a
uidized bed. Trans Inst Chem Eng 39:409413, 1961.
Harrison D, Davidson JF, de Kock JW. On the nature of
aggregative and particulate uidization. Trans Inst Chem
Eng 39:202211, 1961.
Hirsan I, Sishtla C, Knowlton TM. Presented at the 73rd
AIChE Annual Meeting, Chicago, IL, November 1980.
Horio M. Hydrodynamics of circulating uidizationpresent status and research needs. In: Circulating Fluidized
Bed Technology III (Basu P, Horio M, Hasatani M, eds.).
Oxford: Pergamon Press, 1990, pp 314.
Hovmand S, Davidson JF. Chemical conversion in a slugging
uidized bed. Trans Inst Chem Eng 46:190203, 1968.
Hovmand S, Davidson JF. Pilot plant and laboratory scale
uidized reactors at high gas velocities: the relevance of
slug ow. In: Fluidization. (Davidson JF, Harrison D,
eds.) New York: Academic Press, 1971, pp 193259.
Hussein FD, Paltra PP, Jackson R. Heat transfer to a uidized bed using a particulate heat transfer medium. Ind
Eng Chem Process Des Dev 20:511519, 1981.
Iya SK, Geldart D. Flow of coarse particles through a uidized bed of nes. Ind Eng Chem Process Des Dev
17:537543, 1978.
Jackson R. The mechanics of uidized beds: the motion of
fully developed bubbles. Trans Inst Chem Eng 41:2248,
1963.
Kan DD, Sen Gupta P. Relative performance of baed and
unbaed uidized bed classiers. Indian J Tech 16:507
512, 1978.
Kececioglu I, Yang WC, Keairns DL. Fate of solids fed
pneumatically through a jet into a uidized bed. AIChE
J 30:99110, 1984.
Kehoe PWK, Davidson JF. The uctuation of surface height
in freely slugging uidized beds. AIChE Symp Ser
69(128):4148, 1973a.
Kehoe PWK, Davidson JF. Pressure uctuations in slugging uidized beds. AIChE Symp Ser 69(128):3440,
1973b.
Kimura T, Matsuo H, Uemiya S, Kojima T. Measurement of
jet shape and its dynamic change in three-dimensional
jetting uidized beds. J Chem Eng Japan 27:602609,
1994.
Kimura T, Horiuchi K, Watanabe T, Matsukata M, Kojima
T. Experimental study of gas and particle behavior in the
grid zone of a jetting uidized bed cold model. Powder
Technol 82:135143, 1995.
4
Elutriation and Entrainment
Joachim Werther and Ernst-Ulrich Hartge
Technical University Hamburg-Harburg, Hamburg, Germany
INTRODUCTION
DEFINITIONS
freeboard will disappear, and there will be a zone characterized by a sharp decay of the solids concentration.
This latter zone will be called the splash zone.
Experimental observations show that above the
dense bed or the splash zone, respectively, the solids
holdup gradually decays until it becomes constant or
at least nearly constant. The distance between this
point, where the solids concentration becomes (almost)
constant, and the surface of the uidized bed is called
the Transport Disengaging Height (TDH).
3
for the particle ejection, namely the ejection of particles from the bubble roof or nose (Fig. 2a) and the
ejection of particles transported upwards in the wake
of the bubble (Fig. 2b). While there is a general agreement on the important role of the bubbles, no such
agreement exists with regard to the suitable mechanism. While Zenz and Weil (1958) and Chen and Saxena
(1978) consider the nose mechanism to be the most
important, George and Grace (1978) assumed the
wake mechanism to be the governing one. Pemberton
and Davidson (1986) postulated that particles might be
ejected by both mechanisms, depending on the particle
size, the uidizing velocity, and the bed geometry.
From their measurements they deduced that bubbles
bursting one by one eject the solids from the roof while
coalescing bubbles preferably throw the wake into the
freeboard (cf. Fig. 2c). These ndings are also conrmed by Levy et al. (1983) and by Peters et al.
(1983), who took sequences of photographs to study
the mechanism of particle projection. Pemberton and
Davidson (1983) postulated that for particles belonging to the Geldart group A uidized at high velocities,
the ejection from the roof predominates, while for
group B particles uidized with lower velocities, ejection from the wake is governing the process. A dierence between the two mechanisms lies in the amount of
solids thrown into the freeboard and their size distribution. While the solids thrown into the freeboard
from the bubble wake have nearly the same size distribution as the bed material, the solids thrown into the
freeboard from the bubble roof are much ner. The
bubble-roof model gives a much lower solids ux at
the bed surface than the wake model, the dierence
between the uxes being about one order of magnitude. Another dierence between the two mechanisms
is the dispersion of the solids ejected to the freeboard:
while the solids from the wake are thrown into the
freeboard as quite closely packed bulks of solids, the
1 d
xBi t
MB t
A
dt
xBi t
Eih
xBi
EXPERIMENTAL INVESTIGATIONS OF
ENTRAINMENT
kih
For batch experiments, Kih
has to be calculated from
the entrained mass mi;t of size dpi accumulated during
the time t. Therefore Eq. (1) has to be integrated.
Provided the total mass of bed material does not
change signicantly, this gives
K
A
mi;t xBiO
MB 1 exp ih
4
MB
xi;E
Eih xBi
Kih
Eh
Eh
group A powders are concerned there are some indications that there is a critical particle size, below which
the elutriation rate constant Ki1
levels o or even
decreases with decreasing size. Figure 4 shows experimental results measured by Smolders and Baeyens
(1997) with limestone and a-alumina. The experimen
are plotted against the
tally determined values for Ki1
particle size. In spite of some scatter, the general
decrease in the elutriation rate constants below about
10 mm is obvious. Similar results were also presented
by Tasirin and Geldart (1998b), Santana et al. (1999),
and Tasirin et al. (2001).
The same trend was also observed for uidized beds
with binary particle mixturesnes in a coarse particle
uidized bedby Ma and Kato (1998). The reason for
this eect is probably the formation of particle agglomerates in the case of very ne particles for which adhesion forces are large compared to the gravitational
forces. Baeyens et al. (1992) suggest based on their
measurements a correlation for the critical particle diameter dp;crit , at which the inection occurs:
10325
dp;crit
r0:725
s
C
4:5
rs
dp;crit
g
Figure 4 Elutriation rate constant Ki1
versus the particle size for nes and supernes for (a) limestone and (b) a-alumina. (Data
from Smolders and Baeyens, 1997.)
Close to the bed surface the carryover from the uidized bed decreases signicantly with increasing height.
The decay in the entrainment becomes nearly constant,
when the transport disengaging height (TDH) is
exceeded (Fig. 6). Like the entrainment ux, the solids
volume concentration in the freeboard decreases with
increasing distance from the bed surface and becomes
nearly constant above the TDH.
According to measurements by Nakagawa et al.
(1994), the shape of the gas outlet has no signicant
inuence on the solids holdup in the freeboard. Figure
7 shows their measurements in a 0:15 0:15 m2 unit
Figure 5 Inuence of the supercial gas velocity on the elutriation rate, Dt 0:1 m, air at ambient conditions. (Data
measured by Choi et al., 1998.)
5.3.2
Effect of Diameter
While the above cited ndings are all for units with
constant diameter, Briens et al. (1990), Smolders and
Baeyens (1997), and Tasirin and Geldart (1998a)
studied the eect of an enlarged freeboard, i.e., of a
freeboard diameter larger than the uidized bed diameter. In this case the gas velocity is reduced by the
square of the ratio of bed diameter Dbed to freeboard
diameter DFB , which leads to a signicant reduction of
the entrainment ux by increasing the freeboard diameter. Smolders and Baeyens (1997) found
E1 / Dbed =DFB 4 . Tasirin and Geldart (1998a)
found that the elutriation rate constant Ki1
is only
determined by the gas velocity of the expanded section,
provided the expanded section is suciently high. If
there is not sucient length in the extended diameter
section to form a uniform gas velocity prole across
the whole cross-sectional area, the entrainment ux
will be higher than expected for the averaged velocity
in the expanded diameter section.
5.4
Taking into account the role of bubbles for the transport of particles into the freeboard, an inuence of bed
height and also of internals in the uidized bed could
be expected. Baron et al. (1990) studied the inuence of
the bed height, using a uidized bed column of 0.61 m
diameter and silica sand with a surface mean diameter
dp of 65 mm. Their results for the entrainment ux E1
above the TDH and the entrainment ux E0 just above
the bed surface are shown in Figs. 8a and 8b, respectively. The results show a slight increase of the entrainment uxes with increasing bed height, the inuence
being more pronounced for the entrainment ux E0
at the bed surface and higher gas velocities. The reason
for the increase of the entrainment ux is the increase
of bubble sizes with height, which obviously dominates
the decrease in bubble frequency and therefore in the
probability of bubbles to coalesce near the bed surface.
This indicates, that at least for the system studied by
Baron et al., the bubble nose mechanism is the dominating mechanism for particle ejection into the freeboard. Choi et al. (1989), however, found no inuence
of bed height on entrainment during their study of
entrainment from uidized bed combustors. In this
study they used two uidized bed combustors with
cross-sectional areas of 0:3 0:3 m2 and 1:01
0:83 m2 , respectively. The smaller one was equipped
with an extended freeboard of 0:45 0:45 m2 . In
fact, the bed height eect in Fig. 8 is small compared,
e.g., to the eect of the gas velocity.
The picture of the inuence of internals in the freeboard is not very clear. On the one hand, they may
act as a kind of gassolid separators, thus helping to
reduce the TDH (Martini et al., 1976) and the entrainment ux (Baron et al., 1988). On the other hand, they
may increase the entrainment due to the increased gas
velocity in the freeboard (Tweedle et al., 1970), or they
may not aect the entrainment at all (George and
Grace, 1981). Specially designed baes and inserts in
the freeboard may help to reduce the entrainment ow
signicantly (Harrison et al., 1974).
5.6
MODELING
Considering the complex uid mechanics in the freeboard and at the bed surface of the bubbling bed, it is
not astonishing that up to now no rigorous model that
Estimation of TDH
Table 1 Correlations for the Elutriation Constant Ki1
2
1 Ki1
g
dpi
mU uti
1:2
0:0015
Re0:6
t 0:01
Ret
8
!1:88
>
U2
>
7
>
>
1:26
10
>
<
g dpi r2p
Ki1
!1:18
rg
U >
>
U2
>
4
>
>
: 1:31
10
g d r2
pi p
Ki1
U Uti 2
1:52
105
rg U Uti
g dpi
Ki1
U Uti 2
4:6
102
rg U Uti
g dpi
for
U2
< 3:10
g dpi r2p
for
U2
> 3:10
g dpi r2p
!0:5
Re0:725
i
!0:5
Re0:3
t
rs rg
rg
"
0:25 #
u 0:5 U
Ki1
ti
mf
A 130
exp 10:4
rg
U
U
U Umf
!0:15
dp;bed
mm
U, m/s
D, m
Reference
0.31.0
0.071.0
0.11.6
0.30.7
0:05 0:53
0.040.2
0.61.0
0.102
0.7
Wen and
Hashinger (1960)
0.92.8
0.0310.067
0.71.9
Tanaka et al.
(1972)
0.62.4
0:91 0:91
0.061.0
Merrick and
Highley (1974)
0.63.0
0.076
0.3
0.060.35
1.5
Geldart et al.
(1979)
0.10.3
0:61 0:61
00.125
Ki1
u
23:7
exp 5:4 ti
rg
U
U
!
2 1:65
7 K
i1
4 U
9:43
10
rg
U
g dp
U2
for 58
g dpi
!
1000; 0:1m=s U 0:3 m=s;
0 < dp 74 mm
Comments
Ki1
Ki1
kg
u 2
0:011
rs 1 ti
2
U
m s
0.93.7
0:92 0:92
0.31.0
0:3 0:3
2:1
1:6
rs rg 0:54
Ut
kg
2 U uti
Dh
2:8
10
U
rs
m2 s
U
u
10 Ki1
1 ti
1:6
uti
rg
U
U
11
Ki1
3:4
kg
U
u 2
5
r
1 ti
5:4
10
s
2
0:2
U
m s
for dpi
12
Ki1
Sciazko et al
(1991)
0.20.7
0.030.78
Baeyens et al
(1992)
10325
r0:725
s
kg
0:35rs U1 eH
m2 s
0.10.6
0.071
0:08 0:08
0:15 0:15
0.030.2
Nakagawa et al.
(1994)
0.20.8
0.076, 0.152
0.0170.077
Tasirin and
Geldart (1998c)
0.37.0
0.061.0
0.051.0
with
0:64
1 eH 7:41
103 R1:87 A0:55
HFB
t
and R
13
Ki1
X U Uti
for uti < U
xi
Uti
8
< 23:7
r
U 2:5 exp 5:4 uti
for Re < 3000
g
kg
U
u
2
: 14:5
r
U 2:5 exp 5:4 ti
m s
for Re > 3000
g
U
with Re
Dt
U
n
13:1
14 Ki1 dp
0:5
0:303
0:902
Ar exp 6:92 2:11Fg
m
Fd
with Fg g dp rs rg gravity force per projection area)
Fd Cd
rg U 2
(drag force per projection area)
2
Correlation based on a
wide range of dierent
units, materials, and
operating conditions,
e.g., temperature and
pressure
Figure 11 Comparison of predicted elutriation rate constants with measurements by Colakyan and Levenspiel
(1984), Choi et al. (1998), and George and Grace (1981).
Sand, dp 90 mm, for correlation 13 D 0.6 m was
assumed, and the terminal velocity was calculated with the
correlation for the drag coecient by Kaskas and Brauer as
cited by Brauer (1971). The numbers denote the various correlations in Table 1.
Figure 12 Comparison of dierent correlations for the elutriation rate constantinuence of particle size. Sand, U
1 m/s, for correlation 13 D 0.6 m was assumed, the terminal
velocity was calculated with the correlation for the drag coefcient by Kaskas and Brauer as cited by Brauer (1971). The
numbers denote the various correlations in Table 1.
Reference
1:2
or
Eih xBi
Ki1
Ei0
expai h
Figure 13 Transport disengaging height vs. excess gas velocity U Umf for particles of the FCC catalyst type. The
parameter is the bed diameter. The freeboard is assumed to
have the same diameter as the bed. (After Zenz and Othmer,
1960.
12
Table 3 Correlations for the Calculation of the Entrainment Flux E0 at the Bed Surface
Equation
1
E0 3dp
Reference
1 "mf U Umf
bubble nose model
db
E0 0:1
rs 1 "mf U Umf bubble wake model
0:5
A
db U Umf 2:5 r3:5
g
g
2:5
m
E0 3:07
109
3:5
298
T in kelvins
E0 9:6
A
U Umf A
db
2:5
11
NOMENCLATURE
a
A
db
dp
D
DFB
Dh
E1
Greek symbols
rg
rs
m
REFERENCES
Bachovchin DM, Beer JM, Sarom AF. An investigation
into the steady-state elutriation of nes from a uidized
bed. AIChE Symp. Ser. 77:7685, 1981.
Baeyens J, Geldart D, Wu SY. Elutriation of nes from gas
uidized bed of Geldart A-type powderseect of adding
supernes. Powder Technology 71:7180, 1992.
Baron T, Briens CL, Bergougnou MA. Reduction of particle
entrainment from gas-uidized beds with a screen of oating balls. AIChE Symp. Ser. 84:5057, 1988.
Baron T, Briens CL, Galtier P, Bergougnou MA. Eect of
bed height on particle entrainment from gas-uidized
beds. Powder Technology 63:149156, 1990.
Brauer H. Grundlagen der Einphasen- und Mehrphasenstromungen. Frankfurt a. Main: Verlag Sauerlander,
1971, pp 196200.
Briens CL, Bergougnou MA. A new model for entrainment
from uidized beds. AIChE Journal 32:233238, 1986.
Briens CL, Bergougnou MA, Barton T. Reduction of particle
entrainment from gas-uidized beds. Prediction of the
eect of disengagement zones. Powder Technology 62:
135138, 1990.
5
Effect of Temperature and Pressure
J. G. Yates
University College London, London, United Kingdom
INTRODUCTION
Practically all industrial gassolid uidized bed reactors operate at temperatures well above ambient, and
some, such as those used in the production of polyolens, also operate at elevated pressures. It is therefore
important to know how uidized beds behave under
high temperatures and/or pressures and if possible to
predict this behavior from observations made under
ambient conditions. The emphasis here will be on
those aspects of the subject that are of direct relevance
to the design and operation of uidized bed plant; a
comprehensive review of the more academic aspects
can be found in the review by Yates (1996).
As is made clear in other parts of this publication,
solid particles in uidized beds are held in suspension
by the upward ow of gas. The velocity at which the
particles rst become suspended is the minimum uidization velocity, umf , and as the gas velocity through
the bed is increased, the particles pass through a number of ow regimes characterized as bubbling, turbulent, and fast. The transition velocities between one
regime and another, as well as the terminal fall velocity
of single particles, ut , are all inuenced by the physical
properties of the gas, and these in turn are inuenced
by the temperature and pressure under which the bed is
operated. Changes in the surface properties of the uidized solids, particularly with increasing temperature,
can also have an important inuence on bed behavior.
These and other factors will be considered in detail
below.
The review will start by considering the two velocities at the extremes of the ow regimes, i.e. umf and ut .
3
2
fdp
L
e3
e
fdp
where
Ga
dp3 rf rp rf g
;
m2
Remf
dp umf rf
m
1
14
fe3mf
Ga 1650 Remf
pdp2 u2 pdp3
rf
rp rf g
4
2
6
4 d p r p rf g
CD
3
rf u2
and
3
Ga Re2 CD
4
10
CD
Re C1 2 C12
0:75C2 Re2
11
e3
1 1 e1=3
13
i1=3 9
1=3
>
1=3 1=3 >
<
=
1 1 emf
Ree
Re
4=3
CD e mf
CD mf >
:e exp51 emf =9emf >
;
mf
14
Yang et al. (1985) combined Eq. (14) with the standard
drag correlations recommended by Clift et al. (1978)
(Table 1) and found good agreement with experimental
values obtained at pressures of up to 63 bar and temperatures of up to 900 C.
A qualitative appreciation of the way umf is aected
by elevated pressures was given by Rowe (1984), who
rearranged Eq. (3) to express umf in terms of the
operating variables as:
m
42:91 emf
umf
r f dp
8"
#1=2 9
<
=
3
e
mf
Ga
1
1 3:0 104
:
;
1 emf 2
15
Rowe (1984) used Eq. (15) to show how pressure
would be expected to aect umf for particles with a
density of 1250 kg=m3 and a range of sizes (Fig. 1).
In these calculations the minimum uidization voidage
was set at 0.5 and the shape factor at unity; it is clear
that for particles below about 100 mm in diameter
0:01 < Re 20
Correlation
CD 3=16 24=Re
log10
CD Re
1 0:881 0:82w 0:05w2
24
i
24 h
1 0:1315 Re0:820:05w
Re
CD Re
log10
1 0:7133 0:6305w
24
i.e., CD
i.e., CD
(D) 260 Re 1500
24
1 0:1935 Re0:6305
Re
(E)
(F)
eT eamb kT Tamb
16
17
du
dt
pdp3
pdp2 u2
rp rf g CD
r
6
4 f 2
18
At ut
du
0
dt
Hence
"
#1=2
4dp rp rf g
ut
3rf CD
19
ut
18
2:335 1:744fs
2
dp
dp 0:5
#1
21
rp =kg m3
Group
0.0001
0.0005
0.001
1500
1500
2500
0.8
0.8
0.8
A
B
D
dp2 rp rf g
18mf
22
BUBBLING BEDS
4.1
Pressure
Group A Powders
Figure 3 Eect of temperature and pressure on terminal fall velocity of Group A particles in air.
Figure 4 Eect of temperature and pressure on terminal fall velocity in air of particles in Groups A, B, and D.
Group B Powders
Reference
Bed cross
Section, cm
Particle
properties
material
rp ; gcm3
dp ; mm
Gas
Max.
pressure,
kPa
diam: 10:0
(i) diakon
(ii) phenolic
resin
1.18
0.24
125
186
air
1414
diam: 11:4
(i) copper
(ii) shot
(iii) sand
coal
150, 625
160, 800,
2740
1430
air CO2
1000
Knowlton
(1977)
diam: 29:2
(i) coals
1.12
cokes
1.57
(ii) and chars
siderite
3.91
(FeCO3 )
230780
N2
6900
(i) dolomite
(ii) sulphated
dolomite
2.46
3.19
765
717
air
834
290
diam: 15:2
30.5 square
glass
beads
2.48
2.92
650
2600
air
R-12
9000
diam: 2:7
(i) synclyst
(ii) coals
0.9
1.3
63
1963
Ar
CF4
6900
diam: 11:4
(i) copper
shot
(ii) sand
8750
air,
argon
Co2 and
freon
1000
(iii) soda
glass
2450
160, 340,
620
160, 570,
1020, 2370
415
64
211
74
N2
2800
diam: 10
2600
(i) ballotini
2.9
(ii) sand
2.69
CO2
Table 3 Continued
Bed cross
Section, cm
Particle
properties
material
Subzwari et al.
(1978)
2-D bed
46 15
FCC
Powder
0.95
60
air
700
2-D bed
31 1:6
sand
2.65
250295
air
2200
Borodulya et al.
(1980)
diam. 10 cm
sand
2.48
2.60
1261220
air
8000
diam. 10 cm
ballotini
2.9
64
101
475
air
N2
CO2
2500
semicircle
diam.
30.5 cm
2.63
1.16
3.99
438
419
421
Nz
5171
diam.=
20 cm
silicon
carbide
3.19
58
air
400
Xavier et al.
diam. =
10 cm
ballotini
ballotini
ballotini
polymer
2.82
2.97
2.91
0.55
61
475
615
688
CO2
N2
N2
N2
Kawabata et al.
(1981)
2-D bed
30 1
sand
2.63
2.58
2.59
300
430
600
diam.=5
ballotini
2.9
64
101
Reference
(i) sand
(ii) char
(iii) siderite
(FeCO3 )
(i)
(ii)
(iii)
(iv)
rp ; gcm3
dp ; mm
Gas
Max.
pressure,
kPa
2500
air
800
N2
500
diam.=10
(i) ballotini
(ii) sand
(iii) polymer
2.82
2.97
2.66
0.55
61475
N2
CO2
2500
81, 288
688
Rowe et al.
(1982)
diam.=7.6
FCC powder
0.82
70, 82
Ar
air
Co2
2000
diam.=4.5
(i) ballotini
2.3
N2
3500
(ii) alumina
2.8
(iii) pyrrhotite
4.0
125, 177
250
88, 125
177
125
Barreto et al.
(1983)
diam.=10
FCC powders
1.26
0.88
5898
N2
2000
Borodulya et al.
(1983)
diam.=10.5
(i) glass
beads
(ii) sand
2630
2630
2580
2700
3100
1250
1225
794
air
8100
Rowe et al.
(1984)
17:5 12:5
(i) alumina
(ii) silicon
carbide
1.42
3.19
450
262
N2
8000
diam.=11.5
sand
2.60
20300
N2
3100
Verkooijen et al.
(1984)
diam.=10
61
butene
in N2
1000
FCC powder
Table 3 Continued
Reference
Chitester et al.
(1984)
Bed cross
Section, cm
Particle
properties
material
rp ; gcm3
dp ; mm
diam.=10.2
(i) coals
(ii) chars
(iii) ballotini
1.25
1.12
2.47
88361
157376
96374
10:2 1:9
(i) coal
(ii) char
1.25
1.12
195
203
Gas
Max.
pressure,
kPa
N2
6485
13
carbon
carbon
carbon
0.850
0.850
0.850
66
108
171
CO H2
8500
17:8 12:7
alumina
alumina
silicon
carbide
1.417
1.488
3.186
450
695
184
N2
8000
Chan et al.
(1987)
38.1
sand
sand
sand
coke
breeze
char
2.565
2.565
2.565
1.507
100
200
400
400
N2
3200
1.251
400
carbon
0.85
44
122
CO H2
silica
silica
silica
silica
FCC
FCC
resin
sand
0.834
0.706
0.711
0.844
1.172
1.667
1.330
2.580
476
280
165
1057
65
53
566
98
air
9.7
28.4
gel
gel
gel
gel
12,420
800
Carsky et al.
(1990)
13.5
sand
re clay
2.65
2.00
600
600
air
1300
20 30
sand
2.60
700
air
1600
Olsson et al.
(1995)
20 30
sand
2.60
700
air
1600
Wiman et al.
(1995)
20 30
sand
2.60
700
air
1600
Llop et al.
(1996)
sand
2.65
213, 450
728, 1460
air
101
Wiman et al.
(1997)
20 30
sand
2.60
450, 700
air
1600
Wiman et al.
(1998)
20 30
sand
2.60
450
air
1600
Group D Powders
Hfl Hmf
Hfl
S
W
W Nh d
0:89
28
and
Hfl
T
Hfl Nv d
0:27
29
JET PENETRATION
24
Re
1 emf
150
1:75fRe
emf
emf f2 Ar
25
26
27
dor
25:4
dp
30
where dor and dp are the diameters of orice and particle, respectively. Under all other conditions, bubbles
rather than jets would be expected to form.
A considerable number of studies have been
reported on jet penetration, and several correlations
have been developed giving the penetration length as
a function of the physical properties of gas and particles and of operating conditions (Knowlton and
Hirsan, 1980). However, most of these correlations
were produced from experimental data obtained at
ambient conditions, and they fail when applied at elevated temperatures and pressures. Hirsan et al. (1980)
measured jet penetrations in beds of Group B powders
at pressures of up to 50 bar and produced a correlation
for Lmax (Fig. 5) in terms of the Froude number and
the ratio of uid to particle density:
!0:67
!0:34
Lmax
rf
u2o
u 0:24
26:6
31
d0
rp
gdp
ucf
where uo is the orice gas velocity and ucf is the
supercial velocity necessary to completely uidize
the polydispersed powder. The correlation shows that
the penetration length increases with pressure but
decreases as the velocity of the uidizing gas increases,
results also obtained by Yates et al. (1986). Yang
(1981) used the same data as Hirsan et al. (1980) and
produced a slightly dierent correlation:
"
!
!#0:472
Lmax
1
rf
u2o
7:65
32
d0
Rcf rp rf gd0
where
Rcf
ucf pressure
ucf atmosphere
33
This would appear to be the only correlation currently available which draws together results from
measurements at both elevated temperatures and
pressures.
K f20
mf
rf
rf
dp
gdp
35
where Df is the diameter of the uidized bed, the subscript 20 refers to physical properties measured at 20 C
and 1 bar, and
0:211 2:42 103
K
D1:27
D0:27
f
f
!1=0:27
36
1 d
x M i xi
At dt i
37
Figure 6 Eect of pressure and uidizing gas velocity on solids entrainment based on the data of Chan and Knowlton (1984).
(From Kunii and Levenspiel, 1991.)
HEAT TRANSFER
39
40
41
44
45
46
and
hf
mkg
dp
47
48
Effect of Pressure
49
Effect of Temperature
Increasing temperature has two eects (1) by decreasing gas density, the gas convective component of heat
transfer is decreased slightly, and (2) by increasing the
thermal conductivity of the gas, the eectivness of
packets of emulsion phase in contact with the transfer
surface is increased. The overall eect for Group A and
B powders is to increase the convective transfer coecient as was shown by Botterill and Teoman (1980). In
the case of Group D powders where the gas convective
50
1
1
hc
hf he
1
ddp
kg
r1
pt
kc cps rs 1 ec
51
7.3
p1
p3
Cg m
Pr
kg
p4
Cp m
kg
p5
kp
kg
53
hdp
Nu
kg
s
rp cp
u umf
p7 3
kg g
p6
54
kg=m3
Susp
Dens
15
60
W=m2 K
83
56
45
83
56
45
13
13
14
13
13
14
19
19
19
76
76
76
68
48
22
68
48
22
100
80
55
157
137
112
106
89
62
166
144
110
56
57
With certain uidized bed materials, as the temperature of the bed is increased, a point is reached at which
the particles begin to sinter by a softening of the surface and the formation of interparticle bonds. This
temperature, which is often lower than the fusion temperature of the bulk material, is called the minimum
sintering temperature, Ts , and beds operated at temperatures higher than this can suer catastrophic
deuidization through large-scale agglomeration of
particles. On the other hand, increasing the uidizing
velocity at bed temperatures in excess of Ts can prevent
deuidization, as can increasing the initial size of the
particles. Whether beds deuidize would seem to
depend on a balance between the cohesiveness or
stickiness of the particle surfaces and the kinetic
4Gv
2
pdag
58
2Mu0
3pma2
60
tb0
Hmf
u umf
62
K1 K2
m0 expE=RT
63
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
cross-sectional area, m2
Archimedes number
heat capacity, J kg1 K1
drag coecient
bed diameter, m
gas diusivity, m2 Pa s1
diameter, m
diameter of equivalent sphere, m
entrainment ux, kg m2 s1
modulus of elasticity, N m2
emissivity
coecient of restitution
Galileo number
force [Eq. (58)], N
acceleration due to gravity, m s2
heat transfer coecient, W m2 K1
binder thickness, m
length of surface asperities, m
bubble velocity coecient
thermal conductivity, W m1 K1
chemical rate coecient, m s1
height, m
mass, kg
m
Nu
Pr
p
Re
Sh
St
T
tc
U; u0
Uc
Uk
Us
Utr
u
ut
W
X
x
z
zi
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Greek Symbols
a; g
d
e
f
i
m
r
s
t
=
=
=
=
=
=
=
=
=
stoichiometric coecients
bed expansion ratio
voidage
sphericity
elutriation rate coecient, kg m2 s1
viscocity, N s m2
density, kg m3
yield strength, N m2
residence time, s
Subscripts
b
c
cf
D
f
g
gc
max
mb
m
o, or
p
pc
r
s
t
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
bubble
cluster
complete uidization
dense phase
uid
uidized
gas
gas convective
maximum
minimum bubbling
minimum uidization
orice
particle
particle convective
radiant
solids
terminal fall
REFERENCES
Barnea E, Mizrahi J. A generalised approach to the uid
dynamics of particulate systems: general correlation of
uidization and sedimentation in solid multiparticle systems. Chem Eng J 5:171, 1973.
Barnea E, Mednick RL. Correlation of minimum uidization
velocity. Trans I Chem E 53:278, 1975.
Barreto GF, Yates JG, Rowe PN. The eect of pressure on
the ow of gas in uidized beds of ne particles. Chem
Eng Sci 38:19351945, 1983.
Baskakov AP. The mechanism of heat transfer between a
uidized bed and a surface. Int Chem Eng 4:320, 1964.
Borodulya VA, Ganzha VG, Podberezsky AI. Heat transfer
in a uidized bed at high pressure. In: Fluidization (JR
Grace, JM Matsen, eds.). Plenum Press, New York, 1980,
201207.
Borodulya VA, Ganzha VL, Podberezsky AI, Upadhyay SN,
Saxena SE. High pressure heat transfer investigations for
uidized beds of large particles and immersed vertical
tube bundles, Int J of Heat and Mass Transfer 26:1577
584, 1983.
Botterill JSM. Fluid bed heat transfer. In: Gas Fluidization
Technology (D Geldart, ed.). John Wiley, Chichester,
1986.
Botterill JSM, Denloye AOO. Bed to surface heat transfer in
a uidized bed of large particles. Powder Tech 19:197
203, 1978a.
Botterill JSM, Denloye AOO. A theoretical model of heat
transfer to a packed or quiescent uidized bed. Chem Eng
Sci 33:509515, 1978b.
Botterill JSM, Desai M. Limiting factors in gas-uidized bed
heat transfer. Powder Tech 6:231238, 1972.
Botterill JSM, Teoman Y. Fluid-bed behaviour at elevated
temperatures. Fluidization 5 (JR Grace, JM Matsen,
eds.). Plenum Press, New York, 1980, 93100.
Botterill JSM, Teoman Y, Yuregir KR. Temperature eects
on the heat transfer behaviour of uidized beds. AIChE
Symp Ser 77:(208)330340, 1981.
Botterill JSM, Teoman Y, Yuregir KR. The eect of operating temperature on the velocity of minimum uidization
bed voidage and general behaviour. Powder Technology
31:101110, 1982.
Cai P, Chen SP, Jin Y, Yu ZQ, Wang ZW. Eect of temperature and pressure on the transition from bubbling to turbulent uidization. AIChE Symp Ser 85:(270)3743, 1989.
Campos JBLM, Guedes de Carvalho JRF. Drag force on the
particles at the upstream end of a packed bed and the
stability of the roof of bubbles in uidized beds. Chem
Eng Sci 47:40574062, 1992.
Canada GS, MacLaughlin MH. Large particle uidization
and heat transfer at high pressure. AIChE Symp Ser
74:(176)2737, 1978.
Carsky M, Hartman M, Ilyenko BK, Makhorin KE. The
bubble frequency in a uidized bed at elevated pressure.
Powder Tech 61:251254, 1990.
Chan IH, Knowlton TM. The eect of pressure on entrainment from bubbling gas uidized beds. In: Fluidization
(D Kunii, R Toei, eds.). Engineering Foundation, New
York, 1984, pp 283290.
Chan IH, Shishtla C, Knowlton TM. The eect of pressure
on bubble parameters in gas uidized beds. Powder Tech
53:217235, 1987.
Chitester DC, Kornosky RM, Fan LS, Danko JP.
Characteristics of uidization at high pressure. Chem
Eng Sci 39:253261, 1984.
Clift R, Grace JR, Weber ME. Stability of bubbles in uidized beds. I & E C Funds 13:4551, 1974.
Clift R, Grace JR, Weber ME. Bubbles, drops and particles,
Academic Press, New York, 1978.
Compo P, Pfeer R, Tardos GI. Minimum sintering temperatures and deuidization characteristics of uidizable
particles. Powder Tech 51:85101, 1987.
Compo P, Tardos GI, Mazzone D, Pfeer R. Minimum sintering temperatures of uidizable particles. Part Charact
1:171177, 1984.
Crowther ME, Whitehead JC. Fluidization of ne particles at
elevated pressure. In: Fluidization (JF Davidson, DL
Kearins, eds.). Cambridge University Press, 1978, pp
6570.
Davidson JF, Harrison D. Fluidised Particles, Cambridge
University Press, 1963.
Denloye AOO, Botterill JSM. Bed to surface heat transfer in
a uidized bed of large particles. Powder Tech 19:197
203, 1978.
Ennis BJ, Tardos GI, Pfeer R. A microlevel-based characterization of grannulation phenomena. Powder Tech
65:257272, 1991.
Formisani B, Girimonte R, Mancuso L. Analysis of the uidization process of particle beds at high temperature.
Chem Eng Sci 53:951962, 1998.
Findlay JG, Knowlton TM. Final Report for US Dept. of
Energy, Project DE-AC21-83MC20314, 1985.
Fletcher JV, Deo MD, Hanson FV. Fluidization of a multisized Group B sand at reduced pressure. Powder Tech
76:141147, 1993.
Foscolo PU, Gibilaro LG. A fully predictive criterion for the
transition between particulate and aggregate uidization.
Chem Eng Sci 39:16671675, 1984.
Geldart D. Types of gas uidization. Powder Tech 7:285
292, 1973.
Geldart D, Cullinan J, Georghiades S, Gilvray D, Pope DJ.
The eect of nes on entrainment from gas uidized beds.
Trans I Chem E 57:269275, 1979.
Geldart D, Kapoor DS. Bubble sizes in a uidized bed at
elevated temperatures. Chem Eng Sci 31:842843, 1976.
Glicksman LR. Circulating uidized bed heat transfer. In:
Circulating Fluidized Bed Technology 2 (P Basu, JF
Large eds.). Pergamon Press, Oxford, 1988.
Glicksman LR. In: JR Grace, AA Avidan, TW Knowlton,
eds. Circulating Fluidized Beds. Blackie, London, 1997,
pp 261311.
6
Gas Distributor and Plenum Design in Fluidized Beds
S. B. Reddy Karri
Particulate Solid Research, Inc., Chicago, Illinois, U.S.A.
Joachim Werther
Technical University Hamburg-Harburg, Hamburg, Germany
INTRODUCTION
2.2
2.1
TYPES OF GRIDS
Possible Disadvantages
Bed weepage to plenum; can be subject to buckling or
thermal distortion; requires peripheral seal to vessel
shell; requires support over long spans; high pressure
drop required if weepage during operation is to be
minimized
Bubble Caps and Nozzles (Laterally Directed
Flow)
Main Advantages
Simple fabrication; most common; inexpensive; easy to
modify hole size; easy to scale up or down; easy to
edge of the cap. The separation distance sbc is responsible for the sealing eect of the bubble cap.
2.3 Sparger (Laterally or Downwardly Directed
Flow)
Main Advantages
Possible Disadvantages
Expensive; dicult to avoid stagnant regions; more
subject to immediate bubble merger; dicult to
clean; dicult to modify; not advisable for sticky
solids; requires peripheral seal; ports not easily
shrouded.
Details of some nozzles that are currently used in
circulating uidized beds (CFB) combustors are shown
in Fig. 1 (VGB, 1994). There are signicant dierences
between bubble caps (No. 7 in Fig. 1) and nozzles (No.
1 in Fig. 1) with respect to the prevention of solids
back ow: in the case of nozzles, the high velocity of
the gas jet prevents the solids from owing back into
the wind box. On the other hand, in the case of the
bubble cap design, the gas owing out of the bubble
cap into the bed has a rather low velocity. In this case,
the backow of solids is avoided by letting the gas ow
downward from the holes in the inner tube to the lower
Can minimize weeping; good turndown ratio; low pressure drop; can support internals; can undergo thermal
expansion without damage; ports are easily shrouded;
well suited to multilevel uid injection; solids can ow
from above the grid to below.
Possible Disadvantages
Deuidized solids beneath the grid; can be a less
forgiving mechanical design.
Figure 1 Distributors and nozzles used in large circulating uidized bed combustors. (After VGB, 1994.)
2.4
Main Advantages
Promotes solid mixing; prevents stagnant solids
buildup; minimizes solids segregation. Facilitates the
easy discharge of solids.
Possible Disadvantages
Dicult to construct; requires careful design to ensure
good gas distribution; requires high pressure drop for
good gas distribution.
2.5
Possible Disadvantages
Dicult to construct, facilitates only small hole sizes,
requires reinforcement underneath the sheet to support
the bed.
Among the foregoing advantages and limitations,
the designer must select those most pertinent or critical
to the process application. There are, for example,
instances in which solids below the grid level are tolerable, where grid thermal expansion is signicant,
where bed solids are very friable, where pressure
drop, and therefore the cost of compressive horsepower, is critical, where solids are sticky and must
be kept in motion throughout, where internal impellers
or stirrers must be provided, or where grids are
expected to have a short life due to corrosion. These
and many other specics have dictated a host of design
variations, some of which are illustrated below. It
should be emphasized that each application requires
thoughtful engineering consideration before nal
design selection.
3
3.1
Gas owing from the grid holes can take the form of
either a series of bubbles or a permanent jet, depending
on system parameters and operating conditions.
However, a permanent jet prevails for most industrial
conditions. Jet penetration is one of the most important design parameters since it helps in
1. Determining how far to keep the bed internals,
such as feed nozzles, heat exchanger tubes, etc.,
away from the grid to minimize erosion of
internals.
2. Deciding on grid design parameters such as
hole size and the gas jet velocity required to
achieve a certain jetting region.
3. Minimizing or maximizing particle attrition at
grids.
Knowlton and Hirsan (1980) reported that the jet
penetration for upwardly directed jets uctuated
greatly. Karri (1990) noted that jet penetration can
vary as much as 30% for upwardly directed jets.
However, the jet emanating from a downwardly directed grid hole is stable, and its penetration length does
not signicantly uctuate with time. Figure 2 indicates
jet penetration congurations for jets oriented
upwardly, horizontally, and downwardly. According
!0:4
rg;b
rp
!0:2
dp 0:2
dh
The jet penetration lengths for upwardly and downwardly directed jets can be calculated from Eq. (1).
These equations take into account the eects of pressure and temperature on jet penetration. Knowlton
and Hirsan (1980) and Yates et al. (1986) found that
the jet penetration increases signicantly with system
pressure. In addition, Findlay and Knowlton (1985)
found that the jet penetration decreases with increasing
system temperature. Bed internals should not be placed
and
3. Under no circumstances should the pressure
drop across a large-scale commercial grid be
less than 2500 Pa, i.e.,
Pgrid 2; 500 Pa
i.e.,
Ph (Highest hole Ph (Lowest hole
480 9:8 0:9
Ph (Lowest hole 4235 Pa
7
Therefore the lowest grid hole has the lowest pressure
drop, and hence the pressure drop criterion must apply
with respect to the lowest hole on the grid.
3.3
Design Equations
The following equations can be used to design perforated plates, spargers, and bubble cap types of grids:
Pressure drop across the grid:
Pgrid KgrB LB
Pgrid 2; 500 Pa
where K 0:3 for upward and lateral gas entry and 0.1
for downward gas entry.
The gas velocity through the grid hole (orice equation):
s
2Pgrid
Uh Cd
rg;h
pdh2
U
4 h
10
number of grid holes per m2 should be 10. In practice, the number of grid holes per square meter should
be greater than 20.
3.3.2
Hole Layout
To increase the uniformity of uidization, it is common to lay out the holes in triangular or square pitch,
as shown in Fig. 5. All the holes in a grid with triangular pitch are equidistant. This is not the case for a
grid with square pitch. Triangular pitch will also result
in more holes per unit area.
The relationship between the grid hole pitch, Lh ,
and the number hole density (holes per unit area of
the bed), Nd , depends on whether the holes are laid
out in triangular or square pitch.
3.4 Additional Criteria for Sparger Grids
Figure 4 Grid hole discharge coecient design chart.
3.3.1
Hole Size
!2
>5
Figure 5 The relationship between hole density and grid hole pitch for both triangular and square pitch.
11
!2
>5
12
Ds dh
2 tan 5:5
13
dh
particle attrition with shrouds
5
5.1
EROSION
Erosion at Bed Walls and Internals
Erosion in the grid region is primarily due to highvelocity submerged jets impinging on distributor
parts, bed walls, or bed internals. Therefore one should
estimate the jet penetration heights for a given grid
design and check for the following:
Figure 6 Manifold sparger grid showing the denitions of
various parameters.
1. Bed internals should not be placed in the jetting zone near the grid, otherwise the internals
could be severely eroded.
Figure 7 (a) Diverging free jet; (b) shroud too short to contain the jet; (c) minimum shroud length required to contain jet.
5.2
15
second mechanism of erosion was observed by highspeed video in cold models (Hartge and Werther,
1998): even when the gas was owing out of the hole
into the bed, a region near the mouth of the orice
could be observed where the gas jet entrained particles
into the hole. These entrained particles caused erosion
at the outer edge of the hole. Figure 10 shows the
photography of a nozzle that had been painted in
black before the experiment. After 60 hours of operation, the erosion marks were clearly visible. They were
particularly obvious at the lower edges of the holes,
which is due to the fact that the jet issuing from a
horizontal bore tends to bend into the upward direction (see Fig. 2), which gives more surface area to
entrain solids at the lower edge of the hole.
6
WEEPAGE OF SOLIDS
PLENUM DESIGN
Figure 14
POWER CONSUMPTION
"
g1=g #
g
P1
P2 Q 2 1
P2
g1
18
Ws;actual
Z
20
Q Usup
pD2
p132
0:6
79:6 m3 =s
4
4
for a turboblower
for a roots blower
for an axial blower or a twostage reciprocating compressor
Equation (20) can be used not only for power consumption but also to size the correct horse power
motor to drive the blower.
The actual temperature of gas leaving a well-insulated (adiabatic) but not 100% ecient compressor is
then calculated from
"
#
T1 P2 g1=g
1
21
T2 T1
Z1 P1
Since Q N
dh2
U
4 h
Therefore
N
Q 1
79:6 1
1:14
2
2
2
Uh pdh =4 88:6 pdh =4
dh
N
1:14
1
0:0086
2
2
2
p=4D
dh p=413
dh2
10
DESIGN EXAMPLES
10.1
68 pdh2 =4 dh2
Various combinations of N and dh satisfy the pressure drop requirements for the two grid type as
shown in the table:
Pbed 4,234 Pa
60,000
15,000
2,400
600
and
Lh 0:29 m
t
0:025
1
dh 0:025
From Fig. 3
Cd
Lh
dh
0:1
D2head
122 0:0252
0:96
0:025 0:1
Cd 0:96
0:29
0:75 vs. 0.77 (initial guess
!2
>5
!2
>5
or
dh 0:025 m
rp 1440 kg=m3
N 1; 824
emf 0:42
97:13
0:04 m
2401
!0:4
rg;b
rp
!0:2
dp 0:2
dh
% of total length
1
2
3
4
Total =
1.43
3.05
4.68
6.30
8.98
19.16
29.41
39.58
97.13
9.24
19.73
30.28
40.75
Lhor
!0:4
0:64 88:62
5:25
14401 0:429:8 65 106
!0:2
0:5 0:2 65 106
0:025 0:32 m
0:025
1440
Perforated Plate
0:05 0:025
0:13 m
2 tan 5:5
Dh
particle attrition with shrouds
0:05 1:6
3:0
0:025
Thus adding a shroud to the grid reduces the attrition rate to 67% of the rate without a shroud.
10.2
rg;b 17 kg=m3
rp 641 kg=m3
rB 272 kg=m3
Pgrid 0:4Pbed
dh 0:01 m
dp 508 mm
t 0:019 m
emf 0:45
Solution
Determine Pbed and Pgrid
Pbed grB LB 9:8 272 12:2 32,520 Pa
Pgrid 0:4Pbed 13,008 Pa
Determine the gas velocity through the grid hole
(trial and error). Assume Cd 0:8:
s
r
2Pgrid
2 13008
0:8
U h Cd
29:5 m=s
rg;h
19:2
Determine the volumetric ow rate of gas
Q Usup
pD2
p4:62
0:5
8:3 m3 =s
4
4
Q
1
8:3
1
3582
Uh pdh2 =4 29:5 p=40:012
Hole density:
Nd
3582
215 holes/m2
p=44:62
1:9
dh
0:01
Pbed 15 kPa;
Pcyclonesfilters
Pgrid 6 kPa;
12 kPa; pressure at the exit of the filters 350 kPa.
Gas entering the compressor: T1 20 C; P1
101 kPaq; Q1 10 m3 =s.
Use Z 0:85; g 1:4
Solution
Determine compressor discharge pressure, P2 :
!0:4
19:2 29:52
5:25
6411 0:459:8 508 106
!0:2
17 0:2 508 106
0:01 0:55 m
0:1
641
Lh
0:073
0:13 < 1
Lup =2 1:1=2
Power Consumption
Z
0:85
1927 kW (or 2587 hp)
Ws;actual
NOMENCLATURE
Cd
dh
dp
D
Dhead
Dm
Ds
g
Hhigh
Hlow
=
=
=
=
=
=
=
=
=
Ke
LB
Ldown
Lh
Lhor
Lmin
Ls
Lup
N
Nd
Nh
Nm
P1
P2
Q
Q1
Q2
t
T1
T2
Uh
Usup
Ws;actual
Ws;ideal
a
g
rB
rg;b
rg;h
rp
emf
y
Pbed
Pgrid
Ph
REFERENCES
Blake TR, Webb H, Sunderland PB. Chem Eng Sci 45:365,
1990
Chen L, Weinstein H. AIChE J 39(12):1901, 1993.
Findlay J, Knowlton TM. Final Report for US Dept of
Energy, Project, DE-AC21-83MC20314, 1985.
Hartge E-U, Werther J. In: Fan LS, Knowlton TM, eds.
Fluidization IX. Engineering Foundation, 1998, p 213.
Hiby JW. Chem-Ing Techn 36:228, 1964.
Karri SBR. PSRI Research Report No. 60, 1990.
Karri SBR. Grid Design Chapter, PSRI Design Manual,
1991
Knowlton TM, Hirsan I. In: Grace JR, Matsen JM, eds.
Fluidization. Plenum Press, 1980, p 315.
Massimilla L. In: Davidson JF et al., eds. Fluidization.
Academic Press, 1985, p 133.
Merry JMD. Trans Instn Chem Engrs 49:189, 1971.
Mori S, Moriyama A. Inst Chem Eng 18:245, 1978.
Roach PT. Fluid Dyn Res 11:197, 1993.
Shakhova NA, Inzh Fiz Zh 14(1):61, 1968.
Siegel R. AIChE J 25:590, 1976.
Sishtla C, Findlay J, Chan I, Knowlton TM. In: Grace JR,
Shemilt LW, Bergougnou MA, eds. Fluidization VI.
Engineering Foundation, 1989, p 581.
VGB Powertech, Design Report M218H, VGB
Kraftwerkstechnik, 1984.
de Vries RJ, van Swaaij WPM, Mantovani C, Heijkoop A.
Proc 5th Europ Symp Chem Reaction Engng,
Amsterdam, 1972, p 3959.
Whitehead AB. In Davidson JF, Harrison D, eds.
Fluidization. Academic Press, 1971, p 781.
Yang WC, Keairns DL. Ind Eng Chem Fundam 18:317,
1979.
Yates JG, Bejcek V, Cheesman DJ. In: Ostergaard K,
Sorensen A, eds. Fluidization V. Engineering
Foundation, 1986, p 79.
Zenz FA, Othmer DF. Fluidization and Fluid-Particle
Systems. Reinhold, 1960, p 171.
Zenz FA. Inst Chem Eng Symp 30:136, 1968.
Zuiderweg. In: Drinkenburg, ed. Proc Int Symp on
Fluidization. Netherlands University Press, 1967, p 739.
7
Effect of Internal Tubes and Bafes
Yong Jin, Fei Wei, and Yao Wang
Tsinghua University, Beijing, Peoples Republic of China
1.1
eliminating gas short-circuits or enhancing gas interchange between the bubble phase and the emulsion
phase, but for limiting the gas axial backmixing and
promoting the radial movement of the gas and solids.
The transition to a turbulent behavior is related to
particle size and density. The larger the particle size
and density, the higher the transition gas velocity,
Uc . It has been found that the transition from a bubbling to a turbulent regime occurs at a smaller Uc when
suitable internals are applied, which is especially
important for Group B particles. Detailed information
is given in Sec. 3.4.
1.3
2
Figure 4 Axial distribution of solids concentration in a
CFB.
1.4
CLASSIFICATION OF VARIOUS
INTERNALS
Bafes
Baes can prevent bubbles from growing continuously, redistribute bubbles across the cross section of
the bed, strengthen the heat and mass exchange
Table 1
Category
Baes
Type
Conguration
Main features
References
Aspect studied
Application eld
Wire mesh
Eectiveness of baes
Perforated plate
1. Improve radial
solids distribution
2. Increase bubble
frequency
Gelperin (1971)
Kono (1984)
Zheng (1990)
Dutta and Suciu
(1992)
Zhao (1992)
Hartholt (1997)
Classication of a binary
mixture of coal particles
Bubble frequency, dense bed
height
Radial distribution of solids
Eectiveness of baes
Solids backmixing
Mixing and segregation of
solids
Single- or
multiple-turn
plate
Louver plate
Jin (1982)
Bubble behavior
Ring
Chemical reaction
(ozone decomposition)
Horizontal
banks
1. Split bubbles
Olsson (1995)
Lofstrand (1995)
Olowson (1994)
Bayat (1990)
Yates (1987)
Levy (1986)
Jodra (1983, 1979)
Glass and Harrison
(1964)
Vertical banks
1. High heat-exchange
coecient
2. Low erosion
phenomenon
3. Low scale-up eect
Station, irregular
Volk (1962)
Huang (1997)
Pressure drop
Solids holdup
Tubes
2. Reduce carryover of
particles
3. Increase bed expansion
4. Increase chemical
conversion
5. Impede solids motion
Packings
Station, regular
Floating
Inserted
bodies
Pagoda-shaped
bodies
(See Fig. 8)
1. Break up bubbles
2. Enhance gassolid
contact
Bubble behavior
Fluid mechanics
Pilot-plant trials
Table 1
Continued
Category
Type
Ridge-shaped
bodies
Conguration
(See Fig. 9)
Inverse cone
Main features
References
Aspects studied
Bed expansion
Bed homogeneity
Change of ow region
Emulsion phase
behavior
Application eld
Bubble uidized bed
Turbulent uidized bed
distribution
Spiral ow
pates
Swages
Inserted
bodies
(cont.)
Other
congurCenter
ations
circulating
tube
Concentration proles
1. Enhance gassolid
Li (1997)
contact
Cui (1996)
2. Lengthen solids residence
time
Vortex strength
Cocurrent and countercurrent
ow patterns
Bed pressure drop
Average density
Davies and
Graham
(1988)
Pressure drop
Vertical pneumatic
conveying tubes
Milne et al.
Stable operation
(1992, 1994)
Temperature
Liu et al. (1993)
distribution
Wang et al. (1993) Axial solid
Fusey et al. (1986)
distribution
Reasonable Design
Correct Installation
Early at the beginning of the development of uidization (in the 1950s), Rasching rings and Berl saddle
packings were packed in uidized beds to enhance
Figure 6 Vertical internals used in the iron reduction process. 1 vertical tube; 2 crosspiece.
the packed bed. More powder inventory, less gas velocities, and larger voids of the packed beds are of benet
for increasing the powder mass ux and hence for
enhancing the heat and mass transfer.
A GSSCFR exhibits good operating stability and
simple construction and can be operated at a wider
range of supercial gas velocity than a gassolid
solid trickle ow reactor (Yamaoka, 1986; Shibata et
al., 1991; Song et al., 1995). The experimental results of
oxidationdehydrogenation of butene to butadiene
show that the C4 H8 conversion and C4 H6 selectivity
can be signicantly increased in a GSSCFR (Huang et
al., 1998).
Some early Soviet work (Goikhman et al., 1968)
suggests that packings (which are much larger than
the uidized particles) may be more eective if they
are light enough to oat freely in the bed instead of
being essentially xed in position as described above.
This appears to be a peculiar approach, but more work
is necessary before oating packings can be evaluated
with condence.
Despite the considerable amount of research work
that has been carried out in the laboratory on packings
in uidized beds, the technique has not yet found a
place in large-scale industrial applications because of
its fatal weakness of preventing solids movement.
It is well known that, compared with a packed bed
reactor, the most outstanding advantage of a uidized
bed reactor is its strong ability for heat exchange. That
is why it is especially suitable for processes with strong
Figure 10
Figure 11 Diagrams of multistage dryer with spiral ow plates. 1spiral ow bed; 2spiral ow panel; 3rst-stage
cyclone; 4measure tank; 5storage tank; 6buttery value; 7electric heater; 8owmeter; 9second-stage cyclone;
10inlet.
Bubble Behavior
Figure 15 Eect of transverse baes (louver plates) on the behavior of bubbles at dierent gas velocities.
Figure 16
and intensies turbulence. Also, there is an increasingly rapid exchange between the bubble and emulsion
phases throughout the bed.
In general, by catching and breaking up the bubbles,
the internals have the eect of reducing their size. On
one hand, when a bubble is split by an internal, the
sum of the volumes of the daughter bubbles is less than
the volume of the parent one, indicating that some of
the original bubble gas leaks into the emulsion phase
of the bed during the splitting process. Furthermore,
the amount of leakage increases as the number of
daughter bubbles increases. These facts could be of
importance in the design of uidized bed reactors.
With the reduction of the mean bubble size, the
expansion ratio of a bed changes. Figure 18 shows
quite clearly the eects that dierent types of internals
have on the expansion ratio in a two-dimensional uidized bed. On the other hand, small bubbles rise relatively slowly in a uidized bed, and so the result is to
increase the total concentration of bubbles in the bed.
A thick mist of small bubbles has a benecial eect on
the conversion of a chemical reaction.
Two dierent methods have been developed for prediction of the bubble size in a bed with internals
formed by horizontal rows of tubes (Jodra and
Aragon, 1983). One, based on a new correlation,
makes it possible to nd the size of a bubble above
the tube as a function of the operating conditions
and characteristics of the tube arrangement. The
other is based on the general theory of bubble formation in submerged orices. Both methods can be used
to estimate bubble sizes, which are acceptably close to
the experimental values, and can be applied to other
types of internals, such as perforated plates, meshes
and arrangements of horizontal bars of noncircular
cross sections.
Flow Distribution
showing that the conversion of the ozone decomposition reaction is improved with the addition of ring
baes. Arrangement of a narrow seam near the wall
further improves the performance. The gas preferably
ows upward in the annular region so that the radial
distribution becomes more uniform. Similarly, a halfoval blu body forces the gas and solids to ow
upwards in the narrow annular region (Fig. 20).
Hence the dilute core and dense annulus ow pattern
are disrupted (Gan et al., 1990). As a consequence,
gassolid contacting is greatly improved.
However, not all internals provide favorable gas
and solids redistribution. Salah et al. (1996) have
found that ring baes with smaller opening areas
can make the radial voidage distribution less uniform
under low gas velocity. With a ring of 44% open area,
Zheng et al. (1990) reported increased nonuniformity
under certain operating conditions. For risers operated
at high gas velocity and high solids ux, e.g., the riser
in FCC units, particles mainly ow upwards at the wall
so that the installation of ring type baes may actually
induce particle downow, leading to more radial ow
segregation. Fortunately, Zheng et al. (1990) have
found that a plate with larger orices and more opening area near the wall is very eective in ow redistribution. Therefore, properly designed internals can help
to redistribute the gas and solids ow.
At the same time, internals can also change the
average voidage. The results of Zheng et al. (1992)
indicate that ring baes signicantly reduce the average solids concentration. Gan et al. (1990) also found
that their blu body bae reduced the average solids
concentration by scraping o superuous particles
The degree of axial mixing of gas and particles in uidized beds is important for many continuous as well as
batch processes, and control thereof is desirable. It is
very important to limit backmixing and maintain good
gassolid contact while using uidization for solid or
gassolid processing, as these operations need high
solid-phase conversion and gas-phase utilization.
First, gas backmixing is substantially reduced by
immersed obstacles. The higher the gas velocity the
smaller the gas backmixing (Ye, 1987). There is little
gas backmixing between the top and the bottom of
each horizontal bae. The axial gas-mixing coecient
is lowered only in beds of small particles or at low gas
velocities. The larger the particles and the higher the
gas velocity, the less eective is the horizontal baes.
Horizontal tubes completely suppress gas backmixing in a large range of velocities, but vertical
tubes are less eective at higher velocities. A noteworthy result with vertical tubes is the development
of marked nonuniformity of radial gas mixing.
Secondly, most internals are likely to suppress axial
solids backmixing. Subdividing the bed with horizontal
baes is a common and successful practice to improve
gassolid contact and to limit solids backmixing. Many
congurations of baes not only can constrain bubble
growth but also can narrow the RTD of gas and solids.
For example, a 1 m I.D. bed with bae spaces of 100
to 200 mm, with a perforation ratio smaller than 30%
and a small space between the plate and the wall, can
reduce axial solids backmixing and lead to a large temperature gradient along the axis. Zheng et al. (1992)
have found that the axial solids dispersion of a ring
bae is only half of that in an empty column.
The baed uidized bed, when used to carry out a
reaction, is a kind of multistage uidized reactor,
which can be modeled by treating the overall mixing
of solids as a series of perfectly mixed stages with solids
backmixing between them. Consequently, the determi-
3.4
Figure 22 Eect of internals on the transition from bubbling to turbulent regime. (a) Free bed, (b) bed with vertical tubes, (c)
bed with pagoda-shaped bodies.
dp
f
gdp
where n 0:27 and Df is a parameter called performance diameter which has length dimension. The
KDf is found to characterize the geometric structure
of the beds, and its value is obtained as follows:
KDf 3:67 103
KDf 2:32 103
KDf 3:42 104
Dutta S, Suciu GD. An experimental study of the eectiveness of baes and internals in breaking bubbles in uid
beds. J Chem Eng (Japan.) 25(3):345348, 1992.
Gan N, Jiang DZ, Bai DR, Jin Y, Yu ZQ. Concentration
proles in fast-uidized bed with blu-body (in Chinese).
J Chem Eng Chinese Univ 3(4):273277, 1990.
Glass DH, Harrison D. Chem Eng Sci 19:10011002, 1964.
Glicksman LR, Lord WK, Sakagami M. Bubble properties
in large-particle uidized beds. Chem Eng Sci 42(3):479
491, 1987.
Goikhman ID, Oigenblik AA, Genin LS, Filippova LA.
Eect of oating spherical packing on the residence-time
distribution of gas in a uidized bed (in Russian). Khim
Technol Topl Masel 13(10):3638, 1968.
Grace JR. Contacting modes and behaviour classication of
gas-solid and other two-phase suspensions. Can J Chem
Eng 64:353363, 1986.
Grace JR, Jarrison D. Design of uidised beds with internal
baes. Chem Proc Eng 6:127130, 1970.
Harrison D, Grace JR. In: Davidson JF, Jarrison D, eds.
Fluidization. New York: Academic Press, 1971, pp 599
626.
Huang SY, Wang ZW, Jin Y, Yu ZQ. Solids mass ux and
mean dynamic solids hold-up in the packed section of the
gas-solid-solid circulating uidized bed (in Chinese).
Chem Reaction Eng Tech 12:369376, 1996a.
Huang SY, Wang ZW, Jin Y, Yu Z. Axial pressure drop and
powders hold-up distribution in packed section of gassolid-solid circulating uidized bed (in Chinese). J Chem
Eng Chinese University 10:377383, 1996b.
Huang SY, Wang ZW, Jin Y. Studies on gas-solid-solid
circulating uidized bed reactors, Chem Eng Sci 54(13
14):20672075, 1999.
Jiang P, Bi HT, Jean RH, Fan LS. Bae eects on performance of catalytic circulating uidized bed reactor.
AIChE J 37:13921340, 1991.
Jin Y, Yu ZQ, Zhang L, Shen JZ, Wang ZW. Pagoda-shaped
internal baes for uidized bed reactors. Intern Chem
Eng 22(2):269279, 1982.
Jin Y, Yu ZQ, Wang ZW, Cai P. A criterion for transition
from bubbling to turbulent uidization. In: Ostergaard K,
Sorensen A, eds. Fluidization V. Elsinore, Deamarh, May
1823, 1986a, pp 289296.
Jin Y, Yu ZQ, Zhang L, Yao WH, Cai P. Ridge type internal
bae for uidized bed reactor. Petro Chem Eng (China)
15(5):269277, 1986b.
Jodra LG, Aragon JM. Prediction of the bubble-size distribution in uidized beds with internal baes. Inter Chem
Eng 23(11):1830, 1983.
Johnsson F, Andersson S. Expansion of a bubbling uidized
bed with internals. Report A. Department of Energy
Conversion, Chalmers University of Technology,
Goteborg, Sweden, 1990, pp 90184.
Kang WK, Sutherland JP, Osberg GL. Pressure uctuations
in a uidized bed with and without screen cylindrical
packings. Ind Eng Chem Fundam 6(4):499504, 1967.
8
Attrition
Joachim Werther and Jens Reppenhagen
Technical University Hamburg-Harburg, Hamburg, Germany
INTRODUCTION
namely its inuence on the bed particle size distribution: it causes in any case a gradual shrinking of the socalled mother particles from which the nes are drawn
until they are nally lost from the system as well (Kunii
and Levenspiel, 1969; Ray et al., 1987b). In addition,
there might be even a fragmentation of bed particles,
resulting in a signicant increase in the number of
particles and a corresponding decrease in the mean
particle size (Blinichev et al., 1968, Pis et al., 1991).
Whether with a batch process or a continuous one,
attrition will thus always lead to dierences in the
bed particle size distribution in comparison to that of
the feed material. Some authors (such as Kraft et al.,
1956; de Vries et al., 1972) report, for example, on the
eect that the bed material in a considered FCC process got too coarse by the addition of makeup catalyst
in compensation for the lost mass of nes. Braca and
Fried (1956), however, observed the opposite eect,
where the bed material of the FCC process got too
ne since the attrition-produced nes were mostly
kept inside the system.
Since the particle population determines almost all
relevant mechanisms in a uidized bed system, attrition may thus strongly aect the performance of a
uidized bed process. For instance, the elutriation
and entrainment eects (cf. Chapter 4 of this book),
the heat transfer from bed to inserts (Molerus, 1992) or
the conversion and selectivity of reactions (Werther,
1992) are aected either directly or through the bed
hydrodynamics by the particle size distribution.
Therefore some authors (such as Ray et al., 1987a)
even claim that a uidized bed with attritable materials
Figure 1 Attrition modes and their eects on the particle size distribution (q3 mass density distribution of particle sizes dp ).
Material Properties
Particle Structure
Particle Size
2.1.4
pically smooth surface is therefore less prone to breakage, but it may still undergo abrasion. With respect to
the latter, the microscopic surface structure is of interest. Surface asperities may be chipped o and lead to
abrasion.
In their investigation on cyclone attrition,
Reppenhagen and Werther (1999a) compared, for
example, the attrition resistance of two dierent catalysts, the chemical composition of which is identical
but their shape and surface structure is dierent (cf.
Fig. 2). Whereas the catalyst termed 97-G(fresh) is
irregularly shaped and has a rough surface, the catalyst
termed 97-R(fresh) is almost spherical, and its surface
shows only few asperities. Reppenhagen and Werther
(1999a) attributed these morphological dierences to
their nding that the 97-G(fresh) catalyst was almost
twice more attritable than the 97-R(fresh) material.
Both the particle shape and the surface structure
will inevitably change with time inside a process. This
will in any case happen as a gradual eect, which is
discussed in Sec. 2.2.3. But it may also happen instantaneously due to excessive stress, for example the surface fragmentation of catalyst particles that has been
observed by Reppenhagen and Werther (1999a) (cf.
Figs. 19, 4, and 5).
2.1.5
Figure 2 SEM images of two catalyst materials. The surface mean diameter was 112 microns in the case of the 97-R(fresh)
and 103 microns in the case of the 97-G(fresh). (From Reppenhagen and Werther, 1999a.)
Figure 3 SEM images of an equilibrium catalyst and the corresponding fresh material. (Reppenhagen and Werther, 1999a.)
and Granatstein, 2001). In any case, these are gradually acting phenomena, that are strongly related to the
inuence of the solids residence time, the eect of
which is discussed in Sec. 2.2. But there may be also
singular phenomena that aect the materials attritability. A broken steam pipe in the stripping unit of an
FCC plant may for example cause a steam jet that
generates extra stress on the catalyst particles. Such a
singular event would not only cause a higher extent of
attrition during its occurrence but also it would most
probably have a long-term eect. The reason for this is
a serious damage of the morphology of the particles.
Reppenhagen and Werther (1999a) observed such a
phenomenon during their cyclone attrition tests as
they exceeded, for several passes through the cyclone,
a certain threshold energy. The result of this experiment is shown in Fig. 4.
Figure 4 Inuence of surface fragmentation on the cyclone loss rate at conditions of pure abrasion (uc;in 20 m=s, c 0:1).
(From Reppenhagen and Werther, 1999a.)
Figure 5 SEM images of a given catalyst material (a) after subjecting it to pure abrasion, and (b) after subjecting it to abrasion
with additional surface fragmentation. (Note: The bridges between the particles in image (b) are not due to sintering but to an
excess gold layer resulting from the SEM preparation step.) (Reppenhagen and Werther, 1999a.)
Process Conditions
The gas velocity is usually directly related to the particle velocity, which is the most important factor in
generating the mechanical stress by interparticle collisions or by particlewall impacts (British Materials
Handling Board, 1987). The forces involved in the
degradation process may be generated by high-speed
collisions
resulting
preferably
in
breakage.
Alternatively, the energy may be transmitted through
a matrix of comparatively slowly moving particles
resulting mostly in abrasion. Particularly in the distributor region of a uidized bed, where grid jets are
issuing into the bed of particles, the impact velocities
Wall-Hardness
Figure 6 Time dependence of catalyst attrition in a submerged jet test. Dt 0:05 m, uor 100 ms1 , dor 2 mm, and HA-HPV
is a fresh catalyst, whereas FCC is a spent catalyst from a commercial unit). (From Werther and Xi, 1993.)
Temperature
Pressure
Chemical Reaction
ASSESSMENT OF ATTRITION
1 dmb
mb dt
1 dmloss
mb dt
rtot t
dt
n
X
hi1
mh;i
i1
X
mi;k
k1
Figure 7 Changes in the particle size distribution of a FCC catalyst material that was subjected to cyclone attrition. (Zenz and
Kelleher, 1980.) The respective experimental procedure is described in Sec. 5.3.
elements bi;k describe the distribution of attrition products from the intervals i into the intervals k of smaller
particle sizes. With a vector of the feed size distribution
they can be combined in a matrix equation that yields a
size distribution vector of the product. Since these vector and matrix elements must be usually obtained
experimentally, the use of selection and breakage functions is much more cumbersome than working with a
single index number. This concept has therefore been
used almost exclusively for the description of commercially interesting comminution processes but very
seldom for the description of catalyst attrition
(Wei et al., 1977).
However, if attrition exclusively occurs by the mode
of abrasion, the respective particle balance appears to
be less complex. There are several mathematical
models of particle population balances suggested in
the literature that deal with that particular mode
(Levenspiel et al., 1969; Overturf and Reklaitis, 1983;
Ray et al., 1987b; Werner et al., 1995; Reppenhagen,
1999; Reppenhagen and Werther, 2001) for uidized
bed systems, or Hansen and Ottino (1997) for a general description of the abrasion process. A comparison
with the above-discussed concept of selection and
breakage functions shows that all these approaches
assume that the particles of a given size fraction are
either entirely subjected to abrasion or left undamaged.
This means that the selection function is either completely omitted, since all particle size ranges are
assumed to be attritable (Levenspiel et al., 1969;
Overturf and Reklaitis, 1983; Werner et al., 1995), or
is reduced to a step function from zero where no abrasion is assumed to one at a particle size above which all
particles are assumed to be attritable (Ray et al.,
1987b; Hansen and Ottino, 1997). The respective
breakage functions are also comparatively simply
organized in the above approaches. The particle-size
dependence of the attrition extent and thus the mass
of attrition-produced nes per size interval and unit
time is either described by a rst-order kinetics
(Levenspiel et al., 1969) or by a potential dependence
(Hansen and Ottino, 1997). The entire mass of nes is
then either simply assumed to be lost (Levenspiel et al.,
1969) or is assigned to a particular size interval with a
given constant size distribution (Ray et al., 1987b;
Hansen and Ottino, 1997).
In most of the above approaches the balances are
solved analytically and the abrasion-induced shrinking
of the mother particles is thus taken into account by
simply shifting the size distribution in accordance with
the respective material losses. In contrast to this,
Reppenhagen and Werther (2001) recently developed
ATTRITION TESTS
The term attrition test generally covers all experiments in which the eects of attrition are considered
and assessed. This might vary from the observation of
the degradation of only a single particle (Boerejn et
al., 2000) to the consideration of an entire process
(Werther and Reppenhagen, 1999). However, analogously to the division of the assessment procedures,
the large number of experiments can be divided into
two major elds of application, namely the tests of
material friability and the experiments to study attrition phenomena. They will be separately discussed in
the following sections. The relevant test devices will be
discussed afterwards.
4.1
Friability Tests
Examples of such test facilities are the various jet attrition test devices (such as Forsythe and Hertwig, 1949;
Gwyn, 1969; Werther and Xi, 1993), which simulate
the stress generated in the vicinity of a gas distributor.
The results can either be directly extrapolated to a
large-scale process or reveal relationships that may
serve as guidelines in the design of this process. It
should be noted that most of these tests are batch
tests, but most full-scale processes are continuous.
Hence the transfer of the results can be dicult
owing to the time dependence of the degradation.
Furthermore, it must be taken into account that
such a bulk test can only give results that are based
on statistics, i.e., they reveal the average eect on all
particles but cannot indicate a specic mechanism of a
single particle breakage. In any case attention must be
paid when the properties of the bulk sample tested
dier from those in the real process. As an example,
Werther and coworkers (Werther and Xi, 1993;
Reppenhagen and Werther, 1999a) used previously
screened catalysts in their investigation of the mechanisms of jet and cyclone attrition but also evaluated the
eect of the particle size distribution to take the shift in
the particle size into account (Werther and
Reppenhagen, 1999).
In the second attrition-modeling philosophy
(Yuregir et al., 1987; Boerejn et al., 1998, 2000), it
is believed that the mechanics of particle interaction
in process test devices and in the large-scale processes
are not suciently understood to maintain the necessary dynamic similarity and that the analysis of the
experimental data for relating the trend to the largescale operation is therefore not straightforward
(Boerejn et al., 2000). Instead, test devices have to
be used where the material is subjected to a welldened stress like compression, impact, or shear. The
chosen test should preferably apply the type of stress
that is considered as the main source of attrition in the
large-scale process. The complete process attrition can
than be described by coupling the test results with a
model that describes the motion of uid and particles
and with it the frequency and extent of the stress.
Unfortunately, the practical use of this latter approach
is at the limit of our current state of knowledge. First
of all, it is not easy to dene the relevant stress of a
process, i.e., the stress is usually composed of more
than one of the above mentioned pure types of stress,
and it is usually changing in dependence on various
parameters. As an example, Boerejn et al. (2000)
reported a gradual change of the grid-jet-induced stress
from shear to impact stress with increasing orice-toparticle size ratio. Moreover, the necessary models of
Figure 10
(1969).
Comparison of two dierent designs of a submerged jet test (a) after Forsythe and Hertwig (1949); (b) after Gwyn
Cyclone Tests
Figure 11
Experimental setup for the investigation of cyclone attrition. (After Reppenhagen and Werther, 1999a.)
Figure 12 Loss rate obtained with a screened spent FCC catalyst as a function of the number of passes through the cyclone
(uc;in : gas velocity at the cyclone inlet; c : solids to gas loading ratio, 90 mm ID cyclone). (After Reppenhagen and Werther,
1999a.)
m_ loss;c;steady-state m_ a;c;fines
m_ c;in
m_ c;in
must be taken into account. As for the latter parameter, Reppenhagen and Werther (1999a) have given
a simple example. They compared the test results of a
previously screened fraction with those of an
unscreened fraction of the same material. By taking
the dierences in the mean particle size into account
(cf. Sec. 5.3), they observed a distinctly higher loss rate
(approximately 25% higher) in the case of the
unscreened fraction owing to the additional loss of
suciently shrunk mother particles.
In general, it must be recognized that the greater
eort, and the large amount of material required, for
the currently available cyclone attrition test methods
will generally prevent them from being used as a pure
friability test, i.e., a test for a sole ranking of various
materials without any reference to a particular process.
However, the tests are useful for both the study of
general attrition mechanisms in cyclones and the
evaluation of a given materials susceptibility to the
particular attrition stress inside cyclones, which may
be needed as an input parameter in a cyclone attrition
model.
u3or
from both the jets and the bubbling bed. And the jet
attrition rate can be now calculated by subtracting the
bubble attrition rate measured before from the overall
attrition rate. This is a certain oversimplication, since
it assumes that the bubble attrition will always be the
same regardless of the ratio of jet air mass ow to
auxiliary mass ow. However, this method holds fairly
well for low jet gas velocities, where the contribution of
the jet gas ow to the total gas ow is rather small. For
higher jet gas velocities, the jet attrition is so high that
the contribution of bubble attrition may be neglected.
Using the above-described experimental setup,
Werther and Xi (1993) carried out a comprehensive
experimental program to study grid-jet attrition.
They found that at steady-state attrition conditions
(cf. Fig. 6), the jet-induced attrition exclusively occurs
in the mode of pure abrasion. On the basis of this
result they suggested and validated a model that considers the energy utilization of this abrasion process by
relating the surface energy created to the kinetic energy
that has been spent to produce this surface area.
Assuming that there is no interaction between the individual jets of a given distributor, they derived the following relationship for the grid-jet-induced generation
of nes:
2
u3or
m_ abr:;fines;j nor
Kj
rf
dor
uor
10
V_
Ao
11
V_ 3
A2o
12
5.2
Bubble-Induced Attrition
m_ loss;b;steady-state m_ attr:;fines;b
mb
mb
Kb
mnb
u umin
13
14
15
m_ attr:;fines;b
Kb
u umf
mb
16
m_ attr:;fines;b
Kb0
mb
u umf
mb
17
of the grid jets. However, their inuence might be negligible in the present case due to the relatively small jet
velocity.
Ray et al. (1987a) obtained a fairly dierent result
as they considered the attrition of narrow fractions of
limestone with particle sizes between 1.09 and 0.77 mm
in a 0.1 m diameter bed, which was equipped with a
porous plate (Fig. 16). Similar results were obtained by
Xi (1993), who investigated the attrition of ne catalyst
particles with a minimum uidization velocity umf of
0.002 m/s (Fig. 17). As is obvious from Figs 16 and 17,
the attrition rate extrapolates to zero at a uidizing
velocity umin which is signicantly larger than umf .
This means that a minimum kinetic energy or a minimum extent of bubbling is necessary to cause attrition
in these cases.
More recently, Werther and Reppenhagen (1999)
correlated their bubbling bed attrition test results
obtained under pure abrasion conditions in a 200 mm
ID Gwyn-type test plant even to the excess gas velocity
raised to a power of 3 (Fig. 18), which indicates that
the bubbling bed attrition is linearly increasing with
the excess kinetic energy supplied to the system:
rb
m_ loss;b;steady-state m_ abr:;fines;b
mb
mb
Kb
u umf
18
m_ loss;c;steady-state m_ abr:;fines;c
Kc
mnc
u2c;in
m_ c;in
m_ c;in
19
mc
mc
22
In conclusion from the sections above, two main distinctions must be made when attrition in an overall
uidized bed system is considered: Primarily it must
be recognized that there are several attrition sources
in a uidized bed system with distinctly dierent attrition mechanisms, which must be described separately.
20
21
Figure 20 Cyclone attrition test results obtained with various types of catalysts in a 90 mm ID cyclone. (Reppenhagen
and Werther, 1999a.)
the elutriated material. Under conditions of steadystate attrition, the loss rate was then obtained from
the increase in the lter weight per unit time. As bed
material a fresh FCC catalyst material was used. In
order to ensure that no original particle but only
attrition-produced debris is collected on the lter,
the catalyst was previously sieved to remove its fraction of elutriable nes.
The results of two dierent test series are shown in
Fig. 22, where they are plotted as a function of the
supercial gas velocity in a double logarithmic grid.
The rst series of measurements was taken by using a
porous plate as a gas distributor, which was substituted by a perforated plate (640 orices of 0.7 mm
ID) in the second series. For both series a strong sensitivity against the gas velocity can be observed. In the
case of the porous plate distributor, a 10% increase in
the supercial gas velocity leads almost to a doubling
of the loss ow. In case of the perforated plate, the
sensitivity against the gas velocity is lower, but the
absolute values of the loss ows are higher.
In order to describe such parameter dependencies,
the conventional approach would be simply to correlate the measured overall loss ows in the form of a
power law. For the data shown, this would lead to a
dependence on the supercial gas velocity raised to a
power of 4.2 or even 7, respectively. A physically
sound explanation for these exponents is certainly difcult, and it is obvious that such a correlation obtained
with the one system design is not directly transferable
to the other even though only the distributor was
exchanged, which reveals that such a model approach
Figure 22 Measured loss ows from the uidized bed system shown in Fig. 21 operated with previously screened fresh
FCC catalyst. (After Werther and Reppenhagen, 1999;
Reppenhagen and Werther, 1999b.)
u3or
Kb
mb
u umf 3
u2
m_ c;in
Cc
dpc
pc;in
mc
23
D2t
2
nor
dor
24
m_ c;in
rf
uc;in
Ac;in
25
At
Ac;in
26
and
uc;in u
D6t
u3
n2or
4
dor
Kb
mb
u umf 3
Cc
dpc
p p A2:5
m_ c;in
rf
2t
u2:5
Ac;in
27
D6t
u3
4
n2
dor
or
Kb
mb
u umf 3
A3t
A2c;in
X
u
wi
exp 5:4 ti
u3:75
u
i
3:81
Cc
dpc u
rf
Gsi
At
31
On the basis of Eq. (31), the total loss rate and the
contributions of the individual sources of a given system can be a priori calculated in dependence on the
supercial gas velocity. Using the same values for the
materials attrition rate constants and the system
design as in the above considerations, this is demonstrated in Fig. 24.
Obviously, there is a strong sensitivity of cyclone
attrition inside a uidized bed system against the
supercial gas velocity. It is signicantly higher than
the sensitivities of jet- and bubble-induced attrition,
respectively. Furthermore, it is obvious that owing to
these dierent dependencies, the role of the main attrition source changes with the gas velocity. In the lower
velocity rangein this particular example below
0.55 m/sthe gas distributor is the main attrition
source, whereas at higher velocities cyclone attrition
is dominant.
wi
Ki
28
G
G
dpc u P s P s
Gsi
w i
Ki
i xi
i
xi
29
u
Ki 14:5
rf
u2:5
exp 5:4 ti
30
u
Continuous Operation
i 2
n 1
32
Figure 25
m_ loss;i m_ 2;1
dt
i2
33
dmn
m_ loss;n m_ n;n1
dt
34
j b ;
j ;
c
C Cb ; Cj ; Cc
dpj
2
with
b mb
u umf 3 ; j rf
dor
u3or ; c
p
2:5
m_ c;in
rf
Ac;in
uc;in representing the respective inuences of geometry and operating parameters. With the
denition of the surface mean diameter, Eq. (35) can
be written as
X
X
m_ loss
36
m_ loss j C
x i
Q2;i
Abrasion-induced mass transfer between particle size intervals. (From Reppenhagen and Werther, 2001.)
37
rs;a
p
x 3i
m_ loss;i
m_ loss;i
Np;i
Q3;i
mtot
6
38
with Np;i being the number of particles in the size interval Ki . rs;a is the apparent density of the solids, and
mtot is the total mass of solids. Provided that m_ loss;i is
constant in the time interval t, the mass balance for a
single particle can be written as
mi;p;t mi;p;tt m_ loss;i;p
t
This can be transferred to
s
m_ loss;i;p
t
6 p 3
x
x i;t 3
p 6 i;tt
rs;a
39
40
t
x i;t x i;tt
3 1 loss;i
41
Q3;i
mtot
xi xi di
43
with xi being the smallest particle size in the size interval Ki . According to Fig. 26, the transferring mass
fraction can then be calculated as
x x i
mi;i1 mi mloss;i
i
xi1 xi
44
3 1
1
Q3;i
mtot
42
Figure 27 The particle population balance model of a uidized bed system. (Reppenhagen and Werther, 2001.)
Figure 28 Measured and calculated shift in the size distribution when a batch of catalyst was fed 600 times to a cyclone.
(Reppenhagen and Werther, 2001.)
Bx; y
Cb
Cc
Cj
C
di
dor
dp
dp
dpc
dpc u
dpb
dpj
Dt
Etot
Gs
Gsi
NOMENCLATURE
Ac;in
Ao
At
b
Ka
Kb
Kb0
Kb
Kc
Ki
Kj
m_ abr:;fines
mb
mb;0
m_ c;in
m_ crit
m_ fluid
mi
m_ i;i1
= breakage function
= particle size independent rate constant of
bubble-induced attrition, s2 =m3
= particle size independent rate constant of
cyclone attrition, dened by Eq. (21), s2 =m3
= particle size independent rate constant of jetinduced attrition, dened by Eq. (8), s2 =m3
= particle size independent rate constant,
dened by Eq. (35), s2 =m3
= abrasion-induced reduction in the particle
diamter in interval i, dened by Eq. (42), m
= diameter of an orice in a multihole gas
distributor, m
= surface mean diameter, m
= surface mean diameter, dened by Eq. (35),
m
= surface mean diameter of the material that
enters the cyclone, m
= surface mean diameter of the elutriated
material, dened by Eq. (29), m
= surface mean diameter of the bed material, m
= surface mean diameter for jet-induced
attrition, dened by Eq. (35), m
= diameter of the uidized bed column, m
= total attrition extent of a given system,
dened by Eq. (4)
= total rate of solids elutriation from the bed,
kg=m2
s
= fractional elutriation rate for the size interval
i, kg=m2
s
= attrition rate constant in the Gwyn equation,
Eq. (1), sb
= rate constant of bubble-induced attrition,
dened by Eq. (13), m1
kgn
= rate constant of bubble-induced attrition,
dened by Eq. (17), m1
kgn
= rate constant of bubble-induced attrition,
dened by Eq. (18), s3 =m3
= particle size dependent rate constant of
cyclone attrition, dened by Eq. (19), s2 =m2
= elutriation rate constant of the size interval i,
kg=m2
s
= particle size dependent rate constant of jetinduced attrition, dened by Eq. (7), s2 =m2
= mass of abrasion-produced nes per unit
time, kg/s
= bed mass, kg
= initial bed mass, kg
= solids mass ow rate into the cyclone, kg/s
= critical solids mass ow rate, dened by Eq.
(45), kg/s
= mass ow of the uid, kg/s
= material mass in the size interval i, kg
= mass transfer ow rate due to particle
shrinking from the size interval i to i 1,
kg/s
mloss
m_ loss;c
m_ loss;i
m_ loss;i;p
m_ loss;j
m_ loss;tot
mreturn
mtot
Nc
nor
Np;i
p
p
Q2 x
Q2;i
q3 x
Q3 x
Q3;i
rb
rc
rtot
Sx
t
u
uc;in
umin
umf
uor
uti
V_
wi
x
x i
xlimit
Greek Symbols
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
REFERENCES
Anthony EJ, Granatstein DL. Sulfadation phenomena in
uidized bed combustion systems. Progress in Energy
and Combustion Science 27:215236, 2001.
Arastoopour H, Chen C-Y. Attrition of char agglomerates.
Powder Technol 36:99106, 1983.
Arena U, DAmore M, Massimilla L. Carbon attrition during the uidized combustion of coal. AIChE J 29:4048,
1983.
Austin LG. A reviewintroduction to the mathematical of
grinding as a rate process. Powder Technol 5:117, 1972.
Bemrose CR, Bridgwater J. A review of attrition and attrition test methods. Powder Technol 49:97126, 1987.
Blake TR, Webb H, Sunderland PB. The nondimensionalization of equations describing uidization with application
to the correlation of jet penetration height. Chem Eng Sci
45:365371, 1990.
Blinichev VN, Streltsov VV, Lebedeva ES. An investigation
of the size reduction of granular materials during their
processing in uidized beds. Int Chem Eng 8(4):615618,
1968
Boerejn R, Zhang SH, Ghadiri M. Analysis of ISO uidized
bed test for attrition of uid cracking catalyst particles.
In: Fan LS, Knowlton TM, eds. Fluidization IX. New
York: Engineering Foundation, 1998, pp 325332.
Boerejn R, Gudde NJ, Ghadiri M. A review of attrition of
uid cracking catalyst particles. Adv Powder Technol
11(2):145174, 2000.
9
Modeling
Thomas C. Ho
Lamar University, Beaumont, Texas, U.S.A.
INTRODUCTION
PSEUDOHOMOGENEOUS MODELS
ekt E dt
1
CA0
0
and the following equation was proposed to describe
the eective rate constant, i.e.,
k k0 tm
TWO-PHASE MODELS
In most two-phase models, a uidized bed is considered to consist of two distinct phases, i.e., a bubble
phase and an emulsion phase. Each phase is represented by a separate governing equation with a term
in each equation describing mass interchange between
the two phases. A general expression of the two-phase
models, therefore, consists of the following two equations (from Wen and Fan, 1975). For the bubble phase
we have
!
@CA;b
@2 CA;b
@CA;b
FDb
FU
F
@t
@h
@h2
4
F0 CA;b CA;e Fs kCA;b 0
and for the emulsion phase, we have
!
@CA;e
@2 CA;e
@CA;e
f
fU
fDe
@t
@h
@h2
F0 CA;e CA;b fs kCA;e 0
Table 1
Authors
Model
Experimental conditions
Experimental results
b 0
Ue Umf
De 0 or 1
parameter Fo
k 0:06 0:05 (1/s)
De 0
Ue Umf
Lewis et al.
(1959)
Ue 0
De 0 or 1
b 0
Ue Umf
De 0
b 0
Ue Umf
De 0
Orcutt et al.
(1962)
b 0
Ue 0
Kobayashi et al.
(1966a)
Ue Umf
De 0
Kobayashi et al.
(1966b)
Ue Umf
De 0
Kato
(1967)
Ue Umf
De 0
Kobayashi et al.
(1967)
De 0
Ue Umf
De Grout
(1967)
De 0
Ue Umf
Kato et al.
(1967)
Ue 0
De 0:68
U Umf
dp "p
Umf
parameter: Fo , b
k 0:64 (1/s)
parameter: Fo , b
k 1:1 15:6 (1/s)
b 0:05 0:18, F 0:4 0:8
parameter: Fo , b
k 0:75 (1/s)
parameter: Fo
k 0:071 (1/s)
parameter: b
k 0:1 0:8 1/s
b 15 L=Lmf 1
parameter: b
k 0:2 3:5 1/s
b 0:1 0:3
parameters: b
k 1:1 3:3 1/s
b 0:35 0:45
Fo 11=db
Parameter assumed
b 0
Ue 0
De Ds
De 0 or 1
b 0
Muchi
(1965)
Ue Umf
0 < De < 1
Ue Umf
b 0
Ue 0
De 0
De 0
b 6 0
De 0
Method
A steady state analysis of gas
backmixing and residence time
curve and rst-order reaction
by two-phase model
Estimation of conversions for
a rst-order reaction
A study of eect of Fo , b , De
Ue on conversion of a rstorder reaction
Analysis of a rst-order reaction
based on the two-phase cell
model
A study of the eect of k, b ,
De , and Fo on conversion of a
rst-order reaction
Consider bubble-cloud phase
reaction in bubble-cloud
phase with non-rst-order
kinetics
Analysis of backmixing
residence time curve of tracer
gas and the rst-order
reactions
b or F0
Hk
Remarks
Fo L
U
Hk 0:5 2:5; b 0
Fo
5:85D1=2 g1=4
db5=4
4:5Umf
db
Parameter Fo is not
related to the bubble
movement in the bed
Parameter db model
does not account for
bubble growth in the
bed
No relation between
bubble movement and
parameter
Fo = 0:05
shape factor of bubble
Kg;c dc
D
Parameters b , Fo , De
are not related to the
bubble movement
Allow the variation
of bubble population
with height
Parameters b , Fo , Ue ,
are not related to the
bubble growth in the
bed
concentration prole, predicted dense phase concentrations are far too high and too close to the bubble phase
concentrations. They concluded that the mass transfer
rate in the model is too high.
3.1.2
3.1.3
In recent years, several modied versions of the twophase model were proposed for modeling uidized bed
reactors. They include a model proposed by Werther
(1980) for catalytic oxidation of ammonia, in which the
mass transfer process is expressed in terms of lm
theory, as described in Danckwerts (1970); a model
proposed by Werther and Schoessler (1986) for catalytic reactions; a model proposed by Borodulya et al.
(1995) for the combustion of low-grade fuels; a
model proposed by Arnaldos et al. (1998) for vacuum
drying; and a model proposed by Srinivasan et al.
(1998) for combustion of gases. The modications
include the consideration of axial mass transfer prole,
the inclusion of a wake phase in addition to the bubble
and emulsion phases, and the consideration of the
growth of bubbles in the bubble phase.
9
(2) the bed fraction in bubbles, d:
d
U 1 d adUmf U Umf
ub
ub
10
3dUmf ="mf
ubr Umf ="mf
11
12
13
adub
1 d ad
14
15
0:5 0:25
Umf
D g
5:85
db
db1:25
16
17
Using Davidsons theoretical expression for bubblecloud circulation and the Higbie (1935) theory for
cloud-emulsion diusion, the interchange of gas
between bubble and cloud is then found to be
Kbc 4:5
3.2.2
20
21
and
a 0:25 1:0
22
Umf
6
7
"mf
a7
23
gc 1 "mf 6
43
5
umf
ubr
"mf
and
ge
1 "mf 1 d
gb gc
d
24
1
Kbc 2
26
1
3
27
2 gc k
and
3
1
1
Kce ge k
28
ub
1 "mf
ubr
30
Note that since the bubble size is the only quantity that
governs all the rate quantities with the exception of k,
Examples illustrating the model and additional discussions on the model are found in Levenspiel (1972) and
Kunii and Levenspiel (1991). Applications of the
model were reported in several recent studies, including
scale-up studies of catalytic reactors by Botton (1983)
and Dutta and Suciu (1989), a gasication study of
coal carried out by Matusi et al. (1983), and a study
of fast uidized bed reactors by Kunii and Levenspiel
(1998).
3.3 Bubble Assemblage Model
Since the development of the bubbling bed model, various other hydrodynamic models have been proposed
using other combinations of assumptions such as
changing bubble size with height in the bed, negligible
bubble-cloud resistance, negligible cloud-emulsion
resistance, and nonspherical bubbles. Among them,
the bubble assemblage model, proposed by Kato and
Wen (1969), considers changing bubble size with height
in the bed. The model has the following assumptions:
1. A uidized bed may be represented by n compartments in a series. The height of each compartment is equal to the size of each bubble at
the corresponding bed height.
2. Each compartment is considered to consist of a
bubble phase and an emulsion phase. The gas
ows through the bubble phase, and the emulsion phase is considered to be completely
mixed within the phase.
3. The void space within the emulsion phase is
considered to be equal to that of the bed at
the incipient uidizing conditions. The upward
velocity of the gas in the emulsion phase is Ue .
4. The bubble phase is assumed to consist of
spherical bubbles surrounded by spherical
clouds. The voidage within the cloud is
assumed to be the same as that in the emulsion
phase, and the diameter of the bubbles and
that of clouds is given by Davidson (1961) as
3
dc
u 2Umf ="mf
br
ubr Umf ="mf
db
31
Umf
for
ubr
"mf
where
ubr 0:711gdb 0:5
32
Fd Fo K M
8. The bed is assumed to be operating under isothermal conditions since the eective thermal
diusivity and the heat transfer coecient are
large.
36
37
Lmf
0:711g db;a 0:5
38
U Lmf
do
Umf 2
39
Lmf
1 "mf
L
40
33
7. The bubbles are considered to grow continuously while passing through the bed until they
reach the maximum stable size or reach the
diameter of the bed column. The maximum
stable bubble size, db;t , can be calculated by
(Harrison et al., 1961)
u 2 1
t
db;t
34
0:711 g
3.3.1
Lmf
1 "mf
L
L 1 "mf h Lmf
0:5 mf
2LL Lmf
41
42
where
y
L Lmf
L
43
44
3.3.2
0:11
db
45
46
or
2do
2c
47
where
c 0:14rp dp
U
Umf
48
2c
2 c2
2 cn1
2 cn
50
6S" "mf
phn 2 1 "mf
51
52
where
ubr 0:711ghn 0:5
53
54
and
55
h1
hn 2do
d ch1 do
h1 o
2
49
56
57
59
where rA;c and rA;e are the reaction rates per unit
volume of the cloud and the emulsion phase, respectively. Note that for the rst-order reaction, the expressions for rA;c and rA;e are rA;c kr CA;b and
rA;e kr CA;e , respectively.
Figure 3 Main features of bubble assemblage model. (From Kato and Wen, 1969.)
3.3.3
The computational procedure for conversion and concentration prole in a uidized bed reactor is given
below. The following operating conditions are needed,
i.e., particle size (dp ), particle density (rp ), minimum
uidization velocity (Umf ), gas supercial velocity (U),
distributor arrangement (no ), column diameter (dt ),
incipient bed height (Lmf ), reaction rate constant (kr ),
and order of reaction. It should be noted that the
model requires no adjustable parameters.
First, Eq. (38) is used to calculate the expanded bed
height, L. Next, Eq. (50) is used to compute the size of
the nth compartment. Using Eqs. (51) through (55), the
volumes of the cloud, the bubble phase, and the emul-
More detailed discussion regarding the model performance is given by Wen and Fan (1975) and Mori and
Wen (1975). The literature-reported applications of
this model include a combustion study of coal with
limestone injection by Horio and Wen (1975), a coal
gasication study by Mori et al. (1983), a study on
catalytic oxidation of benzene by Jares et al. (1983),
a catalytic ammoxidation of propylent by Stergiou and
Laguerie (1983), a silane decomposition study by Li et
al. (1989), a catalytic oxidation of methane by Mleczko
et al. (1992), and a study on chlorination of rutile by
Zhou and Sohn (1996).
3.4
Comparison of Models
MULTIPLE-REGION MODELS
62
p
64
d2
Qor uor
4 or
and kbe1 represents the diusive mechanism that can be
predicted by the penetration theory expression, i.e.,
kbe1
4D"mf
ptf
0:5
65
When bubbles burst and release their gases at the surface of a uidized bed, it is generally observed that
particles are ejected into the freeboard space above
the surface. These particles can originate either from
the dense phase just ahead of the bubble at the moment
of eruption (Do et al., 1972) or from those solids that
travel in the wake of the rising bubbles (Basov et al.,
1969; Leva and Wen, 1971; George and Grace, 1978).
66
Vcell
67
68
m
rg D
69
u t d p rg
m
70
The model equations were solved for various combinations of the parameters, and it was reported that in
many cases freeboard reactions lead to greater conversion than that achieved by the uid bed itself. The
details of the results are also reviewed by Yates (1983).
CONCLUDING REMARKS
aj
CA
CAh
CA;b
CA;bn
CA;c
CA;cell
CA;e
ab
CAp
Ap
CAo
NOTATION
CA;en
CA;j
CAjJ
CA;or
Db
De
db
db;a
db;t
dc
d0
dor
dp
dt
E
F
Fd
Fo
Fon
0
Fon
Fs
f
f0
fs
g
h
h1
h1
hn
hn
J
K0
Kbc
Kce
k
kbe
kbe1
kg
kje
k0
kr
L
Lfluidized
Lmf
Lpacked
M
m
N
NA
n0
Qor
rA;b
rA;c
rA;e
rA;s
Ret
S
Sex;b
Sc
Sh
t
tf
Vs
a
a0
b
b0
gb
=
=
=
=
=
gc
ge
D
d
"
"A
=
=
=
=
"mf
"packed
y
p
rg
rp
m
1
2
3
o
o
=
=
=
=
=
=
=
=
=
=
=
=
U
Ue
Umf
ub
ubr
ue
uor
us
ut
Vb
Vbn
Vc
Vcell
Vcn
Ve
Ven
Vp
REFERENCES
Arnaldos J, Kozanoglu B, Casal J. Vacuum uidization:
application to drying. In: Fan LS, Knowlton TM, eds.
Fluidization IX, Durango, Colorado. New York:
Engineering Foundation, 1998, pp 709716.
Basov VA, Markhevkd VI, Melik-Akhnazarov TK, Orochko
DI. Investigation of the structure of a nonuniform uidized bed. Int Chem Eng 9:263273, 1969.
Bauer W, Werther J. Proc 2nd World Congress Chem Eng
3:69, 1981.
Behie LA, Kehoe P. The grid region in a uidized bed reactor. AIChE J 19:10701072, 1973.
Borodulya VA, Dikalenko VI, Palchonok GI, Stanchits LK.
Fluidization and combustion of solid biological waste and
low-grade fuels: experiment and modelling. In: Large JF,
Laguerie C, eds. Fluidization VIII, Tours, France. New
York: Engineering Foundation, 1995, pp 351358.
Botton R, Vergnes F, Bergougnou MA.Validation by means
of industrial data of Kunii-Levenspiel type bubble models
which can be used in the scale-up to commercial size of
uidized bed reactors. In: Kunii D, Toei R, eds.
Fluidization IV, Kashikojima, Japan. New York:
Engineering Foundation, 1983, pp 575582.
Chavarie C, Grace JR. Performance analysis of a uidized
bed reactor. II. Observed reactor behavior compared with
simple two-phase models. Ind Eng Chem Fund 14:7986,
1975.
Cooke MJ, Harris W, Highley J, Williams DF. Intn Chem
Engrs Symp Ser 30:21, 1968.
Danckwerts PV. Gas Liquid Reaction. London: McGraw
Hill, 1970.
Davidson JF, Harrison D. Fluidized Particles. Cambridge:
Cambridge University Press, 1963.
De Groot JH. Proceedings of the International Symposium
on Fluidization, 1967, pp 348361.
Do H, Grace JR, Clift R. Particle ejection and entrainment
from uidized beds. Powder Technology 6:195200, 1972.
Dutta S, Suciu GD. Unied model applied to the scale-up of
catalytic uid bed reactors of commercial importance. In:
Grace JR, Shemilt LW, Bergougnou MA, eds.
Fluidization VI, Ban, Canada. New York: Engineering
Foundation, 1989, pp 311318.
Ellis JE, Partridge BA, Lloyd DI. Proc Tripartite Chem Eng
Conf, Montreal, 1968, pp 43.
Fryer C, Potter OE. Bubble size variation in two-phase models of uidized bed reactors. Powder Technology 6:317
322, 1972.
George SE, Grace JR. Entrainment of particles from aggregative uidized beds. AIChE Symp Ser 74(176):6774,
1978.
Gilliland ER, Knudsen CW. Paper 16d, AIChE Annual
Meeting, Chicago, 1970.
Gomezplata A, Shuster WW. Eect of uniformity of uidization on catalytic cracking of cumene. AIChE J 6:454
459, 1960.
Rowe PN, Claxton KT, Lewis JB. Heat and mass transfer
from a single sphere in an extensive owing uid. Trans
Inst Chem Eng 43:1431, 1965.
Shen CY, Johnstone HF. Gassolid contact in uidized beds.
AIChE J 1:349354, 1955.
Sit SP, Grace JR. Interphase mass transfer during bubble
formation in uidized beds. In: Ostergaard K,
Soorensen A, eds. Fluidization V, Elsionore, Denmark.
New York: Engineering Foundation, 1986, pp 3946.
Srinivasan RA, Sriramulu S, Agarwal PK. Mathematical
modeling of the combustion of gases in bubbling uidized
beds. In: Fan LS, Knowlton TM, eds. Fluidization IX,
Durango,
Colorado.
New
York:
Engineering
Foundation, 1998, pp 693700.
Stergiou L, Laguerie C. An experimental evaluation of uidized bed reactor models. In: Kunii D, Toei R, eds.
Fluidization IV, Kashikojima, Japan. New York:
Engineering Foundation, 1983, pp 557564.
Toomey RD, Johnstone HP. Gaseous uidization of solid
particles. Chem Eng Progress 48:220226, 1952.
Van Deemter JJ. Mixing and contacting in gassolid uidized
beds. Chem Eng Science 13:143154, 1961.
Van Deemter JJ. Proceedings of the International
Symposium on Fluidization, Eindhoven, 1967, pp 322.
Werther J. Modeling and scale-up of industrial uidized bed
reactors. Chem Eng Science 35:372379, 1980.
Werther J, Schoessler M. Modeling catalytic reactions in
bubbling uidized beds of ne particles. In: Van Swaaij
WPM, ed. Heat and Mass Transfer in Fixed and
Fluidized Beds. New York: Hemisphere, 1986, pp 355
370.
Wen CY, Fan LT. Models for Flow Systems and Chemical
Reactors. New York: Marcel Dekker, 1975.
Yates JG. Fundamentals of Fluidized-Bed Chemical
Processes. London: Butterworths, 1983.
Yates JG, Rowe PN. A model for chemical reaction in the
freeboard region above a uidized bed. Trans Inst Chem
Eng 55:137142, 1977.
Zhou L, Sohn HY. Mathematical modeling of uidized-bed
chlorination of rutile. AIChE J 42:31023112, 1996.
10
Heat Transfer
John C. Chen
Lehigh University, Bethlehem, Pennsylvania, U.S.A.
INTRODUCTION
2
2.1
Remf 2
1 emf Remf
150
3
fs 2 emf 3
fs emf
Umf dp rg
mg
2.2
2
grg rs rg 3
dp
m2g
dp Ar1=3
6a
dp8=5 rs rg
#5=7
rg2=5 m3=5
g
6b
"
2.2.1
Heat Transfer
Ug Umf
Ut Umf
q
ap Tp Tg
Nup
hp d p
kg
Rep
U g d p rg
mg
10
11
Figure 2 Nusselt numbers for particlegas heat transfer in dense bubbling beds for Prg 0:7.
for
0:1 Rep 50
12
Nup
0:33
1:01Re0:48
p Prg
for
50 < Rep 10
13
It should be noted that the Prandtl number term in
these two equations is somewhat speculative, since
the experimental data did not cover a sucient range
for good correlation. Equations (12) and (13) include
Pr to the 0.33 power, based on the expectation that
dependence is similar to that for single spheres, as
shown in Eq. (11). In applying heat transfer coecients
calculated by Eqs. (12) and (13), it is important to use a
model that considers the particles to be well mixed and
the gas to be in plug ow, in order to be consistent with
the denition of hp as discussed above.
2.2.2
BedSurface Transfer
q
aw Tb Tw
14
where aw is the submerged surface area, Tw is the temperature of the submerged surface, and Tb is the temperature of the particle/gas medium in the bed. For
this analysis, any small temperature dierence between
the particles and the gas is neglected.
Due to its engineering importance, this bed-to-surface heat transfer coecient has been measured by
many investigators for various geometries and operating conditions (see Mickley and Trilling, 1949; Wender
and Cooper, 1958; Vreedenberg, 1960; Botterill and
Desai, 1972; Chen, 1976; Ozkaynak and Chen, 1980).
Typical characteristics of hw are:
hw is several times greater than the heat transfer
coecient for single-phase gas convection.
15
where
It is sometimes necessary to cool or heat dense bubbling uidized beds, and this is usually accomplished
by the insertion of heat transfer tubes carrying cooling
or heating uids into the bed. Heat transfer occurs
between the uidized particle/gas medium (often
referred to as the bed) and the submerged tube surfaces (often referred to as walls). For this situation,
one requires a heat transfer coecient based on the
surface area of the submerged wall:
hw
16
Figure 3 Heat transfer coecients for horizontal tube in bubbling beds of glass spheres, uidized by air at atmospheric pressure.
(Data of Chandran et al., 1980.)
hc dp
0:525Rep 0:75
kg
17
19
!0:44
Re0:44
D
20b
rs
Re 2050
rg p
22
23
where
mg
l
rs rg
!2=3
1 1=3
g
"
B1 1 33:3
B2 1
Ue rs cps
Umf gkg
Ue Ug Umf
#1
1=3
0:28e2s;mf Ue Umf
Umf
B3 1 0:05
Ue
24
25
Ue
rg
rs rg
!1=2
rs cps
gkg
2=3
26
27
28
1=2
30
31
hpa
1=2
32
33
where
Fo Fourier modulus
kpab tpa
cps rs 1 epab dp2
33a
a2 0:398 0:049w
a3 0:022 0:003w
w lnkpab =kg
for
ks
1; 000
kg
35
1 dw
fb
36
37
where
Z
rs cps dp wp
6kg
38
39
40
qr
aw Tb Tw
42
1
1=eb 1=ew 1
44
46
where
I; J = forward and backward radiant uxes,
respectively
A = volumetric absorption coecient
S = volumetric scattering coecient
The one-dimensional transient energy equation completed the system of equations
kpa
@2 T
@T
AI J2AsT 4 rs cps 1 epa
2
@t
@y
47
hc hg
hcb hg
versus
U
Ug Umf
Ut Umf
49
where
hc = the convective coecient in freeboard
hg = the single-phase gas convective coecient at
velocity U
!1=2
D t rg U g
0:00128
mg
!
D t rg U g
50
mg
50
Figure 4 Normalized heat transfer coecients for horizontal tube in freeboard of bubbling beds. H measured from level of
collapsed (packed) bed. (Data of Biyikli et al., 1983.)
Copyright 2003 by Taylor & Francis Group LLC
51
3.1
Heat Transfer
Particle Gas Transfer
d p rg
U
mg t
52
BedSurface Transfer
Convective Heat Transfer. The simplest correlations for the convective heat transfer coefcient empirically relate hc to the solids concentration,
expressed in terms of the mixture density for the gas/
solid medium in the bed (rb ),
rb rs 1 eb rg eb rs es;b rg eb
53
Prg
54
kg
cpg rs
gdp
and the correlation of Divilio and Boyd (1993) is
hc 23:2rb 0:55
55
57
2:3P=Lb
g1 emf rs rg
60
61
62
In contrast to the above correlations for total convective coecient hc , some models estimate the separate heat transfer contributions from lean and dense
e 3:4
64
es;pab 0:57 1 1 s;m
0:57
Once the solids concentration in clusters is determined,
the thermal conductivity of clusters (kpa ) can be estimated by the expression of Chandran and Chen (1985)
as
kpa kpab C 2
where kpab is given by Eq. (34) with
epab 1 es;pab
65
66
tpa
rs d p
L
L
c c
Vpa Repa mg
!
68
0:6
1
1 fl 1 0:018
es;b
for
69
Variable
hc
hl
hd
kpab
es;pab
tpa
fl
Equation No.
16
63
67
34
64
68
69
70
EXAMPLES
Bubbling Fluidized Bed
Density, r (kg/m3 )
Heat capacity, cp (kJ/kg K)
Thermal conductivity, k (W/m K)
Viscosity, m (kg/m s)
Pr
Gas
Particles
1.18
1.00
.0262
1:85 105
0.71
2,480
0.753
0.890
Figure 6 Eective emissivities for fast uidized beds for particles of es 0:7. (Data of Han et al., 1992, 1996.)
4.1.1
Fluidization Regime
grg rs rg 3
dp
m2g
9:811:182480 1:18125 106 3
164
1:85 105 2
Remf 2
1 emf Remf
150
3
fs 2 emf 3
fs emf
Remf 2
1 0:42Remf
150
1:00:423
1:02 0:423
Remf 0:139
from which we obtain the minimum uidization
velocity by Eq. (2),
Umf Remf
mg
r g dp
1:85 105
0:139
1:18125 106
0:0175 m=s
The terminal velocity is obtained from Eq. (6b),
"
#5=7
dp8=5 rs rg
Ut 0:072g
3=5
r2=5
g mg
"
#5=7
125 106 8=5 2480 1:18
0:072 9:81
1:182=5 1:85 105 3=5
0:730 m=s
and the dimensionless velocity is, by denition, from
Eq. (7),
U
Ug Umf
0:160 0:0175
0:20
Ut Umf 0:730 0:0175
4.1.2
If the gas enters the bed 30K cooler than the temperature of the particles in the bed, how long a distance
must the gas travel within the bed before that dierence is reduced to 0.5K? At the supercial gas velocity
of Ug 0:16 m/s, take the average void fraction in the
bubbling bed to be eb 0:585, and the average solid
fraction to be es;b 0:415:
The heat transfer coecient between particle and
gas is given by Eq. (12) or Eq. (13), depending on
the particle Reynolds number, which is dened by
Eq. (10),
Rep
1:28
mg
1:85 105
kg
0:0262
0:33
Nup
0:0282Re1:4
p Prg
dp
125 106
0:0262
125 106
0:0282 1:281:4 0:710:33
7:46 W=m2 K
4.1.3
If a heat transfer tube of a 2.9 cm diameter is submerged horizontally within the uidized medium at
the above operating conditions, what would be the
surface heat transfer coecient estimated by various
empirical correlations?
Since the bed temperature is low, radiative heat
transfer may be neglected; thus only the convective
coecient needs to be calculated. For a horizontal
tube, the correlations of Vreedenberg (1958),
Borodulya (1991), and Molerus and Schweinzer
(1989) can be utilized. For Vreedenbergs correlation,
we rst need to calculate the quantity.
rs
Re
rg p
2480
1:28 2,690 2550
1:18
!!0:3
Re0:3
D
1:85 105 2
9:81 24802 125 106 3
!0:3
Re0:3
D
82:1 Re0:3
D
where
ap np hp
Tp T2
1:99 104 7:42
ln
30
5:23 103 m
0:5
ReD
296
mg
1:85 105
0:0262
kg
82:1 2960:3
82:1Re0:3
D
Dt
0:029
409 W=m2 K
The correlation of Borodulya et al. (1991), Eq. (29), is
also appropriate for this case:
hc dp
r
0:74Ar0:1 s
kg
rg
!0:14
0:46Rep Prg
es2=3
e
cps
cpg
0:24
e2=3
s
0:74 1640:1
2480
1:18
0:14
:24
0:753
0:4152=3
1
0:4152=3
0:585
hc l
0:125 0:58
0:0328
kg
0:0262
2:27
476 W=m2 K
dp
125 106
l
rs rg
2480 1:18
g
1
9:81
1=3
1:77 106 m
"
B1 1 33:3
Ue rs cps
Umf gkg
#1
1=3
"
B2 1
0:28e2s;mf Ue Umf
1 33:3
Ue
#1
1=3
rg
rs rg
0:143
!1=2
rs cps
gkg
2=3
hc 0:0328
4.1.4
kg
0:0262
0:0328
486 W=m2 K
l
1:77 106
dbm dbo
e0:3Lt =Db
where
1:60
0:125es;mf
hc l
0:165Pr1=3
g
kg B1 1 B2 kg =2cps mg
!1=3
rg
1
rs rg
B3
p 20:32 16 1:75
0:65
4
#0:4
19:0 cm
so that
db dbm
dbm dbo
19:0 0:575
19:0 0:30:19=0:203
0:3Lt =Db
e
e
5:09 cm 0:0509 m
The rise velocity of the bubble, relative to the gas/
particle emulsion, is
hc hpa 2C
1=2
0:115 2480 753 0:58 1=2
21
p 0:715
0:415 m=s
466 W=m2 K
0:287
ub 0:558
dw
The frequency of bubbles at moderate-to-high elevations in a bubbling bed is 2 per second. Since bubbles
are equally likely to divert to one or the other side of a
horizontal tube, the frequency of bubble contact by
any spot on the tube would be approximately 2/2, or
1 per second. Equation (36) then estimates the average
residence time of particle packets at the tube surface as
tpa
1 dw 1 0:285
0:715 s
fb
1:0
kpab tpa
cps rs 1 epab dp2
0:115 0:715
4860
0:753 24801 0:42125 106 2
4.1.5
Kn
0:00495
50:4150:58
0:013 m=s
6kg
6 0:0262
17:1
and taking a value of 2.6 for the empirical constant Cc ,
Eq. (37) gives the convective coecient as
kg
0:0262
hd
Z1 eb 1 eN=Cc Z
dp
125 106
17:1 0:415 1 e17:2=2:617:1
477 Wm2 K
hc hg
hcb hg
where
4.1.6
Model
Vreedenberg (1958)
Borodulya et al. (1991)
Molerus and Schweinzer (1989)
Packet model
Martin (1984)
Average deviation
hc calc0 d
(W=m2 K
Deviation
(%)
409
476
486
466
477
3:5
20:5
23:0
18:0
20:8
17.2
0:46
hg 8:29
kg
0:0262
8:29
7:49 W=m2 K
Dt
0:029
"
Density, r (kg/m3 )
Heat capacity, cp (kJ/kg K)
Thermal conductivity, k (W/m K)
Viscosity, m (kg/m s)
Pr
Sphericity, fs
Gas
Particles
1.18
1.00
0.0262
1:85 105
0.71
2,200
0.800
0.900
Ut 0:072g
#5=7
dp8=5 rs rg
3=5
r2:5
g mg
"
#5=7
0:67 m=s
0.95
grg rs rg 3
dp
m2g
9:811:182200 1:18125 106 3
145
1:85 105 2
Remf 2
1 emf Remf
150
3
fs 2 emf 3
fs emf
Remf 2
1 0:45Remf
150
0:950:453
0:952 0:453
Remf 0:144
from which we obtain the minimum uidization
velocity by Eq. (2),
Umf Remf
mg
rg d p
1:85 105
0:144
1:18125 106
0:0182 m=s
The terminal velocity is obtained from Eq. (6b),
Ug Umf
5:0 0:0182
7:63
Fluidization Regime
Ar
U
4.2.2
rb kg=m3 )
hc (W=m2 K
12
24
80
120
Gs 40 kg=m2 s
24
12
0:3
93:7 115 W=m2 K
Ar 144
Gs 60 kg=m2 s
for
12
Gs 40 kg=m2 s
for
0:55
Gs 60 kg=m2 K
Remf 0:145
Ut 0:671 m=s
Ret
5:35
mg
1:85 105
2:3P=Lb
g1 emf rs rg
2:3 235
0:0456
9:81 1 0:452200 1:18
F1 1 f4 Re2mf f1 f2 Remf Ret 189
1 emf f3 1 fAr
1 0:04564 0:1442
0:04561 0:04562 0:145
5:35 1891 0:450:04563
4.2.3
The model of Molerus and Wirth (1993, 1995) correlates the convective heat transfer coecient with bed
pressure drop. Calculate hc for the above case by this
model and compare with the data of Dou et al. (1993)
at Gs 60 kg=m2 s.
Since the heat transfer surface of Dou is 5.5 m
above the gas and particle inlets, acceleration and
wall friction eects would be negligible and bed
pressure drop related directly to bed density:
P
grb 9:81 24 235 Pa=m
Lb
for
Gs 60 kg=m2 s
1 0:0456145c
1:31
2
F2 21 f2 Remf fRet
2
21 0:04562 0:144 0:0456 5:35
0:258
Repa
21 f2 Remf fRet
2f 11 femf
Gs 40 kg=m2 s Gs 60 kg=m2 s
s
F
14 1
F2
120
115 (4%)
133 (11%)
mg
1:85 105
2:48
rg d p
1:18 125 106
0:311 m=s
For the case of interest, the length of heat transfer
surface (Lh ) is smaller than the cluster contact length
(Lc ), so we use the heater length in Eq. (68) to obtain
the cluster contact time,
tpa
Lh
0:032
0:103 s
Vpa 0:311
rb
24
:0110
rs 2200
e 3:4
0:57
es;pab 0:57 1 1 s;m
0:57
"
#
0:0110 3:4
0:0365
1 1
0:57
Next, Eqs. (35) and (34) calculate the thermal conductivity for clusters of this concentration,
0:25
ks
0:900 0:25
fk 0:305
0:305
0:126
kg
0:0262
1 epab
0:0262
kpab kg epab
fk 2kg =3ks
0:0365
1 0:0365
0:126 2 0:0262=3 0:90
0:0318 W=mK
Assuming that the Fourier number is suciently large
to take C 1:0 from Eq. (33), the coecient for dense
phase convection is then obtained from Eq. (67) as
kpa rs Cps es;pa 1=2
hd 2
ptpa
0:0318 2200 800 0:0365 1=2
2
p 0:103
159 W=m2 K
Equation (63) is used for an estimate of the lean phase
convective coecient,
!0:8
Db rg Ug
kg
ht 0:023
Prg 0:33
Db
mg
0:8
0:0262 0:15 1:18 5
0:023
0:710:33
0:15
1:85 105
19:9 W=m2 K
The weighting factor (fl ) is obtained from Eq. (69),
0:6
1
1 fl 1 0:018
0:731
0:0110
and the combined convective heat transfer coecient is
calculated by Eq. (16),
hc fl hl 1 fl hd 0:269 19:9 0:731 159
122 K=m2 K
which is within 2% of experimental value.
4.2.5
Density, r (kg/m3 )
Heat capacity, cp (kJ/kg K)
Thermal conductivity, k (W/m K)
Viscosity, m (kg/m s)
Pr
Sphericity, fs
Gas
Particles
0.307
1.17
0.0743
4:51 105
0.71
2,200
0.800
0.900
0.95
grg rs rg 3
dp
m2g
9:810:3072200 0:307125 106 3
4:51 105 2
636
Remf
Umf
1501 emf
1501 0:45
6:34 103
!
!
mg
4:51 105
Remf
0:00634
rg d p
0:307125 106
0:00745 m=s
"
Ut 0:072g
dp8=5 rs rg
#5=7
3=5
r2=5
g mg
"
#5=7
0:673 m=s
The dimensionless velocity and particle diameter are
then obtained from Eqs. (7) and (4),
U
Ug Umf
5:0 0:00745
7:50
Ut Umf 0:673 0:00745
rb
45
0:0205
rs 2200
Ce es;b es Db
eb 1 exp
dp
0:95 0:0205 0:7 0:5
1 exp
1:00
125 106
Assuming the bed wall to be oxidized steel, a reasonable value for wall emissivity is ew 0:8. The bedwall
combined emissivity is calculated by Eq. (44) as
ebw
1
1
0:80
1=eb 1=ew 1 1=1:00 1=0:8
Tb Tw
ebw sTb4
Tw4
1:22 10 W=m
Note that this high heat ux is representative of operations for FFB combustors. Note also that 61% of the
heat transfer is due to thermal radiation.
NOMENCLATURE
a
A
Ar
cp
C
=
=
=
=
=
Cc
Ce
CD
CR
=
=
=
db
dbm
dbo
dp
dp
D
e
=
=
=
=
=
=
=
f
fb
Fv
Fo
g
h
hc
hp
hw
H
I
J
k
Kn
l
Lb
Lc
Lh
Mg
Nu
Nup
P
Pr
q
r
R
Rb
ReD
Remf
Rep
Ret
S
t
T
U
Ue
U
Vpa
wp
y
Greek Symbols
dw
e
es
fs
m
r
rb
s
t
tpa
Subscripts
b
c
d
g
l
m
mf
p
pa
pab
w
r
s
t
=
=
=
=
=
=
=
=
=
=
=
=
=
=
bed, or bubble
convective
dense phase
gas
lean phase
local gasparticle medium
minimum uidization
particle
packet or cluster
bulk packet
wall or submerged surface
radiative
solid
terminal, or tube
REFERENCES
Andeen BR, Glicksman LR. Heat transfer to horizontal tube
in shallow uidized beds. National Heat Transfer
Conference, St Louis, MO, paper no. 76-HT-67, 1976.
Anderson BA, Lechner B. Local lateral distribution of heat
transfer on tube surface of membrane walls in CFB boiler.
In: Avidan AA, ed. Circulating Fluidized Bed Technology
IV. New York: AIChE, 1993, pp 311318.
Battacharya, Harrison D. Heat transfer in high temperature
uidized beds. Proc of Europ Conf Particle Tech, 1976, p.
K7.
Beaude F, Louge M. Similarity of radial proles of solid
volume fraction in a circulating uidized bed. In:
Fluidization VIII. New York: Engineering Foundation,
1995, pp 245253.
Berg BV, Baskakov AP. Investigation of local heat transfer
between a xed horizontal cylinder and a uidized bed.
Intern Chem Eng 14, no. 3:440443, 1974.
Biyikli S, Tuzla K, Chen JC. Heat transfer around a horizontal tube in freeboard region of uidized beds. AIChE J
29, no. 5:712716, 1983.
Biyikli S, Tuzla K, Chen JC. Freeboard heat transfer in hightemperature uidized beds. Powder Tech 53:187194,
1987.
Borodulya VA, Teplitsky YS, Sorokin AP, Markevich II,
Hassan AF, Yeryomenko TP. Heat transfer between a
11
Mass Transfer
Thomas C. Ho
Lamar University, Beaumont, Texas, U.S.A.
INTRODUCTION
2.1
dNA
i
kg;single Sex;single CA
1
CA
dt
where dNA =dt represents the transfer rate of A from
the particle surface to the gas stream, kg;single represents
the mass transfer coecient of the particle, Sex;single is
i
the exterior surface of the particle, CA
is the concentration of A at the gasparticle interphase, and CA is
the concentration of A in the gas stream. The mass
transfer coecient kg;single under this single-sphere condition has been well established, both theoretically and
experimentally (Froessling, 1938), as
Shsingle
kg;single dsph y
2 0:6Resph 0:5 Sc0:333
D
2
rg udsph
m
m
rg D
Shsingle
5
6
dNA
i
kg;bed Sex;particles CA
7
CA
dt
However, in this equation, dNA =dt represents the combined mass transfer rate from all the particles in the
segment, kg;bed is the average mass transfer coecient
of the particles, and Sex;particles is the total exterior
surfaces of all individual particles in the bed segment.
It is worth pointing out that the average mass transfer coecient for xed bed particles may be higher or
lower than that for single spheres at an identical supercial gas velocity. On one hand, the low xed bed
voidage may enhance the mass transfer coecient
owing to the creation of increased true gas velocities
at the gasparticle interphase for xed bed particles.
On the other hand, the actual particle surfaces involved
in the particlegas mass transfer are smaller for xed
bed particles than those for well dispersed particles
owing to particle blockage in the xed bed
arrangement. The use of total exterior surfaces of all
individual particles in Eq. 7, therefore, overestimates
the active surfaces involved in the mass transfer process. The net eect of this overestimation is to decrease
the average mass transfer coecient evaluated based
on Eq. (7).
There have been several reports on the average mass
transfer coecient for xed bed particles. In general,
for particle Reynolds numbers greater than 80
(relatively large particles), the average mass transfer
coecient for xed bed particles is found to be higher
than that estimated by Eq. (2) [or (5)] for well-dispersed single spheres. On the other hand, for particle
Reynolds numbers less than 80 (relatively smaller particles), the trend is reversed, i.e., the average mass
transfer coecient for xed bed particles is much smaller than that estimated by Eq. (2). It is apparent that
the increase in mass transfer coecient for relatively
large particles in the xed bed setup is due to the eect
kg;bed dp y
2 1:8Rep 0:5 Sc0:333
D
rg Udp
m
10
2
pdsph
fs
3
pdsph =6
6
dsph fs
11
61 ebed
fs dsph
12
The sphericity for dierent solids has been calculated and is given in Table 1. Note that the term
(fs dsph ) appearing in Eq (12) is called the eective particle diameter (deff ) and can be estimated according to
the following approximations:
Sphericity
s
Source
1.00
0.81
(a)
(a)
0.87
0.70
0.58
(a)
(a)
(a)
0.76
0.60
0.47
0.700.90
0.63
(a)
(a)
(a)
(b)
(c)
0.63
0.63
0.65
0.73
0.69
0.65
0.58
0.28
(e)
(e)
(d)
(d)
(d)
(d)
(e)
(d)
0.86
0.66
as high as 0.86
as low as 0.53
0.89
0.85
(e)
(e)
(f)
(f)
(f)
(d)
(a) From geometric considerations. (b) From Leva (1959). (c) From
Uchida and Fujita (1934). (d) From Carman (1937). (e) From Leva et
al. (1948, 1949). (f) From Brown et al. (1949).
Source: Kunii and Levenspiel, 1991.
deff fs dsph fs dp
13
14
15
particles is to conduct the pressure drop experiments to determine the eective particle size.
(c) For very small particles (< 40 mm), there are
two methods to determine the eective particle
size. One is to use the scanning of magnied
photographs of particles and the other is to
conduct sedimentation experiments of particles
in a known uid; the terminal velocity of these
particles will give the diameter of the equivalent sphere.
Method by Carmon (1941). The method proposed by Carmon (1941) for the total exterior areas
of xed bed particles is based on the equation
Sex;particles SLA
S Sa 1 X
17
and
X3
Sa 14
Kt 1 X2
#0:5
18
mLQ
A1 Pp
19
2.3
Sd
dp
20
16
where
"
r
(g=cm3 )
X
()
Sa
(cm2 =cm3 )
S
(cm2 =cm3 )
dp
(cm)
Sd
()
0.646
0.626
0.615
0.651
0.607
0.629
0.391
0.411
0.422
0.389
0.430
0.409
88.1
106.5
167.7
187.8
262.5
53.65
62.73
96.93
114.75
149.63
0.089
0.073
0.045
0.039
0.028
4.775
4.579
4.362
4.475
4.189
4.476
dp;m
(mm)
dt
(cm)
Lm
(cm)
s. naphthalene
Naphthalene
2101700
2.2
4.4
1.32.5
v. water
Silica gel
Alumina
Glass
Lead
Rape seed
Active carbon
Silica gel
Alumina
Celite
3601000
5.9
1015
7101980
10.2
0.39.2
882580
3.0
5dp
18003100
3.8
9.5
11.3
0.67.0
Process
Investigators
Gas
Air
H2
CO2
Air
Air
s. naphthalene
Air
H2
Air
a. CCl4
a. water
s. dichlorobenzene
v. water
v. nitrobenzene
v. n-decane
Kettenring et al.
(1950)
Chu et al.
(1953)
Richardson and Szekely
(1961)
Thodos et al.
(1961, 69, 72)
s.=sublimation, v.=vaporization, a.=adsorption.
Source: Kunii and Levenspiel, 1991.
2.3.2
The procedure for determining the total particle exterior surfaces in a bed segment of a uidized bed is
essentially identical to that in a xed bed, as described
previously, i.e., using Eqs. (10) through (15). The ebed
appearing in Eqs. (10) and (12), however, represents
bed voidage in a uidized bed (or ef ) in this application.
2.3.3
2.4
Another type of mass transfer encountered in a uidized bed involves the transfer between an immersed
active particle and a uidized bed. Fluidized bed combustors are a typical example where active coal particles burn in the matrix of chemically inert particles of
sand and ash which constitute 97 to 99% of the whole
bed mass. The mass transfer between the immersed
active particles and the uidizing gas is expected to
be dierent from that in a bed consisting of only active
particles, because the inert particles may serve as an
impermeable barrier for mass ow.
The subject of mass transfer involving isolated
active particles and the uidizing gas has been studied
by several groups of researchers. Hsiung and Thodos
(1977) measured the mass transfer coecients of a
batch of naphthalene particles uidized in a matrix
of inert particles similar in size and density.
Vandershuren and Delvossalle (1980) experimentally
studied the drying process of moist particles in a uidized bed of similar dry particles. Palchonok and
Tamarin (1983, 1985) measured the mass transfer coefcients for an isolated active particle of various sizes
(dp;a 3:530 mm) freely moving within a uidized
bed of large inert particles (dp;i 1:26.3 mm). Kok
et al. (1986) measured the particlegas mass transfer
26
kg;a dp y
2emf
D
27
to become
Sha 2emf 0:117Ar0:39 Sc0:33
where Ar is the Archimedes number, dened as
rp rg
Ar gdp3 2
n rg
28
29
Figure 4 Mass transfer coecients between immersed isolated active particles and a uidized bed of inert particles.
(Shaded area: range of particlegas mass transfer coecient
measured in a uidized bed of only active particles).
With the assumption that fresh gas enters the bed only
as bubbles, Eq. (7) can be rewritten as
dNA
i
kg;bed Sex;particles CA
CA;b
dt
31
32
dz
ub
37
38
where the terms Sex,particles and Vbubble in the bed
segment can be formulated respectively by
Sex;particles A dz1 ef a 0
or
Sex;particles i
dCA;b
kg;bed
CA CA;b
dt
Vbubble
34
and
Vbubble A dz d
40
0
35
dt
Vbubble
36
or
dp fs
Figure 7 Nomenclature used in developing the model of Kunii and Levenspiel. (From Kunii and Levenspiel, 1991.)
39
41
Vbubble segment
d
Inserting Eq. (41) into Eq. (38) yields
kg;bed 1 ef a 0 i
dCA;b
dCA;b
ub
CA CA;b
dt
dz
d
42
and upon combining this with Eq. (36), the equation
becomes
0
kg;bed 1 ef a
dC
dN=dt
i
CA
ub A;b
CA;b
dz
d
Vbubble
43
With the assumption that equilibrium is rapidly established between CA at the gasparticle interphase and its
i
surroundings, i.e., CA
CA;e CA;c , Eq. (43) can be
expressed as
kg;bed 1 ef a 0
dCA;b
dNA
ub Vbubble
dt
dz
d
d
K
1 ef a 0 GB
D
D
yfs dp2 d
6D1 ef
Kbc 4:5
Umf
D0:5 g0:25
5:85
db
db1:25
49
50
where
q
3p
U d2
4 mf b
51
0:25
g
0:975D
db
52
0:5
46
47
KGB
48
kbc
KGB Kbc
44
and
kg;bed 1 ef a 0
3.1.2
Shbed
yfs dp2 d
K
6D1 ef bc
53
The above equation provides a correlation for estimating Shbed (or kg;bed ) based on particle and uidization
properties, which then can be used to determine the
rate of mass transfer employing Eq. (7) based on the
homogeneous bed approach. In the application, the
following equations provided by Kunii and
Levenspiel (1991) are needed for calculating uidization properties appearing in Eq. (53)
d
U Umf
ub
54
ub U Umf ubr
0:5
ubr 0:711g db
1 ef 1 emf 1 d
55
56
57
3.1.3
3.1.4
6
k
Kbc
dp fs g;single
58
Shsingle D
dp y
59
60
Zd
1
1 a=m
6kg;single tp;mean
f s dp
64
61
62
65
where
Shbed
Shbed aRep b
and
g Sh
6D bc
1 ef b single
66
Table 4
Effect of gb Zd on KGB
db;m (m)
1
KGB;measured (s )
calculated, with gb Zd 0
KGB
calculated, with gb Zd 4 104
0.04
0.06
0.08
0.11
1116
5
918
3
814
2
711
1.6
13
12
12
11
KGC
dNA
dCA;c
Vcloud
kgc Sex;cloud CA;e CA;c
dt
dt
67
where Vcloud is the volume of gas in the bubblecloud
phase, kgc the cloud emulsion mass transfer coecient,
and Sex;cloud the cloud exterior surface area. By analogy
with the transfer from a drop of one immiscible liquid
rising through another, Partridge and Rowe (1966)
proposed the mass transfer coecient to be correlated
by the Sherwood number equation
Shc
kgc dc
2 0:69 Sc0:33 Re0:5
c
D
68
m
rg D
u R d c rg
m
69
3:9emf DShc
Vcloud 0:67
71
Umf
p
73
74
kgt
p
pdb
Interphase Mass Transfer Models and Coefcients for Ozone Exchange in an Air-Fluidized Two-Dimensional
Table 5
Bed
Model
Reference
Coecient
1. Penetration
Chavarie
(1973)
kgc
4Demf ub
dc
2. Cloud two lm
Walker
(1970)
kgc
p
0:93Demf Umf a 1
dc
kgc
s
4De2mf ub a 1
kgc
pdc a
0:33
0:26DRe0:6
c Sc
dc
kgt
equivalent
Calculated
kgt (m/s)
kgc
0.0051
2
k
3 gc
0.0017
"mf kgc
0.0020
kgc
0.0030
kgcQ
kgc Sb Q
0.0051
kgc kgs
0.0056
q=Sb
0.0160
Q q kgb Sb
kgc
6. Toei et al.
7. Murray through-ow,
Murray (1965)
no diusion
Empirical value (from Charvarie and Grace, 1976a)
4Demf ub
pdc
s
r
1:02emf
ub D 1 a 1
kgc
db a
a1
1 2emf =a 1
kgb 0
q Umf db w
0.0160
Umf
4Demf ub 0:5
4
pdb
75
6
k
db gt
1:5Umf
12 Demf ub 0:5
1:5
db
pdb
db
76
dNA
dCA;b
Vbubble
KGT Vbubble
dt
dt
77
CA;e CA;b
Note that this equation is essentially in the same form
as that of Eq. (30) proposed by Kunii and Levenspiel
(1991), i.e., KGT reported by Sit and Grace (1978) is
directly comparable to KGB described previously by
Kunii and Levenspiel (1991).
In a similar attempt, Hatano and Ishida (1986) studied the particlegas mass transfer coecient in a
three-dimensional uidized bed of nonadsorbing glass
beads with dp 0:18 mm using optical ber probes.
Tracer gas concentrations in and around rising single
bubbles was measured continuously by the penetrative
probes, while the bubble boundary and the zone with
the prominent particle movement were detected by
reective probes. The measurements provided necessary data for determining the interphase mass transfer
coecient. The following equation resulted from their
study:
kgt 0:127D0:33 g0:33
u e 0:22
b mf
U
78
4Demf ub;m 0:5
pdb;m
79
1:5
db;m
pdb;m
db;m
According to Yates (1983), Eqs. (80) and (81) would
seem to give the best available values of interphase
mass coecients in three-dimensional beds of nonadsorbing particles. Note again that KGT;m appearing in
Eq. (81) is directly comparable to KGB appearing in
Eq. (30) proposed by Kunii and Levenspiel (1991).
3.4
p
d2
4 or
83
4Demf 0:5
ptf
84
3.5
85
ub emf
Umf
86
hdp
2 1:8Rep 0:5 Pr0:333
kg
87
Cp m
kg
88
kg;bed dp y
2 1:8Rep 0:5 Sc0:333
D
for Rep > 80
89
The above discussion implies that the correlation equations derived for Nubed based on heat transfer experiments can also be converted to the Sherwood number
for estimating kg;particles .
rg udp
m
At ut 69 cm=s;
5
Rep;t
rg ut dp 0:00118690:028
12:66
m
0:00018
Kbc 4:5
43:63 s1
Combining Eqs. (54) through (57), i.e.,
d
U Umf
ub
ub U Umf ubr
ubr 0:711gdb 0:5
EXAMPLE CALCULATIONS
1 ef 1 emf 1 d
we obtain
d
U Umf
1 ef ubr 1 emf
U 1:21
0:711980 0:370:5 1 0:5
U 1:21
6:77
Also
Solution.
Shbed
f dp2 y
d
K
gb Shsingle s
1 ef
6D bc
fs dp2 y 0:40:0282 1
8:04 104
6 0:065
6D
Shbed
fs dp2 y
d
gb Shsingle
K
1 ef
6D bc
U 1:21
0:0054:84 0:00080443:63
6:77
0:0088U 0:011
Then, since
Rep
rg Udp 0:028U0:00118
0:184U
0:00018
m
6.2
Data.
0:37 cm.
NOMENCLATURE
A
Ar
a0
CA
i
CA
CA;b
CA;c
CA;e
CAs
Cp
d
db
db;m
dc
deff
dor
dp
dp
dp;a
dpi
dp;m
dsph
dt
g
h
h
Kbc
Kbe0
KGB
KGC
KGT
KGT;m
Kt
kbc
kbe
kbe1
kg
kg;a
kg;bed
kg;single
kgc
kgt
kgt;m
kje
X
y
z
a
a0
gb
D
d
ebed
ef
emf
ep
m
n
rg
rp
Zd
fs
p
REFERENCES
Agarwal PK, La Nauze RD. Transfer processes local to the
coal particle: a review of drying, devolatilization and mass
transfer in uidized bed combustion. Chem Eng Res Des
67:457480, 1989.
Avedesian MM. Combustion of char in uidized beds. PhD
dissertation: University of Cambridge, 1972.
Bauer W, Werther J. Proceedings of 2nd World Congress
Chemical Engineering, 1981, p 69.
Behie LA, Kehoe P. The grid region in a uidized bed reactor. AIChE J 19:10701072, 1973.
Behie LA, Bergougnou MA, Baker CGJ. Mass transfer from
a grid jet in a large gas uidized bed. In: Keairns DL, ed.
Fluidization Technology, Vol 1. Washington DC:
Hemisphere, 1976, pp 261278.
Bohle H, Van Swaaij WPM. In: Davidson JF, Keairns DL,
eds. Fluidization. London: Cambridge University Press,
1978, p 167.
Brown GG. Unit Operations. New York: John Wiley, 1950.
Carmon PC. Fluid ow through granular beds. Trans Inst
Chem Eng 15:150166, 1937.
Carmon PC. Symposium on New Methods for Particle Size
Determination in the Subsieve Range. Amer Soc Testing
Materials, 1941, p 24.
Chavarie C. Chemical reaction and interphase mass-transfer
in gas uidized beds. PhD dissertation, McGill University,
1973.
Chavarie C, Grace JR. Interphase mass transfer in a uidized
bed. In: Keairns DL, ed. Fluidization Technology, Vol 1.
Washington DC: Hemisphere, 1976a, pp 251259.
Chavarie C, Grace JR. Interphase mass transfer in a gasuidized bed. Chem Eng Sci 31:741749, 1976b.
Chiba T, Kobayashi H. Gas exchange between the bubble
and emulsion phases in gassolid uidized beds. Chem
Eng Sci 25:13751385, 1970.
Chu JC, Kalil J, Wetteroth WA. Mass transfer in a uidized
bed. Chem Eng Prog 49:141149, 1953.
Cobbinah S, Laguerie C, Gilbert H. Simultaneous heat and
mass transfer between a uidized bed of ne particles and
immersed coarse porous particles. Int Heat and Mass
Transfer 30:395400, 1987.
Cooke MJ, Harris W, Highley J, Williams DF. Int Chem
Engrs Symp Ser 30:21, 1968.
Davidson JF, Harrison D. Fluidised Particles. London:
Cambridge University Press, 1963.
Froessling N. The evaporation of falling drops. Gerland
Beitr Geophys 52:170, 1938.
Gupalo YP, Ryazantsez YS, Sergeez YA. In: Davidson JF,
Keairns DL, eds. Fluidization. London: Cambridge
University Press, 1978, p 162.
Hatano H, Ishida M. In: Ostergaard K, Soorensen A, eds.
Fluidization V, Elsionore, Denmark. New York:
Engineering Foundation, 1986, pp 119127.
Hemati M, Mourad M, Steinmetz D, Laguerie C.
Continuous and intermittent drying of maize in a otation
uidized bed. In: Potter OE, Nicklin DJ, eds. Fluidization
VII, Brisbane, Australia. New York: Engineering
Foundation, 1992, pp 831840.
Hsiung TH, Thodos G. Mass transfer in gas-uidized beds:
measurement of actual driving forces. Chem Eng Sci
32:581592, 1977.
Hurt DM. Principles of reactor design: gas-solid interface
reactions. Ind Eng Chem 35:522528, 1943.
Kettenring KN, Mandereld EL, Smith JM. Heat and mass
transfer in uidized systems. Chem Eng Prog 46:139145,
1950.
Kok J, Stark NL, van Swaaij WPM. In: Ostergaard K,
Soorensen A, eds. Fluidization V, Elsionore, Denmark.
New York: Engineering Foundation, 1986, pp 433441.
Krell L, Kunne HJ, Morl L. Flow regimes and heat and mass
transfer in gas-uidized beds of solids. Internat Chem Eng
30:4556, 1990.
Kunii D, Levenspiel O. Fluidization Engineering. New York:
John Wiley, 1969.
Kunii D, Levenspiel O. Fluidization Engineering, 2nd ed.
Boston: Butterworth-Heinemann, 1991.
Leva M. Fluidization. New York: McGraw-Hill, 1959.
Leva M, Grummer M, Weintraub M, Pollchick M.
Introduction to uidization. Chem Eng Prog 44:511
520, 1948.
Leva M, Grummer M, Weintraub M, Storch HH. A study of
uidization of an iron schertropsch catalyst. Chem Eng
Prog 44:707716, 1948.
Leva M, Weintraub M, Grummer M, Pollchick M.
Fluidization of an anthracite coal. Ind Eng Chem
41:12061212, 1949.
12
General Approaches to Reactor Design
Peijun Jiang, Fei Wei,* and Liang-Shih Fan
The Ohio State University, Columbus, Ohio, U.S.A.
INTRODUCTION
Chemical syntheses
Metallurgical and
mineral processes
Other
Heat exchange,
catalyst cooler
Phthalic anhydride
synthesis
Uranium processing
Coal combustion
Solids blending
Propylene ammoxidation
to acrylonitrile
Reduction of iron
oxide
Coal gasication
Particle coating
Maleic anhydride
synthesis
Fluidized catalytic
cracking
Drying (PVC,
MBS, ABS)
Ethylene dichloride
synthesis
Roasting of sulde
ores
Incineration of
solids waste
Adsorption
Methanol to gasoline
and olen processes
Crystalline silicon
production
Cement clinker
production
Granulation,
drying of yeast
Syngas to gasoline
Titanium dioxide
production
Microorganism
cultivation
Binding
Dimethylbenzene (m)
ammoxidation to
dinitrilebenzene (m)
Calcination of
AlOH3
Particle separation
Biomass pyrolysis
Figure 2 Variations of uidized bed reactor congurations. (a) Baed bed, (b) multi-stage bed, (c) multi-stage with countercurrent solid ow, (d) multi-stage with external solid circulation, (e) uidized bed with online solid exchange, (f) multi-cell
bubbling bed, (g) CFB, (h) centrifugal uidized bed, (i) downer, (j) spouted bed.
1.2
Many factors aect optimum uidized bed reactor performance, including hydrodynamics, heat and mass
transfer of interparticles and intraparticles, and complexities of reaction kinetics. The design of uidized
bed reactor processes follows the general approach
for multiphase reactor processes. Krishna (1994) and
Jazayeri (1995) outlined the general procedure for this
process development. The design of the processes can
be described by considering various factors as illustrated in Fig. 3.
The uidized bed reactor design requires
understanding the reaction chemistry. The essential
knowledge for a design engineer may include the
reaction kinetics, conversion or yield, and selectivity,
thermodynamics, and process parameters (e.g., operating temperature and pressure as well as heat of reaction) aecting the reaction.
Information on intrinsic kinetics is an essential element in the analysis of reactor performance. Kinetic
data should be obtained over a range of temperatures
spanning the entire operation conditions. It is also
essential to obtain the kinetic data under much higher
conversion conditions than those anticipated from the
uidized bed reactor, due to the two-phase ow nature
(see Chapter 3), as most reaction occurs in the emulsion/dense phase where the gas reactant is highly
converted. Often the reaction kinetics employed is
Figure 3 Schematic diagram of uidized bed reactor developments. (From Krishna, 1994; Jazayeri, 1995.)
to maleic anhydride in a circulating uidized bed reactor (Contractor et al., 1987) have an attrition problem
that can be treated by spray-drying active catalyst
microspheres together with a polysilicic acid hydrogel
under conditions that allow the silica (SiC or SiO2 ) to
migrate to the outer surface and form a strong porous
layer on the outer surface (Bergna, 1988). This porous
shell allows the reactant and product molecules to diffuse in and out of the catalyst particles without signicantly aecting the maleic anhydride selectivity.
For catalytic reactions, most active sites of the
catalysts are dispersed on the surface of supported
particles. Once appropriate active compositions
are identied, catalyst development can focus on
the selection of support materials. Most reactions in
gas-uidized bed reactors are carried out under hightemperature conditions to overcome high activation
energies and ensure the presence of a gas phase.
Similarly, elevated pressures are typically utilized to
enhance the concentration of reactants on solid surfaces. The catalyst supports should be chemically and
physically stable under high temperature, because the
structure must remain unchanged for a long period of
time. Supports provide an enhanced thermal and
mechanical stability to the catalyst material as well as
reduce the usage of expensive metals. Selection of inert
support can signicantly aect the uidized bed operation. Inorganic materials generally satisfy the rigid
constraint. Only inorganic lattices have been found
to possess a static and stable three-dimensional backbone capable of sustaining a porous structure under
severe reaction conditions. The commonly used
catalysts in uidized bed reactors include zeolites,
amorphous silica, and alumina or other metal oxides.
Aluminum oxides are commercially available in dierent forms. The major phase is g-alumina, which has
small pores, a surface area up to 200 m2 =g and surface
hydroxyl groups. Amorphous silica can be prepared
with high surface area (up to 500 m2 =g). Zeolites are
aluminosilicates of well-dened crystal structure and
regular pore size so they can be engineered to catalyze
reactions with shape selectivity. Each of these materials
can serve as either a catalyst or a support. The activity
and selectivity of catalysts are dictated by the nature of
active sites and may be modied by the presence of
promoter or doping agents. For uidized bed reactors,
the supporting particles are tailored to achieve optimum hydrodynamics for a specic reaction. Catalyst
support particle sizes and pore structures are generally
chosen to ensure minimum intraparticle mass transfer
resistance. Another reason to avoid intraparticle
diusion resistance may result from selectivity
considerations. Minimizing internal transport resistance ensures that desired products rapidly diuse
out of the catalyst particle and avoid further undesirable reactions. For a consecutive reaction scheme, high
diusion resistance decreases intermediate compositions, while nal product conversion increases. The
specic surface area and pore structures are also very
important, and the eects are reaction specic. The
reaction of ammoxidation of propylene to acrylonitrile
preferably uses catalysts with specic surface areas
below 100 m2 =g and a pore radius greater than 50 A.
The catalysts are designed to provide limited oxygen
access to the reactant, and a higher surface area would
only enhance the extent of oxidation. Thus the formation of desired products while restricting further oxidation is possible. Conventional processes for catalytic
cracking of heavy hydrocarbon feedstock to gasoline
and distillate fractions typically use a large-pore molecular sieve, such as zeolite Y, as the primary cracking
component. Upon the addition of a medium-pore zeolite, such as ZSM-5, to the cracking catalyst composition, an increase in the octane number of the gasoline
fraction can be obtained. Conventional ZSM-5 cracking additives are implemented with a crystal size in
excess of 0.2 microns, since smaller crystal materials
reduce hydrothermal stability and hence rapidly lose
activity when exposed to the high-temperature steam
generated during FCC regeneration.
Particles in gassolid reactions in uidized bed reactors require appropriate preparation (e.g., grinding,
surface treatment) to achieve optimum eciency.
Coal must be ground to a particle size range suitable
for operation in a uidized bed combustor. Sorbet
powders are synthesized to obtain desirable pore structure. Limestone is extensively used in the in situ
removal of acid gas species, such as SO2 from uidized
bed coal combustors, and H2 S from advanced dualcycle gasication systems. Removal of SO2 involves
the injection or uidization of dry calcium-based powders in the high-temperature environment (800
1150 C) of the combustor, calcination of the sorbent
to produce CaO, and further reaction with SO2 to form
higher molecular volume CaSO4 .
CaCO3 s $ CaOs CO2 g
CaOs SO2 g
1
2O2 g $ CaSO4 s
Rep
Udp rg
mg
2
Figure 5 Flow regime diagram of uidized bed reactors. (a) Fixed bed, (b) bubbling uidized bed, (c) turbulent uidized bed, (d)
circulating uidized bed, and (e) dilute transport bed.
2.2.1
The bubbling regime is characterized by the coexistence of a bubble phase and a dense/emulsion phase
as shown in Fig. 5b. The majority of the uidizing gas
is present in the form of bubbles, and as a result, the
gas velocity through the dense phase is very low (on the
order of 0.01 m/s for Group A particles). The dense
phase occupies about 4060% of the bed, depending
on the particles and operating conditions. Figure 6
shows typical values for dense phase fraction, bubble
fraction, and overall voidage for FCC particles under
dierent gas velocities. Once formed, bubbles rise,
grow, coalesce, break up, and nally erupt on the
bed surface. Bubbles induce the drift eect on surrounding particles and entrain particles in the wake
region, thereby inducing vigorous local and global circulation of solid particles in the dense phase. Although
bubbles in uidized beds resemble those in gasliquid
systems, the interface of bubbles is permeable to gas
ow. The exchanges of gas between bubble and emulsion phases are accomplished by through-ow, bubble
coalescence and breakup, and diusion. A small fraction of particles may be entrained into the bubble
phase depending on the ne content and operating
conditions. The solid particles in the bubble phase
play an important role in the reactant conversion
when fast reactions and/or slow interphase exchange
are encountered.
The turbulent regime is often regarded as a transition regime between the bubbling and fast uidization
2g
U U
g gp g
#1=3
Figure 7 Generalized map of uidized regimes. (From Kunii and Levenspiel, 1997.)
Figure 8 Bubble congurations and ow patterns around a fast bubble and a slow bubble. (a) Fast bubble (Ub > Vd ), (b) slow
bubble (Ub < Vd ).
kq
1=2
Umf
4Dm emf Ub
3
pdb
4Ddw 4d2w 1 e w
Upon rearrangement of Eqs. (6) and (7), the ratio of
solid particles in the wall region to the overall particle
content can be expressed as
1 e w 4h1 h
1 2h2 1 e c
1
1 e
1 e
Vpc Gs =rp 1 e
Vpc Vpw
h
dw
D
Using the correlation of radial voidage proles developed by Zhang et al. (1991),
2:5
11
r
9
er e 0:191f 3f
f
R
diminishes the agitation or eddy turbulence in the lateral direction. The dense wall layer is of low turbulent
intensity. The strong turbulent actions on the interface
between core and wall regions are not sucient to
sweep accumulated particles away from the wall
under most operating conditions.
2.2.3
Solids Mixing
tp
10
11
Gas Mixing
Figure 13 Inuence of gas velocity on the axial gas dispersion in a turbulent uidized bed. (From Wei et al., 2000a.)
2.2.5
Heat Transfer
Figure 14 Heat transfer variations with gas velocity and solids concentration. (a) Dependence of heat transfer coecient on gas
velocity for three types of particles. (From Molerus, 1992), (b) dependence of heat transfer coecient on solids concentration in a
CFB. (From Shi et al., 1998.)
Table 2 Comparisons of Fluidization Properties of Bubbling, Turbulent, Fast Fluidization, and Dilute Transport Beds
with Group A Particles
Bubbling bed
Turbulent bed
Fast uidization
Flow nature
Bubbles and
emulsion phase
Dispersed dilute
phase and dense
phase
Dispersed dilute ow
with a thin particle
layer on the wall
Solids holdup
0.30.5, decrease
with gas velocity
Axial proles of
solids holdup
Radial proles
of solids fraction
Gas backmixing
(dispersion)
Fig. 16
Fig. 16
Core-annular ow,
particle accumulation
near wall and particle
clustering
Bottom dense region:
0.050.4,
Upper dilute region:
< 0:05, decrease with
gas velocity and increase
with solids ow rate
Fig. 16
Uniform
Large variations
High
Slightly dense in
wall region
Pea 0:10.4
Per 220
Bubble agitation
Well
Per 220
Very well
Very well
Gassolid contact
Temperature prole
Poor
Uniform
Heat transfer
Solids ow
Gas velocity
(throughput)
Residence time
distribution
Very high
Entrainment
Low
Well
May have slight
axial gradient
Very high
High entrainment
Higher
Long/wide
Long/wide
Dilute transport
Fig. 16
Well
Large axial gradient
Low
Solids circulation
Highest
Poor
Solids transport
Highest
Short/relatively wide
Short/narrow
Per 100200
Poor
Poor
undue thermal cracking. Furthermore, for short residence time reactors, the initial gas and solids ow
development is important in order to control the reaction selectivity and product distribution. As a result,
the downer distributor design has a signicant eect on
the gassolid ow structure in the downer, especially in
the entrance section. A good feed nozzle design that
provides excellent gassolid mixing and uniform
distributions of gas and solids over the entrance cross
section would enhance the gas and solid ow development.
Increasing gas velocity or gas throughput in a reactor eventually leads to a high production rate with
transition of ow regimes to turbulent or riser uidization. Some designs and modications have been implemented on bubbling bed reactors to operate as
turbulent bed reactors:
Process Requirements
Successful process developments rely on the understanding of process requirements and constraints.
Fluidized bed reactors have been used in many chemical processes. The typical reactions can be grouped as
catalytic reactions, gassolid reactions, noncatalytic
gas reactions, and polymerization. The following
describes process requirements for these systems.
2.3.1
Catalytic Reactions
Catalyst selection should be based on catalyst reactivity, reaction selectivity, and physical properties such as
particle size, density, and resistance to attrition. For
process development, heat and mass transfer phenomena together with reactivity and physical properties of
catalysts must be taken into account. The catalytic
process begins with gas reactant transferring to the
catalyst outer surface and subsequent intraparticle diffusion of the reactant through the pores of the catalyst.
Reactants then absorb onto the catalyst surface and
react to form product. These products desorb from
the surface, and, through intraparticle diusion, the
products exit from the pores and outer catalyst surface.
Consider the example of the ammoxidation of propylene to produce acrylonotrile over multicomponent
molybdenum/bismuth catalysts:
C3 H6 NH3 12 O2 ! CH2 CH CN 3H2 O
HR;298K 515 kJ=mol
The reaction network of propylene ammoxidation is
not fully understood, but a kinetic model can be
approximated by the network shown in Fig. 18. For
decades, the Sohio acrylonitrile process has utilized
turbulent uidized beds for propylene ammoxidation
GasSolid Reactions
a
P1;0
A M k!
p
Pi;b M k!
Pi1;b
kb
Pi;b Rm;n ! Pim;bn1
c
Pi;b CTA k!
Ri;b A
k
Pi;b ! Ri;b A
d
Pi;b k!
Ri;b
where M is monomer, P is the living (growing) polymer, R is terminated polymer, CTA is chain transfer
agent, and A is catalyst. Typical consumption for modern catalysts is on the order of 10,00050,000 gram of
polymer per gram of catalyst. The polymerization reaction is exothermic on the order of 20 cal/mol.
The UNIPOL process, developed by Union Carbide
in 1968, is a gas phase uidized bed polymerization
process and initially developed to make high-density
polyethylene with Ziegler-Natta catalysts. Along with
the development of these catalysts, enhancements to
process eciency and extension of the method scheme
to produce linear low-density polyethylene, polypropylene, elastomers, and other ethylene-a olen
copolymers have been developed (Burdett et al.,
1998). A ow diagram of the UNIPOL process is
shown in Fig. 19. Four primary operations make up
this process: monomer purication, reaction, resin
degassing, and resin pelleting. The operating gas velocity is relatively high ( 0:6 m=s) with a conversion of
only about 2% (< 5%) per pass. Upon exiting the
reactor, the unreacted/inert gas mixture is cooled, compressed, and then recycled back into the reactor until
the overall conversion reaches nearly 100%.
Polyethylene polymerization is typically carried out
at a pressure of 2030 atm and a temperature of 75
105 C. In order to avoid agglomeration and sheeting,
the operating temperature is limited by the particle
softening temperature. The uidized bed reactor starts
with an initial charge of polymer particles and establishes steady operation by supplying inert gas or a
mixture of inert gas and monomers. After the uidized
bed reaches a certain temperature, catalyst particles are
injected into the bed continuously and initiate poly-
Figure 19
Schematic diagram of uidized bed reactor for gas-phase polymerization. (From Burdett et al., 1998.)
merization to form a broad size distribution of polymer particles. In polymerization, reactant diuses from
the gas phase, through the boundary layer surrounding
the particles, and into the catalyst pores to the active
sites where polymerization through a coordinationinsertion mechanism occurs to form a solid polymer.
The catalyst particles are composed of small metal
fragments and explode into a large number of smaller
fragments during the reaction. In the early stage of
polymerization, these smaller particles are quickly
encapsulated by polymer molecules and grow to large
polymer particles. For polyethylene, the reaction
begins with small particles (20200 mm in diameter)
and nishes, 35 hours later, with a mean particle diameter of 2002000 mm. From gas-phase monomer to
solid-state polymer, ethylene experiences a dramatic
physicochemical transition. Similarly, the particle
morphology undergoes a dramatic change, and the
developmental stages are shown in Fig. 20.
Implementation of uidized beds for gas-phase polymerization also provides a typical example in which
particles grow in size during a short reaction time.
Key factors to uidized bed polymerization processes
are localized/overall temperature control and handling
of sticky particles due to process conditions operated
near the resin melting temperature. Without proper
cooling, the reactor temperature would increase until
the catalysts become inactive or particles fuse to the
reactor wall. Often, the production rate of the gasphase polymerization in a uidized bed reactor is limited by the maximum heat removal rate. The preferred
methods for removing heat from a uidized bed reactor
involving gas-phase polymerization are
1. Recycling unreacted gas through a heat exchanger before returning the gas to the uidized bed aids in
cooling the reactor. The heat removal rate depends on
the external heat exchanger capacity and gas ow rate.
In order to maintain this cooling sink, monomer consumed in the polymerization reaction is replaced by
adding excess monomer gas to the recycle stream.
2. Liquid, with a boiling point lower than the reaction temperature, is intermittently injected into the bed
with a spray pattern and immediately evaporates,
thereby removing heat. The injected liquid can be
monomer to be polymerized or inert liquid hydrocarbon.
3. Fluidized bed reactors are connected in series as
a staging operation. The uidizing gas from the prior
reactor is cooled o before entering the next uidized
bed.
4. Part of the gaseous stream leaving the top of the
reactor may be condensed by means of an external heat
exchanger and then reinjected into the uidized bed in
liquid form. The boiling point of condensable liquid
has to be lower than the operating temperature of
polymerization. Thus the reaction heat can be removed
Figure 20 Morphology developments and particle growth of polyethylene particles during gas-phase polymerization. (From Xie
et al., 1994).
particle surface. This would create a layer of ne particles on the granular surface and reduce the capillary
forces of the granules.
Fluidization quality in a condensing mode of operation can be monitored by pressure drop measurements.
As shown in Fig. 22, the uidized bed with dry polyethylene powders exhibits a typical pressure drop curve
of a normal uidized bed. With moderate liquid contents, the pressure drop decreases with an increase in
gas velocity before the bed reaches the uidized state
due to partial channeling. In the uidization state, the
presence of large agglomerates and intermittent collapse of local channeling cause large uctuations in
pressure drops. The pressure drop curve of dry
particles can be used as a baseline for monitoring uidization quality. Pressure drop can also be coupled
with bed expansion to monitor further the uidization
quality. During uidization, the bed expands, and a
low bed expansion is a characteristic of high moisture
contents in the gas stream (Fig. 23). The low bed
expansion, however, may also reveal the existence of
dead zones, channeling, and/or deuidization.
Evaporative liquid injection is a common practice in
the feed nozzle area of the FCC riser. Similar uidization phenomena can be observed in the uidized bed
processes like particle drying and particle gradulation
(see Sec. 2.3.4). Proper atomization of injected liquid
and dispersion of the ow to enhance mixing is the key
to smooth uidized bed operation with high liquid
injection rates and low liquid vaporization time.
2.3.4
Physical Operations
Figure 21 Particle ow patterns under various liquid contents in a uidized bed reactor. (a) Dry particle uidization, (b) particle
agglomeration, (c) agglomerate segregation, (d) bottom channeling, (e) whole bed channeling, (f) paste or slurry bed.
Figure 22 Eects of moisture content on the overall pressure drop in a uidized bed.
coated particles may collide before the solution completely dries and cause particles to agglomerate. This
can be minimized by using high uidization velocity
and large particles. However, at lower coating ow
rates and higher uidizing velocities, the lm on particles may undergo an onion ring growth pattern. By
taking into account the hydrodynamics and coating
kinetics, optimized conditions can be achieved.
Fluidized beds are often implemented in industry
for drying granular materials. For each individual particle, the drying process involves moisture migration
from the inner core regions to the particle surface.
The surface moisture is then vaporized or evaporated
into the uidizing medium. Meanwhile, heat has to be
transferred from the uidizing medium to the inside of
each particle. Diusion, capillary ow, or internal
pressure increase by vaporization or decrease by drying
could be mechanisms of moisture migration. For drying, in the constant drying rate period, the moisture on
a particle surface is over saturated, and the drying rate
is controlled by uidization characteristics. In the
varying drying rate period, the heat conduction and
moisture diusion to the particle surface are rate-controlling factors. The drying course varies with particle
temperature and heating rate, since drying is often
accompanied by particle shrinkage and deformation.
Optimized uidized bed dryer designs must emphasize
the dierence in drying kinetics. The strategy is to
enhance particle mixing and heat transfer in the constant rate drying period. When materials with a high
internal resistance against moisture migration are
dried, the proper heating rate and system temperature
are essential. Heat can be supplied by the uidization
gas in a uidized bed dryer, but the gas ow need not
be the only source. Heat may be eectively introduced
by heating surfaces such as panels or tubes immersed
in the uidized bed. The system temperature can therefore be adjusted by controlling the external heating
source. Uniform temperature distribution is another
requirement for heat-sensitive materials to avoid
overheating. Special attention to the requirements of
uidized bed design must be made when high moisture
or sticky materials are involved. Using stirred uidized
beds and mixing dried material with fresh feeds are
eective ways to prevent overheating and sticking.
Figure 24
dCba
kq ab eb CdA CbA 0
dz
at z 0
15
14
As discussed previously, the addition of ne particles is commonly practiced in the operation of uidized
bed reactors. In some cases, the addition of ne
particles is necessary in order to provide smooth bed
operations. In systems with ne particles, the assumption of a solids-free bubble phase is no longer valid.
Thus the two-phase model needs to be modied to
account for the ne particle presence and hence reaction in the bubble phase. The mass balance of species
A in the bubble phase can now be expressed by
Ub
dCbA
kq ab eb CdA CbA eb ebs rA 0
dz
19
The mass balance equation for species A in the emulsion phase remains essentially the same as Eq. (16).
Grace (1986b) further simplied the model by assuming that no net gas ows through the emulsion phase,
i.e., Ud 0. Under this assumption, Eqs. (16) and (19)
can be solved analytically for simple reaction kinetics.
Eects of ne particle solids concentration in the
bubble phase on the reactor conversion can also be
examined analytically.
Two-phase models could not be directly applied to
account for the entrance eects of the reactor. For
shallow beds or reactors with strong jetting, the distributor eects dominate the hydrodynamic behavior of
the bed and thus have to be considered dierently from
that of the in-bed region. The two-phase model that
applies to the in-bed region could be modied to
describe the entrance eects, as conceptually depicted
in Fig. 25b. Detailed descriptions of distributor eects
on hydrodynamics are given in Chapter 6.
3.2
d1 eb 1 eds Vd CdA d
dz
dz
dCdA
1 eb 1 eds Dax
dz
kq ab eb CdA CbA 1 eb eds rA 0
21
When the ow properties remain unchanged along
the axial direction in the dense phase, Eq. (21) can be
rearranged as
d 2 CdA Vd dCdA kq ab
eb
C
Dax dz
Dax 1 eb 1 eds dA
dz2
eds
rA
1 eds Dax
kq a b
eb
C 0
Dax 1 eb 1 eds bA
22
23
Q AU
ng Rm T
P
24
r1
k1 Cbta
1 KB Cbta =CO2 KMA Cma =CO2
30
r2
k2 Cbta
1 KB Cbta =CO2 KMA Cma =CO2
31
r3
k3 Cma
1 KB Cbta =CO2 KMA Cma =CO2
32
25
where k1 , k2 , k3 , KB , and KMA are rate constant parameters. The underlying assumptions made in the model
equations are as follows:
26
at
z0
28
Table 3 Correlations Used for Model Simulation of Butane Oxidation in a Turbulent Fluidized Bed
Reactor
Parameters
Correlations
Sources
mg
rp dp g
Ud 0:5711 eds 3
0
dp rg
rg
This chapter
Ud
Ud2 85:7eds
Vd
Vd Ud =1 eds
This chapter
Vb
Vb eb U Ud U Vd 1 eb 1 eds
This chapter
Uc
#0:27
!
0:20 "
rg20 rp rg D
Uc
m20
0:211 2:42 103
p
m
rg
r g dp
D0:27
D1:27
gdp
!0:132
dp 0:02
U1
0:055 rp rg
Ar
exp0:157F45
e 3:0
U2
rg
D
eb
eb
emb
eds
e eds 1
eds
emb
U
Uc
0:3
U
ebs 0:07
exp0:25F45
Ut
efs
Wtf
Wtf
Arp
3
ebs
This chapter
This chapter
ebf 1 e
Hf
This chapter
es 0
assume
1 expaH Hf
a
Hf
Wt rp A1 eHf Wtf
kq ab
Experiments
Dax
Experiments
Figure 28 Eects of ne particle content on the butane conversion in a turbulent uidized bed reactor. (From Jiang and
Fan, 1999.)
1 erA
reDr
@r
r @r
33
r
2
@z
@r
Pea @z
Per r @r
eVgz A
34
Figure 27 Model predictions of concentration in a turbulent
uidized bed. (From Jiang and Fan, 1999.)
CA0
H dCA
z0 Pe
dz z0
a
CA
at
z0
35
@CA
0
@r
@CA
0
@z
at
r0
at
zH
and
D
2
36
37
40
Z0 500 exp69e ea
41
42
Experimental data for gas velocity proles in high-density risers are very limited. Radial proles of the gas
1 exp20r=D 10q0
U
44
2
1 exp1:71491 e 1:263
45
the limiting factor to the reaction. Nonuniform distributions of catalyst in a radial direction develop as the
particle density of the riser increases. The low turbulent intensity and low gas velocity in the wall region
result in poor mass transfer, and the low mass transfer
and high catalyst density yield a reactant-decient zone
in the wall region. Since the reaction network of propylene ammoxidation is represented by a combination
of parallel and series reactions with the intermediate
product being the desired product as shown in Fig. 18,
poor mass transfer favors complete oxidation instead
of acrylonitrile formation.
4
Figure 29 Axial proles of concentrations for propylene,
oxygen, and carbon dioxide in the process of ammoxidation
of propylene. (From Wei et al., 2000a.)
CONCLUSION
A
Ar
ab
m
ng
Mp
M1
=
=
=
=
P
P
Pea
=
=
=
Per
Pr
q
q0
=
=
=
Q
Rep
=
=
Res
R
r
=
=
=
ri
rA
Rm
rp
=
=
Sc
Sh
t
T
T1
Tp
tp
u
u
U
=
=
=
=
=
=
=
=
=
=
U
Ub
Uc
=
=
Ud
Umf
Ut
Vb
=
=
=
=
Vd
Vg
Vgz
Vgr
Vpc
=
=
=
=
reaction order
molar gas ow rate
monomer concentration at particle surface
monomer concentration in surrounding
medium
bed pressure
pressure drop
Peclet number based on axial dispersion
coecient ( UH=Dax )
Peclet number dened based on radial
dispersion coecient ( UD=Dr
Prandtl number ( n=a)
condensated vaporization rate
empirical constant dened in Eqs. (44) and
(45)
total gas volumetric ow rate
Reynolds number based on particle
diameter rg UDp =mg
Reynolds number dened in Eq. (39)
column radius
radial coordinate in cylindrical coordinate
system
reaction rate per unit volume of catalyst for
species i
reaction rate per unit volume of catalyst for
species A
gas constant
polymerization rate per unit volume of
particle
Schmidt number ( mg =rg Dm )
Sherwood number ( ks dp =Dm )
time
temperature
bulk (or surrounding medium) temperature
particle temperature
particle residence time
velocity
average velocity
supercial gas velocity eb Vb
1 eb Vd
nondimensional gas velocity
Ur2g =mg rp rg g1=3 Re=Ar1=3
supercial bubble rising velocity
transition velocity from bubbling to
turbulent regimes
supercial gas velocity in the dense phase
minimum uidization velocity
terminal velocity
interstitial (linear) gas velocity in the bubble
phase
interstitial (linear) gas velocity in the dense
phase
linear gas velocity
linear gas velocity in axial direction
linear gas velocity in the radial direction
particle velocity in the dilute core region
Vpw
Wt
Wtf
z
Z0
Zi
=
=
=
=
=
=
Greek Letters
a
dw
e
e
ea
eb
=
=
=
=
=
=
e c
e w
ed
=
=
=
eds
ebs
efs
emb
emf
e
l
mg
n
rp
rg
t 0 00
=
=
=
=
=
=
=
=
=
tp
thermal diusivity
thickness of the dense wall region
overall or local voidage
cross-sectional averaged voidage
empirical constant in Eq. (39)
volume fraction of the dilute (or bubble)
phase ( bubble volume/bed volume)
averaged voidage in the dilute core region
average voidage in the dense wall region
volume fraction of the dense (or emulsion)
phase ( 1 eb )
volume fraction of solid particles in the
dense phase (per unit volume of densefraction)
volume fraction of solid particles in the
dilute phase (per unit volume of dilute
phase
volume fraction of solid particles in the
free board region
voidage at incipient bubbling
voidage at incipient uidization
empirical constant in Eq. (40)
thermal conductivity
gas viscosity
kinematic viscosity
particle density
gas density
contact time dened by Kunii and
Levenspiel (1991) in Fig. 10
mean particle residence time
Subscripts
A
b
bta
d
in
ma
O2
out
s
REFERENCES
Avidan, AA, Shinnar R. Development of catalytic cracking
technology, a lesson in chemical reactor design. Ind Eng
Chem Res 29:931942, 1990.
Bai DR, Zhu JX, Jin Y, Yu ZQ. Internal recirculation ow
structure in vertical upward owing gassolid suspensions
I, a core/annular model. Powder Technol 85:171178,
1995.
Bergna HE. Method of making maleic anhydride. U.S.
Patent, 4,769,477, 1988.
Bolton LW, Davidson JF. Recirculation of particles in fast
uidized risers. In: Basu P, Large JF, eds. Circulating
Fluidized Bed Technology II. Toronto: Pergamon Press,
1988.
Buchanan JS, Sundaresan S. Kinetics of redox properties of
vanadium phosphate catalysts for butane oxidation.
Applied Catalysis 26:211226, 1986.
Burdett ID, Eisinger RS, Cai P, Lee KH. Recent developments in uidized bed process for olen polymerization.
AIChE Annual Meeting, Miami, FL, 1998.
Cai P, Chen SP, Jin Y, Yu ZQ, Wang ZW. Eect of operating temperature and pressure on the transition from bubbling to turbulent uidization. AIChE Symp Ser
85(270):3743, 1989.
Contractor RM, Sleight AW. Maleic anhydride from C-4
feedstock using uidized bed reactor. Catalyst Today
1:587607, 1987.
Davidson JF, Harrison D. Fluidized Particles. New York:
Cambridge University Press, 1963.
Fan LS, Zhu C. Principles of GasSolids Flows. New York:
Cambridge University Press, 1998.
Fan LS, Ghosh-Dastidar A, Mahuli S. Calcium carbonate
sorbent and methods of making and using same. US
Patent 5, 779,464, 1998.
Froessling N. The evaporation of falling drops. Gerlands
Beitr. Geophys 52:170, 1938.
Ghosh-Dastidar A, Mahuli S, Agnihotri R, Fan LS.
Investigation of high-reactivity calcium carbonate sorbent
for enhanced SO2 capture. Ind Eng Chem Res 35:598
606, 1996.
Gidaspow D. Multiphase Flow and Fluidization-Continuum
and Kinetic Theory Descriptions. San Diego: Academic
Press, 1994.
Grace JR. Contacting modes and behavior classication of
gassolid and other two-phase suspension. Can J Chem
Eng 64:353363, 1986a.
Grace JR. Fluidized beds as chemical reactors. In: Geldart
D, ed. Gas Fluidization Technology. New York: John
Wiley, 1986b, pp 285339.
Horio M, Morishita K, Tachibana O, Murata N. Solids distribution and movement in circulating uidized beds. In:
Basu P, Large JF, eds. Circulating Fluidized Bed
Technology II. Toronto: Pergamon Press, 1988.
13
Fluidized Bed Scaleup
Leon R. Glicksman
Massachusetts Institute of Technology, Cambridge, Massachusetts, U.S.A.
INTRODUCTION
Typically, the development of a new commercial uidized bed process involves a laboratory bench scale
unit, a larger pilot plant, and a still larger demonstration unit. Many of the important operating characteristics change between the dierent size units. The
critical problem of scale-up is how accurately to
account for the performance changes with plant size
to optimize the performance of a full-size commercial
unit. In addition, it would be helpful if the link between
smaller units could be understood, so that the smaller
unit could be used to improve the commercial plant
performance or solve existing problems.
One discouraging problem is the decrease in reactor
or combustor performance that is found when some
pilot plants are scaled up to larger commercial plants.
These problems can be related to poor gas ow patterns, undesirable solid mixing patterns, and physical
operating problems; see Matsen (1985). In most
instances, it is dicult to make any observations or
measurements directly within the commercial uidized
bed. Thus problem solving is hindered by a lack of
understanding of the underlying cause of poor behavior. In the Synthol CFB reactors constructed in South
Africa, rst scale-up from the pilot plant increased the
gas throughput by a factor of 500. Shingles and
McDonald (1988) describe the severe problems initially
encountered and their resolution.
In some scaled-up uidized bed combustors, the
lower combustion zone has been formed into a narrow
Figure 1 Apparent reaction-rate constant vs. reactor diameter and bed height. (From Frye et al., 1958.)
Figure 2 Solid diusivity in axial direction for large units. (From May, 1959.)
3.1
bubbles grew in size and moved toward the centerline, presumably by coalescence, higher up in the bed.
The transition to slugging occurred higher up in the
larger bed at the same supercial gas velocity. The
bubble velocity increased with height until slug ow
was approached, after which the velocity decreased.
For the only case shown, the 100 mm bed, the maximum velocity occurred when the bubble dimension
was about one third of the bed diameter. In larger
beds, the bubble rise velocity was higher for the same
bubble volume, Fig. 7. Hovmand and Davidson
(1971) reviewed data on bubble rise velocity and concluded that when the bubble diameter exceeded onethird of the bed diameter, the bed diameter rather
than the bubble diameter governed the bubble rise
velocity. Note that Werthers results at dierent
supercial velocities are well correlated by the drift
ux form,
p
ub u0 umf c gd
3.2
Mixing
Figure 8 Variation of main chord length with wall spacing. Im 76 cm, probe height 46 cm, X, 1.3 Umf ; *; 1:5Umf ; r;
1:8 Umf ; T denotes with tubes. (From Glicksman and McAndrews, 1985.)
The dierences in behavior between small laboratory beds and larger demonstration units can in part
be attributed to a switch from porous plate distributors
in the small bed to discrete hole or bubble caps in the
larger beds. The porous plates give a better quality of
uidization, e.g., smaller bubbles, for shallow beds and
beds of moderate depth; see Rowe and Stapleton
(1961).
Yerushalmi and Avidan (1985) suggest that the
axial dispersion coecient of solids in slugging and
turbulent ow varies approximately linearly with the
bed diameter, as with Thiel and Potter (1978).
3.3
Arena et al. (1988) measured the hydrodynamic behavior of two circulating uidized beds with riser diameters of 0.041 and 0.12 m ID of roughly the same
Figure 9 Visible bubble ow Qb , Imf 76 cm, probe height 46 cm. X, 1:3Umf ; *; 1:5Umf ; r; 1:8Umf ; T denotes with tubes.
Experiment uncertainty 20%. (From Glicksman and McAndrews, 1985.)
Rhodes et al. (1992) compared the solids ux proles across the cross sections of a 0.152 and a 0.305 m
diameter circulating bed riser. They found a region
where the solid prole, given by the ratio of local
ux to average ux, had a similar variation over the
cross section and was insensitive to the level of solid
ux. The variation of the local solids ux over the
radius was a function of the gas velocity and the riser
diameter. In the larger riser, the proles were somewhat atter and the thickness of the downowing
region relative to the bed radius was smaller. The comparisons were not exact since the cross sections compared for the two beds were at dierent heights.
Zhang et al. (1991) carried out investigations with
three dierent fast bed systems with diameters of 32,
90, and 300 mm. They found that the radial voidage
distribution, given as a ratio to the cross-sectional
average, was independent of bed diameter and solids
recycle rate. The similarity does not hold at transition
Hovmand and Davidson (1971) review Stewarts criterion for the transition from bubbling to slug ow,
u umf
0:2
5
0:35gD1=2
and show it gives good agreement with most experiments. When u umf is larger than the value found
from Eq. (5), the bed will be in slug bed. Thus a small
Figure 11 Empirical correlation and experimental data of thickness of downowing layer at the wall of a CFB as a function of
the equivalent bed diameter. (From Zhang et al., 1995.)
equations. If we can write the proper equations governing uid and particle dynamic behavior, we can
develop the proper scaling relationships even if we cannot solve the equations (at present we cannot). In
essence, if a model is designed that follows the exact
same equations and boundary conditions as the commercial bed, with the same values of the coecients,
then the model and the commercial bed should have
identical behavior. The trick is to nondimensionalize
all of the terms of the equations so that the model and
the commercial beds have identical nondimensionalized values of the parameters. The complete development of this is given in Glicksman et al. (1994) and
Glicksman (1997).
An alternate approach is to design a scale model so
that the ratio of all of the important forces is the same
in the model as it is in the full-scale bed, and a the scale
model uses geometry similar to that of the larger uidized bed. If the ratio of forces acting on a particle in
the scale model bed have the same ratio as they do for
a particle in the full bed, then the trajectory of the
particle motion should be the same for both beds.
The same follows for the motion of the gas and the
motion of clusters of particles or particles in a dense
phase. From this point of view, we must identify all the
important forces in the system. Table 1 lists these
forces.
To characterize the gas ow in the uidized bed,
important forces are viscous and inertia as well as particle-to-gas forces. For the particles, the important
forces include gravity, particle inertia, gas interaction
with the particles such as drag, collisional forces
between particles and between particles and wall, and
particle surface forces such as electrostatic and adhesion forces. For larger particles, in Geldart groups B or
D, the particle surface forces can be neglected. The
question is complicated for smaller particles because
the surface forces are dicult to quantify. We must
Gs =s U0
L=D
PSD
u20 s u0 dp f u0 L s Gs L
; ;
; ;
;
;
m
gL
m
f s u 0 D
For scaling to hold with the full set of scaling relationships, all of the dimensionless parameters given in Eqs.
(6) or (7) must be identical in the scale model and the
commercial bed under study. If the small scale model is
uidized with air at ambient conditions, then the uid
density and viscosity are xed and there is only one
unique modeling condition that will allow complete
similarity. In some cases, this requires a model that is
too large and unwieldy or will not permit simulation of
a very large bed.
In most situations, one would expect that not all of
the parameters are of rst order importance. By reducing the number of parameters that must be maintained in the model it will be possible to model larger
commercial beds with small scale models. This will
involve simplications of the interparticle drag at the
extreme of small and large Reynolds numbers based on
particle diameter. If the same simplication can be
shown to hold in both of these limits, it is reasonable
to consider application of the simplication over the
entire range of conditions.
By representing the particle to air drag force by the
Erqun equation or single particle drag, it can be
shown, Glicksman et al. (1994), that in the limit of
small particle Reynolds numbers the governing parameters can be reduced to
u20 s u0 L1 Gs
; ;
; ;
; bed geometry, ; PSD
g2 f umf L2 s u0
At the other extreme, at large particle Reynolds numbers it can be shown that the same set of governing
parameters, Eq. (8), applies.
Since the same simplied set of dimensionless parameters holds exactly at both high and low Reynolds
numbers, it is reasonable to expect that they hold, at
least approximately, over the entire range of conditions
for which the drag coecient can be determined by the
Erqun equation or an equation of similar form.
For the more general case, Fig. 13 shows the
value of the dimensionless drag coecient relative
to at low Re over a range of conditions when u0 /
umf is 10 and 3, respectively, and Fr and s remain
constant. When u0 /umf and the slip velocity are high
there is a larger variation of dimensionless drag
coecient with Reynolds number. Note that
does not vary with particle Reynolds number
when the Reynolds number remains above about
103 or below about 10. When u0 /umf is 1000, a
condition is approached with very ne particle bubbling beds or circulating beds. The use of the Erqun
relationship is questionable except for the dense
lower part of the bed.
5.1
Figure 13
Ratio of drag coecient to low Reynolds number drag coecient, u0 /umf 10 and 3. (From Glicksman et al., 1993b.)
Figure 14
Exact and simplied models of a pressurized uidized bed combustor. (From Glicksman et al., 1993b.)
Figure 15 Error in drag coecient for simplied scaling relationships, u0 /umf 10. (From Glicksman et al., 1993b.)
The errors in ut /umf is shown in Fig. 17 for simplied scale models at two dierent linear dimensions.
Scaling a combustor with comparatively small particles, 0.2 mm or less, gives good agreement for ut /umf
even at one-sixteenth linear scale, while for large particles a linear scale of one-fourth gives fair agreement
for ut /umf . Since u0 /umf is held constant in the simplied scaling laws, close agreement of ut /umf also results
in close agreement of ut /u0 .
6.1
Viscous Limit
10
Figure 16
Error in cluster drag coecient for xed u0 /umf using Cd for a solid sphere. (From Glicksman et al., 1993b.)
Figure 17
the model, along with the gas pressure and temperature, determines the values of f and m. The particle
density for the model is chosen to match the density
ratio, so that
f
f
s m
p c
Error in terminal velocity using simplied scaling at 1013 kPa and 800 C. (From Glicksman et al., 1994.)
11
12
mf
u0
f
f
f m 2=3
Dm
Dc
f c
13
14
f u 0 d p
mf
dp m
dp c
The particle diameters in the model scale by the same
factor as the bed diameter, by the ratio of the kinematic viscosities to the two-thirds power.
Equating the Froude number and rearranging, we
nd
1=3 1=2
u0m
Dm
fm
15
u0c
fc
Dc
Thus the velocity scales are the square root of the
linear dimension scale.
By satisfying both Eq. (13) and Eq. (15), the
Reynolds number and the Froude numbers are kept
identical between the model and the commercial bed.
Combining Gs =s U0 and the product of the
Reynolds and Froude numbers along with Eq. (13) it
can be shown that
1=3
fm
Gs = m
16
Gs =s c
fc
so that the ratio of solids ow to solids density scales as
the ratio of the cube root of the kinematic viscosity.
Once the model uid and its pressure and temperature are chosen, which sets the gas density and viscosity, there is only one unique set of parameters for the
model that gives similarity when using the full set of
dimensionless parameters. The dependent variables, as
nondimensionalized as X=D, y=D, u=u0 are the same in
the respective dimensionless time and spatial coordinates of the model as the commercial bed. The spatial
variables are nondimensionalized by the bed diameter,
so that the dimensional and spatial coordinates of the
model are proportional to the two-thirds power of the
kinematic viscosity, as given by Eq. (13),
2=3
xm
m
xc
c
17
19
fc
fm
Table 2 gives the values of design and operating parameters of a scale model uidized with air at ambient
conditions which simulates the dynamics of an atmospheric uidized bed combustor operating at 850 C.
Fortunately, the linear dimensions of the model are
much smaller, roughly one-quarter those of the combustor. The particle density in the model must be much
higher than the particle density in the combustor to
maintain a constant value of the gas-to-solid density
ratio. Note that the supercial velocity of the model
diers from that of the combustor along with the spatial and temporal variables.
When modeling a pressurized hot bed (Table 3), the
ambient temperature model uidized with air has
dimensions very close to those of the pressurized combustor. If another gas is used in the model, particularly
a gas with a higher density, the model can be made
much smaller than the pressurized combustor (Table
4). Care must be taken to select a safe modeling gas
and one that yields a solid density for the model that is
available.
7.2 Design of Scale Models Using the Simplied Set
of Scaling Relationships
The simplied scaling relationships, Eq. (8) oer some
exibility in the model design, since fewer parameters
must be matched than with the full set of scaling relationships. When the uidizing gas, and the pressure
and temperature of the scale model, are chosen, the
gas density and viscosity for the scale model is set.
The model must still be geometrically similar to the
commercial bed. There is still one free parameter.
Generally, this will be the linear scale of the model.
For the simplied scaling relationships, the gas-tosolid density ratio must be maintained constant,
f
f
20
s m
p c
Commercial bed
Temperature ( C)
Gas viscosity (105 kg=ms)
Density (kg=m3 )
Derived from scaling laws
Solid density
Bed diameter, length, etc.
Particle diameter
Supercial velocity
Volumetric solid ux
Time
Frequency
21
D
gD
m
so that
1=2
ucm
Dm
u0c
Dc
22
850
4.45
0.314
25
1.81
1.20
3.82sc
0.225Dc
0.225dpc
0.47u0c
0.47Gs =s c
0.47tc
2.13fc
sc
Dc
dpc
u0c
Gs =s c
tc
fc
Gs
s uo
Gs
s u o c
m
Gm
u0m
Dm
sm
fm
Gsc
sc
u0c
fc
Dc
u0
umf
u0
umf
Temperature ( C)
Gas viscosity (105 kg=ms)
Density (kg=m3 )
Pressure (bar)
Derived from scaling laws
Solid density
Bed diameter, length, etc.
Particle diameter
Supercial velocity
Volumetric solid ux
Time
Frequency
24
23
Scale model,
full scaling laws
Commercial bed
850
4.45
3.14
10
sc
Dc
dpc
u0c
Gs =s c
tc
fc
Scale model,
full scaling laws
25
1.81
1.20
1
0.382sc
1.05Dc
1.05dpc
1.01u0c
1.01Gs =s c
1.01tc
0.98fc
25
Commercial bed
Temperature ( C)
Gas viscosity (105 kg=ms)
Density (kg=m3 )
Pressure (bar)
Derived from scaling laws
Solid density
Bed diameter, length, etc.
Particle diameter
Supercial velocity
Volumetric solid ux
Time
Frequency
umf m umf c
ucm
u0c
umf c
Dm
Dc
1=2
26
fc mm
Dc
umf
27
dpm
dpc
28
s
f
29
2
dpc s =f c
Dc
u2mfc
u0c
30
850
4.45
3.14
10
sc
Dc
dpc
u0c
Gs =s c
tc
fc
Scale model,
full scaling laws
20
1.19
4.34
1
1.38sc
.334Dc
.334dpc
.58u0c
.58Gs =s c
.58tc
1.7fc
1=4
dpm
Dm
dpc
Dc
31
Table 5 Scale models of atmospheric commercial hot bed using the simplied
scaling relationship
Commercial beds
dp
umf
1/9
10 m
15
24
49
98
245
8 m
12
20
40
80
198
work was concerned with experimental testing and verication of the scaling relationships. Hot bed behavior
was compared to cold scaled models in several studies.
In others, cold beds of dierent sizes were compared to
each other. For bubbling beds there is the sense of
what constitutes sucient verication of the scaling
relationships. Since bubbles are the prime motive
agents for both gas and solids displacement, a detailed
verication should involve comparison of bubble properties through the large bed and the scale model. This
is preferable to comparison of overall performance
parameters for a bed based on input and exhaust measurements.
7.4
Table 6 Scale model of 10 atm commercial hot bed using the simplied scaling
relationships
Commercial beds
dp
umf
1/9
31 m
46
78
155
310
610
25 m
38
63
126
250
550
Commercial bed t 800 C, P 1 atm, s 2500 kg=m3 gas:air. Model bed s 896 kg=m3
gas:air at STP.
Nicastro and Glicksman (1984) experimentally veried the full set of scaling laws for bubbling uidized
beds. They compared the time-resolved dierential
pressure measurements from a bubbling uidized bed
combustor and a scaled cold model. Good agreement
was obtained between the measurements that were proportional to the bubble size and frequency. Figure 18
presents the comparisons. They concluded that hydrodynamic similarity had been achieved between the hot
combustor and the cold model. When actual hot bed
material was used in the cold model, which was a violation of the scaling laws, the models behavior was
very dierent from that of the hot bed.
Horio et al. (1986) used three geometrically similar
bubbling beds, uidized with ambient air, to verify
their proposed scaling laws. The solid-to-gas density
ratio was maintained constant in the experiments,
although it was not one of the proposed scaling parameters. By maintaining a constant density ratio, they
Figure 19 Comparison of nondimensional bubble frequencies from two cold scaled models. (From Newby and
Keairns, 1986.)
7.5
Almstedt and Zakkay (1990) made scaling comparisons between a hot PFBC with horizontal tubes and
a pressurized cold scale model using the full set of
scaling laws. The cold model had linear dimensions
one-half those of the hot bed. A capacitance probe
was used to measure the mean values of the bubble
frequency, pierced length, bubble rise velocity, and
bubble volume fraction. Scaling comparisons were
made using the dimensionless form of these dependent
hydrodynamic parameters. Three dierent bed materials were used in the cold bed: olivine sand and two
dierent size distributions of the hot bed material,
one properly scaled and one out of scale. Almstead
and Zakkay concluded that behavior that is hydrodynamically similar to that of a pressurized uidized bed
Figure 22 Solid fraction prole comparisons for Tidd PFBC and cold model based on simplied scaling laws (From Glicksman
and Farrell, 1995.)
APPLICATIONS OF SCALING TO
COMMERCIAL BUBBLING FLUIDIZED BED
UNITS
The earliest scaling studies were directed at atmospheric bubbling bed combustors. To date, a rich variety of questions have been addressed.
Jones and Glicksman (1986) constructed a model
of the 20 MW bubbling bed pilot plant jointly sponsored by the Tennessee Valley Authority and the
Electric Power Research Institute (EPRI) at
Paducah, Kentucky. Figure 23 shows a photograph
of the model of the in-bed tubes installed in the scale
model. The model, which is roughly 100 by 120 cm in
cross section, simulates two-thirds of the entire 20
MW pilot plant. Care was taken to match the pilot
plant tube bundle geometry and distributor design.
Steel grit particles with the same dimensionless size
distribution and sphericity as the hot bed material
were used. The full set of scaling parameters was
matched in the model and the combustor. The largest
discrepancy was in the solid-to-gas density ratio,
which was 18% smaller in the model than in the
pilot plant.
The measured bubble velocity for an actively bubbling bed was found to agree closely with the drift ux
form proposed by Davidson and Harrison (1963). In
contrast, the volumetric ow rate of the bubbles was
found to be far less than that predicted by the twophase hypothesis; see Fig. 24.
Figure 23 Model of 20 MW bubbling uidized bed combustor showing tube arrangement. (From Jones and Glicksman,
1986.)
value below umf . In this design, the bed does not contain vertical partitions above the distributor. The scale
model was used to determine the rate of growth of the
xed bed in the deuidized zone along with the heat
transfer to tubes in that region. Figure 26 shows a
typical pattern of particle accumulation in a slumped
zone adjacent to an actively uidized zone. Heat transfer coecients are also shown. Note that tubes near the
upper surface of the solids that experience a downow
of solids have a very high heat transfer rate.
The heat transfer from tubes in the freeboard was
also measured for the 20 MW model. The variation of
measured overall heat transfer coecient in the 20
MW pilot plant versus velocity agreed closely with
that predicted from the scale model test. When the
bed height is lowered, uncovering some tubes, the
heat transfer is reduced because there are fewer particles contacting the tube surface.
Ackeskog et al. (1993) made the rst heat transfer
measurements in a scale model of a pressurized bubbling bed combustor. These results shed light on the
inuence of particle sizes, density, and pressure levels
on the fundamental mechanism of heat transfer, e.g.,
the increased importance of the gas convective component with increased pressure.
A multisolid bed contains a mixture of large solids
that are contained in a dense region at the bottom of
the bed and ner particles that recirculate through the
bed and external cyclone. Ake and Glicksman (1989)
used a cold scale model of a multisolids combustor to
determine the dense bed expansion. The measured
expansion in a properly scaled quarter-scale model
using steel pellets to simulate the coarse particles and
to satisfy the solid-to-gas density ratio gave good
agreement with eld data; see Fig 27. It was also
demonstrated that an improperly scaled cold model,
using the same coarse material as the hot bed, had
an incorrect gas-to-particle density ratio and substantially overpredicted the bed expansion.
Tube erosion has been observed in both atmospheric and pressurized bed combustors. The scaling
analysis presented earlier can be used to construct an
accurate hydrodynamic simulation of the commercial
bed. This can be used to investigate qualitatively the
factors related to tube wear such as the location of
highest wear around the circumference of an individual
tube and the location within the bed of highest wear.
Quantitative wear rates cannot be obtained from
model tests unless the parameters governing both the
hydrodynamics and the wear phenomena are matched
between the model and the commercial bed. The inclusion of more large particles causes an increase in the
Figure 24 Deviation from two-phase theory for model of bubbling uidized bed combustor. (From Jones and Glicksman,
1986.)
tion, bubble frequency, vertical bubble size, jet halfangle, and others. They also found considerable gas
leakage from the bubbles to the emulsion phase. This
data covers much larger bed sizes than had been
examined before.
HYDRODYNAMIC SCALING OF
CIRCULATING BEDS
Figure 25 Bed expansion for TVA 20 MWe FBC and for scale model, shown in Fig. 23. (From Glicksman et al., 1989.)
Figure 27 Expanded bed height model data compared to eld data for multisolids uidized bed for properly scaled steel pellets
and misscaled gravel particles. (From Ake and Glicksman, 1989.)
Figure 28 Range of experimental scaling studies for circulating and bubbling uidized beds.
Figure 29 Solid fraction proles, glass/plastic viscous limit scaling; density ratio mismatched: low velocity case. (From
Glicksman et al., 1993a.)
Figure 30
Solid fraction proles, glass simplied scaling, high velocity case. (From Glicksman et al., 1993a.)
model, and a one-sixteenth scale cold model. The onesixteenth scale model had a cross sectional area of
16 cm2 to simulate a 2.5 MW combustor! The average
solid fraction proles were in good agreement for most
of the conditions tested. The agreement was excellent
between the one-quarter scale cold model, which used
the full set of scaling laws, and the one-sixteenth scale
model that used the simplied set of scaling laws. Thus
any disagreement between the Studsvik bed and the
one-sixteenth scale model is not due to the simplica-
Figure 31 Solid fraction proles, hot bed scaling with simplied scaling laws, low velocity, 49% primary air. (From Glicksman
et al., 1993a.)
Figure 32 Solid fraction prole comparison: between pressurized circulating uidized bed combustor and one-half size scale
model based on simplied scaling. Two dierent operation conditions. (From Glicksman et al., 1995.)
Figure 33 Average solid fraction: 11.5 m high CFB combustor at 7.5 bar compared to 1/6.5 scale cold model.
10
CONCLUSIONS
operating characteristics and the inuence of hydrodynamics factors such as bubble distribution and trajectories. A series of dierently sized scale models can be
used to simulate changes in commercial bed behavior
with bed size.
The scale models must be carefully designed. Failure
to match one or more important dimensionless parameters will lead to erroneous simulation results.
Modeling can be extended to particle convective heat
transfer. Wear or erosion of in-bed surfaces can be
qualitatively studied, although quantitative assessment
requires the identication and simulation of additional
wear-related parameters.
Most of the simulation eort has been applied to
uidized bed combustors that use relatively large particles. Simulation can also be used for other uidization processes in the petrochemical industry. Research
should be undertaken to identify the proper scaling
parameters for beds uidized with smaller particles.
Similar simulations may also apply to components
such as cyclones.
ACKNOWLEDGMENTS
The author would like to acknowledge the contributions
of many present and former M.I.T. students who carried
out scaling studies of uidized beds. Much of the M.I.T.
research mentioned in this chapter has been sponsored
by the Electric Power Research Institute, the National
Science Foundation, the U.S. Department of Energy,
and the Tennessee Valley Authority.
NOTATION
CD
D
dB
dc
dp
dV
GS
g
L
QB
t
u
ub
umf
u0
uf
x; y
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Dimensionless
PSD
Re
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Drag coecient
Bubble volume fraction
Boundary layer thickness in a fast bed
Distributor plate voidage
Void fraction
Cluster void fraction
Void fraction at minimum uidization
Fluid density
Solid density
Fluid viscosity
Fluid kinematic viscosity
Particle sphericity
Greek
e
ec
emf
f
S
m
Subscripts
c
m
=
=
Commercial bed
Model bed
Superscripts
0
=
=
Dimensionless
Vector
REFERENCES
Ackeskog HBR, Almstedt AE, Zakkay V. An investigation
of uidized-bed scaling: heat transfer measurements in a
pressurized uidized-bed combustor and a cold model
bed. Chem Eng Sci 48:1459, 1993.
Ake TR, Glicksman LR. Scale model and full scale test results
of a circulating uidized bed combustor. Proceedings of
1988 Seminar on Fluidized Bed Combustion Technology
for Utility Appl. EPRI, 1989, pp 124.
Almstedt AE, Zakkay V. An investigation of uidized-bed
scaling-capacitance probe measurements in a pressurized
uidized-bed combustor and a cold model bed. Chem Eng
Sci 45(4):1071, 1990.
Arena U, Cammarota A, Massimilla L, Pirozzi D. The
hydrodynamic behavior of two circulating uidized bed
units of dierent size. In: Basu P, Large JF, eds.
Circulating Fluidized Bed Technol II. Oxford:
Pergamon Press, 1988.
Avidan A, Edwards M. Modeling and scale up of Mobils
uid bed MTG process. In: Ostergaard K, Sorensen A,
eds. Fluidization V. New York: Engineering Found, 1986.
Avidan A, Edwards M, Owen H. 50 years of catalytic cracking. Oil Gas Journal 33, 1990.
14
Applications for Fluid Catalytic Cracking
Ye-Mon Chen
Shell Global Solutions US, Houston, Texas, U.S.A.
1
1.1
INTRODUCTION
What Is FCC?
Fluid catalytic cracking (FCC) is the primary conversion unit in most reneries in the United States. It
converts low-value heavy components of crude oil
into a variety of high-value lighter products. A modern
FCC unit consists of three major sections: the reactor/
regenerator section, the main fractionation/gas plant
section, and the ue gas cleaning/power recovery section. This chapter focuses mainly on the reactor/regenerator section of the FCC process, paying special
attention to its relationship with uidization and
uidparticle systems. Readers who are interested in
broader aspects of the FCC process can refer to several
recent publications, such as Sadeghbeigi (1995) and
Wilson (1997).
In the reactor/regenerator section of the FCC process, liquid hydrocarbon feedstock is preheated, mixed
with steam, and injected into the riser reactor through
feed nozzles. Hot regenerated catalyst is drawn from
the regenerator and contacts the hydrocarbon feed.
The two are thoroughly mixed in the lower riser by
ne atomization through the feed nozzles. The temperature of regenerated catalyst drops rapidly as the
feed vaporizes and reacts, converting the feedstock into
lighter products while a coke layer deposits on the
spent catalyst, temporarily deactivating the FCC catalyst. Product vapor and spent catalyst are separated by
riser termination devices, such as cyclone systems, at
The FCC unit usually has the second highest throughput after the crude distillation unit in most U.S. reneries. Crude oil contains hydrocarbons of dierent
boiling points ranging from light gas to residue.
Typically, crude oil is rst distilled in an atmospheric,
or crude, distillation unit to produce a wide range of
hydrocarbon products according to the dierent boiling points in the original crude. The residue from the
atmospheric distillation unit, the atmospheric residue,
is then fed to the vacuum distillation unit, which produces vacuum gas oil and vacuum residue. The vacuum
gas oil from the vacuum distillation unit is the traditional FCC feedstock with typical boiling points ranging from 650 to 1050 F. Depending on the type of
crude and the operating conditions of the vacuum distillation unit, the portion of vacuum gas oil can range
from 25 to 35% of the basic crude. Many reneries use
delayed cokers to take the vacuum residue from the
vacuum distillation unit as feedstock. Liquid products
from cokers, the coker gas oils, are another common
feedstock for the FCC unit.
Over the years, FCC has evolved into a very exible
process that can take a wide variety of other feedstocks, including hydrotreated gas oils, deasphalted
oils, slop oils, and lube extracts. In addition, many
modern FCC units are capable of the direct processing
of atmospheric residue or vacuum residue, which is
called residual or resid uid catalytic cracking
(RFCC).
The FCC process also plays a key role as a feed
preparation unit for downstream rening processes.
For example, light petroleum gas (LPG) produced by
the FCC unit is a common feedstock to the alkylation
unit.
1.3
The FCC unit has several important functions in rening and petrochemical manufacture. First of all, the
FCC unit is a gasoline machine. About 45% of worldwide gasoline production comes either directly from
FCC units or indirectly from FCC downstream units,
such as alkylation units.
The second function of the FCC unit is to reduce
the amount of residue in crude oil. In the U.S. market,
the product slate produced directly from the distillation units contains too little transportation fuel (gasoline and diesel) and too much residue. The FCC unit
enables reneries to convert most of the residue into
lighter products, thus fully utilizing the crude oil to
meet the market demands.
The third function of the FCC unit is to provide
exibility to the rening operation. The FCC unit
can adjust operating conditions to maximize the production of gasoline, middle distillate (LCO), or light
petroleum gas (LPG). This exibility enables the renery to adjust its product slate to meet seasonable
demands in the market. Reneries in the U.S. also
take in dierent crude diets according to changes in
the crude market. The exibility of the FCC unit
enables the renery to process dierent crude diets as
the crude market changes.
The fourth function of the FCC unit is to produce
light olens for downstream rening processes, such as
alkylation and other petrochemical processes, as in
MTBE, PE, and PP plants.
2
2.1
2.2
Heat Balance
Like every system operating under steady-state conditions, an FCC unit must stay in heat balance at all
times. However, because catalyst circulation between
the regenerator and the reactor serves the dual purposes of providing reaction activity and the heat
requirement to the reactor, the heat balance in the
FCC process has special purposes that other processes
do not have. Thus, heat balance is the key to better
understanding of how dierent variables interact with
one another in the FCC unit.
Most FCC units only have a few independent variables. Typically, these independent variables are the
feed rate, feed preheat temperature, reactor/riser temperature, air ow rate to the regenerator, and catalyst
activity. The feed rate and air ow rate to the regenerator are set by ow controllers. The feed temperature
is set by the feed temperature controller. Catalyst activity is set by catalyst selection and fresh catalyst addition rate. Reactor temperature is controlled by the
regenerator slide valve that regulates the catalyst circulation rate. The catalyst circulation rate is not
directly measured or controlled. Instead, the unit relies
on the heat balance to estimate the catalyst circulation
rate. Except for these independent variables, other
variables, such as regenerator temperature, degree of
conversion, and carbon-on-catalyst, etc., will vary
accordingly to keep the FCC unit in heat balance.
These variables are dependent variables.
The overall heat balance requires that the FCC unit
constantly adjust itself to produce and burn just the
right amount of coke in the regenerator to supply the
heat requirements for the entire unit. The major heat
requirements are heating and vaporizing the hydrocarbon feed, heat of reaction, heating air and steam to
system temperature, and heat losses of the unit.
Combustion of coke in the regenerator is the single
source of heat supply to run the FCC unit. The heat
generated in the generator is carried to the rest of the
unit by catalyst circulation, as described by the following equation:
Heat transfer catalyst
catalyst
oil
2.6
Hydrogen Balance
Every chemical reaction must satisfy an overall material balance as well as individual atomic balances. It is
important to discuss the hydrogen balance in the FCC
process, because hydrogen in the feedstock is the limiting component of the cracking reactions.
Typical FCC feedstock contains about 12 wt% of
hydrogen. The average hydrogen content of lighter
cracking products is about 13.5 wt% whereas the average hydrogen content of the heavier liquid cracking
products is about 9 wt%. Since typical hydrogen content in coke is in the range of 5 to 7 wt% and coke yield
is about 5 wt% on feed, a simple hydrogen balance
shows that typical conversion of an FCC unit should
be about 70%. Any deviation from these typical values
will shift the hydrogen balance accordingly. For
instance, an increase in feedstock hydrogen content
will allow the unit to have higher conversion, leading
to higher production of cracking products, coke and
less heavier liquid cracking products.
Hydrogen balance also provides other insights into
the FCC process. For example, it is known that thermal cracking produces more C1/C2 dry gas. Because
these shortest hydrocarbon molecules have the highest
hydrogen contents, more coke and heavy cracked oils
are also produced by thermal cracking to satisfy the
hydrogen balance. Thus thermal cracking is undesirable. Dehydrogenation reaction, which is promoted by
metal contamination, has similar eects to those of
thermal cracking in increasing the production of coke
and heavy cracked oils based on the hydrogen balance.
Another example is to look at the eect of adding
residuals to the feedstock. As the percentage of residual in feed increases, the hydrogen content in the
overall feedstock will drop. Thus hydrogen balance
requires that a residual cracking unit operate at a
much lower conversion and have a higher coke yield
than a typical FCC unit.
The feed injection system is the most critical component of the modern FCC riser reactor design. Several
factors have made the feed injection system increasingly important over the years. The most important
factor is that the FCC reaction time is getting shorter
and shorter. This was not the case for earlier FCC
units when the catalyst was not as active and the reaction time was relatively long, i.e., on the order of minutes. Thus, mixing time and evaporation time of the
feed induced by the feed injection system were not as
critical. However, due to the development of the highly
active zeolite catalyst, typical reaction time has been
shortened to a few seconds in the riser reactor. Some of
the most recent FCC reactor designs have gone to
Figure 2 CCU feed nozzle design. (From Chen and Dewitz, 1998.)
Figure 3 Atomization and spray pattern of the feed nozzle shown in Fig. 2. (From Chen and Dewitz, 1998.)
3.2
Riser/Reactor
of the vertical riser. Thus the reactor vessel in the modern FCC unit downstream of the riser is actually a
containing vessel for spent catalyst/product vapor
separation. In some older FCC units, however, the
reactor vessel is designed to provide additional residence time for the cracking reactions to complete downstream of the riser while using a less active catalyst.
Due to the cracking reactions, the molar ow rate
and the volumetric ow rate of the hydrocarbon vapor
increase as it moves up the riser. In a typical FCC unit,
the volumetric expansion is in the range of 3 to 4 times
of the vaporized feedstock. Thus the riser diameter
may be increased once or twice after feed injection to
keep the vapor velocity within the range of 40 to 80
feet/s. Some modern FCC risers could further include
boundary layer strip rings, e.g., Gwyn (1993), to
reduce catalyst backmixing along the riser wall and
to achieve performance closer to an ideal plug ow
reactor.
A modern FCC riser design also includes a separation device at the end of the riser. The objective is to
achieve a quick and clean separation of spent catalyst
and hydrocarbon vapor in order to minimize post riser
cracking. Post riser cracking is undesirable because the
riser is designed to achieve optimum conversion, and
additional cracking reactions after the riser will push
the conversion outside the optimum range. One riser
termination design, called the prestripping cyclone,
uses a dead end tee at the riser top followed by a special
cyclone, as described in Parker et al. (1987). Figure 5
shows further details of this design in which the upper
part is a rough-cut cyclone for catalyst/vapor separation and the lower part is a prestripper bed where steam
is injected to remove entrained vapor. Catalyst not captured in the prestripping cyclone is separated in downstream secondary cyclones. In most modern FCC
designs, the secondary cyclones are close-coupled with
the primary cyclones, such as examples shown in Wilson
(1997), to minimize post riser cracking. The reactor
cyclones can be designed to operate either inside or outside the reactor vessel, as shown in Fig. 1.
3.3
Stripper
3.4 Regenerator
activity is higher. Potential drawbacks of total combustion regenerators include higher catalyst deactivation owing to higher regenerator temperature, lower
coke burning capacity owing to the requirement of
excess air, and higher heat release owing to total combustion reaction. Catalyst deactivation can be reduced
by using two-stage regeneration; see, e.g., Herbst et
al. (1989). In the two-stage regeneration, the rst
stage is operated at a moderate temperature to burn
o the hydrogen component on the coke, which has a
higher reaction rate, and some of the carbon component. The second stage uses excess air to burn o the
carbon component on the coke at higher temperature
to carbon dioxide. Because of the absence of water
vapor in the second stage regenerator, catalyst deactivation at high temperature can be minimized. Twostage regeneration is more expensive and more complicated to operate. The coke burning issue is typically addressed by having a large air blower, which
adds to higher capital and operational costs. The
third issue in total combustion regeneration is the
heat balance, because more heat is generated in the
regenerator at the same coke burning rate when carbon is reacted to carbon dioxide instead of carbon
monoxide. If more heat is generated in the regenerator than what is required for the reactor, a common
solution is to add a catalyst cooler to the regenerator
to remove the excess heat, as shown in Fig. 1. The
catalyst cooler is a heat exchanger that produces
steam while removing heat from regenerator. This
also adds to capital costs.
The phenomenon of afterburn also occurs in total
combustion regenerators. This is most likely to happen when the spent catalyst is poorly distributed and
a particular area of the regenerator has too much
coke. Although, on average, air is always in excess
in the total combustion mode, local areas could
have insucient air to burn all the carbon to carbon
dioxide if spent catalyst is not well distributed. As a
result, some carbon monoxide will escape through the
regenerator bed from these areas. Excess oxygen is
always present in the freeboard to react with carbon
monoxide and to produce carbon dioxide, causing an
afterburn problem. Afterburn in total combustion
regenerators can be overcome by using higher excess
air, promoter, or a better design of the spent catalyst
distributor, e.g., Fig. 6.
Residual cracking poses another challenge to regenerator design and operation. For a given conversion,
coke yield is much higher in residual cracking compared to the conventional FCC because of heavier
feedstocks and higher metal contents. Catalyst coolers
are required, regardless whether the regenerator operates in partial or total combustion, in order to keep the
unit in heat balance. In addition, metal contamination
on equilibrium catalyst is high, which has detrimental
eects on the catalytic reactions. There are several
ways of dealing with this problem. These include selection of catalyst with metal trapping capability, o-line
removal of metal contamination, and management of
equilibrium catalyst to control the maximum concentration of metals.
3.5
Third-Stage Separator
Standpipe ow is the main driving force behind catalyst circulation between the regenerator and the reactor, which drives the entire FCC process. The objective
of a standpipe is rather straightforward. It is to build a
hydraulic head by holding a column of uidized catalyst. This enables the catalyst to move in the direction
against pressure gradient, from low pressure at the top
to high pressure at the bottom of the standpipe. When
standpipes do not function properly, the FCC unit
cannot circulate catalyst at the design rate and the
unit is forced to cut feed rate. Even with such an
important role in the FCC process, standpipe ow
remains poorly understood. The existing theories on
standpipe ow, e.g., Leung (1977) and Chen et al.
(1984), are overly simplied.
Figure 8 CCU standpipe inlet design. (From Chen and Brosten, 2001.)
REFERENCES
Bedaw RE. Atomizing spray nozzle for mixing a liquid with a
gas. U.S. Patent 5,240,183, 1993.
Chen Y. A theoretical investigation of an aerated hopper
ow. J Chinese Inst Chem Eng 17:195, 1986.
Chen Y. Feed nozzle. U.S. Patent 6,387,247, 2002.
Chen Y, Brosten D. Standpipe inlet. U.S. Patent, 6,228,328,
2001.
Chen Y, Dewitz TS. Feed nozzle. U.S. Patent 5,794,857,
1998.
Chen Y, Rangachari R, Jackson R. Theoretical and experimental investigation of uid and particle ow in a vertical
standpipe. Ind Eng Chem Fundam 23:354, 1984.
Chen Y, Brosten D, Nielson JW. Feed nozzle. U.S. Patent
5,979,799, 1999.
de Kruij GT, van Hattem A. Apparatus for solidsuid
separation. U.S. Patent 5,1998,005, 1993.
Parker WA, Dewitz TS, Hinds GP, Gwyn JE, Bilgic AH,
Hardesty DE. Apparatus for the separation of uid cracking catalyst particles from gaseous hydrocarbons. U.S.
Patent 4,692,311, 1987.
Sadeghbeigi R. Fluid Catalytic Cracking Handbook. 1995.
Wilson JW. Fluid catalytic crackingtechnology and operation. 1997.
15
Applications for Gasiers and Combustors
Richard A. Newby
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
the specic application and the properties of the carbonaceous fuel being processed.
Fluidized bed gasiers dier from uidized bed
combustors in their basic reaction mechanisms, functions, features, and operating conditions. They have in
common, though, several characteristics. Fluidized bed
gasiers and combustors operate at relatively low temperatures, below the fuel ash agglomeration temperature, and they have relatively uniform
temperature throughout their volume. Their designs
are tailored to the properties and behavior of the carbonaceous fuels to be processed. Finally, they must
overcome operational and performance diculties
associated with the phenomena of nonuniform uidized bed gas mixing and bypassing, nonuniform fuel
feeding and mixing, bed overtemperature zones, bed
agglomeration, ash deposition, particle erosion, materials corrosion, and residual carbon conversion.
The many variations in design congurations and
features of the key types of uidized bed gasiers and
combustors have been developed from uidization
principles, chemical reactor principles, test observations, and empirical rules, coupled with conventional
engineering practice. Coals and biomass fuels represent
two contrasting extremes in fuel properties and behavior, as well as in industrial application capacity, and
have thus evolved uidized bed processing technologies having somewhat diering congurations and features. Other carbonaceous fuels, having their own
unique set of properties and behavior, have preferred
Fluidized bed gasication might be applied to carbonaceous fuels to produce steam or process heat, chemicals, or electric power, or combinations of these.
The uidized bed gasier combines processed carbonaceous fuel (dried and sized) with oxidant, steam,
and/or an external heat source to produce a raw reducing gas stream, a raw ash stream, or a raw char
stream, and possibly a raw liquid products stream.
Inert gases, such as nitrogen and recycled product gas
will also be fed to the gasier, functioning as solids
transport, aeration, and purge gases. These inert
gases can contribute signicantly to the makeup of
the raw reducing gas. Sorbent particles or inert bed
makeup particles may also be fed to the gasier.
The uidized bed gasier raw reducing gas stream
may be processed further to generate a syngas used
to produce chemicals, or a fuel gas used to generate
process steam, process heat, or electric power (Ghate
and Longanbach, 1988; Schobert et al., 1998). In some
gasication processes, liquid products (tars and oils)
might also be extracted from the raw reducing gas as
The functional requirements of the uidized bed gasier are to convert eciently and reliably the carbonaceous fuel into a raw reducing gas stream and a raw
char, or raw ash stream meeting all of the product
stream compositional, temperature, pressure, and
ow rate specications that are imposed on the gasier.
The uidized bed gasier is designed to promote a
reaction environment having good gasparticle contacting, good particleparticle mixing, and relatively
uniform temperature conditions, and to avoid operational diculties resulting from the agglomeration,
deposition, erosion, and corrosion characteristics associated with the gasication of carbonaceous fuels. The
design engineer is challenged to select fuel and sorbent/
inert particle feed sizes, uidization conditions, temperature environment, and gasier reactant feed
stream proportions that avoid the nonidealities common to uidization (for example, deuidization, gas
bypassing, excessive particle attriation and elutriation,
particle segregation, generation of hot spots in the
dense bed and freeboard, and nozzle and drain plugging).
The eciency of the gasier conversion might be
dened in a variety of ways, based on the energy content of the raw reducing gas, raw liquid products, raw
Reaction Environment
2.2
Table 2 presents a similar characterization of biomass-fueled gasiers nearing demonstration and commercialization. Partial oxidation at both elevated
pressure and low pressure are included, and both bubbling bed and circulating bed uidization are feasible
regimes. As with coal, the biomass gasiers are primarily fuel gas applications for steam or combustion turbine power.
Figure 1 illustrates the general schemes representative of the uidized bed coal gasication processes of
signicance today that are listed above, and Fig. 2
represents the key biomass fuel gasication processes.
The process schemes shown are oversimplied and
show only the major streams, gasier components,
and process interfaces relating to the primary uidized
bed conversions. Gasier stream interfaces are identied in the gures to emphasize the critical need for
satisfying all interface requirements. The gures also
indicate optional streams that are used depending
on the specic fuel properties and the application
needs. Nonmechanical valve locations for solids
recycle streams are identied, without aeration gas
being shown. In general, feed streams carrying solids
also require an inert transport gas (for example,
nitrogen or recycled product gas), and these transport
gases are not included in either Fig. 2 or 3. Inert gases
are also used at various places within the gasiers for
purging, aeration, local uidization, and stripping, and
these also are not shown in the gures.
Three schemes are shown in Fig. 1: (A) pressurized
bubbling bed partial oxidation of coal; (B) pressurized
bubbling bed carbonization of coal; and (C ) low-pressure bubbling bed pyrolysis of coal. Figure 1A is representative of pressurized bubbling bed partial oxidation
of coal. The sketch indicates that the bubbling bed
uidized bed gasier vessel cross-section is expanded
in the freeboard to minimize particle elutriation, and
this is generally true of all bubbling bed coal gasiers.
Elutriated nes recycling is used in general, with nonmechanical valves to seal the recycle system and to
control the reinjection of nes into the uidized bed
gasier. The raw fuel gas will have a low heating value
of about 3.35.6 MJ/Nm3 (LHV) for air-blown operation, and a medium heating value of about 8.211.2
MJ/Nm3 (LHV) for oxygen-blown operation. The vessel will drain a low-carbon ash product.
The gasier must interface with the pressurized oxidant delivery system (air or oxygen-blown gasiers
being feasible for coals); with the pressurized coal
and sorbent preparation and feed systems; with the
raw fuel gas pressurized processing system; and with
the pressurized ash processing system. Coals may be
Applications
Description
References
Kellogg
KRW gasier
IGT
U-gas gasier
Rheinbraum AG
HTW gasier
Renzenbrink et al.
1998
British Coal
Robertson, 1995
Lu et al., 1996
Table 2
Applications
Carbona Corp
(IGT Renugas gasier)
Cratech gasier
Description
100,000 MJ/h pilot unit operated on
range of biomass fuels;
integrated gas cleaning;
750,000 MJ/h alfalfa-stem plantation
unit in design
10,000 MJ/h pilot unit operating on
cotton gin mill waste;
integrated gas cleaning;
100 kW turbine operation planned
References
DeLong et al.,
1995
Foster Wheeler
Energia Oy
Termiska Processor
AB (TPS) gasier
Hirschfelder and
Vierrath, 1998;
DeLange and Barbucci,
1998
Nieminen and Kivila,
1998
The design of uidized bed gasiers requires that transport models be developed that apply the conservation
principles for mass, energy, and momentum to some
degree of fundamental detail. The mass and energy
balances are closely coupled, and their solutions,
applying empirical reaction kinetic and multiple
phase mixing models, provide estimates of solids and
gas composition and temperature proles and input
and output stream conditions. The momentum balances provide the gasier pressure prole and total
pressure drop when applied with appropriate uidized
bed phase density models. These empirically based
models provide the means needed to size the gasier
equipment and scale pilot plant data.
The design of uidized bed gasiers involves the
selection of several interrelating design, operating,
and performance parameters and requires the consideration of performance and cost trade-os for any specic application. The type of uidized bed gasier, the
fuel properties, and the application requirements
strongly inuence the uidization parameter values
selected for the design. To make such designs properly,
for all of the types of gasiers, fuels, and applications
described in this chapter, it is currently necessary to
Gasication of Coal
A pressurized bubbling bed partial oxidizer for biomass fuel, shown schematically in Fig. 2A, will be
designed much like the pressurized bubbling bed coal
partial oxidizer in Fig. 3. The higher reactivity of the
biomass fuel relative to coal, and the probable elimination of in-bed desulfurization with biomass fuels,
might result in shorter combustion and char gasication zones if the gasier operating temperature is not
limited to low values due to bed agglomeration potential. The high reactivity of biomass fuels makes relatively uniform fuel feed distribution even more
important than it is with coals. Reactive biomass
fuels may not require the use of ne particle recycling
or charash separation features. Secondary air might
be injected into the freeboard of the biomass fuel gasier to promote tar destruction through partial fuel gas
combustion. The biomass fuel bubbling bed would
consist largely of inert sandlike particles containing
small amounts of biomass fuel char and ash.
The same design considerations must be applied
with biomass fuels as with coals. Biomass fuel gasication is relatively simple due to the generally highly
reactive nature of biomass fuels. The greatest risk in
biomass fuel gasication is associated with ash, or ashinert particle agglomeration within the bed that may
lead to deuidization and plugging of nozzles and
drains. With the tendency for biomass fuel bed
agglomeration, relatively low partial oxidation temperatures may need to be selected that result in relatively high carbon losses. The tendency for greater
formation of tars in the raw fuel gas, and increased
gasier drain material carbon content, as the temperature is decreased, must be weighed against the tendency
for greater bed agglomeration, and possibly greater
alkali vapor release, as the gasier temperature is
increased. The reliable, controllable preparation, pressurization, and feeding of biomass fuels are signicant
challenges to be fullled to make the technology a
commercial success.
Pressurized circulating bed air blown partial oxidation of biomass fuels, shown schematically in Fig. 2B,
is feasible because of the high reactively of biomass
fuels. Proper biomass fuel drying and fuel and inert
bed material particle size control (size reduction or
pelletization) are important to the gasier performance. The biomass fuel is fed several meters above
the air distributor to a relatively low-density zone of
the fast uidized bed. The initial pyrolysis of the biomass fuel, with some partial combustion of the volatiles and char, will occur rapidly near the fuel injection
point, and the vessel diameter may be expanded at this
point owing to the large evolution of biomass fuel
volatiles. Steam reforming of char particles will continue throughout the fast bed riser, with the conversion
becoming less eective at the lower density higher elevations of the fast bed. The recirculating solids, containing unconverted char and sand, are injected into
the turbulent, dense uidized bed at the base of the
vessel, near the air distributor, so that unconverted
char can be burned.
major distinction of uidized bed combustors compared to other combustor types is that carbonaceous
fuel combustion occurs at a relatively low temperature
of 760930 C, within an almost uniform temperature
reactor environment. Carbonaceous fuel is fed into a
uidized bed consisting primarily of sulfur-sorbent
(limestone) particles, or inert particles, and containing
smaller portions of fuel char and ash particles. The bed
is uidized by air, fed in excess of that needed for
complete fuel combustion, to perform combustion
within the relatively well-mixed, relatively uniform
temperature of the uidized bed. The uidized bed
combustor may operate at near-atmospheric pressure
for steam generation, process heat production, or
Rankine cycle electric power generation; or at elevated
pressure for combined cycle power generation.
The uidized bed temperature, high enough to perform ecient combustion and sulfur removal, but low
enough to avoid bed agglomeration, is maintained by
rapidly removing heat to generate steam through heat
transfer surfaces immersed in the uidized bed. A raw
combustion gas is produced that is processed, primarily removing entrained particles, so that the product
combustion gas can have further energy extracted
from it. This gas processing is considerably less complex than the gas processing associated with uidized
bed gasication. Bed drain solids, a mixture of the
carbonaceous fuel ash and sulfur-sorbent products,
are produced that may require further processing
(heat removal, chemical/physical processing) before
disposal or by-product use.
The functional requirements for uidized bed combustors are generally more complex than those for uidized bed gasiers: they must eciently and reliably
combust the carbonaceous fuel, achieve specied sulfur
removal performance with economic consumption of
sorbent, generate a raw combustion gas meeting ow,
composition (SO2 ; NOx ; CO; O2 , particulate), and
temperature specications, generate steam meeting
ow, pressure, temperature, and quality requirements,
and generate a raw ash stream meeting its ow, composition, and temperature constraints. The uidized
bed combustors overall thermal eciency is the ratio
of the recoverable thermal energy content of the output streams (steam, raw combustion gas, raw ash) to
the carbonaceous fuel input energy. This thermal eciency is strongly inuenced by the carbonaceous fuel
properties and the nature of the process application
and is sensitive to unburned carbon losses. The combustion ineciency is the fuel value of the waste solids
streams (primarily based on their carbon content) over
the carbonaceous fuel input.
3.3.1
horizontal heat transfer surface within the dense uidized bed, achieving high heat transfer coecients
(250400 W=m2 C) but also facing the possibility of
steam tube erosion. Heat transfer surface is also
placed above the dilute freeboard zone where convective steam generation occurs. Bubbling bed combustor dense beds are generally about 11.5 m deep,
allowing them to achieve acceptable pressure drops
(1525 kPa), while the splash zone and freeboard
height is about 10 m from the top of the dense bed.
The combustion gas velocity above the transport dis-
As with bubbling bed and circulating bed AFBC technology evolution, bubbling bed PFBC has led the
development of PFBC over the still-to-be-demonstrated circulating bed PFBC. Bubbling bed and circulating bed PFBC both appear to be attractive options
for advanced high-eciency power generation systems
for large electric generating applications.
The basic features of bubbling bed and circulating
bed PFBC are analogous to those of bubbling bed and
circulating bed AFBC, except for the following points:
The elevated pressure of PFBC (10001500 kPa)
means that even the bubbling bed combustor
cross-sectional area is reasonable for shop fabrication at relatively large capacities.
The elevated pressure of PFBC means that the dense
bubbling bed can be relatively deep (35 m) and
still meet pressure drop limits, so that the distribution of coal feed points is not as demanding as
it is in bubbling bed AFBC.
The bubbling bed PFBC combustor vessel is typically tapered near the base to provide higher uidization velocities and more intense mixing in the
regions where coal is fed.
Coal may be fed by either dry lock hopper methods
or by more reliable waterpaste methods.
The PFBC vessel, cyclones, solids circulation equipment, and the external heat exchanger may be constructed as independent cylindrical pressure vessels.
Alternatively, at water-walled construction, like that
used in AFBC may be placed, along with cyclones and
many other pieces of process equipment designed for
low-pressure operation, within the pressurized environment of a single large external pressure vessel, allowing
the combustor to retain many of its AFBC features.
Figure 6 illustrates the PFBC power plant conguration, showing the main power plant components.
The gure is representative of bubbling bed PFBC
with the currently commercial process conguration:
multiple stages of cyclones are used to remove particulate from the combustion gas, and the combustion gas
is expanded through a ruggedized turbine expander, an
expander designed to tolerate large particulate content
in the gas. The stack gas must be further cleaned of
particulates in commercial equipment (electrostatic
precipitator or fabric bag lter) to meet environmental
standards. The PFBC unit and several other components are located in a large external pressure vessel that
is pressurized and cooled by the warm compressor air.
This arrangement separates the thermal boundary
from the pressure boundary and provides several practical equipment and process advantages over the use of
separate pressure vessels for each component
(McDonald et al., 1987; Huryn et al., 1987).
PFBC will operate with conventional heavy-duty combustion turbines having higher ring temperatures and
higher pressure ratios than those used in PFBC. Hot
gas lters must be used to remove particulate from
both the combustion gas and the carbonizer fuel gas
to protect the turbine expander from erosion and
deposition damage. The additional challenge of topped
PFBC is to integrate and control the combined operation of the carbonizer and combustor and their interfacing components.
The design of the uidized bed carbonizer and combustor unit for topped PFBC change little from those
design factors already reviewed. Because a higher turbine inlet temperature is used than in PFBC, the operating pressure of the uidized bed combustor would be
higher, resulting in smaller bed diameter and smaller
external pressure vessel diameter than for a comparable PFBC. The uidized bed combustor is fed warm
char from the uidized bed carbonizer at appropriate
char feed distribution points, and the combustor must
be sized to burn eciently the low-volatile char and
convert the CaS from the uidized bed carbonizer
into CaSO4 . Topped PFBC also has the option of feeding coal to the uidized bed combustor as well as to the
uidized bed carbonizer, making much greater power
plant exibility possible.
REFERENCES
Agrawal RK, Haldipur GB. Advanced gasier-desulfurizer
process development for SNG application. Final Report
by KRW Energy Systems, Inc., to Gas Research Institute,
GRI-89/0011, 1988.
Anderson KD, Manaker AM, Stephans EA Jr. Operating
experience of the Tennessee Valley Authoritys 160-MW
atmospheric uidized bed combustion demonstration
unit. Proceedings of the 14th International Conference
on Fluidized Bed Combustion, ASME, 1997, pp 3945.
Barducci G, Daddi P, Polzinetti G, Ulivieri P. Thermic and
electric power production and use from gasication of
biomass and RDF: Experience at CFBG plant at Greve
in Chianti. Proceedings of the Second Biomass
Conference of the Americas, Portland, 1995, NREL/CP200-8098, pp 565575.
Belin F, Flynn TJ. Circulating uidized bed boiler solids
system with in-furnace particle separator. Proceedings of
the 1991 Fluidized Bed Combustion Conference, ASME,
1991, pp 287294.
Bernstein N, Goidich S, Li S, Phalen J. Inuence of fuel type
and steam cycle on CFB boiler conguration. Proceedings
of the 13th Fluidized Bed Combustion Conference,
ASME, 1995, pp 2737.
Brereton C. Combustion performance. In: Grace JR, Avidan
AA, Knowlton T, eds. Circulating Fluidized Beds. New
York: Blackie, 1997, pp 376416.
Bryan BG, Hoppe JA. Status of the Shanghai Coking and
Chemical Companys U-gas coal gasication plant.
Proceedings of the Fifteenth Annual International
Pittsburgh Coal Conference, Pittsburgh (CD-ROM),
1998.
Chau LL, Rovner JM. Wood gasication for combined cycle
power applications. Paper presented at the AIChE
Summer National Meeting, July, 1995.
Craig JD, Purvis CR. A small scale biomass fueled gas turbine engine. Paper presented at the ASME, International
Gas Turbine and Aeroengine Congress and Exhibition,
Stockholm, Sweden, June 1998.
Dawes SG, Mordecai M, Welford GB. Recent design and
cost studies for air blown gasication. Proceedings of
the 14th International Conference on Fluidized Bed
Combustion, ASME, 1997, pp 347358.
DeLong MM, Oelke EA, Onischak M, Schmid MR, Wiant
BC. Sustainable biomass energy production and rural
economic development using alfalfa as feedstock.
Proceedings of the Second Biomass Conference of the
Americas, Portland, 1995, NREL/CP-200-8098, pp
15821591.
16
Applications for Chemical Production and Processing
Behzad Jazayeri
Fluor Daniel, Inc., Aliso Viejo, California, U.S.A.
CHEMICAL SYNTHESIS
1.1
Acrylonitrile
dized bed catalyst discussed above. Details of the catalyst structure are provided by Contractor et al. (1978).
A signicant departure from conventional uidized
bed design is the use of the catalyst as oxygen carrier.
The catalyst is oxidized in a separate regeneration vessel by contact with air. It is then circulated to a riser
Range tested
360420
<5
2501100
410
< 25
Oxychlorination of Ethylene
1
C2 H4 2HCl O2 ! C2 H4 Cl2 H2 O
2
Phthalic Anhydride
Coal tar naphthalene is oxidized with air over a vanadia catalyst in a uidized bed according to the following highly exothermic partial oxidation reaction:
C10 H8 4SO2 ! C8 H4 O3 2H2 O 2CO2
suggested experimenting with internal serpentine cooling coils as a way to increase the heat removal capability of the reactor. This experiment proved to be very
successful. Heat removal was increased so much that
the external cooler was no longer needed. The behavior
of the uidized bed was also markedly improved with
the addition of the coils. The second generation reactors were designed with internal cooling coils only.
Medium to high pressure steam was generated directly
in the coils.
Entrained catalyst leaving the bed was originally
removed by externally mounted ceramic lter elements.
The hot reactor ogas is susceptible to afterburn
(Graham and Way, 1962). Long runs of pipes between
the reactor and lter housings were used to cool the gas
to about 260 C before the catalyst nes were removed.
The recovered dust was blown back to the reactor with
heated air to prevent thermal shock of the lter ele-
GAS TO LIQUIDS
2.1
Sasol
Figure 5
The synthol uidized bed process produces hydrocarbons in the gasoline boiling range using an iron catalyst. The scale-up and reactor design development are
depicted in Fig. 5 (Jones, 1991). Sasol 1A and B used a
circulating uidized bed design based on the original
10 cm pilot plant work of the WM Kellogg company in
the 1940s. These rst commercial reactors have a diameter of 2.3 m and operate at about 2 m/s. They represent a 500-fold increase in gas throughput compared to
the pilot plant. Catalyst is circulated between the
upow leg of the reactor and the downow leg of the
disengaging hopper and standpipe. The reactors were
originally built with two banks of xed tube sheets for
heat removal. The reacting gas and catalyst mixture
owed through the tubes and was cooled by oil circulating on the outside. Operating problems plagued the
startup and successful operation of the Sasol 1A and B
reactors (Shingles and Jones, 1986). The disengaging
hopper cone was too shallow. Funnel ow occurred,
causing bridging of the catalyst. The standpipe was
found to operate in stickslip ow (Chapter 21) and
was subject to bridging and loss of circulation. The
reactors were designed originally to operate with continuous catalyst addition and withdrawal to maintain
catalyst activity. This was never practiced in these rst
generation reactors as the valve technology capable of
reliable operation was not available at that time. The
most serious problem was however with the tube
sheets. Localized pressure uctuations resulted in preferential gas ow through some tubes, leaving the
others with a dense solid pocket that quickly plugged
due to formation of waxy compounds. The increased
gas ow through the open tubes then resulted in tube
erosion.
It took Sasol two years to learn to operate these rst
reactors with acceptable run times between shutdowns.
In 1960s, a third reactor, Sasol 1C, was added; it used
longer tubes for increased heat removal and a taller
standpipe to improve the pressure recovery. In 1974,
Sasol, reacting to the world oil crisis, decided to build
the much larger Sasol 2 complex. A cooperative program was started with Badger to review the Synthol
reactor operation and determine the best options to
eliminate the operating problems. Excellent accounts
of this cooperative development have been provided
by others (Shingles and Jones, 1986; Jones, 1991). A
brief review follows.
Sasol 2 was built using the Synthol concept and
was deemed a major success. The disengaging hopper
cone angle was increased to eect mass ow, eliminating previous bridging problems. The standpipes were
designed with smaller diameters, eliminating stickslip
ow. However, the most signicant design improvement was the replacement of the xed tube sheets
with serpentine cooling coils. By allowing the gas/
solid mixture to ow on the outside of the coils, the
instability problems associated with previous xed
tube sheet design were eliminated, allowing reliable
continuous operation. According to Jones (1991), a
test coil bank was installed in Sasol 1C to demonstrate feasibility and measure erosion. The test proved
successful. Very high heat transfer coecients were
measured, and the mechanical integrity of the design
was proven. The 10 Synthol reactors of Sasol 2
(Jones, 1991) have now operated for over 20 years
with no reported problems with the coils. In the
early 1980s, the Sasol 3 complex, essentially a copy
of Sasol 2, was added. The second generation Synthol
technology was licensed for use in the Mosselbay project that started in 1991. The Sasol 2 reactors are
apparently still operated batchwise. As the catalyst
ages, the density is reduced, and the mass ow
through the reactor must be increased to maintain
constant holdup (Shingles and McDonald, 1988).
2.2
Exxon
alumina (95 wt% minimum) supported catalyst containing 0.5 to 2.5 wt% nickel and substantially no silica
is used. The catalyst density is very high, between 2.4
and 3.9 g=cm3 . Exxon claims that catalyst nes leaving
the reactor cyclone system can lay down in the overhead piping and will promote the reverse reaction of
syn gas to methane, increasing methane slip. The above
catalyst characteristics are considered essential by
Exxon for minimizing nes generation and thus achieving high methane conversion. The preferred reactor
temperature is between 980 and 1000 C. Operating
pressure can be as high as 40 atmosphere, but the preferred range is 20 to 30 atmospheres. The uidized bed
syngas generator is shown in Fig. 8. Natural gas/steam
and oxygen enter the bed via separate spargers. The
reactor ogas is separated from entrained solids using
external cyclones and is then rapidly quenched to prevent the reverse reaction to methane and carbon formation by the Boudard reaction. Exxon claims that a
minimum cooling rate of 150 C and preferably 200 C
is required. A circa 1.5 meter pilot plant was operated
in early 1990s to prove the syngas reactor concept.
Conversion of syngas to hydrocarbons is by the
FischerTropsch synthesis. A multiphase slurry bed
reactor is used. Information on the slurry reactor is
limited, but the general design is expected to be similar to that in Fig. 7. Exxon claims that the addition of
an inert solid enhances bed expansion and reactor
performance.
2.3
Syntroleum
The Syntroleum company has announced the development of an air-based gas conversion process. Syngas is
produced using an air-blown authothermal reformer.
The syngas and diluent nitrogen is then passed on a
once-through basis through the conversion reactor.
Syntroleum is developing xed, uidized, and slurry
bed conversion reactors using cobalt catalyst. The
company claims favorable economics at capacities as
low as 5,000 barrels per day and has joined up with
ARCO and Texaco to develop slurry bed conversion
reactors.
POLYMERIZATION OF OLEFINS
In addition to gasication (Chapter 15), several uidized bed processes are available for upgrading of
heavy oil. These upgrading processes fall into two
groups. The rst group upgrades heavy oil by carbon
rejection. These include resid cracking (see Chapter
14), uid coking, and exicoking. The second group
upgrades heavy oil by hydrogen addition. These
as gas bubbles, pass through cyclones and into a scrubber, where they are quenched before being fractionated. The coke particles from the reactor are rst
steam stripped for product recovery and then ow to
the burner vessel, where a portion of the coke is burnt
with air. Hot coke is then returned back to the reactor
to supply the heat of reaction.
4.2
Flexicoking
Figure 11
Combi-Cracking
4.5
HDH
U-CAN
SEMICONDUCTOR SILICON
MEMC Electronic Materials uses a uidized bed reactor originally developed by the Ethyl Corporation to
produce high-purity silicon for use in the semiconductor industry. The process was piloted (Ibrahim and
Johnston, 1990) using several pilot plant scale units
of increasing size before being demonstrated in a 125
tons per year market development unit. A commercial
plant with 1250 tons per year capacity was started in
1987. The high purity silicon is produced in the uidized bed reactor by decomposition of silane in a
hydrogen atmosphere by the reaction
SiH4 ! Si 2H2
loss. Dust generation increases with higher silane concentration, increased reactor temperature, a larger
reactor bubble holdup, higher gas velocity, and lowered bed height. However, without adequate bubble
fraction, solid mixing is reduced and the particles can
agglomerate when silane is decomposing on the surface. In extreme cases, the whole bed can collapse.
Operating conditions of the reactor are reported as
600 to 800 C and 5 to 15 psig. Average particle size
is 700 to 1100 micron. Operating velocity is 1.2 to 3.5
minimum uidization. Under these conditions, the
reactor operates in a slugging mode.
Ibrahim and Johnston suggest a two-cycle reactor
operation. In the initial cycle, the reactor is operated in
a high productivity condition with silane concentrations as high as possible. This operating condition
results in appreciable homogeneous decomposition
and nes generation. In the nal cycle, the silane concentration is reduced for a short time. This condition
promotes heterogeneous decomposition. A layer of
silicon of possibly no more than 1 micron is deposited
on the particles, which cements the loosely adhering
silicon dust formed in the initial cycle. It is presumed
that product withdrawal occurs immediately after the
nal cycle.
The silicon produced in this process typically contains 10 to 44 ppmw of hydrogen. This hydrogen level
can be further reduced to 2 ppmw by heat treatment in
a second uidized bed reactor.
METALLURGICAL PROCESSING
6.1
Figure 14
7
7.1
OTHER PROCESSES
UOP/Hydro Methanol to Olens
between which catalyst is circulated. Methanol conversion is essentially 100%. The reaction produces a small
quantity of coke that will slowly deactivate the catalyst. To maintain activity, catalyst is circulated to the
regenerator where the coke is burnt o with air. The
heat of reaction is removed by steam generation. As
the reactor severity is increased, the ethylene/propylene
ratio and the coke yield increase, while the overall olen yield drops (Vora et al., 1996). A 0.5 MT/day
demonstration unit was operated in the mid-1990s.
UOP has announced the successful completion of the
demonstration unit and the availability of this process
for licensing.
7.2
MobilBadger Technologies
Figure 16
7.3
7.4 Isophthalonitrile
GB Biosciences (now part of Zenneca) has been operating a dual uidized bed process to produce isophthalonitrile by ammoxidation of m-xylene, by the reaction
C8 H10 2NH3 3O2 ! C8 H4 N2 6H2 O
Figure 18
REFERENCES
Benson JS. Catoxid for chlorinated byproducts.
Hydrocarbon Processing, pp 107108, October 1979.
Bolthrunis CO. An industrial perspective on uid bed reactor
models. Chem Eng Progress, pp 5154, May 1989.
Bunk S. Fluid-bed roasting advances. Randol Gold Forum,
pp 173174, September 1990.
Burdett ID. A continuing success: the UNIPOL process.
Chemtech, pp 616623, October 1992.
Contractor RM. US Patent 4,666,802 to E I DuPont de
Nemours and Company, May 26, 1985.
Contractor RM, Bergna HE, Horowitz HS, Blackstone CM,
Malone B, Torardi CC, Griths B, Chowdry U, Sleight
AW. Butane oxidation to maleic anhydride over vanadium phosphate catalysts. Catalysis Today 1:4958, 1978.
17
Applications for Coating and Granulation
Gabriel I. Tardos
The City College of the City University of New York, New York, U.S.A.
Paul R. Mort
Procter & Gamble, Cincinnati, Ohio, U.S.A.
SUMMARY
INTRODUCTION
Figure 1 Schematic representation of some industrial granulators. (a) Conventional uid bed granulator. (b) Conventional
spouted bed granulator. (c) Spouted bed with draft tube. (d) High shear rotor granulator.
UB a 2
DB
Figure 2 Schematic representation of a uid bed granulation circuit. Shown in the circuit is a uid bed grindergranulator. 1.
Fluid bed houseing, 2. Fluidized bed, 3. Grinding roller, 4. Drive, 5. Spray head, 6. Binder solution, 7. Pump, 8. Fluidization air
fan, 9. Air heater, 10. Air control valve, 11. Cyclone, 12 and 13. Rotary valves, 14. Filter bag-house. (After Dencs and Ormos,
1993.)
MICROSCOPIC PHENOMENA
Figure 3 Granule growth mechanisms (a) Agglomerate formation by nucleation of particles. (b) Agglomerate growth by
coalescence. (c) Layering of a binder-coated granule. (d)
Layering of partially lled binder droplet.
point can exhibit some softness, i.e., malleable character, which would allow surfaces to bond upon contact. From the simplied picture of Fig. 4, all
mechanisms of Fig. 3 can be easily reconstructed if
one allows for variation in size and nature (solid or
liquid) of the colliding particles.
The study of the kinetics of granule growth and consolidation in binder-granulation can be undertaken
using the picture in Fig. 4 that encompasses, in a simplied way, all mechanisms depicted in Fig. 3. The
relative velocity, U0 , between binder-coated particles
is generated by the mixer, i.e., the uidized bed, in
this case. The magnitude of the relative velocity is
mixer dependent and is predicted approximately by
Eq. (1). The kinetics of the process essentially reect
the conditions under which two colliding particles will
stick together or rebound in the shear eld: sticking
will yield growth while rebounding will not. The rate
of growth will depend on the relative sizes of the individual particles involved. The mechanisms of coalescence (Fig. 3b) and nucleation (Fig. 3a) involve
collisions of similar-sized particles and will yield higher
growth rates than layering (Fig. 3c) and engulfment
(Fig. 3d), which incorporate nes particles into a larger
granule. Once formed and grown to a certain size,
granules have to survive in the shearing mass; special
conditions between the magnitude of shear in the outer
shearing mass and the inner strength of the granule
have to be satised in order for this to happen.
Growth is also inuenced by granule consolidation
that is closely related to the viscosity of the binder
after granule formation and the time allowed for deformation of the formed bridges.
Several essential conditions have to be fullled a
priori in order for granule growth and consolidation
to occur. These conditions pertain to the presence of
appropriate amounts of binder on the granular surface at
the points of contact and the binders physical and ow
properties in the bridge; these aspects of granulation
will be examined rst. Following this, granule growth
kinetics and considerations regarding granule consolidation will be presented.
It is intuitively obvious from the schematic representation in Fig. 4 that a certain amount of binder is
required in the powder mass before enough will be
present on the surface to ensure stickiness. This critical, minimum amount of binder is an important characteristic of the system and must be determined
beforehand. Equally important are (1) the initial distribution of the binder within the bulk powder and (2)
the time span over which the binder either spreads on
the granular surface and/or penetrates into the pores of
the powder. Also critical is the time over which the
binder bond increases in strength through evaporation
3U0 a2
a sin2 C0 2
4h
2
Figure 6 Liquid bridge formed between two moving, spherical particles covered with a liquid layer.
Stcoal
Stcoal
2Fvis h
9
acoal
cr
9Stcoal
8p
1=2
A
1=2
5.3
mp U02
2Vp
where y is the yield strength, k is an apparent viscosity, and n is the ow index. A green unsolidied
granule is, from a rheological point of view, a complex
p a2 2
2y
GRANULE CONSOLIDATION
x
1 expSt
2h
11
13
EXPERIMENTAL VERIFICATION
whelm frictional eects, indeed exhibits a linear velocity prole, provided that conditions close to nonslip
can be generated at the walls. The shear rate for the
system can thus be calculated from
Vwall
14
15
scales of granulation equipment. In all cases, a pharmaceutical formulation of an initial size of 45 microns was
granulated using a polymeric binder. The following sections analyze the data in terms of binder dispersion,
growth, and consolidation.
7.2.1
Binder Dispersion
As discussed in the background on microscopic phenomena, the prerequisite for growth by binder granulation is the adequate presence of binder at contact
points. Therefore it is important to consider the initial
step of binder dispersion and its eect on growth.
Inadequate binder dispersion in the agglomeration
device can result in product heterogeneity. For example, some particles with ample binder will grow large in
size while others remain unagglomerated owing to
local scarcity of binder. Binder delivery is measured
in terms of ux of binder per surface wetted area of
powder. The ux is dened by the binder ow rate
relative to the spray coverage area, and the degree of
powder uidization or mixedness that provides fresh
powder surface to the spray.
Watano et al. (1995) report three scales of agitated
uid beds that dier in binder spray area relative to
batch size (Table 1). The batch size was increased in
proportion to the mixer volume, maintaining a xed
powder height/diameter ratio for all scales. In addition, the area of spray coverage was kept in proportion
to the cross-sectional area of the vessel. However, the
ratio of the spray area to the product bulk volume
decreased with scale. This resulted in poorer binder
dispersion and broader granule size distribution at larger scales. In addition, the uidization air velocity was
also varied. At higher air velocity, the turnover of powder in the spray zone increased, leading to improved
dispersion and narrower size distribution. The combined eect of both variables is shown in Fig. 11
where the geometric standard deviation, (GSD) g is
given as a function of relative spray area and gas velocity. As seen, the data is continuously arranged
throughout the eld of the gure: improved dispersion
of binder caused by increase in air ow velocity,
increase in spray surface area, or decrease in mass in
the mixer yields narrower product size distribution.
The data is also continuous in mixer scale as long as
the above conditions are equal.
7.2.2
Agglomerate Growth
Figure 10a Time sequence of powder granulation in the constant shear granulator.
17
n
logacr 1 log log C m log c
2
18
NQ-125
NQ-230
NQ-500
12.5
122.7
38.5
1
0.36
106.9
23
415.5
132.7
6.23
2.23
59.5
50
1963.5
594.0
64
22.9
25.9
Ui
logacr m log 2 c
Uf
19
Here we assumed that the shear rate, , is proportional to the impeller tip speed, Ui , and therefore the
quantity Ui =Uf2 is some modied shear rate. The average value of the slope for all granulators, large and
small, is approximately m 0:25, which gives a rheological ow index of n 1:5. Such values are common
for concentrated slurries of viscous liquids and small
solid particles and characterize a shear thickening
behavior. The most important conclusion from the
above considerations is that both the slope, m, and
the coecient, c, take approximately the same value
for all size granulators operated under dierent conditions, especially at the high-shear range of the impeller
where the power-law term of the viscosity dominates.
This implies that the binder viscosity has the same role
in both small and large units and that the Stokes number, based on deformation, is the relevant dimensionless parameter. It also implies that the Stokes number
may be the appropriate scaling factor when such
devices are scaled up.
7.2.3
Agglomerate Consolidation
20
with K being a parameter that incorporates all variables except the shear rate and where we again assume
the propeller tip speed to be proportional to the shear
rate.
Watano et al. (1995) found that uidization gas
velocity also has an inuence on consolidation, i.e.,
higher consolidation is obtained at lower gas velocities
as shown in the regression in Fig. 13, but this is not
captured directly in Eq. (20). A trend, however, can be
obtained by combining Eqs. (13), (2), and (4) to yield
a g
p U B a 2
p a2 g1=2
34
48
2
g
DB
2
D1=2
B
21
where we used UB 0:71 g DB 2 . It is likely that uidization conditions impose a smaller bubble size, DB ,
at lower gas velocities, and that would result in larger
consolidation; this would explain the result as found by
the above authors. Although it is not clear how the
dimensionless bubble spacing,
, changes under these
conditions, it appears that a combination of the above
two relationships would correctly reect the trend
obtained experimentally,
2
a g
00 Ui
K
g
Uf
22
These authors, with a long history of eminent granulation research starting as early as 1962, published
recently a set of papers reporting on granulations
aimed to recover solids from a concentrated aqueous
solution. They used a uidized bed in which shear
forces are controlled by bubble motion and by a rotating insert. The insert has two cylindrical rollers
mounted on the ends of a rotating arm that keeps
the rollers adjacent to the wall of the uidized bed,
as shown in Fig. 2 (a similar construction has the rollers moving parallel to the distributor plate). The gap
between the rollers and the wall (distributor) can be
adjusted, and by controlling the speed of the rotating
arm, strong shear forces can be generated. The device
is called a uid bed grinder-granulator by the
authors to emphasize the grinding action of the rollers
combined with the granulation function of the uid
bed.
The unit can be run in a semicontinuous mode in
which the bed is started with seed particles from a
previous run and solution is introduced continuously
while solids are discharged through a tube close to the
distributor. The granulator is operated at some elevated temperature above 100 C so that water from
the solution is evaporated and the residual solid material is layered onto the existing solids in the bed. A
constant agglomerate diameter is obtained after some
time of operation, and from there on a steady state sets
Figure 14 Experimental data by Dencs et al. (1993) and Dencs (1998); agglomerate size versus modied shear rate and
uidization velocity. Data for sucrose and zinc sulfate solutions in small and large uid beds.
tions between solid particles that also include the inuence of the liquid surface tension; details of the
procedure are given in the above paper. The most
important addition to the original RGFS simulation
is the inclusion of viscous and capillary forces [as
given in Eq. (2)] into the force balance used to solve
Newtons law (Fext m dU=dt). These forces are added
vectorially to supplement friction, elastic, and sometimes plastic forces used in previous simulation.
While the addition of viscous forces looks like a simple
extension of previous work, it is, in fact, a very serious
complication: this is due to the extensive (long) range
action of viscous forces and to the fact that they
become very large when particles come into close
proximity (and become unbounded when particles
touch). In addition, simple solutions only apply to
very slow movement (low Reynolds number, inertialess
ows) of the uid between the particles.
The simulation was successfully implemented for a
maximum of 40,000 particles, of which up to 30%
carried the sticky binder; a parallel computer system
at the University of Illinois at UrbanaChampaign and
a Sun computer cluster at CCNY were used. To limit
the complexity of the code and to keep computer running time to acceptable levels (1214 hours), it was
necessary to assume that particles covered with binder
(uid) at the onset of binder introduction were the only
ones carrying the binder for the duration of computations. In other words, sticky binder-covered particles
always remained sticky with a constant uid layer
on them, while dry (unsticky) particles remained
always dry, and no binder transfer between particles
was allowed during a particular run. Additional binder
could be introduced at set intervals but only to make
more (or less) particles sticky which then remained
binder covered for the duration of the run.
8.1
The simulation domain with 40,000 particles is presented in Fig. 15. The shear is introduced at a constant
rate through the upper and lower cell boundaries as
shown. Mirror images of the simulation domain ensure
that the inuence of solid boundaries is neglected (see
Khan and Tardos, 1997). The Stokes [St, dened in Eq.
(3)] and capillary (Ca, dened in Fig. 6) numbers are
the dimensionless parameters of the problem that have
to be perceived as dimensionless binder viscosity or
shear rate and uid surface tension, respectively.
Increasing binder viscosity and decreasing shear rate
result in lower Stokes numbers, while increasing the
liquid surface tension results in lower capillary num-
Figure 16 Simulated granulation with increasing amounts of binder. Cumulative particle size distribution vs. relative agglomerate diameter. Stokes number St 0:1 and capillary number Ca 1:0.
dagg
St
dp
St
24
!2:5
2
6
St 4
92r1:65
B
for St < 1
25a
for St > 1
25b
1=4 32:5
h
dp
Ca0:32
7
5
2:25
Ca
for Ca < 10
26
Figure 18 Values of the critical Stokes number, St , during granule deformation and break-up. Dimensionless agglomerate size
vs. Stokes number for values of the capillary number of Ca 10, 1.0 and 0.5. Experimental result reproduced from Fig. 10.
CONCLUSIONS
A comprehensive model of granulation was presented,
one that takes an agglomeration process from binder
selection and testing to granule growth, deformation,
and consolidation. An instrument and procedures were
described to measure the critical amount of binder that
ensures that coalescence and particle growth will
occur. It was also shown how the pendular bridge
apparatus (PBA), developed earlier by the present
authors, can be used to test the bridge-forming characteristics of the binder while at the same time determine binder spreading and penetration rates and the
critical time of binder strengthening. These operations
can be performed before any granulation runs are carried out, and binder characteristics can be modied to
achieve proper wetting, bridge formation characteristics, and binder strengthening (viscosity increase).
A theoretical model of particle/granule growth
based on simple principles of energy dissipation in
the liquid (binder) bridge between two colliding particles was developed. When the initial energy of the particles is smaller than the dissipated energy, particle
coalesce and growth will take place. This concept, ori-
=
=
=
=
=
=
=
=
=
=
D0 D50 [at x 1
Dg
Dd
D L B=L B
DB
bed
=
=
=
=
e
F
g
h
k
L
mp Vp p
m 1 n=2
n
N
=
=
=
=
=
=
=
=
=
=
rB
St
t
t
U
U0
UB
Uf
Ui
Vp
Vwall
=
=
=
=
=
=
=
=
=
=
Greek Letters
=
=
8
"0
a
g
=
=
=
=
=
=
p
y
=
=
=
=
=
c
g
=
=
coalescence
def
cr
vis
cap
wall
ext
=
=
=
=
=
=
=
deformation
critical
dimensionless, limiting value
due to viscous eects
due to capillary eects
at the wall
exterior
REFERENCES
Adams MJ, Briscoe BJ, Kamjab M. The deformation and
ow of highly concentrated dispersions. Adv in Colloid
and Interface Sci 44:143, 1993.
Campbell CS. Rapid granular ows. Ann Rev Fluid Mech
22:57, 1990.
Campbell CS, Brennen, CE. Computer simulation of granular shear ows. J Fluid Mech 151:167, 1985.
Dencs B, Ormos Z. A particle size control in a uidized bed
spray-dryer and granulator during the recovery of solids
from liquids. Hungarian J Ind Chem Veszprem
21:225231, 1993.
Dencs B, Ormos Z. A Orlo-gorgos uidizacios-porlasztos es
granulalo berendezes fojadekok szarazanyag-tartalmanak
kinyeresere. J Hungarian Chem 7, 1994 (in Hungarian).
Dencs B. Personal communication, April 1998.
Ennis BJ, Tardos GI, Pfeer R. A micro-level-based classication of granulation regimes and mechanisms.
Proceedings of the Second World Congress on Powder
Technology, Kyoto, Japan, 1990, p. 409.
Ennis BJ. On the mechanisms of granulation. Ph.D. dissertation, City University of New York, 1990.
Ennis BJ, Li J, Tardos GI, Pfeer R. The inuence of viscosity on the strength of an axially strained pendular liquid
bridge. Chem Eng Sci 45:30713087, 1990.
Ennis BJ, Tardos GI, Pfeer R. A micro-level-based characterization of granulation phenomena. Powder Technol
65:257272, 1991.
Hancock BC, York P, Rowe RC. An assessment of substratebinder interactions in model wet masses. 1: Mixer
torque rheometry. Int J Pharm 102:167176, 1994.
Khan MI, Tardos GI. Stability of wet agglomerates in granular shear ows. J Fluid Mech 347:347368, 1997.
Kristensen HG, Holm P, Schaefer T. Mechanical properties
of moist agglomerates in relation to granulation mechanisms. Part I: Deformability of moist densied agglomerates. Powder Technol 44:227237, 1985.
Kristensen HG, Holm P, Schaefer T. Mechanical properties
of moist agglomerates in relation to granulation mechanisms. Part II: Eect of particle size distribution. Powder
Technol. 44:239247, 1985.
Kristensen HG, Schaefer T. A review on pharmaceutical wet
granulation. Drug Develop Ind Pharm. 13(4,5):803, 1987.
Kristensen HG. Agglomeration of powders. Acta Pharm Soc
25:187204, 1988.
18
Applications for Fluidized Bed Drying
Arun S. Mujumdar
National University of Singapore, Singapore
Sakamon Devahastin
King Mongkuts University of Technology Thonburi, Bangkok, Thailand
INTRODUCTION
Table 1
Criterion
Particle size
Particle size distribution
Drying time
Floor area
Turndown ratio
Attrition
Power consumption
Maintenance
Energy eciency
Ease of control
Capacity
Rotary
Flash
Conveyor
Fluid bed
Large range
Flexible
Up to 60 min
Large
Large
High
High
High
Medium
Low
High
Fine particles
Limited size distribution
1030 s
Large length
Small
High
Low
Medium
Medium
Medium
Medium
500 m10 mm
Flexible
Up to 120 min
Large
Small
Low
Low
Medium
High
High
Medium
1002000 m
Limited size distribution
Up to 60 min
Small
Small
High
Medium
Medium
High
High
Medium
ware as well as the operating conditions. It is important to point out that for removing organic solvents or
handling toxic or ammable solids, the conventional
hot air uidized bed dryer is not a good choice, since
there is danger of re or explosion if ammability limits are exceeded. Also, owing to the inherent need to
handle large volumes of air due to the need to achieve
uidization, complete collection of the toxic material is
not economically feasible.
low values. For certain polymer pellet applications, uidized bed dryers with immersed heat exchange panels
(nonadiabatic) are the only cost-eective means for drying. It should be noted, however, that heat transfer
considerations make this option most attractive only
for ne particles of the order of 100 m in size.
Table 2 summarizes the diverse variants of the uidized bed technique; not all the types noted are
commercially available, however.
Figure 3 shows the basic steps involved in the
selection of a dryer. It is beyond the scope of this presentation to discuss the selection of dryers from a very
large pool of possible dryers. The interested reader is
asked to refer to Mujumdar and Menon (1995) and
Baker (1997) for a more general discussion. Note
that while over four hundred dryer types have been
reported in the literature, over one hundred are sold
commercially. Careful selection of the dryer type is
crucial to the success of the nal design and optimization of the drying system.
For any dryer, the steps shown in Fig. 3 apply. Unless
reliable data exist for the same product it is strongly
recommended that laboratory scale tests be carried
out to verify that the material can be processed in a
uidized bed. The conditions for uidization (e.g., minimum uidization velocity) may be estimated using published correlations, but serious errors may accrue due to
the surface wetness of the particulate solid, which may
generally behave entirely dierently from a surface-dry
particle. Even trace amounts of surface wetness may
render the material unuidizable without mechanical
energy input. In the case of a uidized bed dryer with
immersed heat exchangers, it is obvious that the material must not stick to the heat exchange surfaces.
While the convective heat and mass transfer rates in
wet particle uidization may be estimated reasonably
well with dry particle correlations available readily in
the literature, particle-to-immersed-surface heat transfer rates are very sensitive to the presence of surface
moisture. Evaporation of the surface moisture may
lead to a severalfold increase in the heat transfer rate
until the particle is surface-dry, e.g., it reaches its critical moisture content or attains its equilibrium moisture content at the surface.
It is suggested that the laboratory scale tests be carried out in a chamber of at least 100 mm in diameter so
that the wall eects are not severe. When possible, pilot
scale tests are also recommended owing to the highly
nonlinear nature of the hydrodynamics. Especially
noteworthy is that the product quality (e.g., physical
size, size distributions, structure, texture, color) may
vary with the scale of operation depending on the pro-
Particulate ow regime
Processing mode
Fluidizing gas ow
Fluidizing gas temperature
Heat supply
Fluidization action
Fluidized material
Fluidizing medium
Number of stages
Main consideration
Design specications
Calculate in design; given in analysis problem
Depends on heat sensitivity of product
Maximum product temperature must be specied
Exit gas in equilibrium with bed
Equilibrium moisture content of product must be known
Related to attrition
Quality of product aected
Figure 5 Typical textbook batch drying rate curve under constant drying conditions.
4.2
In removing surface moisture, gas velocity has a dominant eect. For particles with high internal resistance
to moisture transfer, gas velocity has marginal to no
eect.
3.4
The eect is complex and depends on the relative signicance of external and internal resistances to moisture transfer. Higher external heat uxes can lead to
increased bed temperature, which in turn leads to
higher moisture diusivities and hence drying rates
(see Reay and Baker, 1985).
Here we summarize the key features of the more commonly used uidized bed dryers. Some of the more
innovative and specialized uidized bed dryer designs
are discussed in the nal part of this chapter.
4.1
The time required to remove a given amount of moisture increases as the square of the particle diameter for
a Geldarts type B particle, all other conditions being
equal. For type A particles the eect is much smaller.
Geldarts classication of powders is discussed in
Chapter 3, Bubbling Fluidized Beds.
3.3
Figure 7
acteristics. Suggested
stocks are listed in
materials, a suitable
aid of a decision tree
DESIGN CONSIDERATIONS
5.1
WMFBD
PFFBD
VFBD
SBD
Inert
particle bed
No
Inert
particle bed
Inert
particle bed
Wet particles
Free-owing
Cohesive
Yes
No
Yes
No
Yes
No
Yes
No
Particle size
Small
Medium
Large
Polydisperse
Yes
Yes
No
No
Yes
Yes
No
No
Yes
Yes
Yes
Yes
No
Yes
No
No
Moisture
Surface
Internal
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Product specication
Uniform moisture
Low moisture
Fragile
Heat-sensitive
No
No
No
Yes
Yes
Yes
No
No
Yes
Yes
No
No
No
No
No
No
Drying time
510 min
1060 min
60 min
Yes
Yes
No
No
Yes
Yes
No
Yes
No
No
Yes
Yes
Throughput
Low
Medium
High
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Immersed heaters
Yes
No
No
No
Temperature zoning
No
Yes
Yes
No
Flexibility in operation
No
Yes
Yes
No
Feed
Liquid/suspension/slurry
Dryer suggested
Fluid bed, spouted bed of inert particles
WMFBD/backmix with dry product
FBD or SBD
ms Cps
dt
ms Cps
Tin Tp t 0
t
log
Gg Cpg
Tin Tp t
Figure 8 Decision tree for selection of uidized bed dryers for granular feedstock. Note: Other dryer types are also suited for
several of the above applications. If the feedstock contains a solvent, uidized bed dryers are not recommended in general;
indirect or vacuum dryers are preferred instead.
5.2
tii ps pi X Xe Tib
1 1
t
dt
4
X0
Xt exp
m 0
m
where m is the mean residence time of particles in the
bed.
Reay and Allen (1982) discussed the diculty associated with the use of this equation. The batch drying
curve Xt must be obtained at a constant bed temperature corresponding to the temperature selected
for the continuous operation. This value is normally
not known at the start of the development process, and
in any case is extremely dicult to obtain experimentally. The experiment has to be performed with a high
initial inlet air temperature, which must then be
decreased at the later stage of drying in order to keep
the bed temperature constant at the desired value. This
is very dicult to achieve even with a sophisticated
automatic control system.
The problem was overcome, however, by Reay and
Allen (1982) following their derivation of a normalization rule for the eect of bed temperature on batch
drying curves for both type A and B materials. They
proposed, and veried experimentally, the following
steps for constructing a constant bed temperature
batch drying curve at any desired bed temperature
from a batch drying curve measured at any convenient
constant inlet air temperature.
1. Obtain a record of the changing bed temperature Tii during constant inlet air temperature
run.
Table 7
Numerous innovative designs and operational modications have been proposed in the literature while
fewer are readily available commercially.
Tables 7, 8, and 9 summarize the key features of the
innovative dryer concepts relative to the more conventional FBDs. Many of the ideas presented are new
and have not been investigated even at the laboratory
stage. It must be noted that not all the novel ideas are
necessarily better, but many are worth exploring, since
they appear to oer some distinct benets over the
more conventional ones.
6.1 Modied Fluidized Bed Dryers
Depending on special requirements it is possible to
develop modied versions of the traditional uidized
bed dryers. Two-stage uidized bed dryers consisting
of several dryers are now in commercial use. Some
examples of these combinations are given in Table 10.
Conventional
Convective heat transfer
Steady gas ow
Constant gas temperature
Pneumatic uidization
Used for drying of particles
Air/combustion gas as drying medium
Air drag resisted by gravity
Single stage/multistage uid beds
Simultaneous uidization of entire bed
Innovative
Convection conduction (immersed heaters in bed)
Pulsed gas ow
Variable gas temperature
Mechanically assisted uidization (vibration/agitation)
Drying pastes, slurries using inert media
Superheated steam for uidization/drying
Centrifugal uid beds (articial gravity generated by rotation)
Multistage with dierent dryer types
Moving uidization zone (pulsating uidized bed)
Innovative
Pneumatic spouting
Single spout
Constant gas ow/continuous spouting
Constant gas temperature
Drying particles
Spatially xed spout
Convective drying
Axisymmetric
Pulsing the uidizing gas ow has been demonstrated to enhance drying performance in some cases
at the expense of added capital costs. Periodic local
uidization of the bed by moving the distributor
zone at regular intervals (so-called pulsated uidized
beds) has the advantage of reduced energy and air
consumption when batch drying particulate materials
with most of the heat and mass transfer resistances
within the particles (Gawrzynski and Glaser, 1996).
For larger particles (e.g., Geldarts type D), intermittent spouting of the bed with a rotating spouting jet
has been shown by Jumah et al. (1996) to reduce
energy consumption at only a marginal increase in drying time for batch drying. Indeed, in this case, it is also
possible to introduce another type of intermittency,
e.g., periodic heating of the spouting air when drying
in the batch mode.
To conserve energy while reducing the drying time
in batch uidized bed drying one may remove the surface moisture using a higher gas temperature initially;
the gas temperature can be reduced progressively as the
material dries so that it does not exceed its maximum
permissible temperature (Devahastin and Mujumdar,
1999). This can be achieved with a fuzzy logic controller or with a model-based control scheme. In plug-ow
dryers this idea can be extended by temperature-zoning
6.2
Innovative
Variable gas ow
Variable gas temperature
Nonaerated
Combined conduction/radiation/microwave
Vacuum operation
Vertical spiral trough
Variable frequency/amplitude
Remarks
Reduces oor area/bed depth of each stage; low product
moisture content possible
Fast removal of surface moisture; reduced stickiness leads
to easy uidization/smaller uidized bed dryer
Signicantly reduced spray dryer size
Ease of uidization for high moisture content feed; uniform/
low nal product moisture content possible
A higher drying rate above a critical inversion temperature when removing surface moisture
Better quality product under certain conditions
Closed system operation to minimize air pollution
Osetting some of these advantages are some limitations, e.g.,
Higher product temperature
Higher capital costs compared to hot air drying
Possibility of air inltration making heat recovery
from exhaust steam dicult by compression or
condensation
Among industrial scale applications of superheated
steam uidized bed drying one may cite
Drying of pulverized coal (Faber et al., 1986)
Drying of high-moisture lignite using a uidized bed
with internal heat exchangers to improve energy
eciency and reduce dryer size (Potter et al.,
1986)
Drying of pulps, bagasse, sludges, hog fuel, spent
grain from breweries, etc., in a pressurized superheated steam dryer (35 bars) manufactured by
Niro A/S, Denmark (Jensen, 1992)
As may be expected, the net energy consumption of
such dryers can be very small. Up to 90 percent fuel
saving over conventional dryers has been reported for
drying of beet pulp in a high-pressure uidized bed
dryer (Bosse and Valentin, 1988).
6.3
be used to lower the operating temperature, thus reducing the probability of thermal degradation; it may
allow operation in conditions out of the ammability
limits. Arnaldos et al. (1997) have presented experimental results on the hydrodynamic behavior of
vacuum uidized bed dryers and their applications
for drying granular solids and coating of particles.
Note that the minimum uidization velocity increases
as the pressure decreases. The ow behavior at lower
pressures is quite dierent from that at atmospheric or
high-pressure operation. The uidization is progressive, and the concept of minimum uidization velocity
loses some of its signicance. Experimental measurements of the velocity for complete uidization are
needed for design purposes.
Little work is published on vacuum uidized bed
drying. Arnaldos et al. (1997) report vacuum uidized
bed drying data at 200, 400, and 500 mbar pressures.
Silica dp 975 m, s 1650 kg=m3 ; s 0:7) and
millet (dp 1800 m, s 1600 kg=m3 ; s 0:90)
were used as test materials. Interestingly, the highly
porous silica particles dried faster as the pressure was
lowered, while the denser millet particles dried slower
since much of its resistance to drying is internal. Thus
the eect of lower pressure depends on the internal
heat and moisture resistance of the particles. Because
of their inherently higher capital and operating costs,
vacuum uidized bed dryers have found application
mainly in the pharmaceutical industry.
CLOSING REMARKS
Fluidized bed dryers, both conventional and innovative, will continue to nd increasing applications in
various industries. For further details, the reader is
asked to refer to Vanecek et al. (1965), Gupta and
Mujumdar (1983), Pakowski et al. (1984), Reay and
Baker (1985), Hovmand (1995), Keey (1992), and
Mujumdar (1995), which contains an extensive anno-
NOTATIONS
A
B
Cpg
Cps
D
dp
Gg
L
mb
ms
pi
=
=
=
=
=
=
=
=
=
=
=
ps
S
Tib
=
=
Tii
Tin
Tp
t
X
Xe
=
=
=
=
=
X0
U
Umf
=
=
Subscripts
b
e
g
p
s
=
=
=
=
=
bed
equilibrium
gas
particle
solid
Greek Symbols
s
s
m
=
=
=
particle sphericity
particle density, kg m3
mean particle residence time, s
REFERENCES
Arnaldos J, Martra J, Casal J, Kozanoglu B. Vacuum uidization processes: drying and coating. Powder Handling
and Processing 9:315319, 1997.
Bahu RE. Fluidised bed dryer scale-up. Drying
TechnologyAn International Journal 12:329339,
1994.
Bahu RE. Fluidized bed dryers. In: Baker CGJ, ed. Industrial
Drying of Foods. London: Blackie Academic and
Professional, 1997, pp. 6589.
Baker CGJ. Dryer selection. In: Baker CGJ, ed. Industrial
Drying of Foods. London: Blackie Academic and
Professional, 1997, pp. 242271.
Borde I, Dukhovny M, Elperin T. Heat and mass transfer in
a moving vibrouidized granular bed. Powder Handling
and Processing 9:311314, 1997.
Bosse D, Valentin P. The thermal dehydration of pulp in a
large scale steam dryer. Proceedings of the Sixth
International Drying Symposium, Versailles, 1988, pp.
337343.
Devahastin S, Mujumdar AS. Batch drying of grains in a
well-mixed dryer-Eect of continuous and stepwise
change in drying air temperature. Transactions of the
ASAE 42:421425, 1999.
Faber EF, Heydenrych MD, Seppa RUI, Hicks RE. A
techno-economic comparison of air and steam drying.
In Mujumdar AS, ed. Drying86, Vol. 2. Washington:
Hemisphere, 1986, pp. 588594.
Gawrzynski Z, Glaser R. Drying in a pulsed-uid bed with
relocated
gas
stream.
Drying
Technology-An
International Journal 14:11211172, 1996.
Gupta R, Mujumdar AS. Recent developments in uidized
bed drying. In Mujumdar AS, ed. Advances in Drying.
Vol. 2. Washington: Hemisphere, 1983, pp. 155192.
Hovmand S. Fluidized bed drying. In: Mujumdar AS, ed.
Handbook of Industrial Drying. 2d ed. New York:
Marcel Dekker, 1995, pp. 195248.
Jensen AS. Pressurized drying in a uid bed with steam.
Proceedings of the Eighth International Drying
Symposium, Montreal, 1992, pp. 15931601.
Jumah RY, Mujumdar AS. A PC program for preliminary
design of a continuous well-mixed uid bed dryer.
Drying TechnologyAn International Journal 11:831
846, 1993.
Jumah RY, Mujumdar AS, Raghavan GSV. Control of
industrial dryers. In: Mujumdar AS, ed. Handbook of
Industrial Drying. 2d ed. New York: Marcel Dekker,
1995, pp. 13431368.
Jumah RY, Mujumdar AS, Raghavan GSV. A mathematical model for constant and intermittent batch drying of
grains in a novel rotating jet spouted bed. Drying
TechnologyAn International Journal 14:765802,
1996.
Keey RB. Drying of Loose and Particulate Materials.
Washington: Hemisphere, 1992.
19
Circulating Fluidized Beds
John R. Grace and Hsiaotao Bi
University of British Columbia, Vancouver, British Columbia, Canada
Mohammad Golriz
Umea University, Umea, Sweden
INTRODUCTION
The term circulating uidized bed, commonly abbreviated CFB, has been in common usage since the mid1970s, although the origins of the technology date back
to the 1940s for catalytic processes (Squires, 1994) and
to the 1960s for gassolid processes (Reh, 1971). The
term implies two complementary characteristics for
gassolid systems:
1. A conguration where particles, entrained at a
considerable ux from a tall main reactor or riser,
are separated eciently from the carrying uid, usually
external to the reactor, and returned to the bottom of
the riser, forming a recirculation loop for the particles.
Individual particles circulate around this loop many
times before leaving the system, whereas the uid passes
through only once. A typical setup is shown schematically in Fig. 1. As described in Sec. 4 below, many different geometric congurations are utilized in practice.
2. Operation at high supercial gas velocity (typically 212 m/s) and high particle ux (typically 101000
kg/m2 s) so that there is no distinct interface in the riser
between a dense bed and a dilute region above.
Contacting is therefore carried out at gas velocities
beyond the bubbling, slugging and turbulent uidization ow regimes, residing instead in a higher velocity
ow regimefast uidization, dense suspension upow,
or dilute pneumatic conveying (see Sec. 3 below).
APPLICATIONS
FLOW REGIMES
Key References
Comments
FischerTropsch synthesis
Maleic anhydride
Combustion of coal,
biomass, wastes, o-gases
Gasication
Calcination (e.g., of
aluminium trihydrate
and carbonates)
Catalyst regeneration
Roasting of ores
Reduction of iron ore
Smelter o-gas treatment
Flue gas dry scrubbing of
HF, HCl, SO2 , dioxins,
mercury, etc.
Applied in China
Applied in Australia
Lurgi plant in Trinidad
One plant in Australia supplied by Ahlstrom
Commercial units since the 1970s, primarily in
Europe
velocity,
Utr ,
of
A linear relationship between Gs and Ug in the highvelocity range with e constant in Fig. 3 suggests that
Uslip in Eq. (2) approaches a constant value, determined from the intercept and slope of the linear part
of the curve. This critical velocity, designated Use (Bi et
the gas velocity is reduced to reach the minimum pressure gradient point (Matsumoto and Harakawa, 1987).
Radial segregation into a dilute core and dense annulus
results from the balance between the force needed to
suspend the weight of the particles and wall friction,
and the transition velocity is inuenced by the unit
geometry.
Several correlations have been proposed to predict
the transition velocity, UCA (see Bi et al., 1993). The
correlations of Bi and Fan (1991),
q G 0:542
s
UCA 21:6 gdp
Ar0:105
4
g UCA
and Yang (1975),
4:7
2gD "CA
1
Figure 3 Denition of critical velocity, Use , of Bi et al. (1995).
3.2.1
UCA ="CA ut 2
6:81 105
2:2
g
p
0:064
1:064
g g g
UDSU 0:0113G1:192
s
g
7
3.3
A number of ow regime diagrams have been proposed, beginning with Zenz (1949), who plotted the
pressure gradient (dP=dz) vs. supercial gas velocity
(Fig. 5). The cocurrent gassolids ow region spans
the ow regimes encountered in the circulating uidized beds, with the lower limit set by the choking
velocity. There is an inoperable region between the
low-velocity uidization and cocurrent upow,
because the transport line is choked as discussed
above. Such a regime diagram has been further
extended to incorporate more subregions within cocurrent ow (e.g., Drahos et al., 1988). The pressure gradient (dP=dz) is almost proportional to the solids
fraction, since friction and acceleration/deceleration
terms can usually be neglected. One can alternatively
plot solids fraction or bed voidage (e.g., Li and
Kwauk, 1980) vs. supercial gas velocity (Ug ), normalized supercial gas velocity (Ug =ut ), or slip velocity.
490
Grace et al.
4.1
4.1.1
1 t
"r; z lim
"r; z dt
t!1 t 0
1 R
"r; z2r dr
"av z
R2 0
8
9
1
H
"av z
10
11
linear relationship for quartz particles in a water-uidized bed for "s 0:5; they then assumed linearity and
calibrated their probe by comparing the optical signals
to two concentrations obtained by -ray absorption.
Linearity has also been assumed by others (e.g., Kato
et al., 1991; Zhou et al., 1994). Werther et al. (1993),
using solidsliquid suspensions with FCC and sand
particles, suggested an empirical t of the form I
k1 "n and found that n was independent of the uid
properties. Nieuwland et al. (1996) obtained similar
results. Qin and Liu (1982) derived a power law relationship between the light intensity and the solids
holdup for uniform spheres by assuming that the
light is reected from the rst two layers of particles
only. Lischer and Louge (1992) found a similar power
law from in situ calibration using a capacitance probe.
A simulation of the output of a single ber immersed
in a suspension of spheres showed the output to be
inuenced by the particle size and the refractive
index of the suspending medium. Therefore calibration
with liquids may cause errors. Nonlinear calibration
curves were also obtained by Herbert et al. (1994)
and Issangya (1998), who placed an optical ber
ush with the inner wall of a tube and recorded the
signal as FCC particles fell uniformly.
An iterative procedure for obtaining the true (nonlinear) calibration curve based on a particle dropping
test was proposed by Zhang et al. (1998). Nonlinearity
and errors can be caused by particles traveling through
a blind region at the tip of the probe. This can be
avoided by adding a thin transparent window (Cui et
al., 2001; Liu, 2001).
Capacitance Probes are based on the principle that
solids inside a sensing volume alter the effective dielectric constant (permittivity) of the gas/solids mixture. Since uidized bed particles and gas are
typically nonconducting, capacitance is usually easier
to measure than electrical resistance or conductivity.
Various probe congurations have been used, ranging
from parallel plates to needle shapes. The instrumentation is simple, inexpensive, and easy to construct.
However, calibration is required for each bed material. Problems include signal drift, e.g. that due to humidity or temperature variations, powder buildup on
the probe, and difculty in delineating the exact measuring volume. Brereton and Grace (1993) found the
relationship between the voidage and the signal intensity of a needle probe to be nearly linear. Hage and
Werther (1997) applied a water-cooled guarded capacitance probe inside a high-temperature CFB combustor.
dP
12
g p 1 "av z g "av z
dz
1 dP
1 dP
"av z 1
or
"sav
g dz
g dz
13
By comparing values from this approach with those
obtained by other methods, e.g., quick-closing valves
(Arena et al., 1986), -ray absorption (Hartge et al.,
1986a; Azzi et al., 1991), or x-ray computed tomography (Grassler and Wirth, 1999), this approach has
been shown to give reasonably accurate measurements.
However, corrections may be needed in some cases:
The neglect of wall friction is reasonable for
large columns, but it can lead to signicant deviations for D (or equivalent diameter
[ 4 Area=Perimeter for noncircular cross
sections) < 0:1 m.
The neglect of particle acceleration eects is generally reasonable when the net solids circulation
ux, Gs , is less than about 200 kg/m2 s.
However, signicant errors are introduced at
the bottom, and, to a lesser extent, at the top
of the column, when the acceleration term is
ignored at higher Gs . Approaches proposed by
Weinstein and Li (1989) and Louge and Chang
(1990) can be used to estimate the corrections
when accelerational eects are appreciable.
In order to apply Eq. (13), pressure taps should be
provided at regular intervals covering the entire height
of the riser, beginning just above the gas distributor at
the bottom of the riser, with the distance between successive pressure taps small enough that the pressure
gradient term, dP=dz, can be estimated with little
error. The resulting axial prole of voidage (or, alternatively, solids holdup or pressure prole, from which
"av and "sav can be readily derived) then provides a
signature for the riser ow, indicating how the suspension density and solids holdup vary with height.
Some representative studies reporting measurements
of this nature are listed in Table 2. Most reported
measurements have been from small units with air at
atmospheric temperature and pressure as the uidizing
medium, but they have covered broad ranges of particle properties, Ug , and Gs .
Typical proles showing how the solids holdup varies with the net solids circulation rate for constant
supercial gas velocity in a column with a smooth
top exit appear in Fig. 7. Similar proles are seen for
very dierent particles (e.g., Jiang et al., 1994). The
Table 2 Representative Data Showing Axial Proles of Cross-Sectional Mean Voidage or Particle Concentration
Author
Li & Kwauk, 1980
Rhodes & Geldart, 1986
Bader et al., 1988
Mori et al., 1992
Bai et al., 1994
Brereton & Grace, 1994
Karri & Knowlton, 1997
Mastellone & Arena, 1999a
Issangya et al., 1999
Particles
iron, alumina,
FCC, pyrite
alumina
catalyst
FCC
FCC, sand,
alumina
sand
catalyst
glass, sand,
FCC
FCC
dp , m
U, m/s
Gs , kg=m2 s
D, m
H, m
Exit
geometry
54105
0.85.6
16135
0.09
smooth
64
76
54
70, 321,
633
148
76
67310
4
4.3, 9.1
1.01.5
1.53.0
45115
147
2.429
7128
6
12.2
1.55.5
3.0
abrupt
smooth
various
smooth
column geometry
binary mixtures
3.79.2
3.08.1
36
989
98586
35144
0.152
0.305
0.050.1
0.097,
0.15
0.152
0.3
0.12
9.3
13
5.75
various
smooth
smooth
70
48
18425
0.076
6.1
smooth
Figure 7 Axial proles of solids holdup for FCC particles in riser of diameter 0.1 m and height 16 m with a smooth unconstricted exit. dp 67 m, p 1500 kg=m3 , Ug 3:0 m=s. (Huang and Zhu, 2001.)
14
16
0:103
17
"s0
Gs
g
s
p Ug p g 0:082
"s
0:208
18
"s0
Gs
g
where "s0 Gs =p Ug ut is the solids holdup
expected for the ideal case where all particles travel
with a velocity equal to Ug minus the terminal setting
velocity, ut . Huang and Zhu (2001) found that the
distance required to achieve fully developed voidage
or solids holdup proles in a riser of small D increases
with increasing Gs and decreases with increasing Ug .
Models that predict axial voidage proles in addition
to other hydrodynamic properties are covered in Sec.
4.6 below.
4.1.3
Measurements based on a variety of probes (e.g., capacitance and optical), as well as x-rays, -rays, and
tomographic techniques, have shown that particle concentrations near the wall tend to be much greater than
in the interior of CFB risers. Table 3 summarizes some
major sources of experimental data. An alternative
tabulation, also listing the experimental techniques
employed, was provided by Xu et al. (1999). Typical
results appear in Fig. 9. Note that in this gure and
some other measured proles, the shape is like the
letter W (or an M, if the voidage, rather than the solids
holdup, is plotted). There is slight asymmetry in this
particular prole, due to the side top exit (as discussed
below). Except in the bottom zone (Rhodes et al.,
1998) or at very high uxes where a dierent ow
Table 3 Representative Data Showing Radial Proles of Local Voidage in CFB Units
Author
Particles
dp , m
Ug , m/s
Gs , kg=m2 s
D, m
H, m
z, m
sand
sand
FCC
catalyst
FCC
polymer
glass
FCC
glass
56
56
54
76
60
325
110
54
60
3.9
5.0
1.43.7
3.7
4.4
2.4
2.56.5
2.34.6
28
100
75
591
98
2652
721
30107
40180
30600
0.050
0.40
0.09
0.305
0.140
0.102
0.411
0.186
0.19
3.3
7.8
10
12.2
5.1
6.3
8.5
8
15
0.62.2
0.656.0
2.25
4.0, 9.1
4
0:6
36
2.36.3
4.411.6
5:08:8
19
20
21
Measurement Techniques
1
ttot
tot
f1 t f2 t dt
22
Experimental Results
23
Such coreannulus patterns have been observed frequently in small-scale risers, but also in much larger
equipment, even in a 14:7 11:5 m cross section combustor operated by Electricite de France (Caloz et al.,
1999).
Typical lateral proles of time mean local particle
velocity, both upward and downward, along a midplane of a column of square cross section are shown
in Fig. 10. The supercial gas velocity is seen to have
little inuence. Ascending particles are dominant in the
center of the column, whereas there are more descending than ascending particles near the wall (i.e., as y=Y
approaches 1). The magnitudes of the velocities of
rising particles at the axis of the column are similar
Table 4 Representative Data Showing Radial Proles of Particle Vertical Flux or Velocity
Author
Monceaux et al., 1986
Bader et al., 1988
Hartge et al., 1988
Rhodes et al., 1989
Azzi et al., 1991
Herb et al., 1992
Yang et al., 1992
deDiego et al., 1995
Wang et al., 1996
Wei et al., 1997
Nicolai & Reh, 1997
Corenella & Deng, 1998
Issangya et al., 1998
Mastellone & Arena,
1999a, 1999b
Caloz et al., 1999#
Particles
FCC
catalyst
FCC, ash
alumina
FCC
FCC, sand
FCC
sand, coal
sand
FCC
glass
sand
FCC
FCC, glass,
sand
glass
dp m
Ug , m/s
Gs , kg=m2 s
D, m
H, m
z, m
59
76
85, 120
70
75
68, 276
54
400650
120
54
40300
200
70
70310
3.2
4.6
2.9, 3.7
34
6.2
2.46
1.56.5
45.5
2.5
1.910.1
1.59.6
3.45.4
4.57.5
6
3168
147
49, 30
3040
150
2044
625
2898
up to 26
37236
NA
1030
38325
35, 55
0.144
0.305
0.4
0.152
0.190.95
0.05, 0.15
0.140
0.10
0.161
0.186
0.41
0.114
0.076
0.120
8
12.2
8.4
6
11.7
2.7, 10.8
11
4
6.2
8.5
8.5
2.7
6.1
5.75
4:8
9.1
0.94.7
1.74.9
3.116
1.5, 2.1, 5.5
3.3, 6.6
2.13.2
1.3, 4.4
1.8, 3.9
4.04.2
0.7, 2.5
2.8
1.24.2
62 for 0.41
083 m units
1.73.3,
5:5
1798
0.41, 0.83,
14:7 11:5
8.5, 11.2,
35
2.621.5
Studies that report local particle ux data are summarized in Table 4. For the fast uidization ow regime,
radial proles of local particle ux are generally similar
in shape to the radial proles of particle velocity
described aboveupward near the center line, dropping o with increasing r, and negative near the wall.
One such experimental prole, shown in Figure 11, is
seen to be symmetrical around the axis of the column.
Note that the ux is downward in this case only in a
thin region adjacent to the outer wall of the column.
q
z
0:5 1 1:34 1:301 "av z0:2 1 "av z1:4
D
0:80 "av 0:9985
24
Lateral Flux
Gas Velocities
While most CFB hydrodynamic studies have been carried out in risers of circular cross section, some have
been conducted in risers of rectangular (e.g. Saberi et
al., 1998) and square (e.g., Leckner et al., 1991; Zhou
et al., 1994) cross section, geometries that are especially
relevant to CFB combustors. Some work (e.g.,
Yerushalmi et al., 1978; Arena et al., 1992) has also
been performed in thin (two-dimensional) risers.
Phenomena in rectangular columns are qualitatively
similar to those in risers of circular cross section,
with the usual dilute upward moving core and dense
downward ow in an outer annular region.
Unlike low-velocity uidized beds, where the distributor is often regarded as the most critical element in
achieving a successful design, there is much greater
latitude when specifying the distributor and bottom
geometry for CFB risers. In fact, it is possible to operate without any distributor plate whatsoever, so long
as the solids can be blown over at the end of each
run to be stored in the return system (i.e., the standpipe
and storage vessel, if any). Nevertheless, the inlet conguration plays a signicant role, and it would be
wrong to consider the design of this region unimportant.
CFB systems have various congurations for recirculating and reinjecting solids. The feed rate may be
regulated by a mechanical (e.g., slide valve) or nonmechanical valve (see Chapter 11). L-valves, H-valves,
L-valves, V-valves and loop seals are common in CFB
combustors, while J-valves and slide valves are popular
in uid catalytic cracking risers. Nonmechanical valves
tend to give good ow control for Geldart group B and
D particles, while the ow of aeratable group A particles usually needs to be regulated by a slide valve or
other mechanical valve. Some common bottom geome-
tries are illustrated in Fig. 15. Cheng et al. (1998) classify inlets according to the degree of restriction of the
solids circulation and the conguration of the open
area. A strong restriction at the entry helps stabilize
the ow, but adds to the system overall pressure drop
and hence to the operating cost. Weak restrictions
cause the riser to be directly coupled with the entire
return system and with the windbox (if any), whereas a
strong restriction (high distributor pressure drop)
allows the riser to operate somewhat independently.
Hartge and Werther (1998) provide useful information
on dierent nozzle types and their inuence on erosion.
To reduce erosion they recommend nozzles that have a
section where there is downward gas ow with a minimum exit velocity.
The geometry and orientation of the entering gas
can directly aect axial voidage proles over a considerable height beginning in the lower part of the riser
(Weinstein et al., 1996; Cheng et al., 1998; Kostazos et
al., 1998). The bottom zone may behave as a vigorously bubbling bed or as a turbulent uidized bed. It
may be clearly demarcated from the region above,
merge smoothly into the upper part of the riser, or
show a transitional splash zone between a lower
dense bed and the riser above. Addition of secondary
gas hastens the transition from a denser bottom zone
to a more dilute upper region. For low-pressure drop
distributors, as in the Chalmers University CFB boiler,
surging may occur, with pressure oscillations in the
windbox directly aecting the lower zone of the riser,
being a source of turbulence (Sterneus et al., 1999) and
causing exploding bubbles (e.g., Johnsson et al.,
1992; Schouten et al., 1999). This eect can be cor-
4.5.5
Exit Effects
The ow at the top of a riser can be aected profoundly by the geometry of the exit port through
which the gas and solids pass to the primary cyclone
or other separator. Some congurations employed in
practice are illustrated schematically in Fig. 16.
Important factors include
The asymmetry of the exit: An exit at one side tends
to cause asymmetry, as noted above.
The degree of constriction: A severe reduction in the
cross-sectional area leads to a higher exit pressure drop, aecting the overall pressure balance,
also reducing the eux of solids from the riser
for a given gas owrate.
The supercial gas velocity: Exit eects are more
important as Ug increases.
The tortuosity of the ow path: The exit acts like a
crude gassolid separator. Any increase in curvature of the gas streamlines as the gas nds its way
into the exit makes it more dicult for the particles to remain entrained and increases internal
reuxing of solids in the riser.
For horizontal exits, solids may build up in the exit
duct and dump periodically back into the main
riser (van der Meer et al., 1997).
Given these factors, the exit design should be central
and smoothly tapered if one wishes to minimize backmixing of gas and solids and prevent reection from
the top. However, if the designer wishes to increase
solids holdup in the riser to provide maximum resi-
Figure 16 CFB riser top exit congurations: (a) sharp rightangle; (b) exit at side; (c) tapered right angle; (d) long-radius
bend; (e) nozzle and low pressure drop impingement separator [gas exit not shown]; (f) top exit with constriction.
Internals
Geometric Variants
Most approaches assume a sigmoidal voidage prole and/or perfect coreannular ow; as noted
above, these are oversimplied.
Axial symmetry is usually assumed, making it
impossible to deal adequately with asymmetries
such as side feed ports, side reentry of solids, and
asymmetric top exits.
The models do not readily allow for the introduction of secondary or tertiary air (or other gases).
However, mechanistic approaches provide rational
approaches for predicting key hydrodynamic features.
Two useful models will be outlined briey.
4.6
Hydrodynamic Models
Empirical Correlations
Mechanistic Models
Mechanistic models are generally preferred to empirical approaches. Ideally they incorporate the key physical factors aecting the ow, leaving out factors of
lesser importance. Mechanistic models for CFB risers
are often useful, but many suer from weaknesses,
such as
It is frequently dicult to set the boundary condition at the bottom, i.e., at z 0.
U g p
Gs Ug p
25
26
27
5:6
0:47Fr0:41
t
Fr
28
29
Gs
p 1 "av
30
Table 5
Aspects covered
Reference
g dp Ug
Rep
g
D
dp
Gs
p Ug
31
Gs
p Ug
Ug
ut
Rep < 4
32
Gs
p U g
p
g
33
encountered in CFB risers. The omission of the coecients of restitution and friction and the particle size
distribution, all of which relate to particleparticle
interactions, may explain why experimental attempts
(e.g., Glicksman et al., 1991) to verify scaling based
on any of the above lists (or similar sets of groups)
have generally produced disappointing agreement.
Nevertheless, the above sets of groups give a reasonable starting point for experimental attempts to
achieve hydrodynamic scale modeling. For example,
this approach has been used to model the eects of
temperature and pressure on CFB hydrodynamics
(Chang and Louge, 1992; Louge et al., 1999; Younis
et al., 1999).
Scaling may also be based on hydrodynamic models
(e.g., Winter et al., 1999). For example, the mechanistic
models covered in Sec. 4.5.2 incorporate large-scale
data in empirical correlations and mechanistic models
used to predict hydrodynamics of CFB risers covering
a wide range of sizes.
5.1
Early studies of CFB hydrodynamics were dominated by modest net particle uxes (Gs from 10
to 100 kg/m2 s and solids holdups in the upper part
Figure 19 Axial prole of solids holdup for a range of net circulation uxes in a riser of diameter 76 mm with a smooth
unconstricted exit. dp 70 m, p 1600 kg=m3 , Ug 8 m=s. Highest uxes correspond to dense suspension upow. Note that
the entrance region was dierent for the highest solids ux case. (Issangya et al., 1999; Liu, 2001.)
6
toward the outer wall, partially osetting the upward
ux there. Typically the local net ux is 1.52.5Gs at
the axis of the column (Issangya et al., 1998; Karri and
Knowlton, 1999; Liu, 2001) falling to nearly 0 at the
wall. Results from a column of diameter 0.3 m at high
Ug in Fig. 21 indicate that the radial net ux proles
can reverse, showing a minimum at the axis of the
column.
6.1
MIXING
Gas Mixing
Figure 20 Local time mean particle velocity and standard deviation of particle velocity at z 4:2 m in 76 mm diameter, 6.4 m
diameter column operated at Ug 6 m/s, Gs 417 kg/m2 s. (Liu, 2001.)
Figure 21 Radial proles of local solids ux reported by Karri and Knowlton (1999) for Gs 586 kg/m2 s, z 3:7 m in a 0.3 m
diameter, 13 m long column. Numbers shown on gure are supercial gas velocities.
The overall mixing behavior of the riser can be characterized by the residence time distribution (RTD)
curve or an eective dispersion coecient. To obtain
34
Ug2 D2
Dgr
36
where M is a constant which characterizes the nonuniformity of the axial velocity prole.
6.1.2
37
Gas Backmixing
d 2 C Ug dC
0
dz2 "av dz
38
at
z0
39
and
Equation
Pegr 500 1000
Koenigsdor and
Werther (1995)
Gayan et al. (1997)
Namkung and
Kim (2000)
Dgr 43:5
Ug
1 "av
0:7
"av
Ug ="av Gs =p 1 "av
Pegr 6:46"av
Gs g Ug D 0:297
g
1:3
Gs g Ug D 1:14
dp
Gs 1
D1:85
p
g
0:96 3:73
Ug
Udp
dp
153:1
Dgr
D
Usl
Comment
For small columns only.
No column diameter eect.
For small columns only.
That Pegr increases with Gs is inconsistent
with other experimental data.
For dilute ow only with 1 "av < 0:01.
Ug
Gs
"av p 1 "av
C0
at
z 1
40
41
Axial gas dispersion can be studied with pulse or stepchange injection (van Zoonen, 1962; Li and Wu, 1991;
Luo and Yang, 1990; Martin et al., 1992; Bai et al.,
1992b). Transient one-dimensional dispersion can be
described (Li and Wu, 1991) by
related to the axial gas dispersion coecient for relatively low dispersion by
Dgd "av
Dgd "av 2
2t
8
2
45
Ug L
Ug L
2
In all reported pulse tests, the tracer was injected on
the axis, with the sampling port downstream on the
axis. The tracer concentration then reects the local
response on the axis, not the response from the entire
cross section. The dispersion coecient, Dgd , obtained
in this way is the same as the one-dimensional dispersion coecient dened by equation (37) only for risers
of small diameter because of the strong radial variation
of tracer concentrations downstream of the injection
point.
Dgd generally increases with increasing Gs at given
gas velocities for Gs < 200 kg/m2 s. For Gs > 200 kg/
m2 s, van Zoonen (1962) reported a decrease of Dgd
with increasing solids ux. van Zoonen (1962) and
Luo and Yang (1990) found that Dgd increases with
increasing Ug at constant Gs for Gs > 50 kg/m2 s. Li
and Wu (1991) and Bai et al. (1992) reported the opposite trend for Gs < 50 kg/m2 s.
Luo and Yang (1990) found that Dgd decreases with
increasing particle size at given Ug and Gs . The eect of
column diameter is unclear. Li and Wu (1991) correlated their data by
Dgd 0:195"4:11
av
46
@C
@C
@C
Ug
Dgd 2 0
@t
@z
@z
43
For pulse injection with open-open boundary conditions, the solution of equation (43) is
"
#
1
1 2
C p exp
4"av Dgd =Ug L
2 "av Dgd =Ug L
44
where L is the distance between the injection and the
sampling point for single point sampling or the distance between the two sampling locations for doublepoint sampling. The variance of the RTD curve is
48
@Ca
@2 C
2K R
Dgza 2a 2 ca c2 Cc Ca 0
@t
@z
R Rc
49
Solids Mixing
RTD curves of solid tracer particles have been analyzed to obtain an axial dispersion coecient assuming
axially dispersed one-dimensional ow (Smolders and
Baeyens, 2000b),
@C
@2 C
@C
Dpz 2 vp
@t
@z
@z
50
t
and
Pepz
vp L
Dpz
52
where is the mean residence time, E is the dimensionless RTD function, and L is the distance between
the tracer injection and detection points. The mean
18:3
Gs =Ug 0:8
53
predicts the opposite trend. Equation (53) is recommended, although the eect of D is unclear.
6.2.2
55
and
Pepr 150 5:6 104 1 "av
56
57
@Ga fa
G @f
Ica fc Iac fa a a
vpa @t
@z
58
7
7.1
HEAT TRANSFER
Introduction
Favorable heat transfer is a major reason for the success of CFB reactors in such applications as combustion and calcination. CFB heat transfer has previously
been reviewed by Grace (1986b), Glicksman (1988),
Leckner (1990), Yu and Jin (1994), Basu and Nag
(1996), Glicksman (1997), and Molerus and Wirth
(1997). Temperature gradients tend to be small within
the riser due to vigorous internal mixing of particles,
while the gas and particle temperatures are locally
nearly equal, except at the bottom (Grace, 1986;
Watanabe et al., 1991). Because of wear caused by particle impacts, heat transfer surfaces are predominantly
vertical, and only this case is considered here. Many
CFB combustion systems are also equipped with external uidized bed heat exchangers, cooling solids returning to the bottom of the riser while they are being
recirculated in the external loop. However, these operate as bubbling beds exchangers, and the reader is
referred to Chapter 3. In this section we consider the
case where the bulk solids are hotter than the wall, as in
CFB combustion. It is straightforward to change to the
case where the transfer is in the opposite direction.
Suspension-to-wall heat transfer is controlled by the
ow of the solids and gas near the wall. Most studies
have employed small-scale columns at atmospheric
temperature and pressure. For simplicity, the experimental risers have generally been idealized in that they
have been (Grace, 1996):
Smooth-walled, with roughness elements much
smaller than the diameter of the particles
Circular in cross section, i.e., without corners or
protruding surfaces
Vertical, i.e., with no tapered or oblique sections
Supplied with air only through a at multiorice
distributor at the bottom of the riser.
Most boilers operate at 800 to 900 C, but few experimental investigations have been carried out at elevated
temperatures. Industrial CFB combustion systems
have risers that are usually rectangular in cross section,
manufactured from membrane water walls. Surfaces
may be rough, especially where refractory is present.
The lower section is commonly tapered or constricted.
In addition to primary gas from the base, secondary
gas may be injected through nozzles at higher levels.
Given these factors, data obtained in laboratory CFBs
are not always representative of industrial-scale CFB
reactors.
Heat transfer models are usually written in terms of
either clusters or dense wall layers, based on the hydrodynamics of fast uidization. For cluster models (Fig.
26), heat can be transferred between the suspension
and wall by (1) transient conduction to particle clusters
arriving at the wall from the bulk, supplemented by
radiation; (2) convection and radiation from the dispersed phase (gas containing a small fraction of solid
material). The various components are usually
assumed to be additive, ignoring interaction between
the convective and radiation components.
7.2 Convection
Any portion of the wall comes into intermittent
contact with clusters of particles, interspersed with periods, usually brief, where that portion of the wall is in
contact with relatively dilute phase. Heat transfer from
the descending clusters is commonly described by
unsteady, one-dimensional conduction normal to the
wall. Because this transfer relies on renewal of fresh
particles from the bulk, this heat transfer mechanism
is often referred to as particle convection, even though
the wall. Experimental data from large-scale hightemperature units are plotted in Fig. 27, with corresponding operating conditions summarized in Table
7. The wall-to-suspension heat transfer coefcient increases strongly with increasing susp , primarily because of the particle convection term.
the transfer between the cluster and the wall is primarily by conduction through the intervening gaseous
medium.
7.2.1
Figure 27
Heat Transfer Surface: The walls of CFB combustion chambers consist of membrane waterwalls,
i.e., parallel tubes connected longitudinally by ns
(Fig. 28). The protected channel between tubes is a
region of augmented particle downow and greater
particle concentration than the exposed tube surfaces
(Wu et al., 1991; Golriz, 1994; Lockhart et al., 1995;
Zhou et al., 1996b). Particles descend faster and travel further along the ns on average before being
stripped off and returning to the core (Wu et al.,
1991; Zhou et al., 1996b). These ndings help explain
the distribution of local heat transfer coefcients
around membrane surfaces (Andersson and Leckner,
1992; Golriz, 1994; Lockhart et al., 1995).
The conductivity and thickness of the membrane
water wall surface also play a role (Bowen et al.,
1991), because heat transferred to the n must be conducted to the tube surface and then into the coolant
(usually boiling water). In practical systems, the heat
transfer coecient on the coolant side is almost always
much higher than on the suspension side, so that the
overall heat transfer coecient is barely aected by the
coolant ow rate. A model that accounts for the coolant-side resistance, as well as the suspension-side and
Heat transfer coecient vs. suspension density for large-scale CFB risers. For operating conditions see Table 7.
Figure 28
Size, m
Ug , m/s
Tb , C
dp , m
Tw , C
4 4 23
4:6 12 33:5
8:6 11 33
6:9 7:4 34
1:4 1:7 13:5
7:6 5:2 31
5:1 5:1 28
3:0 6:0 20
6.4
35
NA
2.65.1
1.86.1
6.2
6.3
5.1
875
700860
850
774913
640880
850
855
920
200
280
140
150
220440
NS
220
400
500
360
340
330
210
340
340
290
crease in heat transfer coefcient with increased temperature is predominantly due to increased radiation
(see below).
Particle Size: The convective heat transfer coefcient decreases with increasing particle diameter
(Stromberg, 1982; Fraley et al., 1983; Basu, 1990) for
small heat transfer sections where clusters have brief
residence times on the heat transfer surface. The inuence of dp diminishes with longer surfaces, since
the contact times on the surface are then much higher
than the thermal time constant for individual particles (Wu et al., 1989). Andersson (1996) found that
the vertical distribution of heat transfer coefcients
in a 12 MWth CFB boiler are independent of dp for
constant Ug /ut .
Figure 30 Temperature prole at n, crest of tube of membrane tube wall, and refractory wall from Golriz (1992). Ug
3:6 m/s, dp 270 m, Tb 860 C. Dashed line indicates
location of tube crest.
7.2.3
Empirical Correlations
59
Radiation
Some Experimental Investigations
Relatively few studies have been carried out at temperatures high enough (> 600 C) that radiation
plays a signicant role. Basu and Konuche (1988)
and Wu et al. (1989) found that the radiative heat
transfer coecient increased with increasing temperature, while the contribution of radiation to the total
heat transfer increased with decreasing suspension density. In dierent studies, the radiative component has
been found to contribute more than 50% of the total
heat transfer at low particle loadings (e.g., Basu and
Konuche, 1988), whereas the contribution is 15 to
40% for higher suspension densities (e.g., Han and
Cho, 1999; Luan et al., 1999).
7.3.2
Packet Models: Models featuring packets constitute the most common type of model. These usually assume that the radiative component is the sum of
radiation from clusters and from the dispersed phase.
The radiative heat transfer between clusters and the
Table 8
z
T Tw
T
1 0:023Rep 0:094 b 0:294
H
Tb Tw
Tw
x
exp 0:0054
dp
60
For short surfaces, temperature variation in the emulsion layer is ignored. Temperature gradients in the core
are ignored in all cases. The gas and particle temperatures are equal locally. The heat transfer surface is a
gray, diuse surface. The emulsion is taken as a nongray, absorbing, emitting, and scattering medium.
Radiation is neglected in the vertical direction.
Investigators
Correlation
Suspension density,
kg=m3
Bulk temperature
C
hov 400:5
sus
hov 300:5
sus
580
750895
hov 88 9:450:5
sus
770
800850
hc 23:20:55
sus
5500
Cold condition
Andersson (1996)
hov 700:085
sus
> 2 kg=m3
637883
hov 580:36
sus
2 kg=m3
Figure 31
61
4. Estimate the fractional wall coverage by clusters. For this purpose we recommend an equation
(Golriz and Grace, 2002) tted to the data plotted by
Glicksman (1997) that includes the inuence of column
diameter, D, while also approaching the limiting value
of 1 in a rational manners as D increases:
2
f 1 exp 25; 0001 "av 1 0:5D
e
e0:5D
62
5. The gas convection coecient (hcg ) is evaluated
using a standard steady state correlation for internal
Figure 32
Te
Tb R56 Tw R34
R34 R56
and
R56
with R34
R5 R6
R5 R6
66
2Tb4 Te4
2=esus 1Tb Te
67
hrad
ew
4Te4 Tw4
1=esus 1=ew 1Te Tw
68
f
1
Gsh Cpp hrad
be
65
Predicted heat transfer vs. experimental data. Dashed lines indicate 25% deviations.
R3 R4
R3 R4
1
kg =
g hrad
ew
69
Dw Gsh Cpp hrad
be Tb Te
70
where Dw D 2
z, with
z given by Eq. (24).
Beginning at the top of the heat exchange surface
where the boundary condition Te Tb is imposed,
one can integrate this equation downwards to the bottom of the surface. Numerical integration is required
for small risers, not only because of the nonlinearity in
Te of the radiation terms in Eqs. (67) and (68), but also
because of variations of susp and
z with z. For the
special case where all radiation terms are unimportant
and where there is negligible variation of other terms
with height, Eq. (70) can be integrated to give
z=zth
Te Te1 Tb Te1 e
71
where
Te1
72
and
zth
73
74
MASS TRANSFER
dC
a1 "av km C Cs 0
dz
76
Figure 33 Nonuniform radial proles of concentration determined experimentally in CFB reactors: (a) concentration of ozone
undergoing decomposition in 0:25 m dia 10:5 m tall riser (Ouyang et al., 1995a); (b) concentration of NOx in
0:15 0:15 m 7:3 m CFB pilot scale combustor (Brereton et al., 1995).
Table 9 Examples of Reactions Considered in CFB Reactor Studies and Dimensions of Columns Used in the
Experimental Studies
Reaction
M model,
E Exptl
A. Catalytic Reactions
Claus reaction
FischerTropsch process
Fluid catalytic cracking
Methanol-to-olens
Maleic anhydride from n-butane
Oxidative coupling of CH4
Ozone decomposition
Gasication
Roasting
NA not applicable.
M
E
M, E
M
M
M
M
E, M
M
E
M, E
M, E
M, E
E
M
M
References
Puchyr et al., 1996, 1997
Shingles & McDonald, 1988
Fligner et al., 1994
Gao et al., 1999
Schoenfelder et al., 1994
Gupta et al., 1999
Pugsley et al., 1992
Golbig and Werther, 1996
Pugsley et al., 1996
Jiang et al., 1991
Ouyang & Potter, 1993, 1995a,b
Schoenfelder et al., 1996
Mao et al., 1999
Bolland & Nicolai, 1999
Pugsley & Malcus, 1997
Matsukata et al., 1995
E
M, E
M, E
M, E
M, E
M, E
M, E
E
Riser dimensions
NA
0.1 m id 13 m, then scaled up by
factors of 500 & 2.5
0.3 m id 12 m
NA
NA
NA
NA
21 mm i.d.
NA
0.10 m id 6:3 m
0.25 m id 10:9 m
0.4 m id 15:6 m
0.25 m id 10:5 m
411 mm id 8:5 m
NA
NA
0:15 m sq: 7:3 m
0.20 m square
0:20 m sq: 6:5 m
0:51 m sq: 14:3 m
3:96 m sq: 21:4 m
2:5 m2 13:5 m
11 m2 17 m
56 m2 33:5 m
1:6 m sq: 13:5 m
0:4 m sq: 20 m
0:3 m id 8 m
150 mm id 6:0 m
One-Dimensional Models
are completely ignored, with the gas and particle composition, as well as all hydrodynamic variables, treated
as invariant across the entire cross section at any level.
Some such models also assume that the hydrodynamics
are uniform in the axial direction, while others account
in some manner for the usual decrease in suspension
density with height and/or the presence of a relatively
dense zone, possibly subject to a dierent ow regime,
at the bottom. Key gas mixing and hydrodynamic
assumptions in one-dimensional models are summarized in Table 10. Gas mixing assumptions cover a wide
spectrum from plug ow to relatively well mixed.
One-dimensional models are overly simplied
except to provide rough estimates. In practice, the pronounced lateral gradients in CFB risers have important
consequences for reactions. For example, high particle
concentrations in the outer region can lead to elevated
local conversions of gas species toward the wall. When
there is downow of solids at the wall (i.e., fast uidization), concentration gradients can be pronounced,
and substantial backmixing occurs adjacent to the
wall. In CFB reactors with secondary gas addition
above the bottom of the reactor, gradients associated
with the wall jets can persist over considerable heights,
possibly even to the top of the reactor, with lateral
dispersion too slow to obliterate plumes originating
from the secondary jets. Lateral temperature gradients
can also be signicant.
Models that ignore axial gradients are further
removed from reality and are likely to require correction factors (e.g., Ouyang et al., 1993) to t experimental data. Models that account for axial variations in
hydrodynamic variables have found a variety of ways
of doing so, e.g.,
Axial dispersion
PF
PF, ADPF
PF, PM
PF, ADPF& (PM PF)
PF
PF
TIS
TIS
TIS PF
BB TIS
TIS PF
None
None
None
Solids holdup decays exponentially
"av f z
"av f z
"av f z
"av f z (each cell uniform)
Two separate zones
Two separate zones
5 compartments corresponding to regions
bounded by ring baes
ADPF axially dispersed plug ow; BB bubbling bed; PF plug ow; PM perfect mixing; TIS wellmixed tanks in series.
CoreAnnulus Models
2Kca Rc Ca Cc
kn 1 "a Can 0
R2 R2c
78
2
Core:
Ug R dCc 2Kca Ca Cc
Rc
R2c dz
kn 1 "c Ccn 0
79
Conversion
80
Core
Annulus
Axial gradients
PF
PF
Stagnant
Stagnant
PF
PF
PF
None
Allows separately for
ne & coarse particles
PF CA: "av f z
None
None
Stagnant
Stagnant
Intermittent
up and down
Ouyang & Potter, 1994
PF
PF rel. to solids None
Talukdar et al., 1994
PF
PF
TIS (turbt. & FF zones)
Schoenfelder et al., 1994 PF rel. to PF rel. to solids 4 zones in series
solids
Kunii & Levenspiel, 1998
PF
Stagnant
CA 1-D decay zone
Ignored
Puchyr et al., 1997
PF
PF rel. to
descending
solids
Interregion
mass transf.
Core radius/
column radius
Fitted
Rapid
0.85
Varies with height
Fitted
Fitted
ka 0:23 s1
by tting
Fitted
Varied, then tted
From gas mixing
expts
To be tted
Modied Higbie
penetration theory
Presumably tted
Fitted
0.85
NS
NS
Fitted
To be tted
Werther correlation
CA core/annulus; FF fast uidization; NS not specied; PF plug ow; PM perfect mixing; TIS tanks-in-series.
Figure 34 Conversion vs. dimensionless rate constant from simple one-dimensional, two-region, axially uniform model for rstorder chemical reaction with dierent overall voidages of 0.90 and 0.96 and dierent values of the dimensionless coreannulus
interregion mass transfer coecient.
Figure 35 Schematic representation of several multizone models: (a) coreannulus freeboard approach of Kunii and
Levenspiel (1997); (b) four-zone approach of Neidel et al. (1995); (c) ve-zone approach of Schoenfelder et al. (1994).
series of equations are written that combine hydrodynamic balances with empirical correlations to give
methods for predicting the evolution of voidages and
ows as a function of the axial coordinate, z. This
approach also incorporates correlations for a slip factor, discussed in Sec. 4.6.2 above.
9.4
The coreannulus models treated in the previous section improve on the one-dimensional models covered
in the preceding section by making some allowance for
the dierence in behavior between the relatively dense
wall region and the dilute core of fast uidized beds.
However, the hydrodynamics are represented in a relatively crude manner. As illustrated in Fig. 33, experimental results show that reactant concentration varies
continuously across the entire cross section of the riser,
rather than there being a sharp discontinuity at a core
annulus boundary. Hence models are needed that provide for continuous variation across the riser, or at
least a greater number of intervals in the lateral direction. Such models include the following:
The model of Werther et al. (1992) still assumes a
coreannulus structure, but allows for radial dispersion in the central core, while disregarding
gradients in the annular wall zone.
=
=
=
a; b
C0
Cc
Ca
Ci
Cpc
Cpg
Cpp
=
=
=
=
Cs
Dgb
Dgd
Dge
Dgr
Dpr
Dpz
dp
dp
E
=
=
=
=
=
=
=
=
=
ec
ep
esusp
ew
Fo
f
=
=
=
=
=
=
fa
fc
g
Ga
=
=
=
=
Gc
Gs
=
=
Gs;max
Gs;CA
H
hcc
=
=
=
=
hcg
hrc
hrs
htot
Iac
=
=
Ica
K0
Kca
k
kc
=
=
kg
k1
=
=
k1
km
kp
L
=
=
=
Lc
M
m
n
P
Pege
Pegr
Pepr
Pepz
R
r
Rc
Rep
Rw
=
=
=
=
=
=
=
=
=
=
=
=
=
=
T
Tb
Tc
Tw
t
tc
UCA
UCC
UDSU
=
=
=
=
=
=
=
=
=
Ug
Umf
Ump
=
=
=
Use
Uslip
Utr
U
=
=
=
uc
ug
ugc
ut
=
=
=
=
vp
vpa
vpc
vpw
W0
x
z
=
=
=
=
=
=
=
zi
Greek letters
c
g
=
=
c
"
"
=
=
=
"
"a ; "c
"av
=
=
=
"cl
"CA
"c
"mf
"s
"sav
"sd
"s
g
b
c
g
p
susp
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
t
c Cpc kc ; W2 s=m2 K4
thickness of gas lm between cluster and wall
surface, m
dimensionless core radius R=Rc
voidage
overall average voidage for the riser as a
whole
local time average voidage
voidage of annulus, core
cross-sectional average voidage at a given
height
cluster voidage
voidage at UCA
emulsion layer voidage
bed voidage at minimum uidization
solids volume fraction
cross-sectional average solids fraction
solids volume fraction is dense bottom region
solids volume fraction at top exit
gas viscosity, kg/m
s
average bulk bed density, kg/m3
density of cluster, kg/m3
gas density, kg/m3
particle density, kg/m3
suspension density p 1 " g "; kg=m3
dimensionless time, ( t= or t=tm )
mean residence time, s
StefanBoltzmann constant,
5:67 108 Wm2 K4
variance of tracer concentration, s2
REFERENCES
Abba IA, Grace JR, Bi HT, Thompson ML. An integrated
approach to uidized bed reactor modeling, AIChE J, in
press, 2003.
Adams CK. Gas mixing in fast uidized beds. In: Basu P,
Large JF, eds. Circulating Fluidized Bed Technology II.
Oxford: Pergamon Press, 1988, pp 299306.
Ambler PA, Milne BJ, Berruti F, Scott DS. Residence time
distribution of solids in a circulating uidized bed: experimental and modeling studies. Chem Eng Sci
45:21792186, 1990.
Amos G, Rhodes MJ, Mineo H. Gas mixing in gassolids
risers. Chem Eng Sci 48:943949, 1993.
Andersson B-A. Eect of bed particle size on heat transfer in
circulating uidized bed boilers. Powder Technol
87:239248, 1996.
Andersson B-A, Leckner B. Experimental methods of estimating heat transfer in circulating uidized bed boilers.
Int J Heat and Mass Transfer 35:33533362, 1992.
Senior RC, Grace JR. Integrated particle collision and turbulent diusion model for dilute gassolid suspensions.
Powder Technol 96:4878, 1998.
Shen X, Zhou N, Xu Y. Experimental study of heat transfer
in a pressurized circulating bed. In: Basu P, Horio M,
Hasatani M, eds. Circulating Fluidized Bed Technology
III. Oxford: Pergamon Press, 1990, pp 451456.
Shingles T, McDonald AF. Commercial experience with
Synthol CFB reactors. In: Basu P, Large JF, eds.
Circulating Fluidized Bed Technology II. Oxford:
Pergamon Press, 1988, pp 4350.
Simons SJR, Seville JPK, Clift R, Gilboy WB, HosseiniAshra ME. Application of gamma-ray tomography to
gas uidized and spouted beds. In: Beck MS,
Campogrande E, Morris M, Williams M, Waterfall RC,
eds. Tomographic Techniques for Process Design and
Operation. Southampton: Computational Mechanics
Publications, 1993, pp. 227238.
Sinclair JL. Hydrodynamic modeling. In: Grace JR, Avidan
AA, Knowlton TM, eds. Circulating Fluidized Beds.
London: Chapman and Hall, 1997, pp 149180.
Sinclair JL, Jackson R. Gas-particle ow in a vertical pipe
with particle-particle interactions. AIChE J 35:1473-1486,
1989.
Smith TN. Limiting volume fractions in vertical pneumatic
transport. Chem Eng Sci 33:745749, 1978.
Smolders K, Baeyens J. Overall solids movement and solids
residence time distribution in a CFB-riser. Chem Eng Sci
55:41014116, 2000.
Sobocinski DJ, Young N, de Lasa HI. New ber-optic
method for measuring velocites of strands and solids
hold-up in gassolids downow reactors. Powder
Technol 83:111, 1995.
Squires AM. Origins of the fast uidized bed. In: Kwauk M,
ed. Fast Fluidization. Adv Chem Eng 20:137, 1994.
Srivastava A, Agrawal K, Sundaresan S, Reddy Karri SB,
Knowlton TM. Dynamics of gas-particle ow in
circulating uidized beds. Powder Technol 100:173182,
1998.
Sterneus J, Johnsson F, Leckner B, Palchonok GI. Gas and
solids ow in circulating uidized bedsdiscussion on
turbulence. Chem Eng Sci 54:53775382, 1999.
Steynberg AP, Shingles T, Dry ME, Jager B, Yukawa Y.
Sasol commercial scale experience with Synthol FFB
and CFB catalytic FischerTropsch reactors. In: Basu P,
Horio M, Hasatani M, eds. Circulating Fluidized Bed
Technology III. Oxford: Pergamon Press, 1991, pp
527532.
Stromberg L. Fast uidized bed combustion of coal. Proc 7th
Int Conf Fluidized Bed Combustion. Philadelphia. 1982,
pp 11521163.
Sun B, Gidaspow D. Computation of circulating uidizedbed riser ow for Fluidization VIII benchmark test. Ind
Eng Chen Res 38:787792, 1999.
Sun GL, Grace JR. Eect of particle size distribution in
dierent uidization regimes. AIChE J 38:716722, 1992.
Wu RL, Grace JR, Lim CJ, Brereton CMH. Suspension-tosurface heat transfer in a circulating uidized bed combustor AIChE J 35:16851691, 1989.
Wu RL, Lim CJ, Grace JR, Brereton CMH. Instantaneous
local heat transfer and hydrodynamics in a circulating
uidized bed. Int J Heat Mass Transfer 34:20192027,
1991.
Xie D, Bowen BD, Grace JR, Lim CJ. Two-dimensional
model of heat transfer in circulating uidized beds.
Intern J Heat Mass Transf. In press, 2003.
Xu G, Sun G, Nomura K, Li J, Kato K. Two distinctive
variational regions of radial particle concentration proles in circulating uidized bed risers. Powder Technol
101:91100, 1999.
Yang GL, Huang Z, Zhao L. Radial gas dispersion in a fast
uidized bed. In: Kunii D, Toei R, eds. Fluidization.
Cambridge: Cambridge University Press, 1983, pp
145152.
Yang H, Gautam M, Mei J. Gas velocity distribution in a
circular circulating uidized bed riser. Powder Technol
78:221229, 1994.
Yang WC. A mathematical denition of choking phenomenon and a mathematical model for predicting choking
velocity and choking voidage. AIChE J 21:10131021,
1975.
Yang WC. From pneumatic transport to circulating uidized
beds: the importance of operating maps. In: Kwauk M, Li
J, Yang WC, eds. Fluidization X. New York: Engineering
Foundation, 2001, pp 333340.
Yang YL, Jin Y, Yu ZQ, Wang ZW. Investigation on slip
velocity distributions in the riser of dilute circulating uidized bed. Powder Technol 73:6773, 1992.
Yerushalmi J, Cankurt NT. Further studies of the regimes of
uidization. Powder Technol 24:187205, 1979.
Yerushalmi J, Cankurt NT, Geldart D, Liss B. Flow regimes
in vertical gassolid contact systems. AIChE Symp Ser
174(176):112, 1978.
Yoshida K, Mineo H. High velocity uidization. In:
Doraiswamy LK, Mujumdar AS, eds. Transport in
Fluidized Particle Systems. Amsterdam: Elsevier, 1989,
pp 241285.
Younis HF, Glicksman LR, Hyre MR. Measurement
and prediction of the hydrodynamics of binary mixtures in a cold model pressurized circulating uidized
bed. In: Werther J, ed. Circulating Fluidized Bed
Technology VI. Frankfurt: DECHEMA, 1999,
pp 3945.
Yous Y, Gau G. Aerodynamique de lecoulement vertical
de suspensions concentrees gaz-solideI. Regimes
decoulement et stabilite aerodynamique. Chem Eng Sci
29:19391946, 1974.
Yu Z, Jin Y. Heat and mass transfer. Adv Chem Eng 20:
203237, 1994.
Zenz FA. Two-phase uidized-solid ow. Ind Eng Chem
41:28012806, 1949.
Zheng QY, Zhu DH, Feng JK, Fan JH. Some aspects of
hydrodynamics in N-shaped circulating uidized bed
combustor. Manaker AM, ed. Proc 10th Intern Conf
Fluidized Bed Combustion. New York: ASME, 1989,
Vol. 2, pp 157161.
Zheng QY, Xing W, Lou F. Experimental study on radial
gas dispersion and its enhancement in circulating uidized beds. In: Potter OE, Nicklin DJ, eds.
Fluidization VII. New York: Engineering Foundation,
1992, pp 285293.
Zheng Z, Zhu J, Grace JR, Lim CJ, Brereton CMH. Particle
motion in circulating and revolving uidized beds via
microcomputer-controlled colour-stroboscopic photography. In: Potter OE, Nicklin DJ, eds. Fluidization
VII. New York: Engineering Foundation, 1992, pp
781789.
Zhou J. Circulating uidized bed hydrodynamics in a riser of
square cross-section. PhD diss., University of British
Columbia, Vancouver, 1995.
Zhou J, Grace JR, Qin S, Brereton CHM, Lim CJ, Zhu J.
Voidage proles in a circulating uidized bed of square
cross-section. Chem Eng Sci 49:32173226, 1994.
Zhou J, Grace JR, Lim CJ, Brereton CMH. Particle velocity
proles in a circulating uidized bed of square cross-section. Chem Eng Sci 50:237244 1995a.
Zhou J, Grace JR, Lim CJ, Brereton CMH, Qin S, Lim KS.
Particle cross-ow, lateral momentum ux and lateral
velocity in a circulating uidized bed. Can J Chem Eng
73:612619, 1995b.
Zhou J, Grace JR, Brereton CMH, Lim CJ. Inuence of wall
roughness on the hydrodynamics in a circulating uidized
bed. AIChE J 42:11531156, 1996a.
Zhou J, Grace JR, Brereton CMH, Lim CJ. Inuence of
membrane walls on particle dynamics in a circulating uidized bed. AIChE J 42:35503553, 1996b.
Zhu JX, Yu JQ, Jin Y, Grace JR, Issangya AS. Cocurrent
downow circulating uidized bed (downer) reactors. Can
J Chem Eng 73:612619, 1995.
20
Other Nonconventional Fluidized Beds
Wen-Ching Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
SPOUTED BED
D in meters
The maximum value of the minimum spouting velocity, Um , occurs when the bed height is at the maximum spoutable bed depth or
1=3 s
dp D i
2gHm p f
3
Um
D
D
f
The Um is closely related to the minimum uidization
velocity, Umf , of the particles by
Figure 1 A classical and conventional spouted bed.
Um
b 0:9
Umf
to
1:5
p
dp Umf f
p
2
1 35:9 106 Ar 1
dp crit
2
60:6
gp f f
!1=3
8
Spout Diameter
2:00G0:49 D0:68
0:41
b
"
#
G0:433 D0:583 0:133
Ds 5:61
f b g0:283
2.4
10
Voidage Distribution
12
Figure 2
bed.
13
2.7
v20 max p
2g p f
14
0:5
Db
Do
1:68
0:57
tan
2
15
z 2
16
"0 1 E
H
where E is an empirical parameter varying between 0.3
and 0.6 and is empirically correlated as follows
0:12 0:97
Db
Ho
U 0:7 0:25
E 1:20
17
Ums
Do
Di
18
"0 "w
"w
1 exp r rs =27:81r2:41
s
20
The recirculating uidized bed with a draft tube concept is illustrated in Fig. 4. This concept was rst called
a recirculating uidized bed by Yang and Keairns
(1974). Several other names have also been used to
describe the same concept: the uid-lift solids recirculator (Buchanan and Wilson, 1965), the spouted uid
bed with a draft tube (Yang and Keairns, 1983;
Hadzismajlovic et al., 1992), the internally circulating
uidized bed (Milne et al., 1992; Lee and Kim, 1992);
or simply a circulating uidized bed (LaNauze, 1976).
The addition of a tubular insert, a draft tube, in a
spouted uid bed changes the operational and design
characteristics of an ordinary spouted bed. Notably,
there is no limitation on the so-called maximum spoutable bed height. Theoretically, a recirculating uidized bed with a draft tube can have any bed height
desirable. The so-called minimum spouting velocity
will also be less for a recirculating uidized bed with a
draft tube because the gas in the draft tube is conned
and does not leak out along the spout height as in an
ordinary spouted bed.
Figure 4 Recirculating uidized bed with draft tube operated as a pneumatic transport tube.
d Sd
Ad
21
r Sr
Ar
22
d Sd r Sr
Ad
Ar
23
Vsr "mf
1 "mf
24
26
27
28
29
4
2D
p f dp
31
The solid friction factor, fp , can be evaluated with the
equation proposed by Yang (1978):
"
#0:979
1 "r
Ret
1 "r
32
fp 0:0126
Rep
"3r
The lower limit of integration, Upr1 , is derived from
Wsr Upr p 1 "r
33
with "r 0:5, and the upper limit, Upr2 , by the equation
s
2
fp Upr
"4:7
Upr Ugr Ut
1
34
r
2gD
The total pressure drop in the acceleration region can
then be calculated as
P2-3
L
p 1 "r dL
"
0
2
"r Upr
fp p 1
2gD
2
p 1 "r Upr
g
2
2fg f Ugr
dL
gD
dL
#
at
L
35
Both solids circulation mechanisms and the solids circulation rate are important in designing and operating
a recirculating uidized bed with a draft tube. For
commercial applications in the area of coating and
encapsulation of solid particles, such as in coating of
pharmaceutical tablets and in coating seeds for delayed
germination and controlling the release rate of fertilizers, the particle residence time and cycle time are
important considerations. The performance based on
cycle time distribution analysis for coating and granulation was studied by Mann and Crosby (1973, 1975),
Mann (1983), and Turton et al. (1999).
Two mechanisms for solids circulation have been
observed experimentally (Yang and Keairns, 1978a).
High-speed movies (1000 to 1500 frames per second)
taken at the inlet and the midsection of the draft tube
with a sand bed revealed that solids transport inside
the draft tube was not a conventional pneumatic transport, where uniform solid suspension prevailed, but a
slugging type transport. The high-speed movies taken
at the outlet of a 23 m/s air jet showed that the air jet
issuing from the jet nozzle supplying air to the draft
tube was composed of bubbles rather than a steady jet.
The bubble grew from the mouth of the nozzle until its
roof reached the draft tube; then the sudden suction
from the draft tube punctured the roof. A continuous
stream of dilute solids suspension passed through the
roof into the draft tube. Simultaneously, another bubble was initiated. As this bubble grew, it pushed a slug
of solids into the draft tube. The high-speed movies
taken at the midsection of the draft tube exhibited
alternate sections of dilute suspension and solids slug
occupying the total cross section of the draft tube.
A steady jet without bubbling can be maintained in
a sand bed between the jet nozzle and the draft tube
inlet with high jet velocities of the order of 60 m/s and
without downcomer aeration. Once the downcomer is
aerated, the solids circulation rate increases dramatically and the steady jet becomes a bubbling jet.
Apparently, the inward-owing solids have enough
momentum to shear the gas jet periodically into
bubbles.
From the experimental evidence, the design of a recirculating uidized bed with a draft tube involves the
specication of a number of design parameters and
an understanding of the coupling eects between the
design and the operating variables. A procedure is presented here for the design of a bed to give a specied
Ad
Ar
36
Umf
"mf
37
3.5
38
dD
p f gdD
where Uj is the gas velocity issuing from the draft tube
gas supply. For high-temperature and high-pressure
operations, Eq. (41), to be discussed later, should be
used for calculating Lj . Another consideration is that
the jet boundary at the end of jet penetration should
correspond to the physical boundary of the draft tube
inlet. Merrys expression (1975) for jet half-angle can
be used for this purpose:
f dD 0:3
39
cot 10:4
p dp
or
Ld
D dD
2
tan
40
Industrial Applications
tacting operations. Examples are the low-waste conversion of ethylene and chlorine to dichloroethane,
biological treatment of high-strength municipal and
industrial euent, and bioreactors. Critical aspects of
the design and operation of bubble columns with draft
tubes have recently been reviewed by Chisti and MooYoung (1993). Freedman and Davidson (1964) also
carried out a fundamental analysis for gas holdup
and liquid circulation in a bubble column with a
draft tube. Extensive experimentation in a bubble column with a draft tube was conducted by Miyahara et
al. (1986), and an in-depth analysis by Siegel et al.
(1986). The eects of geometrical design on performance for concentric-tube airlift reactors were studied
by Merchuk et al. (1994).
4.1.2
7:65
41
do
Rcf p f gdo
where
Rcf Ucf p =Ucf atm
42
The gas jets can also carry solids and are referred to as
gassolid two-phase jets in this discussion.
4.1.1
Jet Half-Angle The jet half-angle can be calculated from the experimentally measured bubble size
and jet penetration depth as follows:
1 DB d0
43
tan
2Lj
Experimentally observed jet half-angles range from 8
to 12 for the experimental data mentioned above.
These compare to 10 suggested by Anagbo (1980)
for a bubbling jet in liquid.
4.1.3
Bubble Dynamics
To describe the jet adequately, the bubble size generated by the jet needs to be studied. A substantial
amount of gas leaks from the bubble to the emulsion
phase during the bubble formation stage, particularly
when the bed is less than minimally uidized. A model
developed on the basis of this mechanism predicted the
experimental bubble diameter well when the experimental bubble frequency was used as an input. The
experimentally observed bubble frequency is smaller
by a factor of 3 to 5 than that calculated from the
Davidson and Harrison model (1963), which assumed
no net gas interchange between the bubble and the
emulsion phase. This discrepancy is due primarily to
the extensive bubble coalescence above the jet nozzle
and the assumption that no gas leaks from the bubble
phase.
High-speed movies were used to document the phenomena above a 0.4 m diameter jet in a 3 meter diameter transparent semicircular jetting uidized bed
(Yang et al., 1984b). The movies were then analyzed
frame by frame to extract information on bubble frequency, bubble diameter, and jet penetration depth.
The process of bubble formation is very similar to
that described in Kececioglu et al. (1984), but it was
much more irregular in the large 3 m bed. A model was
developed to describe this phenomenon by assuming
that the gas leaks out through the bubble boundary at
a supercial velocity equivalent to the supercial minimum uidization velocity. For a hemispherical bubble
in a semicircular bed, the rate of change of bubble
volume can be expressed as
dVB
D2B
Gj Umf
dt
2
44
2
4Gj 2Umf DB
45
"
Gj
Gj
p
F
2Umf
2Gj Umf
#
p
2Gj DB 2Gj Umf
D3B
p DB
ln
12
2Gj DB 2Gj Umf
48
equation numerically along with the axial velocity proles in the jet obtained with a pitot tube. It is concluded that the gas mixing in a jetting uidized bed
with a permanent amelike jet is due primarily to convection and that diusion plays a negligible role.
The resulting velocity proles and the ow pattern
inside and around the jet indicated that the gas ow
direction is predominantly from the emulsion phase
into the jet at distances close to the jet nozzle. This
ow can be from the aeration ow in the emulsion
phase, as in the cases of high jet velocity or high aeration ow, or from the ow recirculated from the upper
part of the jet. The entrainment of gas into the jet
occurs immediately above the jet nozzle. The extent
of this region depends on both the aeration ow outside the jet and the jet velocity. Increases in aeration
ow and jet velocity tend to increase the height of this
region. Beyond this gas entrainment region, the gas in
the jet is then expelled from the jet along the jet height.
The gas expelled at the lower part of the jet is recirculated back into the jet, setting up a gas recirculation
pattern at the lower part of the jet. The extent of this
recirculation pattern increases with increases in jet
velocity and with decreases in aeration ow outside
the jet.
The axial velocity proles, calculated on the basis of
Tollmien similarity and experimental measurement in
Yang and Keairns (1980), were integrated across the
jet cross section at dierent elevations to obtain the
total jet ow across the respective jet cross sections.
The total jet ows at dierent jet cross sections are
compared with the original jet nozzle ow, as shown
in Fig. 7. Up to about 50% of the original jet ow can
be entrained from the emulsion phase at the lower part
of the jet close to the jet nozzle. This distance can
extend up to about 4 times the nozzle diameter. The
gas is then expelled from the jet along the jet height.
4.2.2
Yang et al. (1986) have shown that, based on the traversing force probe responses, three separate axial
solids ow patterns can be identied. In the central
core of the bed, the solid ow direction is all upward,
induced primarily by the action of the jets and the
rising bubbles. In the outer regions, close to the vessel
walls, the solid ow is all downward. A transition zone,
in which the solids move alternately upward and
downward, depending on the approach and departure
of the large bubbles, was detected in between these two
regions.
The solids circulation patterns were investigated
with a force probe. Since the force probe is directional,
the upward solids movement will produce a positive
response from the probe and vice versa, the magnitude
Figure 7 Gas entrainment into a 3.5 cm semicircular jet. (Adapted from Yang, 1998.)
@XJ0
@X 0
R2i 1 "mf p J 2Ri Wz
@z
@t
XJ0
49
XJ 0
@XJ
@X
R20 R2i 1 "mf p J
@z
@t
0
2Ri Wz XJ XJ 0
50
where
K nVB fw 1 "w p
51
XJ XJ0 0
2
2
dt
Ro Ri 1 "mf p
52
Since both XJ and XJ0 are independent of z, the relationship between XJ and XJ0 can be approximated by
the material balance of the coarse particles injected
into the bed to serve as the tracer.
XJ
1
Pt 1 expPt
XJo tw P
57
Hp R2i 1
Wt
"i R2o R2i 1 "mf
58
nVB
1 "mf
R2i UA
59
where fB is the volumetric fraction of bubbles occupying the bubble street region at any instant; it can be
evaluated from
fB
Figure 9 Schematic of a mathematical model for solids circulation in a jetting uidized bed.
Wt
R2i H1 "mf p
"
Ro
Ri
2
#
1 "mf
1
XJ
1 "i
53
Substituting Eq. (53) into Eq. (52), after some mathematical manipulation we get
dXJ
PXJ Q 0
dt
54
where
P
Q
R2o
2Ri Wz
2Wz
2
Ri 1 "mf p Ri 1 "i p
Ri R2o
2Wz Wi
2
Ri H1 "mf 1
"i 2p
55
56
nVB
R2i UA
60
Scale-up Considerations
The development of commercial uidized bed processors generally requires intermediate stages of testing on
physical models simulating commercial equipment.
Simulation (or scale-up) criteria derived from uidized
bed momentum-conservation relations may be applied
to determine the design and operating conditions for
physical models discussed in Chapter 13 Fluidized
Bed Scale-up. Such criteria, while not totally established at this time, have been applied to simulate a
pressurized uidized bed gasier having a large vertical
central jet for fuel feeding, combustion, and gasication (Yang et al., 1995).
Physical models of commercial uidized bed equipment provide an important source of design information for process development. A physical model of a
commercial uidized bed processor provides a smallscale simulation of the uid dynamics of a commercial
process. While commercial processes will typically operate at conditions making direct observation of bed uid
dynamics dicult (high temperature, high pressure,
corrosive environment), a physical model is designed
to allow easy observation (room temperature and pressure, nonreactive atmosphere, transparent vessel).
Cold ow studies have several advantages.
Operation at ambient temperature allows construction
of the experimental units with transparent plastic
material that provides full visibility of the unit during
operation. In addition, the experimental unit is much
easier to instrument because of operating conditions
less severe than those of a hot model. The cold
model can also be constructed at a lower cost in a
shorter time and requires less manpower to operate.
Larger experimental units, closer to commercial size,
can thus be constructed at a reasonable cost and within
an aordable time frame. If the simulation criteria are
known, the results of cold ow model studies can then
be combined with the kinetic models and the intrinsic
rate equations generated from the bench-scale hot
models to construct a realistic mathematical model
for scale-up.
The need for physical modeling of uidized bed
processors is dictated by the state of the art of uidized
bed scale-up technology. In general, no rational procedure exists for scaling up a new uidized bed processor
concept that precludes the need for physical modeling.
Many empirically developed rules of thumb for uidized bed scale-up exist in specic areas of uidized
bed application that are not generally applicable.
Existing mathematical modeling approaches are themselves based heavily on empirical descriptions of uidized bed uid dynamics. These bubbling bed models
can be applied only where condence exists for the
empirical bubble ow description built into the
model. Fluidized bed processors operate over such a
broad range of uid dynamic regimes that this condence rarely exists for new concepts.
Yang et al. (1995) described the application of this
scale-up approach. Comprehensive testing programs
were performed on two relatively large-scale simulation units for a period of several years: both a 30 cm
diameter (semicircular) Plexiglas cold model and a 3 m
diameter (semicircular) Plexiglas cold model operated
at atmospheric pressure.
The results are highly signicant to the development
of uidized bed technology because they represent a
case study of a rational uidized bed development
approach. The extensive data generated are unique in
their equipment dimensions, pushing existing models
and correlations to new extremes and oering new
insights into large-scale equipment behavior. The
understanding of the hydrodynamic phenomena developed from the cold ow model studies, and the analytical modeling reported was integrated with parallel
studies that investigated coal gasication kinetics, ash
agglomeration, charash separation, and nes recycle
to develop an integrated process design procedure.
4.6
Applications
The primary applications for large-scale jetting uidized beds are in the area of coal gasication as
described by Yang et al. (1995), Kojima et al. (1995),
and Tsuji and Uemaki (1994). Smaller scale applications are for uidized bed coating and granulation discussed in Chapter 17, Applications for Coating and
Granulation.
33:7
0:0408
C2
C1
2
f p f dp3 !2 C3
C1
2
0:5
33:7
C2
C1
63
where
1 1
ri ro
r
C2 ro ln o
r1
C1 r2o
C3
5.2
64
r2o r2i
2
Pressure Drop
ro
1
2 2 1
P Ug ro ln
Ug ro
ri
ri ro
66
where
16501 "
dp2
24:51 "f
dp
67
68
or
P p f 1 "!2
r2o r2i
2
69
NOTATION
Aa
Ad
Ar
Ar
As
C
CDS
Cm
=
=
=
=
=
=
=
=
=
=
=
Db
DB
DB max
dD
Di
=
=
=
=
=
do
Do
dp
dp crit
Ds
Ds
F
=
=
=
=
=
=
=
fB
fp
fw
g
G
Gj
Gr
H
=
=
=
=
=
=
=
=
HF
Hm
Hmf
=
=
=
Ho
Ld
Lj
L
=
=
n
P
P1-2
P1-4
P2-3
P3-4
r
=
=
=
=
=
=
=
ri
ro
rpf
=
=
=
rs
r1=2
=
=
r1=2 c
Ri
Ro
ReDo;ms
=
=
=
Remf
Rep
Ret
Sd
Sr
t
tw
=
=
=
=
U
Ua
=
=
UA
Ucf atm
=
=
Ucf p
Ufr
Ug
Ugd
Ugr
Uj
Ujb
=
=
=
=
=
=
=
Ujm
Ujr
=
=
Um
Umf
Umfc
Umfi
Umf atm
=
=
=
=
Umf p
Ums
Ums 0:5
=
=
Upd
Upt
Uslug
=
=
Usz
Ut
VB
V0 max
vs
=
=
=
=
Vsd
Vsr
vw
W
=
=
Wsd
Wsr
Wt
Wz
Xjo
=
=
X j ; Xj 0
Greek Letters
"
"0
"w
=
=
=
bed voidage
bed voidage at r 0
bed voidage at the wall in conical spouted
bed
"a
"bd
"br
"bs
"d
"i
"mf
"r
"o
"s
=
=
=
=
=
=
=
=
=
=
"w
airp1
=
=
=
b
f
p
s
!
=
=
=
=
=
=
=
d
r
=
=
REFERENCES
Abramovich GN. The Theory of Turbulent Jets. Cambridge,
MA: M.I.T. Press, 1963.
Alappat BJ, Rane VC. Studies on the eects of various
design and operational parameters on solid circulation
rate in a recirculating uidized bed. Can J Chem Eng
73:248252, 1995.
Anabtawi MZ. Minimum spouting velocity for binary
mixture of particles in rectangular spouted beds. Can J
Chem Eng 76:132136, 1998.
Anabtawi MZ, Uysal BZ, Jumah RY. Flow characteristics in
a rectangular spout-uid bed. Powder Technol
69:205211, 1992.
Anagbo PE. Derivation of jet cone angle from bubble theory.
Chem Eng Sci 35:14941495, 1980.
Botterill JSM, Bessant DJ. The ow properties of uidized
solids. Powder Technol 8:213222, 1973.
Buchanan RH, Wilson B. The uid-lift solids recirculator.
Mech Chem Eng Trans (Australia) 117124, May 1965.
Chen YM. Fundamentals of a centrifugal uidized bed.
AIChE J 33:722728, 1987.
Chevray R, Chan YNI, Hill FB. Dynamics of bubbles and
entrained particles in the rotating uidized beds. AIChE
26:390398, 1980.
Chisti MY. Airlift Bioreactors. New York: Elsevier, 1989.
Chisti MY, Moo-Young M. Improve the performance of airlift reactors. Chem Eng Progr 3845, June 1993.
He YL, Lim CJ, Qin SZ, Grace JR. Spout diameters in full
and half spouted beds. Can J Chem Eng 76:702706,
1998a.
He YL, Lim CH, Grace JR. Hydrodynamics of pressurized
spouted beds. Can J Chem Eng 76:696701, 1998b.
Horsler AG, Thompson BH. Fluidization in the development
of gas making process. Proceedings of Tripartite Chem
Engr Conf. Montreal, 1968, p 51.
Horsler AG, Lacey JA, Thompson BH. High pressure
uidized beds. Chem Eng Progr 65:5964, October 1969.
Ishida M, Shirai T. Circulation of solid particles within the
uidized bed with a draft tube. J Chem Eng (Japan)
8:477481, 1975.
Judd MR, Masson H, Meihack W. Solid circulation and
gasication experiments in a uidized bed with a draft
tube. In: Kunii D, Toei R, eds. Fluidization. New York:
Engineering Foundation, 1984, pp 663670.
Kao J, Pfeer R, Tardos GI. On partial uidization in rotating uidized beds. AIChE J 33:858861, 1987.
Keairns DL, Yang WC, Newby RA, Hamm JR, Archer DH.
Circulating bed boiler concepts for steam and power
generation. Proceedings of 13th Intersociety Energy
Conversion Eng Conf. Warrendal: Society of
Automotive Engineers, 1978, pp 540547.
Kececioglu I, Yang WC, Keairns DL. Fate of solids fed
pneumatically through a jet into a uidized bed. AIChE
J 30:99110, 1984.
King DF, Harrison D. The minimum spouting velocity of a
spouted bed at elevated pressure. Powder Technol
26:103107, 1980.
Knowlton TM, Hirsan I. The eect of pressure on jet penetration in semi-cylindrical gas-uidized beds. In: Grace
JR, Matsen JM, eds. Fluidization. New York: Plenum
Press, 1980, pp 315324.
Kojima T, Yoshitake H, Kimura T, Matsukata M, Uemiya
S. Contribution of local reactions in the grid zone to the
performance of a jetting uidized bed gasier of coal char.
Energy Fuels 9:379383, 1995.
Krambrock W. Mixing and homogenizing of granular bulk
material in a pneumatic mixer unit. Powder Technol
15:199206, 1976.
LaNauze RD. A circulating uidized bed. Powder Technol
15:117127, 1976.
LaNauze RD, Davidson JF. The ow of uidized solids. In:
Keairns DL, ed. Fluidization Technology. Vol. 2.
Washington: Hemisphere, 1976, pp 113124.
Lee WJ, Kim SD. Hydrodynamics and CWM combustion
characteristics in an internally circulating uidized bed
combustor. In: Potter OE, Nicklin DJ, eds. Fluidization
VII. New York: Engineering Foundation, 1992, pp
479486.
Li Y. Spouted bed hydrodynamics at temperatures up to
580 oC, M.A.Sc. thesis, University of British Columbia,
Vancouver, Canada, 1992.
Lim CJ, Mathur KB. Modeling of particle movement in
spouted beds. In: Davidson JF, Keairns DL, eds.
21
Standpipes and Nonmechanical Valves
T. M. Knowlton
Particulate Solid Research, Inc., Chicago, Illinois, U.S.A.
INTRODUCTION
STANDPIPES
The gas owing upward relative to the solids generates a frictional pressure drop. The relationship
between the pressure drop per unit length (P=Lg)
and the relative velocity for a particular material is
determined by the uidization curve for that material.
Normally, this uidization curve is generated in a uidization column where the solids are not owing.
However, the relationship also applies for solids owing in a standpipe.
Nearly all standpipe transfer systems use either
Geldart group A or Geldart group B solids. The uidization curve for Geldart group B solids diers from
that for group A solids. For both types of solids, as the
relative gas velocity through the bed increases from
zero, the P=Lg through the bed increases linearly
with vr . This region is called the packed bed region. At
some vr , the P generated by the gas owing through
the solids is equal to the weight of the solids per unit
area, and the solids become uidized. The relative velocity at this point is termed the interstitial minimum
Figure 1
Overow standpipe.
Figure 2
pipes. In small-diameter standpipes, even small bubbles have a diameter that is a large fraction of the
diameter of the standpipe, and these bubbles can hinder solids ow in small standpipes. In large standpipes,
this same small bubble diameter will not signicantly
aect solids ow (Fig. 5)
3. Streaming Flow. Underow standpipes (especially cyclone diplegs) sometimes operate in a dilutephase streaming ow characterized by high voidages.
A substantial amount of gas can be carried down the
standpipe when operating in this mode (Geldart and
Broodryk, 1991). Excessive gas ow down a dipleg is
usually undesirable. It can be minimized by reducing
the mass ux in the dipleg and/or increasing the pressure drop that the dipleg has to seal. This usually
means increasing the immersion of the dipleg into the
uidized bed.
Figure 4 Optimum standpipe operating regions for group A and group B solids.
An
An
An
An
Figure 6
H
Lg mf sp
Figure 11
at the bottom of the standpipe (Fig. 11B). If the material owing in the standpipe is a Geldart group A material, it is required that the aeration be added uniformly
along the standpipe. If the aeration is added at only one
location (i.e., at the bottom of the standpipe), a large
bubble will form in the standpipe at the aeration point
(Fig. 12A). If the bubble is large enough, it can restrict
the ow of solids down the standpipe. The large bubble
forms because it is dicult for the aeration gas to
permeate through the very ne solids moving through
the standpipe. Therefore it requires a signicant area
for the gas to dissipate through the very ne particles at
the same rate that it is being added through the aeration
tap. If the aeration gas is added at several locations,
then the bubble size is signicantly reduced, and standpipe operation is signicantly improved (Fig. 12B).
Note in Figs. 12A and 12B that the bubbles are
shown extending downward from the aeration point
in the direction of ow of the solids. This occurs
because the momentum of the solids is much greater
than the buoyancy force of the bubble and elongates
the aeration bubble in the direction of ow.
For Geldart group B solids, it is often unnecessary
to add aeration at several locations along the standpipe
to maintain the standpipe in uidized ow. Adding
aeration at the bottom of the standpipe operating
One of the simplest types of automatic solids recirculation systems is the cyclone/dipleg system utilized to
collect entrained solids and return them to the bed.
This type of system is shown in Fig. 13 for a uidized
bed containing both primary and secondary internal
cyclones. External cyclone return systems are also
automatic recycle systems. All solids entering the primary and secondary cyclones are returned to the bed
automatically via the primary and secondary diplegs.
The pressure balance around the system for returning solids to the bed via the secondary cyclone dipleg is
10
Figure 13
11
In this unit, hot catalyst is introduced into a dilutephase riser where it is contacted with crude oil. The
hot solids vaporize the oil and catalytically crack it
into lower molecular weight hydrocarbons in the
riser. The catalyst reactivity is signicantly reduced
by carbon deposition during this step. Therefore the
Figure 14
catalyst is transferred to a regenerator where the carbon is burned o of the catalyst. This heats up the
catalyst and also restores its reactivity. However,
before the catalyst is returned to the regenerator via
an angled standpipe/slide valve combination, hydrocarbon vapors are removed from the catalyst in a
steam stripper. From the regenerator, the catalyst is
transferred down another angled standpipe/slide
valve combination and is injected into the bottom of
the riser.
In most FCC units, the solids ow rate around the
system is controlled by a slide valve. FCC units
designed by the Kellogg Brown and Root Company
use another type of valve called a cone, or plug
valve, to control the solids ow rate (Wrench et al.,
1985).
Solids ow in standpipes in FCC units such as those
shown in Fig. 14 are generally controlled in the follow-
13
drop adjusts to exactly balance the pressure drop produced by the sum of the pressure drops on the independent side of the loop. However, there is a maximum
pressure drop per unit length (P=Lg) that the moving
packed bed standpipe can develop. This maximum
value is the uidized bed pressure drop per unit length,
P=Lgmf , for the material.
The independent pressure drop can be increased by
increasing any or all of the pressure drops across the
CFBC, the cyclone, or the L-valve. For a constant gas
velocity in the riser, as the solids ow rate into the bed
is increased, the independent part of the pressure drop
loop increases. The moving packed bed standpipe pressure drop then increases to balance this increase. It
does this by automatically increasing the relative gas
solids velocity in the standpipe. Further increases in
the solids ow rate can occur until the P=Lg in the
moving packed bed standpipe reaches the limiting
value of P=Lgmf . Because of its reduced capacity
to absorb pressure drop, a short standpipe reaches its
maximum P=Lg at a lower solids ow rate than a
longer standpipe. Thus the maximum solids ow rate
through a nonmechanical valve depends upon the
length of standpipe above it.
As indicated above during the discussion of how an
underow packed bed standpipe operates, the pressure
drop in such a standpipe is generated by the relative
velocity, vr , between the gas and solids. When vr
reaches the value necessary for minimum uidization
of the solids, a transition from packed bed to uidized
bed ow occurs. Any further increase in vr results in
the formation of bubbles in the standpipe. These bubbles hinder the ow of solids through the standpipe
and cause a decrease in the solids ow rate.
To determine the minimum standpipe length
required for a particular solids ow rate, it is necessary
to estimate the pressure drop on the independent side
of the pressure drop loop at the solid ow rate
required. The minimum length of standpipe necessary,
Lmin , is
Lmin
Pindependent
P=Lgmf
14
1. The Jones and Davidson (1965) equation relating valve pressure drop to solids ow rate and valve
open area, i.e.,
1
Ws 2
PL valve
15
2p 1 "mf CD A0
2. The relationship between the amount of aeration added to the L-valve (QA ), the amount of gas
owing down the standpipe (Qsp ), and the amount
actually owing around the L-valve bend (QT ):
QT Qa Qsp
16
3. An equation estimating the amount of gas owing down the standpipe (this equation assumes no slip
between the solids and gas in the standpipe):
QT
Ws "mf
p 1 "mf
17
18
transfer these solidsgenerally with less than satisfactory results. The best and simplest way to return the
nes is to use a nonmechanical device. The simplest
nonmechanical device used to recycle cyclone nes
back to a reactor is a straight cyclone dipleg.
However, the dipleg must be immersed in a uidized
dense-phase bed (or tted with a trickle valve or
another control device at its discharge end if it is not
immersed in the bed) in order to perform properly.
Since cyclones in CFBs are generally external to the
bed and located some distance from the CFB, a vertical cyclone dipleg cannot be used. Angled diplegs tend
to slug and perform poorly for essentially the same
reason as angled standpipes (see above). Therefore,
loop seals, seal pots, and L-valves are employed to
return the solids to the bed. A V-valve, another type
of solids return device, can also be used in this manner.
A discussion of the various types of automatic nonmechanical devices used is given below.
2.7.1
Seal Pot
Loop Seal
N-Valve
A novel type of nonmechanical device called the Nvalve was used by Hirama et. al. (1986). However,
this type of device is not recommended because gas
channeling back from the angled section can cause
large bubbles in the standpipe just as was found with
the hybrid standpipes discussed earlier.
2.7.4
V-Valve
the standpipe allows the solids to ow from the standpipe into the diverging section. The included angle of
the diverging section is usually smallbetween 5 and
10 degrees. Greater angles cause uneven distribution of
the gas and solids.
As with the loop seal, the V-valve will not operate if
the solids in the upow (diverging) section are not
uidized. When the solids in the diverging section are
uidized, solids ow from the standpipe, through the
aperture and diverging section, into the uidized bed.
As with the loop seal, the upow section of the V-valve
prevents pressure surges from causing the dipleg to
blow.
2.7.5
L-valve
Figure 25
NOMENCLATURE
Figure 27
A0
CD
D
Dv
g
Hdip2
Hdip2
Hmin
Hsp
Hsp0
L
Lmin
=
=
=
=
=
=
=
=
=
=
=
=
Lsp
P
P0
Pb
Pt
P
PH
PCFB
Pcy
Pcy 1
Pcy2
Pd
Pdip2
=
=
=
=
=
=
=
=
=
=
=
=
=
Pfluid bed
Pindependent
=
=
Plb
Pls
PL-valve
Pregen
Priser
Psp
Psp1
Psp2
Pstripper
Psurge
Psurge hopper
Psv1
Psv2
Pub
Pv
P=Lg
P=Lgmf
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Q
QA
Qsp
=
=
=
QT
U
Umf
=
=
=
vg
vmb
vmf
=
=
=
vr
vs
Ws
"
"mf
dip2
mf
=
=
=
=
=
=
=
p
sk
t
=
=
=
REFERENCES
Bristow TC, Shingles T. Cyclone dipleg and trickle valve
operation. In: Grace JR, Shemilt LW, Bergougnou MA,
eds. Fluidization VI, 1989, pp 161168.
Campbell DL, Martin HZ, Tyson CW. U.S. Patent
2,451,803, 1948.
Chan I, Findlay J, Knowlton TM. Operation of a nonmechanical L-valve in the automatic mode. Paper presented at
the Fine Particle Society Meeting, Santa Clara, CA, 1988.
Eleftheriades CM, Judd MR. The design of downcomers
joining gas-uidized beds in multistage systems. Powder
Technology 21:217225. 1978.
Geldart D, Broodryk N. Studies on the behaviour of cyclone
diplegs. Presented at the Annual Meeting of the AIChE,
Los Angeles, CA, November 1821, 1991.
Geldart D, Jones P. The behaviour of L-valves with granular
powders. Powder Technology 67:163172, 1991.
Geldart D, Kerdoncu A. The behaviour of secondary and
tertiary cyclone diplegs. Presented at the AIChE Annual
Meeting, Miami Beach, 1992.
Grace JR. Contacting modes and behaviour classication of
gassolid and other two-phase suspensions. Can J Chem
Eng 64:353364, 1986.
Hirama T, Takeuchi H, Horio M. Nitric oxide emission from
circulating uidized bed coal combustion. Proceedings of
the 9th International Fluidized Bed Combustion
Conference, Boston, pp 898910, 1986.
Jones DRM, Davidson JF. Rheol Acta, 4:180191, 1965.
Karri SBR, Knowlton TM. Comparison of group A solids
ow in hybrid angled and vertical standpipes. In: Avidan
Ar, ed. Circulating Fluidized Bed Technology IV, 1993,
pp 253259.
22
Cyclone Separators
T. M. Knowlton
Particulate Solid Research, Inc., Chicago, Illinois, U.S.A.
INTRODUCTION
CYCLONES
g0:8ro 9:810:4Db
Cyclone Types
Figure 1
2.2
3.1 Series
is only 0.025 m/s for a ux in the dipleg of 20 kg/sm2 . Gas can easily ow up the dipleg against this very
low velocity. Conversely, in rst stage cyclone diplegs
where the solids ux may be as high as 750 kg/s-m2 ,
the solids velocity is so high (approximately 0.94 m/s
at a dipleg density of 800 kg/m3 ) that gas is carried
down the dipleg with the solids.
Because of these recurring problems with secondstage cyclone diplegs, companies sometimes choose to
use a single high-eciency cyclone in place of two
cyclone stages. Even though the overall eciency
using a single cyclone is slightly less than using two
cyclone stages, they nd that this cyclone arrangement
optimizes the operation of the process by minimizing
upsets and reducing downtime using a single higheciency cyclone. It is also a much simpler system as
well and saves on initial capital costs.
3.2
Parallel
Because of problems in the operation of small parallel cyclones, it is generally found that they result in a
collection eciency less than or (if designed well) equal
to that of a large, single cyclone processing the same
amount of gas and solids.
In the FCC and similar processes there are many
units operating with two cyclone stages in series, but
with several (2 to 18) cyclones in each stage in parallel.
These cyclones can have very high eciencies even
though they operate in parallel. However, each rststage cyclone generally has a dipleg that is immersed in
the bed (there is no common collection chamber). Each
second stage cyclone has a dipleg that usually has a
trickle valve on its discharge. The trickle valve eectively is a check valve that prevents backow of most
of the gas through the dipleg on startup. In addition,
the pressure drop through the cyclone stages is generally sucient to prevent maldistribution.
There has been very little work in the literature on
the operation of parallel cyclones. However, what
work does exist indicates that too low a pressure
drop results in poor gassolids distribution, and that
too many parallel cyclones in a common hopper are
not recommended.
Broodryk and Shingles (1995) studied parallel
cyclone operation with cyclones in the freeboard
above a turbulent uidized bed. They found that too
low an overall pressure drop (caused by inlet gas velocities of less than about 15 m/s) resulted in preferential
ow of gas through two of their three cyclones. They
also found that poor trickle valve design and/or operation could induce an imbalance in the gas ow into the
cyclones. When preferential gas ow occurred, it
resulted in signicant solids carryover and reduced
cyclone eciency.
Koman (1953) studied the eciency of several different small-diameter (38 and 51 mm diameter) parallel
cyclone congurations and reported the relative eciencies shown in Table 1. Komans data show that
placing cyclones in a common hopper reduced the collection eciency, and that adding more cyclones to a
common hopper also caused a reduction in eciency.
Smellie (1942), using three 200 mm diameter cyclones,
and Whiton (1941), using 9 to 20 cyclones 610 mm in
diameter, also reported poorer cyclone performance
with parallel cyclone operation. However, OMara
(1950) reported very little dierence between single
cyclone and parallel cyclone eciencies These results
indicate that parallel cyclones can be designed and
operated with the same eciencies as a single cyclone,
but that it takes good and careful design of the
cyclones to accomplish this.
Eciency, %
96.0
95.3
94.1
92.2
Hugi and Reh (1998) report that continuous acceleration of the solids throughout the inlet is desired for
improved eciency, and that the angled inlet described
above achieves this. If the momentum of the solids is
sucient and the solids are continuously accelerating
along the length of the inlet, they will form the stable,
coherent strand important for high collection eciencies.
Another superior inlet for high solids loadings is the
volute cyclone inlet. At high inlet loadings in a tangential cyclone inlet, the gassolids stream expands rapidly
from its minimum width at the point of contact. This
rapid expansion disturbs the laminar gas ow around
the gas outlet tube and causes ow separation around
the tube. This results in lowered cyclone eciency. But
when a volute inlet is used, the expanding solids stream
is further from the gas outlet tube and enters at an
angle so that the solids do no induce as much ow
separation or asymmetrical ow around the gas outlet
tube. This cyclone eciency is not aected to as great a
degree.
The nature of the gassolids ow in the inlet ducting
to the cyclone can aect cyclone eciency signicantly.
If the solids in the inlet salt out on the bottom and
result in dune formation and the resulting unsteady
or pulsing ow, cyclone eciency is adversely aected.
To minimize the possibility of this occurring, it is
recommended that the inlet line to the cyclone operate
above the saltation velocity (Gauthier et al. 1990b),
which will prevent the solids from operating in the
dune or pulsing ow regime. If this is not possible,
then the inlet line can be angled downward (approximately 15 to 20 degrees) to let gravity assist in the ow
of the solids. Keeping the inlet line as short as possible
can also minimize any pulsing of the solids ow.
Cyclones can collect solids over a wide range of loadings. Traditionally, solids loadings have been reported
as either kg of solids per m3 of gas (kg/m3 ), or as kg of
solids per kg of gas (kgs =kgg ). However, loading based
on mass is probably not the best way to report solid
loadings for cyclones. This is because the volume of
solids processed by a cyclone at the same mass loading
can vary greatly, depending on the density of the
solids. For example, many polymers have a bulk density of approximately 400 kg/m3 and iron ore has a
bulk density of approximately 2400 kg/m3 . This is a
ratio of 6. Therefore, a cyclone operating with polymer
would have to process six times the volume of solids
To prevent solids from the cyclone inlet from bypassing directly into the outlet of the cyclone, a tube the
same diameter as the gas outlet is extended into the
cyclone to a level equal to or below the bottom of
the solids inlet. This prevents solids from bypassing
directly into the outlet of the cyclone. This tube is
called a gas outlet tube (it is also called a vortex nder
or a vortex tube). The gas outlet tube does increase the
eciency of a low-loading cyclone relative to a cyclone
that does not have a gas outlet tube. However, many
Figure 8 Eect of outlet tube penetration on cyclone eciency. (From Lewnard et al., 1994.)
CYCLONE LENGTH
11
CYCLONE DIAMETER
1
dp;th
3.
4.
5.
n
X
xi Eoi
i1
where:
xi the weight fraction of each individual
cut size
Eoi the collection efficiency of the cyclone
for each individual cut size, %:
6.
7.
Modify the low-loading cyclone collection eciency for the eect of loading.
Determine cyclone pressure drop.
12.1.2
Partition Cuts
Estimation of Ns
12.1.3
Figure 9 The eect of gas velocity on the number of solid spirals in a cyclone. For maximum velocity, use greater of inlet or
outlet velocity.
n
X
xi Eoi
i1
12.1.4
12.1.5
Kfi or Ko
0.50
0.47
0.43
0.395
0.35
4Inlet area)
Inlet perimeter
2f g Ui2 Db Ns
dhi
dhi Ui g
10
g Ui2
2
11
Figure 12
13
VORTEX LENGTH
The spinning gas from the outer annulus ows into the
inner rotating core (the vortex) over a certain length of
the cyclone. The length of this vortex, Lv , is taken to be
the length within the cyclone where the vortex gas can
exert a positive driving force (Bryant et al., 1983). This
DB Ao =Ai 1=2
14
13
VORTEX STABILIZERS
14
Eo /
1
Dp;th
s
Ns Up g
9Lw
The eects of temperature and pressure manifest themselves in the way that they aect gas density and gas
viscosity. From Eq. (15), it can be seen that cyclone
eciency is theoretically related to gas density and gas
viscosity as
r
p g
Eo /
16
As pressure is increased, gas density will increase.
However, the term p g does not change with
increases in gas density because particle density is so
much greater than the gas density (typically about 2000
kg/m3 versus approximately 20 kg/m3 at high pressure)
that it dominates this term. Therefore, it would be
expected that gas density would have little or no eect
on cyclone eciency. Conversely, cyclone eciency
would be expected to decrease with system temperature
because gas viscosity increases with increasing temperature.
Knowlton and Bachovchin (1978) studied the eect
of pressure on cyclone performance and found little
change in overall cyclone eciency with pressure over
a pressure range from 0 to 55 barg. However, fractional
eciency curves for the same study (Fig. 14) showed
that cyclone eciency decreased with pressure for particle sizes less than about 20 to 25 microns. For particle
15
=
=
=
Ao
As
=
=
Db
Dp;th
=
=
Do
dhi
=
=
dp;50
dp;th
=
=
Eo
Eoi
f
H
Kfi
=
=
=
=
=
Ko
L
Lv
Lw
Ns
Rehi
=
=
=
=
=
ro
U
Uf
Ui
Uo
Upi
Upf
=
=
=
=
=
=
=
Ut
xi
=
=
P
Pfig =
Pfig
Pbf
Po
p
g
=
=
=
=
=
=
REFERENCES
Akiyama T, Marui T. Dust collection eciency of a straightthrough cyclone: eects of duct length, guide vanes and
nozzle angle for secondary rotational ow. Powder
Technol 58:181, 1989.
Alexander RM. Fundamentals of cyclone design and operation. Proceedings of the Australian Institute of Mining
and Metallurgy 152, 153:203, 1949.
Barth W. Berechnung und auslegung von zyklonabscheidern
aufgrund neuerer untersuchungen. BWK 8(1):19, 1956.
Barth W, Leineweber L. Beurteilung und auslegung von
zyklonabscheidern. Staub 24(2):4153, 1964.
Broodryk NJ, Shingles T. Aspects of cyclone operation in
industrial chemical reactors. Preprints of Fluidization
VIII. Tours, France, May 1419, 1995, p 1083.
Bryant HS, Silverman RW, Zenz FA. How dust in gas aects
cyclone pressure drop. Hydrocarbon Processing June,
pp 8790, 1983.
Dietz PW. Collection eciency of cyclone separators. AIChE
J 27:888, 1981.
Gauthier TA, Briens CL, Bergougnou A, Galtier P. Uniow
cyclone eciency study. Powder Technol 62:217225,
1990a.
Gauthier TA, Briens CL, Galtier P, Bergougnou MA. Gas
solid separation in a uniow cyclone at high solids loadings: eect of acceleration line. In: Basu P, Horio M,
Hasatani M, eds. Circulating Fluidized Bed Technology
III. Oxford: Pergamon Press, 1990b, pp 639644.
Gauthier TA, Briens CL, Crousle O, Galtier P, Bergougnou
MA. Study of the gas circulation patterns in a uniow
cyclone. Can J Chem Eng 70:209215, 1992.
Homan AC, de Groot M, Peng W, Dries HWA, Kater J.
Advantages and risks in increasing cyclone separator
length. AIChE J 47(11):24522460, 2001.
Homan AC, de Jonge R, Arends H, Hanrats C. Evidence of
the natural vortex length and its eect on the separation eciency of gas cyclones. Filt Sep 32:799, 1995.
Hugi E, Reh L. Design of cyclones with high solids entrance
loads. Chem Eng Technol 21(9):716719, 1998.
Iinoya K. Memoirs of the faculty of engineering. Nagoya
University 5(2): September, 1953.
Kaliski H. Vergleich von iehkraftentstaubern. Freiberger
Forschungsheft A93:78100, 1958.
23
Dilute-Phase Pneumatic Conveying
George E. Klinzing
University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
Dilute-phase pneumatic conveying is not an energyecient way of conveying, since generally considerable
horsepower is needed to provide the motive air or gas.
Dilute-phase pneumatic conveying, however, is easy
and convenient to put into operation, even if often it
does not work at optimum conditions, mostly owing to
poor understanding of the overall concept of the
process.
There are ve components to a dilute pneumatic
conveying system:
1.
2.
3.
4.
5.
Conveying line
Air/gas mover
Feeder
Collector
Controls
Pressure system
Vacuum system
Pressure/vacuum system
2
2.1
Shape Analysis
gas
gravity
solid
gravity
solids
friction
effect
gas
friction
effect
particle velocities. In order to be sure that actual measurement of the particle velocity is carried out, the
HinkleIGT (1978) model can be related to several
systems and particle parameters:
0:2 0:54
D
Yang (1977) has developed an implicit equation to predict the particle velocity that is described as
up ug ut
1. Voidage
2. Particle velocity
3. Friction factors
Voidage
4Ws
up p
Particle Velocity
Actually, in our studies we have found that this expression is good up to about 500 micrometers in particle
size. Several other studies have given models for the
2fs 2 4:7 0:5
u "
1
gD p
fp 4fs
2.6
5
6
Friction Factors
Another frictional representation that has been receiving some attention has been developed by Michaelides
(1987):
p
K gD
fs
9
4 up
where K depends on the type of particle conveyed. For
coal it is 0.058.
The eect of dierent pipe wall materials has been
probed by Rizk (1973) and others. This eect, however, appears to be minor and even of little consequence when a ne powder coats the conveying line
wall. For inclined ow Klinzing et al. (1989) has
given a frictional representation that covers all kinds
of inclined ows from horizontal to vertical:
2
4:7
3=4
C
"
u
u
D
g f
p
D
fs 2
g sin
p s d
2up
Acceleration
11
Saltation
13
14
2.10
Compressibility
Figure 1 Bifurcation diagram, including pickup and saltation mechanisms of solid particles and dierent ow regimes observed
in fully developed ow of gassolids suspensions in horizontal pipelines.
2.11
Bends
Choking Conditions
FLOW CLASSIFICATIONS
The ow classication of Geldart (1973) for uidization has been applied also to pneumatic conveying.
Detailed discussion of Geldsarts classication of
powders is presented in Chapter 3, GasSolids
Fluidization. The four materials A, B, C, and D
have a signicant relationship with pneumatic conveying ease and ability. The C type powders, which are
cohesive in nature, are denitely the most dicult to
move, although the piston type ows can be easily
formed with these materials. The free-owing A powders work well in dilute-phase ow. These materials
cause few problems in the coupling of the feeder unit
with the conveying line. The B materials generally do
not move well in the more heavily loaded systems as
Table 1 Design Strategy for Dilute-Phase
Transport.
1. Select the criteria for the transport velocity:
saltation velocity, pickup velocity, specied
velocity.
2. Determine the length of each horizontal and
vertical section.
3. Determine the number of bends.
4. Determine the pressure drop for each section,
horizontal, vertical, and bends.
5. Calculate the acceleration pressure loss for the
entrance of the particles into the system.
6. Calculate the over-pressure loss by summing the
component parts.
7. Decide if the system should have a stepping
pipeline arrangement if the over-distance is
greater than 300 feet.
8. Select the blower, feeder, and collector for the
system.
Figure 2 General Zenz-type diagram of pressure drop per unit length versus velocity for varying owrates. Note dense, dilute,
and saltation regimes.
Figure 3 Phase diagram-type plot of reduced pressure analog versus the volume analog at constant reduced gas ow analog.
Note similarities of strand ow to liquid phase, ow above strand ow to vapor phase, and homogeneous ow to gas phase.
4.1
Air Movers
Fans
Blowers
Compressors
Feeders
4.4 Filters
Filters or other collection devices are required for all
conveying systems. The reserve-jet air ow systems for
cleaning the lter are very common units. For lower
pressure, the air-to-lter cloth ratio is usually approximately 1.1 to 0.3 CMM/m2 (3.5 to 1.0 CFM/sq.ft). For
higher pressure reserve ow systems, a higher air-tocloth ratio is employed, 2.1 to 0.3 CMM/m2 (7.0 to 1.0
CFM/sq. ft). For more detailed discussions, see
Chapters 22 and 28, GasSolids and LiquidSolids
Separation.
4.5
Cyclones
Scrubbers
TROUBLESHOOTING
NOMENCLATURE
CD
D
fg
fs
g
L
P
ug
up
Ws
=
=
=
=
=
=
=
=
=
=
drag coecient
diameter of the pipe
gas friction factor
solids friction factor
gravity constant
length
pressure drop
gas velocity
particle velocity
solids ow rate
Greek
"
g
p
=
=
=
=
=
voidage
angle of inclination of the pipe
gas density
particle density
loading
REFERENCES
Barth W. Stromungstechnische probleme der verfarhrenstechnik. Chem Ing Tech 20, No. 1:2932, 1954.
Blasius H. Grenzschichten in ussigkeiten mit kleiner
reibung. Mitt Forschungsarb 131:140, 1913.
Cabrejos FJ, Klinzing GE. Pickup and saltation mechanisms
of solid particles in horizontal pneumatic transport.
Powder Tech 79:173186, 1994.
24
Electrostatics in Pneumatic Conveying
George E. Klinzing
University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
Higher loadings at low relative humidities also produced increased electrostatic eects. Little or no
electrostatic charging was generated at ratios of kg of
water per kg of solids, which were greater than 0.1.
Smeltzer et al. (1982) and Nieh and Nguyen (1988),
experimenting with a 2 inch copper pipe and glass
beads, found that when the system moisture content
exceeded a relative humidity of 76%, the charge on the
glass beads became neutralized. The charging exhibited
itself as an increase energy requirement for ow, which
could under certain conditions be represented by an
increase in pressure drop of up to 70% (Ally and
Klinzing, 1983).
Boshung and Glor (1980) showed that the particle
size and the material are the two most important parameters associated with charging in pneumatic transport systems.
In solids processing one should be on guard for ne
particles with a high specic surface area of nonconductive material in an atmosphere of low relative
humidity owing in nonconductive or nongrounded
containments. These factors all add up to a troublesome situation.
Adhesion of particles to surfaces due to electrostatics causes problems by coating piping walls. In
these cases the conveyed materials contact the wall
coating rather than the material of construction of
the pipe and can lead to charge buildup. In addition, coating of sensors by the adhesion of particles
by electrostatics will impede the process of measuring properties of the gassolid ow systems. It is
well known that powder and granular materials
generate charging more readily than large diameter
particles.
Not all electrostatic charges should be viewed with
suspicion in the handling of powders. One can engineer
and manage the charging to provide avenues to measure eciently the solid ows and separate powders of
dierent properties.
USEFULNESS OF ELECTROSTATICS
After rst having to deal with the hazards of electrostatic generation in pneumatic conveying and their
adverse eects on electronics, these phenomena can
be channeled into a useful tool for measurement. The
charges generated in the conveying operation can be
measured at two distinct points, and these signals can
be cross-correlated to provide a measure of the particle
velocities. In fact, one can use a coil as an antenna
PRECAUTIONS
Electrostatics and electronic equipment such as computers and A/D converters do not mix. One should be
very careful, when a system is shown to have electrostatic tendencies and the measuring devices employed
have a computer linkage, that the computer should be
isolated from electrostatic charging eects. This isolation is sometimes very challenging, since the charging
eects can travel on the surface of even nonconducting
surfaces to interfere or damage the electronics.
Changing the electric signal to a light signal is sometimes a solution to avoid large discharges interfering
with and damaging computer operations. Sometimes
such conversions will modify the original signal, so
knowledge of this modication is essential before
applying this procedure.
One of the most dangerous procedures when dealing
with electrostatic charging is covering a charged nonconducting surface with a metal foil and shield. This
may appear as a possible solution to the grounding of
the charge generated, but in eect, what has happened
in applying the foil is that one is constructing a large
capacitor that can store energy and discharge this
5.1
Horizontal System
Vertical System
humidity. Prior to eliminating the electrostatics, interesting data was acquired for 97m and 545 m glass
beads.
5.3
Figure 2 Eect of electrostatics charging on the pressure drop behavior of PVC particles in 0.0254 m (I.D.) system
Ws 1:14 kg=min.
Figure 3 Eect of electrostatics charging on the standard deviation of the pressure drop for PVC particles in 0.0254 m (I.D.)
system Ws 1:14 kg=min:
The rst minimum occurring at the lower gas velocity of the double minimum increased with increasing
solids ow rate (Figs. 4 and 5). This increase in pressure drop, as the gas velocity was in the range of 4.5
7.6 m/s, occurred owing to the clustering of the particles caused by the electrostatic generation of the glass
beads with the Lucite tubing. This clustering caused
the pressure drop in the system to be equivalent to
that of larger particles. The greater the charging of
the particles, the larger and more dened the clusters
were. This was also evident by the tracking of the
P=L vs. time for a given Ug and Ws . The pressure
drop would continue to increase until a discharge, of
the system, as evident by sight and sound, occurs. At
this point the pressure drop would fall back to a certain level and continue to increase again with time until
there is another discharge. A certain residual charge
resided in the particles and on the pipe.
Figure 4 Pressure drop for 800 g/min with and without electrostatics present.
Figure 5
Pressure drop for 400 g/min with and without electrostatics present.
citor, since the solids were isolated from any conducting medium due to the Lucite walls, and therefore it
contained a charge. The pressure drop decreases just
prior to the charging portion of the curve. Since the
particles are large (545 m) in comparison with the 9
7 m glass beads, it is hypothesized that the particles
are weakly held together or attracted by electrostatic
charges, causing a higher pressure drop than if they
were individual entities. The dip in the pressure drop
curve prior to the charging and discharging portion of
the pressure drop curve could be due to the particles
separating (behaving as individual particles) and possibly losing some of the charge. Then upon losing the
charge or becoming separated, there is an increasing
force that causes the particles to reattract or cluster
together.
5.5
Summary of Testing
NOMENCLATURE
ti
Ug
Ut
P
P=L
=
=
=
=
=
=
Time, s
Gas Velocity, m/s
Terminal velocity of particle, m/s
Dierential pressure drop, Pa
Pressure drop per unit length, Pa/m
Standard deviation, Pa
Time
610
t1
t2
701
720
t3
discharge occurred in system
595
t4
Times between readings, i.e., ti1 ti t, are approximately the same
REFERENCES
Ally MR, Klinzing GE. Electrostatic eects in gas-solid
pneumatic transport with loadings to 100. J Powder and
Bulk Solids Tech 7(3):1320, 1983.
Ally MR, Klinzing GE. Inter-relation of electrostatic
charging and pressure drops in pneumatic transport.
Powder Tech 44(1):8588, 1985.
Boshung P, Glor M. Methods for investigating the electrostatic behavior of powders. J Electrostatics 8:22052219,
1980.
Cartwright P, Singh S, Bailey AG, Rose LJ. Electrostatic
charging characteristics of polyethylene powder during
pneumatic conveying. IEEE Transactions on Industry
Applications IA21(2):541546, 1985.
Cross JA. Electrostatics: Principles, Problems and
Applications. Bristol, U.K.: IOP, 1987, pp 376382.
Dahn JC. Electrostatic hazards of pneumatic conveying
of powders. Plant/Operations Progress 11(3):201204,
1992.
Dhodapkar SV. Flow pattern classication in gassolid
suspensions. Ph.D. diss., University of Pittsburgh,
Pittsburgh, PA, 1991.
Floyd TL. Electronics Fundamentals: Circuits, Devices
and Applications. Columbus, OH: Merrill, 1987,
pp 334339.
Gajewski A. Measuring the charging tendency of polystyrene
particles in pneumatic conveyance. Journal of
Electrostatics 23:5566, 1989.
Gajewski JB, Glod B, Kala W. Electrostatic method for
measuring the two-phase pipe ow parameters. IEEE
Industry Applications Society Annual Meeting, Seattle,
WA, Oct. 1990.
Gajewski JB, Glod B, Kala W. Electrostatic ow meter for
measuring the two-phase ow parameters in pneumatic
transport. Results of preliminary tests. Proceedings of
25
Instrumentation and Measurements
Masayuki Horio, Rafal P. Kobylecki , and Mayumi Tsukada
Tokyo University of Agriculture and Technology, Tokyo, Japan
INTRODUCTION
Thermocouples
Optical Pyrometer
Table 1 Interrelationships Among Measurement Techniques or Sensors and Process Parameters of Fluidized Beds Applicable for Bench Scale
(B), Pilot Scale (P), and Commercial Scale Tests (C). Case C also Includes Emergency Measurements
Flow regime
& structure
Particle
characteristics
Pressure
B, P, C
B, P, C
B, P, C
Pressure sensors
Thermocouples
Pyrometers
Suction probes
Gs meters
Capacitance probes
Optical ber probe
Laser sheeting
LDV/PIV
Tomography
Solid/gas tracers
Camera observation
B, P
B, P, C
B, P
B, P, C
B
B, P
B
B
B, P, C
B, P, C
Temperature
Voidage
Gas
Velocity
B, P, C
B, P, C
B, P, C
B, P, C
B, P
B
B, P, C
B, P, C
(IR camera)
Solids
mass
ux
Gas/solids
composition
B, P, C
B, P, C
B, P, C
Solids
velocity
B, P
B, P, C
B
B, P
B
B
B, P, C
B, P
B, P
B, P
B, P
B, P
B, P
B, P
B, P, C
B
B, P
B, P
B, P
B, P
B, P, C,
B, P, C
B, P, C
B, P
B
B, P
B, P
B, P, C
C "
1
5 eC2 =
T 1
where
"
is
emissivity
of
an
object,
C1 3:74 1016 Wm2 , and C2 0:0114 mK.
For shorter wavelengths and for lower temperatures, the above equation is simplied to the formula
called Wiens law:
C1 "
C2
exp
E
T
5
X
C2 =
1 T
E1 "
1
5
1 e
C2 =
2 T
E2 "
2
5
2 e
Since "
1 "
2 , Eq. (3) can be rewritten and solved
for T as
!1
1
1
51
4
T C2
ln X 5
2
1
2
2.1.3
Infrared Camera/Thermometer
5
Q E
d
"T 4
where is the StefanBoltzmann constant,
5:67 108 W=m2 K4 .
MgF2 , ZnSe, and Sapphire (Al2 O3 ) glasses are used
as IR transparent media windows, e.g., MgF2 : wavelength 35:4 m, 95% of IR passing through; ZnSe:
0.520 m, 90%; Al2 O3 : 0.24 m, 85% transparency
for IR.
Since the IR thermometer must adapt itself to the
temperature of the surroundings, it may not measure
the temperature accurately without enough time to
adapt. The measurement results are also aected by
the angle of optical axis to the surface as shown in
Fig. 3. When the angle is larger than 40 , an IR
camera placed at position A gives temperature values
higher than the real ones.
IR cameras have been applied to so-called inverse
heat transfer problems, which concern the identication of unknown temperature distribution or thermal
Figure 3 Ball surface temperature versus observation angle, dball 20 mm. (a) Schematic of measurement; (b) horizontal
temperature distribution of the equator; (c) vertical temperature distribution. (Kobylecki and Horio, 2000.)
4
4
Tbed
Twall
hTbed Twall
1="bed 1="wall 1
where h is the heat transfer coecient, T is temperature, " is emissivity, is the StefanBoltzmann constant, and q is the heat ux.
Since heat transfer in uidized beds is dominated by
surface renewal of bed materials, the bed-to-wall heat
transfer coecient for a vertical surface is much larger
at its leading edge. Accordingly, the heat transfer coefcient determined by small probes can be abnormally
large. Furthermore, depending on gas and solids
motion, surface renewal rate, and the time fraction of
coverage by bed material, the heat transfer coecient
varies for dierent segments of the surface, particularly
for horizontal tubes. The requirement to determine the
instantaneous heat transfer coecient is the low thermal inertia of the probe, such as that used e.g., by
Mickley et al. (1961) or Ma and Zhu (1999) and
shown in Fig. 4ab, respectively.
In some experiments it is necessary to separate the
radiative and convective heat uxes. This can be done
by using two probes of dierent emissivities (e.g.,
Gas Pressure
Gas pressure sensing is the most basic issue in uidization and solids transport. The key points of the sensing
system are the volumetric capacity of its tubing and the
Figure 4 Examples of heat transfer probes. (a) Low thermal capacity instantaneous type of Mickley et al. (1961); (b) miniature
heat transfer probe of Ma and Zhu (1999); (c) probe to measure radiative and total heat ux. (Luan et al., 1999.); (d) radiative
heat ux vs. suspension density. (Luan et al. 1999.) [(a): Reproduced with permission of the American Institute of Chemical
Engineers. Copyright # 1986 AIChE. All right reserved. (b)(d): with permission of Elsevier Science]
7
0:5
p
2
0
1 !=!n 2 4h2 !=!n 2
2.3.2
Figure 5 Silicon semiconductor piezo-resistive pressure sensor. (Chau et al., 1999, with permission of CRC Press.)
tan1
2h!=!n
1 !=!n 2
10
Manometer
Solid Pressure
Figure 7 Eect of tube diameter, dpp , on pressure uctuations. Particles: 68 m FCC, purge gas velocity: 1 m/s, pressure port
length: 2 m dead volume, 450 mm3 . (Xie and Geldart, 1997.) (With permission of Elsevier Science)
To determine the bulk solid ow in bubbling uidized beds, force sensors have been used in lab
scale and pilot scale units. Examples can be found
in Miyauchi and Morooka (1969), Tomita and
Adachi (1973), Mann and Crosby (1977), Raso et
al. (1983), Fasching and Smith (1993), and Ito et
al. (1999).
2.3.5
Figure 9 shows two typical probes to measure viscosity: a rotating drum and a rotating bar. It shows also
some measurement results obtained with using a rotating bar. In case the internals are heavily placed in the
bed, the solid-to-internal friction may aect solids circulation. In such cases, solid viscosity also has to be
taken into account in the scaling experiments, as discussed e.g., by Naruse et al. (1996).
The main parameters to describe light-related phenomena are classied as radiant quantities and luminous
quantities; the latter are psychophysical parameters.
Table 2 shows their denitions and units.
Total optical energy input i consists of three components: reection r , transmission t , and a
adsorption:
i r t a
11
The commonly used parameters reectance and transmittance are dened as
12
Reflectance r
i
Transmittance
t
i
13
Figure 8 Frequency response of a particle pressure sensor to the gas pressure change. (a) Setup to test the dynamic response; (b)
gas pressure (measured by a pressure transducer) and the phase shift. (Campbell and Wang, 1990.) (With permission of Institute
of Physics)
Figure 9 Viscosity measurements in uidized beds. (a) Viscosity measurement with rotating drum. (Schugerl et al., 1961.) (With
permission of Elsevier Science) (b) T-type rotating bar to measure viscosity. (Naruse et al., 1996.) (c) Viscosity measured with a
bar viscosimeter. (Naruse et al., 1996.) Dt 105 mm; bed material: SiO2 ; dp 582 m and 150 m:
dI
I"p
dl
14
15
Light Sources
Denition
Radiant
energy
Qe
Radiant
ux
Radiant
exitance
Irradiance
Radiant
Radiance
Luminous quantities
Unit
Parameter
Denition
v dt
Unit
lm
s
Quantity of
light
Qv
e
v
Km e
V
d
lm cd
sr
Me
de =dA
W=m2
Mv
dv =dA
lm=m2
Ee
Ie
de =dA
de =d
W=m2
W=sr
Luminous
ux
Luminous
exitance
Illuminance
Luminous
intensity
Ev
Iv
dv =dA
dv =d
lx lm=m2
cd
Le
d 2 e
dAd cos
W=m2 sr
Lv
d 2 v
dAd cos
cd=m2
Luminance
dA is innitesimally small area, d is an innitesimally small solid angle, is the angle between a normal direction to the surface
and a given direction, and V
is the spectral luminous efciency.
Figure 10 Distributions of scattering angle. (a) Mie scattering; (b) intermediate; (c) Rayleigh scattering.
s
i
p
16
where
s ,
i , and
p are the wavelengths of signal, idler,
and pumping waves, respectively.
Table 3
Photodetectors
Optical Fibers
Application
Halogen
Mercury
Xenon
Metal halide
Light emitting diode (LED)
Laser
2.4.3
Type
Lamp
Gas
Semiconductor
Tunable laser
Reference
Lasers
Material
Peak
wavelength
(nm)
GaP/GaP
GaA1As/GaAs
InGaAlP/GaAs
GaAsP/GaP
700
660
644
635
InGaAlP/GaAs
InGaAlP/GaAs
623
620
CO2
InGaAlP/GaAs
GaAsP/GaP
InGaAlP/GaP
GaAsP/GaP
GaP/GaP
InGaAlP/GaAs
612
610
590
587
570
574
Excimer
GaP/GaP
InGaAlP/GaAs
GaP/GaP
Blue
SiC/SiC
GaN/sapphire
Infrared GaAs
GaAs:Si
565
562
555
470
450
910
940
Color
Red
Orange
Yellow
Green
Yellow
Green
Type
Gas
Solid
Liquid
Material
Wavelength
Power
Operation
HeNe
632.8 nm
(most cases)
0.1100
mW
Continuous
Argon ion
488 nm,
515.4 nm,
and others
10:6 m,
9:4 m,
and others
100 mW
20 W
Continuous
1W15 kW Continuous
> 1 MW Pulse
150 W
Pulse
> 2 kW
Continuous
and pulse
Nd: YAG
1064 nm
Nd:YAG
(OPO)
300950 nm
Continuous
and pulse
Organic
dye
2501100 nm
Continuous
and pulse
Continuous
Figure 11 Example of emission spectrum of GaAlAs LED and lamps. (a) GaAlAs LED. (Hamamatsu Photonics, 1999c.) (b)
Various lamps. (Hamamatsu Photonics, 2000.) (c) Low-pressure mercury lamp. (Hamamatsu Photonics, 2000.) (d) Metal halide
lamp (150 W). (Hamamatsu Photonics, 2000.)
Figure 12 OPO laser conguration and wavelength. (a) Wideband OPO; (b) narrow band OPO; (c) eect of phase matching
angle on idler wavelength of BBO crystal OPO. (Muraoka and Maeda, 1995.) (With permission of Sangyotosho)
17
sin
2
18
Figure 14 Structures and characteristics of typical point sensors. (a) Spectral responses of photoconductive sensors.
(Hamamatsu Photonics, 1999b.) Note: Luminous eciency equals V
of Table 2. (b) Spectral responses of photodiodes.
(Hamamatsu Photonics, 1998b.) (c) Spectral response of a photomultiplier. (Hamamatsu Photonics, 1998a.)
receiving systems, among which a dual-beam forwardscattering mode is the most popular and commercially
available. It is popular because the signal can be
easily detected and processed, and high-frequency
resolution is not required. In the dual-beam forward-scattering method a laser beam is split in two.
The two beams create interference fringes as illu-
2 sin=2
19
where
is laser wavelength and is the angle between
the beams.
Figure 15 Spectral attenuation and transmittance of optical bers. (a) Spectral attenuation of PMMA optical ber (diameter:
0.53 mm, room temperature, 65% RH). (Tolay Industries, 1991.) (b) Spectral transmittance of a 1 m long quartz optical ber.
A: SiO2 core; B: GeO2 =SiO2 core. (Fujikura Ltd., 1988.)
Type
Fiber diameter
Target parameter/s
Reference
Reection type
a
Single ber
df dp
p
Bubble fraction
Two-ber
df > dp
p
Bubble fraction
Solid loading
and velocity
df dp
Three-ber
df dp
vp
(one-dimensional)
Bubble detection
Multiber
df dp
vp
(two-dimensional)
Space lter
df dp
vp
(one or two
dimensional)
Fiber bundle
df dp
Particle passage
vp and p
Bore
scope
Cluster image
Particle image
Table 5 continued
Structure
Type
Fiber diameter
Target parameter/s
Reference
df dp
Gas concentration
Tracer gas passage
Images of clusters
and strands
Transmission type
g
Two in line
Bore
scope
h
Two parallel
df dp
Bubble passage
Curved ber
type
df dp
Bubble passage
Bore
scope
Particle image
Burning char
temperature of
pyrometer
d f dp
Type
Pair
Target parameter
Correlation
Multi or
array
Distribution of bubbles
Bubble shape
Jet shape
Two parameter
measurement
References
sin
20
2
This expression is the same as Eq. (18) obtained from
Doppler shifts. Equation (20) is usually called the
fringe model for the beat frequency in LDV.
Figure 16 Examples of optical ber probes. (a) Optical ber probe for simultaneous measurements of solid and gas velocities
(solid: light reection measurement; gas: transmission measurement with O3 tracer). (Horio et al., 1992a.) (b) Water cooled
optical ber probe. (Werther and Hage, 1996.) (With permission of United Engineering Foundation)
Figure 17 Conguration and characteristics of reection type optical ber probes. (a) Fiber congurations. (b) The intensity of
reected light as a function of the distance to a at target. (Krohn, 1986. Cf. also Reh and Li, 1991.) (With permission of
International Society for Optical Engineering)
v p fD
2 sin=2
21
dp H
f
2 l
22
Figure 18 Calibration curves for FCC particles. (a) Without Larostat. (b) With 0.5 wt% of Larostat. (Herbert et al., 1994.)
(With permission of Elsevier Science)
Figure 19 Fringes and output signals of a dual-beam forward-scattering LDV. (a) Interference fringes and a measurement
volume. (b) Typical signal from freeboard of a uidized bed. (Levy, 1986.) (Reproduced with permission of the American
Institute of Chemical Engineers. Copyright # 1986 AIChE. All rights reserved.)
up to 0.21. More information on LDV and its applications can be found e.g., in Watrasiewicz and Rudd
(1976), Drain (1980), Arastopour and Yang (1992),
van de Wall and Soo (1994), and Onofri and
Tadrist (1999).
2.4.7
Space Filters
Space ltering is a method of determining particle velocity with the spatial light distribution of a gratelike
structure including interference fringes. Dual beam
mode LDV is one type of space lter; the others
include a reticle type lter and a lter with gratinglike
detectors.
In the case of reticle-type lters, the illuminated
particle passes through the measurement volume and
reects light to the grating, thus being a source of a
periodical signal (Fiedler et al., 1997). From the frequency f of this signal, the particle velocity vp is determined as
fg
vp
23
24
25
Figure 20 Processing scheme for gassolid two-phase ow measurement (Tsuji and Morikawa, 1982.) and an example of
experimental data. (Lehner et al., 1999.) (a) Optical arrangement and signal processing system. (b) Time chart of the signals.
(c) Experimental data obtained with LDV. Description of the signals of case (b): A: raw signal; B: frequency tracker signal of A;
C: synchronized pulse, which the tracer gives at every signal processing; D: Pedestal component of A after low-pass ltering; E:
square wave corresponding to the pedestal amplitude larger than the threshold T1 ; F: square wave corresponding to the pedestal
amplitude smaller than the threshold T2 ; G: derived from E and F through an exclusive NOR gate; H: alternative signal of A
(Doppler amplitude, burst signal); I: signal produced by comparator; J: square wave corresponding to the envelope of Doppler
amplitude larger than threshold Td ; K: logical product G and J by AND gate; L: signal obtained from K and C through the Dip-op circuit; M: signal obtained from F and C through the D-ip-op circuit. The main characteristics of the optical system:
15 mW HeNe laser of wavelength 632.8 nm used as light source, 25 ,
1:2 m, w0 76:2 m, l 250 m, 41 fringes in
the scattering volume, frequency-to-velocity conversion factor 0.6838 ms1 MHz1 . (a,b: Reproduced with permission of
Cambridge University Press; c: with permission of DECHEMA e.v.)
Figure 21 Doppler signals corresponding to a single particle passage received by dierent photomultipliers. (van den Moortel et
al., 1997.) (With permission of Elsevier Science)
kA
d
26
k
kvacuum
27
28
Microphones
Figure 22 Schematic of a light reection type space lter (a), tracer particle passage (b), and wave pattern of output signals from
light reection type space lter in a gasliquid system (c). (Kamiwano and Saito, 1984.) (Reproduced with permission of the
American Institute of Chemical Engineers. Copyright # 1984 AIChE. All rights reserved.)
Figure 23 CCD type space lters. Grating constant: 64 m; image scale (i.e., the size at grating divided by the size at measuring
point): 0.0621; sampling frequency: 156 kHz. (a) CCD type space lter probe. (b) Block diagram for CCD type space lter.
(Fiedler et al., 1997.) With permission of Elsevier Science
Acceleration sensors
Electrostatic Sensors/Probes
Temperature,
C
Frequency,
Hz
0
20
20
25
24
23
20
25
20
20
< 3 106
Low
106
106
106
106
106
106
106
106
1.001
80.37
4.58.5
3.33
6.26
2.26
5.6
3.1
3.85
46
Figure 24 An example of a capacitance probe. (a) High-temperature two-needle capacitance probe of Wiesendorf et al. (1999)
(With permission of American Society of Mechanical Engineers). (b) Output signals from a 12 mW (Chalmers) and a 250 MW
(Gardanne) CFB boiler obtained with the probe shown in (a). (Johnsson et al., 1999.) With permission of DECHEMA e.v.
2.9
Humidity Sensors
2.10
Radioactive Sensors
radiation. A popular type is the sodium iodide scintillator activated with thallium (NaI(Tl)). However, this
material is hygroscopic and must be encapsulated.
Other commonly used scintillators contain cesium
iodide activated with thallium or sodium, CsI(Tl),
CsI(Na).
Examples of some radioactive isotopes used for uidized bed tracking are shown in Table 7. As discussed
e.g., by Seville et al. (1995) and Benton and Parker
(1996), the scattering or absorption of the emitted
radiation may lead to incorrect determination of tracer
position. However, since the scattered rays have lower
energy than the correct ones, by removing the scatter data one may avoid this problem. Application of
radioactive particles as tracers is relatively easy and
inexpensive, even for large uidized bed plants.
However, it may sometimes disturb the ow and denitely requires strict safety regulations during the mea-
Figure 25 Faraday cage used by Fasso et al. (1982) (With permission of Elsevier Science) (a) and examples of electrostatic
probes of Soo et al. (1964) (Reprinted with permission from Ind. Eng. Chem. Fundam., 1964, 3, 98106 Copyright # 1964
American Chemical Society) (b) and Boland and Geldart (1971) (With permission of Elsevier Science) (c).
Figure 26 Schematic of a zirconia sensor and an example of the output signal (measurement in a 12 MW CFB boiler).
(Johnsson et al., 1999.) With permission of DECHEMA e.v.
3.
Tracers
4.
Solid Tracers
To obtain information on solids movement and mixing, the solid tracers are fed into a uidized bed and
detected either in situ, as in the case of isotope tracers, or by sampling and analysis. It is assumed that
the tracers behave like the solids in the ow. Some
examples of commonly used tracers are shown in
Table 8.
In choosing the radioactive tracer one should take
into consideration the following aspects:
1.
2.
The purity of radionuclide; to assure that rays are emitted from disintegration of the
selected radionuclide and not from some of its
impurities.
Rb
O15
N13
C11
Ga68
F18
Ti45
Half-life
78 s
122 s
10 min
20.3 min
68 min
110 min
3.1 h
Nuclide
62
62
Zn =Cu
Ga66
Cu64
I124
Ge68 =Ga68
Na22
Half-life
9.2 h
9.7 h
12.7 h
4.2 days
271 days
2.6 years
2.11.2
Tracer
NaCl tracer
Limestone
Phosphor
Color
Color
Colored particles
Colored particles
Fracture
Spheres of dierent
strengths
Paint pigment attached
to coal particles
Fluorescent
Fluorescent
Magnetic
Thermal
Radioactive
Co60
Na22
Na24
Radioactive Na2 CO3
Ga68
Radioactive SiO2
Sc46 oxide
Radioactive tracers of
various sizes and
densities
F18
Au198
Positron emission
Irradiated Cu
Mn56 and cordierite
Remarks
FCC reactor, the tracer detected by electric
conductivity of samplewater solution
Diusivity in a large-scale FB boiler
Eect of tracer size on solids mixing in the
0.14 m i.d., 10.4 m high FCC riser
Semicircular uidized bed, video recording
Video recording of particle velocity in 2D
uidized bed
Kinetic forces acting on particles in a uidized
bed determined from analysis of tracer particles
Particle velocity in the cocurrent downward ow
of coal and air
Tracer detected by LDV, 319 176 mm crosssection FB
Particles illuminated by UV light, reected light
detected by optical bers
Solids mixing in a bed of FCC, Dt 0:15 m i.d.
Bed operated in a bubbling, slugging, and
turbulent regime
Solids mixing studied by magnetic separation
Measurement of the ow of dark colored
particles
Temperature measurements
Observation of particle motion in a FB
Studies on particle motion in a 0.03 m i.d. gassolid uidized bed. Scintillation detector placed
below the uidized column
Particle motion in a FB of glass beads
Scintillation counter placed above the bed
RTD measurement in CFBs of various sizes
Particle tracking in a CFB
Studies on RTD in a 82.8 mm i.d. CFB
Studies on particle motion and solids mixing in a
3D liquidsolid uidized bed
Measurement of local particle velocities in a cold
CFB
Particle tracking in a 0.15 m i.d. sand bed
Studies on the hydrodynamics of a CFB tracer
Tracer velocity measured 250 times per second
Particle tracking, 3D resolution below 15 mm
Studies on particles tracking in the 80 MWth CFB
boiler. 16 NaI(T1) detectors used, data
collected with the frequency of 20 Hz
Solid concentration axial prole and
instantaneous local values measured in a 0.4 m
i.d. 15.6 m CFB riser, Gs up to 50 kg/m2 s
Author(s)
Bader et al. (1988)
Ito et al. (1999)
Du et al. (1999)
Yang and Keairns (1982)
Gbavcic et al. (1990)
Kono et al. (1987)
Brewster and Seader (1980)
Hamdullahpur et al. (1987)
Kojima et al. (1989)
Nowak et al. (1991)
Avidan and Yerushalmi
(1985)
Horio et al. (1986a)
Yamaki et al. (1994)
Westphalen and
Glicksman (1995)
Kondulov et al. (1964)
Borlai et al. (1967)
Schlichthaerle and
Werther (1999)
Tracer type
Comments
CH4
CH4
O3
O3
Ye et al. (1999)
O3
CO2
He
He
He
He
He
He
He
SF6
Li and Wu (1991)
Warm gas
injection,
liquid N2
injection
H2 pulses
Ar pulses
Studies on axial gas mixing and RTD in the CFB containing FCC
particles (dp 58 m)
Studies on gas residence time, tracer detected by two rapid response mass
spectrometers, reactor 0.09 m i.d. CFB
Figure 27 Path of meandering gas in a uidized bed (a) and the variation of tracer concentration (b). (From Jovanovic et al.,
1980.)
Gas Sampling
3
3.1
Remarks
Dust sampling from a PFBC
Nonisokinetic solids sampling from a 16.1 cm i.d. CFB riser operated at various temperatures
Solids suction to measure local solids mass uxes and PSD in a multisolid FB
Application of a water-cooled sampling probe to measure the amount of CaS in a 12 MW
CFB boiler
Solids sampling from a CFB sludge waste combustor to detect solids PSD and composition.
Application of a particle sampling probe to evaluate hot gas cleanup system
Application of a water-cooled solids mass ux meter to measure local Gs at the furnace
outlet of a 75 t/h CFB boiler
13 mm i.d. suction probe successfully applied to a CFB boiler
Autocorrelation
Figure 29 High temperature solids sampling probe of Mattisson and Lyngfelt (1995). (With permission of American Society of
Mechanical Engineers)
Remarks
Flue gas sampling to measure the concentration of Hg, F, and the solids fraction
Continuous He injection into a FB. Gas concentration proles measured
Detection of He injected into an FCC riser
Application of CH4 tracer to a CFB riser
Detection of bubbles in a FB. Bubble passage determined using a dual static pressure probe
Gas sampling probe, shown in Fig. 30c, applied to sample gas from bubble and
emulsion phases separately. Reactor: 0.1 m i.d. 0.7 m FBC; bed material: sand, fuel: coal
Gas sampling system applied to measure the concentration of NH3 and HCN inside a
12 mW CFB boiler
Particle probe,
-probe, gas probe and alkali metal sampling probe used to monitor a
20 MWe CFB combustor
Alkali-sampling probe applied to characterize sorbents for a PFBC.
Alkali concentration measurements in a 10 MW PCFB using laser uorescence and plasmas
spectroscopy sensors
Gas sampling from a 12 MW CFB boiler
Simultaneous measurement of N2 O and NOx concentration, as well as particle
temperature in the vicinity of a burning char particle
Figure 30 Gas sampling system (a), probe details (b), and bubble passage determined from pressure drop (c). (Naruse et al.,
1994.) (With permission of Combustion Institute)
Copyright 2003 by Taylor & Francis Group LLC
Arithmetic
moving
average
Denition/explanation
Signal
Pm
wjxi j
Pm
yi jm
i m 1;
n m
jm wj
wj: window function
yi
xt
Filtering in
frequency region
m
X
1
xi j
2m 1 jm
Fourier
!
1
X!W!
Pressure
Fourier inverse
!
X!
Remarks
S/N improvement,
equivalent to simple
average for random
noise
yt
Reference
Hatano et al.
(1999)
Eective when
noise frequency
is much higher
than the signals
1
Averaging
Root mean
square (RMS)
Probability
distribution
function (PDF)
W!
1! !1 !1 : cut off
0! > !1 frequency
N
1X
x
N i1 i
xrms
px
s
x2
2
x1 x x dx
x2 x1
n
1 lim X
t1
x T!1 i1 T
Pressure
Pressure
uctuation
Phase transition
Light
LDV
Light
Autocorrelation
function
Cross-correlation
function
R
lim
T!1 0
Rxy
lim
xtxt dt
T!1 0
xtyt
S/N improvement,
elimination of
signal from
individual particle
Yerushalmi
and Cankurt
(1979)
Determination of threshold
value to distinguish:
Bubble and emulsion phases Horio et al. (1985)
Gas and particle signals in
Tsuji and
pneumatic transport
Morikawa (1982)
Cluster and lean phase
Horio et al.
(1988)
Pressure
Slugging condition
Pressure
Light
Light
Heat transfer
Bubbling condition
Bubbling regularity
Freeboard turbulence
Dynamics of CFB
Pressure
Pressure
Bubbling condition
Pressure wave velocity
Light
Particle velocity
Broadhurst and
Becker (1976)
Fan et al. (1981)
Greon et al. (1997)
Horio et al. (1980a)
Li et al. (1993)
Fan et al. (1981)
Roy and
Davidson (1990)
Horio et al. (1980a)
Table 12 continued
Examples
Processing
Power spectrum
Fractal analysis
Denition/explanation
Sx !
2
lim 1
XT !
T!1 T
Signal
Bubbling condition
Solid structure in a CFB
Pressure
Phase transition
Heat transfer
Pressure
(Maximum
entropy
method,MEM)
Dynamics of a CFB
Analysis of long range
force
Image
Pressure
Cluster shape
Chaos analysis
Solid
momentum
probe signal
Chaos analysis
Pressure, light,
-ray porosity,
hot wire prove
signal, etc.
Quantication of the
Schouten and van
den Bleek (1992),
dynamics and phase
transition in the BFB,
Marcocchella
CFB, mixing chamber, etc. et al. (1997),
Ji et al. (2000),
Huilin et al.
(1995)
Briens et al.
(1997)
Evaluated by Kolmogorov
entropy, correlation
dimension, Lyapunov
exponent, presence of the
chaotic attractor, Hurst
coecient, etc.
Wavelet
transform
Cross-correlation
Cross-correlation, dened in Table 12, provides information about the statistical mean delay time between
two signals if there is any correlation between them.
Figure 35 shows an example of a bed pressure uctuation signal in a bubbling uidized bed. From crosscorrelation shown in Fig. 35b the obtained mean
delay time is 0.12 s.
The degree of interaction between two signals can
also be determined from cross-power spectrum, coherence (van den Schaaf et al., 1999b), or cross-covariance
function (Greon et al., 1997). In case of wide solid
Reference
Pressure
Pressure
Chaos analysis
3.1.2
Remarks
3.1.3
Power Spectrum
Figure 31 The moving average with second and third polynomial adaptive window. Left: original; center: 7 points averaging;
right: 21 points averaging. (Minami, 1986.)
Wavelet Transform
t
t 2
1
p
exp 0:5
a
a
2
29
j!0 t
exp
a
The wavelet transform of a signal xt), Wx ; a), is
dened by
1
t
1
dt
30
Wx ; a p
xt
a
a 1
where a is a scale dened by using the representative
frequency of the signal !0 , as
a
!0
!
Figure 32 Eect of averaging on the shape of signal: (a) rst, (b) second, (c) Nth, (d) nal. (Minami, 1986.)
31
Figure 33 Elimination of high-frequency noises by low-pass ltering. (a) Original signal, (b) Real part of Fourier transform, (c)
Processed signal. (Minami, 1986.)
Figure 34 Autocorrelation function and power spectrum. Upper row: original pressure signal; middle row: autocorrelation (no.
of lags 20 0:45 s; sampling interval: 22.5 ms; data points: 1000; Dt 0:1 m, u0 0:1 m/s; bed material: silica sand, dp 0:183
mm); bottom row: power spectrum. (a) Smoothly uidized bubbling bed L=Dt 1); (b) slugging bed (L=Dt 2). (Broadhurst
and Becker, 1976.)
Figure 35 Pressure uctuation signal from a slugging uidized bed and their cross-correlation function. (Roy et al., 1990.)
At 0:12 0:12 m2 ; Lmf 1:6 m; silica sand bed, dp 240 m. (a) Original signal from pressure probes (upper row: 1.5 m
above the distributor; lower row: 0.07 m above the distributor. --- is the theoretical signal from Kehoe and Davidson, 1973). (b)
Cross-correlation function. (With permission of Elsevier Science)
Figure 36 Power spectrum density of pressure uctuations in a bubbling uidized bed of sand (van den Schaaf et al., 1999a). Dt :
0.80 m; u0 : 0.44 m/s; dp : 390 m; settled bed height 2:19 m. (With permission of American Society of Mechanical Engineers)
1
1
t dt 0
32-1
2
t dt 1
32-2
1
1
Visual Observation
3.2.2
Laser Imaging
Figure 38 Spectrum of mercury lamp obtained by FFT and MEM. (Kawata et al., 1983.)
Measurement
Derived information
Reference
Hee et al. (1997)
Orthonormal wavelet
transform
Pressure
uctuation
Morlet wavelet
Heat ux
Decomposition of the signals into long-term and shortterm correlation components; one represented by the
self similarity Hursts parameter of fractional Brownian
motion originated by bubble, and another represented
by the intensity parameter of Gaussian white noise
originated by gas jetting, small bubble formation, and
turbulence of uidized beds, etc.
Self-similar bifurcation and trifurcation phenomena
Discrete wavelet
transform based on
Mallet and Zhang
Discrete analog of
wavelet transform
(orthogonal wavelet
basis functions by
dilating and translating
in discrete steps
Mallats pyramidal
algorithm used for
computing onedimensional orthogonal
wavelet transform
Gabor wavelet
Optical ber
probe signal
Pressure
uctuation
Signal denoising
Roy et al.
(1999)
Pressure
uctuation
Lu and Li
(1999)
Signal from an
acceleration
sensor
Fukayama
et al. (2000)
Figure 39 Examples of Gabor wavelet and Gabor function. (a) Gabor wavelet (dilated window); (b) Gabor function (xed
window width). (Sheng, 1996.) (With permission of CRC Press)
3.2.3
As for an example, let us suppose that for the construction of a tomographic image we use light of initial
IL I0 eL
33
35
36
where
aij 1 1 2 2 lij
37
I
bi ln Li 2 2 Li
I0
38
Figure 40 Laser sheet technique, image modication, and laser sheet picturing in dense suspension. (a) Scanning laser sheet
imaging (Tsukada et al., 1997); (b) horizontal proles of gray scales of the images before and after image modication (r0 0:5
m, cellophane tape correction by interpolation) (Tsukada et al., 1997); (c) hood system of Kuroki and Horio (1994); (d)
endoscopic system of Werther and Rudnick (1996, courtesy of Prof. Werther); (e) 3D image obtained from scanning laser
sheet images (Horio and Ito, 1997)upper: turbulent bed (u0 0:58 m/s; Gs 0:045 kg=m2 s), and lower: fast bed (u0 1:1
m/s, Gs 0:087 kg/m2 s).
Pl; f x; y ds
39
L
When f x; y is a continuous function with a continuous rst derivative, Radon (1917) reported that
f x; y) can be reconstructed from an innite set of
line integrals Pl; by
Copyright 2003 by Taylor & Francis Group LLC
1
f x; y 2
4
2 1
0
l
l
x
cos
y sin
1
40
@Pl;
dl d
@l
where l x cos y sin is the perpendicular distance of the point x; y from the line l.
In reality the reconstruction is always done as an
approximation of Radons solution for a certain number of projections. As shown in Fig. 42b, for each angle
of projection , the value of Pl; depends on the
Figure 41 PIV setup of Chen and Fan (1992). With permission of Elsevier Science.
Figure 42 An algebraic approach to image reconstruction in CT (a) and the geometry for CT image reconstruction (b).
(Adopted from Kumar and Dudukovic, 1997.) With permission of Elsevier Science.
42
Normalization of f x; yapprox is achieved by multiplying by 1=2.
The quality of image reconstruction depends on the
number of projections and detectors, the speed of measurement and of the data acquisition system, and the
homogeneity of the measurement volume. More on
image reconstruction can be found, e.g., in Kak and
Slaney (1988); Kalender (2000); and Kumar and
Dudukovic (1997).
The rst reported application of tomography was
probably that of Grohse (1955), who investigated the
variation in density of a bed of silicon powder as a
function of uidizing velocity. At almost the same
time Bartholomev and Casagrande (1957) used
Co60 as -ray source to monitor a 20:4 00 i.d. FCC
unit and detected the -rays by a GeigerMuller
tube.
In capacitance tomography, the measurement
setup consists of an array of electrodes installed
around the external surface of the monitored system,
and measurements of capacitance between various
pairs of electrodes are performed, so that data is
received on the capacitance in various directions
within the system. Then, from the algorithm
described above, the volume fraction of solids or
gas within the system can be calculated. Examples
of various sources for uidized bed CTs are presented
in Table 14. An example setup and results are shown
in Fig. 43.
Positron emission tomography (PET) uses radioactive tracers that decay via the emission of a positron.
It results in the production of two back-to-back rays, from which the tracer position can be determined. More information can be found, e.g., in
Hawkesworth et al. (1991). Another noninvasive
method is nuclear magnetic resonance (NMR) imaging. The measurements are conducted based on the
paramagnetic properties of atom nuclei. During the
experiments, external signals (i.e., radio frequency
pulses and magnetic eld gradient pulses) interact
with the nucleus spin system positioned in a static
magnetic eld. NMR imaging enables us to observe
dense ows, but it can only be performed with watercontaining systems. Details can be found, e.g., in
Pangrle et al. (1992), Shattuck et al. (1997), and
Sederman et al. (2001).
An advantage of tomography is its capability of
monitoring a relatively large volume at the same
moment. However, safety considerations concerning
particularly the emission of x-rays or -rays during
tomographic scanning have to be kept in mind. Using
x-rays instead of -rays enables adjusting the energy of
the emitted radiation by changing the voltage of cathode of an x-ray generator. A disadvantage of the tomographic imaging is the diculty of transforming and
analyzing the huge amount of data from the scanner,
but faster computers should solve this diculty.
3.3.2
Neutron Imaging
Source Type
Remarks
-ray
-ray
-ray, x-ray
x-ray
x-ray
x-ray
Kantzas (1994)
x-ray
x-ray
x-ray
x-ray
x-ray
x-ray
Capacitance
Capacitance
Capacitance
Capacitance
Figure 43 CT scanning system of Halow et al. (1993) (a) (with permission of Elsevier Science) and pseudo 3D image of bubbles and their interface structure (Kai et
al., 1999.) (b)
4.1.1
Dt
Vs
43
butor such as a vertical pressure probe, a wall pressure tap and/or a distributor buried type, or
indirectly from the pressure of the plenum chamber
by subtracting the distributor pressure drop measured
separately. Pressure taps on the freeboard wall are
also necessary in case the gas outlet pressure drop
cannot be negligible. To avoid gas dynamic pressure
the hole of a pressure tap must face parallel to the
main gas ow. To prevent particle entering, one
needs to use some ltering material such as a wire
gauze (Figs. 46a and 46c) or to slant the pressure tap
tube (Fig. 46b). Figure 46d shows immersion type
probes, which can be traversed to measure the pressure distribution.
4.1.2
Deuidization Velocity
Tbi
Tbi Tei
"b P
45
Figure 44 Determination of uidization regime from pressure data. P : theoretical pressure drop bed weight-buoyancy
force)/(total cross section).
The local visible bubble velocity uvb is given by multiplying the voidage by bubble velocity ub measured by
the two needle probes or by a borescope or other image
sensors so that
P
ubi Tbi
uvb ub "b P
(local time average)
Tbi Tei
46
By traversing the point probe across the bed one can
obtain the bubble ow distribution.
4.1.4
Figure 45
Figure 46
47
However, lbi does not immediately give the bubble diameter. To obtain the real bubble diameter, Db , from a
series of data for lbi , let us rst assume (as a simplest
case) that the bubbles are spherical and uniform in
size and that they are uniformly distributed radially
across the bed. The probability, Db ; lb dlb , of
having a pierced bubble length within the interval of
lb lb dlb for bubbles of diameter Db is 8rjdrj=D2b ,
where the radius, r, is dened by r2 D2b lb2 =4
(accordingly we have r dr lb dlb =4. Then, since
8rjdrj=D2b 2lb dlb =D2b , and since lb is less than Db ,
the probability distribution density function for bubble
size Db ; lb is given by
2l
Db ; lb b2
Db
for
0 < lb < D b
48-1)
and
Db ; lb 0
Figure 47
for
lb < 0
or
Db < lb
48-2
Type of signal from capacitance probes. (Lanneau 1960.) With permission of the Institution of Chemical Engineers.
of pierced length lb for the case where Db is distributed, lb : lb can be related to Db and
Db ; lb by
1
Db ; lb dlb Db dDb
52
lb dlb
lb
53
lb
Dierentiating it by lb , we obtain
d
lb ; lb lb
dlb
54
4.2.1
we obtain
@p
p 1 "g
@z
d Db
1
Db ; Db dDb
Db
1
d
1
dlb
0 l ; l dl
b b
b
56
where lb is measured and Db ; lb can be calculated for a bubble of particular shape. Then the average bubble size is given by
1
1=3
3
Db
Db Db dDb
57
0
58
ws
A
kg=m2 s
59
Ws
p 1 "vp dA
60
A
4.1.6
4.3
4.3.1
Author(s)
Lasch et al. (1988)
Bader et al. (1988)
Hartge et al. (1988)
Beaud and Louge (1995)
Ye et al. (1999)
Arena et al. (1988, 1993)
Study on ve ways of
solids mass ux
measurement systems
Semicontinuous solids
mass ux meter
Remarks
Measurements based on periodical opening/closing of the valve(s)
Dicult to apply to hot FB due to valve requirements and its
damage possibility at high temperatures
Measurements of solids mass ux in a CFB of cast steel
Measurements based on switching over the three-way valve
(placed outside the hot reactor) and accumulating solids in a
vessel through the certain time period. No continuous
measurements possible
Measurement method based on eect of electrostatic induction of
charged particles on a surrounding metallic ring
Measurements of solids uxes in a 0.14 m i.d. CFB riser of glass
beads and FCC. The solids were collected in each probe for a
given time period and then weighed. Local particle concentration
was measured using the capacitance probe
Study on application of closing valve in the return leg, observation
of an identiable particle, using a device to record the force
imparted by returning solids form the cyclone, measuring the
pressure drop across the constriction in the return loop, and
estimating solids mass ow from the heat balance on a
calorimetric section in the standpipe
Application of load cell. The solids were accumulated in the device
through a certain time period. Dicult application to hot units
due to load cell requirements
Idea based on measurements of pressure drop and gas velocity.
Assuming that gassolid slip velocity is related to the terminal
velocity of individual particles, the mass ux was calculated.
Idea based on application of weighing chamber and an
electronic balance
Scoop-like device measuring continuously the mass uxes of
various solids including group C
Figure 48 Solids mass ow meters: (a) Horio et al. (1992b), (b) Kobylecki and Horio (2002).
As shown in Fig. 49, channel formation and its collapse and transition to the agglomerating uidization
can be detected from the cyclic response of the bed
pressure drop. However, it should be noted here that
in the cohesive powder uidization the wall eects are
quite important, and the scale eect related to them
cannot be neglected.
Very ne particles are uidized forming agglomerates. Minimum uidization velocity of such agglomerates ua;mf can be measured in the same manner as in the
case of noncohesive particles, i.e., from the decreasing
velocity period (period after #7 in Fig. 49). A particle
recycling device, such as a cyclone or a lter, is necessary for ne powder experiments. From ua;mf , the
apparent agglomerate size da can be determined from
minimum uidization velocity correlations such as
Wen and Yus (1966) by using the particle density
separately determined by mercury porosimetry or
other methods.
Expansion Characteristics
Figure 49 Pressure drop response and uidization behavior of Al2 O3 powder (5.0 m) bed. (Nishii et al., 1993.)
transition velocities should be determined with a column where slugging can be avoided. Furthermore, to
avoid criticism as for the early studies (Geldart and
Rhodes, 1986), the axial solid holdup should be
always monitored, so that the dense bed is kept in
the column, preventing the pressure tap from exposure to the lean phase, which causes the decrease in
pressure uctuation at higher gas velocities and gives
ctitious values for uc and uk . For this issue readers
SUMMARY
Figure 50 Fluidized bed collapse of group A powder (glass ballotini, 26 m), reported by Geldart (1986) (# John Wiley & Sons.
Reproduced with permission) (a), and collapsing of uidized bed reported by Tung and Kwauk (1982) (b).
Figure 51 Expansion curve for group A powder (60 m ballotini). (Abrahamsen and Geldart, 1980.) With permission of
Elsevier Science.
image analysis systems. Data to evaluate sensing systems are included as much as possible for readers convenience. In the last part of the chapter a practical
approach to the diagnostics of uidization systems as
well as powders are discussed.
ACKNOWLEDGMENTS
The authors thank Ms. Malgorzata Kobylecka for her
support in preparing the manuscript.
REFERENCES
Grassler T, Wirth KE. X-ray computed tomographya nonintrusive technique to characterize local solids distribution
in a circulating uidized bed (CFB). In: Werther J, ed.
Proc 6th Conference on CFB. Wurzburg, Germany,
1999, pp 885890.
Greon J, Berben J, Mudde R, van den Akker H. Local bubble properties in 40 and 80 cm diameter uidized beds.
AIChE Symp Ser 93(317):147151, 1997.
Grohse E. Analysis of gas uidized solid systems by x-ray
absorption. AIChE J 1:358365, 1955.
Halow J, Nicoletti P. Observations of uidized bed coalescence using capacitance imaging. Powder Technol.
69(3):255277, 1992.
Halow J, Fasching G, Nicoletti P, Spenik L. Observations of
a uidized bed using capacitance imaging. Chem Eng Sci
48:643659, 1993.
Halow J. Electrical capacitance imaging of uidized beds. In:
Chaouki J, Larachi F, Dudukovic M, eds. Noninvasive
Monitoring of Multiphase Flow. New York: Elsevier
Science, 1997, pp 263307.
Hamamatsu Photonics K.K. CdS Photoconductive Cells.
Catalog No. KCDS 0001E03, 1999b.
Hamamatsu Photonics K.K. Light Sources. Catalog No.
TLSO 0002J04, 2000.
Hamamatsu
Photonics
K.K.
Optosemiconductors.
Condensed Catalog, Catalog No. KOTH 0001E03, 1999a.
Hamamatsu Photonics K.K. Photodiodes. Catalog No. KPD
0001J07, 1998b.
Hamamatsu Photonics K.K. Photomultiplier Tubes and
Related Products. Catalog No. TPMO 0003J02, 1998a.
Hamamatsu Photonics K.K. Solid State Emitters. Catalog
No. KLED 0001E06, 1999c.
Hamdullahpur F, Pegg M, MacKay G. A laser-uorescence
technique for turbulent two-phase ow measurements. Int
J Mult Flow 13(3):379385, 1987.
Hansen P, Bank L, Dam-Johansen K. In-situ measurements
in uidized bed combustors: design and evaluation of four
sample probes. Proc 13th Conf FBC. Orlando, FL, USA,
1995, pp 391398.
Hartge EU, Li Y, Werther J. In: Basu P, ed. Circulating
Fluidized Bed Technology. Oxford: Pergamon Press,
1986, pp 153160.
Hartge EU, Rensner D, Werther J. Solids concentration and
velocity patterns in circulating uidized beds. In: Basu P,
Large J, eds. Circulating Fluidized Bed Technology II.
Oxford: Pergamon, 1988, pp 165180.
Hatano H, Kudo N. Microscope visualization of solid particles in circulating uidized beds. Powder Technol 78:115
119, 1994.
Hatano H, Ogasawara M, Horio M, Hartge EU, Werther J.
Local ow of gas and solids in a circulating uidized bed.
Proc 6th SCEJ Symp Fluidization, 1993, pp 107114.
Hatano H, Matsuda S, Takeuchi H, Pyatenko A, Tsuchiya
K. Local interactive patterns of dispersed and swarm particles in a circulating uidized bed riser. Ind Eng Chem
Res 35:43604365, 1996.
Rhodes M, Wang X. Studying fast uidization using electrical capacitance tomography. Werther J, ed. Proc 6th Conf
CFB. Wurzburg, Germany, 1999, pp 891896.
Rivault P, Nguyen C, Laguerie C, Bernard J. Countercurrent
stripping dense circulating beds eect of the baes. Prep
of the Fluidization VIII, Tours, France, 1995, pp 491498.
Rix S, Glass D, Greated C. Preliminary studies of elutriation
from gas-uidised beds using particle image velocimetry.
Chem Eng Sci 51(13):34793489, 1996.
Ross K, Pence D. Wavelet and fractal analysis of uidized
bed heat transfer, HTD 351. Proc ASME Heat Transfer
Div 1:245253, 1997.
Ross I, Patel M, Davidson J. The temperature of burning
carbon particles in uidised beds. Trans Inst Chem Eng
59:8388, 1981.
Roy R, Davidson J, Tuponogov V. The velocity of sound in
uidized beds. Chem Eng Sci 45:32333245, 1990.
Roy M, Kumar V, Kulkarni B, Sanderson J, Rodes M, van
der Stappen M. Simple denoising algorithm using wavelet
transform. AIChE J 45:24612466, 1999.
Sauer S, Wallen V. Mixing in experiments and modelling of
ABFBC. Proc 15th Conf Fluidized Bed Combustion,
Savannah, Georgia, USA, 1999.
Schlichthaerle P, Werther J. Axial pressure proles and solids
concentration distributions in the CFB bottom zone.
Werther J, ed. Proc 6th Conf CFB. Wurzburg,
Germany, 1999, pp 185190.
Schoenfelder H, Kruse M, Werther J. Two dimensional
model for circulating uidized-bed reactors. AIChE J
40:18751888, 1996.
Schouten J, van den Bleek C. Chaotic hydrodynamics of
uidization: consequences for scaling and modeling of
uid bed reactors. AIChE Symp Ser 88(289):7084, 1992.
Schugerl K, Merz M, Fetting F. Rheological characteristics
of gassolid uidized bed system. Chem Eng Sci 15:138,
1961.
Schuurmans H. Measurements in a commercial catalytic
cracker unit. Ind Eng Chem Process Des Dev
19:267271, 1980.
Sederman A, Alexander P, Gladden L. Structure of packed
beds probed by magnetic resonance imaging. Powder
Technol 117:255269, 2001.
Seo Y, Gidaspow D. An x-ray--ray method of measurement
of binary solids concentrations and voids in uidized
beds. Ind Eng Chem Res 26:16221628, 1987.
Seville J, Morgan J, Clift R. Tomographic determination of
the voidage structure of gas uidized beds in the jet
region. In: Ostregaard K, Sorensen A, eds. Fluidization
V. New York: United Engineering Foundation, 1986,
pp 8794.
Seville J, Simons S, Broadbent C, Martin T, Parker D,
Beynon T. Particle velocities in gasuidized beds. Prep
Fluidization VIII, Tours, France, 1995, pp 319326.
Shattuck M, Behringer R, Johnson G, Georgiadis J.
Convection and ow in porous media. Visualization by
26
LiquidSolids Fluidization
Norman Epstein
University of British Columbia, Vancouver, British Columbia, Canada
ORIGINS
2
2.1
PARTICULATE FLUIDIZATION
Introduction
Let us start with the most conventional of liquiduidized beds shown schematically in Fig. 1. The
cylindrical vertical column is commonly circular in
cross section, though it may also be rectangular, e.g.,
square. The function of the calming or homogenizing
section, usually located upstream of the distributor, is
to produce, in conjunction with the distributor, as
radially uniform a liquid velocity distribution in the
uidized bed as possible, thus eliminating or at least
minimizing any tendency toward channeling or bulk
circulation (gulf streaming). The system shown in
Fig. 1 is a semibatch (or semicontinuous) operation,
i.e., continuous ow of liquid for an unchanging
batch of the uidized solids, which in this case are
bottom-restrained. We shall see later that many other
Figure 1
However, to get from Eq. (7) to Eq. (8), one must write
that
dP
rB g
dz
so that the eective or dynamic buoyancy in the uidized state is considerably greater than under static
conditions. Combination of Eqs. (1) and (2) gives
FD V rp r g
6
while combination of Eqs. (1), (3), and (4) results in
7
FD V rp r eg
Note that, by virtue of the greater value in Eq. (1) of B
given by Eq. (3) than by Eq. (2), FD via Eq. (7) is less
than FD via Eq. (6) by a factor of e(emf < e < 1).
Advocates of Eq. (6) argue that it is the drag force
FD that causes the dissipation of mechanical energy
and that this dissipation must include the energy
required to convert the original static buoyancy of
the particle in the unuidized state to the subsequent
dynamic buoyancy in the uidized state. Partisans on
both sides of the argument agree that the frictional
pressure gradient, dpf =dz, of the uidized bed must
be given by the specic weight of the suspension
corrected for hydrostatic head, i.e.,
dp
f rB g rg
dz
rp 1 e reg rg
8b
1 erp rg
8c
8a
V=1 e
1 erp g
Ve=1 e
1 erp rg
If, as is commonly assumed, dpf =dz represents frictional force per bed volume (particles plus liquid)
rather than simply frictional force per volume of liquid
in the bed, then that is an additional argument in favor
of Eqs. (2) and (6) over (3) and (7).
We will here adopt the buoyancy convention incorporated in Eqs. (2) and (6) rather than (3) and (7).
However, since the latter convention has made signicant inroads into the literature, the reader is referred to
Table 1 of Khan and Richardson (1990) for useful drag
coecient relationships based on the above alternate
conventions, as well as on alternate denitions of the
characteristic liquid velocity, and to conversions by
Jean and Fan (1992) of several equations incorporating
buoyancy as dened by Eq. (3) [derived by Foscolo
et al. (1983, 1989), Foscolo and Gibilaro (1984, 1987)
and Gibilaro et al. (1985a, 1986)] to the corresponding
equations based on buoyancy as dened by Eq. (2).
2.2
Hydrodynamic Representation
10
M
rp AL
pf
M
constant
rp rg rp A
Figure 2 Frictional pressure drop as a function of liquid supercial velocity for monodispersed particles.
12
13
P
Lg
14
MINIMUM FLUIDIZATION
2 2 3
3
fd
f dp emf
p emf
15
150 1 emf
fe3 Ar
0
16
Remf mf
1:75
1:75
f
where Remf dp Umf r m and Ar dp3 rrp rg=m2 :
The physically realistic solution of quadratic Eq. (16) is
Re2mf
17
where
C1
1501 emf
1 emf
42:86
2f1:75
f
18
fe3mf
0:5714fe3mf
1:75
19
and
C2
Chen, 1987
Chyang and Huang, 1988
Tannous et al., 1994
C1
C2
Applicability claims
33.7
25.46
25.7
25.28
25.25
27.2
25.74
31.6
0.0408
0.0382
0.0365
0.0571
0.0651
0.0408
e4:8
mf =0:336
0.0425
28.7
0.0494
33.95
29.5
0.0465
0.0357
32.1
0.0571
25.2
0.0672
33.67f0:1
0.0408/f0:45
33.3
25.83
0.0333
0.043
0:415
f0:483
20
1
N
X
i1
vi =dpi
22
Figure 6 Frictional pressure drop vs. liquid supercial velocity for particles of mixed sizes, shapes, and/or densities.
(After Couderc, 1985.)
rp
Mi
i1
N
X
23
Mi =rpi
i1
4.1 Introduction
Crucial to the design of a uidized bed is a quantitative
knowledge of the bed expansion as a function of the
liquid supercial velocity, i.e., of either L or e [the two
being interrelated by Eq. (13)] as a function of U. The
chaotic behavior of a uidized bed, in contrast, say, to
viscous ow through an ordered (e.g., simple cubic)
array of immobilized spheres, renders measurements
from the former inherently less reproducible than
from the latter.
The equations that have been developed for predicting the frictional pressure drop due to ow through a
xed, including an expanded, packed bed (of which
there are many in the literature) yield a relationship
between dpf =dz, e, and U that, when combined with
Eq. (8) to eliminate dpf =dz, give an equation relating
e to U. Because, as mentioned earlier, particles in a
uidized bed, in contrast to those in a rigid array,
are relatively free to move around and arrange themselves so as to minimize or at least decrease the resistance to ow (Richardson and Meikle, 1961), therefore
for a given value of e, U for a uidized bed will usually
be larger than U for a xed bed under the same pressure gradient. Consequently, the use of a packed bed
equation will in most cases underestimate U for a given
e, or overestimate e and L for a given U (Happel and
Epstein, 1951). If, for example, we remove the minimum uidization (subscript mf) restriction from Eqs.
(15)(19), then Eq. (17) becomes
42:862 1 e2
Re
0:5714fe3 Ar
f2
42:861 e
f
!1=2
24
Theoretical Models
25
26
p
U
27
1 0:75c 1 8=c 3
U0
which, though it breaks down for e 1=3 (i.e., c 2=3,
for which U=U0 0), nevertheless gives reasonable
answers in the uidization region, 0:4 < e < 1, and
even agreement with the lower values of U/U0 reported
in the literature for liquid-uidized beds at low Re and
e=0.51 (Verschoor, 1951). To render Eq. (27) comparable to Eqs. (25) and (26), we rewrite it for the case
of c ! 0, for which it collapses to
p
U
1 1:5 2 c1=2 1 2:12c1=2
27a
U0
A more rigorous cell model, inasmuch as it involves a
solution of the NavierStokes equation for creeping
ow rather than a modication of that equation as in
the case of Brinkman, is that of Happel (1958). In this
case the basic cell is that of a single sphere surrounded
by a concentric spherical envelope of uid, the volume
of which bears the same ratio to the volume of the cell
as bed voidage does to unity. The crucial feature of this
model is that the outer surface of the uid envelope
is frictionless (zero shear stress), so that it is often
referred to as the free surface model. The solution is
U
3 4:5c1=3 4:5c5=3 3c2
U0
3 2c5=3
28
28a
(between 0.5 and 1). Improved agreement with empirical results was again obtained and, unlike the solution
of Jean and Fan (1989) or even Eq. (25) above, the
present solution predicts sucient particle clustering
at low particle concentrations that U=U0 exceeds
unity at c 0:03, approaching 1.07 as c approaches
zero. Values of U=U0 exceeding unity to as high as
1.07 have been observed between c 0:001 and
c 0:03 by Kaye and Boardman (1962), who
explained their sedimentation results by the assumption of cluster formation, which at low particle concentrations overrides the small return ow eect
associated with hindered settling. For even smaller
concentrations (c < 0:001), the values of U=U0 measured by Kaye and Boardman, unlike the theoretical
values of Smith, reduce as expected to unity. The
minor defect in Smiths theory is inconsequential to
uidization practice, in which volumetric particle
concentrations are invariably larger than 0.001.
The zero vorticity cell model has been extended
numerically to intermediate Reynolds numbers for
spheres (LeClair and Hamielec, 1968), and both the
zero vorticity and the free-surface cell models (unmodied) to spheroids at low Reynolds numbers (Epstein
and Masliyah, 1972). However, these extensions are
more applicable to immobilized packed beds than to
uidized beds, and only in the absence of turbulence,
which for unexpanded, xed packed beds of spheres
develops at Re > 110150 (Jolls and Hanratty, 1966).
Grbavcic et al. (1991) have developed an intuitive
variational method of predicting U for any given value
of e that is not restricted to any given ow regime. The
method, which has been applied successfully to water
uidization of glass spheres over the range of Re0
180920 (Grbavcic et al., 1991), is succinctly summarized by Jean and Fan (1998), as
"
U Umf
e3 1 ea0mf
e3mf 1 emf a0
#1=2
29
a
1
e emf
1 B2
1 lB
lB
a0mf
1 emf
2 )1=2
B1
29a
2
2
Umf
B1 41
U 2 e3mf
1 B21 1=2
1 B21 1=2 B1
!2 31=2
5
29b
29c
and
1=2
B1
lB 1 B21
29d
Empirical Equations
30
where
0
B2
31
32
U0
Re0
where k Ue =U0 1.
Figure 7 Expansion characteristics of a stable liquid uidized bed. Concentrated suspensions follow bold solid line
of slope n, which may at e ec change to either slope n 0 > n
all the way to e 1 (bold dotted line), or slope n 00 < n to
e < 1 (bold dashed line).
Ar1=3
which applies up to Re0 260,000 (Haider and
Levenspiel, 1989).
The index n for spheres was originally correlated by
ve empirical equations relating this index to Re0 and
the wall eect ratio, dp =Dc (Richardson and Zaki,
1954). However, the wall eect on n has not been conrmed by subsequent investigators (Chong et al., 1979;
Di Felice and Parodi, 1996), and several investigators
have proposed a single equation relating n to either
Re0 (Wallis, 1969; Garside and Al-Dibouni, 1977;
Rowe, 1987) or Ar (Khan and Richardson, 1989),
the advantage of the latter being that Ar, unlike Re0,
34
35
37
39
Umf
emf
41
Umf
U0 enmf
40
42
42a
"
#a
1 dp =Dc
U00
U0
1 0:33dp =Dc
44
44a
A more accurate two-parameter model was subsequently developed by Kehlenbeck and Di Felice
(1999).
The other dilute suspension route illustrated in Fig.
7 involves an initial slope decrease to n 00 < n, which
usually occurs at Re0 35 (Chong et al., 1979; Fan
et al., 1985; Di Felice, 1995). The slope n 00 has been
correlated with Ar by Fan et al. (1985), as has the
corresponding intercept Ue0 < Ue , obtained by linear
extrapolation to e 1. The nal termination of the
actual (as opposed to the extrapolated) data points
for the very dilute suspensions is at e 1, U U00
(Bena et al., 1963). It should be noted that the rather
abrupt changes of slope on loglog coordinates illustrated in Fig. 7 do not occur when the same data
are plotted on arithmetic coordinates, a fact that is
important for the variational method of Grbavcic et
al. (1991) summarized earlier.
Liquid uidized beds of nonspherical, rough or very
ne particles have been shown by many investigators
(Jottrand, 1952; Lewis and Bowerman, 1952;
Richardson and Zaki, 1954; Whitmore, 1957;
Richardson and Meikle, 1961a; Fouda and Capes,
1977; Chong et al., 1979; Cleasby and Fan, 1981) to
expand according to Eq. (32), with k depending not
only on dp =Dc and possibly on Re0 but also on sphericity f (Dhamarajah and Cleasby, 1986). The dependence of k on dp =Dc is again better represented by
Eq. (36) than by Eq. (35) (Dhamarajah and Cleasby,
1986). The value of n in Eq. (32) is always greater than
that given by Eq. (34). This positive deviation increases
with decreasing sphericity, decreasing particle size, and
decreasing Re0. Thus at Re0 < 0.1, n for particles with
dp > 280 mm increased from 4.8 to 5.8 as f decreased
from unity to 0.7 (Chong et al., 1979) and increased
from 6.9 to 9.5 as dp of methacrylate powder decreased
from 194 to 65 mm (Whitmore, 1977); while n decreased
from about 3.55.5 (over a wide sphericity spread) at
Re0 10 (Ar 310) to about 2.42.6 (over a similarly
wide spread in f) at Re0 1000 (Ar 330,000)
(Dhamarajah and Cleasby, 1986). The last observation
indicates that the particle shape eect on n decreases
markedly as Re0 increases. However, no reliable quantitative correlations between n and f, dp, or Re0 have
yet been developed. Chianese et al. (1992) scored a
Ar1=3
Re0
m
Ar2=3
Ar1=6
45
which is recommended for Re0 < 25,000 and
0:5 < 1.
Another approach to nonspherical particles and/or
rough spheres, originating with Steinour (1944) and
adopted by many other sedimentation workers,
assumes that such particles immobilize liquid around
their surface irregularities and thereby behave as
smooth spheres with an eective volumetric concentration Kc, where the hydrodynamic volume factor K is
the volume of the liquid envelope plus solid divided by
the solid volume. The eective voidage then becomes
1 Kc, the eective particle density reff
rp rK 1=K, and the eective particle diameter
deff dp K 1=3 . By this means, Gasparyan and Ikaryan
(1962) attempted to reconcile their hindered settling
results for nonspherical particles with the original
empirical equations of Richardson and Zaki (1954)
for spheres. The same approach was later applied to
liquid uidization by Fouda and Capes (1977), who
determined n and k for their experiments from the
original empirical equations of Richardson and Zaki
(1954) for spheres, U0 from tables provided by
Heywood (1962), and K iteratively by tting the equation
U
k1 Kcn
U0
46
0:603
0:03
1 0:415=f0:483
48
49
FD0 3pmU0 dp U0
50
51
2
2
FD0 0:44pdp rU0 =8 U02
52
53
54
55
57
3
3m2
Eq. (57) is therefore equivalent to
Are4:7 18Re 2:7Re1:687
59
60
61
64
b 4:7 0:65e1:5log Re
=2
65
66
and densities. A common situation, e.g., in crystallization, is that of particles having uniform shape and
density but diering in size. A constant axial pressure
gradient throughout the bed then denotes perfect mixing of the particles, whereas variation of this gradient
signies segregation, the larger the variation the larger
the degree of segregation (Neuzil, 1964; Scarlett and
Blogg, 1967). For multisized solids in a liquid uidized
bed at supercial velocity U, the voidage at height
z above the distributor may be obtained from the
pressure gradient at that level by means of Eq. (8),
rewritten as
dpf =dz
67
e1
grp r
By means of the bed expansion equations developed
above, e.g., Eqs. (32)(34) and (36)(38), we can relate
this value of e for the given rp, r, m, and U to a given
value of dp, namely dp z; and in general,
e f dp
68
69
Therefore the mass mz of particles between the distributor and the height z is given by
m
z
mz
z
dmz rp A 1 e dz
70
71
L
dpf
dz
dz
72
z
dpf =dz dz
mz 0
M L
dpf =dz dz
0
1 f dp dz
73
1 f dp dz
Figure 8 Mixing and segregation patterns for liquid uidization of binary-sized particles of equal density. (From
Gibilaro et al., 1985.)
In mineral processing, in addition to size, the particle density and possibly the shape may also vary.
Using the measured variation of dpf =dz with z,
the procedure of Di Felice et al. (1987), just described,
can also be applied to determine the changing composition with bed level of two spherical particle species
diering in both diameter and density, by means of an
additional equation for the volume-average density,
rp
c1 rp1 c2 rp2
c 1 c2
74
c 1 c2
c1 =dp1 c2 =dp2
75
5.3
79
80
81
82
If the size dierence between the two species of particles is suciently small that kS kL , nS nL , aS0 aL0 ,
and mS mL , then Eq. (82) can be simplied and
rearranged to
3m=mn
d
eS eL L
83
dS
Therefore from Eq. (79),
"
#
dL 3m=mn
g eL
1
dS
84
85
where Di is the axial dispersion coecient (or dispersivity) of species i and Upi the segregation (or classication) velocity of that species through a swarm of
species j. Kennedy and Bretton proposed that the segregation velocity Upi is the dierence between the local
interstitial (or slip) velocity of the uid in the bed
and the interstitial velocity that would be required to
maintain particles exclusively of that size in suspension
at the voidage existing in the local mixed-particle
environment, so that
Upi
U Ui
e
e
ci
U
dci
c
ni 1
Di ln i
ki U0i e
dz Di
ci
ci0
e
88
ci0
DL
1eL
z
dcL
cL z
DL ln
aL bL z dz
cL
1 eL
0
aL z
1e
S
DS
cS
i L; S
89
and they determined the values of aL, bL, and aS, bS,
accordingly. Eq. (89) could then be integrated as
bL z
2
90
dcS
1 eS
DS ln
aS bS z dz
cS
cS z
86
For a monocomponent system or for the fully segregated regions of either a partially (Fig. 8c) or fully (Fig.
8a) segregated bed, Upi 0, i.e., U Ui , assuming i
signies the particles in the segregated region involved.
Substituting for Ui according to Eq. (32), the result is
Di dci
U
ni 1
ci
87
ki U0i e
dz
e
iL;S
z
cL
bS H 2 z 2
a S H z
2
91
92
93
dcL cL !0
U kL U0L enSL
which yields the initial gradient for the downward
marching solution of Eq. (92) as cS vs. cL, and thence
to an integration of Eq. (87) in parallel with Eq. (92)
from cL 0 to cL 1 eL , thus generating curves of
cL vs. z and the corresponding cS vs. z, for any assumed
value of Di. Thus, by this method, curves are generated
without advance knowledge of the voidage or concentration proles. The nal matching of the results to the
experimental concentration proles requires, in addition to the appropriate choice of Di, satisfaction of
the integral conditions stipulating the total volume
uidized of each solid component:
L
Vi A ci dz
0
i L; S
94
dz
xL 1 x L
DL
97
UpL0 z zL
xL
1 xL
DL
98
1
PeL ZZ L
1e
99
1
PeS ZZ S
1e
100
where PeS LUpS0 =DS and Z Z S z zS L. But,
in accord with Di Felice et al. (1987), cL and cS are
linearly related, so that
cS
cL
1
101
1 eS
1 eL
Equation (101), rst stated as such by Van Duijn and
Rietema (1982), is simply the serial model, Eq. (66),
applied locally to a binary. Thus
ePeL ZZL
1
PeL ZZ L
PeL ZZ L
1e
1e
102
xS 1 xL
7:9 1:1dp2 rp rg U Umf 2:140:05
erU
U0
16:4%
103
in which Umf presumably applies to the larger particles, so that both are fully uidized. When Di Ds ,
this equation thus gives recognition to the properties of
both particle species, directly via dp and rp and indirectly via U0 for the smaller particles, and indirectly via
Umf for the larger. When, however, Di DL , only the
properties of the larger particles are manifested in Eq.
(103), in violation of experiments (e.g., Juma and
Richardson, 1983) that indicate that both species
play a role in the determination of Di.
Although the case of multisized particles of xed density is far more common in industrial situations than
that of particles of uniform size having dierent
104
while in the absence of the heavier particles, the suspension density is given by
rBl rpl 1 el rel
105
107
108
into Eq. (32) for each particle species, and making the
assumption that the density dierence between the two
species of particles is small enough that the indices m
and n, and the coecients k and b, are essentially the
same for each of the two particle species, then since U
is given,
rph r1=m enh rpl r1=m enl
109
or
!1=mn
rpl r
eh el
rph r
110
eS
rp constant
p
dL =dS 2
eL
g
dp constant p
rph r=rpl r 2
el
eh
gh
1.0
0.5
0.866
0.433
0.134
0.067
1.0
0.5
0.930
0.465
0.070
0.181
0.930
0.465
0.070
0.035
1.0
0.5
0.930
0.465
0.070
0.181
2
3
!1=mn
rpl r
rpl r
rpl r
5
gh 1
el 4
rph r
rph r
rph r
111
Since mn equals 4.8 both in the Stokes and the Newton
regimes and is in any case always in excess of unity, it
follows that the expression in square brackets is always
positive, so that gh (in contrast to g) always increases
as the voidage decreases. Thus we can explain the
empirical rule, long known in the ore dressing industry
(Gaudin, 1939), that classication by size, i.e., sizing,
is best performed under dilute (free settling) conditions, while classication by density, i.e., sorting,
prevails under concentrated (hindered settling or
teeter bed) conditions. We shall see below how this
rule manifests itself in the bed inversion phenomenon.
Note also in Table 2 [or in Eqs. (84) and (111)] that,
in contrast to segregation by size, which is attenuated
as one moves from the Stokes to the Newton regime,
i.e., as Re0 increases, there is no such eect in the case
of segregation by density.
Axial dispersivities for particle species of dierent
density, whether or not the sizes are the same, can be
determined by methods similar to those for xed density, mixed size binaries, but their experimental determination has received much less attention in the
literature. Equation (103), to the extent that it is
applicable at all, is restricted to binaries of equal
density particles.
5.5
112
The voidage e can be expressed in terms of the supercial liquid velocity U, explicitly by means of the
RichardsonZaki relation, Eq. (32), or by Eq. (64)
with (65), and implicitly (requiring iteration) by
means of Eq. (62) with (63), applied to the given
solidliquid system. A plot of rBlL vs. U from UmL
to U0lL can thus be generated. Similarly, for the
smaller, denser particles (species hS) subjected alone
to uidization by the same liquid at the same temperature and pressure,
rBhS rphS 1 e re
113
Figure 9 The bed inversion phenomenon of a binary as manifested by the bulk densities of the two layers as a function of liquid
velocity. The dotted line ABD merging with the species hS line to the left of A and the species lL line to the right of D represents
the bottom layer. Assuming vIL 0.4, the two thickened solid lines represent the top layer below and above the inversion
velocity, UB, respectively, while B represents conditions at the singular inversion point. rBlL and rBhS are the bulk densities of the
monocomponent species lL and species hS beds, respectively. (rB)max refers to the bulk densities of the bottom layer along the
dotted line ABD, and clL/(clL chS) are the corresponding compositions of the binary solids in this layer. (After Epstein and
Pruden, 1999.)
followed, respectively, by the two additional segregation patterns (b) and (d) of Fig. 10.
To gain further insight into this phenomenon, apply
Eq. (62) with (63) or, alternately, (64) with (65) to various binary mixtures of the two solids species uidized
by the given liquid, assuming after Wen and Yu (1966)
and Gibilaro et al. (1986a) that dp is the Sauter mean
particle diameter given by Eq. (75), evaluating rp as the
volume-average density by Eq. (74), and noting that
e 1 c1 c2 1 clL chS
114
115
Figure 10 Bed inversion phenomenon. Blackened circles represent higher density, smaller spheres (species hS); open circles
represent lower density, larger spheres (species lL). (Based on experimental observations of Moritomi et al., 1982, and Jean and
Fan, 1986, after Di Felice, 1995.)
118
INVERSE FLUIDIZATION
dpf
1 er rp g 1 erp rg
dz
119
p 1
p
dp2
dp3 rp rg 120
rU02
FD0 0:95
4
2
6
whence
117
1 e 1 elL 1 ehH
s
dp rp rg
U0 1:18
r
121
7.1
Monodisperse Solids
Continuing with our assumption of relatively homogeneous particulate uidization, but switching from
semibatch operation, i.e., from an unchanging batch
of bottom- or top-restrained (the latter in inverse uidization) solids to a continuous vertical ow of both
solids and liquid, we consider rst the case of monosized constant density solids. Studies by Mertes and
Rhodes (1955) and by Lapidus and Elgin (1957) have
shown that, irrespective of the ow direction of either
liquid or solids, the relative or slip velocity, Ur,
between the liquid and the solids in the nonaccelerating region of a given liquidsolids system is always
the same unique function of the voidage, so that in
general,
Ur
Up
U
f 0 e
e 1e
122
123
@U
0
@1 ejUp
124
125
and
U kU0 en 1 n1 e
126
Multidisperse Solids
U Up;i
ki U0i eni 1 ;
ci
e
i 1; 2; 3; . . . ; N
127
(Lockett and Al-Habbooby, 1974), but the small difference in density of the equal-sized binary particle
species on which they based this claim, and experiments by later workers, render such a conclusion generally inadmissible. These experiments and some
theory addressed themselves both to the problem of
species interactions and that of density dierences
among species.
Mirza and Richardson (1979), based on batch sedimentation experiments with several sized binaries
down to e 0.5, concluded that the species interaction
eect could be accounted for by the factor e0.4 applied
to the right-hand side of Eq. (127), so that
Uri k0i U0i eni 1 e0:4 k0i U0i eni 0:6
128
At the same time, Masliyah (1979) addressed the density dierence problem. For a monodisperse system,
i.e., a single particle species, both Eq. (123) and Eq.
(127) are equivalent to
Ur kU0 en1
129
130
it follows that
Ur kU0 en2
rp rB
rp r
131
i1
135
with
N
X
davg
N
X
ci di
i1
N
X
ci
ci di
i1
1 e
136
i1
137
0
where U01
is the terminal free settling velocity of the
fastest moving particles in a uid of density equal to
r2;3
c2 rp2 c3 rp3 er
1 c1
138
and
133a
133b
i1
3
davg
eei 1 1
f1 e1=3 1g
di
0 n2 1
Ur2 k2 U02
e
139
0
where U02
is the terminal free settling velocity of the
intermediate-velocity particles in a uid of density
equal to
r3
c3 rp3 er
1 c1 c2
140
while
Ur3 k3 U03 en3 1
141
Thus only the smallest particles will behave as predicted by Eq. (127). This model was found to agree
well with both multisized sedimentation data
(0.5 < e < 0.85) and the binary-sized countercurrent
results of Lockett and Al-Habbooby (1973).
U Up2 Up1
00
k1 U01
1 c1 n1 1
e c2
c1
142
143
LIQUID DISPERSION
148
Axial Dispersion
e
n1
n
149
151
er
r
152
Radial Dispersion
As shown roughly by Hiby (1967), and more rigorously by Masliyah (1989), this phenomenon can be
explained by the incremental increase of voidage in
the solids upow region (channeling) and the corresponding incremental decrease in the downow region,
the result of which is a net decrease of total drag on the
uidized solids. Masliyah demonstrated experimentally
that the same contraction eect can be brought about
by mechanical stirring of a liquid-uidized bed.
An arguably ubiquitous nonhomogeneity, at least
for U=Umf > 1:5 0:3 (Ham et al., 1990), which has
received much theoretical and experimental attention
in the literature, e.g., by Jackson (1963, 1985),
Anderson and Jackson (1968, 1969), and Homsy and
coworkers (e.g., El-Kaissy and Homsy, 1976;
Didwania and Homsy, 1981; Ham et al., 1990), are
the horizontal striations or bands of low particle concentration, dubbed parvoids by Hassett (1961a,b),
which rise wavelike through a liquid uidized bed
and become more pronounced as they rise. The growth
rate of these voidage waves, which may sometimes be
accompanied by short-lived bubblelike formations
(Cairns and Prausnitz, 1960b; Hassett, 1961a), tend
to increase with increasing voidage for a given system,
and to increase with increasing bubble size for a given
voidage. This type of disturbance, which is associated
with liquid uidization of relatively low-density solids,
e.g., water uidization of 1 mm glass beads (Jackson,
1994), is relatively innocuous, i.e., it is minimally disruptive of the operating characteristics of a nondisturbed bed and its eects are incorporated in the
previously written empirical equations, e.g., in the
equations for predicting bed expansion. Jackson
(1994) has pointed out that this type of instability
should not be confused with the disruptively persistent
bubbling and consequent large-scale particle circulation (Jean and Fan, 1998) characteristic of aggregative
uidization, which is a property of most gas-uidized
beds but for liquid uidization is usually restricted to
relatively high density or large particles, e.g., water
uidization of 3 mm lead spheres (Cairns and
Prausnitz, 1960b). For aggregative uidization, most
of the previously written empirical equations, including those for predicting bed expansion, no longer
apply. These two types of instability are not explicitly
distinguished from each other in Table 1 of Di Felice
(1995), in which are recorded the nonhomogeneous
behavior observed in 26 dierent studies of liquid
uidization from the literature.
A straightforward quantitative criterion which purports to distinguish between particulate and aggregative uidization of both gas and liquid uidized beds
153
156
157
where emf < eb1 < eb2 < 1. Thus particulate uidization prevails from emf to the minimum bubbling voidage eb1, aggregative uidization from eb1 to the
maximum bubbling voidage eb2, and particulate uidization again in the very dilute phase regime from
eb2 to a voidage of unity, e.g., for 0.5 mm copper
spheres uidized by water, for which eb1 0.5 and
eb2 0.95. (3) The LHS of Eq. (155) is zero for
e eb1 < emf and e eb2 < 1 and negative between
eb1 and eb2, in which case vigorous bubbling starts as
soon as the bed is uidized and ceases only at eb2, the
maximum bubbling voidage, above which dilute phase
particulate uidization prevails, e.g., for 2 mm copper
spheres uidized by water (Chen et al., 1999), for
which eb2 0.97.
Gibilaro et al. (1990) have subsequently corrected
Eq. (155) for the added or virtual mass eect due to the
uid acceleration that accompanies particle acceleration. The correction involves adding r/2 to both rp
and r wherever they occur in the equation, as well as
multiplying the second term in the equation by a factor
exceeding unity that approaches unity as r/rp
approaches zero. However, since this correction only
becomes important as r/rp approaches unity, i.e.,
under conditions when the LHS of Eq. (155) will in
most cases be positive whether or not the correction
is applied (and will hence indicate particulate uidization, in agreement with experiment), and given the
experimental uncertainty involved in pinpointing
regime transitions, this added complication can be
safely dispensed with, at least for purposes of predicting aggregative vs. particulate uidization.
The predictions of Eq. (155) have reasonably well
matched the relatively few available regime transition
data on liquid uidization (Gibilaro et al., 1986b), as
well as the much more numerous data on gas uidization (Foscolo and Gibilaro, 1984, 1987; Chen et al.,
1999; these references also include some liquid uidization comparisons). However, Batchelor (1988) has
questioned the basis of Eq. (154) and has proposed
other stabilizing mechanisms, while Jackson (1994)
has also criticized Foscolo and Gibilaro for not properly distinguishing between the stability of a uniform
bed against small perturbations and the phenomenon
of aggregative uidization. The alternative stability criterion developed by Batchelor (1988) is similar in form
to Eq. (155), but it contains two quantities that are
more dicult to evaluate. Until the challenging task
of solving this problem by two-dimensional nonlinear
stability analysis is accomplished (Jackson, 1994), Eq.
(155) can serve as a useful semitheoretical or semiempirical guideline.
10
DISTRIBUTOR DESIGN
UdCA Da d 2 CA
kp Sv CAS CA
dz
dz2
dz
160
CAS CA
UAL
which, when integrated between the limits z 0,
CA CA1 and z L, CA CA2 , results in
CAS CA1
161
kp S=UA
ln
CAS CA2
and is equivalent to
mA QCA2 CA1 kp SCA l:m:
hp Sv Ts T
dz
dz2
and the corresponding solute balance,
158
162
11
159
163
Since CA 1:m: > CAS CA2 , and since the real
situation falls between the two, it follows that determination of kp assuming plug ow will tend to underestimate kp while its determination assuming perfect
mixing will tend to overestimate kp. Arters and Fan
(1986) have solved Eq. (159) for nite Da, assuming
appropriate boundary conditions, and for two-phase
particleliquid uidization, kp, based on Eq. (162), is
much closer to the answer by this solution than kp by
Eq. (163). Measured axial solute proles are also much
closer to those for plug ow than those for perfect
164
165
166
167
Actually, Ar Ga
(rp r)/r, and the Toulouse
group that initiated correlation along these lines
always tted their data to
0
rp r b c
Sc
168
Shp K 0 Rea Gab
r
but even for their spread of rp r=r 0:271:14,
the largest dierence between b and b 0 in any such
correlation developed by this group was only 0.023
(Tournie et al., 1979), while the one study that
extended the upper range of (rp r)/r to 7.07 found
that b b0 0 (Nikov and Delmas, 1987). Thus Eq.
(167) takes precedence over Eq. (168) (Epstein, 1992).
The index on Re in Eq. (167) has been generally
found to be zero (Nikov and Karamanev, 1991,
1992) or very close to zero, e.g., +0.004 by Tournie
et al. (1979) for Re 1.61320 and 0.07 by Riba and
Couderc (1980) for Re 15025,000 and e 0.45
0.90. Within the experimental error it is conveniently
taken as zero, which means that the eect of any
increase in liquid supercial velocity for a given bed
is compensated for by a corresponding increase in voidage. We are then left with an equation similar in form
to that recommended by Calderbank (1967) for mass
transfer into liquids from both solid spheres and small
gas bubbles, namely,
Shp K00 Arb Scc
169
170
on the equivolume sphere diameter, dp, Shp was modied by the factor f1.35 over the sphericity range 0.71.
For spheres with a surface area partly active and partly
inert, Panier et al. (1980) found electrochemically that,
for dierent ratios and geometries of the active and
inert parts, Shp (in which kp is based only on the active
area) increased as the fraction fA of active area
0:25
over
decreased, the modication factor being fA
the range fA 0.0570.98. A qualitatively similar
mass transfer intensication eect occurs when inert
particles of higher density and smaller diameter are
mixed with active particles; the higher the fraction of
inert particles the greater the observed increase in Shp
(Yang and Renken, 1998). Finally, when the solute ux
due to mass transfer from or to the surface of the
particles is not completely overshadowed by the mainstream mass ux (or mass velocity), e.g., for dissolution of very soluble solids, Shp by Eq. (170) must be
corrected owing to distortion of the mainstream velocity and concentration proles, rapid change in the
area of the solidliquid interface, and sharp physical
property variations with concentration near the interface. Any analogy, conventional or otherwise, between
heat and mass transfer breaks down in these circumstances. Chhun and Couderc (1980) found that for
dissolution experiments with B 0 surface mass ux/
mainstream mass ux 0.0130.674,
Shp
Shp;low flux
0:56
1 B0
171
12.1
Heat transfer between a liquid uidized bed and a submerged surface in the bed, most commonly the bed
wall itself, has been the subject of many experimental
studies, the conditions, scope, and results of which
have been summarized by Haid et al. (1994) in their
Tables 1 and 2. A typical set of results is shown in Fig.
11, where it is seen that the increase of the heat transfer
coecient, h, with supercial liquid velocity, U, that
occurs in the xed packed condition continues in somewhat attenuated form when the bed is uidized, eventually achieving a maximum and then declining to
Figure 11 Heat transfer coecient between wall and liquid for packed bed, uidized bed, and single-phase ow as a function of
supercial liquid velocity. (After Haid et al., 1994; data points from Kato et al., 1981.)
Nu / e0:79n1:66 1 e0:12
173
174
n1
n 0:5
175
Mass Transfer
Turning to mass transfer between xed submerged surfaces and liquid-uidized beds, it has been noted by
several authors (King and Smith, 1967; Briens et al.,
1993; Schmidt et al., 1999) that the conventional analogy between heat and mass transfer is inapplicable.
Thus, substituting Sc for Pr and Sh km dp =DA for
Nu in Eq. (172) results in overprediction of Sh by
factors varying from about 5 to 2 as measured Sh
rises from 2 to 30 (Schmidt et al., 1999). Schmidt et
al. attribute this result to the presence of particle convection in heat transfer and its absence in mass transfer, but in the light of the discussion immediately
above, factors of 25 cannot be accounted for by this
which is equivalent to
00
Sh a Re
1b00
b00 1
1 e e Sc
1=3
176b
177
178
The correlation index (or coecient of determination) R2 for this equation was 0.98, as compared to R2
0:8 for the same data correlated by Eq. (176a).
Neglecting the rst term on the right-hand side of
Eq. (178) relative to the second, assuming emf 0.4,
and again invoking Eq. (32), we nd that
Sh / 1 ee 0:4en=3
179
Some applications of liquid-uidized beds, such as particle classication, are over a century old, while many,
Figure 14 Schematic diagram of a uidized bed bioreactor. (After Grady Jr et al., 1999.)
ACKNOWLEDGMENTS
Thanks are due to John Grace, Arturo Macchi, Dusko
Posarac, and Yunbi Zhang for help in locating references, and to the Natural Sciences and Engineering
Research Council of Canada for continuing research
support.
NOMENCLATURE
a
a; b; c
ai
a0
a 00
A
Ac
Ap
Ar
bi
b
b 00
B
B1
B2
B
c
ci
s1
constant for particle species i in Eq. (89),
s1
exponent in Eq. (168), dimensionless
exponent in Eq. (176), dimensionless
buoyant force on single particle in swarm
of particles, N
parameter dened by Eq. (29b),
dimensionless
parameter dened by Eq. (29c),
dimensionless
surface mass ux/mainstream mass ux,
dimensionless
volumetric particle concentration of
monodisperse solids, m3
m3
local or overall volumetric concentration
in liquid of particle species i, m3
m3
ci0
cm
cp
cpp
ct
c
c c
c c
c20
CA
CAS
CA1
CA2
CA 1:m:
CD
CDS
CD0
C 1 , C2
d
davg
deff
di
dp
dpi
dp
ds
dsv
Da
DA
Dc
Dh
K1
specic heat capacity of solid particles,
J
m
K1
local or overall volumetric concentration
in liquid of particle species i, m3
m3
saturation concentration in Fig. 12, wt%
bulk liquid undersaturation for crystal
dissolution points in Fig. 12, wt%
bulk liquid supersaturation for crystal
growth points in Fig. 12, wt%
volumetric concentration of smaller
particles in pseudouid through which
larger particles move, m3
m3
bulk concentration of solute A in liquid
solution at given bed level, kg
m3
concentration of solute A in liquid
solution at particle surfaces, kg
m3
value of CA at z 0, kg
m3
value of CA at z L, kg
m3
logarithmic mean of CASCA1 and CAS
CA2, kg
m3
drag coecient of particle in bed uidized
at supercial velocity U, dimensionless
drag coecient of isolated particle at
velocity U, dimensionless
drag coecient of isolated particle at
velocity U0, dimensionless
constants in Eq. (17), dimensionless
diameter of spherical particle, mm or m
average diameter of spherical particles as
given by Eq. (136), mm or m
eective particle diameter dp K 1=3 , mm
or m
spherical particle diameter of species i
particles, mm or m
diameter of sphere having same volume as
particle, mm or m
equivolume sphere diameter of species i
particles, mm or m
local mass average or mass median
particle diameter, mm or m
diameter of sphere having same surface
area as particle, mm or m
diameter of sphere having same surface-tovolume ratio as particle, mm or m
axial dispersion coecient of liquid,
cm
s1 or m
s1
molecular diusivity of solute A in liquid,
cm2
s1 or m2
s1
column diameter, m
hydraulic diameter 4 cross-sectional
area/wetted perimeter, m
Di
Dia
Dir
DL
Dr
DS
ed
f
f dp
f e
f 0 e
FD
FDS
FD0
g
Ga
h
hmax
hp
jm
k
ki
km
K1
distance between pressure taps in Eqs. (9)
and (10), or thickness of transition zone in
Eqs. (89)(91), m
ChiltonColburn mass-transfer
factor km =USc2=3 , dimensionless
wall-eect factor Ue/U0, dimensionless
wall-eect factor for particle species i
alone, dimensionless
mass transfer coecient between
submerged surface and liquid in uidized
bed, m
s1
kp
K0
K00
L
Lmf
mA
mz
M
Mi
n
ni
np
n0
n 00
N
Np
Nu
Nup
p
pd
pf
dpf =dz
P1
P2
P
P=z
Pea
Pei
PeL
Per
PeS
Pr
Q
R2
Re
Remf
Re0
S
Sv
Sc
Sh
Shp
t
T
Ts
U
Uamf
U A ; UB ;
UC ; UD
Ubf
Ue
Uei
Ue0
UE
Ui
Umf
Umfd
Umff
Umpd
Umpf
Up
UphS
Upi
Up;i
Upi0
UpL0
UpS0
Ur
Uri
Utf
U0
U00
U0i
0
U01
0
U02
00
U01
Ue
vi
Vi
Vp
W
xL
xS
z
zL
zS
Z
Z L
Z S
Greek Letters
a
0
a 0 =amf
b
g
gh
eb
eb1
eb2
ec
eei
ei
emax
emf
k
l
lB
leddy
lp
m
m2
r
rB
rBA, rBB,
rBD
K1
parameter dened by Eq. (29d),
dimensionless
eddy conductivity of heat, W
m1
K1
thermal conductivity of particles,
W
m1
K1
liquid viscosity, Pa
s
pseudouid viscosity for determining
motion of larger particles through
suspension of smaller particles, Pa
s
liquid density, kg
m3
bulk density of liquid-uidized bed,
kg
m3
bulk densities, (rB)max, of bottom layer at
points A, B, and D, respectively, in Fig. 8,
kg
m3
(rB)max
re
rp
r p
rpi
r2
r2,3
r3
P
f
fA
Subscripts
h
hS
j
l
lL
mf
S
1; 2; 3
REFERENCES
Adler IL, Happel J. The uidization of smooth spheres in
liquid media. Chem Eng Prog Symp Ser 58(38):98105,
1962.
Agarwal GP, Hudson JL, Jackson R. Fluid mechanical
description of uidized beds. Experimental investigation
of convective instabilities in bounded beds. Ind Eng Chem
Fundam 19:5966, 1980.
Chilton TH, Colburn AP. Mass transfer (absorption) coecientsprediction from data on heat transfer and uid
friction. Ind Eng Chem 26: 11831187, 1934.
Chitester DC, Kornosky RM, Fan L-S, Danko JP.
Characteristics of uidization at high pressure. Chem
Eng Sci 39:253261, 1984.
Chong YS, Ratkowsky DA, Epstein N. Eect of particle
shape on hindered settling in creeping ow. Powder
Technology 23:5566, 1979.
Chung SF, Wen CY. Longitudinal dispersion of liquid owing through xed and uidized beds. AIChE J 14:857866,
1968.
Chyang CS, Huang WC. Characteristics of large particle
uidization. J Chin Inst Chem Engrs 19(2):8189, 1988.
Cleasby JL, Baumann AR. Backwash of granular lters used
in waste water ltration. Final Report on EPA Project R
802140, Engineering Research Institute, Iowa State
University, 1976.
Cleasby JL, Fan KS. Predicting uidization and expansion of
lter media. J Environ Engng Div ASCE 107:455471,
1981.
Cleasby JL, Lorence JL. Eectiveness of backwashing for
wastewater lters. J Environ Engng Div ASCE
104:749765, 1978.
Cleasby JL, Sejkora GD. Eect of media intermixing on dual
media ltration. J Environ Engng Div ASCE
101:503516, 1975.
Cleasby JL, Woods CF. Intermixing of dual media and
multimedia granular lters. J Am Water Works Assoc
67:197203, April 1975.
Cleasby JL, Stangl EW, Rice GA. Developments in backwashing of granular lters. J Environ Engng Div ASCE
101:713727, 1975.
Clift R. An Occamist review of uidized bed modelling.
AIChE Symp Series 89(296):117, 1993.
Clift R, Grace JR, Weber ME. Bubbles, Drops, and
Particles. New York: Academic Press, 1978, p. 114.
Clift R, Seville JPK, Moore SC, Chavarie C. Comments on
buoyancy in uidized beds. Chem Eng Sci 42:191194,
1987.
Cole LT, Allen CA. Liquid-uidized-bed heat exchanger ow
distribution models. Proc Intersoc Energy Convers Eng
Conf 13:11291134, 1978.
Cooper PF, Atkinson B, eds. Biological Fluidised Bed
Treatment of Water and Wastewater. Chichester: Ellis
Horwood, 1981.
Cooper PF, Wheeldon DHV, Ingram-Tedd PE, Harrington
DW. Sand/biomass separation with production of a
concentrated sludge. In: Cooper PF, Atkinson B, eds.
Biological Fluidised Bed Treatment of Water and
Wastewater. Chichester: Ellis Horwood, 1981, pp 361367.
Cornish ARH. Note on minimum possible rate of heat
transfer from a sphere when other spheres are adjacent
to it. Trans Instn Chem Engrs 43:T332-T333, 1965.
Couderc J-P. Incipient uidization and particulate systems.
In: Davidson JF, Clift R, Harrison D, eds. Fluidization.
2d ed. London: Academic Press, 1985, pp 146.
Richardson JF. Incipient uidization and particulate systems. In: Davidson JF, Harrison D, eds. Fluidization.
London: Academic Press, 1971, pp 2564.
Richardson JF, Aatin E. Fluidisation and sedimentation of
binary mixtures of particles in a liquid. Proc 9th
International
Conference
on
Transport
and
Sedimentation of Solid Particles. Cracow, Poland:
Zeszyty Naukowe Akademii Rolniczej we Wroclawiu Nr
315, 1997, pp 487502.
Richardson JF, Meikle RA. Sedimentation and uidisation
Part III: The sedimentation of uniform ne particles and
two-component mixtures of solids. Trans Instn Chem
Engrs 39:348356, 1961a.
Richardson JF, Meikle RA. Sedimentation and uidisation
Part IV: Drag force on individual particles in an assemblage. Trans Instn Chem Engrs 39:357362, 1961b.
Richardson JF, Shabi FA. The determination of concentration distribution in a sedimenting suspension using radioactive solids. Trans Instn Chem Engrs 38:3342, 1960.
Richardson JF, Zaki WN. Sedimentation and uidisation:
Part I. Trans Instn Chem Engrs 32:3553, 1954.
Richardson JF, Romani MN, Shakiri KJ. Heat transfer from
immersed surfaces in liquid uidized beds. Chem Eng Sci
31:619624, 1976.
Rigby GR, Van Blockland GP, Park WH, Capes CE.
Properties of bubbles in three-phase uidized beds as measured by an electroresistivity probe. Chem Eng Sci
25:17291741, 1970.
Rowe PN. Drag forces in a hydraulic model of a uidised
bedPart II. Trans Instn Chem Engrs 39:175180, 1961.
Rowe PN. A convenient empirical equation for estimation of
the RichardsonZaki exponent. Chem Eng Sci 42:2795
2796, 1987.
Rowe PN, Claxton KT. Heat and mass transfer from a single
sphere to uid owing through an array. Trans Instn
Chem Engrs 43:T321T331, 1965.
Rowe PN, Henwood GA. Drag forces in a hydraulic model
of a uidised bedPart I. Trans Instn Chem Engrs 39:43
54, 1961.
Saman PG. On the settling speed of free and xed susensions. Studies in Appl. Math L11(2):115127, June
1973.
Salas-Morales JC, Evans JW, Newman OMG, Adcock PA.
Spouted bed electrowinning of zinc: Part 1. Laboratoryscale electrowinning experiments. Metall Mat Trans B
28:5968, 1997.
Saxena SC, Vogel GJ. The measurement of incipient uidisation velocities in a bed of coarse dolomite at temperature
and pressure. Trans IChemE 55:184189, 1977.
Saxton JA, Fitton JB, Vermeulen T. Cell model theory of
homogeneous uidization: density and viscosity behaviour. AIChE J 16:120130, 1970.
Scarlett B, Blogg MJ. The motion and distribution of
particles in a liquid-uidized bed. In: Drinkenburg
AAH, ed. Proc Internat Conf on Fluidization.
Amsterdam: Netherlands Univ. Press, 1967, pp 8289.
Schiller L, Naumann A. Uber die grundlegenden berechnungen bei der schwerkragtaufbereitung. Z Ver Deutsch Ing
77:318320, 1933.
Schmidt S, Buchs J, Born C, Biselli M. A new correlation for
the wall-to-uid mass transfer in liquidsolid uidized
beds. Chem Eng Sci 54: 829839, 1999.
Scott CD, Hancher CW, Shumate II SE. A tapered uidized
bed as a bioreactor. Enzyme Eng 3:255261, 1978.
Scott TC, Cosgrove JM, Asif M, Petersen JN.
Hydrodynamic studies of an advanced uidized-bed bioreactor for direct interaction with coal. Fuel
72:17011704, 1993.
Selim MS, Kothari AC, Turian RM. Sedimentation of binary
suspensions. AIChE Symp Series (Fluidization and Fluid
Particle Systems: Theories and Applications, Knowlton
TM, ed.) 79(222):103108, 1983a.
Selim MS, Kothari AC, Turian RM. Sedimentation of multisized particles in concentrated suspensions. AIChE J
29:10291038, 1983b.
Selke WA, Bliss H. Continuous countercurrent ion exchange.
Chem Eng Prog 47:529533, 1951.
Shiau L-D, Cheung S-H, Liu Y-C. Modelling of a uidizedbed crystallizer in a batch mode. Chem Eng Sci 54: 865
871, 1999.
Sibony MJ. Clarication with microsand seeding. A state of
the art. Water Res 15:12811290, 1981.
Slater MJ. A review of continuous counter-current contactors for liquids and particulate solids. Brit Chem Eng
14:4146, 1969.
Slater MJ. Recent industrial-scale applications of continuous
resin ion exchange systems. J Separ Proc Tech 2(3):212,
1981.
Slater MJ. The relative sizes of xed bed and continuous
counter-current ow ion exchange equipment. Trans
IChemE 60:5458, 1982.
Slater MJ, Lucas BH. Flow patterns and mass transfer rates
in uidized bed ion exchange equipment. Can J Chem Eng
54:264268, 1976.
Slis PL, Willemse Th W, Kramers H. The response of the
level of a liquid uidized bed to a sudden change in the
uidizing velocity. Appl Sci Res A8:209218, 1959.
Smith TN. A model of settling velocity. Chem Eng Sci
53:315323, 1998.
Sozen ZZ, Grace JR, Pinder KL. Thermal energy storage by
agitated capsules of phase change material. Ind Eng Chem
Res 27:679691, 1988.
Steinour HH. Rate of sedimentation: suspensions of uniform-size angular particles. Ind Eng Chem 36: 840847,
1944.
Storck A, Coeuret F. Mass transfer between a owing liquid
and a wall or an immersed surface in xed and uidized
beds. Chem Eng J 20:149156, 1980.
Streat M. Recent developments in continuous ion exchange.
J Separ Proc Technol 1(3):1018, 1980.
Struve DL, Lapidus L, Elgin JC. The mechanics of
moving vertical uidized systems III. Application to
27
GasLiquidSolid Three-Phase Fluidization
Liang-Shih Fan and Guoqiang Yang
The Ohio State University, Columbus, Ohio, U.S.A.
INTRODUCTION
under high pressure and high temperature, for example, methanol synthesis (at P 5.5 MPa and T 2
608C), resid hydrotreating (at P 5.521 MPa and
T 3004258C), FischerTropsch synthesis (at
P 1.55.0 MPa and T 2508C), and benzene hydrogenation (at P 5.0 MPa and T 1808C) (Fox, 1990;
Jager and Espinoza, 1995; Saxena, 1995; Mill et al.,
1996; Peng et al., 1999). Fundamental study of ow
characteristics under high-pressure and high-temperature conditions is crucial for the development of these
industrial processes, and much progress has been made
recently regarding high-pressure and high-temperature
systems with relevance to industrial processes.
The fundamental characteristics of three-phase
uidization including bubble characteristics, hydrodynamics, and heat and mass transfer properties along
with many industrial processes have been extensively
reported in Fans (1989) book as well as its companion
book on bubble wake dynamics (Fan and Tsuchiya,
1990). As both books are widely referenced in the
eld of three-phase uidization, this chapter is presented mainly as an update to these two books. The
chapter will cover the continued research progress
made over the past ten years on the fundamentals of
three-phase uidization. Major ndings on uidization
and bubble dynamics under ambient conditions and
the relevant literature reported earlier will be covered.
Furthermore, new research on the high-pressure and
high-temperature three-phase uidization will be highlighted as well as computational uid dynamics.
BUBBLE DYNAMICS
The expansion stage and the detachment stage follow the same force balance equation, although the
expression for the same force in the two stages may
be dierent owing to dierent bubble moving velocities
in the two stages. The expressions for all the forces
under two stages are listed in Table 1. The particle
bubble collision force is merely the rate of momentum
change of the particles colliding with the bubble surface. The suspension inertial force is derived from the
suspension ow eld around an accelerating bubble,
obtained from a particle image velocimetry (PIV) sys-
Expansion stage
p 3
d r rg g
6 b l
p 2
D r u2
4 o g o
db2
r u2
l b
Detachment stage
Same as expansion stage
Same as expansion stage
CD
24
Re
FD
CD
Fs
pDo s cos g
FI;g
d h
p 3 i
r
d u
dt g 6 b b
FBA
Not applicable
FC
p 2
D 1 ees rs u2e
4 o
d
rm um dV
FI;m
dt
Source: Luo et al. (1998a).
3 2 p
d prl ml
2 b
t
0
du=dt
p dt
tt
p 2
d e r u2
4 b s s
z
d h
p 3 i
r
d u
dt m 6 b b
dt
Vc
"
2 #
rb d 2 rb
drb
Pb P0 rm
3 dt2
dt
2s 4m1 1
rg
rb
4
rb
Qo
1
2
4 pDo
!2
Pb P0 rm
3 dt2
dt dt
2
2s 4ml drb 1
Qo
r
rb dt 4 g 1=4pD2o
rb
5a
5b
Figure 4 Simulated (a) bubble formation time and (b) instantaneous gas ow rate through the orice during the bubble
formation process in the liquidsolid suspension (Do=1.63
mm, Vc=650 cm3, es=0.18). (From Yang et al., 2000a.)
2.2.1
The FanTsuchiya equation, generalized for highpressure systems, can be written in a dimensionless
form:
#n
1=4 ("
rl
Mo1=4 r 5=4 02
0
db
ub ub
rl
sg
Kb
0 n=2 )1=n
2c
r db
6
db0
rl 2
where the dimensionless bubble diameter is given by
db0 db
r g1=2
l
s
n
c
0:8
1:6
8a
1:2
1:4
8b
Kb maxKb0 Mo0:038 ; 12
where
Kb0
14:7
10:2
8c
4 grdb
3 rl u2b
24
72
1 0:15 Re0:687 ;
;
Re
Re
10b
10c
(Fig. 6a), suggesting that the viscous force predominates in the bubble rise process. On the other hand, at
the high temperature (Fig. 6b), there is a strong agreement over the bubble size range of db > 2 mm including
the sharp breakpoint/peak. This indicates that the
liquid used tends to behave as a pure inviscid liquid.
Note that over the pressure range from 0.1 to 19.4 MPa,
the liquid viscosity varies from 29 to 48 mPa
s at 278C,
whereas it is almost constant within a range from 4.7 to
5.2 mPa
s at 788C (Lin et al., 1998).
The FanTsuchiya equation applied for the given
liquid, a pure (n 1:6), multicomponent (c 1:4),
and organic solvent (Kb0 10:2), demonstrates good
overall predictive capability except for the sharp peak
existing under the high temperature (Fig. 6b). The
equation by Tomiyama et al. (1995) also has good
general applicability, especially around the peak
point occurring near db 2 mm at 788C; however, it
tends to underestimate the bubble rise velocity over the
rest of the bubble size range.
The consistent dierence in ub prevailing between
0.1 and 19.4 MPa for db > 2 mm is due to the signicant
increase in gas density (as large as a 200-fold increase
with pressure from 0.1 to 19.4 MPa). The density eect
is accounted
for in FanTsuchiya equation in terms of
rrl or in Tomiyamas equation in terms of both
r rl and Eo. As can be seen from the equations
and gure, the density dierence between the continuous liquid phase and the dispersed gas phase plays an
important role in determining bubble rise velocity,
especially for large bubbles.
The ReEo relationship is often utilized in representing the general rise characteristics of single bubbles
in liquids (Clift et al., 1978; Bhaga and Weber, 1981).
Figure 7 shows the ReEo relationship at dierent Mo
values. The thin background lines signify the general,
quantitative trend for the rise velocity of single bubbles
in puried Newtonian liquids under ambient conditions, plotted with constant intervals of log Mo. The
gure shows the general agreement in correlation predictions. The experimental results under four conditions are also plotted in the gure (Lin et al., 1998).
By employing accurate values for the physical properties of the liquid phase and the gas density at given
pressures and temperatures, the experimental results
can be successfully represented over the entire Eo
range by the FanTsuchiya equation. This equation
was also proven to represent single bubble rise velocity
in various liquids under ambient conditions (Tsuchiya
et al., 1997). So it is clear that much more is known
now than before about single bubbles rising in liquids
at high pressure and high temperature.
2.2.2
Figure 8 Eect of pressure on bubble rise velocity in a uidized bed of 0.88 mm glass beads at (a) 26.58C and (b) 87.58C
(solids holdups for , open, and lled symbols are 0, 0.384,
and 0.545, respectively). (From Luo et al., 1997b.)
and
esc 1:3 0:1 tanh0:510 102 ut es0
Figure 9 Eect of pressure on bubble rise velocity in a uidized bed of 0.21 mm glass beads at (a) 26.58C and (b) 87.58C
(solids holdups for +, open, and lled symbols are 0, 0.381,
and 0.555, respectively). (From Luo et al., 1997b.)
11
13a
13b
1977, 1978). There are three steps in the bubble coalescence process (Vrij, 1966; Chaudhari and Homann,
1994): (1) approach of two bubbles to form a thin
liquid lm between them; (2) thinning of the lm by
the drainage of the liquid under the inuence of gravity
and suction due to capillary forces; and (3) rupture of
the lm at a critical thickness. The second step is the
rate-controlling step in the coalescence process, and
the bubble coalescence rate can be approximated by
the lm thinning rate (Vrij, 1966). The lm thinning
velocity can be expressed as (Sagert and Quinn, 1977,
1978)
dl
32l 3 s
dt 3fR2d ml db
14
2.4
Bubble Breakup
Another quantiable area of bubble dynamics is bubble breakup. Bubble breakup is a key phenomenon to
study because it determines the bubble size in the system and hence aects heat and mass transfer between
phases. Bubble breakup has been investigated theoretically and experimentally. There are two causes for
bubble breakup: bubbleparticle collision and bubble
instability. In this section, only bubble breakup due to
bubble instability will be discussed.
It is known that the variation of bubble size with
pressure is the key for understanding pressure eects
on hydrodynamics. The upper limit of the bubble size
is set by the maximum stable bubble size, Dmax , above
which the bubble is subjected to breakup and hence is
unstable. Several mechanisms have been proposed for
the bubble breakup phenomenon, and based on these
mechanisms theories have been established to predict
the maximum bubble size in gasliquid systems. In
this section, the mechanisms of bubble breakup and
the theories to predict the maximum bubble size are
covered.
Hinze (1955) proposed that bubble breakup is
caused by the dynamic pressure and the shear stresses
on the bubble surface induced by dierent liquid ow
patterns, e.g., shear ow and turbulence. When the
maximum hydrodynamic force in the liquid is larger
than the surface tension force, the bubble disintegrates
into smaller bubbles. This mechanism can be quantied by the liquid Weber number. When the Weber
number is larger than a critical value, the bubble is
not stable and disintegrates. This theory was adopted
to predict the breakup of bubbles in gasliquid systems
(Walter and Blanch, 1986). Calculations by Lin et al.
(1998) showed that the theory underpredicts the
maximum bubble size and cannot predict the eect
of pressure on the maximum bubble size.
A maximum stable bubble size exists for bubbles
rising freely in a stagnant liquid without external stresses, e.g., rapid acceleration, shear stress, and/or turbulence uctuations (Grace et al., 1978). RayleighTaylor
instability has been regarded as the mechanism for
bubble breakup under such conditions. A horizontal
interface between two stationary uids is unstable to
disturbances with wavelengths exceeding a critical
value if the upper uid has a higher density than the
lower one (Bellman and Pennington, 1954):
r
s
lc 2p
grl rg
15
16
the bubble surface. This force can suppress the disturbances at the gasliquid interface and thereby stabilize
the interface. Centrifugal force explains the underestimation of Dmax by the model of Grace et al. (1978). On
the other hand, the centrifugal force can also break
apart the bubble, as it increases with an increase in
bubble size. The bubble breaks up when the centrifugal
force exceeds the surface tension force, especially at
high pressures when gas density is high. Levich
(1962) assumed the centrifugal force to be equal to
the dynamic pressure induced by the gas moving at
the bubble rise velocity, that is, kf rg u2b =2kf 0:5,
and proposed a simple equation to calculate the
maximum stable bubble size:
Dmax
3:63s
q
u2b 3 r2l rg
17
9prg u2b a2
p
64 2a
18
q
drc 2 dz2
Fs sL s
ellipse
19
p
4saE 1 a2
Figure 10 Schematic of the internal circulation model for bubble breakup: (a) internal and external ow elds; (b) circulation
velocity on surface S; (c) force balance and 3-D view of surface S and the ow pattern on S. (From Luo et al., 1999.)
ap
3
8a
20
r
s
Dmax 2:53
grg
for a 0:21
23a
for a 0:3
23b
in liquids, and
r
s
Dmax 3:27
grg
1:6sg
rg
!1=4
24
Figure 11 Comparison of maximum stable bubble size obtained experimentally and the predictions by various models. (From
Fan et al., 1999.)
TRANSPORT PHENOMENA
Hydrodynamics
Minimum Fluidization
The rst topic of study for the hydrodynamics of threephase uidization is minimum uidization. For a given
gas velocity, the minimum liquid ow rate required to
uidize a bed of particles (Ulmf) may generally be determined from the change in the bed dynamic pressure
drop behavior that occurs as the bed changes from a
xed bed to a uidized bed. There are considerable
variations on minimum uidization phenomena
among small/light, large/heavy, and mixed particle systems. However, Ulmf in general can be evaluated
mechanistically by considering an intrinsic condition
for minimum uidization where the total pressure
drop over a bed of particles at the xed state is equal
to the total bed weight per unit bed cross-sectional area
as formulated by Song et al. (1989). In this evaluation,
the pressure drop in the xed bed can be described by a
ow model developed by Chern et al. (1983, 1984).
The minimum uidization velocity can also be eectively described based on the gas-perturbed liquid
model (Zhang et al., 1998a). The model assumes that
the solid particles in the bed are fully supported by the
liquid and there is little direct interaction between gas
and solid phases. In this model, the gas phase plays a
simple role in the bed by occupying space with particles
uidized by the liquid. The minimum uidization velocity along with minimum liquid velocity for particle
transport for three-phase uidized beds of relatively
large particles (dp > 1 mm) (Zhang et al., 1998b) can
n
U
xkU g eg
l
c
k
n1
el
n1
Ug Ul
k
1 kU l
el
n 1 eg
U
n
xk U l l
<0
25
el
n1
Here, n is the RichardsonZaki index, k is the ratio of
wake size to bubble size, and x is the ratio of solids
concentration in the wake region to that in the liquid
solid uidized region.
3.1.2.2 Moving Packed Bed Phenomenon. Moving packed bed ow is characterized by the motion
of solids in piston ow in a three-phase uidized bed.
Moving packed bed ow, which usually occurs during start-up, depends not only on the gas and liquid
velocities but also on how they are introduced into
the bed. Moving packed bed ow is caused by the
surface phenomena involving ne bubbles attached
onto particles and the subsequent formation of a ne
bubble blanket under the packed solids; a liquid ow
would move the entire bed upward. This phenomenon is thus associated with the small bubble system.
The moving packed bed ow phenomenon in a threephase uidized bed is a known, anomalous event in
the resid hydrotreating industry. It was observed in
the 1960s in the bench and pilot units during the
development and commercialization of the resid
hydrotreating process (Fan, 1999). The reactor was
typically operated at pressures between 5.5 and
21 MPa and temperatures between 300 and 4258C. In
the early 1970s, moving packed bed ow was
observed in a commercial three-phase uidized bed
reactor. The occurrence of a moving packed bed in a
three-phase uidized bed could simply be circumvented by utilizing a start-up procedure that involves
degassing the bed rst and then introducing liquid
ow to expand the bed prior to commencing gas
ow. Commercial operators of three-phase uidized
bed reactors have long recognized and undertaken a
proper start-up procedure of this nature since observing this anomalous event. As the small bubbles can
also be generated under ambient conditions using
surfactants in an airwater system, the moving
packed bed ow phenomenon was reported in open
literature rst by Saberian-Broudjenni et al. (1984)
l s
usmall
!0:03
3 !0:273
s s rl
rl
2:25
rg
ml
gm4l
eg;tran
26
usmall
27
Figure 14 Visualization of bubbles emerging from the surface of a three-phase uidized bed at (a) P 0.1 MPa; (b)
P 3.5 MPa; (c) P 6.8 MPa; (d) P 17.4 MPa. (From
Fan et al., 1999.)
4:1
1 eg
cosh Mo0:054
m
28
29a
0:011
b 0:096 Mom
29b
For high-pressure bubble columns and slurry bubble columns operated under a wide range of conditions, hydrodynamic similarity requires the following
dimensionless groups to be the same: Ug =umax , Mom ,
and rg =rm (Luo et al., 1999). To simulate the hydrodynamics of industrial reactors, cold models could be
used and milder pressure and temperature conditions
could be chosen, as long as the three groups are similar
to those in the industrial reactor. Safoniuk et al. (1999)
identied eight key variables that aect hydrodynamics
of three-phase uidization systems, yielding ve
independent dimensionless groups according to the
Buckingham PI theorem,
grm4l
r2l s3
rl d p U l
Re0
ml
Mo
Figure 15 Eects of (a) pressure and (b) solids concentration on the gas holdup in a slurry bubble column (nitrogenParatherm NF heat transfer uid100 mm glass beads).
(From Luo et al., 1999.)
grd 2p
s
rp
Ug
bd
bU
rl
Ul
Eo0
32
33
34
In the last several sections, hydrodynamic characteristics of three-phase uidization systems were covered.
Starting from this section, several key transport phenomena will be discussed. This section examines heat
transfer characteristics in three-phase uidization systems; mass transfer will be discussed in a later section.
Comprehensive reviews on the heat transfer behavior in various gasliquid and gasliquidsolid systems
under ambient conditions are given by Kim and
Laurent (1991), and more recently by Kumar et al.
(1992, 1993a,b) and Kumar and Fan (1994). There
are two types of heat transfer coecients, that is,
heat transfer coecients between the column wall
and the bed (wall-to-bed), and heat transfer coecients
Figure 17 Variation of instantaneous heat transfer rate with bubble passage synchronized with the visualization of ow patterns
in the vicinity of the heat transfer probe in a liquidsolid uidized bed of low-density gel beads. (From Kumar et al., 1993a.)
Pressure Effect
h h eg
36
upt;0
Ug0:45
where h 0 is the heat transfer coecient of a liquidsolid
uidized bed with the same solids holdup, and upt;0 is
the particle terminal velocity in the uidizing liquid at
ambient pressure. The units for Ug and upt,0 in the
equation are in m/s. The heat transfer coecient, h 0 ,
can be calculated by the correlation (Richardson et al.,
1976):
Nu 0 0:67 Re0:62 Pr0:33
e0:38
s
1 es
37
Parameter
ml
rl
kl
Cpl
s
db
eg
Eect of the
parameter on heat
transfer coecient
Eect of pressure
increase on
parametric value
Parametric eect
on heat transfer
coecient with
increase in pressure
No direct eect
(small)
(small)
(small)
(small)
(small)
(small)
No direct eect
: Increase; decrease.
Source: Luo et al. (1997a).
transfer coecient decreases appreciably with increasing pressure. The variation in heat transfer coecient
with pressure is attributed to the counteracting eects
of the increased liquid viscosity, decreased bubble size,
and increased gas holdup or frequency of bubble passage over the heating surface as the pressure increases.
In slurry bubble columns, it is observed that the bubble
size reduces signicantly with an increase in pressure,
especially under low pressures, which would result in a
decrease in the heat transfer coecient. Therefore the
bubble size is the most important factor aecting the
heat transfer rate in slurry bubble columns. The addition of ne particles to the liquid phase enhances heat
transfer substantially, and the eect of temperature on
the heat transfer behavior is mainly determined by the
change in liquid viscosity.
It is noted that the pressure eects on the heat
transfer coecient are dierent between large-particle
and small-particle systems (Luo et al., 1997a; Yang
et al., 2000b). Similar observations were also found
for hydrodynamics and bubble characteristics
between large-particle and small-particle systems.
Luewisutthichat et al. (1997) photographically studied bubble characteristics in multiphase ow systems.
They found that large-particle systems (i.e., threephase uidized beds) exhibit appreciably dierent
Figure 19 Eect of pressure and gas velocity on heat transfer coecients in a slurry bubble column (nitrogenParatherm NF heat transfer uid53 mm glass beads).
(From Yang et al., 2000b.)
The consecutive lm and surface renewal model originally developed by Wasan and Ahluwalia (1969) can be
used to analyze the heat transfer behavior in slurry
bubble columns (Yang et al., 2000b). For ne particles,
the liquidsolid suspension can be reasonably treated
as a pseudohomogeneous uid phase. The model
assumes that a thin uid lm (liquid or liquidsolid
mixture) with a thickness d exists surrounding the heat-
6:14L
Re3=4
m
1=3
Prm
41
L
ub
42
Interfacial Area
43
area. The study in a large bubble column with a diameter of 15.6 cm also conrmed this conclusion
(Stegeman et al., 1996). The positive inuence of the
operating pressure on the interfacial area is clearly
attributed to smaller bubble size and higher gas holdup
at high pressures. The increase of the interfacial area
with pressure is more pronounced at higher gas
velocities.
On the basis of Eq. (43) and the observation that the
Sauter mean bubble diameter is proportional to rg0:11 ,
Wilkinson et al. (1992) proposed a procedure to estimate the interfacial area in a bubble column under
high pressure as shown in Eq. (44):
ahigh pressure eg high pressure
aatmospheric
eg atmospheric
"
#0:11
rg high pressure
rg atmospheric
44
eg high pressure n
kl ahigh pressure
eg atmospheric
kl aatmospheric
with
n 1:01:2
45
Phase Mixing
46
under ambient conditions. Studies of gasliquid dispersion behavior in coal liquefaction reactors using a
neutron absorption tracer technique indicated a similar
trend of the pressure eect (Onozaki et al., 2000a,b).
The axial dispersion coecients of the liquid phase
under coal liquefaction conditions (P 16.8
18.8 MPa) are smaller than those estimated from literature correlations. These observations imply a decreasing trend for the pressure eect on liquid mixing.
It has been shown that the steady-state thermal dispersion method is suitable for measuring the axial dispersion coecients of the liquid phase at high
pressures in bubble columns, particularly for hydrocarbon liquids in which the electrical conductivity technique is not applicable (Yang and Fan, 2002). In this
technique, heat is introduced close to the outlet of the
liquid phase and the upstream temperature prole in
the liquid is measured. The dispersion coecient is
then determined from the axial temperature prole
based on the one-dimensional dispersion model. The
eects of pressure and gas velocity on the axial dispersion coecients in the 5.08 cm and 10.16 cm columns
for the nitrogen-Paratherm heat transfer uid system
are shown in Fig. 22 (Yang and Fan, 2002). The axial
dispersion coecient decreases signicantly when the
pressure is increased from ambient to elevated pressures for both columns, indicating distinct ow behaviors for ambient pressure and elevated pressure.
When the pressure is further increased, the decrease
rate of the axial dispersion coecient becomes smaller.
The pressure eect is more pronounced at higher gas
velocities and in larger columns. A decrease in the
liquid mixing at high pressures was also observed by
Tarmy et al. (1984) and Onozaki et al. (2000a,b). Other
researchers found that an insignicant eect of pressure on liquid mixing in small columns (diameter normally less than 10.0 cm) is possibly due to the narrow
operating conditions in their studies, that is, either low
gas velocities or a narrow pressure range (Houzelot et
al., 1985; Holcombe et al., 1983). As shown in Fig. 22a,
in small columns, the variation of the axial dispersion
coecient with pressure is not pronounced at low gas
velocities (i.e., in the homogeneous bubbling ow
regime) and low pressures.
The eects of gas velocity and pressure on liquid
mixing can be explained based on the mixing mechanisms of gross liquid circulation and local turbulent uctuations. The available theories in the literature
describing liquid mixing under atmospheric pressure
are based on liquid turbulence induced by rising bubbles
(Baird and Rice, 1975), large-scale liquid internal circulation (Joshi, 1980), or a combination of these two
creasing pressure. The counteracting eects of gas holdup and bubble rise velocity on the liquid circulation
velocity may result in unchanged or slightly changed
liquid circulation velocity at elevated pressures, as
shown in Eq. (47). On the other hand, when the system
pressure increases, bubbles become smaller, and the
liquid entrainment by the bubble wake and the turbulence induced by the motion of bubbles are reduced.
Therefore liquid mixing is reduced at elevated pressures. The combination of variations of liquid-phase
turbulent uctuations and internal liquid circulation
gives rise to the overall eect of pressure on the liquid
mixing.
Recent study of ow elds and Reynolds stresses in
high-pressure bubble columns using laser Doppler
velocimetry (LDV) also conrmed that the bubbleinduced turbulence of the liquid phase is depressed at
higher pressures (Lee et al., 2001). Figure 23 shows the
eect of pressure on the proles of axial liquid velocity
and Reynolds normal stress measured by the LDV
technique. The ensemble-averaged velocity shows the
gross circulation pattern, and as the pressure increases,
the magnitudes of the averaged velocity of the liquid
phase at the center and in the wall region decrease. The
Reynolds stresses decrease signicantly with increasing
pressure, indicating the smaller extent of liquid-phase
uctuations at high pressures. The uctuations of the
liquid phase are mainly caused by the motion of gas
bubbles, and the uctuations are damped out as the
pressure increases because of a narrower bubble size
distribution with a smaller mean bubble size.
Recently, research in mixing behaviors of both the
liquid and the solid phase in three-phase uidized beds
under ambient conditions has continued. Chen et al.
(1995) and Zheng et al. (1995) studied the axial variations of the gas holdup, axial liquid mixing, and gas
liquid mass transfer. They observed a signicant
increase of the axial dispersion coecients of the liquid
phase and a decrease of the gasliquid mass transfer
coecients with an increase of the axial distance under
certain operating conditions, particularly under high
liquid ow rates. They attributed the signicant axial
variation in part to the coexistence of the dispersed bubble and coalesced bubble regimes along the axial direction. Kim et al. (1992) proposed empirical correlations
to account for the axial dispersion coecient of the
liquid phase over a wide range of operating conditions:
1:03 1:66
d pU l
dp
Ul
11:96
Pez
Ezl
Ul Ug
Dc
for the bubble coalescing regime with a range of variables of 0:004 dp =Dc 0:024 and 0:250 Ul =
Ul Ug 0:857, and
1:03 1:66
d pUl
dp
Ul
20:72
Pez
E zl
Ul Ug
Dc
48a
48b
V UCs
49
E
@t
@z
@z
where Cs is the solids concentration in a solidliquid
mixture, E is the solid phase axial dispersion coecient, and V and U are liquid and solids velocities,
respectively.
A theoretical analysis of the sedimentationdispersion model by Jean et al. (1989) indicated that in a
common expression of the sedimentationdispersion
model with Ul (dened as the cross-sectional averaged
approach in which both the gas and the solid phase are
treated as pseudocontinuous phases. Alternatively, a
second-order moment three-phase turbulence model
can be used. This approach was demonstrated satisfactorily in the simulation of the gasliquid ow in a bubble column (Zhou et al., 2002) and is readily extendable
to the gasliquidsolid ow simulation. There is, however, a practical need for a discrete simulation that
illustrates the inherent discrete ow characteristics of
individual phases such as the bubble wake structure.
The discrete phase simulation method has demonstrated its potential in simulating three-phase ows (Li
et al., 1999; Zhang et al., 2000). In this approach, the
Eulerian volume-averaged method, the Lagrangian
discrete particle method (DPM), and the volume-ofuid (VOF) volume-tracking method can be employed
to describe the motion of liquid, solid particles,
and gas bubbles, respectively. A bubble induced force
(BIF) model, a continuum surface force (CSF) model,
and Newtons third law are applied to account for the
couplings of particlebubble, bubbleliquid, and
particleliquid interactions, respectively. A close distance interaction (CDI) model is included in the
particleparticle collision analysis, which considers
the liquid interstitial eect between colliding particles
(Zhang et al., 1999).
In the following, the computational models for each
individual phase will be discussed. Relevant underlying
equations will be given, and the assumptions of each
model will be explained. Finally, simulated results of
several important phenomena in three-phase ows will
be demonstrated.
4.1
@el v
rl r
el vv rp rt rl g f b
@t
51
where the subscript ij denes the location of a computational cell; and V are the domain and volume of
this cell, respectively; xkp is the location vector of particle k; Ffp is the uidparticle interaction force acting
on any individual particles, which includes the drag
(FD), buoyancy (FB), added mass (FAM), and Basset
force (FBA):
F fp F D F B F AM F BA
53
Liquid-Phase Model
@el
r
el v 0
@t
t is the viscous stress tensor; g is gravitational acceleration; f b is the total volumetric body force acting on
the liquid phase excluding the gravitational force, that
is, the volumetric particleuid interaction force (fpf)
plus the volumetric bubbleuid interaction force (fbf).
Based on Newtons third law of motion, the force acting on a particle from the liquid phase, Ffp, yields a
reaction force on the liquid. Therefore the momentum
transfer from particles to the liquidgas phase is taken
into account in Eq. (51) by adding the volumetric
particleuid interaction force, fpf, given below to
the body force term,
P k
F fp
f pf
xkp 2 ij
52
Vijk
54
where the plus sign is used for the liquid phase and the
minus sign for the gas phase; d(x,t) is the Delta function, which equals 1 at the gasliquid interface and 0
elsewhere; s is the surface tension of the interface;
k(x,t) is the curvature of the free surface; a(x, t) is
the volume fraction of the uid.
The Newtonian viscous stress tensor is used in Eq.
(51) and given as
t 2ml S ml rv rvT
55
4.2
Gas-Phase Model
56
Qm
and
where Q represents a property of the uid, and
Qg represent the properties of the liquidsolid suspension and the gas bubble, respectively. By denition,
a(x, t) 1 in the liquid or liquidsolid mixture,
0 < a(x, t) < 1 at the free surface, and a(x, t) 0
in the gas bubble. Therefore Q is replaced by Qm
when a(x, t) equals 1 in the liquidsolid suspension
or Qg when a(x, t) equals 0 inside the gas bubble.
The advection equation for a(x, t) is
@a
v
ra 0
@t
57
58
where the surface pressure, ps , is the surface-tensioninduced pressure jump across a uid interface. The
dvp
F total
dt
59
60
The drag force acting on a suspended particle is
proportional to the relative velocity between the
phases:
0
rl Av vp v vp
61
F D 12 e2l CD
where A is the exposed frontal area of the particle in
the direction of the incoming ow, and CD0 is the eective drag coecient, which is a function of the particle
Reynolds number, Rep . For isolated rigid spherical
particles the drag coecient CD can be estimated by
the following equations (Rowe and Henwood, 1961)
8
>
< 24 1 0:15Re0:687
63
64
p
2
rp nl fH3 Re
fH Re 0:75 0:105Re;
Udp
Re
nl
66
68
1
2 rp 16 rp h3 dh
4
9 ml f f vp v 9 rl rp vp vvp v
2
2 r p rp
32 rp h4
vp
vp
!
r g
1 l
rp v p
t
9ml Kt tdvp v=dtdt
0
2r2p rp vp
69
"
0:47 #
Rep 0:44 rp 0:19 rp Rep
f exp
1:7
rl
h
f 1
70
0:15Re0:687
p
UaN UbN
e
Ub N UaN
0
71
UaT UbT
72
73
74
75
T
UaT oa0
ra
Uac
0
T0
Ubc UbT ob0
rb
76
Figure 25 Simulation of a single bubble rising in a liquidsolid suspension at high pressure (nitrogen-Paratherm NF heat
transfer uid0.88 mm glass beads, P 17.3 MPa, es 0.384, db 7.5 mm). (From Zhang et al., 2000.)
Paratherm liquids is shown in Fig. 26. In the simulations, an 80 mm (width) 50 mm (height) domain is
used with two orices at the bottom, an exit at the top,
and two sidewalls as boundary conditions. The liquid
initially lls the domain to a certain height. At the
beginning of the simulation, gas is injected into the
liquid through two orices. Paratherm NF heat transfer
uid and nitrogen are used as liquid and gas phases,
respectively. The gas inlet velocity is 14.0 cm/s. The
diameter of each orice is 6 mm. To study the interaction between bubbles formed from dierent orices, the
orice separation distance is set as 20 mm. As can be
seen from the gure, due to the closeness of the orices,
the uctuations of ow eld induced by the bubble
forming from one orice have an eect on the bubble
forming from the other orice. This interactive eect
yields varied bubble shapes and an alternate detachment pattern of bubbles from two orices. Initially
the bubble from each orice rises rectilinearly (Fig.
26a); after detachment, however, the bubbles tend to
move to the centerline, rise in zigzag fashion, and break
up at the free surface (Figs. 26df). Based on simulation
results, it is found that owing to the closeness of the
orices, two forming bubbles induce a high liquid velocity in the area near the centerline. The high velocity
thus results in a lower pressure in this area and draws
bubbles toward the centerline. In other words, bubble
and bubble wake interaction leads to a zigzag bubble
SUMMARY
Figure 26 Simulation of two-bubble formation in Paratherm NF heat transfer uids with bubblebubble interactions
(P 6.6 MPa, Do 6 mm, Lor 20 mm). (From Li et al., 2001.)
NOMENCLATURE
A
a
b
C
CD
CD0
Cpl
Cs
Dc
Dmax
Do
db
db0
dp
dvs
E p
E 1 a2
Eo
Eo 0
Erl
Ezl
e
FAM
FB
FBA
Fbp
FC
FD
Ffp
FG
Fi
FI,g
FI,m
FM
Ftotal
Fx
Fs
Fr
f
fb
fbf
fk
fpf
f db
g
H
h
h0
hi
I
K
Kb
Kb0
kl
ko
L
Lor
l
Mo
Mom
mp
Nc
Nu 0
ni
P
Pb
Pc
Pe
Pez
P0
Pr
Prm
p
ps
pv
Q
Qg
Qg
Qm
Q0
R
Rd
Re
Re 0
Rem
Rep
r
rb
rc
ro
rp
S
Stm
T
T
t
tc
UN
UT
Ub
Ub1
Ug
Ug,tran
Ul
Ulmf
u
ub
ub0
ub
ue
um
umax
uo
upt,0
ur
usmall
ut
ux
Vb
Vc
Vp
v
vp
x
xkp
y
z
hvi
hv 0 v 0 i
Greek Letters
a
acr
bd
bU
eg
eg,tran
el
es
esc
es0
f
k
lc
mg
ml
mm
nl
rg
rl
rm
rp
rs
s
t
o
x
z
V
r
REFERENCES
Allen MP, Tildesley DJ. Computer Simulation of Liquids.
Oxford: Clarendon Press, 1987.
Attia YA, Elzeky M, Bavarian F, Fan LS. Cleaning and
desulfurization of high sulfur coal by selective occulation
and bioleaching in a draft tube uidized bed reactor. In:
Dugan P, Quigly D, Attia YA, eds. Processing and
Utilization of High Sulfur Coals IV. Elsevier, 1991,
pp 181201.
Azbel D. Two-Phase Flows in Chemical Engineering.
Cambridge, UK: Cambridge University Press, 1981.
Baird M, Rice R. Axial dispersion in large scale unbaed
columns. Chem Eng J 9:171174, 1975.
Yang GQ, Luo X, Lau R, Fan LS. Heat-transfer characteristics in slurry bubble columns at elevated pressures and
temperatures. Ind Eng Chem Res 39:25682577, 2000b.
Yoo DH, Tsuge H, Terasaka K, Mizutani K. Behavior of
bubble formation in suspended solution for an elevated
pressure system. Chem Eng Sci 52:37013707, 1997.
Zahradnik J, Fialova M, Ruzicka M, Drahos J, Kastanek F,
Thomas NH. Duality of the gasliquid ow regimes in
bubble column reactors. Chem Eng Sci 52:38113826,
1997.
Zaidi A, Deckwer WD, Mrani A, Benchekchou B.
Hydrodynamics and heat transfer in three-phase uidized
beds with highly viscous pseudoplastic solutions. Chem
Eng Sci 45:22352238, 1990.
Zhang JP, Grace JR, Epstein N, Lim KS. Flow regime identication in gasliquid ow and three-phase uidized
beds. Chem Eng Sci 52:39793992, 1997.
Zhang JP, Epstein N, Grace JR. Minimum uidization velocities for gasliquidsolid three-phase systems. Powder
Technology 100:113118, 1998a.
Zhang JP, Epstein N, Grace JR. Minimum liquid velocity for
particle transport from gasliquid uidized beds. In: Fan
LS, Knowlton TM, eds. Fluidization IX. New York:
Engineering Foundation, 1998b, pp 645652.
Zhang J, Fan LS, Zhu C, Pfeer R, Qi D. Dynamic behavior
of collision of elastic spheres in viscous uids. Powder
Technology 106:98109, 1999.
Zhang J, Li Y, Fan LS. Numerical studies of bubble and
particle dynamics in a three-phase uidized bed at elevated pressures. Powder Technology 112:4656, 2000.
Zheng C, Chen Z, Feng Y, Hofmann H. Mass transfer in
dierent ow regimes of three-phase uidized beds. Chem
Eng Sci 50:15711578, 1995.
Zhou LX, Yang M, Lian CY, Fan LS, Lee DJ. On the
second-order moment turbulence model for simulating a
bubble column. Chem Eng Sci 57:32693281, 2002.
28
LiquidSolids Separation
Shiao-Hung Chiang, Daxin He, and Yuru Feng
University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
These unit operations are integral parts of many industrial processes mainly to serve the purposes of recovering
1.
2.
3.
4.
LiquidSolids Systems
Aqueous Systems
1.1.2
Nonaqueous Systems
A major problem in the treatment of nonaqueous systems is the removal of colloidal particles to produce
acceptable liquid products (e.g., fuel oil). Such a situation is encountered in hydrocarbon production from
tar sands and oil shale. The particles (such as oxides,
silicates, and clay mineral) suspended in the hydrocarbon liquids originate from a rock matrix. Particle
separation problems occur in the solvent extraction of
bitumen with nonaqueous media such as toluene.
Electrostatic forces (bonding forces) play a predominant role in the physical state of these nonaqueous
systems. In many instances, by addition of antisolvents, and selecting the proper temperature and
agitation, these systems can be altered to improve
solid separation. Separation of carbon black particles
suspended in tetralin using Aerosol OT as a surfactant
and by ltration through a bed of sand (deep-bed ltration) is another example of liquidsolids separation
in nonaqueous systems.
1.1.3
Biological Systems
Industrial Applications
As mentioned earlier, liquidsolids separation technology is basic to many manufacturing industries (chemical, mineral, food, beverages, etc.) as well as to
pollution abatement and environmental control. It is
dicult to nd any important engineering enterprises
in which liquidsolids separation does not play an
important part. Major industrial and commercial
applications of four key unit operations for liquid
solids separation are summarized here:
Unit Operation
Applications
Filtration
Sedimentation
Centrifugation
Hydrocyclones
FILTRATION
Figure 1 Mechanisms of ltration: (a) cake ltration, (b) deep-bed ltration, (c) cross-ow membrane ltration. (From Chiang
and He, 1995.)
dt
macV=A Rm
1
krs 1 e
2
A P
dV P1 msA
where
rs
c
1 ms
4a
4b
Many industrial ltration processes can be approximated as constant rate ltration. In this case, the linear
velocity of ltrate, u V=At, remains constant. The
relation between the overall pressure drop, P, and
the ltration time, t, can be expressed as (McCabe et
al., 1993)
2
V
1n
a0 mc
t Kr t
6
P Pm
At
where Pm is the pressure drop across the lter
medium and Kr a0 mcV=At2 is a constant.
In industrial ltration operation carried out under
variable pressure and variable rate conditions, the
method of Tiller must be employed to integrate Eq.
(1) for the general case (Tiller, 1958).
2.2
Deep-Bed Filtration
For a clean lter bed, the pressure drop can be calculated using the CarmanKozeny equation,
@P
150
1 e0 2
2 mu
@L 0 dp
e30
Membrane Filtration
Basic Concepts
C S CP
CS
10
and
R0
Cb CP
Cb
11
Microltration
Ultraltration
13
14
where Vf is the volume of feed. The volume concentration ratio (VCR) is dened as
VCR
Vf
Vr
15
dV
e3
P
12
V f Vp
Jav
16
The conguration of UF is usually designed as polymeric and asymmetric modules for high productivity
and resistance to plugging. Membrane modules used in
the UF process design are similar to those adopted in
MF process design (see previous section).
2.4 Filter Media
After specifying the lter type and the optimum operating conditions, the remaining issue for a lter design
is the selection of the most suitable lter medium. In a
ltration process, the fundamental role of a lter medium is to separate eectively the particulates from
a owing uid to provide a suciently clean ltrate
without clogging and damaging the medium (i.e., low
Figure 3 Membrane modules for microltration and ultraltration. (a) Tubular membrane module. (From Zeman and Zydney,
1996.) (b) Cassette membrane assembly. (From Perry et al., 1997.) (c) Spiral wound membrane module. (From Purchase, 1996.)
(d) Hollow ber membrane module. (From Zeman and Zydney, 1996.)
Company
Location
Flat plate
Dorr-Oliver USA
Ioplate
Flat plate
DDS
Denmark
Tube
Abcor 24.0
USA
Tube
RomiconAmicon
USA
Tube
Berghof
Germany
Tube
Enka/
Membrana
Germany
Product
Codea
Membrane
type
Lab Unit 35
US1
US2
US3
US4
Polysulphone
Polyvinylidene uoride
VC-AN copolymer
Polysulphone
Lab Unit 20
HFA
US5
UT1
Polysulphone
Cellulose acetate
HFM
PM
UT2
UT3
Polyvinylidene uoride
Polysulphone
XM
BM
UT4
UT5
Modacryl polymer
Polyamidimide
Dyna-Sep
Sampler
MT1
Polypropylene
Microdyne
MT2
Polypropylene
Um
(ms1 )
Height or
i.d. (mm)
L
(m)
Membrane
area (m2 )
Rec
103
Rew d
103
Ultraltration
5.012.2
5.012.2
5.012.2
0.681.36b
2.54
2.54
2.54
5.90b
0.37
0.37
0.37
0.22
0.067
0.067
0.067
0.150
1:894:25 104
1.801:42 104
0.951:18 104
0.696:94 105
21.84
21.84
21.84
6.02
48.0108.0
30.036.0
24.030.0
4.041.0
0.100.20b
7.60b
5.19b
2.54
0.076
3.05
0.018
0.204
0.168:10 105
9:43 105
0.78
19.30
6.042.0
11.3622.7b
3.326.48
2.54
1.10
3.05
0.635
0.204
0.022
1.412:82 104
0.1161:62 104
43.26
5.39
36.072.0
1.318.0
1.10
0.60
0.635
0.30
0.0021
5:65 104
1.391:74 104
0.1166:94 105
Microltration
4.217.97
5.50
1.83
0.032
0.3651:62 104
33.44
20.089.0
1.03.0
1.80
0.56
1.2
0.5562:78 104
3.60
10.050.0
2.846.02
0.0110.266
VW
(ms1 )
4.87
0.083
15.019.0
0.074.20
% free area
540
1520
2550
3040
45
2555
5060
80
6095
7085
8090
93
particulates within the lter media. The detailed information regarding the lter media properties and their
selection can be found in the Handbook of Filter
Media (Purchas, 1996) and the Filters and Filtration
Handbook (Dickenson, 1994).
A good lter medium should have the following
features:
1.
2.
3.
4.
5.
6.
2.5
As a general rule, a suitable lter aid should be capable of creating a thin layer structure with a maximum pore size over the mediums external surface
and producing a prespecied ltrate clarity at an optimum ltration rate.
Filter aids may be applied to the ltration operation
in two ways. The rst way is to precoat the lter medium using a precoat lter aid. The precoat is to behave
as the actual lter medium. The function of such an
application is to prevent the lter medium from clogging or fouling as well as to facilitate the removal of
the formed cake at the end of ltration. The second
way is to pretreat the suspension using a lter aid
powder with a coarser size distribution prior to the
ltration process. Such material is called body aid or
admix. The functions of body aid are to increase the
porosity of lter cake and to decrease its compressibility, resulting in a decrease in the cake resistance and in
turn an increase in the ltration rate.
2.5.1
These requirements are all found in the two most common lter aids. The rst one is the diatomaceous silica
type lter aid (also called diatomite, kieselguhr or diatomaceous earth), which contains 90% or almost pure
silica and particle size mostly smaller than 50 mm. Its
bulk density ranges from 128 to 320 kg/m3 . Calcinated
diatomaceous additives display their high retention
ability with relatively low hydraulic resistance. the relative permeability of the calcinated diatomite increases
up to 320 times that of natural diatomite. The disadvantage of diatomite is that it may foul ltering liquids
by dissolved salts and colloidal clays. The second most
common lter aid is expended perlite, which is a
glasslike volcanic rock. The porosity of this lter
aid is in the range of 0.850.9. Its bulk density ranges
Smallest particle
retained, microns*
Subdivisions
Solid fabrications
100
10
5
Metal sheets
(a) perforated
(b) woven wire
20
5
(a)
(b)
(c)
(d)
10
5
1
1
Plastic sheets
Woven fabrics
10
5
Nonwoven media
10
plastics
sintered metals
ceramics & stoneware
carbon
10
10
<0.1
10
5
2
0.5
Cartridges
5
5
3
Loose media
(a) bers
(b) powders
1
<0.1
Membranes
(a) ceramic
(b) metal
(c) polymeric
0.2
0.2
<0.1
2.5.2
Filtration Equipment
Figure 4 Rotary drum vacuum lter. (a) Cut-away view. (From Dickenson, 1994.) (b) Filtration cycle. (From Rushton, 2000.)
2.6.2
This arrangement leads to longer operating cycles, particularly in low-pressure systems. For potable or wastewater ltration, a minimum top eective size of 1.1 to
1.7 mm for the coarse medium (anthracite) is recommended. It has been demonstrated that the use of coarser media can reduce the ltration costs, because such a
lter can be run at a higher ow rate with a lower pressure drop. Obviously, the use of coarser particles
requires deeper bed media to meet the demand of
equivalent media surface area for ltrate quality. In
some cases, a thin layer of dense solids, such as alumina
or gravel, has been used as a third layer situated beneath
the sand in the modied design of a dual layer deep-bed
lter. However, modern lter design still mostly stems
from the design version used in the 1980s. It uses coarse
media at the bottom and ner sand media at the top (see
Fig. 7ac). It was reported that good quality water without chemical pretreatment was produced at a rate of 45
million gallons per acre per day. A comparison of operating characteristics of various deep-bed lters is presented in Table 4 (Rushton et al., 2000).
NOMENCLATURE
A
C
Cb
CP
CS
c
dp
F
J
Jav
Jf
Ji
K
K0
Kr
k
L
Lm
m
n
P
Pc
Pm
R
R0
Rm
R0
Sp
s
t
permeate concentration
retentate concentration at the membrane
surface
mass of solid per unit volume of ltrate
particle (grain) diameter
function describing the eect of particle
deposition on deep-bed ltration coecient
ux
average ux
nal ux at the highest concentration
initial ux
Kozeny coecient
constant
constant
permeability
distance from the top of the bed to the section
under study or bed depth
eective membrane thickness
mass ratio of wet cake to dry cake
compressibility constant of lter cake
overall pressure drop across the lter
pressure drop across the lter cake
pressure drop across the lter septum
true retention coecient
observed retention coecient
resistance of lter medium (e.g., septum)
constant
surface area of a single particle
slurry concentration
ltration time
u
V
Vf
Vp
Vr
VCR
liquid ow rate
ltrate volume
volume of feed
volume of permeate passed through the
membrane
volume of retentate, retained on the membrane
surface
volume concentration ratio
Greek Symbols
a
ao
a
e
e0
l
l0
m
r
rs
s
Continuous upow
Coarse downow
Multimedia
Direction of ow
Cleaning
Media size, mm
upwards
continuous
sand, 12
downwards
sequenced
sand, 13
downwards
sequenced
sand, 0.41
anthracite, 0.82
Sphericity of media
0.8
0.80.9
Media depth, m
1.2
1.52
Vessel depth, m
Surface loading, m3 =m2 h
Hydraulic stability
3.57
814
low
35
618
high
sand, 0.8
anthracite, 0.7
sand, 0.5
anthracite, 0.3
23.5
816
high
Removal eciency
moderate
low
Water, m3 =m2 d
Air, m3 =m2
Time
Relative capital cost
Relative operating cost
Back-wash cycle
19
30
1012 at 7.5 bar
120 at 1 bar
continuous
2 h/d
high
low
high
low
sand: high
anthracite: low
30
50 at 0.5 bar
1 h/d
moderate
low
REFERENCES
Akers, RJ, Ward AS. Liquid ltration theory and ltration
pretreatment. In: Orr, CC, ed. Filtration: Principles and
Practices, Part I. New York: Marcel Dekker, 1977, pp
169250.
Belfor G. Fluid mechanics and cross-ow membrane ltration. In: Muralidhara HS, ed. Advances in SolidLiquid
Separation. Columbus: Battelle Press, 1986, pp 182183.
Cheremisino NP. Solid/Liquid Separation. 2d ed.
Lancaster, PA: Technomic, 1995, p 240.
Cheremisino NP. Liquid Filtration. 2d ed. Boston:
Butterworth-Heinemann, 1998, p 277.
Cheremisino NP, Azbel DS. Liquid Filtration for Process
and Pollution Control. New Jersey: SciTech Publishers,
1989, p 59.
Cheryan M, Ultraltration Handbook, Lancaster, PA:
Technomic, 1986, p 2.
Chiang SH, He DX. Section 62: Fluid/particle separation. In:
Dorf RC, ed. Engineering Handbook. Boca Raton, FL:
CRC Press, 1995, pp 644651.
Dickenson C. Filters and Filtration Handbook. 3d ed. UK:
Elsevier Science, 1994, p 214.
Fitch B. When to Use Separation Techniques other than
Filtration. In: Shoemaker W, ed. AIChE Symposium
SEDIMENTATION
Sedimentation is the separation of suspended solid particles from a uid stream by the action of a body force
on the settling behavior of the particle. The body force
may be either gravitational or centrifugal force. This
section covers gravity sedimentation, represented by
clarication and thickening. The equipment used for
these two operations are called clariers and thickeners, respectively. Centrifugal sedimentation will be
discussed in the next section.
From a unit operation standpoint, clarication and
thickening are essentially based on the same design
principles, and each combines features of the other.
The key objective of a clarication operation is to
remove small quantities of suspended particulates
from the liquid stream to produce a claried euent
or overow stream. In thickening operation, the goal
is to concentrate the dilute suspensions for their
further treatment in lters or centrifuges. Key features of these two types of operation are presented
in Fig. 8.
3.1
Sedimentation Fundamentals
3.2 Thickeners
As shown in Fig. 10, there are four zones existing in a
thickener: a clarication zone, a feed zone, a transition
zone, and a compression zone. As solids thicken, a
critical concentration will be reached. This concentration would aect the passage of solids to the underow, causing a buildup in the compression zone
thickness. The thickener design and operation must
prevent the solid concentration from reaching the critical value.
3.2.1
Thickener Design
1=C 1=Cu
ui
17
where C is the test solid concentration; Cu the underow solids concentration; and ui the initial settling rate
at the test condition.
For a occulated suspension, the unit area of the
thickener is determined by
Ao
tu
C o Ho
18
Vc
tc rs rl
rs rp rl
19
20
Equations
ut
Note
Kdp2 rs r
m
Rep
dp ut r
< 0:1
m
Rep 0:2
u ut en
non-occulated suspension
(Richardson and Zaki, 1954a,b)
n is a function of the particle Reynolds number (see Table 5b)
u ut 1 kf fS 4:7017:8
d =D
occulated suspension
(Scott, 1984)
f kf f s
f settling units
fs volumetric dry concentration
kf the factor allowing for the liquid closely associated with the solids
d the mean volume-surface-length diameter relevant to sedimentation
4:65 19:5 dp =D
4:35 1:75 dp =D Rep0:03
4:45 18 dp =D Rep0:1
4:45 Rep0:1
2.39
4.65
4:35 Rep0:03
4:45 Rep0:1
4:45 Rep0:1
2.39
not greater than 20% of the design level to avoid shutdown due to inevitable process upsets or overloads.
Rake speed requirements depend on the type of solids
entering the thickener. Rake speed ranges used are 3
to 8 m/min for slow-settling solids, 8 to 12 m/min for
fast settling solids, and 12 to 30 m/min for coarse
solids or crystalline materials.
The detailed description of the major mechanical
components used for the thickeners, such as tank,
feed well, drive support structure, drive and lifting
devices, rake structure, underow withdrawal, and
overow collection systems can be found in the litera-
Light
Standard
Heavy
Extra heavy
5
5
15
525
0.21
2555
<0.2
>55
0
100
1.251.5
1575
05
85100
1.52.7
75150
515
5085
2.74.0
150300
>15
<50
>4.0
>300
3.3
Clariers
Figure 11 Conventional thickener with bridge-supported mechanism (Eimco). (From Schweitzer, 1997.)
3.3.1
Clarier Design
Feed
Underow
Unit area,
m2 day/mg
Overow rate,
m3 /(m2 h)
110
34
0.56
312
14
2050
1025
2040
2045
1520
1.23
25
0.51a
0.33a
310
0.070.12
0.71.7
0.50.8
1450
1030
25
4075
4565
4560
0.22
0.41
0.51.5
1.22.4
0.020.05
13
13
0.010.05
0.010.05
510
0.51.5
510
515
47
28
2040
8
1.218a
0.62.5
11.7
11.2
1.22.2
High-rate thickeners using required occulant dosages operate at 10 to 50% of these unit areas.
FGD represents ue-gas desulfurization.
Source: Perry RH, Green DW, Maloney JO. Perrys Chemical Engineers Handbook. 7th ed. New
York: McGraw-Hill, 1997, pp 1872, Table 18-7.
b
into the feed stream without creating too much turbulence, while keeping the maximum overow clarity.
The usual operating range for the upow clarier is 2
to 5 m3 /(h m2 ). The retention time required for the oc
separation in a sedimentation basin can be estimated
using Stokes law or Newtons law, especially for welldened homogeneous systems. It should be noted that
allowances must be made for deleterious eects due to
convection, air-induced surface waves, and inlet and
outlet turbulence.
If the clarier is assumed to operate close to its
capacity limit, the solids ux, G, into the clarier at a
steady state can be determined using a simple material
balance:
G
CQ Qu
A
21
Figure 14 Schematic of a tilted-plate clarier (Heil Process Equipment Co.). (From Schweitzer, 1997.)
NOMENCLATURE
A
Ao
C
Co
Cu
D
G
Ho
kT
Q
Qu
T
tC
tu
ui
VC
settling time
initial settling rate or hindered settling velocity
at solids concentration Ci
unit volume of the compression zone
Greek Symbols
rl
rp
rs
REFERENCES
Mukai T. Solidliquid separations of water and wastewater
treatment systems in Japan. In: Muralidhara HS, ed.
Advances in SolidLiquid Separation. Columbus:
Battelle Press, 1986, p 409.
Osborne DG. Gravity thickening. In: Svarovsky L, ed. Solid
Liquid Separation. 3d ed. London: Butterworths, 1990,
pp 132201.
Perry RH, Green DW, Maloney JO. Perrys Chemical
Engineers Handbook. 7th ed. New York: McGraw-Hill,
1997, pp 1860 to 1872.
Richardson JF, Zaki WN. Sedimentation and uidization:
Part 1. Trans Inst Chem Engrs 32:3553, 1954a.
Richardson JF, Zaki WN. The sedimentation of a suspension
of uniform spheres under conditions of viscous ow.
Chem Eng Sci 3:6573, 1954b.
Rushton A, Ward AS, Holdich RG. SolidLiquid Filtration
and Separation Technology. 2d ed. Reinheim, Germany:
Wiley-VCH, 2000, p 109.
Schweitzer PA. Handbook of Separation Techniques for
Chemical Engineers. 3d ed. New York: McGraw-Hill,
1997, pp 4.1404.156.
Scott KJ. Hindered settling of a suspension of spheres.
Critical evaluation of equations relating settling rate to
mean particle diameter and suspension concentration.
CSIR Report CENG 497. CSIR, Pretoria, South Africa,
1984.
CENTRIFUGATION
become identical with that of the bowl. There is therefore no tangential ow in the bowl, and the uid only
rotates with the bowl and moves upward through the
bowl at a constant velocity, carrying solid particles
with it. As the solid particles move with the uid,
they are subjected to high centrifugal forces and
begin to settle at some position in the uid, say at a
distance rA from the vertical axis, following trajectory
similar to that shown in Fig. 16. Its settling time is
limited by the residence time of the uid in the bowl;
at the end of this time let the particle be at a distance rB
from the axis of the rotation. If rB < r2 , the solid particle leaves the centrifuge with the liquid; if rB r2 , it is
deposited on the bowl wall and removed from the
liquid.
Ambler (1952) introduced the cut point concept in
the sedimentation centrifuge separation. It is dened as
the diameter of the particle that just reaches one-half
the distance between r1 and r2 . If a solid particle is to
be removed from the uid, it must travel the distance
r2 r1 =2 to the bowl wall in the available residence
Copyright 2003 by Taylor & Francis Group LLC
pbo2 rD2pc
r22 r21
18m
ln2r2 =r2 r1
22
3.
4.
5.
6.
fuge is somewhere in between and suitable for moderate feed solids concentrations.
4.2
Filter Centrifuges
Figure 18
23
where A2 is the area of the lter medium, A L the arithmetic mean cake area, A a the logarithmic mean cake
area, Rm the lter medium resistance, a the specic
cake resistance, and mc the total mass of the solids in
the lter. The mean areas A L and A a are dened by the
equations
A a ri r2 pb
24
2pbri r2
A L
lnr2 =ri
25
Figure 20
Peeler centrifuge: (a) feed and separation phase, (b) solids discharge phase. (From Jacobs and Penney, 1987.)
Figure 21 Schematic of (a) single-stage pusher centrifuge, (b) two-stage pusher centrifuge. (From Perry et al., 1997.)
The capacity and performance of the pusher centrifuge depend on the particle size and shape of the feed:
the coarser the feed particle size, the higher the operation capacity and the lower the moisture. The thickness
of the cake is one of the most important factors aecting the operation of the pusher centrifuge. The pusher
centrifuge has three signicant operating characteristics
dependent on the cake thickness: the solid discharge,
the lter rate, and the pushing force. Generally, the
thickness of the cake on the rotor is proportional to
the friction of the cake over the lter medium, the
centrifugal force, and the length of the rotor.
Conical Screen Filter Centrifuges The conical
screen centrifuge is another very commonly used continuous centrifuge. Instead of using the pusher action
to discharge the solids, it uses a helical conveyor that is
turning slightly slower or faster than the rotor to discharge the solids. Such a solid discharge mechanism
can be used for both cylindrical and conical rotors.
The dierential speed of the conveyor controls the
rate at which the solids move through the drainage
zone. A schematic of a conical screen centrifuge is
shown in Fig. 22. The conical screen centrifuges may
have vertical or horizontal axes of rotation.
The conical screen centrifuges are used mainly for
processing coarser particles, for example crystalline
salts, coal, and minerals. These centrifuges are suitable for relatively large throughputs, up to 320,000 kg/
h. The various types of conical screen centrifuges
operate successfully in a wide range of applications.
Among them, the sliding lter centrifuges and the
B. Filtering Centrifuges
Minimum particle size, micron
Maximum particle size, micron
Allowable concentration of feed solid, %
Condition of cake
Typical solids handling rate, lb/h
Tubular
Disc
(nozzle type)
Conveyor
bowl
0.1
200
0.1
Pasty, rm
0.15
0.25
50
220
Fluid
1030,000
2
5,000
260
Firm, pasty
100100,000
Batch
vertical
Conical
basket
Pusher
10
1,000
210
Pasty
0.11.0
250
10,000
4080
Relatively dry
540
40
5,000
1575
Relatively dry
0.55.0
Source: Cheremisinoff PN. Solid/Liquids Separation. Lancaster PA: Technomic, 1995, p 240, Table 6.3.
NOMENCLATURE
A L
A a
A2
b
Dpc
G
mc
Nc
qc
r1
r2
rA ; r B
ri
Rc
Rm
Greek Symbols
a
r
r
m
o
HYDROCYCLONES
The hydrocyclone is based on the principle of centrifugal force causing the separation of solids from a
liquid by the dierences in density and particle size.
A typical hydrocyclone consists of a cylindrical section
and a conical section, as shown in Fig. 23. It does not
have any internal rotating parts. An external pump is
used to transport the liquid suspension to the hydrocyclone through a tangential inlet at high velocity,
which in turn generates the uid rotation and the
necessary centrifugal force. The outlet for the bulk of
the liquid is connected to a vortex nder located on the
axis of the cylindrical section of the vessel. The underow, which carries most of the solids, leaves through
an opening (apex) at the bottom of the conical section.
The principle and basic design of the hydrocyclone
has been known for more than a century (Bretney,
1891), but it did not nd signicant application in
industry until the late 1940s. These separation devices
were rst used in mining and mineral processing, but in
recent years their applications have spread to many
other industries, including chemical manufacturing,
power generation, and environmental cleanup.
5.1 Separation Efciencies
REFERENCES
Alt C. Centrifugal separation. In: Muralidhara HS, ed.
Advances in SolidLiquid Separation. Columbus, Ohio:
Battelle Press, 1986, pp 107139.
Ambler CM. The evaluation of centrifuge performance.
Chem Eng Progress 48:150158, 1952.
Jacobs LJ, Penney WR. In: Rousseau RW, ed. Handbook of
Separation Process Technology. New York: John Wiley,
1987, p 168.
McCabe WL, Smith JC, Harriott P. Unit Operations of
Chemical Engineering. 5th ed. New York: McGraw-Hill,
1993, pp 10111028, 10641072.
Perry RH, Green DW, Maloney JO. Perrys Chemical
Engineers Handbook. 7th ed. New York: McGraw-Hill,
1997, pp 1864.
Svarovsky L. SolidLiquid Separation Processes and
Technology. Amsterdam: Elsevier, 1985, pp 72106.
Total Efciency
Figure 23 Hydrocyclone: (a) schematic diagram, (b) ow pattern. (From Rushton, 2000.)
ET
UCu
QCf
26
There are two problems associated with total eciency. First, if a hydrocyclone delivers both liquid
and solid to the underow and nothing to the overow,
an ideal total eciency of 1 will result. Second, without
any separation, a hydrocyclone, by simply splitting the
feed to an overow and an underow, will result in
certain guaranteed total eciency. In order to overcome these weaknesses, several alternative denitions
of eciency are used. One of the denitions is the
reduced total eciency ET0 :
Copyright 2003 by Taylor & Francis Group LLC
ET0
ET Rf
1 Rf
27
5.1.3
U
Q
28
Grade Efciency
In place of the total eciency, a grade eciency corresponding to a particular particle size is used, since
a hydrocyclone is a size-dependent separator. A
graphical representation of the relationship between
the grade eciency and the particle size is called the
grade eciency curve, as shown in Fig. 24 (Svarovsky,
1985).
The grade eciency curve can be determined by
measuring the total eciency and the particle size
distribution of any two of the three streams (feed,
underow, and overow). One of the following
equations can be used for calculating the grade
eciency:
Gx ET
dFu x
dFf x
29
dFo x
dFf x
30
31
5.1.4
Gx Rf
1 Rf
32
Cut Size
5.2.3
33
5.2.4
Theories of Separation
34
Eu
p
n2 r=2
35
Re
Dnr
m
36
Stk50
Operating Variables
37
As the operating pressure drop increases, the tangential velocity increases. Consequently, the higher
separation eciency or smaller cut size d50 results.
Generally, this relationship only holds to be true
with the operating pressure up to 2 bar. Beyond that,
further increase in the operating pressure has little
eect on the separation eciency.
The second important operating variable is the feed
solid concentration. With increasing feed solid concentration, the separation eciency falls o rapidly owing
to its eect on liquid ow pattern and the interaction
among solid particles. Therefore hydrocyclones are
usually operated with dilute feed solids concentrations
(<2% by volume).
The adjustment of the underow orice (apex)
opening is also a very important operating variable.
Correct adjustment of this opening is vital for the
best operation of the hydrocyclone. In general, an
increase in the underow orice size causes an increase
in operating capacity, but it tends to reduce the cut size
and the underow solids concentration.
5.3.2
Design Variables
38
and
P / Dy
39
40
41
Dzi
42
and
Eu Kp
Renp
43
Scale-up constants
Rietemas design
0.28
D 0:075 m
Bradleys design
0.133
D 0:038 m
Mozley cyclone
0.154
D 0:022 m
Mozley cyclone
0.160
D 0:044 m
Mozley cyclone
0.197
D 0:044 m
Warman 3 00 Model R 0.29
D 0:076 m
R W 2515 (AKW)
0.20
D 0:125 m
0.34
0.40 5
0.20
Kp
Running cost
criteriona
np
(Stk50 4=3 Eu
20
0.0611
316
0.134
2.12
0.33 6.85
0.1111
446.5
0.323
2.17
0.1203 6381
3.20
0.25
0.57 7.71
0.1508 4451
4.88
0.32
0.57 7.71
0.2182 3441
8.70
0.20
0.31 4.0
15
0.1079
0.32
0.8
15
0.1642 2458
6.24
2.618 0.8
0
2.07
6.66
NOMENCLATURE
C1
Cf
Cu
d50
D
Di
Do
Du
ET
ET0
Eu
Ff x
Fo x
Fu x
Gx
G 0 x
Kp
l
L
n
np
p
Q
R
Re
Rf
Stk50
U
v
V
x
y
z
empirical constant
solid concentration in the feed
solid concentration in the underow
cut size
hydrocyclone diameter
inlet pipe diameter
diameter of the vortex nder
underow (apex) diameter
total eciency
reduced total eciency
Euler number
Cumulative percentage of solids with particle
size x in the feed
Cumulative percentage of solids with particle
size x in the overow
Cumulative percentage of solids with particle
size x in the underow
grade eciency
reduced grade eciency
empirical constant
length of the cylindrical section
total length of the hydrocyclone
empirical constant
empirical constant
operating pressure drop
feed volumetric ow rate
radius
Reynolds number
underow to throughput ratio
Stokes number of cut size
underow volumetric ow rate
supercial velocity
tangential velocity
empirical constant
empirical constant
empirical constant
Greek Letters
y
r
m
cone angle
density dierence between solids and liquid
liquid viscosity
REFERENCES
Bradley D. The Hydrocyclone. London: Pergamon Press,
1965.
Bradley D, Pulling DJ. Flow patterns in the hydraulic
cyclone and their interpretation in terms of performance.
Trans Inst Chem Eng 37:3445, 1959.
Bretney E. Water Purier. U.S. patent No. 453,105, May 26,
1891.
Driessen MG. Theory of ow in cyclone. Rev Ind Mining,
Special Issue 4: pp 449461, 1951.
Fahlstrom PH. Discussion (on hydrocyclone) Proceedings of
International Mining Processing Congress. Institute of
Mining and Metallurgy, 1960, pp 632643.
Fontein DF. Cyclone separator of solid mixtures of dierent
grain size and specic gravity. U.S. patent No. 2,550,341,
1951.
Kelsall DF. A study of the motion of solid particles in a
hydraulic cyclone. Trans Inst Chem Eng 30:87104, 1952.
Rietema K. Performance and design of hydrocyclones. Parts
I to IV. Chemical Engineering Science 15:298325, 1961.
Schubert H, Neesse T. A hydrocyclone separation model in
consideration of turbulent multi-phase ow. In: Proc
International Conference on Hydrocyclones. Cambridge,
BHRA Fluid Engineering, Craneld, 13 October 1980,
Paper 3, 2336.
Svarovsky L. Hydrocyclones. New York: Technomic, 1984,
pp 3660.
Svarovsky L, ed. SolidLiquid Separation Processes and
Technology. Amsterdam: Elsevier, 1985, pp 1827.
Svarovsky L. Hydrocyclones. In: Svarovsky L, ed. Solid
Liquid Separation, 3d ed. UK: Butterworths, 1990, pp
203248.