HANDBOOK of

FLCIIDIZATION and
FLUID-PflRTICLE SYSTEMS
 CHEMICAL INDUSTRIES

A Series of Reference Books and Textbooks

Consulting Editor

HEINZ HEINEMANN
Berkeley, California

1. Fluid Catalytic Cracking with Zeolite Catalysts, Paul B. Venuto and E. Thomas Habib, Jr.
2. Ethylene: Keystone to the Petrochemical Industry, Ludwig Kniel, Olaf Winter, and Karl Stork
3. The Chemistry and Technology o f Petroleum, James G. Speight
4. The Desulfurization o f H eavy Oils and Residua, James G. Speight
5. Catalysis o f Organic Reactions, edited by William R. Moser
6. Acetylene-Based Chemicals from Coal and Other Natural Resources, Robert J. Tedeschi
7. Chemically Resistant Masonry, W alter Lee Sheppard, Jr.
8. 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. Swearingen, and Marilyn E. Weightman
9. Metering Pumps: Selection and Application, James P. Poynton
10. Hydrocarbons from Methanol, Clarence D. Chang
11. Form Flotation: Theory and Applications, Ann N. Clarke and David J. Wilson
12. The Chemistry and Technology o f Coal, James G. Speight
13. Pneumatic and Hydraulic Conveying of Solids, O. A. Williams
14. Catalyst Manufacture: Laboratory and Commercial Preparations, Alvin B. Stiles
15. Characterization o f Heterogeneous Catalysts, edited by Francis Delannay
16. BASIC Programs for Chemical Engineering Design, James H. W eber
17. Catalyst Poisoning, L. Louis Hegedus and Robert W . M cCabe
18. Catalysis o f Organic Reactions, edited by John R. Kosak
19. Adsorption Technology: A Step-by-Step Approach to Process Evaluation and Application, edited by Frank L.
Slejko
20. Deactivation and Poisoning o f Catalysts, edited by Jacques Oudar and Henry W ise
21. 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
22. Catalysis o f Organic Reactions, edited by Robert L. Augustine
23. Modern Control Techniques for the Processing Industries, T. H. Tsai, J. W . Lane, and C. S. Lin
24. Temperature-Programmed Reduction for Solid Materials Characterization, Alan Jones and Brian McNichol
25. Catalytic Cracking: Catalysts, Chemistry, and Kinetics, Bohdan W . Wojciechowski and Avelino Corma
26. Chemical Reaction and Reactor Engineering, edited by J. J. Carberry and A. Varm a
27. Filtration: Principles and Practices: Second Edition, edited by Michael J. Matteson and Clyde Orr
28. Corrosion Mechanisms, edited by Florian Mansfeld
29. Catalysis and Surface Properties of Liquid Metals and Alloys, Yoshisada Ogino
30. Catalyst Deactivation, edited by Eugene E. Petersen and Alexis T. Bell
31. 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 W atanabe
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 F C C Catalysts, Julius Scherzer
43. Oxygen in Catalysis, Adam Bielanski and Jerzy Haber
44. The Chemistry and Technology o f 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 Carm o J. Pereira
52. Models for Thermodynamic and Phase Equilibria Calculations, edited by Stanley I. Sandler
53. Catalysis of Organic Reactions, edited by John R. Kosak and Thomas A. Johnson
54. Composition and Analysis o f H eavy Petroleum Fractions, Klaus H. Altgelt and Mieczyslaw M. Boduszynski
55. N M R Techniques in Catalysis, edited by Alexis T. Bell and Alexander Pines
56. Upgrading Petroleum Residues and H eavy Oils, Murray R. Gray
57. Methanol Production and Use, edited by W u-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 o f Coal: Second Edition, Revised and Expanded, James G. Speight
60. Lubricant Base Oil and W ax 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. Scaros 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,
W illiam 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 o f Porous Materials: Zeolites, Clays, and Nanostructures, edited by Mario L.Occelli and Henri
Kessler
70. M ethane 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 o f 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. The Desulfurization o f H eavy Oils and Residua, Second Edition, Revised and Expanded, Jam es 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 Param eter Estimation for Chem ical 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, Jam es R. Couper, O. Thom as
Beasley, and W . Roy Penney
84. Transport Phenom ena Fundamentals, Joel L. Plawsky
85. Petroleum Refining Processes, Jam es G. Speight and Baki Ozum
86. Health, Safety, and Accident M anagem ent in the Chemical Process Industries, Ann Marie Flynn and Louis
Theodore
87. Plantwide Dynamic Simulators in Chem ical Processing and Control, W illiam L. Luyben
88. Chemicial Reactor Design, Peter Harriott
89. Catalysis o f 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 W en-Ching Yang
92. Conservation Equations and Modeling o f Chemical and Biochemical Processes, Said S. E. H. Elnashaie and
Parag Garhyan
93. Batch Fermentation: Modeling, Monitoring, and Control, Ali Qinar, Gulnur Birol, Satish J. Parulekar, and
Cenk Undey
94. Industrial Solvents Handbook, Second Edition, Nicholas P. Cheremisinoff

ADDITIONAL VOLUMES IN PREPARATION

Chemical Process Engineering: Design and Economics, Harry Silla

Petroleum and Gas Field Processing, H. K. Abdel-Aal, Mohamed Aggour,, M. A. Naim

Process Engineering Economics, Jam es R. Couper

Thermodynamic Cycles: Computer-Aided Design and Optimization, Chih W u

Re-Engineering the Chemical Processing Plant: Process Intensification, edited by Andrzej Stankiewicz and
Jacob A. Moulijn
 HANDBOOK of
FLUIDIZATION and
FLUID-PflRTICLE SYSTEMS

edited by

Wen-Ching Yang
S ie m e n s W e s tin g h o u s e P o w e r C o rp o ra tio n
P itts b u rg h , P e n n s y lv a n ia , U .S .A .

CRC Press
Taylor Si Francis Group
Boca Raton London New York

CRC Press is an im print of the

Taylor & Francis Group, an informa business
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2003 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group

No claim to original U.S. Government works

Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2
International Standard Book Number-13: 978-0-8247-0259-5 (Hardcover)

This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are
indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the
publisher cannot assume responsibility for the validity of all materials or for the consequences of their use.

Except as permitted by U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage
or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact
the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system
of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation
without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Preface

Every chemical engineer, whether a student or practicing, has looked up technical information in Perry's 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 field’s most useful general-purpose
reference books. It was in this spirit of serving the profession that I undertook the task of compiling the Handbook
o f Fluidization and Fluid-Particle Systems. Through future revisions and additions, I sincerely hope that this hand­
book will become an archivable reference volume for every practitioner in this field, spanning the boundary of
various disciplines. Fluidization and fluid-particle system engineering is being applied in industries as diverse as
basic and specialty chemicals, mineral processing, coal and biomass gasification and combustion for power gen­
eration, environmental technologies, resource recovery, FCC petroleum refining, pharmaceuticals, biotechnology,
cement, ceramics, and other solids handling and processing industries. The first focused handbook ever published in
this extended field, 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
fluidization and fluid-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 dynamics—important in all aspects of particle
production, manufacturing, handling, processing, and applications—are discussed in Chapter 1. Chapter 2 presents
the flow through fixed beds and summarizes packing characteristics of spherical and nonspherical particles, pres-
sure-drop correlations for flow through fixed beds, and heat and mass transfer. Bubbling fluidized beds are pre­
sented in detail in Chapter 3, which covers all important aspects including jetting phenomena and particle
segregation, topics not addressed extensively in other books on fluidization. Other important design considerations
are treated in separate chapters: elutriation and entrainment in Chapter 4, effect of temperature and pressure in
Chapter 5, gas distributor and plenum design in Chapter 6, effect of internal tubes and baffles 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 fluidized bed reactors are elucidated in Chapter 12, “General
Approaches to Reactor Design,” and Chapter 13, “Fluidized Bed Scaleup.”
Important industrial applications for fluidized bed reactors are also discussed, including fluid catalytic cracking
(Chapter 14), gasifiers and combustors (Chapter 15), chemical production and processing (Chapter 16), coating and
granulation (Chapter 17), and fluidized bed drying (Chapter 18).
The important variation of bubbling fluidized beds—the circulation fluidized beds— are discussed in detail in
Chapter 19. Chapter 20 summarizes other nonconventional fluidized beds, including spouted beds, recirculating
fluidized beds with a draft tube, jetting fluidized beds, and rotating fluidized beds. The solids handling, transport
and circulating devices are described in Chapter 21, “Standpipe and Nonmechanical Valves,” and Chapter 22,

iii
iv Preface

“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 fluidized beds and fluid-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 fit the page limitation. The final 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 o f 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 effort. The staff 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 in
Contributors vii

1 Particle Characterization and Dynamics 1

Wen-Ching Yang
2 Flow Through Fixed Beds 29
Wen-Ching Yang
3 Bubbling Fluidized Beds 53
Wen-Ching Yang
4 Elutriaton and Entrainment 113
Joachim Werther and Ernst-Ulrich Hartge
5 Effect of Temperature and Pressure 129
J. G. Yates
6 Gas Distributor and Plenum Design in Fluidized Beds 155
S. B. Reddy Karri and Joachim Werther
7 Effect of Internal Tubes and Baffles 171
Yong Jin, Fei Wei, and Yao Wang
8 Attrition 201
Joachim Werther and Jens Reppenhagen
9 Modeling 239
Thomas C. Ho
10 Heat Transfer 257
John C. Chen
11 Mass Transfer 287
Thomas C. Ho
12 General Approaches to Reactor Design 309
Peijun Jiang, Fei Wei, and Liang-Shih Fan
vi Contents

13 Fluidized Bed Scaleup 343

Leon R. Glicksman
14 Applications for Fluid Catalytic Cracking 379
Ye-Mon Chen
15 Applications for Gasifiers and Combustors 397
Richard A. Newby
16 Applications for Chemical Production and Processing 421
Behzad Jazayeri
17 Applications for Coating and Granulation 445
Gabriel I. Tardos and Paul R. Mort
18 Applications for Fluidized Bed Drying 469
Arun S. Mujumdar and Sakamon Devahastin
19 Circulating Fluidized Beds 485
John R. Grace, Hsiaotao Bi, and Mohammad Golriz
20 Other Nonconventional Fluidized Beds 545
Wen-Ching Yang
21 Standpipes and Nonmechanical Valves 571
T. M. Knowlton
22 Cyclone Separators 599
T. M. Knowlton
23 Dilute-Phase Pneumatic Conveying 619
George E. Klinzing
24 Electrostatics in Pneumatic Conveying 631
George E. Klinzing
25 Instrumentation and Measurements 643
Masayuki Horio, Rafal P. Kobylecki, and Mayumi Tsukada
26 Liquid-Solids Fluidization 705
Norman Epstein
27 Gas-Liquid-Solid Three-Phase Fluidization 765
Liang-Shih Fan and Guoqiang Yang
28 Liquid-Solids Separation 811
Shiao-Hung Chiang, Daxin He, and Yuru Feng

Index 851
Contributors

Hsiaotao Bi Department of Chemical and Biological Engineering, University of British Columbia, Vancouver,
British Columbia, Canada
John C. Chen Department of Chemical Engineering, Lehigh University, Bethlehem, Pennsylvania, U.S.A.
Ye-Mon Chen Shell Global Solutions US, Houston, Texas, U.S.A.
Shiao-Hung Chiang Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh,
Pennsylvania, U.S.A.
Sakamon Devahastin Department of Food Engineering, King Mongkut’s 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 Department of Applied Physics and Electronics, Umea University, Umea, Sweden
John R. Grace Department of Chemical and Biological Engineering, University of British Columbia, Vancouver,
British Columbia, Canada
Ernst-Ulrich Hartge Technical University Hamburg-Harburg, Hamburg, Germany
Daxin He Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh,
Pennsylvania, U.S.A.
Thomas C. Ho Department of Chemical Engineering, Lamar University, Beaumont, Texas, U.S.A.
Masayuki Horio Department of Chemical Engineering, Tokyo University of Agriculture and Technology, Tokyo,
Japan
Behzad Jazayeri Fluor Daniel, Inc., Aliso Viejo, California, U.S.A.
viii Contributors

Peijun Jiang Department of Chemical Engineering, The Ohio State University, Columbus, Ohio, U.S.A.
Yong Jin Department of Chemical Engineering, Tsinghua University, Beijing, People’s Republic of China
S. B. Reddy Karri Particulate Solid Research, Inc., Chicago, Illinois, U.S.A.
George E. Klinzing Office of the Provost, University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.
T. M. Knowlton Particulate Solid Research, Inc., Chicago, Illinois, U.S.A.
Rafal P. Kobylecki* Department of Chemical Engineering, Tokyo University of Agriculture and Technology,
Tokyo, Japan.
Paul R. Mort Procter & Gamble, Cincinnati, Ohio, U.S.A.
Arun S. Mujumdar Department of Mechanical Engineering, National University o f Singapore, Singapore
Richard A. Newby Science & Technology Center, Siemens Westinghouse Power Corporation, Pittsburgh,
Pennsylvania, U.S.A.
Jens Reppenhagen^ 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 Department of Chemical Engineering, Tokyo University of Agriculture and Technology,
Tokyo, Japan
Yao Wang Department of Chemical Engineering, Tsinghua University, Beijing, People’s Republic of China
Fei Wei Department of Chemical Engineering, Tsinghua University, Beijing, People’s Republic of China
Joachim Werther Technical University Hamburg-Harburg, Hamburg, Germany
Guoqiang Yang Department of Chemical Engineering, The Ohio State University, Columbus, Ohio, U.S.A.
Wen-Ching Yang Science & Technology Center, Siemens Westinghouse Power Corporation, Pittsburgh,
Pennsylvania, U.S.A.
J. G. Yates Department of Chemical Engineering, University College London, London, United Kingdom

* Current affiliation: Energy Engineering Department, Czestochowa Technical University, Czestochowa, Poland.
' Current affiliation: BMH Claudius Peters, Buxtehude, Germany.
 HANDBOOK of
FLCIIDIZfiTION and
FLUID-PflRTICLE SYSTEMS
1

Particle Characterization and Dynamics

Wen-Ching Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.

1 INTRODUCTION 1.1.1 Definitions of Particle Size

The particle size is one or more linear dimensions
Particle characterization is important in all aspects of
appropriately defined to characterize an individual
particle production, manufacturing, handling, proces­
particle. For example, an ideal particle like a sphere
sing, and applications. Characterization of particles is
is uniquely characterized by its diameter. Particles of
the first necessary task required in a process involving
regular shapes other than spherical can usually be
solid particles. The required characterization includes
characterized by two or three dimensions. Cubes can
not only the intrinsic static parameters (such as size,
be uniquely defined by a single dimension, while
density, shape, and morphology) but also their
cuboids require all three dimensions, length, width,
dynamic behavior in relation to fluid flow (such as
and height. Two dimensions are required for regular
drag coefficient and terminal velocity). In this chapter,
isotropic particles such as cylinders, spheroids, and
the characterization of single particles with different
cones.
available techniques is first introduced. The dynamic
Irregular particles of practical interest, most often,
behavior of a single particle in the flow field at Stokes
cannot be uniquely defined. Their sizes are usually
flow regime, the intermediate flow regime, and the
defined based on certain reference properties. The
Newton’s law regime is then discussed. The chapter is
choice of any particular diameter for characterization
concluded with coverage of multiparticle systems.
of an irregular particle depends, in many cases, on the
intended application. Unfortunately, in most cases, the
1.1 Characterization of Single Particles correct choice of a representative diameter is uncertain.
Many diameters have been defined to characterize the
The complete characterization of a single particle irregular particles. The more common ones are sum­
requires the measurement and definition of the particle marized below.
characteristics such as size, density, shape, and surface
morphology. Because the particles of interest are Volume Diameter
usually irregular in shape and different in surface The volume diameter, dv, is defined as the diameter of
morphology, there are many different ways and tech­ a sphere having the same volume as the particle and
niques to characterize the particles. Depending on the can be expressed mathematically as
methods employed, the results may not be completely
/ 6 V \ 1/3
consistent. Some methods may be more appropriate dyj = I — - I where Vp = volume o f
than others for certain selected applications. V 71 / the particle (1)

1
2 Yang

Surface Diameter Projected Area Diameter

The surface diameter, ds, is defined as the diameter of a The projected area diameter, d&9 *s defined as the
sphere having the same surface area of the particle. diameter of a sphere having the same projected area
Mathematically it can be shown to be as the particle viewed in a direction perpendicular to
the plane of the greatest stability of the particle (see
1/2
Fig. 1).
where surface area of
the particle (2)
Feret Diameter
Surface-Volume Diameter The Feret diameter, dF, is a statistical diameter repre­
The surface-volume diameter, dsv, also known as the senting the mean value of the distances between pairs
Sauter diameter, is defined as the diameter of a sphere of parallel tangents to a projected outline of the p ar­
having the same external-surface-area-to-volume ratio ticle, as shown in Fig. 1. The Feret diameter is usually
as the particle. This can be expressed as used in particle characterization employing the optical
imaging technique to be discussed later.
A _ ^ P _ d
l (3)
sv ~ S ~ d l
Martin Diameter
Sieve Diameter The M artin Diameter, dM, is also a statistical diameter
The sieve diameter, dA, is defined as the width of the defined as the mean chord length of the projected out­
minimum square aperture in the sieve screen through line of the particle, which appropriately bisects the area
which the particle will pass. The sieve diameter will be of the projected profile, as shown in Fig. 1. Like the
discussed in more detail when particle size analysis by Feret diameter, the M artin diameter is also employed
sieving is discussed later in the chapter. most often during particle characterization using the
optical imaging technique to be discussed later.
Stokes Diameter Only four of the above particle size definitions are o f
The Stokes diameter, dst, is the free-falling diameter of general interest for applications in packed beds and
the particle in the Stokes law region and can be calcu­ fluidized beds. They are the sieve diameter, dA, the
lated from volume diameter, dv, the surface diameter, ds, and
the surface-volume diameter, dsv. The most relevant
terminal diameter for application in a fluidized bed is the sur­
/ 18pC/t velocity face-volume diameter, dsw. For applications in cataly­
dct — where Ut = (4)
(pP - P fk of the tic reactors with different isometric catalyst shapes,
particle Rase (1990) suggested that we use the equivalent dia­
meters summarized in Table 1.
Free-Falling Diameter
The free-falling diameter, df, is the diameter of a sphere
having the same density and the same free-falling velo­
city (or terminal velocity) as the particle in a fluid of
same density and viscosity. In the Stokes law region,
the free-falling diameter is the Stokes diameter defined
earlier.

Drag Diameter
The drag diameter, J D, is defined as the diameter of a
sphere having the same resistance to m otion as the
particle in a fluid 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 Figure 1 Illustration for projected area diameter, Feret dia­
the particle. meter, and Martin diameter.
Particle Characterization and Dynamics 3

Table 1 Suggested Equivalent Particle Diameters for Catalysts in Catalytic

Reactor Applications

Shape Equivalent particle diameter, dp = dsv = 6/as

Sphere dp = diameter of a sphere

Cylinder with length (/y) dp = dy, the diameter of a cylinder

equal to diameter (dy)

Cylinder, dy ^ ly a ^y
p ~ 4 + 2dy/ly

Ring with outside diameter of dQ, dp = 1.5(d0 —dx)

and inside diameter of dx

Mixed sizes , 1
p E(*//4»-)
i
Irregular shapes with <$>= 0.5 to 0.7 dp ==: (pdyj dy ~ d

Source: Rase (1990).

1.1.2 Definitions of Particle Shape For a true sphere, the sphericity is thus equal to 1. For
nonspherical particles, the sphericity is always less than
N atural and man-made solid particles occur in almost
1. The drawback of the sphericity is that it is difficult to
any imaginable shape, and most particles of practical
obtain the surface area of an irregular particle and thus
interest are irregular in shape. A variety of empirical
it is difficult to determine c|) directly. Usually the more
factors have been proposed to describe nonspherical
the aspect ratio departs from unity, the lower the
shapes of particles. These empirical descriptions of
sphericity. The sphericity, first introduced as a measure
particle shape are usually provided by identifying two
of particle shape, was subsequently claimed to be use­
characteristic param eters from the following four: (1)
ful for correlating drag coefficient (Wadell, 1934).
volume of the particle, (2) surface area of the particle,
There is some theoretical justification for the use of
(3) projected area of the particle, and (4) projected
sphericity as a correlating param eter for creeping
perimeter of the particle. The projected area and
flow past bodies whose geometric proportions resem­
perimeter must also be determined norm al to some
ble a sphere. But for other circumstances its use is
specified axis. For axisymmetric bodies, the reference
purely empirical (Clift et al., 1978). Leva (1959) and
direction is usually taken to be parallel or normal to
Subram anian and Arunachalam (1980) suggested
the axis of symmetry.
experimental methods using the Ergun equation for
All proposed shape factors to date are open to criti­
evaluation of the sphericity. This methodology will
cism, because a range of bodies with different shapes
be discussed when Ergun equation is introduced in
may have the same shape factor. This is really inevit­
C hapter 2, “ Flow through fixed beds.” For regularly
able if complex shapes are to be described only by a
shaped solids, the sphericities can be calculated from
single parameter. Thus in selecting a particular shape
Eq. (5), and they are presented in Table 2a. As for
factor for application, care must be taken to assure its
commonly occurring nonspherical particles, their
relevance.
sphericities are summarized in Table 2b.
Sphericity
Circularity
Wadell (1933) proposed the “degree of true sphericity”
Wadell (1933) also introduced the “ degree of circular­
be defined as
ity” , defined as
^ Surface area of volume —equivalent sphere
Circumference of circle having same cross-
Surface area of particle sectional area as the particle
Actual perimeter of the cross-section
(6 )
 Yang

Table 2a Sphericities of Regularly Shaped Solids

Shape Relative Proportions (j) dsv/ dy

Spheroid 1 1:2 0.93

1 2:2 0.92
1 1:4 0.78
1 4:4 0.70
Ellipsoid 1 2:4 0.79
Cylinder Height = diameter 0.87
Height = 2 x diameter 0.83
Height = 4 x diameter 0.73
Height = ;\ x diameter 0.83
Height = I\ diameter 0.69
Rectangular parallelpiped 1 1:1 0.81
1 1:2 0.77
1 2:2 0.77
1 1:4 0.68
1 4:4 0.64
1 2:4 0.68
Rectangular tetrahedron 0.67
Regular octahedron 0.83

Source: Adapted from G eldart (1986).

Unlike the sphericity, the circularity can be determined dependence of flow behavior on particle orientation
more easily experimentally from microscopic or photo­ (Cliff et al., 1978).
graphic observation. For an axisymmetric particle
projected parallel to its axis, is equal to unity. Use
Operational Sphericity and Circularity
of <f_ is only justified on empirical grounds, but it has
Since the sphericity and circularity are so difficult to
the potential advantage of allowing correlation of the
determine for irregular particles, Wadell (1933) p ro ­
posed that c|) and <f_ be approxim ated by “operational
sphericity and circularity:”
Table 2b Sphericities of Commonly
Occurring Nonspherical Particles V 1/3
Volume of particle J
Material Sphericity Volume of the smallest circumscribing I
sphere /
Sand
Round sand 0.86 (7)
Sharp sand 0.66
v 1/2
Crushed sandstone 0.8-0.9 Projected area of particle |
Coal
Area of the smallest circumscribing I
Pulverized coal 0.73
circle /
Crushed coal 0.63-0.75
Activated carbon 0.70-0.90 (8)
Mica flakes 0.28
Fischer-Tropsch catalyst 0.58 F or the ellipsoids, the operational sphericity, c|)op, can
Common salt 0.84 be expressed as
Crushed glass 0.65
Silica gels 0.70-0.90
<l>op - (e \ e i T 1/3 (9)
Tungsten powder 0.89
Sillimanite 0.75
F or the rounded particles, it can be approxim ated by
Wheat 0.85
Eq. (9). The ex and e2 are called the flatness ratio and
Source: A dapted from G eldart (1986). the elongation ratio, respectively, and are defined as
Particle Characterization and Dynamics 5

terizing particle shape are under development. For a

e] = - Flatness ratio ( 10)
t review, see Kaye (1973).
Heywood (1962) suggested that k may be estimated
e2 = T Elongation ratio ( 11) from the corresponding value, fce, of an isometric par­
ticle of similar form by
where t = thickness, the minimum distance between
two parallel planes tangential to opposite surfaces.
k =- (14)
One of the two planes is the plane of the maximum
stability, b = breadth, the minimum distance between
two parallel planes that are perpendicular to the planes Values of k Qfor some regular shapes and approximate
defining the thickness and tangential to opposite sur­ values for irregular shapes are given in Table 3.
faces. / = length, projected on a plane norm al to the Equation (14) is exact for regular shapes such as
planes defining t and b. spheroids and cylinders. Heywood suggested that k
The three characteristic dimensions have an increas­ be employed to correlate drag and terminal velocity,
ing order of magnitude t < b < 1. using da and the projected area to define Re and CD,
However, (j)op is not generally a good approxim a­ respectively. There are justifications for this approach
tion to (j). Aschenbrenner (1956) showed that a better because many natural particles have an oblate shape,
approxim ation to is given by a “working sphericity,” with one dimension much smaller than the other two.
<t>w, obtained from the flatness and elongation ratios: Over a large Reynolds num ber range in the “intermedi­
ate” regime, such particles present their maximum area
1/3
12.8(eie|) to the direction of motion, and this is the area char­
<1> W =
(12) acterized by da. There is also evidence that the shape of
1 + e2(l + ex) + 6^1 + £ 2(1 + ^ i) this projected area, which does not influence k, has
little effect on drag (Clift et al., 1978).
The working sphericity has been found to correlate
A more m odern approach uses fractal analysis, or
well with the settling behavior of naturally occurring
Fourier transform ation. The latter is a bit involved,
mineral particles.
requiring several coefficients for complex definition.
Wadell (1935) suggested that ^ op provides an esti­
mate of $ based on a two-dimensional projection of a
particle, and thus is sometimes called the “projection Table 3 Approximate Values of ke and k
sphericity.” The operation circularity, however, does
not approximate § for regular bodies and has virtually ke for isometric irregular shapes
no correlation with settling behavior of natural irregu­ Rounded 0.56
Subangular 0.51
lar particles (Clift et al., 1978).
Augular
tending to a prismoidal 0.47
The Heywood Shape Factor tending to a tetrahedron 0.38
The Heywood shape factor is sometimes called the ; for selected natural particles
volumetric shape factor. Heywood (1962) proposed a Sand 0.26
widely used empirical param eter based on the pro­ Bituminous coal 0.23
jected profile of a particle as follows: Limestone 0.16
Gypsum 0.13
Talc 0.16
k - ^ (13) for regular shape
£
Sphere 0.524
where da — y/ 4 A p/n. Cube 0.696
The projected area diameter, <7a, is the diameter of Tetrahedron 0.328
the sphere with the same projected area as the particle. Cylinder with an aspect ratio of 1
viewing along axis 0.785
A number of methods have been suggested for obtain­
viewing normal to axis 0.547
ing an estimate for da. Even if da is available, the
Spheroids
Heywood shape factor can only be evaluated if Vp is with an aspect ratio of 0.5 0.262
known. For naturally occurring particles or if a distri­ with an aspect ratio of 2 0.370
bution of particle sizes or shapes is present, Vp may not
be readily available. Autom atic techniques for charac­ Source: Heywood (1962).
6 Yang

1.1.3 Definitions of Particle Density W hen a porous particle is broken into smaller
pieces, the particle density of the smaller pieces will
There are several particle density definitions available.
usually be larger than the original particle density by
Depending on the application, one definition may
virtue of the elimination of some pores. W hen the
be more suitable than the others. For nonporous
particle size becomes finer, the particle density will
particles, the definition of particle density is straight­
approach that of a nonporous particle because all the
forward, i.e., the mass of the particle, M , divided by
pores will be completely eliminated. In the process o f
the volume of the particle, Fp, as shown in Eq. (15).
handling porous particles, this tendency should be kept
Mp mass of the particle in mind, and the particle density should be carefully
Pp = TFp
T = volume that the particle would displace evaluated.
if its surface were nonporous Knight et al. (1980) described a method of deter­
(15) mining the density of finely divided but porous p ar­
ticles such as fluid bed cracking catalyst. They
For porous particles with small pores, the particle suggested to set a known mass and measurable
volume in Eq. (15) should be replaced with the volume of powder in resin and then measure the voi-
envelope volume of the particle as if the particles dage of the sample sections of the resin. The results
were nonporous as shown in Fig. 2. This would be were satisfactory, though the method was quite
more hydrodynamically correct if the particle behavior tedious. Buczek and Geldart (1986) proposed to use
in the flow field is of interest or if the bulk volume of very find dense powders as the pycnometric fluid to
the particles is to be estimated. For total weight esti­ determine the density of porous particles. F or a pow­
mation, then the skeleton density should be known. der to be a suitable pycnometric fluid, it should be
The skeleton density is defined as the mass of the p ar­ nonporous and free-flowing. Its density should be
ticle divided by the skeletal volume of the particle. In much larger than the porous particles to be measured,
practice, the pore volume rather than the skeletal and its diameter should be much smaller than the
volume is measured through gas adsorption, gas or porous particles and greater than the biggest pores
water displacement, and mercury porosimetry. These of the porous particles. It should also be nonreactive
techniques will be discussed in more detail later. There toward the porous particles. Akapo et al. (1989) also
are also porous particles with open and closed pores. proposed a novel method of determining particle den­
The closed pores are not accessible to the gas, water or sity of porous aeratable powders. The m ethod
mercury and thus their volume cannot be measured. In depends on measurement of bed expansion of a gas
this case, the calculated skeleton density would include fluidized bed of the powder in the region between the
the volume of closed pores as shown in Fig. 2. For minimum fluidization and the minimum bubbling.
nonporous particles, the particle density is exactly The Richardson and Zaki bed expansion equation
equal to the skeleton density. For porous particles, was then applied to back out the particle density.
the skeleton density will be larger than the particle The Richardson and Zaki bed expansion equation
density. and the minimum fluidization and minimum bubbling
regimes will be discussed in Chapter 3, “Bubbling
Fluidized Beds.”
Hydrodynamic envelope
and particle volume V p

2 PARTICLE CHARACTERIZATION
TECHNIQUES

There are many techniques that can be employed to

Open pores
characterize particles, some simple and primitive and
some complicated and sophisticated. Almost every
technique is associated with intrinsic experimental
Closed pores
errors and implicit assumptions. Thus care must be
' exercised to select proper techniques for your specific
applications. The available techniques are reviewed in
Figure 2 Skeleton particle density. this section.
Particle Characterization and Dynamics 7

2.1 Methods for Direct Characterization of Particle mended. Detailed discussion of sieve analysis techni­
Size and Shape ques can be found in standard textbook such as Allen
(1975) and Kaye (1981).
2.1.1 Sieve Analysis
2.1.2 Imaging Technique
The most commonly used m ethod for classifying
powders is to sieve the particles through a series of Direct measurement of particle dimensions is also pos­
screens with standardized mesh size by sifting, swirling, sible from enlarged photographic or electronic images
shaking, or vibrating. Two standard mesh sizes, U.S. of microscopes. There are three types of microscopes
sieve size and Tyler sieve size, are usually used in the commonly employed, i.e., the optical microscope, the
U.S. In the European practice, the British Standard scanning electron microscope (SEM), and the trans­
and Germ an D IN sieve sizes are also employed. The mission electron microscope (TEM). The optical
mesh number of a sieve refers to the num ber of parallel microscope is employed for particles from 1 pm to
wires per inch in the weave of the screen. The about 150 pm. Both SEM and TEM make use of elec­
American Society of Testing M aterials (ASTM) speci­ tron beams and can be used for particles from 5 pm
fications for standard mesh sizes for those systems are down to as small as 0.01 pm. They are especially useful
compared in Table 4. The mesh size was designed so for revealing the surface m orphology of extremely
that the aperture or opening of the alternating mem­ small particles.
bers in the series has a factor of 21/2. For example, the Particles to be imaged in an optical microscope are
U.S. 12 mesh with an opening of 1.68 mm is 21/2 times usually dispersed in a drop of viscous fluid on a glass
the U.S. 16 mesh with an opening of 1.19 mm. The slide. Their images are then visually compared with a
result of the sieve size analysis is commonly plotted set of standard circles, geometric shapes, or linear grids
in a logarithmic-scale graph expressing the cumulative to derive their actual sizes and shapes. The M artin and
weight percentage under size as the abscissa and the Feret diameters are also often used to characterize a
particle size as the ordinate, as shown in Fig. 3. There, particle. The M artin diameter is defined as the magni­
the 21/2 factor in the mesh size arrangement allows the tude of the chord that divides the image into two equal
size data points to be almost equally spaced on the areas with respect to a fixed direction (see Fig. 1). The
logarithmic scale. Also, experimentally, many crushed Feret diameter of a particle image is defined as the
materials yield a straight line if plotted as in Fig. 3. length of the image as projected with respect to a
The inaccuracies and uncertainties of sieve analysis specific reference direction (see Fig. 1). Both M artin
stem from the discrete steps of the mesh size arranged and Feret diameters are intended to be statistical dia­
at an approximate factor of 21/2 between successive meters, i.e., after characterization of many images.
mesh sizes. Sieve analysis does not provide the infor­ Thus any slight departure from the true randomness
mation for the largest and the smallest particle sizes. in the field of view of particle images can produce bias
The size cut provides an approximate value for the in the particle size characterization. M odern instru­
mean particle size within the cut. Sieve analysis also ments couple television cameras interfaced with com­
does not differentiate the particle shape. A needle- puters and sophisticated software can speed up the
shaped particle can either pass through a mesh or be imaging analysis considerably.
retained on the screen, depending on its orientation M artin’s statistical diameter is also referred to as the
during sifting. The result of sieve analysis is also mean linear intercept or mean chord and has been
dependent on the time of sieving action, the particle shown to relate to the specific surface in the following
loading on the sieve, and sieve blinding (also called m anner (Herdan, 1960):
pegging). Enlargement of aperture due to wire erosion 4
of a sieve can cause discrepancy as well. For small M artin diameter = ——- (16)
particles, agglomeration due to static electricity or P^v
moisture can also occur. where p = the density of packing and S v = particle
The smallest mesh size for the Tyler Series is 400 surface area per unit volume of particle.
mesh, equivalent to a 38 pm opening, while the smallest Based on the experimental evidence, on the average,
mesh size for the U.S. Series is 635 mesh, equivalent to the M artin diameter is usually smaller than the mean
a 20 pm opening. For particles finer than 20 pm, the projected diameter and the Feret diameter larger. Since
surface and electrostatic forces become im portant, and both M artin and Feret diameters depend on the
particle classification by sieve analysis is not recom­ particle shape, the ratio of Feret diameter to M artin
8 Yang

Table 4 Summary of Various Types of Standard Sieve

U.S. Standard Tyler British Standard German Standard

mesh No. opening, mm mesh No. mesh No. opening, mm DIN No. opening, mi

3- 5.66 3I — — 1 6.000
2
4 4.76 4 — — — —
5 4.00 5 — — — —
6 3.36 6 5 3.353 2 3.000
7 2.83 7 6 2.812 — —

8 2.38 8 7 2.411 2- 2.400

10 2.00 9 8 2.057 3 2.000
12 1.68 10 10 1.676 4 1.500

14 1.41 12 12 1.405 — —

16 1.19 14 14 1.204 5 1.200

18 1.00 16 16 1.003 6 1.020

20 0.84 20 18 0.853 — —
— — — _ — 8 0.750
25 0.71 24 22 0.699 — —

30 0.59 28 25 0.599 10 0.600

— — — — — 11 0.540
35 0.50 32 30 0.500 12 0.490
40 0.42 35 36 0.422 14 0.430
45 0.35 42 44 0.353 16 0.385

50 0.297 48 52 0.295 20 0.300

60 0.250 60 60 0.251 24 0.250
70 0.210 65 72 0.211 30 0.200
80 0.177 80 85 0.178 — —
100 0.149 100 100 0.152 40 0.150

120 0.125 115 120 0.124 50 0.120

140 0.105 150 150 0.104 60 0.102
170 0.088 170 170 0.089 70 0.088

200 0.074 200 200 0.076 80 0.075

230 0.062 250 240 0.066 100 0.060
270 0.053 270 300 0.053 — —
325 0.044 325 — — — —

400 0.038 400 — — — —

635 0.020 _ — — — —

diameter is usually fairly constant for the same m ate­ between 300 to 500, needs to be measured to minimize
rial. For Portland cement, it is about 1.2; for ground statistical error.
quartz and ground glass, approximately 1.3 (Herdan,
1960).
2.1.3 Gravity and Centrifugal Sedimentation
The microscopic measurement technique is most
suitable for particles relatively uniform in size and The falling speeds of particles in a viscous fluid under
granular in shape, because a large number of particles, the influence of gravity are used to measure the particle
Particle Characterization and Dynamics 9

Particle Size Distribution

m
m
°c
m
m
§
m
m
•©
C
m
m
m
3
JC
m
m
2

Cumulative Weight % Undersize

Figure 3 Graphic representation of sieve size analysis—logarithmic-scale graph expressing cumulative % oversize or undersize.

size in the gravity sedimentation technique. The m ea­

VOQ = Vm{ \ + 2 A d^ j (17)
sured 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 irre­ According to Eq. (17), even if the column is 50 times
gularly shaped particles fall with different orientations the diameter of the particle, there is still a 5% reduc­
in the vertical direction and thus different settling velo­ tion in the falling speed of the particle.
cities, similar irregularly shaped particles can have a Two basic suspension systems, the line start system
range of Stoke diameters. Disturbances caused by the (or the two-layer sedimentation system) and the initi­
presence of other particles, the concentration effect, ally homogeneous system are employed in the gravity
can also be im portant. Recommendations for the sedimentation technique. In the line start system, a thin
upper limit particle concentration to be used during layer of particles is placed at the top of the sedimenta­
the sedimentation analysis varies from 0.01 to 3.0% tion column, and its settling behavior is analyzed by
(Kaye, 1981). An upper limit of 0.05% was recom­ different techniques. In the initially homogeneous sys­
mended by Kaye, and in certain practical cases up to tem, the column is homogenized first and its settling
0.2% by volume is permissible. The probability of pattern is subsequently studied.
forming clusters increases with particle suspension Classical techniques for measuring the sedimenta­
concentration. The clusters tend to fall at a higher tion behavior include taking samples with a pipette,
speed and thus introduce measurement error. measurement of height of sediment layer at the bot­
The hindering effect of the containing wall on the tom, and use of balance pan to measure the weight of
falling speed of the particles cannot be ignored either. settled particles. M odern sedimentometers make use of
F or a spherical particle, the effect can be expressed by the diffraction pattern of a light beam, the power loss
the Landenburg equation as of an x-ray, or a Doppler shift of a laser beam. The
10 Yang

modern techniques are primarily applied to m onitor As in sedimentation, the concentration of the p arti­
the sedimentation kinetics of an initially homogeneous cles in the elutriators affects the results of measure­
suspension. The photosedimentometers measure the ment. Also the velocity in the elutriators tends to be
beam with a small-angle detector. This application is parabolic rather than uniform and thus introduces
based on the Lam bert-Beer law, which is errors in measurement. Centrifugal force can also be
applied like that in the sedimentation to enhance the
Ic = I(>e~aCL ( 18) performance. Different designs, including horizontal
where IQ = the original light beam intensity, / c = the elutriators, are available.
intensity of a light beam after passing through a p ar­
ticle suspension of concentration C, L = the length of 2.1.5 Cascade Im paction Technique
the light beam path, and a = an empirical constant The cascade impaction technique is based on similar
depending on equipment and particle characteristics. principles employed in the elutriation technique and is
In the x-ray sedimentometers, the x-ray absorption also based on the inertia of the particles. The particles
of a sedimentation suspension is used to measure the with terminal velocities smaller than the flow velocity
concentration gradient in the suspension. The law out­ will be carried along by the gas flow. In addition,
lined by Olivier et al. (1970/1971) is employed: smaller particles will tend to follow the streamlines o f
\n T = - $ W s (19) the flow better than the larger particles owing to their
smaller inertia. When the flow changes direction
where T = the measured transmittance, X c/ X 0, X 0 = because of the presence of a plane surface, larger
the intensity of the x-ray beam passing through the particles with larger inertia will impact on the plane
sedimentation column filled with clear fluid, X c = the surface and be collected. By successive decreases in
intensity of the x-ray beam passing through a particle flow velocity from stage to stage, the particles can be
suspension of concentration C, Ws = the weight frac­ collected and classified into different particle size
tion of particles in the x-ray pass, and (3 = an empirical fractions. Earlier designs employed slides coated with
constant depending on equipment, particle, and sus­ adhesive to collect particles for microscopic analysis.
pending fluid characteristics. M odern devices have many more variations. The cas­
Alternatively, centrifugal force instead of gravita­ cade impaction technique is usually applied for particle
tional force can be created to enhance the sedimenta­ sizes between 0.1 and 100 pm.
tion performance. Depending on the size of the
centrifugal arm and the speed of rotation, many 2.1.6 Resistivity and Optical Zone Sensing
times the gravitational force can be applied. Techniques
The modern nonintrusive sedimentometers are effi­
The resistivity and optical zone sensing techniques
cient, can be adapted easily for hostile environments,
measure the particle size by measuring the changes o f
and can provide inform ation in situ and thus are ideal
resistivity or optical properties when the particles are
for quality control in m anufacturing processes. The
passed through the sensing zone of the instruments.
resolution and sensitivity allow measurement of
The well-known Coulter counter is a resistivity zone
Stokes diameters down to about 0.5 pm.
sensing instrum ent (see Fig. 4). The particles to be
analyzed are first suspended and homogenized in an
2.1.4 Characterization by Elutriation
electrolyte and then are forced to pass through a
Particle size characterization by elutriation makes use cylindrical orifice placed between two electrodes. The
of the same kind of principles employed by the sedi­ passages of particles through the orifice generates vol­
mentation. Instead of letting particles settle with grav­ tage pulses that are amplified, recorded, and analyzed
ity, the particles are actually carried out against gravity to produce a particle size distribution. The instrum ent
during elutriation. In vertical elutriators, the particles is usually calibrated with standard particles such as
with terminal velocities less than the vertical fluid velo­ latex spheres of known size. The data are analyzed
city will be elutriated out. By operating the elutriator by assuming that the sensing zone is isotropic (i.e.,
at different flow conditions, the particle size distribu­ the exact location of the particles in the sensing zone
tion of the sample can be calculated. The flow in the is unim portant) and the pulse height is proportional to
elutriators is usually lam inar flow, and the Stokes the volume of the particle. The results from this ana­
equation is used to estimate the Stokes diameter of lysis are thus the equivalent spherical diameters of the
the particle by assuming that the particle is spherical. particles. F or the simple voltage-pulse-to-particle-
Particle Characterization and Dynamics 11

To Vacuum Pump used to illuminate the particle. Generally, there are two
basic designs: the scattered light can either be collected
in a narrow forward direction in the direction of the
illuminating light or be collected in a wide angular
range. The forward scattering systems are more suit­
able for sizing particles with light-absorbing proper­
ties, since for these particles there exists a monotonic
relationship between the intensity and the size. For
nonabsorbing particles, however, multiple values exist
for particles larger than 1 pm. Thus most of the exist­
ing particle-sizing instrum ents make use of the second
scattering geometry to collect the scattered light into a
large angular range oriented either perpendicularly or
axially with respect to the light beam direction, see Fig.
Sample Beaker
5. The instrum ent is usually calibrated with nonab­
Figure 4 Electrical zone sensing (Coulter counter). sorbing spherical polystyrene latex particles, and the
scattered light intensity is a m onotonic function of
the particle size. However, the presence of absorptivity
volume relationship to hold, the particle size to be can reduce substantially the sizing sensitivity of the
analyzed needs to be less than about 40% of the orifice instrument.
diameter. Particles smaller than 3% of the orifice The light scattering theories employed for this tech­
diameter do not produce a reliable result. Thus for a nique are by Mie, Rayleigh, and Fraunhofer (Bohren
sample of wide size distribution, changes of orifices of and Huffman, 1986). When the dimensions of a
different diameters are usually necessary. particle are of the same order of magnitude of the
Possible errors during measurement arise from the wavelength of the incident light, Mie theory is used.
possibility of more than one single particle occupying When the particle is much smaller than the wavelength
the sensing zone at the same time. This can easily be of the light, the appropriate light-scattering theory is
solved by diluting the suspension or making multiple that of Rayleigh. F or particles much larger than the
measurements with increasingly dilute suspensions. If wavelength of the light, Fraunhofer diffraction is
the particles substantially deviate from the spherical employed. F or more detailed discussions of principles
shape, the equivalent spherical diameters obtained
with this technique may not have any physical signifi­
cance. Another potential error stems from the fact that
particles may not all pass through the axis of the ori­
fice. Similar size particles will generate different voltage Solid Particles
t\
signals depending on whether they pass through the
orifice at the axis or close to the cylindrical wall. - ......... ^ .. - ........

M odern instruments are increasingly capable of editing Laser Light

the signals to reject those stray signals. The electrical Lens
zone sensing technique can be applied for particles Focal
ranging from 0.6 to 1200 pm. Plane
The optical sensing technique measures the scat­ Measuring Cell
tered light from a particle passing through a sensing
volume illuminated by a light source, such as a white
light or a laser. The intensity of the scattered light is
then related to the size of the particle. In an ideal
situation, a m onotonic relationship exists between the
intensity of the scattered light and the particle size and
thus allows unique determination of the size of the
particle. In reality, the light-scattering properties of a
particle depend in a very complex way on its refractive
index and shape, and on the wavelength of the light Figure 5 Laser-diffraction spectrometry.
12 Yang

behind this application, see Fan and Zhu (1998). An cially available models based on the electrical sensing
innovative instrum entation using three lasers to pro­ zone principle, eight commercially available models
duce scattered light through an angular range from based on the laser diffraction and scattering principle,
close to 0° up to 169° in one continuous pattern was three commercially available models based on the x-
described by Freud et al. (1993). Bonin and Queiroz ray sedimentation principle, seven commercially avail­
(1991) also described an application in an industrial- able models based on the photosedim entation princi­
scale pulverized coal-fired boiler measuring local p ar­ ple, and one commercially available model based on
ticle velocity, size, and concentration. the light attenuation principle were employed in a
Besides for particle sizing, the optical sensing tech­ round-robin evaluation. They found that the effect o f
nique can also be employed for measurement of the anisotropic particles such as flaky and rodlike particles
particle num ber concentration. Discussions mentioned on the particle size measured by the commercial equip­
earlier regarding the measurement errors for the resis­ m ent is much larger than that of the isotropic block­
tivity sensing zone technique apply here as well. When shaped particles. In the x-ray sedimentation technique,
more than one particle is present in the sensing zone, the equivalent volume diameter is usually measured.
error can occur. The effect of particle shape is thus small, because the
transm ittance intensity of x-rays is independent of the
2.1.7 Techniques for Particle Surface particle shape and its orientation. Similarly, the effect
Characterization of particle shape was also found to be small when the
light attenuation technique was employed, because the
In many applications related to chemical reactions, the particles were dispersed at random in the agitated sam­
total surface area is more im portant than the particle ple suspension. The laser diffraction and scattering
size and shape. To measure the particle surface area, technique, on the contrary, resulted in a large range
two commonly used techniques are the permeability of size distributions, because the particles tended to
technique and the gas adsorption technique. The per­ orient along the shear flow, especially the anisotropic
meability technique is a method for the determination particles. The effect of particle shape also produced a
of the power surface area by measuring the permeabil­ large range of size distributions in the coarse-size range
ity of a powder bed. The permeability is defined in of the distribution in the photosedim entation techni­
Chapter 2 in relation to D arcy’s law. The derivation que owing to the turbulence of liquid created by the
of pressure drop equations through power beds is quite particle orientation, which resulted in intensity fluctua­
involved and will be discussed in C hapter 2, for mea­ tion of the transm itted light at the initial stage of grav­
suring the powder surface area the reader can refer to itational sedimentation.
Allen (1975). If the particle size distributions are characterized by
Another technique commonly used to measure the the average diameters at the cumulative masses of 10,
powder surface area and the pore size is the physical gas 50, and 90%, the data scattering measured by different
adsorption technique based on the well-known BET techniques can be represented by the diameter ratios at
(B runauer-E m m ett-T eller) method on monolayer the cumulative masses of 10, 50, and 90%, or D R 10,
coverage of adsorptives such as nitrogen, krypton, D R 50, and D R 90. Since the particle diameter measured
and argon. The application is very well established, by the electrical sensing zone technique is to be the
and detailed discussions are available in Allen (1975). equivalent volume diameter and is independent of the
particle shape, its particle diameter is defined to be one.
2.2 Effect of Particle Shape on Size Distribution The particle diameters measured by other techniques
Measured with Commercial Equipment are then ratioed with this diameter. The experimental
results of D R 10, D R 50, and D R 90 are summarized in
The effect of particle shape on particle size distribution Table 5. It can be seen that the results of particle
was investigated by Naito et al. (1998) with commer­ analysis from different techniques can be quite differ­
cial particle size analyzers based on five different ent.
measuring principles: electrical sensing zone, laser dif­ Because of differences in measuring techniques
fraction and scattering, x-ray sedimentation, photose­ based on different properties of irregular particles,
dimentation, and light attenuation. The particles used the particle size and size distribution obtained by dif­
are blocky aluminum oxide and barium titanate parti­ ferent methods from the same sample are usually
cles, flaky boron nitride particles, and rodlike silicon different. The size and distribution measured by sieve
whisker ceramic powders. Altogether, four commer­ analysis will be different from that by laser diffract­
Particle Characterization and Dynamics 13

Table 5 Experimental Results of D R 10, D R50, and D R90

Silicon
Aluminum Barium Boron nitride
Sensing technique oxide titanate nitride whisker

DRio
Electrical sensing zone 1.00 1.00 1.00 1.00
Laser diffraction and scattering 0.61 0.77 0.94 0.61
X-ray sedimentation 0.82 1.10 — 0.69
Photosedimentation 0.82 0.95 0.58 0.75
Light attenuation 1.00 1.15 0.96 0.96
D R 5o
Electrical sensing zone 1.00 1.00 1.00 1.00
Laser diffraction and scattering 0.97 1.11 1.32 1.41
X-ray sedimentation 0.84 1.05 — 0.86
Photosedimentation 0.78 0.96 0.74 1.05
Light attenuation 1.33 1.20 1.11 1.04
D R90
Electrical sensing zone 1.00 1.00 1.00 1.00
Laser diffraction and scattering 1.15 1.33 1.56 2.32
X-ray sedimentation 0.90 1.14 — 1.01
Photosedimentation 1.01 1.17 1.53 2.43
Light attenuation 1.20 1.46 1.78 0.98

Source: N aito et al. (1998).

ometer, x-ray sedimentation, and so on. Conversion The values of the H G I usually range from 15 to 140.
factors are available to convert particle size distribu­ The higher the H G I, the higher is the grindability of
tions obtained by one m ethod to that measured by the material. The H G I has been found to correlate
another technique. Austin (1998) suggested a technique with the attrition characteristics of the particles in
to perform this correction and provided an equation fluidized beds and in pneumatic transport lines
for conversion between laser diffractometer and x-ray (Davuluri and Knowlton, 1998). The H G I does not
sedigraph measurements. directly relate to hardness. For example, some m ateri­
The applicable particle size ranges for various p ar­ als such as plastics are difficult to grind.
ticle sizing techniques are summarized in Table 6.

2.3 Measurement of Mechanical Properties of 2.3.2 A ttrition Index

Particles F or application in fluidization and fluid-particle
systems, the attrition index is probably the most
2.3.1 Hardgrove Grindability Index (HGI) im portant particle characteristic. The particle attrition
The Hardgrove Grindability Index was originally can affect the entrainment and elutriation from a flui­
developed by Babcock and Wilcox to measure the rela­ dized bed and thus subsequently dictate the design of
tive ease of pulverizing coal. This ASTM D409-71 downstream equipment. The attrition in a pneumatic
method has since been employed for characterizing transport line can change the particle size distribution
other particles as well. The m ethod processes 50 g of of the feed material into a fluidized bed reactor and
air-dried particles screened to a size of 16 x 30 mesh thus alter the reaction kinetics. Davuluri and
(1180 x 600 pm) in a small ball-and-race mill for 60 Knowlton (1998) have developed standardized proce­
dures to evaluate the A ttrition Index employing two
revolutions. The am ount of m aterial (B^oo) passing
the 200 mesh (75 pm) screen is then measured. The techniques, solids impaction on a plate and the
result is then compared with a standard index to arrive Davison jet cup. The two test units used are shown
at the H G I or the H G I is calculated from the equation in Figs. 6 and 7. They found that these two test tech­
niques are versatile enough to be applicable for a wide
HGI = 13 + 6 .9 3 ^ 0 0 (20) range of materials, such as plastic, alumina, and lime­
14 Yang

Table 6 Applicable Particle Size Ranges for Various Particle Sizing Methods

Applicable particle size

Particle sizing methods range (pm) Measured dimension

Sieving
Dry >10 Sieve diameter
Wet 2-500
Microscopic examination
Optical 1.0-100 Length, projected area,
Electronic 0.01-500 statistical diameters
Zone sensing
Resistivity 0.6-1200 Volume
Optical 1.0-800
Elutriation
Laminar flow 3-75 Stokes diameter
Cyclone 8-50
Gravity sedimentation
Pipette and hydrometer 1-100
Photoextinction 0.5-100 Stokes diameter
X-ray 0.1-130
Centrifugal sedimentation
Mass accumulation 0.5-25
Photoextinction 0.05-100 Stokes diameter
X-ray 0.1-5
Centrifugal classification 0.5-50
Gas permeability 0.1-40
Gas adsorption 0.005-50
Cascade impaction 0.05-30

Source: Adapted from Pohl (1998) and Lloyd (1974).

stone. For a more recent review on attrition, see under controlled conditions. The weight loss of the
W erther and Reppenhagen (1999). coupon is an indication of the erosiveness of the p arti­
cular coal and the potential damage to the processing
2.3.3 Abrasiveness Index and handling equipment, and other boiler components.
In pulverized coal combustion, the abrasiveness of the
particles severely limits the life of the pulverizer grind­
3 FLUID DYNAMICS OF A SINGLE PARTICLE
ing elements. The Abrasiveness Index in this applica­
tion is usually determined by the Yancey-Geer Price
F or a particle moving in a fluid, the force acting on the
apparatus (Babcock and Wilcox, 1992). In this test,
surface of a particle depends only on the flow of the
four metal test coupons attached to a rotating shaft
fluid in its immediate vicinity. F or the simplest case, let
are rotated at 1440 rpm for a total of 12,000 revolu­
us consider a single particle moving at a velocity Ur
tions in contact with a sample of 6350 pm xO
relative to its immediate fluid around the particle. It is
(0.25 in. x 0) coal. The relative Abrasiveness Index is
also assumed that the fluid is newtonian and that the
then calculated from the weight loss of the coupons.
Ur is constant. The fluid dynamic parameters can then
Babcock and Wilcox has estimated the wear of full-
be evaluated as follows.
scale pulverizers on the basis of the Yancey-Geer
Price Index.
3.1 Definition of Particle Drag Coefficient
2.3.4 Erosiveness Index
The drag coefficient is defined as the ratio of the force
Babcock and Wilcox (1992) has also developed a on the particle and the fluid dynamic pressure caused
method of quantifying the erosiveness of coal by sub­ by the fluid times the area projected by the particles, as
jecting a steel coupon to a stream of pulverized coal shown in Eq. (21) and Fig. 8.
Particle Characterization and Dynamics 15

F = ^ C DPfUr2Ap

Figure 8 Definition of particle drag coefficient.

three different regimes distinguishable depending on

the magnitude of the particle’s Reynolds number.

Figure 6 Schematic of attrition apparatus employing the 3.1.1 The Stokes Regime
solids impaction principle.
24
CiD for (Re) < 0.2 (23)
(Re)p
Cr> — (21)
( l / 2 ) p f L r2T p The Stokes law regime is also commonly known as the
creeping flow regime. In this regime, the viscosity of
or
the fluid is dominating.
F = - C Dp{U fA p (22)
3.1.2 The Intermediate Regime
The drag coefficient, CD, is a function of particle’s
Reynolds number, (Re)p = UYdppf/p , only. There are CD = /|_(Re)pJ for 0.2 < (Re)p < 500 (24)

In the intermediate regime, the drag coefficient is a

function of the particle’s Reynold number.

3.1.3 The N ew ton’s Law Regime

CD = 0.44 for (Re)p > 500 (25)

■Air
Tangential Inlet Tube In the N ew ton’s law regime, CD is relatively constant,
(0.15 cm diameter)
and the force F is largely due to the inertia of the fluid
rather than to the viscosity of the fluid.

3.2. The Stokes Law Regime

T Cylindrical Only in the Stokes law regime, (Re)p < 0.2, have
E
o
in
Housing theoretical methods of evaluating CD met with much
success. The theoretical analysis starts with the viscous
1 •Air flow around a rigid sphere, which can be expressed as

Figure 7 Schematic of attrition apparatus employing

Davidson jet cup. Fs = 3tcjidp Ur (26)
16 Yang

Two-thirds of the drag force is due to the viscous shear 3.5 Corrections to the Stokes Approximation
stresses acting on the particle surface (skin friction).
The other one-third is due to the difference in pressure In 1910, Oseen provided the first correction term to the
on its surface (form drag). Stokes approxim ation as well as a mathematically self-
Equating Eqs. (22) and (26), we have consistent first approxim ation, as shown in Eq. (33):

1
Fs = 3npdpUv = - C Dpf t/r - n d p
9 1 9
(27) 24 3(Re)
Cr> — (33)
(Re)p 16
or
M ore recently, Proudm an and Pearson (1957) p ro ­
24 vided the key to a more straightforward calculation
C n = 24 (28)
dpUr Pj (Re) of further correction terms. The results for the drag
force on the sphere in accordance with Stokes’,
The linear relation between the Fs and UT is an im por­ Oseen’s, and Proudm an and Pearson’s calculations
tant property of the Stokes law. It enables considerable are given in Table 7. A very large and im portant col­
simplifications in many problems. lection of solutions of such problems can be found in
the book by Happel and Brenner (1965).
3.3 The Stokes Free Fall Velocity

Equating the drag force and the gravitational force for 3.6 Empirical Drag Coefficient Expression
a single spherical particle, we have
The preferred correlations to be used in different
nd„ ranges of particle Reynolds numbers were recom­
- C Dpf Ur dp _ — k (p p - pf)g (29)
mended by Clift et al. (1978) and are summarized in
Table 8. In 1986, Turton and Levenspiel proposed a
or
single correlation applicable for the complete range o f
4 d v (pp - p f ) g Reynolds numbers and considerably simplified the cal­
CD — - (30) culation of the single-particle drag coefficient.
pf V}
The Turton and Levenspiel (1986) equation is
or
24
Cr j l +0.173(R e)p657]
flp (p P - P f ) # (R ef
uT= ut \8p
(31) (34)
0.413
Equation (4) for the Stokes diameter can be derived 1 ■b 16.300(Re)p-1 .0 9
from Eq. (31).
H aider and Levenspiel (1989) subsequently improved
3.4 Wall Effect the equation to cover the nonspherical particles and
proposed
When the fluid is of finite extent, there are two effects.
The fluid streamlines around the particle impinge on
the walls and reflect back on the particle, causing
Table 7 Summary of Theoretical Expressions for the
increasing drag. Also, since the fluid is stationary at
Drag Force [(Re)p 1]
a finite distance from the particle, this distorts the
flow pattern and increases drag. The simple correction Stokes (1851)
can be expressed as : 3ndp[iUT
Oseen (1910)
Fs = 3ndp[iUr I 1 + k c (32) F = Fo 1 + l ( R e ) r + ()|(Re,;}
Proudman and Pearson (1957)
where Lw = the distance from the center of the particle 3 9 , f (Re)D
F = Fs 1 + 7 7 (Re)p + (Re)p In + O J(Re)p }
to the walls, k c = 0.563 (for a single wall), k c = 1.004 16
(for two walls), and k c = 2.104 (for a circular cylinder).
Particle Characterization and Dynamics 17

Table 8 Recommended Empirical Drag Correlations w = log10(Re)p

_ 3 24
(Re)p < 0.01 D “ I6 + (R^)T

0.01 < (Re)p < 20 'Cp(Re)p

logi, - 1 = —0.881 + 0.82w —0.05w
24

24
Cn = [l +0.1315(R e)^82_0-05M;)]

20 < (Re)p < 260 Cp(Re)p

logi. - 1 = -0.7133 + 0.6305w
24

24
Cr) — [l +0.1935(Re)®'6305]

260 < (Re) < 1500 log,o CD = 1.6435 — 1.1242w + 0.1558w

1.5 x 103 < (Re) < 1.2 x 104 logjo CD = -2.4571 + 2.5558tv - 0.9295w2 + 0.1049+

1.2 x 104 < (Re) < 4.4 x 104 log10 CD = -1.9181 + 0.6370w - 0.0636+

4.4 x 104 < (Re)p < 3.38 x 105 logjo CD = -4.3390 + 1.5809w - 0.1546w2

3.38 x 105 < (Re) < 4 x 105 CD = 29.78 - 5.3w

4 x 105 < (Re)p < 106 CD = O.lw —0.49

106 < (Re) 8 x 104

CD = 0.19 -
(Re)p

Source: A dapted from Clift et al. (1978).

24 no fully satisfactory method is available for correlating

Cp —

(Re)P [>+(f \ l \ 6 e -4.06554) ) m ° P0964+0.5565(J)

the drag.
Pettyjohn and Christiansen (1948) determined the
7 3 .6 9 (V 5-07484,)(R e)p free-settling rates of isometric particles of the following
+ shapes and sphericities: spheres (cj) = 1), cube octa­
(Re) + 5.378<?6-2122(f)
hedron (cj> = 0.906), octahedron (cj) = 0.846), cube
(35) (cj> = 0.806), and tetrahedron (cj> = 0.670). Their results
suggest that the correction factor should be
For spherical particles, Eq. (35) reduces to

24 0.4607(Re)p K = 0.8431 log ^ (R e X < ° -° 5 , (3 7)

■3.3643(Re)'0.3471 + 0.065
m : 0.67 < d> < 1
(Re)p + 2682.5
(36) where K is the ratio of the settling velocity of the
volume-equivalent sphere to the settling velocity of
the particle. Results on other isometric particles have
3.7 Corrections for Nonspherical Particles also been published. These include the experimental
results of Heiss and Coull (1952) using solid cylinders
The book by Clift et al. (1978) contains an extensive and rectangular parallelpipeds, Becker (1959) using
review on this subject. The treatm ent is, however, prisms and cylinders, Christiansen and Barker (1965)
mostly for the axisymmetric particles such as spheroids using cylinders, prisms, and disks, Isaacs and Thodos
and cylinders and orthotropic particles such as rectan­ (1967) using cylinders, and Hottovy and Sylvester
gular parallelepipeds. For particles of arbitrary shape, (1979) using roundish irregularly shaped particles.
18 Yang

In the Stokes law regime, homogeneous symmetrical Karam anev and Nikolov (1992) and K aram anev
particles can take up any orientation during their sett­ (1994) showed recently that the relationship between
ling in a fluid of infinite extent. Ellipsoids of uniform CD and (Re)p for particles with density much smaller
density and bodies of revolution with fore and aft sym­ than the fluid follows the standard drag curve only
metry can attain spin-free terminal states in all orienta­ when (Re)p is less than 135. Otherwise the CD is a
tions. The terminal velocities of the ellipsoids will, constant equal to 0.95. Karam anev (1996) proceeded
however, depend on their orientation. A set of identical to suggest the following drag coefficient equations on
particles can therefore have a range of terminal velo­ the basis of the Archimedes number rather than the
cities, though the range is generally fairly small. Reynolds num ber for calculation of the terminal
Asymmetric particles such as ellipsoids and discs do falling or rising velocities.
not generally fall vertically unless they are dropped F or free falling spheres,
with a principal axis of symmetry parallel to a gravity
field. They tend to drift from side to side. CD = f (1+ (4„
In the intermediate flow regime, particles adopt
preferred orientations. Particles will usually align F or free rising spheres,
themselves with their maximum cross section norm al 432 0.517
to the direction of relative motion. There is no appre­
ciable secondary m otion in the intermediate flow
Cr
=^(.+0.0470^ )+ ^ ^ <«>
regime, so results for flow past fixed objects of the for Ar < 1.18 • IQ6;/3
same shape can be used if the orientation corresponds for
CD = 0.95 Ar > 1.18 • 1064 (43)
to the preferred orientation.
Secondary motion associated with wake shedding J p being in meters.
occurs in the N ew ton’s law regime, and CD is insensi­
tive to (Re)p. In this regime the density ratio, X, plays
4 CALCULATION OF TERM INAL VELOCITY
an im portant role in determining the type of motion
FO R A SINGLE PARTICLE
and the mean terminal velocity. For particles of arbi­
trary shapes,
The terminal velocity of a single particle is an intrinsic
CD = 7T1/I8(5.96 - 5-51(f>) for 1.1 < 7. < 8.6 characteristic of the particle, and its calculation and
(38) measurement are as im portant as other . intrinsic
particle properties, such as particle size and density.
Pettyjohn and Christiansen (1948) also suggested a Calculation of the terminal velocity of a single particle
correction factor for nonspherical particles: used to be an iterative process. M ore recent develop­
Cn = 5 .3 1 - 4 . 2 • 103 < (Re)t < 2 • 105, ments allow direct calculations without trial and error.
0.67 < 4> < 1 (39) These methods are introduced here.

CD here is based on the cross-sectional area of the 4.1 Equation by Haider and Levenspiel (1989)
volume-equivalent sphere. The terminal velocity, Uu
can be calculated by From Eq. (30), the terminal velocity for a single
spherical particle can be obtained as
Ut = 0.49X1/36 g(pP - P f H (40)
pf(l .08 — 4>) 14 ^p(Pp Pf )g
(44)
Ur Ut V P fC D

3.8 Drag Coefficient for Particles with Density By combining Eq. (44) and the recommended drag
Lighter than the Surrounding Fluid coefficient correlations in Table 8 or drag coefficient
equations proposed by Turton and Levenspiel (1986),
All the above discussions are for particles with density Eq. (34), and Haider and Levenspiel (1989), Eq. (35),
larger than the surrounding fluid. F or particles with the terminal velocity can be calculated. Haider and
density smaller than the surrounding fluid, it has Levenspiel further suggested an approxim ate m ethod
long been assumed that the free rising velocity is for direct evaluation of the terminal velocity by defin­
governed by the same equations, the only difference ing a dimensionless particle size, d£, and a dimension-
being that of particle movement direction. However, less particle velocity, 17*, by
Particle Characterization and Dynamics 19

1/3 1/3 acceptance. The tables, however, are not very conven­
Pf(pp - Pf)s
d; = dp = A r^ CD(Re)2 ient to use, especially for applications using a com pu­
ter. The tables are now fitted by Fouda and Capes
(45) (1976) in the form of

1/3
pf 1/3_ (Re)p _ [4 (R e )p1
u* = u H r1/3 cDJ
Y = J 2 a nX n (49)
_p(pP - Pf)s_ n=0

(46) where Y = log10CP4) or log10( E (50)

The terminal velocity of the irregular particles can then
be calculated by and X = log10 ( E ) or log10(7Wt) (51)
-1
18 2.335 - 1.744(j)
U* = 0.5
0.5 < 4> < 1 The diameter dt is the particle diameter of a sphere
LK*) P p*) having the same terminal velocity as the particle. P,
(47) Q, Uu and dt are related in the following two dimen-
sionless equations derived by Heywood:
For spherical particles, Eq. (44) reduces to
-l
18 0.591 CDRe2 = Pf)-pf4 3 = P3d t2 (52)
= 1 (48) |P
0.5
_(dp*) (d*) 2
Re Pf ut u[
(53)
CD M pp-pf) p e3
4.2 Terminal Velocity by Polynomial Equations
Fitted to Heywood Tables The resulting polynomial coefficients are summarized
in Table 9.
Fouda and Capes (1976) proposed polynomial equa­ The polynomial equations were reported to have an
tions fitted to the Heywood (1962) tables for calculat­ average deviation less than 0.15% and a standard
ing large numbers of terminal velocities. The Heywood deviation of less than 4.3%.
tables were recognized as being a simple and accurate Similar types of equations were also proposed by
method for calculating both the terminal velocity and H artm an et al. (1994) for nonspherical particles.
the equivalent particle diameter, and they enjoyed wide They introduced the equations

Table 9 Polynomial Coefficients for Polynomial Equations Fitted to Heywood Tables

[see Eq. (49)]

Polynomial coefficients

Terminal velocity calculation log10(Ut/Q) log 10(Pdt) a0 = -1.37323

a! = 2.06962
a2 = -0.453219
a3 = -0.334612 x 10“
2
a4 = -0.745901 x 10"
a5 = 0.249580 x 10“2
Equivalent diameter calculation log10(PJt) log10(£/t/g ) ao —0.785724
ai = 0.684342
a2 = 0.168457
a3 = 0.103834
a4 = 0.20901 x 10_1
a 5 = 0.57664 x 10“2
 Yang

log(Re)t( Y, 4>) = log(Re)t( V, 1) + P( Y, cj>) (54)

(59)
and
where k is the permeability and R is the radius of the
log(Re)t(F , 1) = 0.77481 - 0.56032 log Y
sphere. The resistance ratio, Q, is normally less than
+ 0.024246(log Y)2 - 0.0038056 unity. F or impermeable spheres, Q approaches unity
because P approaches infinity. F or infinitely permeable
(log Y f (55) spheres, P tends to zero and so does Q.
P(Y, ct>) = -0.10118(1 - 4>) log Y + 0.092944 Based on Eq. (58), the ratio of terminal velocities o f
permeable and impermeable spheres can be written as
(1 - 4»)(log Y)2 - 0.0098356(1 - 4>)
(V )P _ l
(60)
(log Y 'f - 0.12666(1 - c]))2 log Y Wt)ip «
(56) and the ratio of dimensionless drag coefficient can be
expressed as
where
(CDRe)
g(pP ~ Pf)n (61)
Y= 2 (57) (CDRe)ip
Ut3pj
where
These equations are good for 0.01 < (Re)t < 16000
and 0.67 < 4> < 1. The equation fitted the experimental 4rf2g[(l - e)(pp - pf)]
(CDRe) = (62)
terminal velocities of limestone and lime particles to 3p(V)P
better than 20%
and
4.3 Calculation of Terminal Velocity of Porous 4dpg(pp - pf)
Spheres (CDRe)iT = 24 (63)
3p(t/t)ip

Calculation of the terminal velocity of a porous sphere Equation (60) shows that a permeable sphere with the
is useful and im portant in applications in water treat­ same diameter and bulk density as the impermeable
ment where settling velocities of a floe or an aggregate sphere will have a higher terminal velocity than that
are estimated. It is also im portant in estimation of of the impermeable sphere. At higher Reynolds num ­
terminal velocities of clusters in fluidized bed applica­ ber, the experiments by Masliyah and Polikar (1980)
tions. The terminal velocity of a porous sphere can be suggested the following equations:
quite different from that of an impermeable sphere. F or 15 < p < 33
Theoretical studies of settling velocity of porous 24Q
spheres were conducted by Sutherland and Tan Cr> — 1 0.1315(Re)),0'82~005M')l
(Re)p . (64)
(1970), Ooms et ah (1970), Neale et al. (1973),
Epstein and Neale (1974), and M atsum oto and 0.1 < (Re)p < 7
Suganuma (1977). The terminal velocity of porous
and
spheres was also experimentally measured by
Masliyah and Polikar (1980). In the limiting case of a 24Q
Co — [l +0.0853(Re)!(1.093—0.105w)
very low Reynolds number, Neale et al. (1973) arrived m ~ P
'P
(65)
at the following equation for the ratio of the resistance
experienced by a porous (or permeable) sphere to an 7 < (Re)p < 120
equivalent impermeable sphere. An equivalent where w = log10(Re)p.
impermeable sphere is defined to be a sphere having This confirms the theoretical study of Neale et al.
the same diameter and bulk density of the permeable (1973) that the drag experienced by a porous sphere
sphere. at low Reynolds numbers is less than that for an
2ft [1 —(tanh P)/p] impermeable sphere of similar diameter and bulky den­
Q =- (58) sity. The effect of inertia at high Reynolds numbers is
2(32 + 3[1 —(tanh (3)/(3]
higher for a porous sphere than for an impermeable
where P is the normalized sphere radius expressed by sphere of similar diameter and bulk density.
Particle Characterization and Dynamics 21

As far as the wall effect is concerned, the experi­ £ « /4

ments show that, at low Reynolds numbers, the wall _j_____ (67)
dR=
effect for a porous sphere is of the same order of m ag­ £«<
nitude as that for an impermeable sphere. At high
Reynolds numbers, however, the wall effect become
smaller and less significant for a porous sphere. 5.1.3 Volume M ean
The volume mean diameter is the diameter of a
hypothetical spherical particle whose volume multi­
5 CHARACTERIZATION OF MULTIPARTICLE
plied by the total num ber of particles in the assemblage
SYSTEM S
would be equal to the total volume of the assemblage.
In most applications, the systems and processes con­
tain large am ounts of particles with size distribution; E ^ '4
( 68)
each size may also possess a distinguished shape. To T > r
describe properly these systems and processes for
design and analysis, they need to be adequately
characterized to reflect their physical and chemical 5.1.4 Volume-Surface M ean
potentials. In the following sections, different average
particle diameter definitions are introduced along with The volum e-surface mean is also known as the Sauter
statistical descriptions of particle size distribution. mean. This defines the average particle size based on
Depending on applications, one definition may be the specific surface area per unit volume or per unit
more suitable than others. Thus care must be exercised weight, as shown in Eq. (69):
to select the proper characterization for each process.
E M p ,-
5.1 Different Definitions of Average Particle r _ i
(69)
Diameter £ nidpi £-
■*pi
As mentioned earlier, naturally occurring particles are
usually irregular in shape and also different in size. 5.1.5 Weight Mean
Different definitions of average particle diameter
The weight mean is also known as the DeBroucker
have been used to characterize an assemblage of par­
mean. This defines the average particle diameter
ticles. Some may be hypothetical while others may
based on the unit weight of the particles. The weight
have physical significance. They are summarized here.
mean particle diameter is the diameter of a sphere
whose surface area times the total num ber of particles
5.1.1 Arithmetic Mean
equals the surface area per unit weight of the assem­
The arithmetic mean is defined as the sum of all dia­ blage.
meters divided by the total num ber of particles
E Mp;
E H (70)
(66) 4
where
where can be the num ber of particles or the weight
percentage. Widpi
Vi = (71)
£ « ;4
5.1.2 Surface M ean
The surface mean diameter is the diameter of a
5.1.6 Length M ean
hypothetical spherical particle whose surface multi­
plied by the total num ber of particles in the assemblage The length mean is the average diameter obtained by
would be equal to the total surface area of the assem­ dividing the total surface area of all particles with the
blage, expressed mathematically as summation of the diameters, as shown here:
22 Yang

E nidli
(72)
E>v 'pz

5.1.7 Geometric M ean

The geometric mean is the logarithmic equivalent of
the arithmetic mean. Being a logarithmic average, the
geometric mean is always smaller than the arithmetic
mean.
_ njmx _jn2 ' mn (73)
ag ~ y u p l a p2 Pn

E h log K i) ]
d„ — - (74)
Yd

5.1.8 Harm onic Mean Particle S i z e --------

This is an average diameter based on the summation of Figure 9 Skewed particle distribution of typical dust.
the reciprocals of the diameters of the individual com­
ponents of an assemblage and can be expressed m ath­
ematically as 4 V = log 1(log dg + 1.151 log2 <7g) (76)
E ni , d\/2 ~ dg (77)
= (75)
E dr\. dm = log-1 (log dg - 2.303 log2 a g) (78)

The harmonic mean is actually related to the average Because of its m athematical properties, the standard
spherical particle corresponding to the particle surface deviation a is almost exclusively used to measure the
per unit weight. M athematically, the harmonic mean is dispersion of the particle size distribution. When the
similar to the volume-surface mean or Sauter mean. skewed particle size distribution shown in Fig. 9 is
replotted using the logarithm of the particle size, the
5.2 Statistical Characterization of Particles with a skewed curve is transform ed into a symmetrical bell­
Size Distribution shaped curve as shown in Fig. 10. This transform ation
is of great significance and importance in that a sym­
Statistically, the particle size distribution can be metrical bell-shaped distribution is amenable to all the
characterized by three properties: mode, median, and statistical procedures developed for the norm al or
mean. The mode is the value that occurs most gaussian distribution.
frequently. It is a value seldom used for describing In the log-normal particle size distribution, the
particle size distribution. The average or arithmetic mean, median, and mode coincide and have an identical
mean diameter, J av, is affected by all values actually value. This single value is called the geometric median
observed and thus is influenced greatly by extreme particle size, J g, and the measure of dispersion, the geo­
values. The median particle size, <71/2, is the size that metric standard deviation, a g. Thus the log-normal
divides the frequency distribution into two equal areas. particle size distribution can be described completely
In practical application, the size distribution of a typi­ by these two characteristic values. To determine
cal dust is typically skewed to the right, i.e., skewed to whether the particles have a distribution close to log­
the larger particle size. The central tendency of a norm al distribution, the particle cumulative frequency
skewed frequency distribution is more adequately data can be plotted on a logarithmic probability graph
represented by the median rather than by the mean paper. If the particle size distribution is log-normal, a
(see Fig. 9). M athematically, the relationships among straight line will result. The geometric median particle
the mean, median, and mode diameter can be size is the 50% value of the distribution as shown in
expressed as Fig. 11. The geometric standard deviation is equal to
Particle Characterization and Dynamics 23

soda, ash, sodium bicarbonate, alumina, and clay all

were observed to fit the log-normal distribution.
M athematically the log-normal distribution can be
expressed as

1 ( ^ ) 2‘
/ = exp (79)
o 2V2n 2 cj2

where

d\\f zd\\f
/ = dz ’ z = In ((Li), (80)
E d\\f
and \|/ can be number, surface, or weight.
Mathematically, the log-normal distribution can
Particle Size (Log Scale) -------- also be derived from the arithmetic distribution, by
substituting x by In x = z to give Eq. (81):
Figure 10 Log-normal particle size distribution.
1 (x —x)2
/ = / ^ exP (81)
av2n 2a2
the ratio of 84.1% value divided by the 50% value or
50% value divided by the 15.9% value. Although a2 =
E (X - X ?
(82)
68.26% of the particles will lie in the particle size n —1
range between dg + a g and dg —a g. If the particle size
reduction is due to comminution such as crushing, An im portant characteristic of the log-normal distri­
milling, and grinding, the resulting particle size distri­ bution is that the transform ation among the various
bution very often tends to be log-normal distribution. particle size definitions and the statistical diameters
can be performed analytically and graphically (Smith
Pulverized silica, granite, calcite, limestone, quartz,
et al., 1929). The H atch-Choate (1929) transform ation
equations are summarized here:

log dgm = log d„c + 6.908 log (83)

log dav = logd„c + 1.151 log Og (84)

log ds log dgc + 2.303 log' a g (85)

log dv log dgc + 3.454 log a g (86)

CO
o
00 log dvs log dgc + 5.757 log2 a g (87)

logdw log dgc + 8.023 log2 dg (88 )

log dgc log dgm - 6.908 log2 cig (89)

•■e
CO
Ol
log dav = log d L - 5.757 log2 c g (90)

log ds log dgm —4.605 log CTg (91)

log log dgm - 3.454 log d g (92)

Cumulative Less Than, %
log dvs log dgm - 1.151 log2 d g (93)
Figure 11 Log-normal distribution plotted on a logarithmic
probability graph paper. lOg dyj : log dgm + 1-151 log2 dg (94)
24 Yang

5.3 The Rosin-Rammler Distribution

The Rosin-Ram m ler (1933) distribution was first

developed for broken coal, but it has since been
found to be applicable to many other materials, such
as cement, gypsum, magnetite, clay, dyestuffs, quartz,
flint, glass, and ores. From probability considerations,
the authors obtained

dW^ Sa) = 100 nbS"~l exp(-M Z) (95)

where n and b are constants: b is a measure of the range

of particle sizes present, and n is a characteristic of the
substance. Integrating Eq. (95), we have
Figure 12 A two-dimensional device for management of
Wf = 100 exp(—b$a) (96) angle or the repose and the angle of internal friction.
or

l o g - l o g 0 ^ j = constant+ nlog(5a) (97) when the same powder is pulled on top of a flat sur­
face. N orm al powders have angles of repose of around
If log-log(100/IFf) is plotted against log Ya, a straight 35°. The angles of repose for commonly occurring
line results. The peak of the distribution for n = 1 is at powders are summarized in Table 10.
100/e = 36.8%. This is used to characterize the degree The angle made by the free surface of the powder
of comminution of the material. In Eq. (97), Wf is the still remaining in the test device with the flat bottom o f
weight percent of material retained on the sieve of the test device, the angle a in Fig. 12, is called the angle
aperture *Sa. If the mode of the particle size distribution of internal friction of the powder. The angle of internal
curve is (Ya)m, Eq. (97) gives b — l/(*Sa)m. Since the friction is also an intrinsic property of the powder and
slope of the line on the Rosin-Ram m ler graph depends can be considered as a shearing plane of the powder. In
on the particle size range, the ratio of tan -1 (n) and (S a)m order for the powder to flow, the angle has to be higher
is a form of variance. This treatm ent following Rosin than the angle of internal friction of the powder.
and Rammler is useful for m onitoring grinding opera­ Norm al powders have angles of internal friction o f
tions for highly skewed distributions. It should be around 70°. The angles of internal friction for com­
employed carefully, however, since taking the loga­ monly occurring powders are also listed in Table 10.
rithm always reduces scatter, taking the logarithm Both the angle of repose and the angle of internal
twice as in this case, tends to obscure the actual scatter friction can be modified and reduced by flow aids
in the distribution. such as aeration, vibration, and the addition of other
powders.
5.4 Measurement of the Angle of Repose and the
Angle of Internal Friction

Two im portant characteristics of powder rheology are Table 10 Typical Angles of Repose and Angles
the angle of repose and the angle of internal friction. of Internal Friction
Simple devices can be constructed to measure both. Angle of
Figure 12 depicts a simple two-dimensional bed with Angle of repose internal friction
transparent walls and a small orifice at the bottom of Material degree degree
the bed. After filling the bed with the powder to be
examined, the powder is allowed to flow out of the FCC catalyst 32 79
test device to the surface of the test stand. The angle Sand 36 64
Resin 29 82
of the powder-free surface measured from the flat sur­ —
Wet ashes 50
face of the test stand, the angle P in Fig. 12, is called
Wheat — 55
the angle of repose. This angle is an intrinsic charac­ Oats 21 —
teristic of the powder and should be close to a constant
Particle Characterization and Dynamics 25

NOMENCLATURE /c = intensity of light beam after passing through a

particle suspension of concentration C
Ap = projected area of a particle I0 = original light beam intensity
k = Heywood shape factor
Ar = Archimedes number, Ar = k = permeability
= K = ratio of the settling velocity of the volume-
exterior surface area/volume of catalyst pellet
= equivalent sphere to the settling velocity of the
b breadth
= particle
C concentration of particle suspension
kc = a constant
CD drag coefficient
= projected-area diameter ke = Heywood shape factor for an isometric particle
da. of similar form
d_A = sieve diameter
/ = length
daw = arithmetic mean particle diameter
ly = length of a cylinder
dc = perimeter diameter
= drag diameter L = length of light beam path
d-o
Lw = the distance from the center of a particle to the
df = free-falling diameter
= Feret diameter walls
d^?
M = mass of a particle
dg = geometric mean particle diameter or geometric
nt = number of particles of size i or weight percent
median particle diameter
of particles of size i
dgc = geometric median diameter by count
R = radius of a sphere
dgm = geometric median diameter by weight
(Re)p = Reynolds number based on the particlediameter
4 = harmonic mean particle diameter
(Re)t = Reynolds number based on the terminal velocity
di = inside diameter of a ring
of the particle
di = length mean particle diameter Sa = sieve aperture
dM = Martin diameter
(Sa)m = sieve aperture at mode of particle size
dm = mode particle diameter distribution
4 = outside diameter of a ring Sp = surface area of a particle
4 = equivalent particle diameter Sv = particle surface per unit volume of particle
dpt == particle diameter of size i t = thickness
d*pu = dimensionless particle diameter
T = measured transmittance, Xc/X 0
4 = surface diameter U* = dimensionless particle velocity
4 = surface mean particle diameter Ur = relative velocity between the particle and the
4t = Stokes diameter fluid
4v = surface-volume diameter Ut = terminal velocity of a single particle
4 particle diameter of a sphere having the same (Ut)iv = terminal velocity of an impermeable particle
terminal velocity as the particle (Ut)p = terminal velocity of a permeable (or porous)
dy = volume diameter particle
4 = volume mean particle diameter Vm = measured velocity in a column of diameter D
dys = volume-surface mean particle diameter Vp = volume of a particle
4 = weight mean particle diameter Foo = falling speed of a sphere in an infinite fluid
dy = diameter of a cylinder Wf = weight fraction
D = sedimentation column diameter Ws = weight fraction of particles in the x-ray pass
D R i0 = particle diameter ratio at the cumulative mass < W2oo = weight of material passing through 200 mesh
10% screen in grams
d r 50 — particle diameter ratio at the cumulative mass < X/ = weight fraction of particle size dpi
50% Xc = intensity of x-ray beam passing through a
d r 90 — particle diameter ratio at the cumulative mass < particle suspension of concentration C
90% X0 = intensity of x-ray beam passing through the
4 /2 = median particle diameter sedimentation column filled with clear fluid
= flatness ratio ji = viscosity of fluid
^2 = elongation ratio p == packing density
/ = frequency of observation pf = density of fluid
F = force exerting on a particle pp = density of particle
Fs = force exerting on a sphere c|) = sphericity
g = gravitational acceleration 4>op = operational sphericity
26 Yang

<l>w = working sphericity LS Fan, C Zhu Principles of Gas-Solid Flows. Cambridge:

£ = circularity Cambridge University Press, 1998, pp 13-16.
<£0p = operational circularity AE Fouda, CE Capes. Calculation of large numbers of
a = an empirical constant depending on equipment terminal velocities or equivalent particle diameters using
and particle characteristics polynomial equations fitted to the Heywood tables.
a = angle of internal friction Powder Technol 13:291-293, 1976.
P = an empirical constant depending on equipment, PJ Freud, MN Trainer, AH Clark, HN Frock. Unified scatter
particle, and suspending fluid characteristics technique for full-range particle size measurement. Proc.
P = normalized sphere radius Pittsburgh Conference, Atlanta GA, March 8-12, 1993.
P = angle of repose A Haider, O Levenspiel. Drag coefficient and terminal
X = density ratio velocity of spherical and non-spherical particles. Powder
Q = ratio of resistance experienced by a porous Technol 58:63-70, 1989.
sphere to an equivalent impermeable sphere J Happel, H Brenner. Low Reynolds Number Hydro­
4/ = number, surface, or weight dynamics. New Jersey: Prentice Hall, 1965.
a = standard deviation M Hartman, O Trnka, K Svoboda. Free settling of non­
= geometric standard deviation spherical particles. Ind Eng Chem Res 33:1979-1983,
1994.
T Hatch, S Choate. Statistical description of the size proper­
REFERENCES ties of non-uniform particulate substances. J Franklin Inst
207:369-387, 1929.
S Akapo, TM Khong , CP Simpson , B Toussaint, JG Yates. IF Heiss, J Coull. Effect of orientation and shape. Chem Eng
A novel method for determining the particle density of Prog 48:133-140, 1952.
porous aeratable powders. Powder Technol 58:237-242, G Herdan. Small Particle Statistics. 2nd ed. New York:
1989. Butterworths, 1960.
T Allen. Particle Size Measurement. London: Chapman and H Heywood. Uniform and non-uniform motion of particles
Hall, 1975. in fluids. In: Proc Symp Interaction Fluids and Parts.
B Aschenbrenner. A new method of expressing particle London: Inst Chem Eng, 1962, pp 1-8.
sphericity. J Sediment Petrol 26:15-31, 1956. ID Hottovy, ND Sylvester. Drag coefficients for irregularly
LG Austin. Conversion factors to convert particle size dis­ shaped particles. Ind Eng Chem Process Des Devel
tributions measured by one method to those measured by 18:433^136, 1979.
another method. Part Part Syst Charact 15:108-111, 1998. JL Isaacs, G Thodos. The free-settling of solid cylindrical
Babcock & Wilcox. Steam. 40th ed. Barberton: The Babcock particles in the turbulent region. Can J Chem Eng
& Wilcox Company, 1992, pp 12-8-12-9. 45:150-155, 1967.
HA Becker. The effects of shape and Reynolds number on DG Karamanev. On the rise of gas bubbles in quiescent
drag in the motion of a freely oriented body in an infinite liquids. AIChE J 40:1418-1421, 1994.
fluid. Can I Chem Eng 37:85-91, 1959. DG Karamanev. Equations for calculation of the terminal
CF Bohren, RD Huffman. Absorption and Scattering of velocity and drag coefficient of solid spheres and gas
Light by Small Particles. New York: Wiley-Interscience, bubbles. Chem Eng Comm 147:75-84, 1996.
1986. DG Karamanev, LN Nikolov. Free rising spheres do not
MP Bonin, M Queiroz. Local particle velocity, size, and con­ obey Newton’s law for free settling. AIChE J 38:1843-
centration measurements in an industrial-sale, pulverized 1846, 1992.
coal-fired boiler. Comb and Flame 85:121-133, 1989. BH Kaye. Automated decision-taking in fine particle science.
B Buczek, D Geldart. Determination of the density of porous Powder Technol 8:293-306, 1973.
particles using very fine dense powders. Powder Technol BH Kaye. Direct Characterization of Fine Particles. New
45:173-176, 1986. York: John Wiley, 1981.
EB Christiansen, DH Barker. The effect of shape and density MJ Knight, PN Rowe, HJ Macgillivray, DJ Cheesman.
on the free settling of particles at high Reynolds numbers. On measuring the density of finely divided but porous
AIChE J 11:145-151, 1965. particles such as fluid bed cracking catalyst. Trans Instn
R Clift, JR Grace, ME Weber. Bubbles, Drops, and Chem Engrs 58:203-207, 1980.
Particles. New York: Academic Press, 1978. M Leva. Fluidization. New York: McGraw-Hill, 1959.
RP Davuluri, TM Knowlton. Development of a standardized PJ Lloyd. Particle characterization. Chem Eng 120-122,
attrition test procedure. In: LS Fan, TM Knowlton, eds. April 29, 1974.
Fluidization IX. New York: Engineering Foundation, JB Masliyah, M Polikar. Terminal velocity of porous
1998, pp 333-340. spheres. Can J Chem Eng 58:299-302, 1980.
N Epstein, G Neale. On the sedimentation of a swarm of K Matsumoto, A Suganuma. Settling velocity of a permeable
permeable spheres. Chem Eng Sci 29:1841-1842, 1974. model floe. Chem Eng Sci 32:445-447, 1977.
Particle Characterization and Dynamics 27

M Naito, O Hayakawa, K Nakahira, H Mori, J Tsubaki. P Rosin, EJ Rammler. The laws governing the fineness of
Effect of particle shape on the particle size distribution powdered coal. J Inst Fuel 7(31):29—36, 1933.
measured with commercial equipment. Powder Technol WO Smith, PD Foote, PF Busang. Capillary rise in sands of
100:52-60, 1998. uniform spherical grains. Phys Rev 34:1271-1279, 1929.
G Neale, N Epstein, W Nader. Creeping flow relative to P Subramanian, Vr Arunachalam. A simple device for the
permeable spheres. Chem Eng Sci 28:1865-1874, 1973. determination of sphericity factor. Ind Eng Chem
IP Olivier, GK Hickin, C Orr Jr. Rapid automatic particle Fundam 19:436-437, 1980.
size analysis in the sub-sieve range. Powder Technol DN Sutherland, CT Tan. Sedimentation of a porous sphere.
4:257-263, 1970/1971. Chem Eng Sci 25:1948-1950, 1970.
G Ooms, PF Mijnlieff, H E Beckers. Frictional forces exerted R Turton, NN Clark. An explicit relationship to predict
by a flowing fluid on a permeable particle, with particular spherical particle terminal velocity. Powder Technol
reference to polymer coils. J Chem Phys 53:4123-4134, 53:127-129, 1987.
1970. R Turton, O Levenspiel. A short note on the drag correlation
CW Oseen. Uber die Stokessche Formel und uber die for spheres, Powder Technol 47:83-86, 1986.
verwandte Aufgabe in der Hydrodynamik. Arkiv for H Wadell. Sphericity and roundness of rock particles. J Geol
Mathematik Astronomi och Fysik. Vol. 6, No. 29, 41:310-331, 1933.
1910. H Wadell. The coefficient of resistance as a function of
ES Pettyjohn, EB Christiansen. Effect of particle shape on Reynolds number for solids of various shapes. J
free-settling rates of isometric particles. Chem Eng Prog Franklin Inst 217:459-490, 1934.
44:157-172, 1948. H Wadell. Volume, shape, and roundness of quartz particles.
M Pohl. Technology update: particle sizing moves from J Geol 43:250-280, 1935.
the lab to the process. Powder and Bulk Eng. 39-46, J Werther, J Reppenhagen. Attrition in fluidized beds and
February 1998. pneumatic conveying lines. In: WC Yang, ed.
I Proudman, JRA Pearson. Expansions at small Reynolds Fluidization, Solids Handlings, and Processing
numbers for the flow past a sphere and a circular cylinder. Westwood, NJ: Noyes Data Corporation, 1999,
I Fluid Mech 2:237-262, 1957. pp 435-491.
HE Rase. Fixed-Bed Reactor Design and Diagnostics.
Boston: Butterworths, 1990, pp 118-119.
2

Flow Through Fixed Beds

W e n -C h in g Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.

Fixed beds, also called packed beds, play a vital role in angle and the triangular or the simple rhombic layer of
chemical processes. Their simplicity induced applica­ an angle of 60° (see Fig. 1). The vertical stacking of
tions in many unit operations such as adsorption, dry­ these layers yields six possible regular packings. For
ing, filtration, dust collection, and other catalytic and the six possible regular packings of spherical particles,
noncatalytic reactors. The prim ary operating cost for a the voidage is only a function of packing arrangement.
fixed bed is the pressure drop through the packed bed The voidage is also independent of the particle size.
of solids. Thus to understand the design and operation The six arrangements are cubic, two orientations of
of a packed bed requires study of packing character­ orthorhom bic, tetragonal-spheroidal, and two orienta­
istics of particles and their effect on pressure drop tions of rhom bohedral. They are graphically presented
through the bed. in Fig. 2. Form ulae for calculating the voidage of these
six arrangements have been developed and the packing
characteristics of these ordered arrangements are sum­
1 PACKING CHARACTERISTICS OF marized in Table 1 (Herdan, 1960). The closest packing
MONOSIZED SPHERES is the rhom bohedral with a voidage of 0.2595, and the
loosest, the cubic with a voidage of 0.4764. The stabi­
The simplest case of a packing system is a bed that lity of the systems increases as the voidage decreases.
consists of only uniform and regular monosized sphe­ The systems tend towards the orthorhom bic state,
rical particles. For this simple system, the voidage of especially if a mechanical disturbance such as vibration
an ordered arrangement of monosized spheres can be is applied to the systems. The voidage of a bed of
derived satisfactorily by m athematical considerations. monosized particles after prolonged shaking is usually
R andom packing of monosized spheres is more com­ about 0.395, approaching the characteristic of an
plicated and can be described mathematically through eight-point contact packing of orthorhombic.
a coordination number. However, the voidage deter­
m ination of random packing of monosized spheres 1.2 Random Packing of Uniform Monosize Spheres
should be done through experiments.
Random packings of uniform monosize spheres are
1.1 Regular Packing of Uniform-Sized Spheres created by irregular and random arrangements of p ar­
ticles. According to Scott (1960), two reproducible
For convenience, the regular packing can be consid­ states of random packing, dense random packing and
ered to be constructed from regular layers and rows. loose random packing, can be experimentally created.
The two basic layers are the square layer with a 90° The dense random packing can be developed by

29
30 Yang

Top View Side View Top View Side View

Top View

Figure 2 Six possible arrangements of monosize spheres—one cubic (Figure 2a), two orthorhombic (Figures 2b and 2c), one
tetragonal-spheroidal (Figure 2d), two rhombohedral (Figure 2e).
Flow Through Fixed Beds 31

Table 1 Packing Characteristics of Ordered Uniform Monosized Spheres

Points of Surface of spheres Cross-sectional

Packing contact Porosity, % per cm3 pore area per cm2

Cubic 6 47.64 1.57/R 0.2146

Orthohombic
(two orientations) 8 39.54 1.81 /R 0.2146
Tetragonal-Spheroidal 10 30.19 2.10/R 0.093
Rhombohedral
(two orientations) 12 25.95 2.22/R —

pouring the spheres into cylindrical containers and A param eter called the layer spacing, (3Jp, where dp is
then shaking for several minutes. The loose random the diameter of the sphere, has also been used to char­
packing is created by tipping the container horizontally acterize the packing. F or most common packings, (3 is
and rotating slowly about its axis, and then returning between 0.707 and 1.0, the limits corresponding to the
slowly to its vertical position. The dense random pack­ rhom bohedral packing and cubic packing, respectively.
ing density obtained by several authors ranged from Though the layer spacing has little physical meaning in
0.61 to 0.63, equivalent to a voidage of 0.39 and 0.37, random packing, the concept is useful for describing
corresponding closely to orthorhom bic packing with a the packing. It is actually related to the bed voidage in
voidage of 0.3954. The loose random packing obtained the following way:
has a voidage of 0.40 to 0.42, corresponding to a pack­
1/3
(2)
ing between the cubic packing (voidage = 0.4764) and
orthorhom bic packing. Haughey and Beveridge (1969) P= _3V2(1 - s).
classified the packing into four different modes: very
loose random packing, loose random packing, poured
random packing, and close random packing. The loose 1.4 Specific Surface Area of the Bed
random packing and poured random packing corre­
spond to Scott’s loose and dense random packings. The specific surface area of the bed is defined as the
The very loose random packing corresponds to the ratio of total particle surface area to the total bed
state where the bed is first fluidized and then the gas volume. Since the num ber of spheres per unit volume
is slowly reduced until it is below the minimum fluid­ of bed is
ization. The bed so formed usually has a voidage of
about 0.44. The close random packing is formed by 6 (1 - E )
(3)
vigorously shaking or vibrating the bed, and then the %dl
voidage is usually approaching 0.359 to 0.375.

1.3 Properties of Regular and Random Packing of Table 2 Correspondence of

Uniform Monosized Spheres Coordination Number and
Voidage
The coordination number, defined as the number of Coordination Number Voidage
spheres in contact with any neighboring spheres, is
used to characterize the voidage of packing. Table 2 3 0.7766
lists the coordination numbers from 3 to 12 and its 4 0.6599
associated voidage. The coordination numbers of 4, 5 0.5969
6 0.4764
6, 8, 10, and 12 correspond to arrangements that are
7 0.4388
regular as shown in Table 1.
8 0.3955
A model derived by Haughey and Beveridge (1966) 9 0.3866
relates the average coordination num ber with the mean 10 0.3019
bulk voidage as 11 0.2817
12 0.2595
n = 22.47 - 39.39s 0.259 < s < 0.5 (1)
32 Yang

The specific surface area can be evaluated as the interstices of the big particles. W hen the fraction
of the small particles decreases, segregation of small
(4) particles becomes inevitable, as shown in Fig. 4.
Equations were proposed to calculate the average
voidage of packed bed with binary mixtures.
F or 0 < X h < X h9
2 PACKING CHARACTERISTICS OF BINARY
1 - eR
MIXTURES OF SPHERICAL AND Sov — 1 (5)
( l _ Xb) + a l b ( l - £s)
NONSPHERICAL PARTICLES
where
The packing of binary mixtures of spheres is the sim­
plest 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
a = 1+ /c
(S (6)

the mixtures. Following M cGeary’s (1967) develop­ / c = l for = 0.5 (7)

4
ment, the minimum fraction of the coarse particles
for closest packing is 73%, and the closest packing fc = 1.2 for ~ = 0.25 (8)
voidage is 0.14. 4
The dependence of maximum observed binary
for — < 0.125 (9)
II
mechanical packing of spheres on the ratio of dia­
4
meters between the large and small spheres shows
that when the size ratio, or diameter ratio, decreases, For Ab < Ab ^ Vnin?
the changes in voidage increase. With a diameter ratio
( l - s s)(Xb - X b)
larger than about 7 to 1, the change in voidage is only Cfiv — 1
very slight. This ratio is very close to the critical Ab{ l - A b[ l - ( l - s s)a]}2
diameter ratio of 6.5 for the smaller sphere to pass
( 10)
W(1 - £s)
through the triangular opening formed by three larger
spheres in close packing (see Fig. 7). Z b{ l - l b[ l - ( l - s s)a]}
Abe et al. (1979) carried out theoretical analysis of a where
packed bed of binary mixture and proposed equations
to estimate voidage at different degree of mixing of ( 11 )
binary mixtures of spherical and nonspherical parti­ a+ (1 — c m ) ( 1 — s s)

cles. If cohesive forces are absent, the binary mixture and

will yield a packed bed of minimum voidage at Xmin, as
shown in Fig. 3. In the region X h < Xmin, large parti­ (1 - Sb) (12)
a = -
cles are distributed randomly and evenly in the packed [i + i . 5 ( 4 / 4 ) r
bed. This state is called complete mixing. In the region
X h > Amin, there are two different states possible. In For Xmm < X b < 1,
the first state, the small particles are distributed in (1 - eb)
(13)
* b( l + PsPc) 3

where

(14)

and

4V
Ps = (15)
d j
Figure 3 Packed bed of minimum voidage for binary par­ n equals 1/3 for crushed stones and 1/2 for round
ticle mixtures. sand.
Flow Through Fixed Beds 33

X b > X min X b > X min

Figure 4 States of binary particle mixtures.

The intersection of Eqs. 10 and 13 gives the binary — = 1.355rp1.566 for rp < 0.824 (17)
mixture of large and small particles where the voidage
is the minimum, or X min and smin: and

8av = Average voidage of the binary mixture -= l for rp > 0.824 (18)
smin = The minimum voidage of the binary mix­
ture where
bs = Voidage of a packed bed with small parti­ do
cles alone (19)
rp = 4
eb = Voidage of a packed bed with large parti­
cles alone F or the limiting case where rp = 1, the voidage of the
X b = Volume fraction of large particles packed bed will not change through mixing of the
X min = Volume fraction of large particles where the binary particles. W hen rp approaches 0, the interstices
voidage of the binary mixture is the mini­ between large particles can be filled with the small
mum particles as discussed earlier. The voidage and the
ds = Diameter of the small particles specific solid volume of the packed bed become
db = Diameter of the large particles 8 8b8s (20)
Vbvs
A common approach to predict the voidage of binary (21)
Vh + Vs - 1
particle mixtures is to make use of the empirical corre­
lation developed by Westman (1936) based on the total 1 eb
Xh = (22)
bed volume occupied by the specific volume of solid 1 8bSs
material V = 1/(1 —s). The equation was later m odi­
The effect of changing the particle size ratio on the
fied by Yu et al. (1993) as shown in the following
packing of binary particles is summarized in Fig. 5.
equation to apply to both spherical and non-spherical
For nonspherical particles, Yu et al. (1993) sug­
particles.
gested to substitution of the particle diameter by the
packing equivalent particle diameter calculated by
v - r hXb -2 G V -V bX b
K K 3.1781 - 3.6821 2 + 1.5040 T
V — X b — VSXS
+
v — x h — VSXS
= 1
(16) d*pe
-( <t> <r
(23)

Vu-l Vh - 1 where dYQ is the volume equivalent particle diameter

and c)> is the W adell’s sphericity described in C hapter
The empirical constant G is independent of the com­ 1. This approach provides good agreement for packing
position of the mixtures but depends on the size ratio characteristics of both spherical and nonspherical
of the particles. Yu et al. (1993) gave the following binary particles, though evaluation of the packing
values for G: equivalent particle diameter is somewhat cumbersome.
34 Yang

m
■&_
m
'm

o
CO
*5
<D
E
J3
5
o
<D
CL
CO

Volume Fraction of Big Component, X b

Figure 5 Effect of changing particle size ratio on the packing of binary particles. (Adapted from Finkers and Hoffmann, 1998).

An approach employing the structural ratio to have also been studied by Ouchiyama and T anaka
predict the voidage of binary particle mixtures for (1989) and by Hoffmann and Finkers (1995). The
both spherical and nonspherical particles is also pro­ packing characteristics of spheres of unequal sizes
posed recently by Finkers and Hoffmann (1998). The have also been investigated (Herdan, 1960). In rhom-
structural ratio based on packing of spherical and non­ bohedral packings of different spherical particle sizes,
spherical particles is defined as the voidage may be reduced to less than 0.15.

fe -p ' (24)
1 —So 3 CRITICAL RATIO OF ENTRANCE AND
CRITICAL RATIO OF OCCUPATION IN
The structural ratio is then used to calculate the BINARY SYSTEM S
empirical constant G for use in Eq. (16).
When the monosized spheres are arranged in norm al
o-o (25)
loosest square or tightest rhom bohedral packings,
there are critical smaller spheres, which can pass
A value of -0 .6 3 was suggested for k in Eq. (25). The
through the openings formed by the larger monosized
proposed approach is good for both spherical and
spheres, as shown in Figs. 6 and 7. These critical dia­
nonspherical particles. F or binary particles with size
meter ratios are called the “critical ratio of entrance.”
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 M oreland (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 6 Critical ratio of entrance for square packing.
Flow Through Fixed Beds 35

rings, and Berl saddles. Coulson (1949) studied pack­

ings 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
Figure 7 Critical ratio of entrance for rhombohedral pack­ sphericity of the particles and can be related as
ing. shown in Fig. 8 and Table 3b based on experimental
findings. F or small grains observable under the micro­
scope, the sphericity can be obtained following Wadell
For the loosest square packing, this critical diameter (1935) and the average values obtained from the fol­
ratio can be expressed as lowing equation after observing a sample of particles.
4 = 1
= 2.414 (26) (31)
4 \ f l —1
F or the tightest rhom bohedral packing, the critical where dc = diameter of a circle equal in area to the
ratio can be found from projected area of the particle when resting on its larger
i face, Dc = diameter of the smallest circle circumscrib­
4
= 6.464 (27) ing the projection of the particle.
4 (2/V3) - 1 When the packed bed is made up of a mixture of
There also exists a critical size sphere that, though it particles with different shapes, the sphericity is usually
cannot pass through the opening made by the m ono­ calculated from the arithmetic mean of the various
sized spheres, can occupy the volume enclosed by the sphericities.
monosized spheres without disturbing the basic pack­ Zou and Yu (1996) studied both loose and dense
ing. To make this possible, the smaller sphere has to be random packing of monosized nonspherical particles
in the position already during packing of the larger and found that the porosity was strongly dependent on
spheres. This critical diameter ratio is named the both the particle shape and the packing method. The
“critical ratio of occupation.” For the loosest square initial porosity of loose and dense random packings
packing, the ratio can be calculated as (Cumberland can be expressed as
and Crawford, 1987)
For loose random packing:
ab
1 = 1.366 (28)
iu 0.732 In Soi,cylinder = <l>5'58 exp[5.89(l - ()>)] In 0.40 (32)
For the tightest rhom bohedral packing, there are two In Soi.disk = ^°'6° exp[0.23(l - <J>)0-45] In 0.40 (33)
values:
F or dense random packing,
4 1
= 2.415 (29)
4l '0.414 In Eod,cylinder = exp[8.00(l - cj))] In 0.36 (34)
4_ 1
. = 4.444 (30) In £od,disk = 4>a63 exp[0.64(1 - c^)0'54] In 0.36 (35)
dM 0.225
The initial porosity of the nonspherical particles can be
where dM is the particle diameter required during approxim ated by the proper use of the packing results
critical ratio of occupation. of cylinders and disks shown in Eqs. (31) through (34)
and expressed as
4 PACKING OF NON SPH ERICA L PARTICLES
80 — ' ^0, cylinder Y " ”£o,disk (36)
L + It
There has been very little theoretical and experimental
work performed in this area owing to the complexity of where Ic is called the cylindrical index and / d, the disk
such a system. Oman and W atson (1944) conducted index. They are defined as follows.
experiments using particles of various shapes and 7C = |<(> - <j)c| (37)
found the voidage of the packing increased in the
following order: cylinders, spheres, granules, Raschig k = (38)
36 Yang

Table 3a Rough Estimate of Bed Void Fraction

Particle of pellet Normal charge Dense packed Multiplier for small tubes

Tablets 0.36 0.31 1 + 0.43 d J D .

Extrudates
Short 0.40 0.33 1 + 0.46dp/D
Long 0.46 0.40 1 + 0.46 dp/D
Spheres
Uniform 0.40 0.36 1 + 0 .4 2 Jp/D
Mixed sizes 0.36 0.32
Irregular 0.42 (average) 0.42 (average) 1 + 0.3 dp/D

Table 3b Voidage of Randomly Packed

Beds with Uniformly Sized Particles
Larger than 500 jam

Voidage

Sphericity Loose packing Dense packing

0.25 0.85 0.80

0.30 0.80 0.75
0.35 0.75 0.70
0.40 0.72 0.67
0.45 0.68 0.63
0.50 0.64 0.59
0.55 0.61 0.55
0.60 0.58 0.51
0.65 0.55 0.48
0.70 0.53 0.45
0.75 0.51 0.42
0.80 0.49 0.40
0.85 0.47 0.38
0.90 0.45 0.36
0.95 0.43 0.34
1.00 0.41 0.32

Source: A dapted from Brown, 1966.

where c|)c is an equivalent sphericity or aspect ratio Based on the work of Zou and Yu (1996), the
defined as the ratio of the maximum length to the Hausner ratio (HR) and the sphericity have the rela­
diameter of the circle having the same projected area tionship
normal to the maximum length, and c|)d, a sphericity
defined as the ratio of the shortest length to the dia­ H R = 1.478 • io -0-136^ (39)
meter of the circle having the same projected area.

4.1 Relationship Between the Hausner Ratio and the 5 FACTORS AFFECTING PACKING DENSITY
Sphericity IN PRACTICE

The Hausner ratio is defined as the ratio of tapped Particles, containers, and filling and handling methods
density to loose density, it is a measurement of the contribute to packing density in practice. The im por­
compressibility and cohesiveness of the powder. tant factors can be summarized as
Flow Through Fixed Beds 37

Figure 8 Relation between packing voidage and sphericity. (Adapted from Zou and Yu, 1996).

Particle shape, particle size, particle size distribu­

1 + o .6 ( '| -0 .6 (4 (40)
tion, particle coefficient of restitution, particle
surface properties (friction)
Container shape, container size, container surface Sc — [8 0.3(1 —s)] 1 + 0 . 6 ( |
properties (friction)
Deposition method, deposition intensity, velocity of
where s is the average voidage experimentally mea­
particle deposition
sured or the ratio of void volume to the total volume
Vibratory compaction, pressure compaction
of the packed bed. When the ratio of particle diameter
to the bed diameter is less than 0.02, the correction for
wall effect becomes negligible.
6 CORRELATIONS FOR FIXED BED BULK Figure 9 is a plot obtained by Benenati and
VOIDAGE Brosilow (1962) in a bed of spheres showing the radial
oscillations of voidage away from the containing
The voidage is the param eter most frequently cylindrical wall. The observation was confirmed by
employed to characterize the pressure drop in a fixed Propster and Szekely (1977). Experiments performed
bed. For accurate determination, experimental tech­ by Goodling et al. (1983) also indicated that for
nique such as the water displacement is usually used. uniformly sized spherical particles, the oscillations in
In the application of catalytic reactors, the catalysis voidage can be up to 5 particle diameters from the
packing is im portant not only for pressure drop but wall. The oscillations in voidage are down to 2 to 3
also for heat and mass transfer. Furnas (1929) was diameters from the wall, for a mixture of two spherical
probably the first to study the effect of wall on the sizes, and down to 1 particle diameter from the wall for
packing of the particles. He proposed the following a mixture of three particle sizes.
correlations for voidage close to the wall and at the Propster and Szekely (1977) also found that a
core. marked local minimum in voidage existed in the inter­
38 Yang

Figure 9 Schematic radial oscillations of voidage away from the containing cylindrical wall. (Adapted from Propster and
Szekely, 1977.)

facial region when small particles were placed on top of have correlations for fixed bed bulk voidage for com­
a bed of larger particles because of penetration of small monly used catalysts such as spheres, cylinders, and
particles into the interstices of large particles. rings. The following correlations are those proposed
However, when the large particles were placed on top by Dixon (1988). A more comprehensive review was
of the smaller particles, the local minimum in voidage conducted by Haughey and Beveridge (1969).
was not as pronounced. When the particle ratio of the F or spherical particles
lower layer to the upper layer is less than 2, relatively
little particle penetration was observed. Ready pene­ e = 0.4 + 0 . 0 5 0 ^ + 0 - 4 1 2 0 ^ ^ < 0.5 (42)
tration of small particles could be observed when the
ratio equals 6, corresponding to a critical ratio of
entrance of 6.464 for the tightest rhom bohedral pack­ e = 0.528 + 2 . 4 6 4 ^ - 0.5^ 0.5 < ^ < 0.536
ing (see Sec. 3). This observation will have practical
application in the iron blast furnace or cupola where (43)
alternate layers of coke and iron ore of different
particle sizes are charged into the reactors. 1 —0.667| - 1 ^ > 0.536
Fixed beds of very low tube-to-particle diameter
ratio have also been proposed and studied. For these (44)
reactors, the effect of wall and particle shape on bulk
voidage becomes im portant. It will be convenient to F or full cylinders
Flow Through Fixed Beds 39

Carm an (1937), and Ergun (1952). The discrete particle

£ = 0.36 + 0.10m + 0.7f ^ - * < 0.6 model assumes that the packed bed consists of an
assembly of discrete particles that possess their own
(45) boundary layer during the flow through the packed
bed. The prim ary developments are due to Burke
8 = 0.677 - 9 ( ^ - 0.625^ 0.6 < < 0.7 and Plummer (1928), Ranz (1952), Happel (1958),
Galloway and Sage (1970), and Gauvin and K atta
(46) (1973). Conceptually, the discrete particle model is clo­
ser to the physical description of the flow through the
s = 1 —0.7631 -g^ -£ > 0 .7 (47) packed bed, but the pipe flow analogy is historically
more widely employed. The correlations developed
through the pipe flow analogy are usually applicable
where <7pv is the diameter of a sphere having the same
for particles with sphericities larger than about 0.6.
volume as the cylinder.
The correlations developed through the discrete parti­
For hollow cylinders
cle model have wide applications, including particles of
sphericities less than 0.6.
(1 ^hc) —
1+2I| t 0-5)!(1145-^ 7.1 Darcy’s Law

(48) The theory of laminar flow through homogeneous por­

dyo > 0.5
ous media is based on a classical experiment originally
where subscripts he and sc denote the hollow and solid performed by Darcy in 1856. D arcy’s experiment is
cylinders respectively, and dyi and dyo are the inside described in Fig. 10. The total volume of the fluid
and outside diameter of the hollow cylinder, respec­ percolating through the fixed bed, Q, can be expressed
tively. in terms of the height of the bed and the bed area as
K A (h2 - AO
Q= (49)
h
7 FLOW THROUGH PACKED BEDS
where A is a constant depending on the properties of
There have been numerous investigations into flow the fluid and of the porous medium. The minus sign
through packed beds. Scheidegger (1961) critically indicates that the flow is in the opposite direction of
reviewed the earlier models. A more recent review is increasing height, h. D arcy’s law can be restated in
by Molerus (1993). There were two main approaches, terms of the pressure P and the density pf of the liquid.
the discrete particle model and the pipe flow analogy. Assuming that the liquid density is constant, we have
Both approaches gave reasonable predictions of pres­
k ' a (p 2 - P x+ pfgh)
sure drop of flow through packed beds of spherical and e = - (50)
near-spherical particles but were inadequate for beds
with particles substantially different from the spheres. and
The pressure drop through the packed bed is due not
P\ = Pfg{h\ - Z \ ) Pi = Pfg(h2 - zi) (51)
only to the frictional resistance at the particle surface
but also to the expansion and contraction of flow Equations (49) and (50) are equivalent statements of
through the interstices among the particles Darcy’s law. It is valid for a wide domain of flows. The
The most popular approach is the pipe flow analogy flow of newtonian fluid at low Reynolds number is
model, also called the capillary tube model or channel known to follow D arcy’s law. Thus it is valid for arbi­
model, which approximates the flow through the trary small pressure differentials for liquids. It has been
packed bed by the flow through a bundle of straight used to measure flow rates by determining the pressure
capillaries of equal size. Further refinement produced drop across a fixed porous bed. For liquids at high
the constricted tube model. In this model, an assembly velocities and for gases at very low and very high velo­
of tortuous channels of varying cross sections simu­ cities, D arcy’s law becomes invalid.
lates the varying dimensions and curvatures of pores Darcy’s law in its original form is rather restricted in
in the packed bed. The m ajor contributions following its usefulness. The physical significance of the constant
this approach include Blake (1922), Kozeny (1927), K \ known as the permeability constant and having
40 Yang

Datum 2 =0

Figure 10 Darcy’s experiment.

dimensions M “ 1L3T 1, has to be elucidated. The depen­ pf = pfCP) (54)

dence of K ' on the porous medium and on the liquid
and the other is the continuity equation
has to be separated for practical applications. Nutting
(1930) proposed that . 9pf
- e -^7 = div ( p f * ) (55)
r = ^ (52)
M- By eliminating all unknowns except the pressure, we
have
where k s is the specific permeability, having dimensions
L2, and the darcy is used as a unit for specific perme­ 9Pf A.
s ——= div (grad P — pf ~g) (56)
ability (1 darcy = 9.87 x 1CT9 cm2). By letting h dt
becomes infinitesimal, D arcy’s law can be written in
the differential form
7.2 Blake’s Correlation
Q. (grad P — pf( (53)
A ' Blakes (1922) may be the first one to suggest using a
modified dimensionless group incorporating the voi­
The differential form of D arcy’s law is by itself not dage, s, in a particulate system. He proposed the use
sufficient to determine the flow pattern in a porous of the interstitial velocity, instead of the superficial
medium for given boundary conditions, as it contains velocity, and the reciprocal of the total particle surface
three unknowns, ~q , P, and pf. Two more equations area per unit volume, instead of the particle diameter,
are required to specify the problem completely. One is as the characteristic length. The Reynolds num ber and
the dependence of the density on the pressure: friction factor proposed by Blake are
Flow Through Fixed Beds 41

Udppf or
(Re)B = (57)
n(l - 8)
3—n
(1 - 8)
AP cL fo e (67)
/ b — (58) (Re)
L Pf U^g (1 s)
These dimensionless groups have been used ever since At low Reynolds numbers, where the pressure drop
as the basis for the pipe flow analogy in almost all does not depend on the fluid density, n — 1, and
packed bed correlations.
From dimensionless analysis, the pressure drop ( 1 - s )2
through a packed bed of particles can be obtained as /o c (68)
s3(Re)p
AP •~n rJl—1 ( ji—3
-oc {i pf (59) At high Reynolds numbers where the viscosity is not
im portant, n — 2, and
The interstitial fluid velocity in the average direction of
fluid motion is used here. The D is a length analogous (1 - 8)
/o c (69)
to the hydraulic radius of a conduit and is defined as
Mean cross-sectional area of flow Because of the large dependence on the voidage, a 30-
Q = channels through bed
to-70-fold increase in f is predicted as the voidage
Mean wetted perimeter of flow channels changes from 0.3 to 0.7 in the turbulent and viscous
(60) ranges, respectively. The expression found above for f
at low Reynolds numbers has been experimentally
Multiplying the num erator and the denom inator by L,
proved to be correct. At high Reynolds numbers,
the height of the bed, we have
however, the expression seems to predict a somewhat
(Total bed volume) x (voidage) stronger dependence of the friction factor on the
D= (61)
(Total bed surface) voidage than is actually found experimentally. For a
single particle, voidage becomes 1 by definition, and
or the Blake’s Reynolds num ber approaches infinity.
This is not surprising, since the Blake analogy based
n ___ £ __ £^P (62)
- S 6(1 - 8) on a capillary flow analogy breaks down in this
range, because such an analogy has no physical mean­
where S is the total surface of solids per unit bed ing.
volume assuming spherical particles, it can be
expressed as
7.3 The Brownell and Kats Correlation
6( 1 - 8)
S = (63)
A nother correlation incorporating the voidage is that
by Brownell and Katz (1947). They introduced a
Equation (63) is essentially similar to Eq. (4).
Reynolds num ber defined as
Substituting into Eq. (59), we obtain
Udppf
f< x p2- " p n
f lun{- — —A a; - 3 (64) (Re)pe = (70)
J-j 8
F or conventional pressure drop through a packed bed where the exponent m depends on the ratio of the
of particles, n — 2. Equation (64) becomes sphericity to the porosity and ranges from 2 to 20.
The use of dimensionless analysis in correlating the
ap __2fP(u2 (65) data is justified as long as the identified dimensionless
L gdp
groups represent true similarity. The Brownell and
The friction factor, / , is proportional to Kats correlation clearly does not represent true simi­
larity, because it predicts that the Reynolds num ber
_ ji.. 2 — n pf
jr oc ^ n —2 m —2 j j n —2
dp U *0
(66)
decreases with increasing voidage, contradicting the
,3 experimental findings.
42 Yang

7.4 The Carman and Kozeny Correlations this type is by Ergun (1952) employing Blake’s defini­
tion of drag coefficient and the Reynolds number, as
Carm an (1937) studied extensively the fluid flow shown in Eqs. (57) and (58).
through various packings in the viscous range and
found that AP gdp§ s (1 - e )
= 75 •0.875 (74)
~T 2pf U2 " (1 —e) c(>(Re)p
90(1 - s)2
/ = for 0.26 < s < 0.S (71)
e3(Re)p F or general applications, including irregular particles,
He also found that Eq. (71) can be applied to other the Ergun equation shown in Eq. (74) is expressed with
regular shapes as long as their surfaces can be deter­ sphericity by substituting §dp for dp, where dp is the
mined accurately and provided that dp is expressed as diameter of the irregular particle obtained by particle
6Vp/ S p. Vp is the particle volume, and Sp is the surface measurement techniques, such as sieving or the Coulter
of the particle. F or mixtures of various sizes and counter, described in Chapter 1. It can be seen that the
shapes, Vp and S p should be taken as averages of all Ergun equation reduces to the B lake-K ozeny-Carm an
particles in the bed. equation at low Reynolds number, and at high
The Kozeny (1927) equation is usually expressed as, Reynolds number, to the Burke-Plum mer equation
by combining Eqs. (65) and (71), for turbulent flow. M any extensions and modifications
of the Ergun equation have been proposed, such as
AP 180(1 & f\iU 5(1 - z f \ i U that by Handley and Heggs (1968), Hicks (1970),
(72) Tallmadge (1970), Leva and coworkers (1947), and
~ L : ' g z \ V J S v)2
Rose and Rizk (1949). Among them the correlations
Kozeny derived the equation by assuming that a suggested by Tallmadge and Leva et al. are for high
granular bed is equivalent to a group of parallel simi­ Reynolds numbers, where the Ergun equation fails to
lar channels, commonly known as the pipe flow ana­ fit the data well.
logy. He started from the general equation for
streamline flow through a uniform channel and 7.5.1 Use of Ergun Equation to Determine
assumed that the hydraulic radius of the channel Sphericity Factor
is e/S . This is equivalent of assuming that the total
Subram anian and Arunachalam (1980) suggested a
internal surface and the total internal volume of the
technique making use of the Ergun equation to deter­
group of parallel similar channels are equal to the
mine the sphericity of irregular particles. For very low
particle surface and the void volume of the bed. To
flow rates, the viscous forces predominate, and the
find the ratio of the surface and volume of the parti­
Ergun equation in Eq. (74) can be rearranged explicitly
cles, Carm an proposed to measure the pressure drop
for the sphericity as
through the bed of the irregular particles experimen­
tally. The specific surface, S p/ V p, can then be evalu­ 1/2
ated from 1501,(1 - s)2fW(</p2e3pfg)
<l> = (75)
(1 - L / H x)
g AP z3
(73)
5(1 - &) \xUL with

The pipe flow analogy breaks down beyond the creep­

ing flow range. Thus beyond the creeping flow range, (76)
purely empirical models must be employed, the only
remaining link with the pipe flow is the structure of the By carrying out the experiment with a packed bed o f
dimensionless groups. irregular particles and by draining the liquid in a
lam inar flow from the height H 0 to 77), the time, t,
7.5 The Ergun Correlation required can be determined. Substituting into
Equations (75) and (76), the sphericity of the irregular
There are many other correlations such as that by particles can be obtained. A simple and accurate
Oman and W atson (1944) and Happel’s correlation device was described in Subram anian and
(1958). The m ost widely used empirical correlation of Arunachalam (1980).
Flow Through Fixed Beds 43

7.6 Modified Ergun Equation A P d pe.4.8 17.3

Jp —
-0.336 (82)
PfU2L{\ —s) (Re)p
Gibilaro et al. (1985) modified the Ergun equation and
proposed an alternative pressure drop equation on the Equation (82) gives a significantly better representa­
basis of theoretical considerations: tion for data at higher Reynolds numbers, where
Ergun equation consistently overestimates the
AP = + 0.336 ) l( i _ s)g-4-8 (77) observed friction factor. Wentz and Thodos (1963)
.(Re). also proposed a general equation for friction factor
for packed and distended beds of spheres:
The proposed equation compared well with published
experimental data obtained from high-voidage fixed 0.351
beds of spheres; it represented a significant improve­ /w = 0.05
(83)
Re' 1.2
ment over that of Ergun. The equation is a com bina­
tion of two equations, one for the lam inar regime and Equation (83) is good for voidage between 0.354 and
the other for fully turbulent flow. The laminar flow 0.882 and Reynolds numbers between 2,600 and
regime equation is derived to m atch the Blake- 64,900.
Kozeny equation at s equal to 0.4 and can be expressed
as 7.8 Drag Coefficient for a Particle in an Array

AP = The drag coefficient for a particle in an array was also

^ + o -336r ^ ( i - e)s“ 4'8 (78) derived by Gibilaro et al. (1985) as

Equation (78) gives an accurate prediction of fluidized

Qm — ^ / P£ 3-8 — O d8 3‘^ (84)
bed expansion characteristics for the lam inar regime
and applies equally well to fixed and suspended particle
systems. where CD = drag coefficient for a particle in an
The equation for the fully turbulent flow regime was infinite fluid.
also derived as Equation (84) provides accurate quantitative predic­
tions of particulate fluidized bed expansion character­
istics in both lam inar and turbulent flow regimes. For
A P = 0.336 9{V L (l - s)s“4-8 (79)
the intermediate flow regime, only a qualitative trend
was observed. Equation (84) is slightly different in
The constant 0.336 is the result of matching s = 0.4 in
dependence of voidage from the drag coefficient sug­
the Burke-Plum m er (1928) equation, Eq. (80), which
gested by Wen and Yu (1966) as shown here:
describes the norm al packed bed pressure drop well:
-4 .8
p{U2L ( l - s ) 0)1 — 0 ) 8 (85)
A i> = 1.75 (80)

Equation (79) well represents the steady-state expan­ 7.9 The General Correlation by Barnea and Mednick
sion characteristics of a turbulent regime in fluidized
beds. A general correlation for the pressure drop through
fixed beds of spherical particles, based on a discrete
7.7 General Friction Factor Correlations particle model corrected for particle interaction, was
proposed by Barnea and Mednick (1978). They
The pressure drop equations can be converted to extended the standard CD versus Re curve for single
general friction factor correlations as follows. From spheres to multiparticle systems by incorporating
the Ergun equation, the friction factor correlations proper functions of the volumetric particle concen­
will be tration. The modified drag coefficient and Reynolds
number they suggested are
AiVpS 150(1 - s )
/e = + 1.75 (81)
pf f/2L(l —8) (Re)p
1
(Re)^ = (Re)p ( 86)
The friction factor correlation resulting from the
equation by Gibilaro et al. (1985) becomes eexP[^3F^]
44 Yang

8/ e 2 ss = 0.8038(1 —s) (94)

(C d )* =
3(1 - e)[l + K{\ - e)1/3] where Ks is the ratio of the mean projected area of a
A P gdp particle (sphere, cylinder, ellipsoid) to that of a sphere
(87) of the same volume, and Kd is the ratio of the mean
L ' 2pf U2
projected area of a prism to that of a disc with the base
and having the same area as the larger face of the prism.
Galloway and Sage (1970) reported that the stag­
4T nant voidage, es, varied from 0.172 to 0.157 when the
(CD) , - 0.63- (88)
(Re), Reynolds num ber was varied from 10,000 to 35,000
during their experiments.
where K is a constant
The correlation allows the prediction of pressure 7.11 Permeability of Packed Beds
drop or drag in single-particle and multiparticle system
with a single curve. The correlation provides good Permeability of a packed bed can usually be estimated
agreement with data in the creeping flow and inter­ from the rearranged form of K ozeny-C arm an equa­
mediate regimes. Deviation is observed in the turbulent tion:
regime. For the data in the distended bed, the correla­
tion also provides a good fit in the highly turbulent R2 75(1 - e)2
range. The range of applicability may be extended by (95)
k 2s3
the application of average particle size definitions,
shape factors, and wall effect correlations. where R is the radius of particle and k is the perme­
ability.
7.10 The Concept of Stagnant Voidage Equation (95) has been found to provide a good
estimation of permeability for packed beds of voidage
Happel (1958) introduced the interesting concept of a between 0.26 and 0.80. Carm an (1956) also found th at
stagnant voidage, a part of the voidage in the packed if the hydraulic radius was used to replace the particle
bed that is occupied by the wake of the particles and is radius in Eq. (95), the equation was also good for
thus not available for the fluid flow. The concept is mixtures of different particle sizes. The hydraulic
primarily employed in the discrete particle model but radius for a bed of spherical particles can be calculated
may be useful for other modeling effort. from
Kusik and Happel (1962) derived this equation for
estimating the stagnant voidage for the packing of (96)
6( 1 - 8)
spheres:
cs = 0.75(1 - e)(e - 0.2) (89) To account for wall effect, M ehta and Hawley (1969)
modified the equation for hydraulic radius to give
Gauvin and K atta (1973) suggested a slightly different
equation for a bed packed with spheres: e<7n
Rh = (97)
6(1 - s ) M
es = 1.6(1 —e)(e —0.2) (90)
and
They also proposed equations for packing of other
isometric particles. 4 cL
M = 1+ (98)
6D( 1 - s)
F or packing of cylinders,
where D is the diameter of packed column.
ss = 1.95Zs( e - 0 .2 ) ( l - s ) (91) The permeability of a packed bed with polydisperse
For packing of ellipsoids, spherical or nonspherical particles can also be esti­
mated using Eq. (95) with particle size calculated
8S = 2.5ifs(s - 0.2)(1 - s) (92)
through the harmonic mean or the Sauter mean if
For packing of prisms and wafers, the size distribution is not very broad. For wide size
distributions, Li and Park (1998) have proposed equa­
ss = 0.9 3 ^8 (1 - s) (93)
tions for calculating the permeability for both spherical
For wood chips, and nonspherical particles.
Flow Through Fixed Beds 45

8 GAS VELOCITY DISTRIBUTION IN PACKED because it evened out the nonuniformities introduced
BEDS upstream of the bed.
Cohen and Metzner (1981) indicated that for new-
Because of the uneven distribution of voidage across tonian fluids, wall effect corrections could be neglected
the packed bed created during the filling process, the if the bed-to-particle-diameter ratio is larger than 30.
radial gas velocity distribution in the packed bed is F or nonnewtonian fluids, it is 50. They also proposed a
neither parabolic, as in an empty pipe, nor uniform. model dividing the bed into three regions— a wall
As pointed out earlier, the voidage close to the wall is region, a transition region, and a bulk region. The
usually higher, owing to the wall effect; the gas flow wall region extends a distance of one particle diameter
tends to be higher close to the wall. Schwarz and Smith from the wall. The transition can be up to six particle
(1953) measured the radial gas velocity profiles in a diameters from the wall where appreciable voidage
2 in. pipe and a 4 in. pipe filled with 1/8 in. cylinders oscillation occurs, as discussed in Sec. 6. In the bulk
and found it indeed was the case (see Fig. 11). When region, the remainder of the bed region, the voidage is
the packed bed diameter becomes larger, the influence essentially constant. For packed beds of small b e d : -
of the wall effect decreases. Thus for a large-diameter t o : diameter ratios, the use of the single-region model
bed, the assumption of uniform radial velocity distri­ based on the average voidage tends to over-predict
bution can be a first approximation. Theoretical pre­ the average mass flux. The triregional model proposed
diction of radial velocity distribution is not possible for by Cohen and M etzner (1981) fitted the experimental
a randomly packed bed. The radial gas distribution data much better. For bed-to-diameter ratios less than
can also be profoundly affected by the design of gas 30, the transition region represent the largest fraction of
inlet and outlet regions of the packed bed due to the total bed cross-sectional area. The mass fluxes in the
bypassing. Szekely and Poveromo (1975) employed wall and transition regions were estimated to be larger
the vectorial differential form of the Ergun equation than that of the bulk region by as much as 50% for
to predict the flow m aldistribution in a packed bed. fluids with power indexes of 0.25 and by 10% for fluids
The experimental measurements were found to be in with power indexes of 1. Saunders and Ford (1940),
reasonable agreement. They also found that a uni­ employing a pitot tube, found the velocity in a ring
formly packed bed with a height/diameter ratio larger about one particle diameter from the wall about 50%
than one could also be used as a flow straightener higher than the bulk gas velocity. Schwartz and Smith

Center Wall
Radial Position

Figure 11 Radial gas velocity profiles in a 50.8 mm and 101.6 mm diameter bed. (Adapted from Schwarz and Smith, 1953.)
46 Yang

(1953), in a bed with a bed-to-particle-diameter ratio qp = hsap(Ts - T) (99)

less than 30, also found a peak velocity at about one
particle diameter away from the wall approximately The heat transfer coefficient can be evaluated from the
30% to 100% higher than the bulk velocity. For bed- equation
to-particle-diameter ratio larger than 30, the bulk
M P
region has the largest cross-sectional area and thus = 2 + l . l P r 1/3(Re);\0.6 accuracy to 25%
has the largest contribution to the flow. The wall region
is actually larger than the bulk region for packed beds (100)
with the bed to diameter ratios less than 18. This phy­
sical division is an im portant consideration for packed where (Re)p = dpUpf /\i and dp is the diameter of a
bed design in the laboratory and for attempting to use sphere or an equivalent sphere. U is the superficial
the laboratory data for scale-up. Since the voidage in velocity in the packed bed.
the bulk region is constant and independent of the bed-
to-particle-diameter ratio, its average flux decreases 9.1.2 H eat Transfer Through Wall— One-
with increasing bed to a particle diameter ratio. Dimensional Model—Axial
F or a cylindrical vessel with spherical particle packing,

q = h1A l( T - T w) (101)
9 HEAT TRANSFER IN PACKED BEDS
where = inside surface of the cylindrical vessel and
As for the solid-fluid heat transfer coefficient for flow
Tw = wall tem perature of the fluid at axial position of
through a randomly packed bed, large variations of up
interest. F or the homogeneous model, T of fluid and o f
to two- to five-fold have been observed (Barker, 1965).
bed are assumed identical.
The primary difficulties have been the attem pt to
model the packed bed with simple average parameters h{D 6dr
k -2.26(Re)“'8JPr0-33e x p ^ - ^
that describe the complex local variations in packing (102)
voidage and the effects of particle shape, distribution, <L
7p <
20 < (Re)p _ 7600 and 0.05 _< D < 0.3
and velocity. Heat transfer in gas-solid packed bed
systems was critically reviewed by Balakrishnan and F or a cylindrical vessel with cylindrical particles
Pei (1979). The overall heat transfer in the packed packing,
bed is quite complex and consists of the following
mechanisms: (1) the conduction heat transfer between 6dl
l.40(Re)p95/ >;'0'33 exp
particles in both radial and axial directions, (2) the ka ^ "~D
convective heat transfer between the bed particle and
the flow gas, (3) the interaction of mechanisms (1) and 20 < (Re)p < 800 and 0.03 < J - < 0.2
(2), (4) heat transfer due to radiation between the bed
particles, between the particles and the flowing gas, (103)
and between the flowing gas and the bed wall, and
(5) heat transfer between the bed wall and bed parti­
cles. M ost of the studies in the literature were directed 9.1.3 Wall H eat Transfer Coefficient—Two-
Dimensional
toward developing correlations for the total heat trans­
fer rate in terms of dimensionless parameters such as
Reynolds number. q = Aw^ i(T R - Tw) (104)

9.1 Rase Heat Transfer Correlations were TR = tem perature at inside radius of the vessel.
F or spherical particle packing,
Based on the recommendations by Rase (1990), the
hwdp
heat and mass transfer correlations to be used for = 0.19(Re)p0.79 iV
D 0.33

packed bed calculations are summarized here.

dn
9.1.1 Particle to Fluid Heat Transfer 20 < (Re)p < 7600 and 0.05 < < 0.3
D
Heat transferred to a single particle can be expressed as (105)
Flow Through Fixed Beds 47

F or cylindrical particle packing, For voidage values between 0.35 and 0.5, Eq. (112)
gives a range of Nusselt numbers between 5 and 10.
h\ud-t
Wwp
: 0.18(Re)p0.93 P0 r 0.33 For fluid heating by percolating through a particle
array,

20 < (Re)p < 800 and 0.03 < < 0.2 (Pe)d^p _ ^ 0:1(1 - s ) z
7.83- (113)
(106) Lr,

where
9.1.4 Effective Radial Thermal Conductivity
Pf ^pg^p
(Pe)d = (114)
5\\ or /Jr=B (107)

d J J pf Equation (113) indicates that the heat transfer between

, £ = 0.27 500 < ? < 6000 the particles and the percolating fluid is very fast.
L 1 —8 |l(l - 8)
Assuming s = 0.4 and L p = ndp and substituting into
Eq. (113) we have
and 0.05 < -^ < 0.15 (108)
D
n = 0.14(Pe)d (115)
9.2 Heat Transfer Correlations Recommended by
For (Pe)d < 10, Eq. (115) implies that just one particle
Molerus and Wirth (1997)
layer is enough to heat the percolating gas to the sur­
face tem perature of the particles.
In a recent book by Molerus and W irth (1997), the
recommended heat transfer correlations for packed
beds can be summarized as follows. For fully devel­ 9.3 Analytical Models for Heat Transfer with
oped laminar flow, an approxim ation formula for Immersed Surfaces
the mean Nusselt number, derived from the pipe flow
analogy, was proposed as When the heat removal is entirely due to the flowing
gas and there is no convective particle movement,
G abor (1970) proposed the following simple model.
(N u)d = ^ 2 p = B .6 6 3 + 1.613(Pe)p 2p
F or heat transfer from a flat plate of length L h
immersed in the packed bed,
(Pe)p7)p
for 0.1 < 104
dT _ k P d2T
(116)
(109) dz CpgC dy2
where (Pe)p is the pipe flow Peclet number and L p is
with the boundary conditions
the pipe length, equivalent to the bed depth. The Dp
and (Pe)p can be expressed as
T = Tw at y = 0 and 0 < z < Lh
2s
D„ = -----------«„ (110) (117)
p 3(1 - s) p
T=T0 at y = 00 (118)
(Pe)p = (ill) T=Tn z = 0
at (119)

where U is superficial flow velocity. F or heat transfer from a cylindrical heater of length L h
For heat transfer from a single particle inside a and radius R y immersed in the packed bed,
particle array,

^ = hdp = 5 4 9 P s dT ke 32T 1 3T )
( 112 ) (120)
\ s dz CpgG 1 d r 2 r dr j
48 Yang

with the boundary conditions dT l a 2^

Pf U\ C pg ~~Z~~ — at z> 0
dz dr2 r dr2
T = Tw at r = Rv and 0 < z < Lh
( 121) and Rt < r < j i?t + ”
T = T0 at r = oo ( 122) (128)
T = Ta at z= 0 (123)
In the region outside the wall region, region 2, the
governing equation is
The solutions for the average heat transfer coefficients
for the flat plate and cylinder cases were found by
G abor (1970). n T j r dT*
PfF^Cpg ^ - t/^e2 j ^2 T m
1927V (129)
For flat plate case, dr2 r dr2

Heat transfer in region 1 was assumed to take place by

, _ /4 k eCpgG
(124) three different mechanisms, neglecting the heat transfer
aV V" through radiation. The three mechanisms are (1) heat
transfer by turbulent diffusion in the radial direction,
F or the cylindrical heater case, £a, (2) heat transfer by molecular conduction in the
fluid boundary layer, k h, and (3) heat transfer
, 4 k eCpgG 1 ke through the thin film near particle contact points,
(125)
av V71 2i?y k c. Mechanisms (1) and (2) operate in series while
mechanism (3) operates in parallel. The resulting
The effective thermal conductivity of the packed bed, effective bed conductivity can thus be expressed as
ke, can be estimated from
^el K ____ 1
(130)
k e = k° + 0 A( CpgdpG) (126) kg kg (kg/k^) T (kg/ k h)

The effective therm al conductivity of the packed bed = 0.045(Re)! Pr (131)

fCcy
expressed in Eq. (126) includes two terms, the conduc­
tivity term with no gas flow, k°, and the convective where
term. The factor 0.1 was recommended by Yagi and
Kunii (1957) for spherical particles. The conductivity cLU i pf , Cngu
(Re)! = p 1Pf and Pr =
term with no gas flow can be calculated below, follow­ p kg
ing that suggested by Swift (1966) for orthorhom bic
particle packing with a voidage of 0.395. The gas velocity in region 1, the wall region, is taken to
be 50% larger than that in region 2, the bulk region.
K = 0.9065 0.5
.(i/k g ) ( l / k s). k o —ka = 1.7 (Re) (132)
0.0935k„
K , (i ^w) (133)
(127)
k r _£w + 2(t)w + ( 2 /3 ) ( V k s)
Botterill and Denloye (1978) conducted an even more
where
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 •= — K — — 1134')
4 \ \ n K — {K — \ ) / K I 3K k„y ’
of higher voidage within half a particle diameter
from the immersed cylindrical surface, and the region (1 -s)(0 .7 2 9 3 + 0.5139 Y)
of constant voidage outside the wall region. The radial Ew = 1 - (135)
1+ Y
transfer of heat for the wall region (region 1) is
governed by where Y = cL/2R.
Flow Through Fixed Beds 49

The effective conductivity in region 2 can be evalu­

ated as ( 145)

= -f- + 0.075(Re)2Pr for (Re)2 < 100 F or more general equations in terms of Re, tortuosity,
fC-pr /Cq and £, see Wen and F an (1975).
(137)

^e2 A? NOM ENCLATURE

+ 0.125(Re)2Pr for (Re)2 > 100
A inside surface of the cylindrical vessel
(138) ap surface area of a particle
CD drag coefficient of a particle in an infinite fluid
Equations (137) and (138) need to be solved num eri­ Cdi drag coefficient of a particle in an array
cally. The agreement with experimental data was (CoXj) drag coefficient for multiparticle systems
reported to be within 25%. The prediction of Cj concentration of component j in bulk region
G abor’s model was consistently higher than the r'-'is concentration of component j at particle
experimental data. The theoretical predictions show surface
an almost linear dependence of the heat transfer coef­ r heat capacity of gas at constant pressure
ficient on the particle Reynolds number. Because of D column diameter
D diameter of the cylindrical vessel
its simplicity, G abor’s model may be used as a first
D defined in Eq. (60)
approximation.
dh diameter of the large particles
dc diameter of a circle equal in area to the
9.4 M ass Transfer in Packed Beds projected area of the particle when resting on
its larger face
Again based on recommendation by Rase (1990), the Dc diameter of the smallest circle circumscribing
following correlations for a particle to fluid mass the projection of the particle
transfer have accuracy to within 25% dM particle diameter required during critical ratio
Mass transfer to a single particle can be calculated of occupation, see Sec. 3
from equivalent particle diameter
Pe
packing equivalent particle diameter
wp = k siap(Cjs - Cj) (139) d■^pv diameter of sphere with equivalent volume of
the cylinder
The mass transfer coefficient can be obtained from diameter of the small particles
dve volume equivalent particle diameter
— — 2 + L lS c 1/3 Re06 (140) inside diameter of the hollow cylinder
P ffy outside diameter of the hollow cylinder
f friction factor
with accuracy to 25%. 7b friction factor proposed by Blake, Eq. (58)
The effective radial diffusivity is /e general friction factor from Ergun (1952)
/p general friction factor from Gibilaro et al.
a f dC i (1985)
7Vjr = (141)
/w general friction factor from Wentz and
Thodos (1963)
efl-j. i 0.38
(142) gravitational acceleration
uL m ~Re vector form of gravitational acceleration
G mass flow rate of gas
For dp/ D > 0.1, G an empirical constant
hav average heat transfer coefficient, averaged over
m = 11 for Re > 400 (143) the length of the heater
m = 57.85 - 35.36 log Re + 6.68(log Re)2 h heat transfer coefficient at inside surface of the
cylindrical vessel
for 20 < Re < 400 (144) heat transfer coefficient to a single particle
K heat transfer coefficient at the wall
For dp/D < 0.1, divide $ r calculated from the above A fluid column height 1
equation by hi heat transfer coefficient in region 1
50 Yang

h2 fluid column height 2 R radius of sphere

h heat transfer coefficient in region 2 Rt radius of the heater
h cylindrical index Rh hydraulic radius
h disk index Re Reynolds number, = D U pf/ja
k permeability (Re)s Reynolds number proposed by Blake, Eq. (57)
k an empirical constant (R e)p Reynolds number based on particle diameter,
K ratio of thermal conductivity of solid and gas, dv U pf/n
K = ke/ k g (Re)pe Reynolds number proposed by Brownell and
K a constant Kats, and defined in Eq. (70)
K' permeability constant (Re)* Reynolds number for multiparticle systems
ks heat transfer by turbulent diffusion in radial (R e)i Reynolds number in region 1
direction (R e), Reynolds number in region 2
h heat transfer by molecular conduction in fluid Rv radius of cylindrical heater
boundary layer specific particle surface area; total particle
kc heat transfer through the thin film near surface area per unit volume of bed
particle contact point Sc Schmidt number, = p/ffjpf
Kd ratio of mean projected area of a prism to that SP particle surface
of a disc with the base having the same area T temperature
as the larger face of the prism T0 initial gas temperature
h effective thermal conductivity of packed bed Tr temperature at inside radius of the vessel
hi effective bed thermal conductivity in region 1 Ts
1 temperature of a single particle
hi effective bed thermal conductivity in region 2 Tw temperature at the heater wall
K effective thermal conductivity of the packed Rw temperature at wall
bed with zero gas flow Ti temperature in region 1
h thermal conductivity of gas T2 temperature in region 2
h specific permeability u superficial fluid velocity
h thermal conductivity of solid Ui superficial fluid velocity in region 1
ksj mass transfer coefficient of component u2 superficial fluid velocity in region 2
Ks ratio of mean projected area of a particle to V specific volume of solid material,
that of a sphere of the same volume F = l/(l-e )
L bed height Vb volume of big particles
Lh length of heat transfer surface Rp particle volume
h pipe length h volume of small particles
m exponential coefficient volume fraction of big particles
mp mass transfer rate Y mu
yx - volume fraction of large particles where the
n coordination number voidage of the binary mixture is the minimum
n an exponential coefficient *3 volume fraction of small particles
number of spheres per unit volume of bed y horizontal coordinate
Nu Nusselt number Z vertical coordinate
(N u )d pipe flow Nusselt number Z axial distance along the heat transfer surface
(Pe)d pipe flow Peclet number defined in Eq. (114) Z\ vertical coordinate at point 1
(Pe)p pipe flow Peclet number defined in Eq. ( I ll) Zl vertical coordinate at point 2
P pressure diffusivity
Px pressure at point 1 p layer spacing
Pi pressure at point 2 fluid viscosity
AP pressure drop pf fluid density
Pe Peclet number Pp particle density
q heat transfer rate Wadell’s sphericity, the ratio of the surface
vector quantity of volumetric flow rate per areas of a volume equivalent sphere and the
unit bed area actual particle
volumetric fluid flow rate an equivalent sphericity or aspect ratio defined
heat transfer rate to a single particle as the ratio of the maximum length to the
radial coordinate diameter of the circle having the same
diameter ratio of small particle to large projected area normal to the maximum length
particle
structural ratio defined in Eq. (24)
Flow Through Fixed Beds 51

= a sphericity defined as the ratio of the shortest PC Carman. The flow of Gases Through Porous Media. New
length to the diameter of the circle having the York: Academic Press, 1956.
same projected area Y Cohen, AB Metzner. Wall effects in laminar flow of fluids
8 = voidage of the fixed bed through packed beds. AIChE J 27:705-715, 1981.
Sav = average voidage of the binary mixture JM Coulson. The flow of fluids through granular beds: effect
£b = voidage of a packed bed with large particles of particle shape and voids in streamline flow. Trans Inst
alone Chem Eng 27:237-257, 1949.
U = voidage at core region DJ Cumberland, RJ Crawford. The Packing of Particles.
She = voidage of the hollow cylinder packing Amsterdam: Elsevier, 1987, p 60.
Smin = the minimum voidage of the binary mixture H Darcy. Les Fontaines publiques de la ville de Dizon. Paris:
^0 = initial packing voidage Dalmont, 1856.
So, cylinder = voidage of initial packing of cylinders AG Dixon. Correlations for wall and particle shape effects on
So,disk = voidage of initial packing of disks fixed bed bulk voidage. Can J Chem Eng 66:705-708,
Sol,cylinder = voidage of initial loose packing of cylinders 1988.
Sol,disk = voidage of initial loose packing of disks S Ergun. Fluid flow through packed columns. Chem Eng
Sod, cylinder = voidage of initial dense packing of cylinders Prog 48:89-94, February 1952.
Sod,disk = voidage of initial dense packing of disks HJ Finkers, AC Hoffmann. Structural ratio for predicting
Ss = voidage of a packed bed with small particles the voidage of binary particle mixtures. AIChE J
alone 44:495-498.
Ss — stagnant voidage CC Furnas. Flow of gases through beds of broken solids.
Ssc = voidage of the solid cylinder packing Bull 307. US Bureau of Mines, 1929.
sw = voidage at wall region JD Gabor. Heat transfer to particle beds with gas flows less
8 = average voidage than or equal to that required for incipient fluidization.
Chem Eng Sci 25:979-984, 1970.
TR Galloway, BH Sage. A model of the mechanism of trans­
port in packed, distended, and fluidized beds. Chem Eng
REFERENCES Sci 25:495-516, 1970.
WH Gauvin, S Katta. Momentum transfer through packed
E Abe, H Hirosue, A Yokota. Pressure drop through a beds of various particles in the turbulent flow regime.
packed bed of binary mixture. J Chem Eng Japan AIChE J 19:775-783, 1973.
12:302, 1979. L Gibilaro, RDI Delice, SP Waldram, PU Foscolo.
AR Balakrishnan, Pei DCT. Heat transfer in gas-solid Generalized friction factor and drag coefficient correla­
packed bed systems, a critical review. Ind Eng Chem tions for fluid-particle interactions. Chem Eng Sci
Process Des Dev 18:30-40, 1979. 40:1817-1823, 1985.
JJ Barker. Heat transfer in packed beds. Ind Eng Chem JS Goodling, RI Vachon, WS Stelpflug, SJ Ying, MS
57(4):43-51, April 1965. Khader. Radial porosity distribution in cylindrical beds
E Barnea, RE Mednick. A generalized approach to the fluid packed with spheres. Powder Technol 35:23-29, 1983.
dynamics of particulate systems part III: general correla­ D Handley, PJ Heggs. Momentum and heat transfer mechan­
tion for the pressure drop through fixed beds of spherical isms in regular shaped packings. Trans Inst Chem Eng
particles. Chem Eng J 15:215-227, 1978. 46:T251-T264, 1968.
RF Benenati, CB Brosilow. Void fraction distribution in beds J Happel. Viscous flow in multiparticle systems: slow motion
of spheres. AIChE J 8:359-361, 1962. of fluids relative to beds of spherical particles. AIChE J
EC Blake. The resistance of packing to fluid flow. Trans Am 4:197-201, 1958.
Inst Chem Eng 14:415, 1922. DP Haughey, GG Beveridge. Local voidage variation in a
JSM Botterill, AOO Denloye. A theoretical model of heat randomly packed bed of equal-sized spheres. Chem Eng
transfer to a packed or quiescent fluidized bed. Chem Eng Sci 21:905-916, 1966.
Sci 33:509-515, 1978. DP Haughey, GG Beveridge. Structural properties of packed
GG Brown. Unit Operations. New York: John Wiley, 1966. beds—a review. Can J Chem Eng 47:130-140, 1969.
LE Brownell, DL Katz. Flow of fluids through porous media G Herdan. Small Particle Statistics. New York: Academic
I. Single homogeneous fluids. Chem Eng Prog 43:537- Press, 1960.
548, October 1947. RE Hicks. Pressure drop in packed beds of spheres. Ind Eng
SP Burke, WB Plummer. Gas flow through packed columns. Chem Fundam 9:500-502, 1970.
Ind Eng Chem 20:1196-1200, 1928. AC Hoffmann, HJ Finkers. A relation for the void fraction
PC Carman. Fluid flow through granular beds. Trans Inst of randomly packed particle beds. Powder Technol
Chem Eng 15:150, 1937 82:197-203, 1995.
52 Yang

J Kozeny. Uber Kapillare Leitung des Wassers im Boden. CE Schwartz, JM Smith. Flow distribution in packed beds.
Sitzungsber Akad Wiss Wien 136:271, 1927. Ind Eng Chem 45:1209-1218, 1953.
CL Kusik, J Happel. Boundary layer mass transfer with GD Scott. Packing of spheres Nature 188:908-911, 1960.
heterogeneous catalysis. AIChE J 8:163, 1962. HY Sohn, C Moreland. The effect of particle size distribution
M Leva. Pressure drop through packed tubes: Part I. A on packing density. Can J Chem Eng 46:162-167, 1968.
general correlation. Chem Eng Prog 43:549-554, 1947. P Subramanian, Vr Arunachalam. A simple device for deter­
Y Li, CW Park. Permeability of packed beds filled with poly- mination of sphericity factor. Ind Eng Chem Fandam
disperse spherical particles. Ind Eng Chem Res 37:2005- 19:436-437, 1980.
2011, 1998. DL Swift. The thermal conductivity of spherical metal pow­
RK McGeary. Mechanical packing of spherical particles. J ders including the effect of an oxide coating. Int J Heat
Am Ceram Soc 44:513-522, 1961. Mass Transfer 9:1061-1074, 1966.
D Mehta, MC Hawley. Wall effects in packed columns. Ind J Szekely, JJ Poveromo. Flow maldistribution in packed
Eng Chem Proc Design Dev 8:280-282, 1969. beds: a comparison of measurements with predictions.
O Molerus. Principles of Flow in Disperse Systems. London: AIChE I 21:769-775, 1975.
Chapman and Hall, 1993. JA Tallmadge. Packed bed pressure drop—an extension to
O Molerus, KE Wirth. Heat Transfer in Fluidized Beds. higher Reynolds numbers. AIChE J 16:1092-1093, 1970.
London: Chapman and Hall, 1997. H Wadell. Volume, shape, and roundness of quartz particles.
PG Nutting. Bull Amer Assoc Petrol Geo 14:1337, 1930. J Geol 43:250-280, 1935.
AO Oman, KM Watson. Pressure drops in granular beds. CY Wen, LT Fan. Models for Flow Systems and Chemical
Refinery Mat Petrol Chem Tech 36:R795, 1944. Reactors. New York: Marcel Dekker, 1975.
N Ouchiyama, T Tanaka. Predicting the densest packings of CY Wen, YH Yu. Mechanics of fluidization. Chem Eng Prog
ternary and quaternary mixtures of solid particles. Ind Symp Ser 62(62): 100-111, 1966.
Eng Chem Res 28:1530-1536, 1989. CA Wentz Jr, G Thodos. Pressure drops in the flow of gases
M Propster, J Szekely. The porosity of systems consisting of through packed and distended beds of spherical particles.
layers of different particles. Powder Technol 17:123—138, AIChE J 9:81-84, 1963.
1977. AER Westman. The packing of particles: empirical equations
WE Ranz. Friction and transfer coefficients for single parti­ for intermediate diameter ratios. J Am Ceram Soc 19:127-
cles and packed beds. Chem Eng Prog 48:247-253, 1952. 129, 1936.
HE Rase. Fixed-Bed Reactor Design and Diagnostics. S Yagi, D Kunii. Studies on effective thermal conductivities
Boston: Butterworths, 1990, pp 89-90. in packed beds. AIChE J 3:373-381, 1957.
HE Rose, MA Rizk. Further researches in fluid flow through AB Yu, N Standish, A McLean. Porosity calculation of
beds of granular material. Proc Inst Mech Eng London binary mixtures of nonspherical particles. J Am Ceram
160:493-503, 1949. Soc 76:2813-2816, 1993.
OA Saunders, N Ford. Heat transfer in the flow of gas RP Zou, AB Yu. Evaluation of the packing characteristics of
through a bed of solid particles. J Iron Steel Inst monosized nonspherical particles. Powder Technol 88:71-
141:291-316, 1940. 79, 1996.
AE Scheidegger. The Physics of Flow Through Porous
Media. New York: MacMillan, 1961.
3

Bubbling Fluidized Beds

W e n -C h in g Yang
Siemens Westinghouse Power Corporation, Pittsburgh, Pennsylvania, U.S.A.

1 IN TROD UCTION endothermic reactions. This in turn provides an easier

reactor control as well. The rigorous mixing also
For chemical reactor applications, the fixed (or improves solids and fluid contacting, and it enhances
packed) beds described in Chapter 2 have some heat and mass transfer. However, fluidized beds also
major disadvantages. If the reactions are fast and possess some serious disadvantages. Rigorous solids
highly exothermic or endothermic, hot or cold spots mixing in the bed produces solid fines through
will form in the packed beds and render the reactor attrition. Operating at high fludiization velocities,
ineffective; or it increases the chance or unwanted fines elutriation and entrainm ent can become a serious
byproduct production. By nature, the fixed bed opera­ operational problem. Also, because of the rigorous
tion has to be in batch mode, which is much less effi­ mixing in the bed, a fluidized bed is essentially a
cient than a continous operation. Sintering, plugging, continous stirred tank reactor with varying solids resi­
and fluid m aldistribution can also occur much more dence time distribution. These deficiencies may or may
readily in packed beds. Comparing to fixed beds, flui­ not be resolved through design. In practice, fluidiza­
dized beds have many advantages. Once the solids in tion is still an empirical science (Yang, 1998a). Care
the bed are fluidized, the solids inside the bed will must be exercised to select proper correlations for
behave just like liquid (Gelperin and Einstein, 1971; design and scale-up. A recent review of hydrodynamics
Davidson et al., 1977). The bed surface of a bubbling of gas-solid fluidization can be found in Lim et al.
fluidized bed resembles that of a boiling liquid, and it (1995).
can be stirred easily just like a liquid. Objects with a There are many different variations of fluidized beds
density lighter than the bulk density of the bed will in practice, which are covered in different chapters in
float and those heavier will sink. If there is a hole on this handbook. F or conventional bubbling fluidized
the side of a fluidizing bed, the solids will flow out like beds, the essential elements are depicted in Fig. 1.
a liquid jet. The gas bubble size, shape, formation, Depending on the applications, not all elements
rising velocity, and coalescence in the fluidized beds shown in Fig. 1 will be necessary. The design of the
have quantitative similarity with those of gas bubbles individual element can also vary substantially from one
in liquids. design to another design. For example, the cyclone can
The liquid like behavior of a fluidized bed thus be arranged either internally or externally; the heat
allows the solids to be handled like a fluid, and transfer tubes can be either vertical or horizontal,
continous feeding and withdrawal therefore become etc. These design aspects and various phenomena
possible. The rigorous mixing in a fluidized bed results occurring in a bubbling fluidized bed will be discussed
in a uniform tem perature even for highly exothermic or throughout this book.

53
54 Yang

2 POWDER CLASSIFICATIONS AND TYPES

OF GAS FLUIDIZATION

The fluidization phenomena of gas-solids systems

depend very much on the types of powders employed.
There are several classifications available in the litera­
ture, all based on the original work by Geldart (1973).

2.1 Geldart Classification of Powders

Geldart (1973) was the first to classify the behavior of

solids fluidized by gases into four clearly recognizable
groups characterized by the density difference between
the particles and the fluidizing medium, (pp — pf) and
by the mean particle size, dp. G eldart’s classification
has since become the standard to demarcate the types Gas
Feed
of gas fluidization. The most easily recognizable fea­
tures are

Group A The bed particles exhibit dense phase

expansion after minimum fluidization
Figure 1 Essential elements of a bubbling fluidized bed.
and before the beginning of bubbling.
Gas bubbles appear at the minimum
bubbling velocity.
Group B Gas bubbles appear at the minimum 8 x 1 0 4(pp - p{)gd.
(5)
fluidization velocity.
Group C The bed particles are cohesive and difficult
to fluidize. Equation (5) is the demarcation between G roup B and
Group D Stable spouted beds can be easily formed D powders. The powder classification diagram for the
in this group of powders. fluidization by air at ambient conditions was presented
by Geldart, as shown in Fig. 2.
Demarcation f or Groups A and B Powders. For Minimum Bubbling Velocity. For gas fluidization
G roup A, of fine particles, the fluidization velocity at which the
gas bubbles first appear is called the minimum bub­
u tnb
> 1 bling velocity, U mb- Geldart and Abraham sen (1978)
observed that
8 x i < r V P2(P p - P f)
(1) Umb 4.125 X 104p°'9pf'1
(6)
mf (pP - P f ) H
(2)
where the units are in kg, m , and s. Equation (6) gives
Umb < Umf, which is generally true for G roups B and
D powders. F or fine particles (Group A), the ratio
^mb — Kdp (3) Umh/ U mf > 1 was predicted and observed to increase
with both tem perature and pressure. The ratio also
Thus the powders are G roup A powders if increases with smaller particles and lighter particles.
8 x 10 4gip(pp - pf)
< 1 (4) 2.2 Molerus’ Interpretation of Geldart’s
K\i Classification of Powders
F or air at Room Temperature and Pressure, K
100
M olerus (1982) classified the powders by taking into
account the interparticle cohesion forces. Free particle
Urnf
F or Group D Powders, UB < m otion for G roup C powder is suppressed by the
s mf
Bubbling Fluidized Beds 55

dp, \im
Figure 2 Geldart classification of powders.

dom inant influence of cohesion forces, and its dem ar­ are good also for gases other than air and for tem pera­
cation can be expressed by the equation ture and pressure other than ambient.
Boundary Between Group A and Group B
l 0 x (pP... p{)dPg = Kl 10 -2 (7) -1.275
Pp-Pf
A r = 1.03 x 10e (10)
where FH is the adhesion force transm itted in a single Pf
contact between two particles. For polypropylene For (pp - pf)/p f = 1000 to 2000, Eq. (10) reduces to
powder, Fu = 7.71 x 10~7 newtons, and for glass Ar = 125 for air at atmospheric pressure. This com­
beads and cracking catalyst, F H — 8.76 x 1CT8 new­ pares to a value of A r = 88.5 as suggested by Goossens
tons. K x is found experimentally. (1998).
For Group A and G roup B transition, the following
equation can be used. Boundary Between Group B and Group D
Ar = 1.45 x 105 (11)
TC(Pp - PfMpg
(8)
6 Fh This G roup B/D boundary can be compared with Ar =
176,900, as suggested by Goossens (1998).
For G roup B and G roup D, the transition equation is

(Pp Pf)^p<^ 2.4 Goossen’s Classification of Particles by

15 (9) Archimedes Number
P f^ n f

Based on the hypothesis that the relative importance of

23 Particle Classification Boundaries Suggested by lam inar and turbulent phenomena governs the fluid­
Grace ization behavior, Goossen (1998) classified the
powders on the basis of Archimedes number. The
Based on additional data beyond that analyzed by proposed classification is of general application, apply­
Geldart, Grace (1986) suggested new boundaries ing equally well in both liquid fluidization and gas
between Groups A and B, and between Groups B fluidization. The four boundaries he suggested are as
and D of G eldart’s classification. The new boundaries follows:
56 Yang

Group C boundary Ar = 0.97 (12) tracking and data processing to characterize powders.
G roup A/C boundary Ar = 9.8 (13) When the gas is abruptly turned off for an operating
G roup A/B boundary Ar = 88.5 (14) fluidized bed, the bed collapses in three distinct stages:
G roup B/D boundary Ar = 176,900 (15) (1) bubble escaping stage, (2) hindered settling stage,
and (3) solids consolidation stage. The bed collapsing
The demarcations between groups of powder compare stage can be conveniently represented grahically in Fig.
well with that of G eldart’s and M olerus’ when air is 3. M athematically, the bed collapsing stages can be
used as the fluidization medium, except for the G roup described as follows.
A/B boundary.

2.5.1 Bubble Escaping Stage, 0 < t < t\>

2.5 Powder Characterization by Bed Collapsing
D uring the bubble escaping stage, the change of bed
The bed collapsing technique has been employed to surface is linearly dependent on the time, or
study fluidization for various objectives (Rietema,
1967; M orooka et al., 1973; Geldart, 1986). Kwauk H X= H 0 - U x (16)
(1992) has instrum ented the bed for autom atic surface

Hi
o /•VO*
y.A;-V HB
' p.y-

’.V 6;V
•W ;
& ':'S.

s/A ii//
t =0 t < tb t = t„ t„< t < i. t = t.
(i) (Si) (sis) (iv) (v)

Figure 3 Bed collapsing stages during bed collapsing characterization of powders. (Adapted from Kwauk, 1992.)
Bubbling Fluidized Beds 57

/ b(1 + /w se)^B + £e(l ” / b ^ w/ b) ^ For small powders whose terminal velocity can be
where U\ =
calculated by the Stokes law, 0S can be expressed by
(17)
(24)
H0 - H e 4>(pp pf)#«
a n d /B = (18)
Ha The dimensionless subsidence time of a powder was
found to be related to the ratio of minimum bubbling
and minimum fluidization velocities
2.5.2 Hindered Settling Stage, % < i < tc
During the hindered settling stage, the bed surface set­ ^mb = 40!,0.25 (25)
In
tles at a constant velocity, and the settled layer at the
bottom, layer D in Fig. 3, also increases in height. The Thus for G roup B and D powders, Umh/ U mf = 1, the
change in bed height for both layers can be expressed corresponding subsidence time is zero, and hence tc is
as zero. Stages 1 and 2 occur almost instantaneously. For
H, = H e - Ur (19) larger values of 0S, the ratios of Umh and U mf become
larger, and the bed exhibits the particulate fluidization
Hn more prominently.
Hn = Urt Yang et al. (1985) have performed extensive experi­
He H0
ments employing the bed collapsing technique to study
+ {1 —exp the modification of fluidization behavior by addition of
fines. They found that improvement in fluidization
U2H e H„ increases monotonically with increases in fines concen­
(20)
(He - He - H a tration in the bed. A more interesting study is the series
for binary mixtures of G roup B-C powders. They
where K2 is the rate constant for the increasing of
found that addition of cohesive G roup C powder
height in layer D expressed as a first-order process:
into G roup B powder will improve the fluidization
behavior of G roup B powder. Conversely, addition
(21)
of G roup B powder into the cohesive G roup C powder
will also improve the fluidization quality of a cohesive
2.5.3 Solids Consolidation Stage, tc < t < t0 powder. There is, however, some limitations on the
effective am ount to be added in the mixture.
The consolidation stage is essentially the consolidation Addition beyond the maximum am ount will adversely
of layer D and thus the change in bed height can be affect the fluidity and fluidization behavior of the
expressed as mixture.
dH3 dHr
^ r = ^ r = - ^ Ha - * ,* ,) (22)
3 DIFFERENT REG IM ES OF FLUIDIZATIO N
or
H 3 = H d = ( Hc - H oc) X exp[ - K 3(t - tc)] + I I ^ Fluidization regimes can be classified into two broad
categories—particulate (smooth) and aggregative
(23)
(bubbling) (Harrison et al., 1961). M ost liquid fluidized
N ot all powders exhibit all three stages during bed beds under norm al operation exhibit the particulate
collapsing experiments. Only G eldart’s Group A fluidization. In particulate fluidization, the solid
powders exhibit all three stages. For Group B and D particles usually disperse relatively uniformly in the
powders, the first bubble escaping stage and the second fluidizing medium with no readily identifiable bubbles.
hindered settling stage are practically instantaneous. Thus the particulate fluidization sometimes is also
The transition between the hindered settling stage called homogeneous fluidization. In the heterogeneous
and the solids consolidation stage, the point (tc, H c) or aggregative fluidization, voids containing no solids
in Fig. 3, is called by Yang et al. (1985) “the critical are usually observed. Those voids are called bubbles.
point.” The better to qualify the powders through the Those voids can be well defined as in a bubbling flui­
bed collapsing tests, Yang et al. (1985) also defined a dized bed or in a slugging bed, or they can be small
dimensionless subsidence time of a powder, 0S. voids where particle clusters dart to and from like in a
58 Yang

turbulent bed or in a fast fluidized bed. There are evidence of maximum bubble size was observed by
a number of criteria available to determine whether M atsen (1973).
a particular system will exhibit particulate or aggre­ For gas-solid systems, there are at least five distin­
gative fluidization. Those criteria are summarized in guishable regimes of fluidization observable experi­
Table 1. mentally: fixed bed, particulate fluidization, bubbling
It was suggested by Harrison et al. (1961) that fluidization, slugging fluidization, and turbulent fluid­
aggregative fluidization may be expected if the ratio ization. When the operating velocity is higher than the
of maximum stable bubble size to particle diameter, transport velocity such that recycle of entrained p arti­
(DB)max/ J p, is larger than 10, and particulate fluidiza­ cles is necessary to m aintain a bed, additional fluidiz-
tion, if the ratio is less than or equal to unity. A transi­ ing regimes are possible. The regimes of fluidization for
tion region exists with the ratio between 1 and 10. They circulating fluidized beds is discussed in more detail in
proposed to calculate the ratio as follows: C hapter 19, “ Circulating Fluidized Beds.” The differ­
ent regimes for gas-solid fluidization are summarized
{DB)n p in Table 2 and Figs. 4 and 5. N ot all of these regimes
■= 71.3 can be observed in all systems, however, because some
gdlvh
regimes are also dependent on the size of the equip­
(26)
(Pp/(Pp Pf)) £mf
m ent employed. For a recent review of fluidization
1 — £mf regimes, see Bi and Grace (1995a).

, . ^ p P f/
54|i2 p
^
P i
1/2

-1
3.1 Transition among Fixed Bed, Particulate
Fluidization, and Bubbling Regime
F or large particles, Eq. (26) reduces to
F or G eldart’s G roup B and G roup D powders, the bed
(A*)n ■= 1.32 Pp ~ Pf -( p p / ( p P - P f ) ) “ e mf ‘ transfers from the fixed bed into a bubbling fluidized
Pf . 1 — £mf bed when the gas velocity is increased beyond the m ini­
(27) mum fluidization velocity of the system. For G roup A
powders, no bubbles will be observed, instead the bed
Eqs. (26) and (27) are only applicable when the will expand homogeneously. The bubbles only appear
maximum stable bubble size is much smaller than the when the gas velocity is increased beyond the minimum
bed diameter, i.e., without wall effect. Experimental bubbling velocity. Thus the transition point from the

Table 1 Criteria for Transition Between Particulate and Aggregative

Fluidization

Criteria for particulate fluidization Reference

Fr = - ^ < 0 . 1 3 Wilhelm and Kwauk (1948)

gap

(T^fO----
max- < jw 2U{
30;, (E>B)max
V^B/max — Harrison et al. (1961)

Umf < 0.2 em/s Rowe (1962)

Fmf(pp —Pf )Hmf Romero and Johanson (1962)

< 100
g[iD
(Pp ~ Pf) C{ depending on s and emf Verloop and Heertjes (1974)

0.5
7pKtvKp
3Pf(pP -P f) Pmf Doichev et al. (1975)
<c.
Pf

m and C2 depend on range of operation

Bubbling Fluidized Beds 59

Table 2 Summary Description of Different Regimes of Fluidization

Velocity range Fluidization regime Fluidization features and appearance

0<U<Umf Fixed bed Particles are quiescent; gas flows through interstices
umf<u< umb Particulate regime Bed expands smoothly and homogeneously with small-scale
particle motion; bed surface is well defined
umb < u < ums Bubbling regime Gas bubbles form above distributor, coalesce and grow; gas
bubbles promote solids mixing during rise to surface and
breakthrough
Ums<u< uc Slug flow regime Bubble size approaches bed cross section; bed surface rises
and falls with regular frequency with corresponding
pressure fluctuation
uc< u< uk Transition to turbulent Pressure fluctuations decrease gradually until turbulent
fluidization fluidization regime is reached
uk<u< utt Turbulent regime Small gas voids and particle clusters and streamers dart to
and fro; bed surface is diffused and difficult to distinguish
U>UtI Fast fluidization Particles are transported out of the bed and need to be
replaced and recycled; normally has a dense phase region
at bottom coexisting with a dilute phase region on top; no
bed surface
u » uu Pneumatic conveying Usually a once-through operation; all particles fed are
transported out in dilute phase with concentration varying
along the column height; no bed surface

fixed bed to the bubbling regime is the minimum fluid­ 3.3 Transition Between Bubbling and Turbulent
ization velocity for G roup B and G roup D powders Regimes
while for Group A powders, it is the minimum bub­
bling velocity. The homogeneous expansion, also When the gas velocity is continuously increased, the
called particulate fluidization, occurs only in Group bubbles grow bigger owing to coalescence, and the
A powders for gas-solids systems. The particulate bubbling bed can transfer into a slugging bed if the
fluidization is especially im portant for liquid fluidized bed diameter is small and the particle diameter is
beds because where most of the occurrence is observed. large, or into a turbulent bed if the bed diameter is
Liquid-solids fluidization will be discussed in Chapter large and the particle diameter is small. If the standard
26, “Liquid-Solids Fluidization.” deviation of pressure fluctuation is measured and
plotted against the superficial fluidization velocity,
two characteristic velocities, Uc and C/k, first suggested
3.2 Transition Between Bubbling and Slugging by Yerushalmi and C ankurt (1979), can be identified.
Regimes The velocity Uc corresponds to the bed operating con­
ditions where the bubbles or slugs reach their maxi­
A slugging regime occurs only in beds with bed height mum diameter and thus have the largest standard
(H) over bed diameter ratio (.D) larger than about 2. deviation of pressure flucuation. Continuing increases
With large H / D ratios, the bed provides enough time beyond this velocity, large bubbles start to break up
for bubbles to coalesce into bigger ones. When the into smaller bubbles with smaller pressure fluctuation,
bubbles grow to approximately 2/3 of the bed dia­ and eventually the standard deviation of the pressure
meter, the bed enters the slugging regime with periodic fluctuation reaches a steady state. This velocity is
passing of large bubbles and regular large fluctuation denoted as Uk, also a characterization velocity for
of bed pressure drop corresponding to the bubble transition from the bubbling regime to the turbulent
frequency. There are several correlations available to regime. Subsequent study by Chehbouni et al. (1994),
predict this transition, they are discussed in Sec. 10; employing differential and absolute pressure transdu­
“ Slugging Beds.” cers and a capacitance probe, concluded that the onset
60 Yang

Beds With Little Entrainment Transport & Circulating Fluidized Beds

(For transition velocities, see Bi & Grace, 1995)

&
o
o
2

(m < 0.66 D

Figure 4 Regimes of fluidization for nontransporting systems.

of turbulent fluidization was at Uc, and the velocity Uk

(Re)k = JpPT - = 1.46Ar°'472
actually did not exist. The existence of Uk is an artifact
due to the use of differential pressure transducers for for C anada et al. (1978) data (29)
experiments. They maintained that there is only one
transition velocity, i.e., Uc. The turbulent fluidization
starts at Uc and ends at the transport velocity, f/tr, the
(Re)k = ^ T > £ = l.41A r0-56
velocity capable of transporting all particles out of the
reactor. There is still controversy in the literature on for Yerushalmi et al. (1978) data (30)
the actual transition boundary between the bubbling
and the turbulent regimes. The extensive literature data based on absolute pres­
In 1986, Horio proposed the following equations to sure fluctuation and bed expansion measurements up
calculate Uc and Uk (see also Horio, 1990). to 1989 were correlated by Cai et al. (1989) to be

( R e ) = gpPf Uc = 0 936A ro.472

(28) (Re) = dPp{Uc = 0.57Ar0,46 (31)
|4
Bubbling Fluidized Beds 61

0.941

(Re)1 = y ^ = 1 0 1 A r- ( ^ )

for Ar < 125 (35)

0.0015
(Re)k = = 2.274Ar'.0 .4 1 9 /

for Ar > 125 (36)

<7pPf ^tr
(Re), = 2.28Ar.0.419 (37)
\i
Bi and Grace (1995b) correlated the literature data
based on the differential pressure fluctuation and
arrived at the following transition equation:
dp Pf Uc
(Re)c = = 1.24Ar,0.45

for 2 < Ar < 1 x 10 (38)

F or gas fluidization of large particles, Catipovic et al.

(1978) further subdivided the regimes into slow bub­
bles, fast bubbles, and rapidly growing bubble regimes.
The transition equations are summarized here.
Regime Between Fast and Slow Bubbles

A r1' 3
(39)
v‘ - ~
Figure 5 Regimes of fluidization for transporting and non­
transporting systems. Regime of Rapidly Growing Bubbles
In this regime, the bubble growth rate is of the same
magnitude as the bubble rising velocity, and for
shallow beds,
Perales et al. (1990) also suggested equations to calcu­
late Uk and UtT as follows. d(DB) d(DB) dh
(40)
dt dh dt

(Re)k = dpP{Uk = 1.95Ar0'453 (32) or

djD*) (41)
dh
dpPfUtr
(Re)tr = = 1.41Ar.0.483 (33)
3.4 Transition to Fast Fluidization

Since the £/k and Utv are very similar, they suggest that Continuing increases in operating velocity beyond that
the following equation alone may be used to calculate required at turbulent fluidization, a critical velocity,
both I/k and Utr for simplification: commonly called the transport velocity Utr, will be
reached where a significant particle entrainment
(Re)k t = _ ^ p P f^ tr _ 1.45Ar
■= /1C A ^0.484
(34) occurs. Beyond this point, continuing operation of
P the bed will not be possible without recycle of the
entrained solids. The bed is now said to be in the fast
Bi and Fan (1992) affirmed the existence of turbulent fluidization regime. The transition velocity has been
regimes in gas-solid fluidization and suggested the fol­ correlated by Bi et al. (1995) as
lowing criteria for transition to turbulent fluidization.
U,r = 1.53Ar05 for 2 < Ar < 4 x 10'.0.5 (42)
62 Yang

For G roup A and B particles, the transition velocity, than perfect design of the gas distributor, to the finite
Utr, calculated from Eq. (42) is larger than the terminal dimension of the containing vessel, and to the possibi­
velocity of the individual particles, while for G roup D lity of channeling. At the point of minimum fluidiza­
particles, the transition velocity equals essentially the tion, the voidage of the bed corresponds to the loosest
terminal velocity of the individual particles. packing of a packed bed. The loosest mode of packing
However, a fluidized bed operating at a high gas for uniform spheres is cubic, as discussed in Chapter 2
velocity alone (> Utr) does not make it a fast fluidized or smf = (6 —7i)/6 = 0.476. Substituting into the origi­
bed. A generally accepted definition of a fast fluidized nal Carm an equation, we obtain
bed is the Coexistence of a dilute phase and dense phase
regimes (see review by Rhodes and Wang, 1998). AP 12 pU (l - e)2
(43)
_COS2(y)_ dl s3

and assuming
4 T H E O R E T IC A L A N D E M P IR IC A L
P R E D IC T IO N S O F M IN IM U M 72
180; Y is usually from 48 to 51° (44)
F L U ID IZ A T IO N V E L O C IT Y cos2(y)
we have
The phenom enon of fluidization can best be character­
ized by a A P / L versus U plot such as the one shown in
Fig. 6. Below a characteristic gas velocity known as the (45)
L <a
minimum fluidization velocity, a packed bed of solid
particles remains fixed, though a pressure drop across At the point of minimum fluidization, the pressure
the bed can be measured. At the minimum fluidization drop is enough to support the weight of the particles
velocity, all the particles are essentially supported by and can be expressed as
the gas stream. The pressure drop through the bed is AP
then equal to the bed weight divided by the cross- — (Pp Pf)0 ^mf) (46)
sectional area of the bed, AP — W / A . Further
increases in gas velocity will usually not cause further Combining Eqs. (45) and (46) with the voidage at the
increases in pressure drop. In actual practice, however, minimum fluidization £mf = 0.476, we find
pressure drop at minimum fluidization velocity is actu­
ally less then W / A because a small percentage of the Umf = 0.00114 gdp(Pp ~ Pf) (47)
bed particles is supported by the wall owing to the less

_W
A
ABDEF, AB'DEF
Powder in a narrow tube or
Powder has been compacted
Increasing Bed AC'EFG, AC'FG
Particle Size Powder of wide size distribution
Q Distribution Partial fluidization indicating
a> particle segregation
2 ACDEFG
a. Normal fluidization curve
FG
Fully fluidized region

U*
Fluidization Velocity

Figure 6 Pressure drop vs. fluidization velocity plot for determination of minimum fluidization velocity.
Bubbling Fluidized Beds 63

Leva (1959) employed the experimental values of In a simplified form, the Ergun equation, in terms of
voidage at minimum fluidization and arrived at the Reynolds num ber and Archimedes number, can be
equations reduced to

d(Pp ~ Pf)
UM = [0.0007(Re)-?°63] (Re)mf = y c f CoAr - Ci (55)

VnfPf
(Re)mf = (48) There are many sets of values suggested for the con­
P stants Cj and C2 based on different databases used for
Equations (47) and (48) are comparable because correlation. A simplified set is shown in the table.
(Re)“ f'°63 is on the order of unity for most fluidized
systems where the Reynolds num ber at minimum flui­
dization, (Re)mf, is generally between 10-2 to 102. Reference Ci C2
Rowe (1961) performed experiments on water flow Wen and Yu (1966a) 33.7 0.0408
through a regular array of spheres and found that the Richardson (1971) 25.7 0.0365
force on the single sphere in the array was 68.5 times Saxena and Vogel (1977) 25.3 0.0571
the force on an isolated sphere at the same superficial Babu et al. (1978) 25.25 0.0651
velocity. If we assume that the same factor of 68.5 is Grace (1982) 27.2 0.0408
also applicable at the minimum fluidization condition Chitester et al. (1984) 28.7 0.0494
when the drag force of the fluid on the particles is just
balanced by the net downward force on the particles, According to Grace (1982), a change of Q from the
we have, for the low-Reynolds number case, 33.7 suggested by Wen and Yu to 27.5 improves the fit
F = 3%\iUdp (49) for fine particles. F or the limiting cases for small p ar­
ticles where the kinetic energy term is not im portant,
and and for large particles where it is dominant, the sim­
plified equations become
(Pp - P fM p£
68.5 x 3n\iUmfdv = (50) F or Ar < 10

or (Re)mf = 7.5 x 10 4 Ar or
0.0008 l(p . - p{) s 4
Umf = (51) Um{ = 0.00075^Pp ~ F .gd? (56)

Another widely employed approach is by simplifica­

F or Ar > 1 0 \
tion of the Ergun equation for the packed bed. By
combining the original Ergun equation as shown in
(Re)mf = 0.202VAr or
Eq. (52) and the pressure drop equation at minimum
fluidization as shown in Eq. (46) /(Pp - Pf)gdp
UmS = 0.202 (57)
AP g$ d p (1 - 8) Pf
75 ■0.875 (52)
L 2pf 172 ( l - e ) ^(Re)p
It is worthwhile to note that at both small and large
we obtain the equation for calculating the minimum Archimedes numbers, the minimum fluidization velo­
fluidization velocity, including both the kinetic energy city, Umf, is directly proportional to the corresponding
term and the viscous term, as terminal velocity, Uu of a single spherical particle of
diameter dp in an infinite medium. For small spheres
Ar = 150 (1 - Smf) (Re)mf + 1.75-4>s
L (R e)mf (53) where the Stokes law applies, Ut/ U mf approaches 74.
t>2< f ,3
mf F or large spheres in the N ew ton’s law regime, Ut/ U mf
Wen and Yu (1966a, 1966b) suggested a simplified approaches 8.6.
form of the Ergun equation by assuming the followings Since in practical applications the particles are
based on experimental data: rarely in narrow distribution, the average particle size
used in all equations dicussed so far for minimum
1 ^ -■
14 and- (1 - £mf) (54) fluidization velocity is recommended to be the
(t)£,mf
3 4>2 s 'mf
3 surface-volume mean:
64 Yang

dc\r -- (58) R Y l = 90 c = 0.3 (64)

xi Um{
2 L ,j.
The above equations are valid for spherical isometric
where x t is the weight fraction of particle size dpi. All particles with negligible wall effect. If the terminal
correlations available for calculating the minimum velocity of a particle is available, the above equations
fluidization velocity have been reviewed by Babu et can be used to estimate the minimum fluidization velo­
al. (1978). A method of estimating the minimum city as well.
fluidization velocity at elevated temperatures and pres­
sures was proposed by Yang (1998a). The effect of
6 TW O -PH A SE THEORY OF FLU ID IZA TIO N
temperature and pressure on minimum fluidization
velocity is discussed in more detail in Chapter 5,
The two-phase theory of fluidization was first p ro ­
“ Effect of Temperature and Pressure.”
posed by Toomey and Johnstone (1952). The model
assumed that the aggregative fluidization consists of
5 THE RATIO OF TERM INAL VELOCITY TO two phases, i.e., the particulate (or emulsion) phase
M IN IM U M FLU ID IZIN G VELOCITY and the bubble phase. The flow rate through the emul­
sion phase is equal to the flow rate for minimum fluid­
F or spherical particles, Bourgeois and Grenier (1968) ization, and the voidage is essentially constant at 8mf.
proposed semitheoretical correlations relating the ratio Any flow in excess of that required for minimum fluid­
of terminal velocity to minimum fluidizing velocity ization appears as bubbles in the separate bubble
with the Archimedes number, Ar, as shown in Eqs. phase. Mathematically, the two-phase theory can be
(59) through (64). Depending on whether the fluidized expressed as
bed is fluidized by air or water, the velocity ratios are
slightly different. The terminal velocity of a solid (65)
particle has been discussed in Sec. 4 of Chapter 1,
“Particle Characterization and Dynamics.” where GB is the average visible volumetric bubble flow
For fluidization by air and 102 < Ar < 4 x 104, across a given cross section of the bed. For some sys­
tems, the two-phase theory is good and simple approx­
* = #-=135.7 45.0(log Ar) + 4.1 (log Ar)2 imation, but there is also much experimental evidence
Lmf that correction may be necessary, especially at high
a = 0.69 for 20 < R < 60 (59) pressures. Experimental evidence indicated that the
equation should be
F or fluidization by air and 4 x 104 < Ar < 8 x 106,
^ = U ~ K 4Um{ with K a > 1.0 (66)
R = ^ - = 26.6 - 2.3(log Ar)
Um{
c = 0.52 for 10 < R < 20 (60) The reason for this deviation was suggested by Grace
and Clift (1974) to be due primarily to two factors: an
For fluidization by air and Ar > 8 x 10 increase in interstitial gas velocity in the emulsion
Ut phase above that required for minimum fluidization
R = 10.8 0.5 (61) and possibly through flow inside the bubbles. This
U,mf
led to their modified or n-type two-phase theory.
F or fluidization by water and 50 < Ar < 2 x 10 ,
= U — £/mf(l + «£b)
R = = 132.8 - 47.1 (log Ar) + 4.6(log A r)2
Umf = U — [Umf(l —£b) + (n + l)U mfsB] (67)
a = 2.9 for 20 < R < 60 (62)
where sB is the average volume fraction of visible
F or fluidization by water and 2 x 104 < Ar < 1 x 106, bubble phase in the bed. The flow through particulate
phase is
R = - F - = 26.0 - 2.7(log Ar)
Vnf Flow through particulate phase = Umf(l —sB)A
0.67 for 9 < R < 20 (63) (68)
F or fluidization by water and A r> 1 x 10 , and the through flow relative to bubbles is
Bubbling Fluidized Beds 65

Through flow relative to bubbles = (n + l)U mfsBT bed. The minimum effective am ount of fines required
(69) to reduce the viscosity is thus that quantity necessary
to coat each coarse particle with just a single layer of
For the classical two-phase theory, n = 0. The model fine particles, as shown in Fig. 7.
implicitly assumes that the superficial velocity in the In an actual fluidized bed, the bed is usually
emulsion phase remains at I7mf, while the velocity expanded during operation, and the voids occupied
inside the rising voids in (n + l)U mf relative to the by the gas bubbles must be subtracted from the volume
boundary of the bubbles. It was hoped by the authors to give the weight fraction of the “film,” which repre­
that n would turn out to be a universal positive con­ sents more closely the true minimum weight required.
stant and thus the m agnitude of n would become a Assuming similar density for ds and <7B, the weight
measure of the overall through flow in freely bubbling fraction of fines can be calculated as
beds. Experimental evidence available so far indicates
that n is not universal. It is different from system to
system and may be even different from observer to J u )+y _
(1 —s) 3
~(ck\ f ds\ 2~ (tf-tfo )]
H°
observer. Within a given system, n is also dependent
on height and superficial gas velocity. The difficulty

1 &£
’(sM S"
1 + (1 —e) 3

i
arises from the inability to measure the invisible flow
components of the total flow, since the invisible flow (70)
component is substantial and its distribution remains
unknown. The possibility of substantial through flow where H 0 and H represent the initial and the expanded
through the uneven distribution of bubbles was ana­ bed height. Assuming the particles to be perfect
lyzed by Valenzuela and Glicksman (1981) theoreti­ spheres, the optimum two-component system with
cally. the maximum surface area has been computed to be
the one in which the diameter of the fines is 22.5% that
of the larger particles, and in which the smaller parti­
7 VISCOSITY OF A FLU ID IZED BED cles constitute 25% of the mixture by weight. A study
by G eldart (1972) employing the surface-volume par­
A number of investigators have measured the apparent ticle diameter seemed to confirm this as well. Some of
viscosity of fluidized bed using methods available for the data by M atheson et al. (1949) and replotted by
ordinary liquids. For example, M atheson et al. (1949) G eldart are presented in Fig. 8.
and Furukaw a and Ohmae (1958) employed a rotating The importance of fines content in the fluidized bed
paddle viscometer and found that a fully fluidized bed reactors was also investigated by Grace and Sun (1990)
had a viscosity from 0.5 to 20 poises. The viscosity from a different aspect. They found that fines spent a
increases with the size of the particles in the bed. much longer time in the bubble phase and considerably
Other researchers, Kram ers (1951), Dickman and enhanced the concentration inside the bubbles due to
Forsythe (1953), and Schugerl et al. (1961), used visc­ the through flow of fluidizing gas through the bubbles.
ometers of slightly different designs and obtained prac­ This enhanced particles-gas contact and improved the
tically similar results. For small shear stress, the fully reactor performance.
fluidized beds behave as newtonian fluids. The viscos­
ity of the fluidized bed is very high at close to minimum
fluidization condition and decreases sharply with
increases in gas flow.
As early as 1949, M atheson et al. reported that the
addition of relatively small amounts of fines to a coarse
bed will decrease the viscosity of the bed substantially.
It remained for Trawinski (1953) to find an analytical
explanation for this phenomenon. He assumed that the
viscosity in a fluidized bed is primarily due to the rub­
bing of coarse particles among one another. The addi­
tion of fines will reduce the friction between coarse Figure 7 Schematic of coating of fine particles on coarse
particles by coating a thin layer of fines on each coarse particles—Trawinski’s model of reduction in fluidized bed
particle and thus decrease the viscosity of a fluidized viscosity upon addition of fines.
66 Yang

Mean Particle Size, microns

Figure 8 Matheson’s data of fluidized bed viscosity. (Replotted by Geldart, 1972.)

Grace (1970) reviewed the literature available in this are remarkably similar to that obtained by others using
area in 1970 and concluded that the direct a rotating cylinder viscometer. For more discussion on
measurement techniques commonly used to determine the viscosity of fluidized beds, see that by Schugerl
the viscosity of newtonian liquids tended to alter the (1971).
behavior of the bed, e.g., the regions underneath and The dense phase viscosity of fluidized beds at ele­
above the probe tended to have high and low voidages. vated pressure was studied by King et al. (1981) at
Thus the viscosity obtained by those methods may not pressures up to 20 bar by measuring the velocity of a
be the true viscosity of a fluidized bed. He, in turn, falling sphere. They found that for particles less than
proposed an indirect method based on the behavior about 100 pm, increases in pressure caused substantial
of bubbles in fluidized beds. decreases in viscosity. For larger particles, however,
The spherical-cap bubbles observed in fluidized beds the viscosity is almost independent of pressure.
can best be characterized by the included angle as
shown in Fig. 9. The angle was found to be dependent
only on the bubble Reynolds number, as shown in Eq.
(71)
A ^ B P f = 23c -0.0040 200° < 0 < 260°
(R e )e =
P
(71)
where De is the equivalent bubble diameter and can be
calculated as
1/3
(72)

The viscosity of the fluidized bed can then be estimated

by the equation
29eFBpp(l £mf)
p= (73)
(R e )e

The estimated range of the viscosity of fluidized beds is Figure 9 Schematic of spherical-cap bubbles for determina­
from 4 to 13 poises by using this method. These values tion of fluidized bed viscosity.
Bubbling Fluidized Beds 67

8 BED EXPANSION tfmaxO - ^Tmax) - H mi{\ " ^Tmf) U - Utmf

(80)
H m{(l - FTmf) + //max EXmax Un
8.1 Expansion of an Aggregative Fluidized Bed
8.1.3 Bed Expansion for the Case with Increases in
The bed expansion of an aggregative fluidized bed can
Bubble Size Due to Coalescence
be calculated using the equation
Xavier et al. (1978) employed the bubble coalescence
L (1 - Smf)
(74) correlation by D arton et al. (1977) and suggested to
L mf (1 -e ) calculate the bed expansion from the following equa­
Since the bed surface is oscillating because of constant tion. The bubble coalescence will be discussed in more
bubble breaking through the bed surface, the bed detail in a later section.
height and the voidage are the time-averaged values.
0.6
. b 0-6]
8.1.1 Bed Expansion with a Constant Bubble Size
- 5b2[(H + B f 2 - 5 0'2] (81)
The bed height oscillation for a bed with a constant 0.2 >0.2
\ H + B)' E
bubble size can be calculated from the absolute bubble 5b2 51 tan 1 tan -l bo.i
velocity expressed as in Eq. (75), and the time it b°‘
requires for the bubble to travel through the bed
where
shown in Eq. (76) (Xavier et al., 1978).
0.8
UA = ( U - Umf) + UB (75) (U -U rt)' and = 0
b = 1.917 rrQA
b ajt

Hrr (82)
T = ~- (76)
UA The A 0 is the so-called catchment area for a bubble
The bed expands from its minimum height, H mf, to its stream at the distributor plate. It is usually the area
maximum height, H max, with a total bubble flow of of distributor plate per orifice.
(U — Umf) A , assuming that the two-phase theory Geldart (1975) also developed an equation for bed
holds, or expansion in freely bubbling beds with “ sandlike”
powders as follows:
(iTmax - tfmfM = (U ~ Umf) A T (77)
(c + dH) 1/2 - cM 2
The bed expansion equation is obtained by combining
H,mf
Eq. (75) through (77) to be
(83)
(78) where d = 0.027(U - Umff 94 and c = 0.915 ( U -
H mf
r
Um(f A for porous distributor plates. Y is a correction
If the bubble velocity, f/B, is known, the maximum bed for deviation from the two-phase theory. Quantitative
expansion can be calculated from Eq. (78). values of Y are shown in Fig. 26.
For a distributor with N holes per cm2, c can be
8.1.2 Bed Expansion with a Constant Bubble Size calculated by
and an Array of Rods in the Bed 0.4
1.43 ( u - a mf
From volumetric balance, it can be written that c = ■t0.2 (84)
N
^max - //mf = (U - Umf) T All units in Eqs. (83) and (84) are in cm and second.
T H mSLXEXmax //mf
mf D
v Tmf (79)

where EXmax = the fraction of bed occupied by rods 9 BUBBLE PHASE IN FLUIDIZED BEDS
when bed height is at the maximum, / / max, and FXmf =
the fraction of bed occupied by rods when bed height is The bubbles in a fluidized bed have two basic proper­
at the minimum, H mf. ties. In general, they rise at a finite velocity, and they
Eliminating the UA and T among Eqs. (75), (76), usually grow in size owing primarily to static pressure
and (79), we find or to coalescence. It has been found that there is a
68 Yang

striking similarity between the behavior of large gas profile. The bubbles evidently rise relative to the fastest
bubbles in liquids and those in a fluidized bed moving liquid in the middle of the tube.
(Davidson et al., 1977). F or a swarm of bubbles, the same thing applies:
UA = U + U B (89)
9.1 Bubbles in Liquids
From the continuity of gas flow it can be derived th at
In a liquid of small viscosity, the rate of rise of large
bubbles depends primarily on inertial forces and Ub= H0
(90)
surface tension. The viscous effect is negligible in com­ U ( H - H 0)
parison. The shape of the bubble will adjust itself to where H is the liquid height when gas velocity is U\
m aintain the pressure inside the bubble constant. An and H q is the liquid height when gas velocity is zero.
approximate solution by Dumitrescu (1943) for a long Equation (90) provides a relationship for estimating
bubble in a tube gives UB from experimentally observable quantities.
Equations (89) and (90) can be applied to a bubbling
UB = 0.35y/gD (85) gas-solid fluidized bed by replacing U with U — I/mf,
Equation (85) was verified experimentally by Nicklin assuming that the two-phase theory applies.
et al. (1962) for application to a finite bubble or to a
slug rising in a tube. Davies and Taylor (1950) also 9.2 Bubbles in Gas-Solid Fluidized Beds
provided a solution with a slightly different empirical
constant. Bubbles in gas-solid fluidized beds usually are sphe-
rical-capped as shown in Fig. 9 with the included angle
UB = 0 J U ^ D e (86) equal to 240° found by experiments as compared to
120° derived theoretically. The bubbles in air-w ater
where De is the diameter of the sphere having the same systems have an angle of 100°.
volume as the bubble and can be calculated from Eq. The rising velocity of a single bubble in a quiescent
(72). bed has been found experimentally to be
Uno and K intner (1956) measured the rising
velocity of gas bubbles in liquid contained in tubes of UB = 0 J l g 1/2
l^ V1/1/6
}
B (91)
various diameters and found that the wall effect pre­
This compares with the experimental value of Davis
dominates when De is more than 1/3 of the bed
and Taylor (1950) for bubbles in liquids shown in
diameter. The wall effect becomes negligible only
Eq. (86).
when De is less than 0.1 of the bed diameter. The
Equation (91) implies that the bubble rising velocity
regime where the wall effect is dom inant is generally
is independent of the type of bed materials used in the
called the slugging regime.
fluidized beds. This has been experimentally proven by
In the case of a stream of bubbles in a vertical tube
Yasui and Johanson (1958).
generated continously by blowing air in at the bottom,
A spherical-cap bubble with an included angle, 0, as
the absolute upward rising velocity of each bubble is
shown in Fig. 9, has a volume of
greater than the velocity of a similar size single bubble
rising in a stagnant liquid. By making a simple material
Vn = nRi j — cos(9) + tc o s 3(0) (92)
balance, it is possible to derive the absolute bubble
velocity as
If Eq. (91) is written in a general form, it can be
expressed as
UA — — + 0.35^/gD (87)
UB = Cgl/2V^/6 (93)
Nicklin et al. (1962) found experimentally that the
absolute bubble velocity is Combining Eqs. (92) and (93), we have
- 1/6
G 1/6
/ - c o s ( 0 ) + t c o s 3(0) (94)
UA = 1.2 0.35y (88)

The factor 1.2 stems from the fact that the peak velo­ F or a gas-solid fluidized bed where the bubble velocity
city at the middle of the tube is about 1.2 times the can be expressed as in Eq. (91), i.e., C — 0.71, the
average velocity, owing to the nonuniform velocity included bubble angle is 120°.
Bubbling Fluidized Beds 69

When the bubbles are large enough to exceed Since D ’Arcy’s law (see Chapter 2) is good only for low
approximately 25% of the column diameter, the bub­ Reynolds numbers, the same limitation also applies
ble velocity is affected by the presence of the vessel here.
wall. The bubble velocity can then be approximated
by Eq. (85) for a long bubble in a tube containing 3. Fluidizing fluid is assumed to be incompressi­
liquid. The relative regime of application of Eqs. (85) ble, and thus the continuity equation can be written as
and (91) for gas-solid bubbling fluidized beds is shown
in Fig. 10. (97)
dx dy

with the assumption that the voidage is constant at smf.

9.3 Davidson’s Isolated Bubble Model Eliminating the velocities from Eqs. (95) through
(97), we have
In 1961, Davidson developed a simple theory that was
capable of explaining a lot of phenom ena relating to a2p d2p
bubbles in fluidized beds observed experimentally. His 1 + ~—T — 0 (98)
dx2 dy2
development involves the following assumptions.
1. The particulate phase is an incompressible fluid Equation (98) is exactly similar to that for a fixed bed,
with a bulk density similar to that of a fluidized bed at and thus the pressure distribution in a fluidized bed is
minimum fluidization. The continuity equation of the unaffected by the motions of the particles in a fluidized
particles can thus be expressed as bed.
4. The pressure throughout the bubble is constant.
9J% h
d 5. The particulate phase behaves as an inviscid
—0 (95)
dx dy liquid.
6. The bubble has a circular cross section.
2. The relative velocity between the particles and Davidson solved these equations in terms of particle
the fluidizing medium is assumed to be proportional to m otion, the pressure distribution within the fluidizing
the pressure gradient within the fluid, and thus fluid, the absolute velocities of the fluidizing fluid, and
D ’Arcy’s law is applicable. the exchange between the bubble and the particulate
phase. He derived the fluid stream function as
dP , dP
UY = VX- K and U y = V y ■K-
dx dy Um r sin 0
x|/f = J7] i - if (99)
(96) smf

Figure 10 Relative regions of applications for bubble velocity equations.

70 Yang

i ? y = ^/B + 2 ( ^ mf/ s mf) thin. In most of the fluidized beds of pratical interests,
( 100) this is the case. The theoretical findings here has been
A bJ UB - ( Umf/ 8mf)
experimentally verified.
He found that the geometry of the stream function is The stream function for the particles was obtained
crucially affected depending whether the bubble velo­ to be
city UB is larger or smaller than interstitial minimum
fluidization velocity, Umf/s mf. r2 sin2 I
1- (102)
r
9.3.1 Fast Bubbles Regime—W hen UB > Umf / emf
In this case, the particulate phase streams past a sphere
The fluidizing fluid in this case moves downward rela­
of radius R B as shown in Fig. 11. The relative velocity
tive to the bubble motion. The fluid flows past the
of the particulate flow to the void is —UB at r = oo.
fictitious sphere of radius R with velocity —( U B—
Umf / emf) at r = oo. Inside the sphere of penetration
9.3.2 Slow Bubbles Regime—When UB < Umf/s mf
of radius R, the fluid leaves the roof of the buble and
recirculates back to the base of the bubble as shown in In this case, the fluidizing fluid moves upward relative
Fig. 11. The radius of the penetration for a 3-D bed can to the bubble motion. This case is usual for beds o f
be calculated from Eq. (101). large particles and small bubbles. Since R is less than
zero in this case, the majority of fluidizing fluid enters
1/3
UB the void at the base and leaves from the roof. The
+ 2
fluidizing fluid, in essence, uses the bubble void as
R = Rv (101)
the shortcut. The fluidizing fluid penetrates the p arti­
- 1
, ( ^ m f / s mf culate phase freely from the bubble except for a small
fraction of fluid in the shaded area, as shown in Fig. 12
In this case, the bubbles in the fluidized bed are accom­ in a circle of radius a \ expressed in the following equa­
panied by a “cloud” while rising through the bed. For tion.
fast bubbles where the bubble velocity is large, or for
fluidized beds of fine powders when the minimum
a' = t § = 0.8|*| (103)
fluidization velocity is small, the cloud is usually very

Gas Gas
Streamlines Streamlines

Fast Bubbles UB > Umf / emf

Figure 11 Davidson’s bubble model—fast bubbles.

Bubbling Fluidized Beds 71

regime that the similarity between the form ation of

bubbles in a fluidized bed and that in an inviscid liquid
is most applicable, because the surface tension in a
fluidized bed is zero. A t still higher gas rates, the
m om entum of gas issuing from the orifice can manifest
as a jet before breakup into bubbles. The existence of a
“jet” in a fluidized bed, however, has become a con­
troversial issue in the literature that will be discussed in
more detail later.
Using the analogy of bubble formation in an in­
viscid liquid, Davidson and H arrison (1963) derived
equtions for both the bubble frequency and the bubble
size (volume), assuming there is no gas leakage from
the bubble to the emulsion phase.
0.2
j _ _ /6 g ;
t = ZOjS (105)

1.2
Figure 12 Davidson’s bubble model—slow bubbles. G
Vn 1.138 ~ 0.6
(106)

At high gas flow rates where the bubble sizes are inde­
The stream function for the particles in this case can be pendent of the bed viscosity, the inviscid liquid theory
similarly expressed as in Eq. (102). can predict the bubble sizes satisfactorily. At low gas
In both these cases, i.e., UB > Umf/e mf and flow rates where the viscosity effects are quite pro­
UB < Umf/e mf, the gas exchange between bubble and nounced, the inviscid liquid theory underestimates
particulate phase was obtained by Davidson through the bubble sizes. In this case, the following equation
an analogy to a fixed bed: by Davidson and Schuler (1960) should be used.
q = 3nRBUmf for a 3-D bed (104) U.2
Vu = 1.378- r0.6 (107)
The gas exchange rate q expressed in Eq. (104) is the
total volumetric flow rate of gas passing through the Equation (107) fits the data by Harrison and Leung
bubble void. It can either pass through the bubble void (1961) and the data by Bloore and Botterill (1961)
or be recirculated back to the bottom of the bubble, fairly well. There is, however, increasing evidence
depending on the relative m agnitude of the bubble showing that the gas leakage from the bubble to the
velocity and the minimum fluidizing velocity. emulsion phase can be substantial, especially for large
There are other models for the bubble motion in particles. Nguyen and Leung (1972) performed experi­
fluidized beds of more complexity, primarily those by ments in a 2-D bed with a fluidizing velocity of 1.2
Jackson (1963) and M urray (1965). They are more times the minimum fluidizing velocity and found that
exact than the D avidson’s model but give essentially the bubble volume could be better approxim ated by
the same results. the following equation:
C1
9.4 Bubble Formation in a Fluidized Bed VB = 0.53— / n = bubble frequency (108)
Jn

Bubble formation in a fluidized bed was found experi­ From the computer enhanced video images of rising
mentally to be very similar to that in an inviscid liquid. bubbles in fluidized beds, Yates et al. (1994) observed
At a very low gas flow rate, the frequency and size of that the bubbles are surrounded by a region of emul­
the bubbles formed are primarily governed by a bal­ sion phase in which the solids concentration is lower
ance between the surface tension of the liquid and the than that in the emulsion phase far from the bubbles.
buoyancy force of the bubble. The inertia of the liquid This region of increasing voidage was called the “ shell”
moved by the rising bubbles becomes more im protant by Yates et al. The volumetric gas in the bubble and in
than the surface tension at higher gas rates. It is in this the shell can be correlated as
72 Yang

-0.42
K = 16n (109) For orifice plates, the following equation can be used:
,0.4
Based on this equation, the volume of void following
D b = 1.43 ryO.2
coalescence of two bubbles is about 28% larger than
the volumes of the two constituent voids, because of 0.94
(113)
the incorporation of gas from the shells.
The constant 0.027 is a dimensional constant with
9.5 Coalescence of Bubbles in Fluidized Beds units of (cm/s)-0'94. Equation (113) gives reasonable
agreement with data from fluidized beds using indus­
The bubbles in fluidized beds grow in size due prim ­ trial types of orifice distributor plates. Porous distribu­
arily to three factors: tor plates, as expressed in Eq. (112), behave as though
they contained approximately 1 hole per 10 cm2 of bed
1. The effective hydrostatic pressure descreases
area. The principal effect of adding fines to a fluidized
toward the top of the fluidized bed.
bed of group B powders is the reduction of the mean
2. Bubbles coalesce in the vertical direction with
particle size. At equal values of excess ( U — t/mf), this
the trailing bubble catching up the leading
results in increased bed expansion and solid circulation
bubble, and
rates but produces no decrease in mean bubble size.
3. Bubbles coalesce in the horizontal direction
M ori and Wen (1975) assumed that all gas above
with the neighboring bubbles.
the minimum fluidizing velocity went to form a single
The effect of the hydrostatic pressure is usually small, train of bubbles rising along the center line of the bed
and the bubbles grow in size owing largely to coales­ and calculated the diameter of bubble that would exist
cence. There are a num ber of bubble coalescence as
models in the literature. A few of the more well-
D bm = 0.652[v4(t/ - UmS) f /5 (114)
known ones are discussed here.
Geldart (1972) found that the fluidization behavior The bubble diameter at bed height H can then be esti­
of G roup B powders was independent both of the m ated as
mean particle size and of particle size distribution. In
particular, the mean bubble size was found to depend D BM - A l ^ 6X / Q3g
(115)
only on the type of the distributor, the distance above D bm “ D Bo \ D
the distributor plate, and the excess gas velocity above
For perforated plates, the initial bubble diameter D Bo
that required at the minimum fluidization condition,
is expressed as
U — Umf. M athematically, it can be expressed as
2/5
Dp : DBo + K H n(U - Umff ( 110) A ( u - u mfy
D Bo = 0.347 (116)
Nn
According to the theory by Davidson and Harrison
(1963), the size of a bubble issuing from a single orifice where A is the area of the bed and N 0 is the total
in a fluidized bed at the minimum fluidization condi­ num ber of orifices. For porous plates, the following
tion can be calculated from Eq. (106). The same idea expression should be used to estimate the initial bubble
was extended by G eldart to multiorifice distributor sizes.
plates by replacing G with (U — Umf ) / N Q and by
D Bo = 0.00376(1/ — Dmf)2 (117)
replacing VB with (1 —/ w)tcDBo/ 6 . By assuming the
wake fraction / w to be 0.25, we have The validity of the above equations has been tested
within the ranges of the following parameters:
u - u mf\ 0A
Nn 30 < D < 130 cm 0.5 < Umf < 20 cm /s
A* = 1.43 Q2 _ + K H *(U - ~ Um£)m (111)
0.006 < dp < 0.045 cm U — Dmf < 48 cm /s

Experimentally, it has been found that for the porous Rowe (1976) suggested the following equation to esti­
plates, the following equation applies. mate the bubble size in a fluidized bed:

Z>B = 0.915(17 — Um(f A + 0 m i H { U - f/mf)0'94 (U - UM ) l/2( H + A0)3/4

(118)
( 112)
Bubbling Fluidized Beds 73

Here hQis an emperical constant and is a characterises Substituting Eq. (120) into Eq. (119), we have
of the distributor plate. The h0 is effectively zero for a
2/5
porous plate buy may be more than a meter for large \ U - Umf) A c
De(0) — 1.63 ; A c = catchment area
tuyeres.
D arton et al. (1977) assumed that the bubbles are
(121)
lined up as close together as possible, as shown in Fig.
13. They also defined a so-called “catchment area” for The catchment area is defined as the area of distributor
each particular bubble track. The bubble frequency plate per hole. When two bubbles of equal volume
can then be calculated by UB/ 2 R B with the bubble from the nth stage coalesce to form a bubble of the
velocity UB = 0.711(gDe)1/2. The bubble flow in each (n + l)th stage, we have
track can be calculated as follows assuming the two-
phase theory. ,3
D ;e (n + l)
-
2D:e(n) (122)

If we further assume that the height of each bubble

coalescence stage is proportional to the diameter of
the catchment area, we have
(0.71 l yf eDl ) (119)
A = -a d 2c and : = ADr (123)
If the bubbles are hemispheres, the volume of each
individual bubble can be calculated by the equation where Dc is the diameter of a circular catchment area
for each bubble stream. We have from Eq. (123) that
jiDi
(120)
: A D C0 + AT) Cl -A D ,c ( n - l ) (124)

Fully Coalesced
Bubble Stream

y V

Bubble
Coalescence

Bubble Bubble

i \

.9 . .9 . 9 9
r r ~\ r ~\ r \ r \ r \ r \ r a
o o o o o o o o o o o o o o o o
Figure 13 Dalton’s bubble coalescence model.