Influence of Particle Size Distribution On The Performance of Fluidized Bed Reactors

INFLUENCE OF PARTICLE SIZE DISTRIBUTION ON T H E
P E R F O R M A N C E OF FLUIDIZED BED REACTORS

By
Guanglin Sun
M. A. Sc. (Chemical Engineering) Zhejiang University
A THESIS S U B M I T T E D IN PARTIAL F U L F I L L M E N T OF
T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF
D O C T O R O F P H I L O S O P H Y
in
T H E F A C U L T Y OF G R A D U A T E STUDIES
C H E M I C A L E N G I N E E R I N G
W e accept this thesis as conforming

to the required standard
T H E UNIVERSITY OF BRITISH C O L U M B I A
1991
© Guanglin Sun, 1991

In presenting this thesis in partial fulfilment of the requirements for an advanced
degree at the University of British Columbia, I agree that the Library shall make it
freely available for reference and study. I further agree that permission for extensive
copying of this thesis for scholarly purposes may be granted by the head of my
department or by his or her representatives. It is understood that copying or
publication of this thesis for financial gain shall not be allowed without my written
permission.
Department
The University of British Columbia

Vancouver, Canada
DE-6 (2/88)
Abstract
The effect of particle size distribution (PSD) on the performance of afluidized bed reactor
was investigated using the ozone decomposition reaction, combined with the study of
hydrodynamics, for fresh and spent fluid cracking catalysts, each having three particle
size distributions - wide, narrow and bimodal - all with nearly the same mean diameter
(60 ^m), the same particle density and the same B E T surface area. The superficial
gas velocity was varied from 0.1 to 1.8 m/s to include the bubbling, slugging, turbulent
and fast fluidization regimes. The catalytic rate constant, based on the volume of the
particles, ranged from 2 to 10 s , while the static bed height was varied from 0.15 m
- 1
to 1 m. Four different multi-orifice gas distributors with different hole diameters (2.2 to
5.1 mm) and hole numbers (4 and 21) were also tested to evaluate the influence of gas
distributor on the performance of fluidized bed reactors.
The particle size distribution was found to play a larger role at higher gas velocities
than at lower velocities. At low gas velocities (Uf < 0.2 m/s), the reaction conversion was
not greatly affected by the PSD. However, with an increase in gas velocity the PSD effect
became larger. The wide size distribution gave the highest reactor efficiency, defined as
the ratio of the volume of catalyst required in a plug flow reactor to that required in the
fluidized bed reactor to achieve the same conversion, while the narrow blend gave the
lowest. The differences are not solely a function of the "fines content".
The influence of particle size distribution on the hydrodynamics of fluidization was
evaluated by measuring particle concentrations in voids, bubble sizes, and dense phase
expansion. When the superficial gas velocity exceeded 0.1 m/s, the bed with the wide size
distribution usually gave the highest particle concentration inside the voids, the smallest
n
bubble size and the greatest dense phase expansion at the same operating conditions.
There is evidence that there is a greater proportion of "fines" present in the voids than in
the overall particle size distribution. This has been explained in terms of the throughflow
velocity inside bubbles being of the same order as the terminal velocity of typical "fines",
causing these particles to spend longer periods of time inside the voids.
The effect of the PSD on the fluidization regime and its transitions was determined by
measuring pressure fluctuations along the column. The earliest transition from bubbling
or slugging to turbulent fluidization occurred in the bed of wide size distribution, while
the latest corresponded to the narrow PSD.
For particles of wide size distribution, higher conversion was achieved for the turbulent
and fast fluidization regimes than for the bubbling fluidization regime under otherwise
identical conditions, while for particles of narrow size distribution, the dependence of
conversion on regime was small. Hence, for reactors of wide PSD, the performance can
be improved significantly by operating in the turbulent or fast fluidization regime, while
for particles of narrow size distribution, the benefit of operating at high gas velocity is
slight at best.
The PSD influence should be considered in modelling fluidized bed reactors. The
"Two-Phase Bubbling Bed Model" has been modified to account for PSD effects. For
the reactor of wide particle size distribution operated at high gas velocities, a single-phase
axial dispersion model with closed inlet and open outlet boundary conditions appears to
be suitable to predict the performance.
It was also found that a high pressure drop across the gas distributor was not sufficient
to maintain good performance of the distributor. The reactor efficiency in the entry region
was higher for a distributor with a greater number of orifices, even though it had a lower
pressure drop, than for a distributor plate with fewer larger holes.
iii
Acknowledgement
I would like to express my great appreciation to Dr. J.R. Grace for his excellent super-
vision and overall support.
Special thanks are also due to the ladies and gentlemen of the Department of Chemical
Engineering, especially to the Workshop and the Stores, for their invaluable assistance.
This project would not have been possible without the financial assistance of the
donors of the Petroleum Research Fund, administered by the American Chemical Society;
a Graduate Fellowship awarded by the University of British Columbia; a Fellowship
provided by Alcan; a Scholarship awarded by Pao; research funding from the Natural
Sciences and Engineering Research Council of Canada. To each of these goes a very
special thank-you.
The assistance of the Esso Petroleum IOCO refinery in Port Moody, B.C. in supplying
the catalyst particles needed in this study is also gratefully acknowledged.
iv
Table of Contents
Abstract ii
Acknowledgement iv
List of Tables vi
List of Figures vii
1 Introduction 1
1.1 PSD effect on the hydrodynamics of fluidized beds 2

1.2 PSD effect on the performance of fluidized bed reactors 6
1.3 Some basic problems with studies of PSD effect 8
1.3.1 Can the PSD effect be separated from the influence of the mean
particle size? 8
1.3.2 Is the "fines" content sufficient to express the particle size distri-
bution? 9
1.3.3 Is the influence of the PSD the same in different hydrodynamic
regimes and bed regions? 11
1.4 Objectives of the present study 11
2 Experimental Particles and Apparatus 13
2.1 Preparation of Particle Blends . 13

2.1.1 Particle separation and blending 13
2.1.2 Catalyst activation and characterization 14
v
2.1.3 Relevant properties of the particles 16
2.2 Experimental Apparatus 26
2.2.1 Gas supply system 26
2.2.2 Reactors 28
2.2.3 Particle collection and return system 30
2.2.4 Gas Distributors 32
2.2.5 Instrumentation 35
3 Reactor Performance at Low Gas Velocities 36
3.1 Rate Constant Determination 36
3.2 Stability of Catalyst Activity 40
3.3 Definition of Efficiency for Fluidized Bed Reactors 40
3.4 Effect of P S D on Conversion and Reactor Efficiency 41
3.5 Influence of Whether Fines or Coarse Particles are Active 49
3.6 M o d e l Predictions of P S D Effect at Low Gas Velocities 52
3.7 Influence of Overall Catalyst Activity on the P S D Effect 56
4 Reactor Performance at Higher Gas Velocities 60
4.1 Previous Work on Fluidization Performance at High Gas Velocities . . . 61
4.2 Dependence of P S D Effect on the Superficial Gas Velocity 64
4.3 Influence of Fluidization Regime on Reactor Performance 68
4.3.1 Regime influence for particles of wide size distribution 69
4.3.2 Regime influence for catalysts of narrow size distribution 73
4.4 Prediction of P S D Effect at High Gas Velocities 75
5 Effect of Particle Size Distribution on Two-Phase Behaviour 81
5.1 Particle Concentration i n Voids 82
vi
5.1.1 Theory underlying the measurements 83
5.1.2 Laser photometer and its calibration 86
5.1.3 Particle and fines concentration in voids 90
5.2 Effect on the Bubble or Void Size . . . 105
6 Pressure Fluctuations and Regime Transitions 117
6.1 Brief review of previous work 117

6.2 Experimental Equipment and Procedure 120
6.3 Influence of PSD on Pressure Fluctuations 121
6.4 Influence of PSD on Regime Transitions 126
6.5 Axial Voidage Distribution . 132
7 Reactor Performance in the Distributor Region 139
7.1 Brief Review of Previous Work 139

7.2 Experimental Method 144
7.3 Influence of Distributor Geometry on Reaction Conversion 147
8 Conclusions and Recommendations for Further Work 161
8.1 Overall Conclusions 161

8.2 Recommendations for Further Work . 163
Nomenclature 165
References 169
Appendices 185
A Minimum integral reactor length of minimize axial diffusion effects 185
vii
B Photometric measurement for spent F C C particles 187
C Statistical Analysis of Experimental Results 191
Cl Significance test of the influence of the PSD 191
C.2 Confidence intervals of experimental results 193
D Raw Data from Chemical Reaction Tests 195
E Typical measurement results of particle size and size distribution 204
vm
L i s t of Tables
2.1 Measured particle and bulk densities for fluid cracking catalyst 17
2.2 Surface area of FCC particles 19
2.3 Distributors tested in the experiments 33
3.1 Determination of the rate constant for 10 g samples of N-FCC catalysts

of wide size distribution 37
3.2 Experimental test on the interphase transport gradient for N-FCC parti-
cles of wide size distribution 39
5.1 Void size and bed expansion in two dimensional fluidized bed 108
6.1 Transition velocity U for N-FCC particles of different size distributions

c
over intervals Az (30 to 280 mm), Az (430 to 680 mm), Az (680 to 930
x 2 3
mm) and Az (930 to 1180 mm)

4 127
C. l Reaction conversions measured in the fluidized bed reactor for particles of

different size distributions; U = 0.5 m/s, k — 9 s~ , catalyst inventory =
T
l
5 kg 192
D. l Raw conversion data of ozone decomposition for particles of wide size

distribution 196
D.2 Raw conversion data of ozone decomposition for particles of narrow size
distribution 196
ix
D.3 Raw conversion data of ozone decomposition for particles of bimodal size
distribution with coarse fraction active 197
distribution with fine fraction active 197
distribution with different fraction active; catalyst inventory: 5.05 kg . . 198
D.6 Raw conversion data of ozone decomposition for particles of different size
distributions; catalyst inventory: 5.05kg 198
D.7 Raw conversion data of ozone decomposition for particles of wide size
distribution with different activities; catalyst inventory: 5 kg 199
D.8 Raw conversion data of ozone decomposition for particles of narrow size
D.10 Raw conversion data of ozone decomposition in fluidized beds with differ-
ent gas distributors, for particles of wide size distribution; catalyst inven-
tory: 1.2 kg 201
D. 11 Raw conversion data of ozone decomposition in fluidized bed with different
gas distributors, for particles of wide size distribution; catalyst inventory:
3 kg 201
D.12 Raw conversion data of ozone decomposition in fluidized bed with different
gas distributors for particles of narrow size distribution; catalyst inventory:
1.2 kg 202
D.13 Raw conversion data of ozone decomposition in fluidized bed with different
distributors for particles of narrow size distribution; catalyst inventory: 3
kg 202
x
D. 14 Axial dispersion model test for particles of wide size distribution at higher
gas velocities; catalyst inventory: 5 kg 203
E. l Size and size distribution for N-FCC particles of wide PSD 205
E.2 Size and size distribution for N-FCC particles of wide PSD after being
activated 206
E.3 Size and size distribution for N-FCC catalysts of wide PSD after being
fluidized in the reactor for 5 hours 207
xi
List of Figures
1.1 Coke particle size distributions. Identifying numbers on curves correspond

to coke materials of several particle size distributions obtained by grinding.
(Matsen, 1973) - 10
2.1 Dimensions of the infrasizer 15

2.2 Size distributions of FCC particles 18
2.3 SEM photograph for N-FCC particles of wide size distribution 20
2.4 SEM photograph for N-FCC particles of wide size distribution after being
activated 21
2.5 SEM photograph for N-FCC particles of bimodal size distribution . . . . 22
2.6 SEM photograph for N-FCC particles of narrow size distribution 23
2.7 SEM photograph for S-FCC particles of bimodal size distribution .... 24
2.8 SEM photograph for S-FCC particles of narrow size distribution 25
2.9 Schematic of experimental equipment 27
2.10 Detailed drawing of the fluidization column 29
2.11 Detailed drawing of the particle collection system 31
2.12 Detailed drawing of gas distributor 34
3.1 Variation of apparent reactant fraction unconverted with height above the
distributor for narrow size distribution with kinetic rate = 9.1 s~ ; catalyst
l
inventory: 5 kg 43
xii
3.2 Effect of PSD on ozone conversion for N-FCC with k = 4.1 to 4.3 s ;
T
_1
catalyst inventory: 5.05 kg. PF and PM are corresponding results for
single-phase plug flow and perfect mixing respectively 45
3.3 Effect of PSD on ozone conversion for N-FCC with k = 2.1 to 2.2
r s' ; 1
catalyst inventory: 2.75 kg. PF and PM as in Fig. 3.2 46
3.4 Effect of PSD on contact efficiency for N-FCC with A: = 4.0 to 4.3 a ;
r
- 1
catalyst inventory: 5.05 kg; "fa" and "ca" denote fines active and coarse
active, respectively 47
3.5 Effect of PSD on contact efficiency for N-FCC with fc = 2.1 to 2.2 a ;
r
- 1
catalyst inventory: 2.75 kg; "fa" and "ca" denote fines active and coarse
fraction active, respectively 48
3.6 Effect of which fraction is active on conversion for bimodal catalyst; k =

T
4.0 to 4.2 a ; catalyst inventory: 5.05 kg

-1
50
3.7 Influence of overall catalyst activity on the PSD effect; catalyst inventory:
5 kg 57
3.8 Influence of overall catalyst activity on the PSD effect; catalyst inventory:
5.05 kg 58
4.1 Influence of gas velocity and PSD on reactor conversion; catalyst inventory:
5 kg, k % 2.4 s'

r
1
65
5 kg, fc « 4.5 s'

r
1
66
5 kg, k K 9 s'
T
1
67
4.4 Influence of hydrodynamic regime on ozone decomposition for catalysts of
wide size distribution 70
xm
4.5 Influence of hydrodynamic regime on ozone decomposition for catalysts of
narrow size distribution 71
4.6 Influence of hydrodynamic regime on contacting efficiency for particles of
wide size distribution 72
4.7 Influence of hydrodynamic regime on contacting efficiency for particles of
narrow size distribution 74
4.8 Axial dispersion model for fluidized bed reactors 77
5.1 Dimensions of the two-dimensional fluidization column 87

5.2 Measurement apparatus used to determine particle concentration in voids 88
5.3 Particle concentration calibration curve 91
5.4 Photometric measurement for N-FCC particles with wide PSD 92
5.5 Photometric measurement for N-FCC particles with narrow PSD 93
5.6 Photometric measurement for N-FCC particles with bimodal PSD . . . . 94
5.7 Influence of PSD on solids hold-up in voids at lower gas velocities . . . . 95
5.8 Influence of PSD on solids hold-up in voids at a gas velocity of 0.4 m/s . 96
5.9 Influence of PSD on solids hold-up in voids at higher gas velocities . . . . 97
5.10 Influence of PSD on the frequency of detectable voids (i.e. with the solids
concentration less than 5% by volume) at different gas velocities and at z
= 0.85 m 100
5.11 Particle size distribution for an industrial fluid cracking catalyst, /j , and
|0
corresponding calculated size distribution inside voids, fiiV 103

5.12 Schematic of experimental apparatus for observing bubble phenomena . . 107
5.13 PSD effect on the properties of bubbles in fluidized beds of N-FCC particles
at U = 0.6 m/s. Left and right photographs correspond to wide and narrow
size distributions respectively 109
xiv
at U — 0.4 m/s. Left and right photographs correspond to wide and narrow
size distributions respectively 110
size distributions respectively Ill
at U — 0.05 m/s. Left and right photographs correspond to wide and
narrow size distributions respectively 112
5.17 PSD effect on the properties of bubbles in fluidized beds of S-FCC particles
size distributions respectively. 113
5.18 PSD effect on the properties of bubbles in fluidized beds of S-FCC particles
size distributions respectively. 114
5.19 PSD effect on the bed expansion for N-FCC particles. Open symbols are
experimental values; filled in symbols are the corresponding predictions
from eqn. 5.13 with experimental values of Db, a • 116
6.1 Effect of PSD on pressure fluctuations with H f = 0.7 m over intervals

m
Az! (30 to 280 mm), Az 2 (430 to 680 mm) and A-z (680 to 930 mm). . .
3 123
6.2 Effect of PSD on pressure fluctuations with H f = l m over intervals A z j ,
m
Az 2 and Az (as in Fig. 6.1)

z 124
6.3 Effect of PSD on U for fluidized beds of N-FCC particles with different
c
H f and at different bed heights

m 129
xv
6.4 Influence of H j on U for N-FCC particles with narrow PSD, over inter-
m c
vals A z ! (30 to 280 mm), A z (430 to 680 mm), A z (680 to 930 mm) and
2 3
Az 4 (930 to 1180 mm) 130

6.5 Influence of H f on U for N-FCC particles with wide PSD, over intervals
m c
Azi (as in Fig. 6.4). 131

6.6 Effect of PSD on the overall voidage in fluidized bed; H ; = 0.7 m. In-
m
tervals A z and A z are as in Fig. 6.4

x 2 134
6.7 Effect of PSD on the overall voidage in fluidized bed; H f = l m . Intervals
m
Azi, Az 2 and A z are as in Fig. 6.4

3 135
6.8 PSD effect on the voidage increment in fluidized bed; H f = 0.7 m. In-
m
tervals A z i and A z are as in Fig. 6.4

2 136
7.1 Schematic showing regions where pressure drops and their fluctuations
were determined 146
7.2 Influence of distributor free area and pitch on conversion: N-FCC particles
of wide PSD, H mf = 0.15 m 148
of narrow PSD, H mJ = 0.15 m 149
of wide PSD, H mf = 0.4 m 150
of narrow PSD, H mf = 0.4 m 151
7.6 Distributor effect on reaction conversion at different bed heights: N-FCC
particles of wide PSD; Di denotes distributor-i, (see Table 2.3) 154
7.7 Distributor effect on reaction conversion at different bed heights: N-FCC
particles of narrow PSD; Di denotes distributor-i, (see Table 2.3) 155
xvi
7.8 Location of taps between which differential pressures were measured. . . 156
7.9 Effect of pitch on the pressure drop between ports 1 and 2: N-FCC of wide
PSD, U = 0.2 m/s 158
7.10 Effect of pitch on the pressure drop between ports 1 and 2: N-FCC of
narrow PSD, U = 0.2 m/s 159
B.l Photometric measurement for S-FCC particles with narrow PSD 188
B.2 Photometric measurement for S-FCC particles with wide PSD 189
B.3 Photometric measurement for S-FCC particles with bimodal PSD . . . . 190
XVll
Chapter 1
Introduction
Fluidized bed reactors are widely used in the process industries for solid-catalysed gas
phase reactions (e.g. in catalytic cracking, acrylonitrile manufacture, phthalic anhydride
production, polyethylene and polypropylene processes, etc.) and in gas-solid reactions
(ore roasting, combustion, gasification, catalyst regeneration, calcination reactions, etc.).
Advantages of fluid bed reactors include favourable heat transfer, temperature uniformity,
capability of continuous addition and removal of particles, and ability to scale up to very
large reactor sizes. A thorough review of industrial processes in fluidized bed reactors
was given by Yerushalmi (1982). For a general introduction to fluidized beds and their
hydrodynamic regimes see Geldart (1986).
The particle size distribution (PSD) is a key factor in determining the performance
and operation of fluidized bed reactors. Although many workers have made reference
to the importance of fines in determining the behaviour and performance of fluidized
beds (e.g. Matheson et al, 1949; Zenz and Othmer, 1960; Agarwal and Davis, 1966;
Abrahamsen and Geldart, 1980; Yates and Newton, 1986, etc.), and oil companies and
other operators or designers of fluidized bed reactors commonly apply empirical criteria
to ensure a level of fines which assists in establishing favourable yields and overall perfor-
mance, the role of fines content and the effect of the PSD are poorly understood. There
are at least three questions which need to be answered:
1. In what ways and to what extent do the PSD and the "fines" content affect the
performance of fluidized bed reactors?
1
Chapter 1. Introduction 2
2. Why do the PSD and the "fines" content affect the performance offluidizedbed
reactors?
3. Is it possible to characterize the influence of the PSD simply in terms of a "fines"
content, as assumed by some authors (e.g. Abrahamsen and Geldart, 1980; Dry et al,
1983; Avidan and Edwards, 1986, etc.)?
The performance of a fluidized bed reactor mainly depends on the hydrodynamics of
fluidization, transfer phenomena within the reactor and the kinetics of the reaction(s).
We are primarily concerned in this thesis with catalytic applications of fluidized bed
reactors. In such applications, the fluidized particles are generally small (almost always
< 100 fim in average size, falling in Group A of the Geldart (1973) classification) with
the result that catalyst effectiveness factors are very nearly one. The PSD primarily
influences the reactor performance, therefore, via its effect on the hydrodynamics of
fluidization and via its effect on the mass transfer in the reactor. In the present study,
both of these interrelated aspects are investigated. The third question is resolved by
separating the effect offinesand the effect of mean particle size in our experiments. The
role of the PSD on the operation of fluidized bed reactors is modelled and discussed.
Optimum PSD's for catalytic reactions are also considered.
1.1 P S D effect on the h y d r o d y n a m i c s of fluidized beds
The performance of fluidized bed reactors is strongly related to the hydrodynamics of

fluidization, to those fundamental factors controlling the behaviour of the fluid-solid
systems. Several years after the successful application of large scale commercial fluidized
bed reactors composed offinelypowdered solids in the cracking process of petroleum, the
effect of "fines" on the character of fluidized beds began to interest a few investigators.
In an early study, Matheson et al (1949) measured the torque necessary to rotate a
paddle suspended in fluid beds made up of binary mixtures of synthetic cracking catalyst
to explore the effect of the PSD on bed "viscosity". Relatively small amounts of fines
added to a coarse cut were found to decrease the "viscosity" markedly. This trend
was maintained up to fines concentrations of about 30%, after which the effect became
less marked. Morse and Ballou (1951) investigated the uniformity of fluidization using
capacitance signals and found that uniformity was better for non-uniform mixtures of
particle sizes than for narrow cuts.
Zenz and Othmer (1960) evaluated early research work on the effects of the PSD on the
pseudo viscosity offluidizedbeds and concluded that the introduction of fines between the
coarse grains could act as a lubricant to reduce the friction between the coarse paticles,
thereby reducing the bed viscosity and helping to maintain "good fluidity". They also
noted that PSD influences the efficiency and loading of fines recovery facilities, as well
as the operation of pneumatic conveyers.
The effect of fines content on the hydrodynamic character of fluidized beds has also
attracted some attention. Agarwal and Davis (1966) compared the minimum fluidization
velocity for closely sized fractions of iron ore to that with a wide size range, and found
that the greater the "fines" content, the lower the minimum fluidization velocities. In
another study, de Groot (1967) investigated the effect of the amount of "fines" on bed
expansion and reported that the expansion increased with addition of "fines", especially
in columns of diameter larger than 0.5 m. While measuring the distribution of voidage
in the particulate phase near rising bubbles by a capacitance technique, Lockett and
Harrison (1967) observed that it became more difficult to define a real boundary between
the bubble and particulate phases as the particle size distribution increased. That is, the
difference between the two phases appeared to be less sharp for the system containing
more fines.
More recently, Abrahamsen and Geldart (1980) measured hydrodynamic properties

of fluidized beds composed of powders with different mean particle size, density and
proportion of fines (defined as particles smaller than 45 ^m). and found that the minimum
bubbling velocity, the dense phase interstitial gas velocity and the bed expansion all
increased with the addition of fines. Similar results were reported by Sun et al (1983)
with different gas properties and Dry et al (1983) with wider size range. In the latter
case, the demarcation between "fines" and the rest of the material was taken to be 22
fim. In their correlations for minimum bubbling velocity, dense phase voidage and dense
phase gas velocity, "fines" content was taken into account and played a significant role.
Avidan and Yerushalmi (1982) investigated the bed expansion of fine powders over a
range of gas velocities spanning different fluidization regimes and noted that the presence
of fines has far-reaching effects on the 'quality' of fluidization. They found that fresh
catalyst with a higher proportion offinesshowed less tendency to segregate, lower cluster
tenninal velocities and lower transition velocities to turbulent fluidization than older
catalysts from which most of the fines had been lost. Werther (1984) measured the
local visible bubble flow, including local bubble size and average bubble rise velocities
in fluidized beds and discovered that increased fines (defined as particles with d < 44
p
fim) could increase the mass transfer rate due to decreasing equiUbrium bubble size
with an increase in the proportion of fines. Nacastro and Glicksman (1984) investigated
pressure fluctuations within a fluidized bed and noted that variation in the particle
size distribution could cause non-similarity of fluidization. The PSD was said to be an
important parameter for modelling fluid dynamic behaviour.
While scaling up the methanol-to-gasoline process, Avidan (1986) found that bubble
suppression was achieved in a turbulent fluidized bed with a minimum of 15% fines, where
fines were defined as particles smaller than 40 fim. Kono et al (1986a) investigated the
emulsion phase characteristics for fluidization of fine powder fluidized beds and reported
that the minimum complete fluidization velocity and bed expansion ratios were somewhat
larger for beds containing more fines than for beds with few fines.
Since the quality of fluidization is supposed to be substantially determined by the
characteristics of the emulsion phase of the bed, such as its viscosity (Grace, 1970), the
fluid bed hydrodynamics can be related, to a certain extent, to the rheological charac-
teristics of the powders. Generally speaking, the less rigid the structure of the emulsion
phase, the more homogeneneous the fluidized system. The PSD effect on the powder rhe-
ology, therefore, has also been studied by researchers, not only in the area of fluidization,
but also in other powder handling systems.
By measuring the tensile strength of powders around the minimum bubbling point,
Kono et al (1986b) observed that both the tensile strength and the plastic deformation
coefficient evaluated from the bed expansion could be decreased by adding fine powders
to a narrow cut of particles. Lloyd and Webb (1987) measured the shear stress of powders
with different fines content in both an aerated and non-aerated state. They noted that
the rate of reduction of the shear stress with increased aeration is dependent on the fines
content, and the addition of particles around 30 pm in size had a large effect. They also
found the addition of fine particles could cause the powder to behave as 'liquid like' at
lower shear velocities.
For spent FCC particles, investigators at UBC (Ip, 1988; Khoe et al, 1991), based
on collapse tests, observed that t/ j,/C/ ^ for particles of wide size distribution was 20%
m TO
greater than for narrow PSD. They reported that for a given average surface-volume
diameter, a wide size distribution gives a greater air retention capacity than a bimodal
distribution which in turn gives greater retention than a narrow distribution. By mea-
suring the pressure fluctuations in a fluidized bed, Ip also found that the magnitude of
pressure fluctuations increases as the particle size spread become more narrow. However,
the frequency of fluctuations is independent of the size spread.
Fines not only affect the hydrodynamics of fluidization, but also the flow character
of particles (Zenz and Othmer, 1960). Changing the particle size distribution allows the
particle flow properties to be modified (Kurz and Miinz, 1975; Geldart and Radtke, 1986).
Reterman (1985) investigated catalyst circulation problems in fluid catalytic cracking
units due to the loss of fines through a damaged cyclone and presented a parameter,
Fprop, which depended on the fines fraction (fines being defined as particles with d p <
45 fim), to express the flow properties and predict the onset of circulation problems.
Kwauk and his co-workers (Tung et al, 1989) carried out an extensive series of bed
collapse experiments for different gas-solid systems and presented a dimensionless param-
eter, 0 , mainly depending on the collapsing time of fluidized beds at certain conditions,
to express the hydrodynamic performance of fluidization. Based on the comparison of
0 values for different powders, they suggested that blending of particle fractions with
different sizes can be used to improve the homogeneity of fluidization.
1.2 P S D effect on the performance of fluidized bed reactors
Very little work has been performed to determine the influence of the PSD on the per-
formance of catalytic fluidized bed reactors. The performance of catalytic fluidized bed
reactors depends not only on the hydrodynamics of fluidization, but also on other fac-
tors, such as diffusion resistance between gas and particles, catalyst activity, reactant
conversion level and the nature of the reaction process. A number of these factors may
be related to the mean particle size and to the particle size distribution.
Yates and Newton (1986) investigated the oxidative dehydrogenation of butene to
butadiene in a reactor of 0.11 m in diameter at superficial gas velocities from 0.06 m/s to
0.32 m/s. The fines content (fines being defined as smaller than 45 pm) of the catalyst
was 0%, 16% and 27%. The conversion of reactant was found to increase with the
increase in the proportion of fines. The authors suggested that more gas flowed through
the emulsion phase of the bed, at the expense of gas flowing through the bubble phase,
in fluidized beds with more fines.
Pell and Jordan (1988) studied the PSD effect in a pilot scale reactor where propylene
was reacted with air and ammonia to produce acrylonitrile. The experiments were carried
out at three different superficial gas velocities and at five different fines levels: 23, 27,
32, 40 and 44% by weight of particles smaller than 45 pm. The proportion of fines was
varied by adding -45 pm enriched material to the existing charge. It was found that the
conversion usually increased as the fines content increased. At higher gas velocities, the
improvement in conversion due to addition of fines was much less than that at lower gas
velocities.
The importance of the PSD on the performance of fluidized bed reactors has been
recognized by industrial companies. It is common to control the PSD in fluidized beds
during long periods of operation by adding fines from time to time to make up for those
lost during the operation and to increase the collection efficiency of cyclones. It has
been found (Sun and Chen, 1986) that the yield of a commercial fluidized bed reactor
producing about 26000 tonnes of acrylonitrile per year can be increased by 2-3 per cent
if the content of fines (less than 45 pm) is raised from 15% to 32% by weight.
Although high "fines" level can often give a better performance in fluidized beds,
caution is necessary at such generalizations. Temperature runaway once happened in the
freeboard region of a commercial fluidized bed reactor of propylene ammoxidation due
to a large quantity of fine particles entrained from the bed with too high "fines" content
(CGCPC Report, 1985). Moreover, if the fines become predominant, the behaviour of
the powder becomes determined by interparticle forces, i.e. the powder behaves as a
group C (cohesive) powder in the classification scheme of Geldart (1973).
Cooper and Clough (1985, 1986) have recently attempted to establish experimental
methods to obtain on-line, real-time measurements of PSD in fluidized bed reactors.
Their apparatus includes a continuous flow particle sampling system, an optical spec-
trometer and a process computer system. The sampling system is closed-loop so that all
sampled particles can return to the bed immediately after being analyzed.
1.3 Some basic problems with studies of P S D effect
Although the importance of fines on the performance of fluidized bed appears to have
been established by the studies cited, the PSD effect on the performance of fluidized
bed reactors has not previously been subjected to systematic investigation. Several basic
problems connected with previous studies are considered in this work. Some of the key
questions to be addressed are as follows:
1.3.1 C a n the P S D effect be separated from the influence of the mean par-
ticle size?
Almost all studies cited above on the effect of particle size distribution discussed the
problem in terms of the influence of "fines" content. It should be noted, however, that
addition of fines affects both the mean particle size and the size distribution. The effect of
"fines" content, therefore, can be divided into two aspects: the effect of average particle
size and the effect of particle size distribution. It is clear that the influence of the
mean particle size on fluidized bed performance is very significant, especially for group A
materials, typical of fluid bed catalytic reactors. This effect has been considered in most
correlations and results published on fluidization.
By measuring the bed expansion and bubble frequency, Matsen (1973) investigated
the influence of mean particle size on the maximum stable bubble size in fluidized beds.
Six kind of particles with different mean particle size (from 26 pm to 150 pm) were
tested. Each kind of particles had different "fines" content (from 0% to 40% by weight)
where "fines" are defined as those less than 20 pm. Matsen found that the maximum
stable bubble size was very sensitive to mean particle size, while the nature of the size
distribution was also very critical. This suggests that the PSD effect should be separated
from the influence of mean particle size if we want to explore the PSD effect without
interference of other relative factors.
1.3.2 Is the "fines" content sufficient to express the particle size distribu-
tion?
No general method of expressing the PSD has been accepted by investigators of fluidized
beds. Although most investigators characterized PSD by "fines" level, the definition of
"fines" has differed from worker to worker. Some workers (e.g. Geldart and Abrahamsen,
1980; Yates and Newton, 1986) have defined fines as those particles which pass through
a sieve of opening 45 pm. Dry et al (1983), however, defined the demarcation between
the fines and the rest of the material to be 22 pm, while Avidan and Edwards (1986)
distinguished on the basis of 40 pm. Evidently, different definitions of "fines" will lead to
different results. Moreover, any such definition will be entirely arbitrary, and it is clear
that addition of significant volumes of material lpm larger than the chosen limit will
give nearly the same result as adding the same volume of material 1 pm smaller than the
limit.
To illustrate the problem, consider the PSD of powders, as shown in Fig. 1.1, used
by Matsen (1973) in a study of maximum stable bubble size for particles with different
size and size distributions. If "fines " are defined as those less than 45 pm, coke 4 and
coke 5 have the same "fines" level. If the "fines" are defined as those less than 20 pm or
15 pm, "fines" concentrations in coke 4 and coke 5 are then different. The experimental
maximum stable bubble diameters determined by Matsen were different for these two
distributions of particles (64 mm for coke 5 and 127 mm for coke 4), even though the
Chapter 1. Introduction
10
CUMULATIVE WEIGHT PERCENT
Figure 1.1: Coke particle size distributions. Identifying numbers on curves correspond to
coke materials of several particle size distributions obtained by grinding. (Matsen, 1973)
mean particle diameters were almost the same for the two powders.
1.3.3 Is the influence of the P S D the same in different hydrodynamic regimes
and bed regions?
Most previous experimental studies of chemical reaction in fluidized beds (e.g. Lewis et
al, 1959; Hovmand and Davidson, 1968; Chavarie and Grace, 1975; Bauer and Werther,
1981; Yates and Newton, 1986) have been restricted to relatively low gas velocities, seldom
exceeding 0.3 m/s. Commercial fluidized beds, however, usually operate at higher gas
velocities (above 0.45 m/s) where different hydrodynamic regimes (e.g. turbulent and
fast fluidization) may be encountered (Grace, 1986). It is possible that the PSD effect
may differ in the various regimes, as indicated by the experimental result of Pell and
Jordan (1988).
It is now well known that reactor characteristics differ widely between the region
immediately above the gas distributor (grid region), the bed proper and the freeboard
region. Since mass transfer resistance appears different in different regions, the extent of
the PSD effect may also change from region to region.
1.4 Objectives of the present study
The objectives of this study are to analyze and explain the basic problems mentioned
above. Briefly the objectives are:
1. To evaluate the PSD effect on the performance of catalytic fluidized bed reactors.
2. To examine the PSD effect in different fluidization regimes and different bed
regions.
3. To investigate the PSD effect on the hydrodynamics of fluidization.
4. To explore the reasons for the influence of "fines" and of particle size distribution.
5. To express the PSD effect on the performance of fluidized bed reactors in different
regimes and regions by suitable reactor models.
6. To test other factors related to the performance of catalytic fluidized bed reactors,
such as the kinetic rate and the gas distributor.
Because PSD effects appear to be most significant for materials which belong in group
A of the Geldart (1973) classification and fluid bed catalysts almost always fall into this
category, this study will be limited to materials of this group.
Chapter 2
Experimental Particles and Apparatus
The main preparation work in this study can be divided into two parts: (a) experimental
materials, including particles and gas, and (b) experimental equipment including the
fluidization column, fixed bed column, instrumentation, data acquisition system and
other accessories. Special attention has been paid to the catalyst powder preparation in
order to achieve the objectives of this study. Two of the key points are:
1. separating the PSD effect from the influence of mean particle size, and
2. keeping the PSD within the experimental system as constant as possible during
each operation period, even at high gas velocities.
2.1 Preparation of Particle Blends
Two kinds of particles, spent fluid cracking catalyst (S-FCC) and fresh fluid cracking
catalyst (N-FCC), were used in the experiments. These were obtained from the Esso
loco refinery in Port Moody, B.C. The mean diameter, defined as lfL(xi/dpi), was about
60 fim in each case. For each kind of particles, three batches were prepared, labelled
"wide", "narrow" and "bimodal", each with the same mean diameter and nearly the
same BET surface area and particle density, but different particle size distribution.
2.1.1 Particle separation and blending
The original spent and fresh cracking catalysts were separated into six different size frac-
tions respectively using the Haultain Infrasizer in the UBC Coal and Mineral Processing
13
Chapter 2. Experimental Particles and Apparatus 14
Centre. This device is based on air elutriation and is suited to split the particles into
multiple fractions below 200 mesh (i.e. 75 fim) at one time (Haultain, 1937). The in-
frasizer consists of six steel conical chambers arranged in series. Their dimensions are
given in Fig. 2.1. The separation involved two passes to minimize overlapping between
adjacent fractions. The respective air flowrates for the first and the second passes were
0.0021 and 0.0022 m /s. The original bulk FCC particles were charged into chamber
3
1 of the infrasizer during the first pass. The fraction collected in cone number 3 was
then passed into the sizer during the second run. Four size fractions of FCC labelled
"wide", "coarse", "narrow" and "fine" were collected for our experiments. The "wide"
distribution was the unseparated (i.e. original) FCC powder. The "coarse" and "fine"
fractions were the powders collected in cone numbers 2 and 4, respectively, from the first
pass. The "narrow" fraction was collected in cone number 3 from the second pass, i.e.
all particles in this fraction ended up in cone 3 in two successive passes.
A horizontal rotating drum mixer was used to blend particles of different fractions or
different activities. Bimodal batches were prepared by combining the "coarse" and "fine"
fractions. The catalyst activity was adjusted by changing the weight ratio of activated
to non-activated particles.
2.1.2 Catalyst activation and characterization
FCC particles were activated as the catalyst for ozone decomposition by being impreg-
nated with ferric nitrate solution at room temperature in a stirring mixer for one hour,
and then roasted in an electric oven at 450°C for one and a half hours. Both pho-
tographs and size distribution measurements before and after this activation procedure
(see Figs. 2.3 and 2.4, and Appendix E) indicate that there was negligible agglomeration
or change in the physical properties as a result of the impregnation and roasting. The
concentration of the ferric nitrate solution used was 10% by weight. The activity of the
Figure 2.1: Dimensions of the infrasizer

resulting catalysts was around 15 s - 1

at 20°C. Keeping the impregnation and roasting
conditions consistent for each batch of particles was essential to ensure uniform quality of
the catalysts prepared. The activities of the catalysts used in the experiments prepared
by blending activated and non-activated catalysts, were measured before and after each
run in a small fixed bed reactor with the same range of mean gas residence time as in
the fluidized bed.
Some properties were found to differ appreciably between the fresh and spent cata-
lysts. For example, the spent FCC particles can be used as a catalyst without activation
for ozone decomposition, with a kinetic rate constant of 0.3 to 0.5 s - 1
at room tempera-
ture, whereas the fresh FCC particles were found to be inactive for ozone decomposition.
However, once activated, the activity of the fresh FCC was much more stable than that
of the spent catalyst. The activity change of the activated N-FCC catalysts during a run
was usually less than 3%. The experimental results on the performance of fluidized bed
reactors were therefore collected mainly for the N-FCC particles.
2.1.3 Relevant properties of the particles
Relevant properties of the particles used in our experiments, including the particle den-
sity, particle diameter, particle shape and surface area, were measured.
Two kinds of density, the particle density, p , and the bulk density, />(,, were measured
p
for each particle fraction and particle blend. The particle density, p , was measured with
p
the "wet cake" method proposed by Abrahamsen and Geldart (1980). The bulk densities
were determined after the bulk volume of a loose packed bed of a pre-weighted sample
of particles was determined using a graduated cylinder. The resulting densities of spent
and fresh FCC are shown in Table 2.1.
The particle size distributions measured by an Electrone Size Analyser, based on the
same principle as that of the Coulter counter, appear in Fig. 2.2.
Table 2.1: Measured particle and bulk densities for fluid cracking catalyst
Powder Size distribution PP (kg/m ) p (kg/m )

3 b
3
S-FCC wide 1444 756
S-FCC coarse 1384 743
S-FCC narrow 1455 735
S-FCC fine 1556 782
S-FCC bimodal 1440 761
N-FCC wide i. 1582 860
N-FCC coarse 1591 895
N-FCC narrow 1586 823
N-FCC fine 1557 847
N-FCC bimodal 1573 868

20-
S-FCC S-FCC S-FCC
A1 wide B1 bimodal C1 narrow
15-
10
OB • Trrr>n
20-
N-FCC N-FCC N-FCC
wide bimodal narrow
rc 15
E
>»
•°10h
c
«
" 5r
J •
Q.
40 80 120 40 80 120 20 60 100

um
Figure 2.2: Size distributions of FCC particles

Chapter 2. Experimental Particles and Apparatus 19 .
Table 2.2: Surface area of FCC particles
Powder Size distribution Activated? Surface area (m /g)

2
N-FCC wide No 212
N-FCC narrow No 208
N-FCC bimodal No 210
N-FCC fine No 251
N-FCC coarse No 144
N-FCC wide Yes 172
N-FCC narrow Yes 164
N-FCC bimodal Yes 169
S-FCC wide No 88
The particle surface area was measured by a Micromeritics Surface Analyser based on
the BET isotherm theory, with sample tubes of type P / N 230-61001-00. The results are
given in Table 2.2. Some reduction in surface area resulted from the impregnation / roast-
ing procedure, presumably due to pore blockage by the ferric oxide.
The particle surface shape was measured by the sweep electron microscope (Hitachi
S-570). SEM pictures of different fractions and different blends of FCC particles are
presented in Figs. 2.3 to 2.8.
Figure 2.3: SEM photograph for N-FCC particles of wide size distribution
Figure 2.4: SEM photograph for N-FCC particles of wide size distribution after being
activated
Chapter 2. Experimental Particles and Apparatus
22
Figure 2.5: SEM photograph for N-FCC particles of bimodal size distribution
Figure 2.6: SEM photograph for N-FCC particles of narrow size distribution
Figure 2.7: SEM photograph for S-FCC particles of bimodal size distribution
Figure 2.8: SEM photograph for S-FCC particles of narrow size distribution
2.2 Experimental Apparatus
Most experiments on the performance and hydrodynamics of fluidized bed reactors were
carried out in a fluidized bed system designed and constructed for this project. Some
hydrodynamic research was also carried out in a two-dimensional fluidization column
which was modified for this investigation. Reliability and safety of the experimental
equipment, flexibility during operation and economy of manufacture and operation were
the main design considerations.
The ozone decomposition reaction,
0 —> 1.5 0 + 143 kJ/mole,

3 2
is a valuable laboratory reaction because it can be carried out under room temperature
and pressure conditions. In addition, it is nearly first order, and ozone can be readily
analysed using UV absorption. Several research workers (e.g. Frye et al, 1958; Van
Swaaij and Zuiderweg, 1972; Chavarie and Grace, 1975; Bauer and Werther, 1981) have
used this reaction to investigate the character of fluidized bed reactors, generally at low
gas velocities and for Group B or D particles.
The apparatus is shown schematically in Fig. 2.9. It includes five major parts: (1)
ozone generation and gas purification system; (2) fluidization column; (3) outlet gas
treatment and particle recirculation system; (4) instruments and data acquisition system;
(5) fixed bed apparatus.
2.2.1 G a s s u p p l y system
Ozone is generated by passing a stream of cylinder oxygen (breathing grade), after being
further dried by calcium sulphate, through an OREC electric discharge ozone generator.
The resulting 0 - 0 mixture, with ozone concentration 1 - 3% by volume, was then
2 3
OZONE
FILTER ROTAMETERS COLLECTOR FIXEO BED RECORDER
BLOWER FILTER ATOR
G E N E R
Figure 2.9: Schematic of experimental equipment

combined with the fluidizing gas before entering the bed. The fluidizing gas was blower
air, purified by passing through a column containing activated carbon and silica gel.
It is found from our pretest that the rate of ozone decomposition is very sensitive
to the humidity of fluidizing gas, especially at room temperature. The catalyst activity
decreases very quickly if the catalysts are fluidized by unpurified air from the blower. Frye
et al (1958) reported that the activity of ozone decomposition varied with the reciprocal
third order of water vapor concentration.
2.2.2 Reactors
The fluidization column (Fig. 2.10) is 2.6 m tall and 0.10 m in inside diameter. There is
an enlarged section above the column, 0.15 m in diameter and 0.5 m in height, to reduce
carryover, with a sloping section in between having a steep enough angle that particles
do not collect thereon. The fixed bed is 19 mm in diameter and 200 mm tall. The ozone
concentration at different positions along the fluidized bed is sampled isokinetically by
five probes. Since fines with diameters less than 20 pm can plug the filter of the sampling
probe, a nitrogen backflow system is set up to purge the filter before each sampling at
lower gas velocity and every 2 to 5 minutes for higher gas velocities (U > 0.5 m/s). The
purge frequency was also dependent on the fines content of the bed material.
Since ozone is a very powerful oxidant and reacts with most substances at 25° C (Razu-
movskii and Zaikov, 1984), the material of the reactor vessels and connecting tubes had
to be carefully chosen. The fluidized and fixed bed vessels are constructed of aluminum.
The sampling and connecting tube between the manometers and the reactor are pure
copper. It has been found (Horvath et al, 1985) that at room temperature and for ozone
concentrations up to 7% by volume, decomposition of ozone is not catalysed by pure
aluminum, lead, copper and tin, although most metals are strongly oxidized by ozone
and promote its decomposition.
TO I C Y C L O N E S
2500
H C
2200
\ / MEASUREMENT PORT
P PRESSURE
T TEMPERATURE
1680
H T C CONCENTRATION
fK)2
_ J 1 8 0 _
H C
_ _ 9 3 0 _
H COLLECTED
PARTICLES
_ 6 8 0 _ _
-p4- T,
_ 4 3 0
-p-H
_ 2 8 0
-p-l- H C
z=0 mm H T
P ^
1
H C
H T
GAS INLET
Figure 2.10: Detailed drawing of the fluidization column

2.2.3 Particle collection and return system
To keep the PSD within the experimental system as constant as possible under various
operating conditions, that is, to prevent fine particles from escaping from the fluidized
bed at higher gas velocities and static bed heights, a double-decked flapper system was
designed (Fig. 2.11). Two factors, the space limitation and the feeding location of par-
ticles collected by the cyclones, required that this structure be used. The overall height
of the experimental column was limited by the height of the ceiling. This limited the
vertical length of the diplegs of the two cyclones, so that they were not long enough to
provide the necessary pressure drop to prevent gas bypassing for some operating condi-
tions. This pressure drop should be higher than the pressure drop across the cyclones,
which, in turn, is related to the collection efficiency of the cyclones, especially for fine
particles.
To ensure that the particles cohected by the cyclones did not immediately become
entrained from the reactor again, these particles were returned to the reactor in the lower
dense region, not far above the distributor. This return location increased the difficult}'
in balancing the pressure drop of cyclones and made the flapper valve at the end of
the dipleg operate inefficiently. During operation, the typical hold-up of material in the
collection and return system (i.e. cyclones and standpipe) was of the order of 5 - 20 %
of the inventor}' of solids in the system.
Gas leaving the reactor was passed through two cyclones to separate entrained par-
ticles. The cyclones were designed according to standard cyclone dimensions (Sinnott,
1983), except for modification of the inlet ducts, where two tapered joints were used to
ensure high collection efficiency over a wide range of superficial gas velocity At lower
superficial gas velocities, the collection efficiency is mainly maintained by the second
cyclone, whose inlet area is only half of that of the first cyclone. At U = 0.5 m/s, for
Figure 2.11: Detailed drawing of the particle collection system.
example, the inlet gas velocity of the second cyclone is 14 m/s. The collecting efficiency
of this cyclone system is quite high, usually more than 99.99%, if the carryover rate is
estimated from the correlations of Wen and Chen (1982).
Separated particles returned to the bed through two diplegs equipped with a double-
decked flapper system (three flapper valves altogether) to prevent gas bypassing. At the
outlet of the second cyclone there is a bag filter and a fixed bed of diameter 300 mm and
height 600 mm filled with activated carbon to adsorb and react any remaining ozone. The
maximum mass of particles collected by the bag filter during each run was less than 0.1%
of the total initial mass in the bed so that the PSD and the mean diameter of particles
remained nearly constant within the system, even at high gas velocities. This confirmed
that the designed particle collection and return system was effective. In addition, there
was negligible attrition over the duration of the experimental tests. Typical PSD results
before and after operation are given in Appendix E.
2.2.4 Gas Distributors
Four sets of gas distributors with different free area ratio, orifice diameter and number
of orifices, as detailed in Table 2.3, were used in the experiments to test the influence of
distributor parameters. Each distributor was composed of two multiorifice plates with
different free areas as shown in Fig. 2.12. The free area ratio of the two plates was
between 3 to 15, with the main pressure drop produced by the upper plate. The angle
a<i was maintained less than 60% of the angle of repose to prevent leakage of particles
into the windbox.
Table 2.3: Distributors tested in the experiments
type" hole diameter, mm hole number free area, % pitch distance, mm
Tl 2.2 21 1.0 20
T2 3.8 21 3.0 20
T3 5.0 21 5.0 20
T4 5.0 4 . 1.0 45
B l - B3 8.8 21 15.5 20
B4 19.5 4 15.2 45
* Ti is the top plate and Bi is the bottom plate of distributor-i.

102 mm
02.2
| I I I <
/II / / A W / / I: / / /111
J I I \
06.8
Figure 2.12: Detailed drawing of gas distributor

2.2.5 Instrumentation
The main instruments used in our experiments include a U V photometer to analyze
ozone concentrations, pressure transducers to measure pressure fluctuations, a laser pho-
tometer to indicate the particle concentration inside voids and other void characteristics,
a computer data-logging system and other instruments to measure gas flow rate and
temperature.
Continuous analysis of ozone concentration along the column was carried out by means
of U V adsorption measurements, using a Dupont Model 400 Photometric Analyser at a
wavelength of 254 nm. The U V beam passed through a quartz cell with a 100 mm path
length and a diameter of 20 mm flushed by the gas being analysed. Calibration of the
analyser was carried out by the K I absorption method suggested by the O R E C operation
manual (see also Razumoskii and Zaikov, 1984) over a wide range of ozone concentration
(40 to 1400 ppm).
Four differential pressure transducers (PX162 - 01D) were used to measure pressure
fluctuations along the column. Each transducer had two input channels and produced
an output voltage proportional to the pressure difference, with a range of 0 to 7 kPa.
The lines which connected the taps and the transducers were checked experimentally to
ensure that no significant damping was introduced.
Differential pressure signals at different bed heights and different operating conditions
were logged to an I B M - X T minicomputer via an A / D convertor (Tecmar board) and
recorded simultaneously in analogue format on a U V chart recorder.
The particle concentration inside voids, the void size and void frequency in fluidized
beds were detected by means of light absorption in a "two-dimensional" plexiglass col-
umn. Detail description is provided in Chapter 5.

Chapter 3
Reactor Performance at Low Gas Velocities
Product yield is the most important indicator of the performance of commercial catalytic
fluidized bed reactors. For a simple reaction, such as ozone decomposition, the yield
is simply determined by the conversion. In our study, the effect of the PSD on the
performance of fluidized bed reactors is therefore mainly evaluated by the measurement
of ozone concentration for different PSD's under different operating conditions.
As discussed in Chapter 6, the fluidization regime and the regime transition from
the bubbling or slugging regime to the turbulent fluidization regime are dependent on
the PSD and on other operating parameters such as the static bed height (Fig. 6.4).
The term "lower gas velocities" used here refers to the range of gas velocities at which
all the fluidized beds discussed operated in the bubbling or slugging fluidization regime,
regardless of the PSD and static bed height. In this study, low gas velocity signifies U <
0.6 m/s. In this chapter, we mainly evaluate the PSD effect at low gas velocities, that
is, in the bubbling and/or slugging fluidization regimes.
3.1 Rate Constant Determination
The rate constant of ozone decomposition for each batch of catalysts was determined in
a fixed bed reactor (19 mm in diameter). If this integral microreactor is assumed to be
operated isothermally with transport gradients minimized and plug-flow attained, the
rate constant of ozone decomposition which can be taken as a first order reaction (Frye
36
Chapter 3. Reactor Performance at Low Gas Velocities 37
Table 3.1: Determination of the rate constant for 10 g samples of N-FCC catalysts of
wide size distribution
C in, ppm
A CAOUI, ppm Contacting time, s fc , s
P
{Vcat/V ) g
330 185 0.26 2.23
509 292 0.26 2.14
686 374 0.26 2.33
et.al, 1958), is readily obtained from the equation:
fc r= ^l„£d!IL (3.1)
where W is mass of catalyst, v the volumetric gas flow rate, p the bulk density, and
g p
CAin/CAout the ratio of inlet to outlet ozone concentration.

Justification for the assumptions underlying equation 3.1 are as follows:
a. Reaction order
The reaction order of ozone decomposition is dependent on the decomposing mech-

anism (Horvath et al, 1985). To determine the reaction order in our experiments, both
inlet gas flow rate and inlet ozone concentration were changed to test the linearity of
the rate constant calculated from Eqn. 3.1. As shown in Table 3.1, the rate constant for
different concentrations in our experimental range is nearly constant, implying that the
decomposition can be taken as a first order reaction.
b. A x i a l dispersion effects
The effect of axial dispersion can be minimized in packed-bed reactors by increasing

the ratio of reactor length, H, to equivalent particle diameter, d . The deviation from
p
plug flow is less than 5% (Mears, 1971a) if
H 20n CAIN
— > —— In —
dp Pe p UAout
where n is the reaction order, i.e. n = 1 in our experiments. Pe p is the axial Peclet
number, Pe = dpU/D , while D is an effective axial dispersion coefficient, which can be
p e e
determined as a function of Reynolds number and Schmidt number (Levenspiel, 1972).

Considering a wide range of GMUICA<m.t a n
d gas velocity in our experiments, we
should have H/d p > 140 (see Appendix 1) to satisfy the Mears criterion, eqn. 3.2. In our
kinetic measurements, H/d p was more than 600. Hence, the effect of axial dispersion is
negligible.
c. Interphase and interparticle gradients
Interphase transport gradients include the concentration and the temperature gradi-
ents around the particle. Since the size of catalysts used in our experiments is small, cri-
teria for minimizing the heat and mass transport gradients in the fixed bed (Rase, 1977)
were readily achieved. This was confirmed by varying the flow rate for four measure-
ments in which V /v was maintained constant. The interphase gradients were negligible
p g
since the resulting conversions at different flow rate were almost the same, as shown in
Table 3.2.
Mears (1971b) suggests that, for a first order reaction, the intraparticle transport
concentration gradient is negligible if
Kd 2
-^<3 (3.3)
This criterion is satisfied for most particles in catalytic fluidized bed reactors. For our
Table 3.2: Experimental test on the interphase transport gradient for N-FCC particles
of wide size distribution
v , ml/min.
g Wcot, g Vcat/Vg, S
1000 6.7 0.26 2.1
1500 10 0.26 2.2
2000 13.4 0.26 2.1
2800 18.7 0.26 2.3
experimental system, k d /D
T
2
< 0.1, i.e. the criterion is easily satisfied,
d . Isothermicity
Isothermicity in integral reactors is favoured by having a low-activity catalyst. For

catalysts used in this study, the activity was varied over a wide range, including some
relatively high values (around 10 s ) . Diluting of the catalysts utilized in the fixed
_1
bed with inert solids, however, was not necessary to achieve isothermal operation in our
work due to the low ozone concentration of the inlet gas (< 0.1% by volume). The
isothermicity of our fixed bed was confirmed experimentally by changing the position of
a thermocouple in the bed.
3.2 Stability of Catalyst Activity
It is important to keep both the PSD and the catalyst activity as stable as possible
during each run to ensure accuracy of the experimental results. Previous results (Frye
et al, 1958; van Swaaij and Zuiderweg, 1972) have indicated that the rate of ozone
decomposition can be influenced strongly by the composition of the incoming gas, due to
adsorption of gaseous species at ambient temperatures. Frye et al noted that the ozone
decomposition was reciprocal third order with respect to water vapor concentration over
the experimental range investigated. The influence of water vapor and other impurities
can be reduced by increasing the operating temperature (van Swaaij and Zuiderweg, 1972)
and by purifying the fluidization gas. The latter method was used in our experiments.
In this study, the fluidization gas was purified by passing through a fixed bed of
diameter 300 mm filled with activated carbon and silica gel, each with 250 mm height, to
improve the catalytic activity stability. The activated carbon was mainly used to remove
impurities such as oil droplets, while the silica gel was to remove water vapor.
Considerable work was done to keep the catalyst activity constant during each exper-
imental run. It was found that catalyst-decay was very significant for catalysts prepared
from S-FCC particles. For example, the activity could decrease from 3.5 s" to 2.1 s
1 - 1
after only 60 min, that is the deactivation rate was 40% per hour. The deactivation rate
for the N-FCC particles, however, was less than 3% per hour.
3.3 Definition of Efficiency for Fluidized B e d Reactors
For a first order irreversible simple reaction, the volume of catalyst, V , required to
CJ>
achieve a certain conversion in afixedbed reactor, with plug flow and negligible interphase
and intraparticle transport resistance can be obtained from eqn. 3.1, as
(3.4)
•r Âout
When the bed is fluidized at the same superficial gas velocity as in thefixedbed, the
mixing and contacting deviate from those of the plug flow reactor. As a consequence, the
actual volume of catalyst, V j, required to achieve the same conversion with the same
c
flow rate and catalyst activity is larger than V . The ratio of V j to V

c>p c CtP is dependent on
the performance of fluidized bed reactors. It is convenient to compare the performance
of different batches of catalysts with different PSD in terms of a "contact efficiency", E,
defined as
E = V JV
C CJ (3.5)
E can be regard as the utility efficiency of catalysts in fluidized bed reactors compared
with that in the plug flow reactor. The lower the E, the poorer the reactor performance,
i.e. the more the reactor deviates from that of a fixed bed with negligible axial dispersion.
3.4 Effect of P S D on C o n v e r s i o n a n d Reactor Efficiency
Experiments on the reactor efficiency were carried out over a wide range of operating
conditions to test not only the effect of PSD on the reactor efficiency but also the influence
of other factors. The kinetic rate constant, based on the volume of the particles, ranged
from 2 to 9 s , while the static bed height was varied from 0.15 to 1 m. The superficial
-1
gas velocity was varied from 0.1 to 1.8 m/s in order to cover a range from bubbling
through the slugging and turbulent regimes just into the fast fluidization bed regime.
Attempts were made in some runs to measure ozone concentration profiles by with-
drawing gas samples from three different ports along the column in order to evaluate the
PSD effect on the performance of fluidized beds in different reactor regions. For the sam-
pling probe immersed in the dense bed, however, a plug of particles would build up on
the tip of the probe very quickly just after beginning to withdraw the sampling gas. This
plug may cause further reaction to occur, distorting the measurement results. For ex-
ample, Figure 3.1 indicates that the apparent ozone concentration increases with height,
an impossible result. A similar problem was reported by Yates and Newton (1986). For
probes above the dense bed, the formation of the particle plug which appeared to be
much slower than in the dense bed, depended on the vertical location of the probe, the
particle size distribution and the superficial gas velocity. To reduce the interference of
the particle plug on the measurement, or the frequency of gas backfiow to remove the
plug formed, all conversion experimental results reported here were determined with the
the sampling probe at the highest level, z = 2.5 m, just before the outlet of the expansion
section. The concentration measured after the cyclone was found to be very nearly the
same as that measured at z = 2.5 m.
Experimental conversions for two catalyst inventories, 5.05 and 2.75 kg, correspond-
ing to static bed heights of 750 and 420 mm, are shown in Figs. 3.2 and 3.3. The
corresponding conversion for single phase plug flow (PF) reactors:
x = 1 — exp(—kj) (3-6)
and for perfect mixing (PM), i.e. CSTR reactors:
k'
j
(3-7)
are shown for reference purposes. The measured conversion never reaches the former, but
may or may not exceed the latter, depending on the PSD and the operating conditions.
Each experimental point is the average of five individual readings over a period of at
Chapter 3. Reactor Performance at Low Gas Velocities
43
i i i i i i
U m/s
o 0-10
0.4 • 0-30
• 050
CA bed surface
• 075
CA.in 1
•
! •
0.2 i • • -
•
• 1 • •
•
o
! °
o
i i i i i i
0 1 2 3
m
Figure 3.1: Variation of apparent reactant fraction unconverted with height above the
distributor for narrow size distribution with kinetic rate = 9.1 s" ; catalyst inventory 5
1
kg-
least 20 min. at stable operating conditions. Statistical considerations are discussed in

Appendix C. There it is shown that the gap between the error bars for each condition
would be smaller than the size of the symbols on the graphs. It is clear from Figs. 3.2
and 3.3 that the PSD does affect the performance of fluidized bed reactors. Among
the three blends used in our experiments the wide size distribution generally gives the
highest conversion, while the narrow distribution gives the lowest. The bimodal PSD
gives intermediate results, although conversions for the bimodal distribution tend to be
closer to those of the wide distribution than to the narrow.
It can be seen from Figs. 3.2 and 3.3 that the effect of PSD on the performance of
fluidized bed reactors is related to the gas velocities and the fluidization regimes. In
the bubbling and slugging regimes, the PSD effect is enhanced by the increase of gas
velocities. After transition to the turbulent fluidization, the PSD effect is reduced as the
gas velocity increases. Clearly, the reason for the PSD effect may differ from regime to
regime. It is better, therefore, to discuss the influence of PSD for each hydrodynamic
regime, separately.
Contact efficiencies are plotted against the unconverted gas fraction in Figs. 3.4 and
3.5 for the same conditions as in Figs. 3.2 and 3.3. The wide size distribution gives the
highest contact efficiency followed by bimodal distributions, with the narrow distribution
again giving the poorest results except for one point in Fig. 3.4 corresponding to a very
high gas velocity (U = 1.75 m/s) where the contact efficiency was higher for the narrow
size distribution than for the bimodal distribution with the fine fraction active. The
contact efficiency at low gas velocities is usually lower than for a single phase CSTR due
to the two-phase nature of fluidized bed reactors.
Although the mean diameters of the wide and bimodal blends are the same and
the bimodal blend contains more "fines" (about 21% by weight for N-FCC if fines are
defined as particles < 45 fim compared with about 16% by weight for the wide blend), the
- 1
— < 1 L_
0 5 10 15
krHmf [1 - €mfI/U
Figure 3.2: Effect of PSD on ozone conversion for N - F C C with = 4.1 to 4.3 s
catalyst inventory: 5.05 kg. P F and P M are corresponding results for single-phase plug
flow and perfect mixing respectively.
Figure 3.3: Effect of PSD on ozone conversion for N-FCC with k = 2.1 to 2.2
T
catalyst inventory: 2.75 kg. PF and P M as in Fig. 3.2.

1 I — i — i i ^
c
0.8
A +y
mmmm A ^
£o 0.6 A
—
o
£c 0.4
A O ,
• A wide
A
o O bimodal-fa
O • bimodal-ca
# narrow
0.2 • model*
<> model**
i i i
0-2 0.4 0.6 0.8

1-x
Figure 3.4: Effect of P S D on contact efficiency for N - F C C with k = 4.0 to 4.3 a " ; cata-
T
1
lyst inventory: 5.05 kg; "fa" and "ca" denote fines active and coarse active, respectively.
* with 4> — 0.05 e and S = 25;

b b b
* * with cf>b = 0.01 e and 5b = 5 .

b
Figure 3.5: Effect of PSD on contact efficiency for N - F C C with k = 2.1 to 2.2 a " ;
T
1
catalyst inventory: 2.75 kg; "fa" and "ca" denote fines active and coarse fraction active,
respectively.
with <p = 0.05 e and S = 25;

b b b
with (f> = 0.01 e and S = S.

b b b
performance of the wide blend is usually better. This suggests that the "fines content"
is not sufficient to characterize the PSD and that the performance of fluidized particles
depends not only on the quantity of "fines", but also on their nature and on the overall
i
size distribution. It is notable that Geldart and Buczek (1989) found that fines content
alone was insufficient to explain their hydrodynamic data.
3.5 Influence of W h e t h e r Fines or Coarse Particles are A c t i v e
To gain greater understanding of the importance of different size fractions and of the
reasons for the influence of the PSD, the bimodal distribution was separated into two
fractions, a finer fraction (mean size 40 pm) and a coarser fraction (mean size 90 ^m), as
shown by the shaded and unshaded portions of Fig. 2.2(b). Due to the imperfect nature
of the separator, there was a small overlap between the two fractions. Three catalysts
were then prepared:
(1) Both fractions (i.e. all particle sizes) activated with ferric nitrate, as described in
Chapter 2;
(2) Finer fraction activated, coarser fraction inactive, designated "fa";
(3) Coarser fraction activated, finer fraction inactive, designated "ca".
In each case, the overall activity, determined in the packed bed, was nearly the same,
4.0 to 4.2 5 _ 1
or 2.0 to 2.2 The mixing ratio of activated particles to inactive particles
was nearly the same, around 0.75. Conversions for these three catalysts appear in Fig. 3.6.
It is seen that for U < 1.1 m/s, the catalyst with only the finer fraction activated gives
higher conversion than that with only the coarser fraction active. Intermediate results
were achieved with all sizes activated. Some of these results also appear in Figs. 3.4 and
3.5 in terms of contact efficiency, again showing that it is more important that the fines
be active than that the coarser fraction be active. For 0.1 < U < 0.7 m/s, i.e. within
Figure 3.6: Effect of which fraction is active on conversion for bimodal catalyst; k — 4.0
T
to 4.2 5 _ 1
; catalyst inventory: 5.05 kg.
the bubbling and slugging regimes, the influence of which fraction is active is enhanced
as the overall catalyst activity and as the gas velocity are increased.
The different conversion in beds of bimodal PSD FCC with different fraction ("fines"
or "coarse") active can be explained by the disproportionate fraction of "fines" reporting
to the dilute (or dense) phase. The sensitivity of conversion to which fraction is active
suggests that more fines are present in the dilute phase, since catalyst particles in the
dilute phase can play a more important role, because they are exposed to gas of higher
ozone concentration than particles in the dense phase. No other explanation can account
for the results since the hydrodynamics, mixing and contacting are all independent of
which fraction is activated. In the bubbling and slugging fluidization regimes, gas is usu-
ally assumed to pass through the dilute phase in plug flow (e.g. Davidson and Harrison,
1963). In fluidized beds of Group A particles, if U ^> U b, as in this study, most of the
m
gas flows in the dilute phase, with exchange back and forth to the dense phase.
Based on the experimental results in this chapter and those in Chapter 5, it could
be concluded that the most important effect of PSD on the performance of fluidized bed
reactors operated at relatively low gas velocities is the influence on the void (bubble or
slug) character, including void type (bubble or slug), void size, particle concentration
in the voids and the relatively higher affinity of fine particles, compared with coarse
particles, for the dilute phase. All these effects appear to be enhanced as the gas velocity
increases. For example, with an increase of gas velocity, both the mean bubble size
and solids concentration inside voids in a fluidized bed of wide PSD appear to increase
gradually, while for beds of narrow PSD some bubbles grow large, becoming slugs, and
there is much less particle hold-up within the voids.
3.6 Model Predictions of P S D Effect at Low Gas Velocities
Many models have been proposed for bubbling fluidized bed reactors. Among the subset
of models which have been shown to give good predictions for the overall conversion in
pilot plant and industrial-scale equipment (Grace, 1986; Johnsson et al, 1987) the two-
phase bubbling bed model (Grace, 1984) is simpler than the others. For a first order
irreversible reaction, the outlet concentration derived from this model can be written as:
= E X P {
X MKH/U)
+ > (3
' 8)
where fa is the fraction of the bed volume occupied by bubble phase solids, fa is the frac-
tion occupied by solids assigned to the dense phase and X is a dimensionless interphase
mass transfer group, given by
X = (3.9)
The interphase mass transfer coefficient, k , can be calculated (Sit and Grace, 1981)
q
from:
^ = + ADe^v
{ ±)1/2 ( 3 i o )
6 rd eq
It is straightforward to show that the contact efficiency predicted from this model is given
where kj is a dimensionless rate constant,
k' } = ^ r n / a - W ) ( 3 . 1 2 )
If X —> co and fa —»0, i.e. in the limit where the two phaseflowfluidizedbed reactor
simplifies to a single phase plug flow reactor (or fixed bed reactor), E —»1 as expected.
From eqn. 3.11 it is clear that changes in the contact efficiency, E, may be caused by
changes in fa, fa or X with the different size distributions, or by some combination of two
or more of these factors. Clearly, the extent of the change also depends on kf. To express
the PSD effect on the performance of bubbling fluidized bed reactors, it is necessary to
account for the PSD effect on the bubble size, on the total particle concentration within
the dilute phase and on the particle size distribution inside voids. We do this by extending
the two-phase bubbling bed model described above.
An increase in conversion for the bimodal PSD with fines fraction active (see section
3.5 above) can be explained by postulating that a disproportionate fraction of fines reports
to the dilute phase. Let S be the overall ratio of the fines fraction by volume in the bed
to that of the coarser fraction, Sb be the corresponding ratio in the bubble or dilute phase
and Sd the ratio in the dense phase. A mole balance on gaseous component A for the
dilute or bubble phase, assuming plug flow and a first order reaction and ignoring axial
variations in hydrodynamic properties is
dC
U — h k a e (C
q b b Ab - C d)
A + Kb<t>bCAb — 0 (3.13)
dz
The corresponding mole balance for the dense phase, assuming perfect mixing in this
phase, is
k a e {C
g b b Ad -C ) Ab + Kd4>dC Ad = 0 (3.14)
Here fc^ is the reaction rate constant of particles in the dilute or bubble phase, while Kd
is the corresponding rate constant for the dense phase. When thefinesfraction is active
Kb (3.15)
and
krd (3.16)
while
where k is the kinetic rate constant of active particles. S<$ can be determined from
rp
c = g m / (l- £ m / )5(W5 )-g^S (l b b + 5)

d
H<f» [\ + S ) ( l + S ) + H<j> S {\ + S) - H {l - e )S(l + S )
d b b b mJ mf b
1
• '
If the coarse fraction is active, then
= ^ (3.19)
and
K, = ~ r (3.20)
J- — Jd
while
K -^ (3.21)
Equations 3.13 and 3.14 can be integrated with the boundarj' condition C A b = C in
A
at z = 0. For a first order irreversible reaction, the outlet concentration becomes:
c Aln
expi
X + MKH/U)^ 1 [
' " }
while the contact efficiency is given by
1 Xjfa^+fa^+kjMkrbkrd/k?).
E = f { L y a f e
- ' ^ J
} / T 6 v r e f a r w
(3.23)
If S b = Sd = 5 or fc b = r fcrd =fcr,that is, if both size fractions of catalyst particles are
uniformly active, E q n . 3.23 reduces to E q n . 3.11.
For the fine fraction active, the efficiency can be expressed as
while for the coarse fraction active.
771 _ f ~
T~ ft ' i-rJh \ ' t f i }\ ~r h} ~| / n <1C)
^ca — , 1
J J 1 J i J
v ,/ 1+ S / [J.ZO)
Since Ef was always higher than E

a ca in the bubbling regime for our experiments, com-
paring Eqn. 3.24 with Eqn. 3.25 it would appear that Sb/S > 1 and Sd/S < 1, so that
Sb > Sj. This suggests that more fines associate with the bubble phase, consistent with
the discussion in Section 3.5 and the theoretical prediction discussed in Section 5.1.3.
The fact that the ratio Ef /E a ca increases with superficial gas velocity and bed height
implies that both Sb and fa may increase with gas velocity and bed height. This is also
consistent with the experiments involving photometric measurements (Chapter 5), where
it was found that the influence of PSD on the particle concentration in voids is strongly
enhanced by an increase in gas velocity for U > 0.2 m/s.
In previous modelling of bubbling fluidized bed reactors (e.g. Orcutt et al, 1962; Kunii
and Levenspiel, 1969; Grace, 1984; Werther, 1984), the particle concentration in bubbles
was usually taken to be in the range 0 < fa/eb < 0.01. However, even if we take fa =
0.01 eb, the upper limit suggested by Kunii and Levenspiel, the predicted Efa/Eca from
eqns. 3.24 and 3.25 is still much less than the experimental result, even if Sb approaches
infinity, i.e. if all of the particles dispersed in the bubble phase were fines. This suggests
that fa should be larger than O.Olej, for Group A particles. As shown in Figs. 5.7 to
5.9, more than 70% of the voids contained more than 2.5% particles by volume in a two-
dimensional bed, even for FCC particles of narrow size distribution. It is notable that the
experiments which led Kunii and Levenspiel to suggest that fa/eb < 0.01 were conducted
with coarse or Group B particles at low gas velocities (Toei et al, 1966; Kobayashi, 1966;
Kunii et al, 1967). Hence fa/eb < 001 may be only suitable for Group B particles. For
the fluidized bed reactor (Section 3.5) of bimodal particles with different fraction active,
if fa were taken as 0.05eb, values of Sb from 2 to 5 would explain the observed results,
as shown in Figs. 3.4 and 3.5. Both Sb andfa,of course, are likely to be related to the
gas-solid characteristics, including the PSD, as well as superficial gas velocities and bed
depth.
3.7 Influence of Overall C a t a l y s t A c t i v i t y on the P S D Effect
The influence of overall activity of catalysts on the PSD effect in fluidized bed reactors
was also tested in our experiments. The catalyst rate constant was varied from 2 to 9 s _1
by changing the concentration of the ferric nitrate solution from 3% to 10% by weight.
It can be seen from Figs. 3.7, 3.8 and 4.1 to 4.3 that the PSD effect is generally enhanced
by the increase of the catalytic activity, especially at high gas velocities. It should be
noted, however, that the enhancement of PSD effect on the performance of fluidized bed
reactors was only significant at lower kinetic rate, that is, when k was increased from 2.2
T
s _1
to 4.5 if the bed was operated in the bubbling or slugging fluidization regime.
The influence of a further increase in k (from 4.5 to 9 s" ) on the PSD effect seems
r
1
negligible. These results suggest that mass transfer between the two phases (dilute and
dense) appears to be the controlling step in determining the reaction conversion in a
bubbling or slugging fluidized bed reactor if the kinetic rate is sufficiently high (e.g. k > T
5s _1
in our experiments).
For a simple irreversible reaction, if the resistance to chemical reaction and mass
transfer between two phases are of similar magnitude, the reaction conversion in a flu-
idized bed reactor would be dependent both on the kinetic rate constant and on the
interphase mass transfer. In this range of catalyst activity, the PSD effect, via its influ-
ence on the mass transfer rate, will be enhanced, by an increase in the kinetic rate, as
k was increased from 1 s" to 5 s
r
1 _1
in our experiments at relatively low gas velocities.
When the reaction is fast enough that conversion is mainly determined by the interphase
mass transfer rate, the influence of an increase in activity on the PSD effect becomes
insignificant.
Based on our experimental results at low gas velocities and the prediction from the
model expressed by eqns. 3.15 to 3.25, it is found that if k'j/X > 2 (for H f < 1 m and m
57
Figure 3.7: Influence of overall catalyst activity on the PSD effect; catalyst inventory- 5
kg-
0.2 0-4 0.6 0.8 1

1 -X
Figure 3.8: Influence of overall catalyst activity on the PSD effect; catalyst inventory:
5.05 kg.
U < 0.8 m / s ) , the P S D effect appears to be not influenced by the change of kinetic rate;
if k'j/X < 2, however, the P S D effect is enhanced by an increase i n k .r

Chapter 4
Reactor Performance at H i g h e r Gas Velocities
In most previous experimental studies of chemical reaction in fluidized beds, whether for
Group A or for Group B particles (e.g. Lewis et al, 1959; Massimilla and Johnstone,
1961; Hovmand and Davidson, 1968; Chavarie and Grace, 1975; Bauer and Werther,
1981; Yates and Newton, 1986), restricted ranges of gas velocity, seldom exceeding 0.3
m/s, have been investigated. Contacting efficiency tends to decrease as the gas velocity
increases over this range, since bubbles or slugs grow larger with increasing gas flow.
However, most successful applications of fluidized beds in the petrochemical and chemical
industries, such as fluid catalytic cracking, acrylonitrile synthesis and oxychlorination
of ethylene, utilize higher gas velocities, generally sufficiently high that clearly defined
bubbling or slugging has given way to a more turbulent hydrodynamic region where gas-
solid contacting can be improved (Yerushalmi and Avidan, 1985). Further shifts from
bubbling mode to high velocity modes are expected for many processes (Horio et al,
1989).
Partially due to technical difficulties which stand in the way of research on high gas
velocity fluidized beds, both fundamental research into the underling phenomena and
systematic compilation of design data lag well behind commercial applications which are
developing rapidly (Yerushalmi, 1986). Modelling work on high gas velocity fluidized
bed reactors is even less developed, especially for reactors operated in the turbulent
fluidization regime, despite their practical importance (Edwards and Avidan, 1986).
The objective of this chapter is to evaluate the effect of particle size distribution
60
Chapter 4. Reactor Performance at Higher Gas Velocities 61
on the performance of fluidized bed reactors operated at relatively high gas velocities
(U > 0.6 m/s), including the transition region from bubbling (or slugging) to turbulent
fluidization, the turbulent regime and a small part of the fast fluidization regime. Some
modelling work on fluidized bed reactors operated at relatively high gas velocities with
different PSD is also carried out.
4.1 P r e v i o u s W o r k on F l u i d i z a t i o n Performance at H i g h Gas Velocities
Generally speaking, five distinct hydrodynamic regimes can be observed in gas-solid sys-
tems involving Group A particles (Yerushalmi, 1986; Yang and Chitester, 1988) beyond
Umf. In order of increasing gas velocities, they are:
1. bubble-free expansion between U f and U b,
m m
2. bubbling/slugging,
3. turbulent fluidization,
4. fast fluidization, and
5. dilute-phase flow.
When the gas velocity is increased, bubbling beds of fine solids undergo a transition
to a turbulent regime. This transition is usually assumed to occur when the amplitude of
the pressure fluctuation reaches a maximum. The corresponding superficial gas velocity,
U , can be taken as a "pseudo" critical gas velocity to characterize the performance of
c
fluidized beds (Yerushalmi, 1986). Below U , the two-phase appearance of the fluidized
c
bed is clear, and voids grow larger with increasing gasflowat a rate depending on the gas-
solid properties and equipment characteristics. Around U , the heterogeneous, two-phase
c
character of the bed peaks. Above U , the fluidized bed gives way, usually gradually,
c
to a condition of increasing uniformity, culminating in the "turbulent" state. Complete

transition to the turbulent regime may occur at Uk, where large discrete bubbles or voids
are on the whole absent. Since Uk is much harder to measure, most people take U as c
the onset of the turbulent regime, which extends to the transport velocity UTR- The
approach to the transport velocity is accompanied by a sharp increase in the rate of
carryover. The transport velocity may be regarded as the boundary between turbulent
fluidization and fast fluidization if the solid flow rate is sufficiently high, or to a dilute
phase flow regime if the solid rate is low.
Although fluidized bed reactors have advantages in many chemical processes com-
pared with other reactors, their application is restricted by the limited mass transfer
between two phases. Two approaches have been used in commercial practice to improve
the fluidization performance. One is to restrict bubble growth and control gas flow by
using internal baffles, or by dividing the bed into several stages. Another is to suppress
the bubbling phenomena by adjusting the gas-solid properties, such as "fines" content,
and increasing gas velocities (Avidan and Edwards, 1986). The latter method is more
effective and interesting for most practical applications of catalytic fluidized bed reactors.
Numerous investigations have been carried out of gas-solid behaviour in fluidized beds
operated at high gas velocities. Most of the studies have been concerned with the solid
movement and distribution (e.g. Li and Kwauk, 1980; Hartge, 1986, Weinstein et al, 1986;
Brereton, 1987; Herb et al, 1989) and the gas flow pattern - dispersion and backmixing
(e.g. Cankurt and Yerushalmi, 1978; Yang et al, 1984a; Avidan and Edwards, 1986;
Krambeck et al, 1987; Brereton, 1987; Weinstein et al, 1989). For catalytic fluidized bed
reactors, the gas flow pattern is usually more important to the reaction results.
Cankurt and Yerushalmi (1978) conducted experiments in a 152 mm diameter column
on gas backmixing in fluid cracking catalysts fluidized at gas velocities spanning the
bubbling, turbulent and fast fluidization regimes, by measuring the tracer (methane)
concentration upstream of the injection point. The degree of gas backmixing was high
in the bubbling regime and diminished gradually over the turbulent regime. In the fast
fluidized bed, there was almost no backmixing. Yang et al (1984a) measured the tracer
(helium) concentration profiles radially and axially in a fast fluidized bed of 115 m m i . d .
and reported that axial mixing could be neglected i n a fast fluidized bed. They also
proposed an emprical correlation to express radial gas diffusion.
Gas tracer experiments were also carried out i n a pilot scale reactor (100 m m i.d.)
by A v i d a n and his colleagues (Avidan and Edwards, 1986; Krambeck et al, 1987) to
study the gas residense time i n bubbling and turbulent fluidization. Solids used i n their
experiments were typical F C C particles, while SFe was used as tracer. Pulse responses
were measured at gas velocity from 0.2 to 1.2 m / s . They reported that at low gas
velocities (U < 0.3 m/s), there existed a radial gradient i n tracer concentration. This
gradient decreased as the gas velocity was increased and appeared to be negligible at a
gas velocity of 0.6 m/s. They concluded that an axial dispersion model can be used to
predict conversion i n a turbulent fluidized bed reactor of Group A particles with enough
fines (> 15% if fines are defined as smaller than 40 ^ m ) , that the two-phase appearance of
the bed is diminished and overall homogeneity is approached. They wondered (Krambeck
et al, 1987) why so many models containing bubble properties were used extensively in
the fluidization literature. These models are seldom claimed to be more than "learning
models", while actual design models i n industry are simpler. They noted that i n the
turbulent fluidization regime, voids are i n a constant state of coalescence and break u p ,
so that models based on discrete voids have little relation to the true physical picture,
especially for most commercial applications of fine powders.
In brief, even though many investigations have been conducted on high velocity fluid
beds, most results are concerned with hydrodynamics and dispersion. Information on
the effect of catalyst size distribution on conversion results or contacting efficiency i n
a fluidized bed reactor operated at relatively high gas velocities, typical of industrial
applications of catalytic fluidized bed reactors, is still not available from the literature.
4.2 Dependence of P S D Effect on the Superficial Gas V e l o c i t y
To test the dependence of the PSD effect on the superficial gas velocity, our experiments
on ozone decomposition were carried out over a wide range of gas velocities, from 0.06
to 1.8 m/s, to include the bubbling, slugging, turbulent and fast fluidization regimes. It
was found from our experiments that the PSD effect on the performance of fluidized bed
reactors is related to the superficial gas velocity, as shown in Figs. 4.1 to 4.3. The PSD
effect appears to be most significant at gas velocities between about 0.3 and 1.4 m/s. The
conversion difference between the particles of wide and narrow size distribution appears to
increase with gas velocity for U < 1.2 m/s and to decrease slightly at higher gas velocities.
These results are consistent with those from hydrodynamic experiments (Chapters 5 and
6) and further suggest that different dependence of fluidized bed performance on gas
velocity is associated with different particle size distributions.
As discussed in Chapter 6, at lower gas velocities, that is at U < 0.6 m/s in our
experiments, for all three different particle size distributions tested in this study, the
lower part of the fluidized bed (z < 0.3 m) mainly operated in the bubbling regime, while
the upper part of the bed was mainly subject to the slugging regime. The characteristics
of voids (i.e. bubbles and slugs), such as void size and solid concentration, however, are
influenced by the PSD. The influence of the PSD appears to increase with increasing gas
velocity. Although for the three blend particles of different size distribution, the void size
increased with the increasing gas velocity for U < 0.6 m/s, the growth rate of voids was
higher for the narrow PSD than for the wide PSD. At U > 0.6 m/s, fluidized bed reactors
of wide size distribution undergo a transition from bubbling or slugging to turbulent
fluidization and the turbulent fluidization regime. In these regimes, bubbles or slugs
break up frequently to smaller voids of irregular shape containing a greater percentage
of the total solids. For particles of narrow size distribution, however, the transition to
l 1 r
I 1 I l i I L_
0 0.4 0.8 1.2

U m/s
Figure 4.1: Influence of gas velocity and PSD on reactor conversion; catalyst inventory:
5 kg, k = 2.4 s~ .
r
l
1.0r
A
o
O A
o O A
o O A
o
0-5
A wide
O narrow
O bimodal
0
0.5 1.0 1.5
U m/s
Figure 4.2: Influence of gas velocity and PSD on reactor conversion; catalyst inventory:
5 kg, kr % 4.5 s . _1
1.0i—•—•—•—•—-
A
o A
OO
A
O A
o O
O A
o O O
0-5
A wide
O narrow
O bimodal
0-5 1.0 1.5

U m/s
Figure 4 . 3 : Influence of gas velocity and P S D on reactor conversion; catalyst inventory:

5 kg, k % 9
T
turbulent fluidization not only occurred much later, but also was much less significant,
that is, there were still some slugs or large bubbles of lower particle concentration at
least in the upper part of the bed, even at high gas velocities (e.g. at U = 1.2 m/s).
The difference in particle concentration inside voids should also be noted. For particles
of wide size distribution, the solids content inside bubbles appears to increase with gas
velocity if U > 0.2 m/s (see section 5.1.3). For particles of narrow size distribution,
however, the solid concentration in voids usually decreases with gas velocity, until U >
0.6 m/s, in contrast to the wide PSD. More reaction, therefore, is likely to occur in the
voids for particles of wide size distribution than for those of narrow PSD. This difference
increases with gas velocity until the two-phase character disappears in fluidized beds of
both particle size distributions.
4.3 Influence of Fluidization Regime on Reactor Performance
The influence of the fluidization regime on the performance of fluidized bed reactors was
investigated by changing the gas velocity and the kinetic rate constant at the same time
to keep the dimensionless rate constant, k'j, constant. The gas velocity was changed from
0.1 to 1.8 m/s to include the bubbling, slugging, and turbulent regimes, and to just reach
the.fast fluidization regime. Further increases of the gas velocity were limited by the
column height, and the requirement that the pressure drop across the dipleg be balanced
against that across the bed and cyclones. The regimes tested in this study, therefore,
were the bubbling, slugging, turbulent and part of fast fluidization regimes. The kinetic
rate constant, k , was changed from 2 to 10 s~ .
r
l
4.3.1 Regime influence for particles of wide size distribution
The conversion in catalytic reactors operated in different fluidization regimes is shown

in Figs. 4.4 and 4.5 for particles of wide and narrow size distribution respectively. The
evaluation of regime effect on reactor performance was based on the comparison of con-
versions achieved in fluidized beds operated in different regimes but with the same k'j.
The corresponding results for single phase plug flow and perfect mixing are also shown
for reference.
It is seen from Fig. 4.4 that for particles of wide size distribution the conversion
obtained in the turbulent and fast fluidization regimes appears to be 10% to 20% more
than in the bubbhng and slugging fluidization regimes at the same k'j. Conversions
obtained in bubbhng and slugging fluidization were usually lower than in single phase
perfect mixing reactors, suggesting the predominant two phase character, i.e. high mass
transfer resistance between a dilute and dense phase. Conversions achieved at higher gas
velocities (U > 0.6 m/s), or in the turbulent and the fast fluidization regimes, however,
were usually higher than that in the corresponding perfectly mixed reactors, implying that
the mass transfer resistance between gas and solids decreases due to the regime changes,
i.e. due to break-up of the typical two-phase character. The effect of hydrodynamic
regime on the contacting efficiency of fluidized bed reactors is also related to the PSD.
For particles of wide size distribution, the contacting efficiency in the turbulent and fast
fluidization regimes is higher than that in the bubbhng and slugging regimes as shown
in Fig. 4.6. Operating at higher gas velocities, therefore, can improve the performance
of fluidized bed reactors of Group A particles of wide size distribution.
The effect of fluidization regime on the reactor performance should be considered in
modelling fluidized bed reactors. The higher efficiency attained at higher gas velocities
for particles of wide PSD, with values between those obtained for single phase perfect
Chapter 4. Reactor Performance at Higher Gas Velocities
70
krHmf (1 - € ) / U mf
Figure .4.4: Influence of hydrodynamic regime on ozone decomposition for catalysts of

Chapter 4. Reactor Performance at Higher Gas Velocities
krHmf (1 - €mf J/U
Figure 4.5: Influence of hydrodynamic regime on ozone decomposition for catalysts

narrow size distribution
Figure 4.6: Influence of hydrodynamic regime on contacting efficiency for particles of

mixing and plug flow reactors, is underestimated by the two-phase bubbling bed model.
Other models, such as a single phase axially dispersed model or a countercurrent flow
model, may be more suitable to evaluate the performance of fluidized bed reactors which
operate, at high ga.s velocities (Avidan and Edwards, 1986; Krambeck et al, 1987).
4.3.2 Regime influence for catalysts of narrow size distribution
For particles of narrow size distribution, it can be seen from Figs. 4.5 and 4.7 that the
difference among conversions obtained at the same dimensionless kinetic rate, kj, but
different hydrodynamic regime is relatively minor, in contrast to the result for the wide
size distribution. These phenomena can be explained by our hydrodynamic experimental
results discussed below in Chapter 6. A s shown in F i g . 6.1, the dependence of F on the

p
gas velocity for particles of narrow P S D differs from that of the wide size distribution
catalyst. For particles of wide size distribution, U . corresponding to the regime transition
c
from bubbhng to turbulent fluidization. is about 0.6 m/s, while the transition is completed
at around 1 m / s (or Uk). For particles of narrow size distribution, although the pressure
fluctuations appear to decrease gradually with gas velocity after reaching apeak at around
0.75 m / s , the significant levelling-off, which is seen for particles of wide P S D , has not
been found at the upper part of the- bed i n the entire range of gas velocity i n this study.
This suggests that even at high gas velocities, the two-phase character still appears to be
significant in fluidized beds of narrow P S D . Hence, considerable voids of large size and low
particle concentration exist in the bed. T h e proper particle size distribution, therefore,
is a necessarj' condition to obtain optimum performance of fluidized bed reactors at high
gas velocities.
The conversion achieved in fluidized bed reactors containing particles of narrow size
distribution was generally less in our experiments than i n a perfectly mixed single phase
reactor, whether the bed operated i n the bubbling, slugging or turbulent regime, except
Figure 4.7: Influence of hydrodynamic regime on contacting efficiency for particles of

narrow size distribution
for U > 1.4 m/s, as shown in Fig. 4.5. The two phase flow model, therefore, may be
suitable to predict the conversion for particles of narrow size distribution at both low
and high gas velocities.
4.4 P r e d i c t i o n of P S D Effect at H i g h Gas Velocities
The effect of the fluidization regime and the particle size distribution on the reactor
performance should be considered in modelling fluidized bed reactors. For the regime
where two-phase phenomena are predominant, that is at low gas velocities for the wide
PSD or over a wider range of gas velocity for the narrow PSD, the conversion achieved in a
fluidized bed can be successfully predicted by the Two-Phase Bubbling Bed Model (Grace,
1984, 1986) with two modifications to express the PSD effect. First, it is neccessary to
account for the effect of PSD on particle concentration in voids (as discussed in Chapter
3). Secondly, PSD has a significant influence on bubble size and/or slug length (as
discussed in Chapter 5). At higher gas velocities, however, the higher efficiency attained
in fluidized beds of wide PSD, with values between those obtained for single phase perfect
mixing and plug flow reactors, as shown in Fig. 4.4, will usually be underestimated by
the two-phase bubbling model.
The suppression of gas bubbles in fluidized bed reactors suggests the use of a homo-
geneous dispersion model. From the experience of scaling-up the MTG process, Avidan
and Edwards (1986) suggested that the axial dispersion model can be used to predict con-
version in a turbulent fluidized bed reactor under certain conditions. These conditions
- high gas velocities, substantial fines contents and sufficient bed height, were almost
achieved in our experiments for particles of wide PSD and at high gas velocities.
The axial dispersion model, a one-dimensional model, is often used to describe phe-
nomena that cause a distribution of residence times. In the case of an unpacked tubular
reactor the dispersion is attributable to molecular diffusion and turbulent transport or

eddy diffusion. In the fluidized bed, it is largely the result of local backmixing and by-
passing, local acceleration, deceleration and trapping, as fluid elements are retained in
a random manner for varying periods in the voids. These phenomena are conveniently
described using an effective axial diffusivity, D , a Fickian-type term. D is usually de-
e e
termined from residence time distributions, which, in turn, are often derived from tracer
tests.
A fluidized bed reactor can generally be divided into three sections along the column,
as shown in Fig. 4.8. For a steady state flow reactor with axial diffusion and first or-
der reaction, if reactions in the windbox and outlet section are negligible, the following
differental equations result from mole balances on reactant A for these three sections:
£,.2^r " ~ M l " V )CA = 0, 0< z < H (4.2)
d?C A dC A
where D i, D 2 and D ^ are gas axial dispersion coefficients in the windbox, reaction
e< e< e
section and outlet section respectively and is the average voidage in the reaction
section. The axial dispersion coefficient may differ between sections for reasons of gas
velocity, presence of particles and the like.
In practical fluidized bed reactors, in order to keep the pressure drop across the
distributor sufficiently high, the orifice gas velocity is maintained high and gas dispersion
across the distributor is negligible. The concentration in the windbox can then be taken
inlet
Figure 4.8: Axial dispersion model for fluidized bed reactors

a s
C'x.in- Equation 4.1, therefore, can be deleted. This model can then be solved with
the following four boundary conditions (see F i g . 4.8):
(a) CA |Z=O-= C
A | * = O + = C ,in
A (4.4)
(b) (UC
A - D ,2^f)
e !.=*-= (UC A - D , ^f)
e s \ Z = H + (4.5)
(c) CÛ -=Ûif H + (4.6)
(rf) |z=+oo= C ,out

A (4.7)
Condition (b) results from the conservation of reactants at the exit of the bed, taking
into account flow and diffusion. Conditions (a) and (c) arise from the intuitive argument
that at steady state the concentration should be continuous between sections at the bed
entrance and exit.
The general solutions for eqns. 4.2 and 4.3, containing four arbitrary constants, Ni,
are:
= N xp-^-{l
ie + a)z + N exp~^-{1
2 - a)z 0<z<H (4.8)
and
C a
- A / + N expV±-
3 A z>H (4.9)
After inserting boundary conditions, the complete solutions for equations 4.2 and 4.3
are:
CA ex
PJtfi[( a
~ l)expPe -f (1 — a)expA ) 2 — exp^Ka -f l)expPe + (1 + a)expAi]
C Aiin (1 + a)exp(Pe(l + a)) + (1 - a)ea:p(Pe(l - a)) - 2ea;pPe
(4.10)
0 < z < H
C
A,out = 2aea;pPe
CA,™ ( l + a)exp(Pe(l + a ) / 2 ) - ( l - a ) e x p ( P ( l - a ) / 2 ) '

e * ~ ^" J
where
Pe = £ 2 . (4.12)
/ 4Jfe'"
fl
= V 1
+ -jr- (- )
4 13
k'f =^ L z i L l (4.14)
, 4 = ( l + a)Pe
a (4.15)
i4 = ( l - o ) P e -
2 (4.16)
This model was tested against our experimental results. The axial dispersion coeffi-
cients used were from the tracer experiments of Guo (1987). Guo applied hydrogen as
tracer gas to measure the residence time distribution and gas axial dispersion coefficient
in a fluidized bed (DT = 0.07 m) of F C C particles. The tracer flowed through a solenoid
valve into the windbox of the column as an impulse input, and the sampling probe was
installed in the expansion section. The measurements were carried out at superficial gas
velocities from 0.1 to 1.2 m/s with H f from 0.15 to 0.9 m. These conditions are similar
m
to those of our systems. Using the data reported by Guo for H f = 0.6 m, the Peclet
m
numbers calculated for our experimental systems were from 6 to 13, similar to values
measured by Avidan and colleagues (Avidan and Edwards, 1986; Krambeck et al, 1987).
As shown in Fig. 4.4, the conversions achieved in the turbulent fluidization regime for
particles of wide size distribution are in good agreement with the prediction results from
this simple model, with an average deviation of only 5.1%.
Conversions predicted from the axial dispersion model with boundary conditions of
open inlet and closed outlet, as used by many previous investigators (e.g. Danckwerts,
1953; Avidan and Edwards, 1986), are also shown in Fig. 4.4. In this case, the average
deviation between the experimental and the predicted results was larger, 8.4%.
Chapter 5
Effect of Particle Size D i s t r i b u t i o n on T w o - P h a s e B e h a v i o u r
The performance of fluidized bed reactors is strongly related to the hydrodynamics of

fluidization. In this study, the influence of particle size distribution on the hydrodynamics
of fluidization was investigated in two different fluidized bed columns. In this chapter,
the influence of PSD on the void phenomena - particle concentration in voids, void size
and void frequency, mainly determined in the two dimensional column - is examined. The
effect of PSD on pressure fluctuations, regime transitions and bed expansion, in the 102
mm diameter cylindrical column in which the ozone decomposition reaction was studied
(see Chapters 3 and 4), is considered in Chapter 6. The distribution of different particle
size fractions between the dense and dilute phases is discussed in Chapter 3.
The performance of bubbhng and slugging fluidized beds is dominated by the dilute
phase behaviour, i.e. by the characteristics of gas voids. At low gas velocities, that is, in
the regime of bubbhng fluidization, gas voids are conveniently termed "bubbles", while-at
higher gas velocity, "bubbles" become "slugs" or appear in a more irregular or transitory
manner in the turbulent fluidization regime (Rowe and MacGillivray, 1980).
Considerable work has been done on the behaviour of bubbles, and a number of
correlations have been proposed to predict mean bubble size. Most of them, however, are
restricted to coarse particles, i.e. group B or D powders according to the Geldart (1973)
classification (Horio and Nonaka, 1987). The dependency of dilute phase character on
the PSD is poorly understood, with direct measurements almost totally missing from the
literature.
81
Chapter 5. Effect of Particle Size Distribution on Two-Phase Behaviour 82
5.1 Particle Concentration in Voids
An assumption made by some investigators in modelling fluidized bed reactors is that

the bubble or dilute phase is completely free of particles (e.g. May, 1959; Van Deemter,
1961; Orcutt et al, 1962; Werther, 1980). Although this assumption appears suitable for
systems of coarse particles and for slow reactions, solids dispersed in the bubbles need to
be considered for a fast reaction system, even when the volumetric particle concentration
is small (Grace, 1982). Previous experimental work carried out with coarse particles
(Toei et al, 1966; Kobayashi, 1966; Kunii et al, 1967), has demonstrated that particles
are present in bubbles. In most catalytic fluidized bed reactors, such as in catalytic
cracking of petroleum, ammoxidation of propylene and chloroxidation of ethylene, not
only is the reaction fast, but also the catalyst is composed of fine (Group A) particles.
The particles inside the void phase, therefore, may play an important role in determining
the gas-solid contacting efficiency and reactor performance.
Hiraki et al (1965) investigated the amount of particles dispersed in bubbles in a

two-dimensional freely bubbling fluidized bed by means of photograpy. Microspherical
catalysts (p — 1.54 g/cm , d = 150 nm) were tested at a gas velocity just above
p
3
p
U f.
m Particles dispersed in a bubble were illuminated using the Tyndall phenomenon
and recorded on film. The volume fraction of particles dispersed in a rising bubble was
calculated by counting the number of particles on the photograph, multiplying by the
average volume of each particle and dividing by the bubble volume. The volume fraction
occupied by particles was found to be 0.4 to 0.6%.
Toei et al (1966) measured the particle concentration in bubbles by tomography in
a two dimensional bed (cross-section 24 x 250 mm). Particles used were glass beads
(150 fim <d <
p 180 fim and 710 fim < d <p 1000 pm) and sand (125 pm < d <
p 180 ^m).
The gas velocity was varied from 0.06 to 0.2 m/s. The measured volumetric concentration
of particles in bubbles ranged from 0.18% to 1.2%.

Kobayashi et al (1966) determined the particle concentrations in bubbles by using a
detection probe consisting of a small electric bulb and a phototransistor. The particles
tested were commercial silica gel particles with size range 180 to 250 pm. The gas
velocity ranged from 0.05 to 0.22 m/s. The average particle concentration in bubbles was
determined to be 0.8 to 4% by volume, while the ratio of particles contained in bubbles
to the total solids in the fluidized bed was 0.4 to 1.2%, increasing with increasing gas
velocity.
All the above experiments were carried out with coarse or Group B particles at low
gas velocities. These results, however, have often been assumed to apply tofluidizedbed
of Group A particles, often containing large quantities of fines, and operated at high gas
velocities.
In this study, the particle concentration in voids in fluidized bed of Group A particles
was estimated using a laser photometric technique, and the effect of the PSD on the
particle concentration was evaluated for a wide range of gas velocities.
5.1.1 T h e o r y u n d e r l y i n g the measurements
Photometric techniques, depending upon fight absorbance, are applicable in many indus-
trial process units since the experimental procedure is simple and non-interfering. The
extinction of a colhmated beam passing through suspended particles can be used to deter-
mine the line-of-sight average particle concentration or spray concentration. Fenton and
Stukel (1976) constructed a miniaturized optical probe system utilizing the attenuation of
a colhmated fight beam passing through a particulate suspension to measure the particle
concentration. Brown et al (1988) measured the size distribution and relative concen-
tration of solid particles in liquids by small-angle laser fight diffraction. By combining
the Beer-Lambert Law with a theorem of Cauchy, the absolute concentration of particles
could also be evaluated. Schmidt et al (1989) developed a laser photometric probe to

measure the local concentration in liquid dispersions. The technique consisted of a laser
light sourse, a fiber-optic probe and an electronic device with a photodetector. Their
measurements for different concentrations of oil-blue droplets in toluene demonstrated
that the Lambert-Beer law holds over the range investigated.
Scattering and absorption characteristics of many particles in fluidized bed can be
obtained from the single particle characteristics based on a size parameter a (a = irdpi/X)
and the particle volume fraction (Tien, 1985). The measurement system can be consid-
ered to be in the independent scattering regime if a > 20 (Hodkinson, 1966), and the
particle volume fraction < 0.15 (Tien, 1988).
For the independent scattering regime it can be assumed that each particle in the
assembly scatters and absorbs radiation unaffected by other particles. Thus the extinction
and scattering by the system is expressed by simple algebraic addition of the energy
extinquished and scattered by each primary particle. The cross-section for the system of
N particles is the sum of the cross-section of each particle, and the individual particles
are assumed to scatter and absorb radiation independently of the others.
With the helium-neon laser (A = 0.632 /xm) and particles used in this study (dpi >
15 fim), a is always greater than 60. For the dilute phase investigated, therefore, the
condition for independent scattering is easily met.
For a measurement system in the independent scattering regime, the ratio of the light
intensity measured before and after the particles are placed in the laser beam is related to
the total projected cross-sectional area of the particles, expressed by the Lambert-Beer
Law, that is
I m
\n- = -J2N A Q L
i i i (5.1)
o
J
i=l
where / and I are the light intensities with and without the particles respectively, Ni
0
is the number of particles of size interval i per unit volume, Ai is the cross-sectional
area of the same particles, Cji is the extinction efficiency of these particles and L is the
optical path length. Qi is related to the diameter of the particles, to the absorptivity and
emmissivity of the particles, to the refractive index of the medium, and to the wavelength
of the radiation.
For a spherical particle,
YL =
(5.2)
A p 3
so that
N i A i =
2k~ (5
- 3)
where Vi is the volume of particles in size range i , per unit volume.

Combining eqns. (5.1) and (5.3), we obtain
T W-
m
(5.4)
Since Vi = V Ptt Xi, and V Ptt = 1 - e, where V Pit is the total volume of all particles in
the beam per unit volume,
In j- = - ( l ' e ) L ± ^ Q i (5.5)
li a > 50, the effect of particle diameter on the extinction efficiency is negligible
(Brown et al, 1988), so that Qi can be taken as Q. The particle concentration in the
voids, expressed as 1 - e, therefore, can be estimated from
2d , ln( J Q / I )
p v
1 _ e
T 3LCJ ( 5
' 6 )
where d
PtV = l/^2{xi/dpi) ,v is the mean particle diameter of the particles in voids.
The theory developed thus far implies that the particle concentration in voids in-
creases logarithmically with the ratio of fo to /. For a measured value of (fo/I), the
particle concentration is directly proportional to the mean diameter of particles inside
the void.
5.1.2 Laser photometer and its calibration
The experiment was carried out in the two dimensional column (12 x 560 mm in cross-
section and 2.1 m in height). There is a tapered expansion section (Fig. 5.1) at the top
of the column to reduce carryover. Above this section, there is a filter bag to collect fines
escaping from the column.
The measurement equipment consisted of a laser light source, a Metrologic photometer
and a UV chart recorder, as shown in Fig. 5.2. The laser was a 5 mW He-Ne tube
(Spectra Physics Stablite) emitting light at a wavelength 632 nm. Power was supplied
by a Model 255 Exciter. The Metrologic Photometer (Model 45-230), designed for laser
measurement, includes a photo-detector and a circuit amplifier. The measurement range
of the photometer was from 0.003 mW to 10 mW, adjusted hy a range selector. The
output of the photometer (0-1 vDC) was recorded continuously on the UV chart recorder.
Both N-FCC and S-FCC particles, each with three different PSD but with the same
dp for each type, were tested in the two-dimensional fluidized bed at gas velocities from
0.06 to 0.8 m/s. The static bed height was 0.9 m, and the measurement location was 0.85
m above the distributor along the center line of the front face of the column. Compressed
air was used as the fluidizing gas.
Most of the experiments were carried out in darkness at night to reduce the influence
of extraneous light on the measurements. Pretesting showed that interference could be
prevented if both walls of the bed, except for apertures for the beam, were covered by
black plastic film (900 x 500 mm). No signal was detected by the photometer if the laser
Figure 5.1: Dimensions of the two-dimensional fluidization column

Chapter 5. Effect of Particle Size Distribution on Two-Phase Beha\iour 88
1. EXCITER
2- LASER
3. BLACK FILM
4- FILTER
5- 2 - D COLUMN
6- DETECTOR
7- PHOTOMETER
8- UV RECORDER
Figure 5.2: Measurement apparatus used to determine particle concentration in voids

was shut off, even in the most sensitive range (0.003 mW).
The operation procedure was as follows. The alignment of the laser tube and the
detector was checked carefully before each measurement. After the laser was turned on
for a half hour, particles were added to the column and fluidized by air at a predetermined
gas velocity for 15 minutes. After the fluidized gas was shut off, so that the laser beam was
blocked by a fixed bed, the photometer output voltage was nulled. Then approximately
1000 ml of particles were withdrawn from the bottom, so that the static bed height fell
below the height of the laser beam. The photometer voltage output was thus recorded
as J (at zero concentration). This procedure was repeated about 3 times to make sure
0
that I was nearly constant. After I was determined, the withdrawn particles and any
0 0
collected by the filter were added to the column to give a static bed height of around
0.9 m. The particle concentration at different gas velocities was then measured, and the
voltage output was recorded on the UV recorder continuously.
To determine the absolute particle concentration, the laser photometer was calibrated
in a small two-dimensional rectangular chamber (12 x 150 x 250 mm principal dimen-
sions) for each kind of particles measured. To ensure that the particles were dispersed
uniformly, water was used as the suspending medium.. The experimental results pre-
sented by Fenton et al (1976) and Brown et al (1988) establish that the change in the
extinction coefficient due to using air instead by water as the medium is insignificant.
Glass beads, with mean particle size 11 pm and 22 pm respectively, were used by them,
and almost the same extinction coefficient was obtained, even though Fenton et al used
air as the suspending medium, while Brown et al used water.
The range of particle concentration calibrated was from 0.5 to 8% by weight (around
0.4 to 6% by volume). A weighed quality of particles was suspended in 400 g water, and
maintained in suspension by vigorous hand stirring. For each sample measured, both 7 , 0
recorded after all particles had settled, and f, recorded while particles were uniformly
dispersed, were read 5 times to obtain an average value. The calibration results, expressed
by five lines, are shown i n F i g . 5.3. T h e linearity between 1 - e and l n ( / / J ) is satisfactory.

0
Comparing lines 1 and 3, it is seen that the extinction coefficient depends on the colour
of the particles. For S - F C C particles with dark-grey colour, Q was about 2, while for the
N - F C C particles (white colour), Q was around 1. T h e calibration results seem consistent,
with the theory expressed by E q n . 5.6. The difference between lines 2, 3 and 5 is caused
by the change of mean particle size. For particles with the same mean size but different
size distribution, the dependency of (1 - e) on the ratio of light intensity, Io/I, (hues 3
and 4) is almost the same, as predicted by E q n . 5.6.
5.1.3 P a r t i c l e a n d fines c o n c e n t r a t i o n i n v o i d s
To keep the P S D inside the bed as contant as possible during each measurement in the
presence of particle carryover, the maximum gas velocity used in the two-dimensional
column was 0.8 m / s . For particles tested i n this study, the m a x i m u m measurable solid
concentration by this photometric technique was around 2% by volume for S - F C C and
5.2% for N - F C C . as shown i n F i g . 5.3, since the m a x i m u m I /I 0 which can be determined
accurately is around 10 (with reading error less than 5%). T h e measurement range for
4
N - F C C particles was wider than for S - F C C particles. Most experimental results discussed
here, therefore, are for the N - F C C particles. The results for the S - F C C particles are given
in A p p e n d i x B .
Results for N - F C C particles are shown in Figs. 5.4 to 5.6. The intensity measured by
the photometer when voids allow light to pass, is seen to be affected by both the particle
size distribution and the superficial gas velocity. A t low gas velocity (U < 0.2 m / s ) , the
P S D influence was small. A t higher gas velocities, however, the influence o f . P S D became
significant.
The effect of P S D on the solids hold-up i n voids i n fluidized beds is shown i n Figs. 5.7
Chapter 5. Effect of Particle Size Distribution on Two-Phase Behaviour
10000
100 L 1 1 1 1 1 1
1 2 3 4 5 6
SOLIDS CONCENTRATION (vol%l
Figure 5.3: Particle concentration calibration curve
1 s
800
U =0-1 m/s
=24i
V, = 2400.—
U-
>( / I — i
0- 2 m/s
0-4 m/s
y UL
0- 6 m/s
IL J ;
0-8 m/s
Figure 5.4: Photometric measurement for N-FCC particles with wide PSD
1 s
U = 0 . 1 m/s
—W\=800 —
0-2 m/s
-lo/|=2400-
0-4 m/s
0-6 m/s
0-8 ny s
Figure 5.5: Photometric measurement for N-FCC particles with narrow PSD
Figure 5.6: Photometric measurement for N-FCC particles with bimodal PSD
SOLIDS CONCENTRATION (vol%l

3 4
• NARROW 0-1 m/s
O " 0-2 »
%8o\ A WIDE 0-1 "
o A » 0-2 »
o
CC
UJ
I 601
I
5 40[
3
O
20
J i i t i i -1 1 I I I I u
500 1000 2000 4000 8000

lo/l
Figure 5.7: Influence of PSD on solids hold-up in voids at lower gas velocities
SOLIDS CONCENTRATION I vol %l

3 4 5
• NARROW 0-4 m/s

o A WIDE
g 80[ • BIMODAL
o
cc
UJ
O 60[
5 401
Z3
O
20
.A'
J i i i i i i J I I I I i_
500 1000 2000 4000 8000

lo/|
Figure 5.8: Influence of PSD on solids hold-up in voids at a gas velocity of 0.4
SOLIDS CONCENTATION Ivol%]

2 3 4 5
T , 1 : r
• NARROW 0.6 m/s
500 1000 2000 4000 ' 8000

lo/l
Figure 5.9: Influence of PSD on solids hold-up in voids at higher gas velocities
to 5.9 where the abscissa gives both I /I, and the solids concentration, while the ordinate
0
is the ratio of the cumulative number of voids with solids concentration lower than the
corresponding (I /I)k
Q to the total number of voids (with solids concentration lower than
5%) for a measurement interval of 60 s. It should be noted that the solids concentration
given on the top axis of Figs. 5.7 - 5.9 is for the overall PSD's. This scale should be
accurate for the narrow size distribution, but there will be a small error for the wide and
i
bimodal distributions where we have shown that there is a greater concentration of fines
in the voids than in the overall bed.
For gas velocities less than 0.2 m/s, the dependency of cumulative number percent,
on the particle and fines concentration in bubbles was similar for the wide and narrow
particle size distributions. Even at a gas velocity of 0.2 m/s, the ratio of the cumulative
number of voids with solids concentration less than 3.3% by volume to the total number
of voids for the narrow size distribution is higher than for the wide PSD. With an increase
of gas velocity, however, there was a significant difference between the two PSD's. This
difference increased with increasing gas velocity for U < 0.7 m/s in our experiments. The
number of voids with lower particle and fines concentrations was much greater for the
narrow particle size distribution than that for the wide PSD. For the wide particle size
distribution, J^Ni/Nx only increased with gas velocity if U < 0.2 m/s, and decreased
dramatically with U if U > 0.2 m/s, except for (I /I)
0 k = 5000, i.e. solids concentration
less than 4.5% by volume. For the narrow particle size distribution, however, this change
only occurred for U > 0.6 m/s. For U > 0.2 m/s, J2 {Io/I)i<1000
Ni/NT decreased very
significantly with gas velocity for the wide PSD, which means the concentration of parti-
cles or "fines" in the voids increased due to enhancement of relative movement between
the two phases in fluidized bed of particles having a wide size distribution. For the bed
with a narrow PSD the change of ^ ( /
»/ )i<
/ 1 0 0 0
Ni/Nj with gas velocity appears to be less
significant in this range.
Comparing the measurement results for particles of bimodal size distribution with
those for the wide size distribution, as shown in Figs. 5.8 and 5.9, we found that the
difference of solids concentration in voids is insignificant, although at higher gas velocities
solids concentration in voids for a bimodal PSD appears to be lower than for a wide PSD.
The effect of the PSD on the frequency of voids in fluidized bed is dependent on
the superficial gas velocity, as shown in Fig. 5.10. For U < 0.4 m/s, the PSD effect on
the frequency of voids seems insignificant, consistent with the experimental results of Ip
(1988). For higher gas velocities, however, the frequency of the detectable voids in the
bed of wide PSD was much smaller than that of the narrow size distribution, suggesting
a progressive change towards a structure of greater homogeneity in the bed of wide PSD.
When an isolated bubble rises through a fluidized bed there is an upwardsflow(often
called a throughflow) from bottom to top of the void relative to the bubble, because the
void offers a path of low resistance to gas seeking to rise from the bottom of the bed to the
bed surface under the overall pressure gradient. The average magnitude of the relative (or
throughflow) velocity can be taken as 3 Umf for three-dimensional beds (Davidson and
Harrison , 1963). This value only applies to isolated spherical bubbles (Clift et al, 1972)
and for Group B powders, since the dense phase is assumed to correspond to the bed
under minimum fluidization (rather than minimum bubbling) conditions. Considering
the difference between Group A and Group B powders, we assume a typical throughflow
velocity of 3 $ U f in a freely fluidized system, where $ = Umb/Umf is unity for Group
m
B powders and > 1 for Group A. The coefficient 3 allows for some bubble interaction.
Consider a typical void rising in a fluidized bed. Imagine that into this void a small
group of particles having a size distribution identical with the overall bed size distribution
is introduced (either from above or below) at a particular time. If these particles are
separated widely enough that hindered settling effects are of second order importance,
the magnitude of the velocity of a particle relative to the void will tend towards a value
100
T 1 1 1 1 1 1 r
L 1 1 1 1 I I I l _
0 0.2 0.4 0.6 0.8

GAS VELOCITY m/s
Figure 5.10: Influence of PSD on the frequency of detectable voids (i.e. with the solids
concentration less than 5% by volume) at different gas velocities and at z = 0.85 m
of order
(U i)i
re =\ i W m J - U
Ti | (5.7)
where Un is the terminal settling velocity of a particle in the i th

discrete size interval.
For the Group A powders which are characteristic of fluid bed catalytic processes, the
minimum fluidization velocity is given (Grace, 1982) by
U mf = 7.5 x 10- gApdl/p

A
(5.8)
For fines in these Group A particles, furthermore, the terminal settling velocities are
given by Stokes law, i.e.
U Ti = gApd^/ilSp) (5.9)
Substitution of Eqns. 5.8 and 5.9 into Eqn. 5.7 leads to
(Urel)i = ^ I (0.0405$3p - «£) I (5.10)
The volumetric concentration of each particle size fraction inside the voids will be pro-
portional to the time spent by particles of that size inside the bubbles, and this, to a
first approximation, will be proportional to (Uî)^ . 1
Assuming that the overall mass of
particles in the voids is much less than that of the total mass of particles in the system,
the mass fraction of particles of size i in the void phase can then be estimated from
(fi), = J'f^fK (5-11)

L, Ji,o/{Urel)i
It is clear from the above analysis that particles with d^ considerably smaller than d , p
and especially particles of order $°" d /5 in diameter, where the term within the absolute
5
p
value signs in Eqn. 5.10 approaches 0, will spend much longer in the bubble phase and
therefore will have considerably enhanced concentration within the bubbles. Note that
particles finer than this will tend to be carried upwards through the bubbles, as in the
experiments reported by Garcia et al (1973), while larger particles will fall downwards if
dislodged from the roof.
It should be noted, therefore, from Eqn. 5.6, that the dependency of particle concen-
tration, 1 - e, on the ratio of light intensity, Io/I, is affected by the mean particle size
in the voids. The light intensity after passing through the suspension, /, may therefore
be affected by PSD in two ways. One is via the change of particle concentration, that
is, more particles are contained in voids in fluidized beds of wide PSD than for a narrow
PSD. Secondly, the mean particle size in voids may differ from that of bed material since
more fines are contained in voids in fluidized beds of wide size distribution, as predicted
from Eqn. 5.11. Hence d pv is smaller than d . For FCC particles of wide PSD, $ = 3, as
p
determined by Ip (1988). The corresponding size distributions of the particles inside the
voids in a bed composed of such a size distribution, calculated from Eqns. 5.10 and 5.11,
are shown in Fig. 5.11, compared with the size distribution of the overall bed material,
and dp iV 0.65 d.p For a fluidized bed of narrow size distribution, however, the mean
particle size in voids is bound to be nearly the same as that of the overall bed material.
The decrease of f therefore may also be caused by d PiV being less than d.
p It is found from
our experiments on the PSD effect on the conversion of ozone decomposition achieved in
fluidized bed reactors (Chapter 3), that both of these factors appear to be important.
It would be best therefore to be able to distinguish the two factors, i.e. to determine
both (1 - e) and dp . tV Unfortunately, it is difficult to solve this problem, i.e. to measure
the particle concentration and the size distribution simultaneously. In some related
work reported recently (Brown et al, 1988; Hayashi, 1988), a photometric technique,
including a complex semiannular silicone photodetector array light diffraction system,
could only be used to determine the volumetric concentration and size distribution of
particles suspended in liquids.
60 80 100 120
dp,i ium
Figure 5.11: Particle size distribution for an industrial fluid cracking catalyst, / i o , and
corresponding calculated size distribution inside voids, /,_„.
In another study, a gamma-radiation absorption densitometer was used to try to mea-

sure the concentration and mean size of particles in voids simultaneously. For radiation
attenuation (Weimer et al, 1985)
1-«=*M> (5.12)
pSu
where p is the attenuation coefficient. It is clear from Eqn. 5.12 that the measured I is
not influenced by d .
PtV An attempt was made in the present work to use a gamma-ray
densitometer, purchased for work in the UBC circulating fluidized bed combustion pilot
plant, to determine the volumetric concentration of particles dispersed in voids in the
two-dimensional column. This attempt, however, failed due to the fact that the difference
between the absorbance in the empty column and with particles present was too small to
be measured, since the gauge was designed to be used in a system having a much higher
total absorbance.
Even if the particle concentration and particle size effects could not be separated, the
following conclusions can be drawn from the experimental photometric measurements:
1. The particle concentration in most bubbles or voids in a fluidized bed of Group
A particles was higher than what has been determined previously for Group B particles,
i.e. 1 - e was higher than the value of 1% reported for Group B particles (Kunii and
Levenspiel, 1969). More than 70% of the voids contained more than 2.5% particles by
volume, even for FCC particles of narrow size distribution.
2. The value of I /I depends on the PSD. The PSD affects both the overall particle
0
concentration in voids and/or the relative fines content in the dilute phase. The exper-
imental results are also consistent with the finding from Chapter 3 that more particles
and a greater proportion of fines are carried by gas in voids in beds of wide PSD than in
beds of narrow PSD.
3. The results of these measurements also provided information on other related

parameters, such as void frequency. As shown in Fig. 5.10, for U > 0.5 m/s, the frequency
of the detectable voids in the bed of wide PSD was much smaller than for the narrow
size distribution.
5.2 Effect on the Bubble or Void Size
Bubble or void size is one of the most important indices in the design of bubbling fluidized
bed reactors, since it is related to various important bed properties such as bed expansion,
gas interchange between phases, particle circulation rate, heat transfer, and elutriation
of fine particles from the bed surface. Numerous measurements of bubble size have been
made, and many correlations have been presented (e.g. Kato and Wen, 1969; Mori and
Wen, 1975; Rowe, 1976; Darton et al, 1977; Werther, 1978). These correlations have
mainly been based on experimental results in fluidized beds of coarse (Group B or D)
particles. In fluidized beds of fine (Group A) powders, recent work (Werther, 1984;
Kai and Furusaki, 1985; Kai et al, 1987; Horio and Nonaka, 1987) has shown that the
bubbles grow in size until a steady bubble size, or so-called maximum stable bubble size,
is reached as a result of an equiUbrium between bubble coalescence and sphtting. The
latter can be explained on the basis of bubble stabiUty in fluidized beds (Grace, 1970;
CUft and Grace, 1972; CUft et al, 1974). Although some modifications of the pubhshed
bubble size correlations have been made to account for the bubble size in fluidized beds
of Group A particles (Werther, 1984; Horio and Nonaka, 1987), less information has been
reported on the effect of PSD on the void size in fluidized beds.
The effect of the particle size distribution on the bubble size, or void size at higher gas
velocities, was investigated in this study by means of two methods: visual measurement
and indirect evaluation from the determination of pressure fluctuations in a fluidized bed.
The former is discussed here, while the latter is described in Chapter 6.

The PSD. effect on the void size was visually investigated in the two-dimensional
fluidized bed. The static bed height was 0.9 m, and the superficial gas velocity ranged
from 0.05 to 0.6 m/s. For each kind of particles and at each gas velocity, bubble or void
phenomena were recorded by a video recorder (SONY SL-2000 with SONY HVC-2800
camera) at 30 frames/s for a period of 10 minutes. The phenomena were also recorded
by a CANON T-70 still camera with 1/250 sec. shutter speed and F1.8 aperture. For
each operating condition, 8 photographs were taken. The film used was Kodacolor Gold
400. The experimental schematic is shown in Fig. 5.12.
The average bubble size measured at a height 0.85 m above the distributor and the
maximum bubble size in the bed are shown in Table 5.1. The effect of PSD on the bed
expansion is also given in Table 5.1. Some photographs portraying bubble phenomena
in beds of particles with different PSD under various conditions appear in Figs. 5.13 to
5.18.
Figs. 5.13 to 5.18 indicate that there is a significant effect of PSD on the size of
bubbles in the fluidized bed, except at low gas velocity (U < 0.2 m/s) for the gas-solid
system used in our experiments. The void size was usually smallest for the wide size
distribution, while the difference between the beds with bimodal and wide PSD was
small. Voids achieved the largest size in the bed with a narrow PSD, probably due to
their high bed "viscosity" (Grace, 1970) or poor "fluidity".
Experimental bubble sizes measured by Werther (1984) using capacitance probes in
a three-dimensional fluidized bed of wide size distribution FCC particles in a column
of 0.45 m diameter at gas velocities < 0.2 m/s were lower than values obtained in the
two-dimensional bed in the present study. The void sizes observed in our experiments,
however, are similar to the results of Rowe and MacGillivray (1980) in a fluidization
column of 140 mm diameter obtained using X-ray cinephotography. It is likely that
1- CAMERA
2- FILTER
3- 2-D COLUMN
4- LAMP (300 W]
1 b
GAS
Figure 5.12: Schematic of experimental apparatus for observing bubble phenomena

Table 5.1: Void size and bed expansion in two dimensional fluidized bed
Particle Size distribution Gas velocity, m/s D ,mm*

bta D max,mm**
bt
N-FCC wide 0.05 1.23 27 31
N-FCC wide 0.2 1.37 51 81
N-FCC wide 0.4 1.55 58 83
N-FCC wide 0.6 1.80 55 70
N-FCC narrow 0.05 1.12 31 40
N-FCC narrow 0.2 1.26 67 95
N-FCC narrow 0.4 1.41 70 101
N-FCC narrow 0.6 1.65 72 107
S-FCC wide 0.2 1.33 44 74
S-FCC narrow 0.2 1.26 63 93
S-FCC bimodal 0.2 1.30 48 82
S-FCC wide 0.6 1.73 52 76
S-FCC narrow 0.6 1.63 71 105
• D is the average equivalent bubble or void size at z = 0.85 m.

ba
** Db.max is the maximum size observed over the 10 minutes period at z = 0.85 m.
1 1
d S
"•+3
I
s-
a.
1
O
o
S i
O
T3 C
QJ
IS] re
<u
'i
o
the wall effect in the two dimensional bed or interference of the capacitance probe are
responsible for differences. However, in our study, we are less interested in the absolute
value of bubble size in the fluidized bed than in the relative change caused by the PSD
influence.
The following qualitative and quantitative conclusions can be drawn from the pho-
tographs:
1. The bubble size is dependent on the PSD, especially as the gas velocity increases,
as shown in Table 5.1.
2. As the gas velocity is increased, bubble shape becomes irregular, and the boundary
between the dense phase and the dilute phase is less distinct. For FCC particles of wide
and bimodal PSD, these features are especially noticeable at gas velocities beyond about
0.5 - 0.6 m/s.
3. More particles are seen inside voids of fluidized beds of Group A particles than
has previously been observed in beds of Group B particles (see Section 5.1).
4. The bed expansion is also affected by the PSD, and is lowest for bed of narrow
PSD. The equation
H m} ~ *-ni{ D )^
9 b
( 5
- 1 3 )
based on the two-phase theory (Davidson and Harrison, 1963), is only suitable to express
the bed expansion for particles with narrow PSD at lower gas velocities, as shown in
Fig. 5.19. For FCC particles of wide and bimodal PSD, and at higher gas velocity, the
traditional two-phase theory must be modified, for example, by the n-type two-phase
theory, as suggested by Grace and Harrison (1969), to avoid overestimation of the visible
bubble flow (Grace and Clift, 1974) and underestimation of the dense phase voidage
(Abrahamsen and Geldart, 1980; Sun et al, 1983).
Figure 5.19: PSD effect on the bed expansion for N-FCC particles. Open symbols are
experimental values; filled in symbols are the corresponding predictions from eqn. 5.
with experimental values of Db, - a
Chapter 6
Pressure Fluctuations and R e g i m e Transitions
Pressure fluctuations have been used as an index of the "quality" of fluidization. The fluc-
tuations are closely related to the gas-solid movement and depend in a complex manner
on the gas and particle properties, bed geometry, and operating conditions. In this study,
the character of pressure fluctuations and fluidization regime transitions, especially the
transition from bubbhng and/or slugging to turbulent fluidization, determined from the
fluctuation measurement, were evaluated to compare the hydrodynamic performance of
fluidized beds containing particles of different size distribution, but equal mean diameter
and nearly the same density.
6.1 B r i e f review of previous w o r k
Pressure fluctuations have been used to define the quality of fluidization by many inves-
tigators and plant operators (e.g. Shuster and Kisliak, 1952; Winter, 1968; Schnitzlein
and Weinstein, 1988), with small amplitude and rapid fluctuations considered to be as-
sociated with good quality fluidization. Based on the analysis of the probabilistic and
statistical characteristics of pressure fluctuations along a fluidized bed, it is also possible
to estimate the size and the rising velocity of bubbles and slugs where these occur (Lirag
and Littman, 1971; Fan et al, 1981; Fan et al, 1983). The average bubble size appears
to be directly related to the standard deviation of the amplitude of pressure fluctuations
(Sadasivan et al, 1980; Kai and Furusaki, 1985; Geldart and Rhodes, 1986).
Pressure fluctuations have also been used to monitor operating conditions and to
117
Chapter 6. Pressure Fluctuations and Regime Transitions 118
distinguish between hydrodynamic regimes in fluidized beds, and to explore the charac-
teristics of these regimes (e.g. Yerushalmi and Cankurt, 1979; Li and Kwauk, 1980; Satija
and Fan, 1985; Jin et al, 1986; Brereton, 1987; Lee and Kim, 1988; Sun and Chen, 1989).
Generally speaking, the transition from bubbling (or slugging) to turbulent fluidization
is gradual with increasing superficial gas velocity.
The character of fluidized beds operated at high gas velocities and the dependency
of fluidization regime on the superficial gas velocity have been the subject of a number
of investigations. In early work, Lanneau (1960) investigated bubbling phenomena in a
fluidized bed (DT — 0.075 m) of alumina catalysts (d — 75 pm) over a wide range of
p
gas velocity (0.03 to 1.5 m/s) with the use of capacitance probes. He reported that the
performance of fluidized beds appeared different in three gas velocity ranges:
1. low gas velocities (0.03 m/s < U < 0.3 m/s), where two-phase character of the
fluid bed was prominent, and discrete bubbles containing few solids were present.
2. intermediate gas velocity (0.3 m/s < U < 0.9 m/s), where the bubbles began to
break up as the velocity was increased, and smaller voids containing a greater percentage
of the total solids were observed.
3. higher gas velocities (0.9 m/s < U < 1.5 m/s), where the condition of almost
uniform or "particulate" fluidization was approached, coinciding with rapid entrainment
of solids from the column.
Kehoe and Davidson (1971) detected bubbles (or slugs) with increasing gas velocity
in 50 and 100 mm diameter fluidized beds of fluid cracking catalyst. They described the
breakdown of slugs to a so-called "turbulent" fluidization regime at gas velocities between
0.45 and 0.5 m/s. Massimilla (1973) evaluated the bubble size and size distribution
inside a fluidized bed with catalysts of mean size and size distribution close to typical
industrial FCC catalysts using a capacitance probe at superficial gas velocities up to 0.65
m/s. At gas velocities above 0.4 m/s,flowconditions were found to degenerate into the
turbulent regime, where the probe detected a sequence of extended low density signals,
intermittently followed by. bursts of short perturbances. He also provided a laboratory
demonstration of the favourable solid-gas contacting efficiency of the turbulent fluidized
bed reactors for propylene conversion to acrylonitrile in a pilot reactor. Conversion for the
turbulent bed was higher than would have expected from a two-phase bubbhng model.
Yerushalmi et al (1978) characterized the transition from the bubbhng (or slugging)
regime to the turbulent regime by two gas velocities, U , the velocity at which the pressure
c
fluctuations reach a maximum, and Uk, the velocity at which the decay of the pressure
fluctuations begins to level off. Some investigators (Yerushalmi and Cankurt, 1979; Horio
et al, 1989) view U as the end of the bubbhng or slugging regime and Uk as the onset
c
of the turbulent regime; others (e.g. Yang, 1984; Jin et al, 1986; Lee and Kim, 1988;
Judd and Goosen, 1989) have treated U as the gas velocity corresponding to the onset
c
of turbulent fluidization. Still others (Satija and Fan, 1985; Brereton, 1987) found that
it is difficult to define Uk directly from the amplitude of pressure fluctuations, at least
for some particles.
In the bubbhng and slugging regimes, pressure fluctuations are mainly associated with
the formation, motion and splitting of bubbles or slugs as they rise along the bed (Fan et
al, 1981), and the amplitude of the pressure fluctuations depends on the size of bubbles
and/or slugs (Sadasivan et al, 1980; Kai and Furusaki, 1985; Geldart and Rhodes, 1986).
U can therefore be regarded as the gas velocity at which bubbles reach their maximum
c
size, dependent on the character of the gas-solid system and the geometry and size of the
fluidization column (Sun and Chen, 1989). Previous work has indicated that U increasesc
with particle size and density (Yerushalmi and Cankurt, 1979; Grace, 1982; Jin. et al,
1986) and decreases with column size (Sun and Chen, 1989; Judd and Goosen, 1989).
Lee and Kim (1988) calculated various statistical properties such as mean amplitude,
standard deviation, skewness and flatness of pressure fluctuations to determine U . They c

found that U values obtained from these various methods were very similar.
c
The dependence of fluidization regime on the gas velocity is influenced by such factors
as column diameter. Experimental results (Kehoe and Davidson, 1971; Thiel and Potter,
1977; Sun and Chen, 1989) indicate that for a given gas-solid system the transition from
the bubbling (or slugging) regime to the turbulent regime in columns of smaller diameter
appears later than in columns of larger diameter.
It is clear that the transition velocity from bubbling to turbulent fluidization increases
with particle size and density. Little information, however, is available on the effect
particle size distribution. Kehoe and Davidson (1971) addedfinerparticles (15 pm < d < p
43 pm) to coarser fractions (40 ^m < d < 90 pm) and found that the transition velocity
p
decreased from 0.5 m/s to 0.32 m/s. When "fines" are added to "coarse" particles,
however, both the particle size distribution and the mean size are changed. Results on the
effect of size distribution for solids having comparable mean size are lacking (Yerushalmi,
1986).
6.2 Experimental Equipment and Procedure
Pressure taps were installed vertically along the 102 mm diameter column with a spacing
between adjacent taps of about 0.25 m, as shown in Fig. 2J.0. The inside end of each
tap was covered with a screen to prevent particles from entering. The two ports were
always covered by dense phase, to avoid apparent regime transitions noted by Geldart
and Rhodes (1986). Each tap was connected to a differential pressure transducer, which
had two input channels and produced an output voidage proportional to the pressure
difference between the two input channels. Differential pressure signals at different bed
heights and different operating conditions were logged to an IBM-XT computer via an
A / D converter. Each signal was logged for 40 s at a rate of 100 points per second and
recorded simultaneously in analogue format on a UV chart recorder.

Static bed heights of 0.15, 0.40, 0.70, 1.0 m were used in the experiments. A wide
range of superficial gas velocities (from 0.04 to 1.75 m/s) was tested so that the bubbhng,
slugging, turbulent and fast fluidization regimes would all be included.
6.3 Influence of P S D on Pressure F l u c t u a t i o n s
The differential pressure fluctuations allow the two phase character of the fluidized bed to
be investigated. A dimensionless standard deviation of the pressure drop or the pressure
fluctuation intensity, defined as
r
F
(6-1)
is used to express the amplitude of the pressure drop fluctuation, where N is the total
number of data points, while A P is the mean pressure drop across the measured section
and APi is the instantaneous pressure drop across the same section.
Since bubbles (or slugs) are the major cause of the pressure fluctuation in aggregative
fluidized beds (Fan et al, 1981), while the average bubble size appears to be directly
related to the amplitude of pressure fluctuations (Kai and Furusaki, 1985; Geldart and
Rhodes, 1986), analysis of the pressure fluctuations is useful in understanding the influ-
ence of the particle size distribution.
The average bubble size tends to increase with distance above the distributor due to
coalescense and interaction between the bubble phase and the emulsion phase. The local
bubble size in bubbhng fluidization is commonly (e.g. Kato and Wen, 1969; Darton et
al, 1977; Werther, 1978) expressed by an equation of the form:
d =d
b bo + A{U-U ) z
mf
B c
(6.2)
where db is the initial bubble diameter at the distributor and A, B and C are constants
0
which may depend on mean particle size (Kato and Wen, 1969). Given the dependence
of pressure fluctuations on bubble behaviour, it is reasonable to seek an expression for
the amplitude of pressure fluctuations of analogous form, i.e.
F p = F
po + a(U-U ) z
rnf
b c
(6.3)
where Fpo is the pressure fluctuation intensity at z = 0 and a, b and c are constants.
Figures 6.1 and 6.2 show the variation of F with superficial gas velocity for N-FCC
p
powders with different size distributions. It is clear that the amplitude of differential
pressure fluctuations is influenced by the PSD. Among the three size distributions used
in our experiments, the narrow blend gives the highest F at given operating conditions,
p
except at very low gas velocities (U < 0.2 m/s), where the values of F are very similar p
for the different PSD's.

It can also be observed from Figs. 6.1 and 6.2 that F first increases with the gas p
velocity, then reaches a peak, and finally decreases gradually. This behaviour is similar
to results reported by other workers (e.g. Yerushalmi, 1986) and to our findings in a two-
dimensional bed column described in Section 5.2. It was also found that F is dependent p
on the location of the measuring interval, even when all the pressure ports are immersed
in the dense bed. For H j < 0.7 m, F appears to increase with the vertical position
m p
coordinate z. However, when the static bed height was increased to 1 m, F did not p
always increase with z. For the wide blend, F increased with z in the lower part, but
p
decreased with z in the upper part of the bed. For narrow blend particles, the behaviour
was more complex: for U < 0.7 m/s, F was found to increase with z, while F for the
p p
higher measurement interval reached a maximum, then increased with z over the lower
part of the bed, while decreasing over the upper section. This suggests that if the bed
is deep enough, there may be two different regions, depending on the PSD, among other
i i i 1 1 1 1 r
1 1 — — i r
• wide AZ2
a A » A Z i
LL
0.5F • narrow A Z2
o » A Zi
• bimodal A Z2
LL
1 ± m
-A
O
O' -A- -A-
- -A A-
0.1 -A
_i 1 1 1 i_ I L.
0-5
_1
1.0 1.5
Superficial Velocity m/s
Figure 6.1: Effect of PSD on pressure fluctuations with H f = 0.7 m over intervals A z m 1
(30 to 280 mm), A z (430 to 680 mm) and A 2 3 (680 to 930 mm).
2
AZi AZ2 AZ3
0.5
wide A A A
narrow o • »
0.4
Fp
0.3 3
3 3
3
3 3
•
0.2 A & A A
A A
O
0.1
0.4 0.8 1.2

U m/s
Figure 6.2: Effect of PSD on pressure fluctuations with H mf = 1 m over intervals A z ,

a
A z and A z (as in Fig. 6.1).

2 3
parameters. For particles of wide PSD in the lower region, bubble size increases with
z. Bubble coalescence and growth are predominant in this region, and the coefficient c
in eqs. 6.2 and 6.3 is positive. However, in the upper part of the bed, bubble splitting
becomes significant and bubble size (and hence pressure fluctuations) approach limiting
values.
It can be seen that the difference between the narrow and wide size distributions
increased with both gas velocity and vertical position, while F N
P< was always greater
than Fw Pi • Therefore, we can obtain from equation 6.3,
FPtN - F,p w = a (U
N - U ) z^
mi
hN
- a (U w - U ) zmf
bw cw
> 0 (6.4)
where the subscript N denotes the narrow blend and W the wide blend. That is,
— {U - U ) -mf
bN
~
bw CN
Z
CW
> 1. (6.5)
aw
The assumption that the effect of PSD on F P I C I is negligible has been used to obtain the
above relationships. When the bed was operated in the bubbhng regime where eqn. 6.2
is suitable, it can be found from Figs. 6.1 and 6.2 that both fcjv and c/v are larger than
bw and cw- According to eqn. 6.5, ajv, should therefore also be larger than aw, since
in the range discussed here, both (U - U f) and z are smaller than 1. That is, all three
m
constants in eqns. 6.2 and 6.3 are related to the PSD. At the same operating conditions
and height, the bubble size is dependent not only on the mean particle size, as some
investigators have stated, but also on the size distribution. The PSD effect, however, has
not been taken into account in any of the published correlations for bubble size.
The difference in F between the bimodal and wide blends seems small. At low gas
P
velocities (before reaching the maximum value of F ) , F is similar for these two blends
P P
over the lower portion of the bed, while for the upper part F was a little higher for P
the wide blend than for the bimodal blend. After F reached its peak, however, F was P P
always greater for the bimodal blend than for the wide blend, in both the upper and
lower portions of the bed.
Based on the results that the average bubble size appears to be directly related to the
standard deviation of the amplitude of pressure fluctuations (Sadasivan et al, 1980; Kai
and Furusaki, 1985), we can assume that the bubbles (or slugs) reach their maximum
sizes at the gas velocity, U , where F reaches a maximum. The maximum bubble size
c p
and superficial gas velocity at which this is achieved can be taken as two parameters to
characterize the hydrodynamic performance of fluidized systems, since at given operating
conditions, gas-solid contacting can be improved by reducing the bubble (or slug) size
(Grace and Clift, 1974).
Comparison of F for different PSD's reveals that F

p PtTnax of the narrow blend is much
higher than for the other size distributions, especially in the upper portion of the bed,
while F PiTnax for the wide blend is similar to that of the bimodal distribution. This
suggests that the maximum stable bubble size in the fluidized bed with a narrow PSD
is larger than for a wide or bimodal PSD, as discussed in Section 5.2. This may well be
caused by the lower aeration capacity of the narrow blend (Ip, 1988) and the higher bed
"viscosity" for the narrow PSD .
6.4 Influence of P S D on Regime Transitions
The PSD effect on £ / m6 , the transition velocity from the particulate to bubbling flu-
idization, has been reported by Abrahamsen and Geldart (1980), Sun et al (1983) and
Ip (1988). For the same spent F C C particles as tested in our experiments, Ip (1988)
observed that U bJUm}forparticles of wide size distribution was 20% more than for a
m
narrow PSD.
The gas velocity, U , at which F reaches a maximum, can be taken as the end of the
c p
Table 6.1: Transition velocity U for N-FCC particles of different size distributions over
c
intervals A z (30 to 280 mm), A z (430 to 680 mm), A z (680 to 930 mm) and A z (930
a 2 3 4
to 1180 mm).
U, c m/s {Hmj = 0.7 m) U,

c m/s (H mf = 1 m)
PSD
Az a Az 2 Az 3 Azi Az 2 Az 3 Az 4
wide 0.72 0.62 0.50 0.71 0.62 0.50 0.38
narrow 0.90 0.75 0.65 0.78 0.75 0.63 0.50
bimodal 0.75 0.65 0.55 0.75 0.64 0.53 0.45
bubbling or slugging regime and the onset of the transition to the turbulent fluidization
regime (Yerushalmi, 1986). The physical meaning of U is that around this velocity bub-
c
bles approach a maximum size which depends on the bed "viscosity" and the aeration
capacity of the particles. At velocities below U , bubble coalescence appears to be pre-
c
dominant, while above U , splitting becomes frequent. The value of U is therefore related
c c
to bubble growth and coalesence. Low U implies that bubbles split more readily, result-
c
ing in an improvement in gas-solid contacting. The transition velocity, U , determined c
from the pressure fluctuation curves (as shown in Figs. 6.1 and 6.2) for different particle
size distributions, with different H f and at different intervals is listed in Table 6.1.
m
The influence of the PSD on U is shown in Fig. 6.3. It is clear that the PSD does
c
affect the transition from bubbling to turbulent fluidization. The highest U is attained
c
for the narrow blend particles, while the lowest U occurs for the wide distribution.
c
From Fig. 6.3 it can also be observed that U decreased with increasing z, whereas F
c p
increased with z. Transition to the turbulent regime appears to occur sooner for deeper
beds. This suggests that the growth of bubbles in fluidized bed may trigger the regime
transition from bubbling to turbulent fluidization. It also implies that there does exist
an "effective" maximum stable size (Harrison et al, 1961), at least for some Group A
particle systems, even if this size is actually a statistical mean value.
The influence of the static bed height on pressure fluctuations has also been examined.
The measurement of pressure drop between two ports which are always immersed in the
dense bed avoids the effect discussed by Geldart and Rhodes (1986), whereby a change
in bed level leads to an apparent transition to turbulent fluidization. It is found that
the effect of static bed height on F p (Figs. 6.1 and 6.2) and U (Figs. 6.4 and 6.5) is
c
negligible, except for H f m — 0.35 m. In this range of static bed height, U decreases,
c
while F increases with H j, implying that the bubble size has not developed fully. For
p m
shallow beds, therefore, bubbles (or slugs) approach their "maximum size" at higher gas
velocities than for deeper beds, consistent with results reported by Canada et al (1976)
who found that there was no apparent effect of static bed height over a range of 250
to 700 mm. The velocity, U , at which the local pressure fluctuation peaks, however,
c
appears to decrease with height. These conclusions can be explained by the nature of
turbulent fluidization: with an increase of gas velocity, bubbles reached their maximum
bubble size break up more readily.
The effect of bed effective viscosity can be taken as a convenient starting point to
discuss the effect of particle characteristics on bubble behaviour in fluidized beds since
the fluidization performance may be substantially influenced by bed viscosity (Grace,
1970; Clift, 1986). The maximum stable size a bubble can attain before splitting usually
• i i 1 — • — i 1 i
Hmf cm
O narrow 100 '
- 70
& » 40
A wide 100
• „ 70
A » 40
• bimodal 70
^4
1 1 i i
0.4 0.8 1.2 1.6
Height, z, m
Figure 6.3: Effect of PSD on U for fluidized beds of N-FCC particles with different H
c m
and at different bed heights.

i 1 1 1 1 1
• -
m/s
•
0-8 •
-
O o
o
-
• •
0-4 - • AZi
O Az
<0>
2
AZ 3
+ Az 4
I iI iI -. 1I 1I 1I L_
0 0.4 0-8 1.2
Hmf m
Figure 6.4: Influence of H f on U for N-FCC particles with narrow PSD, over intervals
m c
Azi (30 to 280 mm), Az (430 to 680 mm), Az (680 to 930 mm) and Az (930 to 1180
2 3 A
mm).
i i i i i 1
m/s
•
0-8 -
• •
o O o
0.4 • A z-j • •
O AZ
<C> A Z
2
• Az 4
i i 1 1
0 0.4 0.8 1.2
Figure 6.5: Influence of H mj on U for N-FCC particles with wide PSD, over intervals
c
A z ; (as in Fig. 6.4).

increases with the effective dense phase "kinematic viscosity" (Clift et al, 197-'4). The
lower the "viscosity", the smaller the maximum bubble size and the earlier the transition
from bubbling to turbulent fluidization. In a fluidized bed of Group A particles, the
effective viscosity increases with particle diameter and decreases with fines content (Kono
et al, 1986a). For the gas-solid system tested in our study, Khoe et al (1991) based on
the results of collapse tests, reported that for a given average surface-volume diameter,
a wide size distribution gives a better air retention capacity than a narrow or a bimodal
distribution, while the effective viscosity of a fluidized bed was inversely proportional to
the air retention capacity of particles.
6.5 Axial Voidage Distribution
Bed expansion is one of the most important and often reported properties of fluidized
beds. Better "fluidization quality" is often associated Avith higher dense phase voidages
and smaller voids, both of Avhich result in higher bed expansion. For given operating
conditions, higher dense phase voidages probably correspond to higher average interstitial
gas flow relative to particles in the dense phase and lower gas flow rates within the bubble
phase (Grace and Clift, 1974). Rowe et al (1978) measured the dense phase voidage of
a bubbling fluidized bed at IOAV gas velocities (17 < 0.15 m/s) using an X - r a y technique
coupled with a densitometer. They found that there were large changes in dense phase
A'oidage with both gas velocity and fines (< 45 /xm) content of the poAvder.
Combined with the measurement of pressure drop fluctuations along the column, the
P S D effect on the axial bed voidage distribution i n the fluidized bed was also investigated.
The actual voidage, ey can be expressed (Weinstein and L i , 1989) as
(6.6)
based on momentum conservation of gas and solids i n the bed. If the acceleration and/or
the flow rate of the particles is not high enough, as in the bubbling, slugging and turbulent
fluidization regimes, the average local voidage between two vertical ports along the bed
can be approximated by
AZi p g
p
where AP{ is the mean pressure drop between ports at heights z,- and Zi+\.
The PSD effect on e^,- and its axial distribution is shown in Figs. 6.6 and 6.7. For gas
velocities less than 0.4 m/s, the narrow blend gave the highest voidage in the lower part of
the bed, partially because its value of is higher than for the other size distributions.
If the gas velocity > 0.4 m/s, the overall voidage for the narrow blend is usually the
lowest at all heights. The voidage increment, - e f, due to bed expansion, always
m
appears lowest for the narrow blend, as shown in Fig. 6.8. These results are in good
agreement with the observations of the influence of PSD on the average amplitude of
pressure fluctuations and bed expansion in the two dimensional bed (Table 5.1). This
verifies that the highest amplitude and the lowest voidage increment, corresponding to
the largest bubbles, occur for particles of narrow particle size distribution.
The axial distribution of overall voidage is also influenced by the PSD. It can be seen
from Fig. 6.7 that for the wide blend powder, the local voidage always increased with
height. For the narrow blend, on the other hand, this was only found for U > U . Below c
U , the change of the overall voidage along the column was small above a certain height
c
(around 0.35 m above the distributor), where the overall voidage was always lowest.
The effect of PSD on the bed expansion can be explained by the deviation of fluidiza-
tion behaviour from the two-phase theory for Group A particles of wide size distribution.
Consider a section of fluidized bed, across which the visible bubble flow rate is Q\> and
the average bubble velocity is u-b. The bed expansion caused by the visible bubble flow
•
•
0.8 ••
o
CD
• •
•
O)
CO
•o
'o
> • u
o ^ •
0.7
o A • o narrow A.Z2
•
• n AZ-i
wide Az 2
9 •
•
» Az-t
bimodal A z
0-6
2
• » • AZ1
0.5
-J 1 1 1 L_
0 0.5 1.0 ,, 1.5

U m/s
Figure 6.6: Effect of PSD on the overall voidage in fluidized bed; H mJ = 0.7 m. Intervals
Azi and Az are as in Fig. 6.4.
2
1 1 1 1 1
WIDE NARROW
0.9 -
A o
A A Z2
• • Az 3
0.8 —
8 A
f
0.7 • • t ft
•
A
2
dm
A
6
•t o
0-6
6 6
-
0.5
i 1 i 1 1
0.2 0-4 0.6 0.8 1
U m/s
Figure 6.7: Effect of PSD on the overall voidage in fluidized bed; H mf = 1 m. Intervals
A z , A z and A z are as in Fig. 6.4.
a 2 3
0.4
• CP o_+
c o •
E 0-3
o g
c
a>
a
cn
• o•
co
•
;u
'o> • o •
• o
•
•
•
0.2
6 o • o •
•
•
f
•
o O
•
•
•
o narrow AZ 2
0-1 • » . AZi
• wide Az 2
• " A Zi
• bimodal A z 2
• »i A Zi
• — ' — J — i — i — i — i i i i i i i i . i i , I
0 05 1.0 1.5
U m/s
Figure 6.8: PSD effect on the voidage increment in fluidized bed; H mf = 0.7 m. Intervals
Azj and Az are as in Fig. 6.4.
2
can then be expressed as
r* Q
AH b = H - H d = / -^-dz, (6.8)
Jo Au b
while the expansion caused by expansion of the dense phase is given by
AH
d = H- d H mJ = [ (e
H
d - e )dz mi (6.9)
Jo
The total expansion of the bed
AH = AH b + AH d = H-H , mt (6.10)
that is,
AH
It is found from Figs. 6.6 to 6.8 that for particles of wide size distribution,
^>0, (6.12)
dz
that is,
^ > 0. (6.13)
dz
Therefore,
de d du b Q
b dQ b 1
• ( 6
- 1 4 )
If the two-phase theory holds, de /dz d = 0 and dQ /dz b = 0, so that du /8z

b < 0,
which means that bubble size decreases with height. Obviously, this cannot express
our experimental results for particles of wide PSD, as shown in Figs. 5.13 to 5.15 and
Figs. 6.1 and 6.2. Equation 6.14 therefore suggests that for Group A particles with wide
size distribution, Q , u and e vary along the bed height, and the increment of e may
b b d d
be higher than that of u and/or the decrement of Q along the column.

b b
For the narrow PSD particles with U < U and z > 0.35 m, it is found from Fig. 6.7, c
that dej/dz ~ 0, or dAH/dz w 0. If the assumption of the two-phase theory holds,

then the sign in equation 6.4 becomes an equals sign and dub/dz fa 0, which means that
the bed mainly operated in the slugging regime. This is consistent with the pressure
fluctuation measurements along the bed with particles of narrow PSD (Figs. 6.1 and
6.2).
Chapter 7
Reactor Performance in the Distributor Region
The distributor region plays an important role in determining the contacting efficiency of
fluidized bed reactors (Grace and de Lasa, 1978). Multiorifice distributors are commonly
used in large gas fluidized beds, and their effective design has been the object of a
number of studies. In this study, the PSD effect on the reactor performance in the
distributor region was investigated by comparing the contacting efficiency for different
distributor geometries and operating conditions. Four different multiorifice distributors
with different hole numbers and hole diameters were tested for particulate materials
having three different PSD's to evaluate the influence of gas distributor on the fluidized
bed reactors and its interdependence with the PSD effect.
7.1 Brief Review of Previous Work
Early work on gas distributors mainly concerned the effect of distributor type and pres-
sure drop on the "uniformity" of fluidization (e.g. Morse and Ballou, 1951; Walker, 1970;
Whitehead, 1971). Compared with the types of distributors often used in the commer-
cial fluidized bed reactors, porous plate distributors used in some small-scale laboratory
tests appeared to give more "uniform" fluidization, that is smaller bubble diameters and
higher bed expansion, while the influence of distributor type could persist for depths
of up to 1.5 m (Walker, 1970). The reaction conversion was also found to be signifi-
cantly better with a porous plate than with a sieve plate distributor (e.g. Hovmand and
Davidson, 1968; Bauer and Werther, 1981). Despite these favourable characteristics, the
139
Chapter 7. Reactor Performance in the Distributor Region 140
porous plate has rarely, if ever, been used in commercial fluidized bed reactors due to
cost, difficulty in manufacture, structural considerations, pore blocking and less flexibil-
ity in operation compared with other types of distributor, such as the perforated plate
or multitube distributor.
The minimum pressure drop across the grid to ensure that the distributor introduces
the fluidizing gas evenly through each hole, thereby keeping the solid particles above it
in constant motion and preventing the formation of defluidized zones within the bed, is
generally considered in relation to that of the pressure drop across the fluidized bed itself.
The required ratio of AP D to APn depends on not only the fluidization performance of
particles but also the ratio of H to (Clift, 1986).
From experience in the development of the Nu-Iron fluidized bed reduction process,
Agarwal et al (1962) suggested that for deep fluidized beds of high density material,
the pressure drop across the distributor plate should be at least 10% of the pressure
drop across the bed and in no case should be less than 0.5 psi (3.5 kPa). For relatively
shallow beds of low density material, they recommended a minimum pressure drop across
the distributor of 2 to 3 kPa.
Based on measurements by Hiby (1964) and Whitehead and Dent (1967), Zuiderweg
(1967) proposed that for "a perforated plate or nozzle distributor the pressure drop should
be related to both the minimum fluidization velocity, U f, and the minimum velocity at
m
which all nozzles operate, Um. For U /U f m m from 1 to 2, they advocated APD = 0.15
AP ,
B while for U /U
M MF » 1, APD = 0.015 AP . B
The height of the entrance region for a multiorifice distributor was investigated by
Fakhimi et al (1983) by measuring the differential pressure between the gas spout and
adjacent dead zone. The principal factors influencing the height were found to be U f, m
mean particle size, orifice spacing and gas flow rate. They also presented an expression
for the minimum APD/AP B to ensure even gas distribution.
Ho et al (1984) studied the characteristics of the distributor pressure drop in a flu-

idized bed corresponding to perforated distributor plates with regular and irregular (i.e.
with different hole size and/or non-uniform hole location) design. They reported that
abnormal conditions, e.g. hole blocking and channeling, can be detected from the anal-
ysis of the distributor and bed pressure drop data. Criteria for uniform gas distribution
across the distributor holes strongly depended on the minimum fluidization velocity of
the fluidizing particles, with Um/Umj higher for particles of lower U f. m
Wen et al (1978) examined the formation of dead zones using two and three-dimensional
fluidized beds. The dead zone height was found to be affected by the gas velocity, dis-
tributor type, orifice pitch, orifice diameter and particle size. The behaviour of the two-
dimensional bed could not be readily extrapolated quantitatively to three-dimensional
cases. Wen et al (1980) detected the presence and disappearance of dead zones in a
cylindrial bed of diameter 262 mm using self-heating thermistor probes. Empirical cor-
relations were proposed to calculate the critical orifice gas velocity required to eliminate
the dead zone as a function of pitch, orifice diameter, particle size and particle density for
a perforated plate distributor. Decreasing the pitch and increasing the orifice diameter
was found to help to ehminate dead zones. Rathbone and Davidson (1986) also found
that the dead zone could be avoided by suitable orifice spacing.
Patrose and Caram (1984) studied the influence of particle density (plastic and glass
beads of diameter 600 to 1000 pm) and inlet gas velocity on the velocity of solids and
dead zone formation by using an optical fiber probe in a two-dimensional column. They
reported formation of a steady spout for the variety of bed conditions studied. The size
of the dead zone decreased with increasing inlet gas velocity. The inlet grid jet was not
a hollow gas cavity, but a region of active solids motion.
Briens et al (1978, 1980) investigated factors that influence grid leakage, such as
gas velocity, static bed height, baffles, grid thickness and hole shape. They found that
grid leakage is caused by pressure fluctuations at the top of the grid holes, which may
result from bubble formation at the grid orifices and sloshing of the bed. Inserting
wavebreakers (such as baffles and tubes) and selecting the proper grid thickness could
reduce grid leakage significantly.
Oki et al (1980) investigated discharge of gas into a three-dimensional fluidized bed
from a multiple-orifice gas distributor using an optical-fiber array probe to test the effects
of distributor orifice configurations and gas flow. They reported that orifice spacing is
an important factor among the variables which affect the mode of gas discharge from the
distributor.
Hsiung and Grace (1978) evaluated the effects of orifice diameter, gas velocity through
the orifice, particle size and plenum chamber volume on mean frequency of bubble for-
mation by using a sensitive pressure transducer. For each of the orifices studied, the
frequency of pressure fluctuations increased with increasing gas flow rate and then lev-
elled off at a value of 19 to 25 Hz. The gasflowfor onset of multiple bubble formation
decreased with decreasing orifice diameter and appeared to be independent of particle
size for the range investigated (glass ballotini with size ranges of 88 to 125 pm, 149 to
250 pm, and 250 to 420 /xm). The frequency spectrum also became more complex as the
particle size was reduced.
Some studies have been carried out to investigate how the distributor influence reactor
performance. Bauer and Werther (1981) tested the influence of distributor type (porous,
perforated and nozzle plate) and free area ratio. They reported that the conversion
obtained with the porous plate distributor was much better than for the perforated plates.
By comparing conversion data from two perforated plates of different free area (8% and
4%), they also found that smaller holes yielded a better dispersion of the fluidizing gas
and hence better conversion.
Yang et al (1984) measured the reaction conversion (butene-1 to butadiene) in the

distributor region for three different distributors, each with same proportion of free area,
but with different hole numbers and diameters. They observed that a zone existed near
the distributor in which the efficiency of chemical conversion is high compared with the
region near the bed surface. A progressive increase in conversion was found as the initial
bubble size decreased.
It has been discovered that bubbling-bed models do not properly represent reaction
near the gas distributor (Chavarie and Grace, 1975) or in the freeboard (region above
dense bed) (Furusaki et al, 1976). To consider the reaction and transfer process in the
vicinity of the grid, Behie and Kehoe (1973) proposed a two-phase model which extended
a well-known model by Orcutt et al (1962). A plug flow jet region devoid of particles
was assumed to act in series with the bubble phase. Mass transfer coefficients between
the jets and surrounding dense phase were based on measurements in a 0.61 m diameter
bed of 60 pm cracking catalyst. Grace and de Lasa (1978) extended this model to two
simple alternate models and allowed for solids dispersed in the jet phase.
Sit and Grace (1986) investigated interphase mass transfer during bubble formation
in the entry region of fluidized bed by measuring tracer concentration at different radial
positions and vertical levels. Interphase transfer was found to occur by two mechanisms
(convection and diffusion), when bubbles formed directly at an orifice at the base of a
fluidized bed of non-adsorbing particles.
Heat transfer in the grid zone between bed and tubes has also been studied (Virr
and Williams, 1985; Ho et al, 1987). Heat transfer coefficients appear to be higher than
in the bubbhng zone of a gas-solid fluidized bed. Ho et al (1987) observed that in the
bubbhng zone the heat transfer coefficient increased and then decreased with the air flow,
while in the grid zone region, the heat transfer coefficient increased with the air flow over
the entire range of flow rates. They also determined that the heat transfer coefficient
increased with decreasing open area ratio and hole pitch distance.
Most previous work on gas distributors used in industrial fluidized bed reactors has
been concerned with three major issues:
1. Macroscopic criteria: What is the minimum pressure drop across the distributor
and the minimum orifice gas velocity to maintain uniform gas distribution?
2. Microscopic understanding: What is the gas-solid motion and mass transfer rate
at or near the grid?
3. Influence on reactor behaviour: How can the distributor be best designed to
optimize the reactor performance?
Although considerable work has been carried out on issues 1 and 2, there is no clear
answer to the question of how these two issues affect each other and influence the per-
formance of a fluidized bed reactor, particularly at high gas velocities, since in previous
work on grid region, U was usually less than 0.25 m/s. In this study, all three questions
above are considered, at the same time as the ozone decomposition reaction is used to
see the interacting influence of the PSD.
7.2 Experimental Method
In most of our reactor experiments (Chapters 3 and 4), No. 2 gas distributor (see Table
2.3) was used, in order to ensure that the pressure drop across the distributor at high gas
velocities could be provided by the blower available in oiir building. Four multi-orifice
plates with three different proportions of free area, three hole diameters and two pitch
spacings were used to test the influence of the pressure drop across the distributor and
the geometric nature of the grid. Their design specifications appear in Table 2.3.
Investigation of the effect of the distributor on reaction in the grid region could be
impeded by particle leakage through orifices of the distributor, since significant change in
catalyst inventory could occur, especially for low bed heights. A screen with very small
holes, less than the minimum size of particles tested, is often used to prevent leakage.
For Group B particles with narrow size distribution, the pressure drop across the screen
may be neghgible. However, for Group A particles containing fines, the pressure drop
across the dense screen may be higher than that across the distributor plate at high gas
velocities.
In this study, particle leakage was prevented by using distributors consisting of two
multi-orifice plates with very different free areas. The ratio between them was from 4
to 15, while the angle OLD (Fig. 2.12) was kept less than 60% of the angle of particle
repose. Since distributor pressure drop is approximately proportional to the square of
gas velocity through the orifice, the pressure drop across the lower plate is less than 10%
of the total pressure drop across the distributor. The amount of particle leakage during
each run was less than 0.3% of the total initial mass in the bed.
Experiments were conducted in the aluminum fluidized bed reactor at two gas veloc-
ities, 0.2 m/s and 0.6 m/s, and with two catalyst charges, 1.2 kg and 3 kg, for each of
four distributors and two particle size distributions - wide and narrow. The gas velocity
through the orifices was from 4 m/s to 60 m/s, while the static bed height was 0.15 m
and 0.4 m. The kinetic rate constant was around 8 s~ . The mixed feed gas samples
x
withdrawn below the distributor were analysed to ensure ozone concentrations in air
around 600 ppm. Outlet gas samples were withdrawn through the port at the top of
the reactor (see Fig. 2.10). The distributor performance was determined by the reaction
conversion or the contacting efficiency.
To evaluate the influence of gas distributor on the hydrodynamic phenomena in the
grid zone, three differential pressure transducers were used to measure the pressure drop
across the distributor, across a region of the bed just above the distributor and across
the upper section, as shown in Fig. 7.1.
Chapter 7. Reactor Performance in the Distributor Region
146
710
-AR b2
460
310
-AP,b1
Z= 60 mm
AP C
GAS
Figure 7.1: Schematic showing regions where pressure drops and their fluctuations
were
determined
7.3 Influence of Distributor Geometry on Reaction Conversion
Results from the conversion measurement of ozone decomposition for different gas dis-
tributors, gas velocities and catalyst inventories are shown in Figs. 7.2 to 7.5. It can be
seen from Figs. 7.2 and 7.3 that for a given gas velocity, higher conversions are achieved
in beds fitted with distributors with both higher orifice gas velocities and smaller pitch.
It should be noted, however, that the influence of pitch is more significant than that of
free area. Although the pressure drop across distributor-4 could be much higher than that
of distributor-3 or distributor-2, while the hole diameter is the same for distributors 3 and
4, the conversion was higher than for distributor-4. For columns fitted with distributor-3,
the actual pressure drop across the distributor at U = 0.2 m/s was less than the APr} min
y
required by Agarwal et al (1962). For beds fitted with distributor-4, however, the pressure
drop across the distributor in our experiments was always higher than that required by
Agarwal et al. This suggested that high pressure drop across the gas distributor in
fluidized bed is not enough to maintain good performance of the distributor. Other
factors, such as the pitch or the number of holes, should also be considered.
Based on the measurements near the grid using a self-heating miniature glass bead
thermistor, Wen et al (1980) proposed the following empirical correlation to calculate
the minimum orifice gas velocity needed to eliminate the dead zone completely for a
perforated plate distributor of a three dimensional fluidized bed:
( n _ A A0.16J0.44
U ie = U ,+0.0m[
jm
[P
p P
} (7.1)
where Uj is the minimum gas velocity through the orifice to eliminate the dead zone, p
c
the orifice pitch, d„ the orifice diameter, and Uj f the orifice gas velocity corresponding
m
to minimum fluidization.
If the above correlation is applicable to our experimental system, the dead zone
1 2 3 4 5 %
FREE AREA OF DISTRIBUTOR
Figure 7.2: Influence of distributor free area and pitch on conversion: N-FCC particles
of wide PSD, H = 0.15 m.
mf
149
1 1
1 l l
1
0.8
• — _
— — • • —
0.6
x •- .
o
•
-
• U m/s
•
hole number
• 0-2 21
0.2 - • 0-6 21
o 0-2 4
• 0-6 4
I 1
I I
1 1
I ! Li_ i
0 1 2 3 4 5
of narrow PSD, H f = 0.15 m.
m
150
i i i i i i
0.8 •
n
i i . •
•
o • — •
0.6 •
U m/s
0.4 hole number
• 0.2 21
• 0.6 21
0.2 O 0.2 4
• 0.6 4
i i i i i i
0 1 2 3 4 5 %
of wide PSD, H = 0.4 m.
mf
of narrow PSD, H f = 0.4 m.
m
above the distributor-4 should be almost eliminated since U/aj^ > Uj for all running c
conditions, where a d4 is the free area fraction of distributor-4. For distributor-3, however,
U/ d3
a
< Ujc, which means that a significant dead zone should occur above the plate
according to eqn. 7.1. These conclusions are not consistent with our experimental results,
as shown in Figs. 7.2 to 7.5. This suggests that single criteria based on hydrodynamic
experiments are insufficient to determine the distributor performance in fluidized bed
reactors.
The deviation of our experimental conversion results from what one would expect
from Eqn. 7.1 and the criteria of Agarwal et al (1962) and Zuiderweg (1967) may be due
to several reasons. One is that most previous work was carried out with Group B particles
with a filter or packed material (e.g. glass wool was used by Wen et al (1978, 1980) )
below the distributor plate to prevent weeping or dumping of particles. For example, the
influence of pitch distance may be underestimated by Eqn. 7.1 due to the pre-distribution
effect by the glass wool below the plate, and the first term on the right hand of Eqn. 7.1
can become insignificant for Group A particles due to the small magnitude of U f • It is m
also possible that the gas-solid movement in the grid zone was influenced by measuring
probes in this region used by some investigators (e.g Wen et al, 1980; Old et al, 1980).
Our experimental ozone conversion results reveal that the total influence of multi-
orifice distributors in fluidized beds on reactor performance may consist of several aspects.
Possible issues include:
1. Uniformity of gas distribution; This may be related to the pressure drop across
the distributor (Agarwal et al, 1962; Zuiderweg, 1967; Fakhimi, 1983);
2. Dead zone above the distributor plate: This is related to the pitch, hole diameter
and orifice gas velocity (Wen et al, 1978, 1980; Oki, 1980);
3. Initial bubble size: This is dependent on the volumetric gas flow rate per orifice
(Davidson and Harrison, 1963; Yang et al, 1984).
Comparing the conversions for the column fitted with distributor-1 to that with dis-
tributors 2 and 3 (all with the same centre-to-centre pitch but different hole diameter and
free area), onefindsthat reaction conversion in the grid zone decreases with decreasing
grid pressure drop and increasing free area. The influence of the fractional free area
seems small in our experiments, as shown in Figs. 7.2 to 7.5. Based on comparison of the
conversions for the column fitted with distributor-4 and distributor-1 (same free area,
but different hole diameter and pitch) and comparison between results for distributors
4 and 3 (same hole diameter but different pitch and free area), we see that the effect of
pitch is more significant than that of pressure drop or free area.
At low gas velocity, e.g. 0.2 m/s in our experiment, the mass of particles entrained
in the freeboard region is negligible compared with the total mass of particles. If the
voidage in a fluidized bed is assumed to be uniform and equal to e/, except in the dead
zone, the following equation can be written, based on a mass balance of particles in the
grid region and other regions of the fluidized bed, as shown in Fig. 7.8:
= (1 - e )H
f 12 + (1 - e )H (l f m - ID) + (1 - e )H .
mf 0ll D (7.2)
pA
p
where 7 D is the ratio of dead zone volume in the grid region to the total bed volume
between levels 0 and 1, H and H are bed heights between levels 0 and 1, and 1 and
m 12
2, respectively, as shown in Fig. 7.8. Hoi is higher than H„, the height of the dead zone,
which was reported to be usually less than 60 mm (Wen et al, 1978).
From eqn. 7.2, since H = H 0 1 -f H , we can write
12
Wcot
(e -e
/ rn/ )/f 0l7l? -f(l-e7)iY (7.3)
Therefore,
ID = , [ f f f - (1 - e,)H]
1
(7.4)
-noi(e/ - em/j Ap p
0.8
0.6 o
•
o Hmf, m
0.4
• 0-16 D1
• 0«4 »
0.2 O 0-16 D4
• 0-4
0 0.2 0.4 0.6 0.8 m/s

U
Figure 7.6: Distributor effect on reaction conversion at different bed heights: N-FCC
particles of wide PSD; Di denotes distributor-i, (see Table 2.3).
u m/s
Figure 7.7: Distributor effect on reaction conversion at different bed heights: N-

particles of narrow PSD; Di denotes distributor-i, (see Table 2.3).
r~ LEVEL 2
E
E
o
m
LEVEL 1
o
CD
LEVEL 0
GAS INLET
Figure 7.8: Location of taps between which differential pressures were measured.
For a given mass of particles, the lower the H Y the more the 7^, i.e. the greater the
proportion of dead zone in the grid region.
Combining eqns. 7.4 and 6.7, we obtain
ID = -77-72 —\{—J - (1 - e/)#oi - — — ] (7.5)

#01 (e/ - <W) p A p pg
p
For columns fitted with different distributors but operated at the same conditions, the
above equation can be written as
1 t W cat A(P ) i a f l
iDri^-js-jz r~^TT~ ( 1 _ e
>^ 0 1
-^
(7 6
-noi(e/,i - em/J p A P pg p
Therefore, the difference of dead zone volume in beds fitted with distributors 4 and 1 can
be evaluated from
V ,4 - Vb,i = AH (
D 01 LDT4 - LDA ) =- — - [(AP12)c,i - ( A P ) 1 2 A 4 ] (7.7)
From our pressure drop measurements along the column it is inferred that there were
some dead or poorly fluidized zones above some distributors, such as distributor-4, used
in our experiments. The volume of dead zone, ^DAHQI, is dependent on some parameters
of the distributor, as shown in Figs. 7.9 and 7.10. Both pitch and hole diameter (or free
area) influence the volume of the dead zone. The effect of pitch appears, however, to
be more significant than that of hole diameter. Based on the above model, the dead
zone volume in columns fitted with distributor-4 can reach 40 - 50% of the grid zone,
while for distributors 2 and 3, the dead zone volume was around 20%. This suggests that
approximately 15 to 20% of the initial mass of particles may be subject to malfluidization
in the dead zone on the grid for H f m = 0.15 m and distributor-4, if distributor-1, which
had the best performance in our experiments, had all particles in the grid zone being
utilized fully. The reaction conversion (or contacting efficiency) achieved in vessels fitted
with distributor-4 is therefore less than for distributor-1, consistent with the conversion
1 2 3 4 5 %
Figure 7.9: Effect of pitch on the pressure drop between ports 1 and 2: N-FCC of wide
PSD, U = 0.2 m/s.
Hmf hole number

• 0.4 m 21
CO • 4
Q.
o 0.16 m 21
3.2 • » 4
3.0
CL
< 2.8
1.2
O
1.0
0-8 _L
1 2 3 4 5 %
Figure 7.10: Effect of pitch on the pressure drop between ports 1 and 2: N-FCC of
PSD, U = 0.2 m/s.
results in Figs. 7.2 and 7.3. For H f — 0.15 m, therefore, the effect of a dead zone or of
m
poorly fluidized particles may play a key role in determining the overall performance.
For H f = 0.4 m, the dead zone influence tended to be reduced, since the ratio of
m
particles in the dead zone to the total initial mass of particles was only around 5 - 8%
for distributor-4, if there was no dead zone for distributor-1. Comparison of the reaction
results shown in Figs. 7.4 and 7.5, suggests that the influence of initial bubble size caused
by different gas flow rate through the orifice also plays an important role, as discussed
by Yang et al (1984).
By comparing the results in Fig. 7.9 with those in Fig. 7.10, we find that the particle
size distribution also affected, to a certain extent, the volume of dead zone in the grid
region. The dead zone volume with distributor-4 appears to be more extensive in beds
of narrow PSD than in beds of wide PSD.
Comparing the conversion obtained in fluidized beds with H f = 0.15 m to that with
m
H f — 0.4 m, it is clear that the "entrance effect" due to the grid zone is related to the
m
PSD, as shown in Figs. 7.6 and 7.7. For particles of wide PSD, the ratio of conversion
achieved in the grid zone for H j = 0.15 m to that obtained for H f = 0.4 m is higher
m m
than for the narrow PSD. This implies that the influence of the PSD on the conversion
achieved in the first 150 mm of bed was greater than that on the subsequent 250 mm.
Moreover, the "entrance effect" appears to be more significant for the wide PSD than for
the narrow PSD.
Chapter 8
Conclusions and Recommendations for F u r t h e r W o r k
8.1 O v e r a l l Conclusions
1. The effect of particle size distribution (PSD) on the performance of a fluidized

bed reactor was investigated by combining measurements of chemical conversion (ozone
decomposition reaction) with study of key aspects of the hydrodynamics for different size
distributions of FCC catalyst, each having a mean particle size of about 60 fim. Fresh and
spent FCC catalysts were prepared of "narrow", "bimodal" and "wide" size distributions
for comparative experiments. Most of the experiments were carried out in a column of
diameter 102 mm, with some of the hydrodynamic tests conducted in a two-dimensional
column.
2. The influence of PSD on hydrodynamics and conversion appears to apply to a
number of hydrodynamic regimes, but to be most prominent for gas velocities corre-
sponding to the turbulent fluidization regime in the wide size distribution catalysts. In
bubbhng and slugging fluidized beds, the influence of the PSD became more significant
with increasing superficial gas velocity.
3. Both the void size and the dense phase expansion in fluidized beds were found to
be influenced by the PSD, especially for U > 0.2 m/s. The void size was usually smallest,
while the bed expansion was highest, for the wide size distribution. Particles of narrow
PSD produced the largest void size.
4. The particle concentration inside bubbles or voids in a fluidized bed of Group A
161
Chapter 8. Conclusions and Recommendations for Further Work 162
particles was higher than what has been reported in the literature for Group B particles.
More than 70% of the voids contained more than 2.5% particles by volume.
5. The overall particle size distribution affects both the particle concentration in
voids and the particle size distribution in the dilute phase. A greater proportion of fines
was found to be present in the dilute phase than in the overall particle size distribution.
The solids concentration in the voids in fluidized beds of wide PSD is higher than for the
narrow PSD, especially at higher superficial gas velocities.
6. The transitions between fluidization regimes are influenced by the PSD. The ear-
liest transition from bubbhng or slugging to turbulent fluidization occurred in the wide
size distribution catalyst. Both the amplitude and the frequency of pressure fluctuations
are related to the PSD. A narrow size distribution gave the most vigorous fluctuations,
while the wide blend corresponded to the lowest amplitude of pressure fluctuations. In
the upper part of the bed, the lowest frequency of fluctuations occurred for the narrow
PSD.
7. A contacting efficiency of fluidized bed catalytic reactors was defined to evaluate the
influence of PSD on the reactor performance. The wide size distribution gave the highest
conversion and reactor efficiency, while the narrow blend gave the lowest conversion and
efficiency at the same operating conditions.
8. The commonly adopted practice of utilizing "fines content" to characterize Group
A powders is not sufficient to express the effect of particle size distribution. The perfor-
mance of fluidized particles depends not only on the quantity of "fines" (however defined),
but also on their nature and on the overall size distribution.
9. For particles of wide size distribution, the reactor performance can be improved
significantly by operating in the turbulent or fast fluidization regime. However, for par-
ticles of narrow size distribution, the benefit of operating at high gas velocities is greatly
reduced.
10. The influence of the particle size distribution and the hydrodynamic regimes
should be considered in modelling fluidized bed reactors. At lower gas velocities for
wide PSD or over a wider range of gas velocity for narrow PSD, the conversion can
be successfully predicted by the "Two-Phase Bubbhng Bed Model" with modifications
to express the hydrodynamic effect of PSD. For beds of wide PSD and operated at
higher gas velocities, a single phase Axial Dispersion Model gave good predictions of
the observed results, with closed and open boundary conditions at the inlet and top
of the bed, respectively, and with the effective axial dispersion coefficient derived from
measurements of Guo (1987).
11. A larger pressure drop across the gas distributors is not enough to assure good
performance of the distributor. The reactor efficiency in the distributor region was found
to be higher for a distributor with more holes, than for one with fewer holes even though
the latter had a higher pressure drop.
8.2 Recommendations for F u r t h e r W o r k
1. All the tests done in the current project were for Group A catalysts. It is recom-
mended that the kind of experiments here be extended to Group B catalysts.
2. In the apparatus used in the current work, it was not possible to exceed a superficial
gas velocity of about 1.8 m/s for FCC type particles. It would be useful to extend the
work to higher gas velocities. This could be accomplished by modifying the apparatus
employed in this project to provide a controllable particle circulation system, or by using
a different set-up so that the influence of PSD within the fast fluidization regime can be
explored.
3. It would be helpful to obtain more quantitative hydrodynamic results in a three-
dimensional column, e.g. by using optical fibres or capacitance probes to determine void
sizes for different particle size distributions.

4. Another technique, such as energy attenuation of 7-rays, should be developed to
investigate the PSD effect on solid concentrations in voids to avoid the interference of
mean particle size.
5. The effect of PSD on the freeboard phenomena in fluidized beds should be evaluated
by measuring the entrainment rate, solids hold-up and its distribution by particle size in
the freeboard region for different particle size distributions and different gas velocities.
165
Nomenclature
A cross-sectional area of fluidized bed m 2
df, bubble surface area per unit volume of bubble m~ x
A p surface area of a particle m 2
CAb CAd } concentration of reacting species in bubble, dense phase kmol/m 3
C ini C ut
A Ao inlet, outlet concentration of reacting component kmol/m 3
D gas molecular diffusivity m /s 2
Db diameter of bubbles (volume - equivalent) m

D e gas axial dispersion coefficient m /s 2
d p average particle diameter m

dpi particle diameter representing i th
size interval m
DT bed diameter m
E contact efficiency of fluidized bed reactor defined by eqn. 3.5
E , Ef
ca a contact efficiency with coarser fraction, finer fraction active
/ fraction of particles by mass in given size interval
F p dimensionless standard deviation of pressure drop fluctuations
fp frequency of pressure fluctuations s -1
g acceleration due to gravity m/s 2
G, solid mass flux kg/m s 2
H mean expanded bed height m

Hf m bed height at minimum fluidization m
f intensity of transmitted light W
I 0 intensity of transmitted light with no particles present W
166
k factor levels of the test
k^ dimensionless kinetic rate constant, k H f{\

T m - e f)/U
m
k q interphase mass transfer coefficient m/s
k r reaction rate constant based on the volume of the particles s~ l
k b, k
r rd effective activity of the bubble, dense phase solids s~ l
L path length for light transmission m
rii number of independent observations for one test
Ni number of voids with (I /I)i

0
NT total number of voids with solids concentration < 5 % (vol.)
Pe Peclet number, UH/D e
Pe p axial Peclet number based on the particle size
Q extinction coefficient
Qb visible bubble flow rate m /s3
S ratio of finer fraction to coarser fraction by volume in the bed
Sb, Sd ratio of finer to coarser fraction by volume in bubble, dense phase
U superficial gas velocity m/s
Ub bubble velocity m/s
U c superficial gas velocity at onset of turbulent fluidization regime m/s
Uj c minimum gas velocity through the orifice to eliminate the dead zone m/s
Umb minimum bubbling superficial velocity m/s
Umf minimum fluidization superficial velocity m/s
Ui re particle velocity relative to rising void m/s
U s particle velocity m/s
UT particle terminal settling velocity m/s

167
UTR transport velocity of particles m/s
v cat volume of catalyst particles m 3
Vcf volume of catalyst particles in fluidized bed reactor m 3
v cp volume of catalysts to achieve same conversion in plug flow reactor m 3
V
9 volumetric gas flow rate m /s 3
volume of a particle m 3
w cat mass of catalysts in fluidized bed kg
X dimensionless interphase mass transfer group -
X conversion of reacting component -
Xi mass fraction of particles having characteristic diameter dpi -
z height coordinate measured from distributor m
a size parameter, Ttd /\

p -
angle defined in Figure 2.12 degrees
€ voidage inside the voids in fluidized beds -
fraction of bed volume occupied by bubble phase -
ed dense phase voidage -
€
f overall voidage in fluidized beds -
bed voidage at minimum fluidization -
e
P voidage in the fixed bed -
gas viscosity Ns/m 2
A wave length m
ID : - ratio of volume of grid region dead zone to total bed volume -
Pb bulk density of particles kgjm 3
Pg gas density kg/m 3
PP particle density kg/m 3

168
4> ratio U b/U f

m m
<f>b, (j>d fraction of bed volume occupied by bubble, dense phase solids
APg, APD pressure drop across bed and distributor, respectively kN/m 2
Subscripts:
b bubble phase
B bimodal size distribution
d dense phase
9 gas
i i t h particle size interval
mb minimum bubbhng
mf minimum fluidization
N narrow size distribution
o overall
P particles
V void or dilute phase
W wide size distribution

169
References
Abrahamsen, A.R. and Geldart, D . (1980). Behaviour of gas fluidized beds of fine
powders, Powder Technol., 26, 35-55.
Agarwal, J.C., W . L . Davis, and D . T . King, (1962). Fluidized-bed coal dryer, Chem.
Eng. Progr., 58, No. 11, 85-90.
Agarwal, J . C . and Davis, W . L . (1966). The dynamics of fluidization of iron and its
ores, Chem. Eng. Progr. Symp. Ser., No. 67, 62, 101-110.
Avidan, A . and Yerushalmi, J. (1982). Bed expansion in high velocity fluidization,
Powder Technol., 32, 223-232.
Avidan, A . and Edwards, M . , (1986). Modelling and scale-up of Mobil's fluid-bed
M T G process, in Fluidization, ed. K. Ostergaard and A . Sorensen, Eng. Foundation,
New York, pp.457-464. '
Baeyens, J. and Geldart, D. (1974). An investigation into slugging fluidized beds,
Chem. Eng. Sci., 29, 255-265.
Bauer, W . and Werther, J. (1981). Scale-up of fluid bed reactors with respect to size
and gas distributor design - measurements and model calculations, Proceedings of 2nd
World Congress of Chemical Engineering, Montreal, Canada. Vol. 3, pp.69-72.
Behie, L . A . and P. Kehoe, (1973). The grid region in a fluidized bed reactor, AIChE.
J., 19, 1070-1072.
Brereton, C. (1987). Fluid mechanics of high velocity fluidized beds, Ph.D. Thesis,
The University of British Columbia.
Briens, C.L., M . A . Bergougnou and C.G.J. Baker, (1978). Leakage of solids (weeping,
dumping) at the grid of a 0.6 m diameter gas fluidized bed, in Fluidization, ed. J.F.
Davidson and K . L . Keairns, Cambridge University Press, Cambridge, pp.38-43.

170
Briens, C.L., M . A . Bergougnou and C.G.J. Baker, (1980). Grid leakage in gas flu-
idized beds: the effect of bed height, grid thickness, wave breakers, cone-shaped grid
holes and pressure drop fluctuations, in Fluidization, ed. J.R. Grace and J . M . Matsen,
Plenum Press, New York, pp.413-420.
Brown, D.J., E.J. Weatherby and K. Alexander, (1988). Shape, concentration and
anomalous diffraction effects in sizing solids in liquids, in Optical Particle Sizing Theory
and Practice, ed. G . Gouesbet and G . Grehan, Plenum Press, New York, pp.351-362.
Brown, G.W. (1990). Direct measurement of F C C fluidizability helps spot problems,
Oil k Gas J., Jan. 15, 46-49.
Canada, G.S., McLaughlin, M . H . and Staub F . W . (1978). Flow regimes and void
fraction distribution in gas fluidization of large particles in beds without tube banks,
AIChE Symp. Ser. 176, 7 4 , 14-26.
Cankurt N . T . and J. Yerushalmi, (1978). Gas backmixing in high velocity fluidized
beds, in Fluidization, ed. J.F. Davidson and D.L. Keairns, Cambridge University Press,
Cambridge, pp.387-392.
Carberry , J.J. (1976). Chemical and Catalytic Reaction Engineering, McGraw-Hill,

r
New York.
C G C P C Report, (1985). Innovation of an acrylonitrile fluidized bed reactor, China
General Corporation of Petrochemical Industry.
Chavarie, C . and Grace, J.R. (1975). Performance analysis of a fluidized bed reactor,
Ind. Eng. Chem., Fundam., 1 4 , 75-91.
Clift, R., (1986). Hydrodynamics of bubbling fluidized beds, in Gas Fluidization
Technolog}', ed. D . Geldart, Wiley, New York, pp.53-95.
Clift, R. and Grace, J.R. (1972). The coalescence of bubble chains in fluidized beds,
Trans. Instn. Chem. Engs. 5 0 , 364-371.
Clift, R., Grace, J.R., Cheung, L. and Do, T . H . (1972). Gas and solids motion around
171
deformed and interacting bubbles i n fluidized beds, J . F l u i d Mechanics, 5 1 , 187-205.
Clift, R . , Grace, J . R . and Weber, M . E . (1974). Stability of bubbles i n fluidized beds,
Ind. E n g . Chem. Fund., 1 3 , 45-51.
Cooper, D . J . and D . E . Clough, (1985). Experimental tracking of particle-size distri-
bution i n a fluidized bed, Powder Technol., 4 4 , 169-177.
Cooper, D . J . and D . E . Clough, (1986). Optimal, real-time monitoring of particle size
distribution i n a fluidized bed, A I C h E . J . , 3 2 , 389-396.
Danckwerts, P . V . (1953). Continuous flow systems distribution of residence times,
Chem. E n g . Sci., 3 , 1-13.
Darton, R . C , R . D . Lanauze, J . F . Davidson and D . Harrison, (1977). Bubble growth
due to coalescence i n fluidized bed, Tran. Inst. Chem. E n g . 5 5 , 274-280.
Davidson, J . F . , and D . Harrison, (1963). Fluidised Particles, Cambridge University
Press, Cambridge, England.
de Groot, J . H . (1967). Scaling-up of gas-fluidized bed reactors, Proc. Intern. Symp.
on Fluidization, ed. A . A . H . Drinkenburg, Netherlands Univ. Press, Amsterdam, pp.348-
361.
Dry, R . J . , J u d d , M . R . and Shingles, T . (1983). T w o phase theory and fine powders,
Powder Technol., 3 4 , 213-223.
Dry, R . J . and M . R . J u d d , (1985). Fluidised beds of fine, dense powders: scale-up and
reactor modelling, Powder Technol., 4 3 , 41-53.
Edwards, M . and A . A v i d a n , (1986). Conversion model aids scale-up of Mobil's fluid
bed M T G process, Chem. E n g . Sci., 4 1 , 829-835.
Fakhimi, S., S. Sohrabi and D . Harrison, (1983). Entrance effects at a multi-orifice
distributor i n gas-fluidized beds, C a n . J . Chem. E n g . 6 1 , 364-369.
Fan, L . T . , Tho-ching H o , S. Hiraoka and W . P . Walawender, (1981). Pressure fluctu-
ations i n a fluidized bed, A I C h E J . 2 7 , 388-396.

172
Fan, L . T . , Tho-ching Ho and W.P. Walawender, (1983). Measurements of the rise
velocities of bubbles, slugs and pressure waves in a gas-solid fluidized bed using pressure
fluctuation signals, AIChE. J., 2 9 , 33-39.
Fenton, D.L. and J.J. Stukel, (1976). Measurement of the local particle concentration
in fully turbulent duct flow, Int. J . Multiphase Flow, 3 , 141-145.
Frye, C . G . , W . C . Lake and H.C. Eckstrom, (1958). Gas-solid contacting with ozone
decomposition reaction, AIChE. J., 4 , 403-408.
Furusaki, S., T. Kikuchi and T. Miyauchi, (1976). Axial distribution of reactivity
inside a fluid bed reactor, AIChE. J., 2 2 , 354-359.
Garcia, A . , J.R. Grace and R. Clift, (1973). Behaviour of gas bubbles in fluidized
beds, AIChE. J., 1 9 , 369-370.
Geldart, D. (1973). Types of gas fluidization, Powder Technol, 7 , 285-292.
Geldart, D. (1986). Gas Fluidization Technology, John Wiley k Sons Ltd., New York.
Geldart, D. and A.R. Abrahamsen. (1980). The effect of fines on the behaviour
of fluidized beds of small particles, in Fluidization, ed. J.R. Grace and J . M . Matsen,
Plenum Press, New York, pp.453-460.
Geldart, D. and D.J. Pope, (1983). Interaction of fine and coarse particles in the
freeboard of a fluidized bed, Powder Technol., 3 4 95-97.
Geldart, D. and A . C . Wong, (1984). Fluidization of powders showing degrees of
cohesiveness-I. bed expansion, Chem. Eng. Sci., 3 9 , 1481-1488.
Geldart, D. and A . C . Wong, (1985). Fluidization of powders showing degrees of
cohesiveness, Chem. Eng. Sci., 4 0 , 653-661.
Geldart, D. and A . L . Radtke, (1986). The effect of particle properties on the be-
haviour of equihbrium cracking catalysts in standpipe flow, Powder Technol., 4 7 , 157-
165.
173
Geldart, D. and M.J. Rhodes (1986). From minimum fluidization to pneumatic trans-
port, in Circulating Fluidized Bed Technology, ed. F. Basu, Pergamon Press, pp.21-31.
Geldart, D. and B. Buczek. (1989). The effect of the size of the fines on the fluidization
behaviour of equihbrium cracking catalyst, in Fluidization, ed. J.R. Grace, L.W. Shemilt
and M . A . Bergougnou, Engineering Foundation, New York, pp.179-186.
George S.E. and J.R. Grace, (1978). Entrainment of particles from aggregative flu-
idized beds, AIChE Symp. Ser., No. 176, Vol. 74, 67-74.
Grace, J.R. (1970). The viscosity of fluidized beds, Can. J. Chem. Eng., 48, 30-33.
Grace, J.R. (1981). Fluidized bed reactor modeling: an overview, American Chemical
Society, Symp. Ser., No. 168, 3-18.
Grace, J.R. (1982). Fluidized-bed hydrodynamics, Chapter 8.1 in Handbook of Mul-
tiphase Systems, ed. G . Hetsroni, Hemisphere, Washington, pp.8-5 - 8-62.
Grace, J.R. (1984). Generalized models for isothermal fluidized bed reactors, Chapter
13 in Recent Advances in Engineering Analysis of Chemically Reacting Systems, ed. L . K .
Doraiswamy, Wiley Eastern, New Delhi, pp.237-255.
Grace, J.R. (1986a). Fluid beds as chemical reactors, Chapter 11 in Gas Fluidization
Technology, ed. D . Geldart, Wiley, New York, pp.287-341.
Grace, J.R. (1986b) Contacting modes and behaviour classification of gas-solid and
other two-phase suspensions, Can. J. Chem. Eng., 64, 353-361.
Grace, J.R. and R. Clift, (1974). On the two-phase theory of fluidization, Chem.
Eng. Sci., 29, 327-334.
Grace, J.R. and H.I. de Lasa, (1978). Reaction near the grid in fluidized beds, AIChE.
J., 24, 364-366.
Grace, J.R. and D. Harrison, (1969). The behaviour of freely bubbhng fluidized beds,
Chem. Eng. Sci., 24, 497-508.
Guo, F. (1987). Gasflowand mixing behaviour in fine-powder fluidized bed, AIChE

174
J., 33, 1895-1898.
Harrison, D., J.F. Davidson and J.W. de Kock, (1961). On the nature of aggregative
and particulate fluidisation, Trans. Instn Chem. Engrs. 39, 202-211.
Hartge, E.-U., Y. Li, and J. Werther, (1986). Flow structures in fast fluidized beds,
in Fluidization, ed K. Ostergaard and A. Sorensen, Engng. Foundation, New York,
pp.345-352.
Haultain, H.E. (1937). Splitting the minus-200 with the superpanner and infrasizer,
The Canadian Institute of Mining and Metallurgy, pp.229-240.
Hayashi, S. (1988). Measurements of absolute concentration and size distribution of
particles by laser small angle scattering, in Optical Particle Sizing Theory and Practice,
ed. G . Gouesbet and G . Grehan, Plenum Press, New York, pp.549-558.
Herb, B., K. Tuzla and J.C. Chen, (1989). Distribution of sohd concentrations in cir-
culating fluidized bed, in Fluidization, ed. J.R. Grace, L . M . Shemilt and M . A . Bergoug-
nou, Engng. Foundation, New York, pp.65-72.
Hiby, J.W., (1964). Untersuchungen iiber den kritischen mindestdruckverlust des
antrombodens bei fluidalbetten, Chem. Ingr. Tech., 36, 228-229.
Hiraki, I., Yoshida, K. and Kunii, D . (1966). Behaviour of bubbles in a two- dimen-
sional fluidized bed, Kagaku Kogaku, Vol. 4, No. 1, 139-142.
Ho, T - C , T . K . Chen and J.R. Hopper, (1984). Pressure drop across the distributor
in fluidized beds with regular and irregular distributor design, AIChE Symposium Ser.,
No. 241, Vol. 80, 34-40.
Ho, T . C . , R.C. Wang and J.R. Hopper, (1987). Characteristics of grid zone heat
transfer in a gas-solid fluidized bed, AIChE. J., 33, 843-847.
Hodkinson, J.R. (1966). The optical measurement of aerosols, in Aerosol Science, ed.
Davies, C.N., Academic Press, New York, pp.931-932.
Hogg, R.V. and Ledolter, J. (1987). Engineering Statistics, Macmillan Publishing

175
Company, New York.
Horio, M . , M . Nishimuro, and H . Ishii, (1989). On the nature of "turbulent fluidized
beds", Proceedings of the 3rd SCEJ Symp. on Circulating Fluidized Beds, Tokyo, Japan,
54-61.
Horio, M . and A. Nonaka, (1987). A generalized bubble diameter correlation for
gas-solid fluidized beds, AIChE. J., 33 1865-1872.
Horvath, M . , L. Bilitzky and J . Huttner, (1985). Ozone, Elsevier Science Pubhshing
Co., New York.
Hovmand, S. and J.F. Davidson, (1968). Chemical conversion in a slugging fluidized
bed, Trans. Instn. Chem. Engrs., 46, 190-203.
Hovmand, S., Freedman, W. and Davidson, J.F. (1971). Chemical reaction in a
pilot-scale fluidized bed, Trans. Instn. Chem. Engrs, 49, 149-162.
Hsiung, T.P. and J.R. Grace, (1978). Formation of bubbles at an orifice in fluidized
beds, in Fluidization, ed. J.F. Davidson and D.L. Keairns, Cambridge University Press,
Cambridge, England, pp.19-24.
Ip, T . (1988). Influence of particle size distribution on fluidized bed hydrodynamics,
M.A.Sc. Thesis, The University of British Columbia.
Jin, Y., Yu, Z., Wang, Z. and Cai, P. (1986). A criterion for transition from bubbling
to turbulent fluidization, in Fluidization, ed, K. Ostergaard and A . Sorensen, Engng.
Foundition, New York, 289-296.
Johnsson, J.E., J.R. Grace and J.J. Graham, (1987). Fluidized-bed reactor model
verification on a reactor of industrial scale, AIChE. J., 33, 619-627.
Judd, M.R. and Goosen, R. (1989). Effects of particle shape on fluidization character-
istics of fine particles in freely bubbling and turbulent regimes, in Fluidization, ed. J.R.
Grace, L.W. Shemilt and M . A . Bergougnou, Engng. Foundation, New York, pp.41-48.
Kai, T . and Furusaki, S. (1985). Behaviour of fluidized beds of small particles at

176
elevated temperatures, J. Chem. Eng. Japan, 18, 113-118.
Kai, T . , Shirakawa, Y., Takahashi, T . and Furusaki, S. (1987). "Change in bubble
behaviour for different fluidizing gases in a fluidized bed", Powder Technol., 51, 267-271.
Kato, K. and C.Y. Wen, (1969). Bubble assemblage model for fluidized bed catalytic
reactors, Chem. Eng. Sci., 24, 1351-1369.
Kehoe, P.W.K. and Davidson, J.R. (1971). Continuously slugging fluidized beds, Inst.
Chem. Eng. Symp. Ser., 33, 97-116.
Kehoe, P.W.K. and J.R. Davidson, (1973). Pressure fluctuations in slugging fluidized
beds, AIChE Symp. Ser., No. 128, Vol. 69, 34-40.
Khoe, G.K., Ip, T . and Grace, J.R. (1991). Rheological and fluidization behaviour of
powders of different particle size distribution, Powder Technol., in press.
Kobayashi, H . , Arai, F. and Chiba, T . (1966). Behaviour of bubbles in a gas-solid
fluidized bed, Kagaku Kogaku, Vol. 4, No. 1, 147-150.
Kono, H.O., Chiba, S., Ells, T . and Suzuki, M . (1986a). Characterization of emulsion
phase in fine particle fluidized beds, Powder Technol., 48, 51-58.
Kono, H.O., S. Chiba, T . Ells, P. Daniell and M . Suzuki, (1986b). The effect of emul-
sion phase characteristics on Fluidization - selection of powder properties, in Fluidization,
ed. K . Ostergaard and A. Sorensen, Engng. Foundation, New York, pp. 143-150.
Krambeck, F.J., A . A . Avidan, C.K. Lee and M . N . Lo, (1987). Predicting fluid-bed
reactor efficiency using absorbing gas tracers, AIChE. J., 33, 1727-1734.
Kunii, D. and 0. Levenspiel, (1969). Fluidization Engineering, John Wiley & Sons,
New York.
Kurz, H.P. and Miinz, G . (1975). The influence of particle size distribution on the
flow properties of limestone powders, Powder Technol., 11, 37-40.
Lanneau, K.P. (1960). Gas-solids contacting in fluidized beds, Trans. Instn. Chem.
Engrs, 38, 125-143.

177
Lee, G.S. and Kim, S.D. (1988). Pressure fluctuations in turbulent fluidized beds, J.
Chem. Eng. Japan, 2 1 , 515-521.
Levenspiel, 0. (1972). Chemical Reaction Engineering, 2nd ed. Wiley, New York.
Lewis, M . K . , Gilhland, E.R. and Glass, M . (1959). Solid-catalyzed reaction in a
fluidized bed, AIChE. J., 5, 419-426.
Li, Y . and M . Kwauk, (1980). The dynamics of fast fluidization, in Fluidization, ed.
J.R. Grace and J . M . Matsen, Plenum Press, New York, pp.537-544.
Lirag, R . C , and H. Littman, (1971). Statistical study of the pressure fluctuations in
a fluidized bed, AIChE Symp. Ser. No. 116, Vol. 67, pp.11-22.
Lloyd, P.J. and P.J. Webb, (1987). The flooding of a powder - the importance of
particle size distribution, Powder Technol., 51, 125-133.
Lockett, J. and Harrison, D. (1967). The distribution of voidage fraction near bub-
bles rising in gas-fluidized beds, Proc. Intern. Symp. on Fluidization, ed. A . A . H .
Drinkenburg, Netherlands Univ. Press, Amsterdam, 257-270.
Massimilla, L. (1973). Behaviour of catalytic beds of fine particles at high gas veloc-
ities, AIChE Symp. Ser. Vol. 69, No. 128, 11-15.
Matsen, J . M . (1973). Evidence of maximum stable bubble size in a fluidized bed,
AIChE Symp. Ser. No. 128, Vol. 69, 30-33.
Matheson, G.L., W . A . Herbst and P.H. Holt, (1949). Characteristics of fluid-solid
systems, Industrial and Engineering Chemistry, 4 1 , 1099-1104.
May, W . G . (1959). Fluidized-bed reactor studies, Chem. Eng. Prog. 55(12), 49-56.
Mears, D . E . (1971a). The role of axial dispersion in trickle-flow laboratory reactors,
Chem. Eng. Sci., 26, 1361-1366.
Mears, D . E . (1971b). Tests for transport limitations in experimental catalytic reac-
tors, I&EC. Process Des. Develop., 10, 541-547.
Monceaux, L., M . Azzi, Y. Molodtsof and J.F. Large, (1986). Particle mass flux
178
profiles and flow regime characterization in a pilot-scale fast fluidized bed unit, in Flu-
idization, ed. K. Ostergaard and A. Sorensen, Eng. Foundation, New York, pp.337-344.
Mori, S. and C.Y. Wen, (1975). Estimation of bubble diameter in gaseous fluidized
beds, AIChE. J., 21, 109-115.
Morse, R.D. and Ballou, C O . (1951). The uniformity of fluidization - its measurement
and use, Chem. Eng. Progr. 47, 199-204.
Nicastro M . T . and L.R. Glicksman, (1984). Experimental verification of scaling rela-
tionships for fluidized bed, Chem. Eng. Sci., 39, 1381-1391.
Oki, K., M . Ishida, and T . Shirai, (1980). The behaviour of jets and particles near
the gas distributor grid in the three-dimensional fluidized bed, in Fluidization, ed. J.R.
Grace and J . M . Matsen, Plenum Press, New York, pp.421-428.
Orccutt, J . C , Davidson, J.F, and Pigford, R.L. (1962). Reaction time distributions
in fluidized catalytic reactors, Chem. Eng. Prog. Symp. Ser. 58(38), 1-5.
Patrose B. and H.S. Caram, (1984). The mechanics of particle motion in a grid jet,
AIChE Symp. Ser. No. 241, Vol. 80, 48-56.
Pell, M . and S.P. Jordan, (1988). Effects of fines and velocity on fluid bed reactor
performance, AIChE Symp. Ser. No. 262, Vol. 84, 68-73.
Rase, H.F. (1977). Chemical Reaction Design for Process Plants, Vol. 1, John Wiley
& Sons, New York.
Rathbone, R.R. and J.F. Davidson, (1986). Design of distributors for air gasification
of char in fluidized bed, in Fluidization, ed. K. Ostergaard and A . Sorensen, Engng.
Foundation, New York, pp.63-70.
Raterman, M . F . (1985). F C C catalyst flow-problem predictions, Oil & Gas J. Tech-
nology, Jan. 7, 87-92.
Razumovskii, S.D. and G.E. Zaikov, (1984). Ozone and its Reactions with Organic
Compounds, Elsevier Science Pub., New York.

179
Rhodes, M.J. and D. Geldart, (1986). Transition to turbulence?, in Fluidization, ed.
K. Ostergaard and A. Sorensen, Eng. Foundation, New York, pp.281-288.
Rhodes, M.J., T . Hirama, G . Cerutti and D . Geldart, (1989). Non-uniformities of
solids flow in the risers of circulating fluidized beds, in Fluidization, ed. J.R. Grace,
L.W. Shemilt and M.A. Bergougnou, Engng. Foundation, New York, pp.73-80.
Rietema, K. (1967). Application of mechanical stress theory to fluidization, Proc.
Intern. Symp. on Fluidization, ed. A . A . H . Drinkenburg, Netherlands Univ. Press,
Amsterdam, pp.154-175.
Rowe, P.N. (1976). Prediction of bubble size in a gas-fiuidized bed, Chem. Eng. Sci.,
31, 285-288.
Rowe, P.N. and D.J. Everett, (1972). Fluidised bed bubbles viewed by X-rays, Trans.
Instn. Chem. Engrs., 50, 43-60.
Rowe, P.N. and H.J. MacGilivray, (1980). The structure of a 15 cm diameter gas
fluidised bed at up to 1 m/s and seen by X-rays, in Fluidization, ed. J.R. Grace and
J . M . Matsen, Plenum Press, New York, 545-553.
Sadasivan, N., Barreteau, D . and Laguerie, C. (1980). Studies on frequency and
magnitude of fluctuations of pressure drop in gas-sohd fluidized beds, Powder Technol.,
2 6 , 67-74.
Satija, S. and L-S. Fan, (1985). Characteristics of slugging regime and transition to
turbulent regime for fluidized beds of large coarse particles, AIChE. J., 3 1 , 1554-1562.
Schmidt, H . , C. Tsouris, E . Eggert, L.L. Tavlarides, (1989). Laser photometric probe
for concentration measurements in liquid dispersions, AIChE. J., 35, 507-510.
Schnitzlein, M . G . and H . Weinstein, (1988). Flow characterization in high-velocity
fluidized beds using pressure fluctuations, Chem. Eng. Sci., 43, 2605-2614.
Sinister, W.W., and P. Kishak, (1952). The measurement of fluidization quality,
Chem. Eng. Prog., 48, 455-460.

180
Sinnott, R.K. (1983). An Introduction to Chemical Engineering Design, Vol. 6 in
Chemical Engineering, ed. J . M . Coulson and J.F. Richardson, Pergamon Press, Oxford,
pp.354-360.
Sit, S-P., and J.R. Grace, (1981). Effect of bubble interaction on interphase mass
transfer in fluidized beds, Chem. Eng. Sci., 36, 327-335.
Sit, S-P., and J.R. Grace, (1986). Interphase mass transfer during bubble formation
in fluidized beds, in Fluidization, ed. K . Ostergaard and A. Sorensen, Eng. Foundation,
New York, pp.39-46.
Sun, G . , Wang, Z. and Chen, G . T . (1983). The incipient bubbling velocity and
voidage of gas-solid particle systems, JI. Zhejiang University, China, 4, 97-108.
Sun, G . and Chen, G.T. (1986). Development of novel reacting flow technology in
fluidization systems, A Report to China General Corporation of Petrochemical Industry.
Sun, G. and Chen, G . T . (1989). Transition to turbulent fluidization and its prediction,
in Fluidization, ed. J.R. Grace, L.W. Shemilt and M.A. Bergougnou, Eng. Foundation,
New York, pp.33-41.
Thiel, J. and Potter, O.E. (1977). Slugging in fluidized beds, Ind. Eng. Chem.
Fundam., 16, 242-247.
Tien, C.L. (1985). Radiation properties of particulates, in Handbook of Heat Transfer
Fundamentals, 2nd ed. Hemisphere, New York, pp.14.83-14.91.
Tien, C.L. (1988). Thermal radiation in packed and fluidized beds, A S M E JI. of Heat
Transfer, 110, 1230-1242.
Toei, R., Matsuno, R., Kojima, H . , Nagai, Y., Nakagawa, K. and Yu, S. (1966).
Behaviour of bubbles in the gas-solid fluidized bed, Kagaku Kogaku, Vol. 4, No. 1,
142-146.
Tung, Y., Z. Yang, Y. Xia, W. Zhang, Y. Yang and M . Kwauk, (1989). Assessing
fluidizing characteristics of powders, in Fluidization, ed. J.R. Grace, L . W . Shemilt and

181
M.A. Bergougnou, Engng. Foundation, New York, pp.169-178.
Van Deemter, J.J. (1961). Mixing and contacting in gas-solid fluidized beds, Chem.
Eng. Sci., 13, 143-154.
Van Deemter, J.J. (1967). The counter-current flow model of a gas-solids fluidized
bed, Proceedings Intern. Symp. on Fluidization, ed. A . A . H . Drinkenburg, Netherlands
University Press, Amsterdam, pp.334-347.
Van den Aarsen, F . G . , A . A . C M . Beenackers and W.P.M. van Swaaij, (1986). Bubble-
to-dense-phase mass transfer kinetics in gas-solid fluidized beds, in Heat and Mass Trans-
fer in Fixed and Fluidized Beds, ed. W.P.M. van Swaaij and N.H. Afgan, Hemisphere,
Washington, pp.397-415.
Van Swaaij, W.P.M. and F.J. Zuiderweg, (1972). Investigation of ozone decomposition
in fluidized beds on the basis of a two-phase model, in Proceedings of 5th European/ 2nd
International Symposium on Chem. Reaction Engineering, Amsterdam, B 9-25 - B 9-36.
Virr, M.J. and H.W. Williams, (1985). Heat recovery by shallow fluidized beds, Chem.
Eng. Prog., 81(7), 50-56.
Walker, B.V., (1970). Gas-solid contacting in bubbhng fluidized beds, Ph.D. Thesis,
Cambridge University, Cambridge, England.
Warren, H.V., C. Madsen and W . H . White, (1939). Some studies with the Haultain
superpanner and infrasizer, Canadian Institute of Mining and Metallurgy, pp.53-56.
Wehner, J.F. and Wilhelm, R.H. (1956). Boundary conditions offlowreactor, Chem.
Eng. Sci., 6 , 89-93.
Weimer, A . W . , D . C . Gyure and D . E . Clough, (1985). Apphcation of a Gamma- radi-
ation density gauge for determining hydrodynamic properties of fluidized beds, Powder
Technol, 44, 179-194.
Weinstein, H . , M . Shao, M . Schnitzlein and R.A. Graff, (1986). Radial variation in
void fraction in a fast fluidized bed, in Fluidization, ed. K. Ostergaard amd A. Sorensen,
182
Engng. Foundation, New York, pp.329-336.
Weinstein, H. and J. Li, (1989). An evaluation of the actual density in the acceleration
section of vertical risers, Powder Technol., 57, 77-79.
Weinstein, H., J. Li, E . Bandlamudi, H.J. Feindt and R.A. Graff, (1989). Gas back-
mixing of fluidized beds in different regimes and regions, in Fluidization, ed. J.R. Grace,
L.W. Shemilt and M . A . Bergougnou, Eng. Foundation, New York, pp.57-64.
Wen, C.Y. and Chen, L . H . (1982). Fluidized bed freeboard phenomena; entrainment
and elutriation, AIChE. J., 28, 117-128.
Wen, C.Y., R. Krishnan, R. Khosravi and S. Dutta, (1978). Dead zone heights near
the grid offluidizedbeds, in Fluidization, ed. J.F. Davidson and D.L. Keairns, Cambridge
University Press, Cambridge, England, pp.32-37.
Wen, C.Y., R. Krishnan, and R. Kalyanaraman, (1980). Particle mixing near the
grid region of fluidized beds, in Fluidization, ed. J.R. Grace and J . M . Matsen, Plenum
Press, New York, pp.405-412.
Werther, J. (1978). Effect of gas distributor on the hydrodynamics of gas fluidized
beds, Ger. Chem. Eng. 1, 166-174.
Werther, J. (1980). Mathematical modelling of fluidized bed reactors, Int. Chem.
Eng., 20, 529-541.
Werther, J. (1984). Hydrodynamics and mass transfer between the bubble and emul-
sion phases in fluidized beds of sand and cracking catalyst, in Fluidization, ed. D. Kunii
and R. Toei, Eng. Foundation, New York, pp.93-101.
Werther, J. and M . Schoessler, (1986). Modeling catalytic reactions in bubbling
fluidized beds of fine particles, in Heat and Mass Transfer in Fixed and Fluidized Beds,
ed. W.P.M. van Swaaij and N.H. Afgan, Hemisphere Washington, pp.355-370.
Whitehead, A.B. and D . C . Dent, (1967). Behaviour of multiple tuyere assemblies in

183
large fluidized beds, in Proceedings Intern. Symp. on Fluidization, ed. A . A . H . Drinken-
burg, Netherlands University Press, Amsterdam, pp.802-820.
Winter, 0. (1968). Density and pressure fluctuations in gas fluidized beds, AIChE.
J., 14, 426-432.
Yacono, C . and H . Angelino, (1978). The influence of gas distributor on bubble be-
haviour: comparison between "ball distributor" and "porous distributor", in Fluidization,
ed. J.F. Davidson and D.L. Keairns, Cambridge University Press, Cambridge, England,
pp.25-31.
Yang, G . , Z. Huang and L. Zhao, (1984a). Radial gas dispersion in a fastfluidizedbed,
in Fluidization, ed. D. Kunii and R. Toei, Engng. Foundation, New York, pp.145-152.
Yang, W . C . and Chitester, D . C . (1988). Transition between bubbhng and turbulent
fluidization at elevated pressure, AIChE Symp. Ser. No. 262, Vol. 84, 10-21.
Yang, X., D . G . Home, J . G . Yates and P.N. Rowe, (1984b). Distributor-zone reaction
in a gas-fluidized bed, AIChE Symp. Ser. No. 241, Vol. 80, 41-47.
Yates, J . G . and Newton, D . (1986). Fine particle effects in a fluidized bed reactor,
Chem. Eng. Sci., 41, 801-806.
Yerushalmi, J. (1982). Applications of fluidized beds, Chapter 8.5 in Handbook of
Multiphase Systems, ed. G. Hetsroni, Hemisphere, Washington, pp.8-152 - 8-214.
Yerushalmi, J. and A. Avidan, (1985). High velocity fluidization, in Fluidization, ed.
J.F. Davidson, R. Clift and D. Harrison, Academic Press, London, pp.226-293.
Yerushalmi, J., Cankurt, N . T . , Geldart, D. and Liss, B. (1978). Flow regimes in
vertical gas-sohd contact systems, AIChE Symp. Ser., Vol. 74, No. 176, 1-13.
Yerushalmi, J. and N.T. Cankurt, (1979). Further studies of the regimes of fluidiza-
tion, Powder Technol., 24, 187-205.
Yerushalmi, J. (1986). High velocity fluidized beds, Chapter 7 in Gas Fluidization
Technology, ed. D. Geldart, Wiley, New York, pp.155-196.

184
Zenz, F . A . and Othmer, D.F. (1960). Fluidization and Fluid-Particle Systems, Rein-
hold Pubhshing Corp., New York.
Zuiderweg, F.J. (1967). "Session report" in Proceedings of Intern. Symp. on Fluidiza-
tion, ed. A . A . H . Drinkenburg, Netherlands University Press, Amsterdam, pp.739-750.

Appendix A
Minimum integral reactor length to minimize axial diffusion effects
For fixed bed reactors, it is generally considered that backmixing effects can be negligible
if the ratio of reactor bed length to equivalent particle diameter is more than 50 for
gaseous reactions (Carberry, 1976). A more stringent criterion should, however, be used
for laboratory fixed bed reactors, or integral reactors, because of the small particle size
used, except for differential reactors where disperion dose not affect the measurement
results. This criterion can be derived by combining the solution of the continuity equation
including the axial diffusion term with the solution for plug flow. A minimum reactor
length required to maintain the deviation from plug flow at less than 5% was obtained
(Mears, 1971a), giving
H 20n C
Aiin
T =
~P7 l n
r— (A>1)
where n is the reaction order, i.e. n — 1 in our experiments. Pe p is the axial Peclet
number, corresponding to the particle size,
Pe
r - ^ < >
A2
The effective axial dispersion coefficient, D , can be estimated after the Reynolds
e
number and Schmidt number have been calculated (Levenspiel, 1972). The calcula-
tion was based on the harshest conditions, that is, the lowest superficial gas velocity
(U = 0.089 m/s) and the highest reaction conversion (a; = 0.99) in the fixed bed during
the experiments. Other typical operating conditions used were:
185
Appendix A. Minimum integral reactor length to minimize axial diffusion effects 186
operating temperature T = 20 ° C ;
operating pressure P = 0.15 bar;
bed voidage e = 0.46.

p
Hence
._ p 9 1.8 x 10- 5
5 c
- - 1.2 x 0.204 x 1 0 - ~ ° - 7 3 5 ( A
- 3 )
and
_ dpUpg 60 x l O " x 0.089 x 1.2 6
n o c c . .
i2e = ^ £ =
p _ _ _ — = 0.355 (A.4)
Pg 1.8 X l O " 5
while these correspond (Levenspiel, 1972) to:
De e p
0.7 (A.5)
That is,
P e
' = ij
§- =
h = a 6 5 7 ( A
- 6 )
Hence, from equation A . l , we can obtain:
=- > 140 (A.7)

dr.
In our experiments, the height of fixed bed H > 40 mm, so that
H
-3 > 600 (A.8)
the axial dispersion effects, therefore, was negligible.

Appendix B
Photometric measurement for spent F C C particles
187
endix B. Photometric measurement for spent FCC particles
U = 0-1 m/s is
0-2 m/s
0-4 m/s
0.6 m/s
/W
0-8 m/s
0-9 m/s
1^, = 800 -
1^ = 2400
W W
Figure B . l : Photometric measurement for S-FCC particles with narrow P

Appendix B. Photometric measurement for spent FCC particles 189
,1s
U =
0-1 m/s
JiL L
0.2 m/s
lo/| =800- —
0-4 m/s
Io/j=2400- —
0-6 m/s
0-8 m/s
Figure B.2: Photometric measurement for S-FCC particles with wide PSD
pendix B. Photometric measurement for spent FCC particles
U = 0-1 m / s
0-2 m/s
lr/V
0-4 m/s
0 . 6 m/s
0-8 m/s
1^ = 8 0 0 -
1^=2400-
Figure B.3: Photometric measurement for S-FCC particles with bimodal PSD
Appendix C
S t a t i s t i c a l A n a l y s i s of E x p e r i m e n t a l Results
C.l Significance test of the influence of the PSD
T h e significance of the effect of the P S D on the performance of fluidized bed reactors
at certain operating conditions (e.g. same U, k , W T cat and d etc.)

p has been examined
using the //'-test based on the assumption of randomization (Hogg and Ledolter, 1987).
T w o type sources of variation may occur, the "between" and the " w i t h i n " , originating
from the influence of the P S D and from the experimental error, respectively. The ratio
of the mean squares, F, can be expressed as
F = MSPSD E- iMFi-F) /(fe-i) 2
(C.l)
= =
MS oR
ERR E t i E£i(*y - * ) / ( Y 2 N
-)
k
(i = 1,2, j = 1,2, . . . , Ti{)
where
Yi = -EYii (C2)
n i
i=i
7
4 Ei=ij=i
E^
i V
(C3)
N = J2 n i (CA)
i=l
and is the experimental value (e.g. reaction conversion) obtained i n the j t h trail for
the i t h P S D .
191
Appendix C. Statistical Analysis of Experimental Results 192
Table C . l : Reaction conversions measured in the fluidized bed reactor for particles of
different size distributions; U = 0.5 m/s, k = 9 s , catalyst inventory = 5 kg.
r
- 1
PSD X
i,j ^2j=l{ i,j
X X
i ) 2
wide 0.875 0.882 0.870 0.893 0.886 0.881 0.000327
narrow 0.802 0.811 0.805 0.791 0.803 0.802 0.000211
bimodal 0.858 0.832 0.850 0.845 0.849 0.847 0.000363
The above mean square (or F) ratio has an F distribution with (fe — 1) and (N — fc)
degrees of freedom. If F > F(a;k — 1, JV — fe), we accept the hypothesis that there is
a significant influence of the PSD on the reactor performance. On the other hand, if the
F-ratio is smaller than the critical value, then there is not enough evidence in the data to
reject the null hypothesis, i.e. we must consider the effect of the PSD to be insignificant.
As typical example, consider a set of experimental results obtained under typical
operating conditions for three different particle size distributions shown in Table C . l .
In this case,fe= 3 and n; = 5, i.e. there were three size distributions and five measured
conversions for each size distribution. According to eqns. C . l to C.4, F = ,î SpsD
=
104.4. It can be seen from the F distribution table, F (0.01, 2, 12) = 6.93 < F. Thus,
the effect of PSD on the reactor performance at these operating conditions is significant
at significance level a = 0.01, i.e. at the 99% probability level.
If we compare the results of the bimodal PSD with those of the wide PSD, similar con-
clusions are obtained. The F-ratio calculated is 33.5, while F(0.01, 1, 8) = 11.26 < F.
This means that for the operating conditions of Table C . l the difference between reaction
conversions for the two particle size distributions were again significant at a significance
level a = 0.01. In most cases (U > 0.2 m/s), it is possible, as in the example given here,
to shown that the differences between the results for the three particle size distributions
is statistically significant.
C.2 Confidence intervals of experimental results
For a random sample x x ,... 1: 2 ,x n arises from a distribution with unknown mean p and
variance a 2
> 0, the computed 100(1 — a) percent confidence interval can be expressed
(Hogg and Ledolter, 1987) as:
5 ±
*(?)4r ( c
- 5 )
where x is the sample mean and the probability (1 — a) is called the confidence coeffi-
cient. If the population mean p is unknown, in many situations, the population standard
deviation a is also unknown. In this case, if the sample size is quite large (> 30), the
sample standard deviation can be taken as tr. If the sample size is small, however, a
100(1 — a) percent confidence interval for p can be expressed by
z±f(f;n-l)-^=, (C.6)
provided that the random sample arises from the normal distribution. Here t is the
"student r" tabulated in most statistics texts.
The confidence intervals for our experimental results were examined against eqn. C.6.
As an example, conversion data for the bimodal particle size distribution in Table C l
were analyzed to evaluate the. confidence interval.
In this case, n = 5, x = 0.847, and £ ( 0 . 0 5 / 2 , 4 ) = 2.776, while

t
s = Jî~' i)a
= 0.00953
Therefore, a 95 percent confidence interval for the mean of the reaction conversion
for the bimodal PSD in Table C . l is 0.847 ± 0.012. The interval breadth is of the same
order as the height of the symbols used on the graphs. Note, however, that the different
points used to determine the variance are from the same run, with the points taken at
different times. In tables D . l to D.6, results are shown for six conditions that were totally
repeated for U = 0.1 and 0.11 m/s. These show that the reproducibility is nearly as good
for complete repeats as for different points taken during the same experiments.
Appendix D
R a w D a t a f r o m C h e m i c a l Reaction Tests
195
Appendix D. Raw Data from Chemical Reaction Tests 196
Table D . l : Raw conversion data of ozone decomposition for particles of wide size distri-
bution
operating condition Av=2.155-\ VF CQt =l.lkg . k =2.25s-\

r W' =2.75kg
CQt
U, m/s mean standard deviation mean standard deviation

0.10 0.528 0.006 0.736 0.007
0.21 0.351 0.004 0.602 0.006
0.31 0.290 0.005 0.518 0.009
0.53 0.208 0.004 0.404 0.002
0.77 0.306 0.005
0.10 0.533 0.005 0.729 0.004
Table D.2: Raw conversion data of ozone decomposition for particles of narrow size
distribution
operating condition k =2.128~\ W = l.lkg

r cat k =2A7s-\
T iy =2.75kg
cat

0.10 0.455 0.005 0.686 0.006
0.21 0.300 0.002 0.466 0.008
0.31 0.225 0.003 0.392 0.005
0.53 6.153 0.004 0.298 0.004
0.77 0.237 0.002
0.10 0.450 0.006 0.682 0.005
Table D.3: Raw conversion data of ozone decomposition for particles of bimodal size
distribution with coarse fraction active
operating condition Av=2-.21*-\ W = l . l k g

cat k =2.25s-\
T M/ =2.75kg
CQt

0.10 0.476 0.006 0.733 0.007
0.21 0.338 0.003 0.549 0.008
0.31 0.264 0.004 0.447 0.005
0.53 0.185 0.005 0.345 0.006
0.77 0.270 0.003
0.10 0.490 0.005 0.724 0.006
distribution with fine fraction active
operating condition k =2.16s-\

r W =l.lkg cat k =2.20s-\
r W =2.75kg
caf
U, m/s mean standard devaition mean standard deviation

0.10 0.513 0.007 0.740 0.009
0.21 0.371 0.003 0.562 0.006
0.31 0.295 0.005 0.485 0.005
0.53 0.222 0.004 0.391 0.008
0.77 0.259 0.003
0.10 0.518 0.006 0.728 0.010
A p p e n d i x D. Raw Data from Chemical Reaction Tests 198
distribution with different fraction active; catalyst inventory: 5.05 kg
active both fractions coarse fraction fine fraction

condition 4.03s' 1
K= 4 . 1 4 s -
1
K= 4 . 1 0 s -
1
U, m / s mean deviation mean deviation mean deviation

0.11 0.911 0.009 0.897 0.006 0.906 0.011
0.31 0.844 0.007 0,814 0.007 0.848 0.005
0.51 0.767 0.006 0.714 0.004 0.783 0.005
0.71 0.689 0.008 0.663 0.006 0.712 0.006
0.91 0.585 0.007 0.600 0.003 0.628 0.008
1.10 0.512 0.005 0.547 0.010 0.539 0.002
1.35 0.445 0.003 0.495 0.002 0.460 0.004
1.70 0.406 0.004 0.431 0.004 0.398 0.003
0.11 0.902 0.008 0.908 0.006 0.914 0.008
Table D.6: R a w conversion data of ozone decomposition for particles of different size
distributions; catalyst inventory: 5.05kg
PSD Wl de {kr=4.27s- ) 1
narrow (& =4.20s )
r
A
U, m / s mean standard deviation mean standard deviation

0.11 0.915 0.009 0.900 0.008
0.31 0.875 0.005 0.798 0.007
0.51 0.822 0.008 0.676 0.004
0.71 0.789 0.004 0.623 0.007
0.91 0.724 0.011 0.556 0.003
1.10 0.655 0.003 0.508 0.002
1.35 0.580 0.005 0.455 0.006
1.70 0.486 0.006 0.410 0.004
0.11 0.904 0.007 0.895 0.010
Table D.7: Raw conversion data of ozone decomposition for particles of wide size distri-
bution with different activities; catalyst inventory: 5 kg
condition k = 2.41s"1
T k = 4.46s- 1
T k = 8.95s- 1
T
U, m/s mean deviation mean deviation mean deviation

0.10 0.838 0.008 0.923 0.013 0.940 0.007
0.30 0.741 0.009 0.872 0.006 0.911 0.010
0.50 0.622 0.004 0.846 0.003 0.881 0.009
0.75 0.515 0.005 0.775 0.006 0.845 0.004
0.95 0.456 0.007 0.728 0.005 0.813 0.004
1.15 0.420 0.002 0.664 0.009 0.764 0.006
1.40 0.596 0.004 0.722 0.008
1.75 0.503 0.007 0.650 0.005
Table D.8: Raw conversion data of ozone decomposition for particles of narrow size
distribution with different activities; catalyst inventory: 5 kg
condition k = 2.32s-1
T k =4.49s- 1
r k = 9.10s- 1
T
U, m/s mean deviation mean deviation mean deviation

0.10 0.816 0.009 0.902 0.007 0.916 0.011
0.30 0.665 0.006 0.816 0.008 0.855 0.007
0.50 0.551 0.007 0.723 0.010. 0.802 0.007
0.75 0.430 0.008 0.651 0.004 0.730 0.008
0.95 0.382 0.003 0.590 0.009 0.700 0.005
1.15 0.325 0.005 0.529 0.003 0.634 0.004
1.40 0.482 0.006 0.609 0.006
1.75 0.435 0.005 0.571 0.003
distribution with different activities; catalyst inventory: 5 kg
condition fc =4.625"
P
1
fc =9.04s-
r
1

0.10 0.913 0.008 0.931 0.015
0.30 0.846 0.011 0.890 0.007
0.50 0.804 0.005 0.847 0.010
0.75 0.720 0.003 0.793 0.009
0.95 0.663 0.009 0.750 0.006
1.15 0.594 0.005 0.692 0.008
1.40 0.520 0.007 0.648 0.006
1.75 0.455 0.003 0.606 0.004
Table D.10: Raw conversion data of ozone decomposition in fluidized beds with different
gas distributors, for particles of wide size distribution; catalyst inventory: 1.2 kg
test condition 0.2 m/s

U= U= 0.6 m/s
distributor No. mean deviation mean deviation
No. i (k^sms- )1
0.703 0.006 0.516 0.007
No. 2 (fc^S.lOs- ) 1 0.645 0.005 0.495 0.008
No. 3 (Av=7.95s- ) a 0.622 0.007 0.469 0.004
No. 4 (fc =8.23«- )
P
1 0.581 0.009 0.440 0.006
Table D . l l : Raw conversion data of ozone decomposition in fluidized bed with different
gas distributors, for particles of wide size distribution; catalyst inventory: 3 kg
test condition U= 0.2 m/s U= 0.6 m/s

No. 1 (ife =8.16-5- )
P
1
0.808 0.007 0.716 0.009
No. 2 (fc^S.OOs- ) 1
0.792 0.004 0.700 0.005
No. 3 (A: =7.93s- )
r
1
0.775 0.010 0.684 0.004
No. 4 (A: =8.04s- )
r
1
0.707 0.008 0.610 0.006
Table D.12: Raw conversion data of ozone, decomposition in fluidized bed with different
gas distributors for particles of narrow size distribution; catalyst inventory: 1.2 kg
test condition U = 0.2 m/s U = 0.6 m/s

No. 1 (A; =7.96^- )
r
1
0.650 0.009 0.441 0.008
No. 2 ( A : ^ . ^ - ) 1
0.631 0.006 0.437 0,009
No. 3 (fc^S.OOs- ) 1
0.592 0.007 0.402 0.005
No. 4 {k^LVOs- ) 1
0.485 0.006 0.348 0.003
Table D.13: Raw conversion data of ozone decomposition in fluidized bed with different
distributors for particles of narrow size distribution; catalyst inventory: 3 kg
test condition U = 0.2 m/s U = 0.6 m/s

No. 1 (fc =8.1l5- )
T
1
0.770 0.011 0.633 0.004
No. 2 ( f c ^ . o g s )
-1
0.764 0.005 0.620 0.005
~No. 3 (K^.OOs- ) 1
0.732 0.002 0.584 0.006
No. 4 (A; =7.965- )
r
1
0.681 0.008 0.540 0.007
Table D.14: Axial dispersion model test for particles of wide size distribution at higher
gas velocities; catalyst inventory: 5 kg.
. test condition model of this study Danckwerts' model
U, m/s ks
Tj Xexp. 2- cal. (•êarp. X l)/x p
ca eX X al.
C
(x .
exp xi
ca )/X p
eX
0.75 2.41 0.515 0.516 - 0.041 0.603 - 0.171

0.95 2.41 0.456 0.483 - 0.059 0.540 - 0.184
1.15 2.41 0.420 0.410 + 0.024 0.465 + 0.107
0.75 4.46 0.775 0.772 + 0.004 0.821 - 0.059
0.95 4.46 0.728 0.715 + 0.018 0.749 - 0.029
1.15 4.46 0.664 0.619 + 0.068 0.650 + 0.021
1.40 4.46 0.596 0.564 + 0.070 0.582 + 0.023
1.75 4.46 0.503 0.417 + 0.171 0.435 + 0.135
0.75 8.95 0.845 0.886 - 0.049 0.937 - 0.104
0.95 8.95 0.813 0.865 - 0.064 0.896 - 0.093
1.15 8.95 0.764 0.780 - 0.021 0.824 - 0.079
1.40 8.95 0.722 0.730 - 0.011 0.758 - 0.050
1.75 8.95 0.650 0.608 + 0.065 0.630 + 0.031
average deviation 0.051 0.084

Appendix E
T y p i c a l measurement results of particle size and size d i s t r i b u t i o n
204
Table E . l : Size and size distribution for N-FCC particles of wide PSD
Same:9 Io: Date: ib.11.68 380 micron Orifice Tuoe, serial * 937 Vciuaetricj 2007.9 microliters
True L O C I Dia. Cn i. i-ia. Ch 128 Current, Gain Count;; Preset Tine Elapsed Time Preset Peak Preset Tot-a
14.3 ("14) . • 6.63 181.51 5.5 1.0 1574 0 .0 . 3.7 25600 8000000
Total - 147679 he an at 59.20 Median at 59.20 Mode at 6't.Ol
Size Volume Xlargiir Chnl Size -Wums '/.larger Crml Sire Volume /larger Chnl Size Vek XLsroc-r
iOO.OO 15.27 97.24 36.61
100.00 15.67 96.87 32w.5
ii
ii 3W*
li 11 IC
38 17.39 66 08 ii": 94 74.84 3831 21.2-;
39 17.85 67 7.03 1231 95 76.82 273'! 18.65
40 13.32 68 38.01 i486 96 78. S 4 3225 I6.T-0
41 18.81 69 39.01 1635 97 80.92 2616 14.61
42 19.30 70 40.04 1814 98 83.06 3300 12. i: 4
43 15.81 71 41.10 2069 99- 85.25 2209 10.61
44 20.34 72 42.19 2608 100 87.5,0 1653 9.11
45 20.87 73 43.30 2875 101 89.81 1539 7.9?
16 21.43 74 44.44 2865 102 92.19 1575 6.91
47 21.99 75 45.62 3323 103 94.62 1548 5.85
48 22.57 76 46.82 3425 104 97,12 1005 4,
3 49 23.17 77 48.06 3825 105 99.68 815 4.12
50 23.78 78 49.33 4223 106 102.31 ?7? 3.57
4 51 24.41 79 50.63 4247 107 105*01 1164 2.90
2 52 25,05 80 51.56 5079 108 107.7? 687 2.12
4 53 25.71 31 53.34 5451 66.05 109 110.63 247 1.65
5 54 26.2? 82 54.74 5084 62.36 HO 113.55 669 1 .Al;
5 rc
27.0? 83 56.1? 5108 59.91 111 116.55- 434
6 JJ
27,60 34 57.67 5699 55.45 112 119.63 156 1.03

.40 56 85 59.20 4967 113 122.7? 338 .74
8 rr-i
w/
28.54 51.60 .63
56 29.2? 36 60.76 5371 40.23 114 126.03 183
30.06 7 c
62.36 5095 44,5? 115 129,36 198 . ;0
10 '-:p
53 lie 132,77 .26
60 XM> 64.01 64v0 41.14 2i-v
Table E.2: Size and size distribution for N-FCC particles of wide PSD after being acti-
vated
Sample Id: Date: i..33 360 micron Orifice Tube, serial t 937 Volumetric 2007.9 microliters
True Log Dia. Ch I Dia. Ch 128 Current Gain Counts Preset Tine Elapsed Time Preset Peak Preset Tot-.il o
14.3 ("14) 4.63 181.51 ' 5.5 1.0 686 0 .0 3.6 25600 8000000
CD
Total = i50435 Mean at 65.70 Median at 67.43 Mode at 67.43 Co
C
i-l
Chnl Volume ZLarger Chnl Size Volume ZLarger Chnl Size Volume ZLarger Chnl Size Volume ZLargw. 1
5 7.36 6 100.00 33 15.27 5 99.92 41 31.67 88 99.39 89 65.70 5775 54.07
6 7.55 22 100.00 34 15.67 3 99.92 62 32.51 148 99.34 90 67.43 4400 50.23 g
si
7 7.75 16 99.98 35 16.09 5 99.92 63 33.37 160 99.24 91 69.21 5620 45.98 cn
8 7.96 8 99.97 36 16.51 7 99.91 64 34.25 249 99.13 92 71.04 5215 42.24 B
9 8.17 4 99.97 37 16.95 10 99.91 65 35.15 247 98.97 93 72.91 5590 38.78
10 8.38 3 99.96 38 17.39 12 99.90 66 36.08 363 " 93.80 94 74.84 4613 35.06
11 8.60 1 99.96 39 17.85 12 99.90 67 37.03 424 98.56 95 76.82 5332 32.00 a.
12 8.83 2 99.96 40 18.32 11 99.8? 68 38.01 605 98.28 96 78.84 . 4836 28.45
13 9.06 1 99.96 41 18.81 .n 99.88 49 39.01 729 97.38 97 80.92 4492 25.24 fa
14 9.30 1 99.96 - 42 19.30 16 99.87 70 40.04 329 97.39 98 33.06 3262 22.25 ST.
15 9.55 0 99.96 43 19.81 10 99.86 71 41.10 1221 96.84. 99 85.25 2900 20.08 CL
16 9.80 i 99.96 44 20.34 19 99.65 72 42.19 1634 96.03 100 87.50 3560 18.15
17 CD
18 10.06 2 99.96 45 20.87 15 99.84 73 43.30 1941 94.94 101 89.81 3024 15.79 tsj
19 10.33 2 99.96 46 21.43 19 99.83 74 44.44 2099 93.65 102 92.19 3072 13.73
20 10.60 •j 99.95 47 23.99 15 99.32 75 45.62 2102 92.26 103 94.63 2572 11.74 B
21 10.83 2i 99.95 48 22.57 26 99.31 76 46.82 2683 90.86 104 97.12 2549 10.03 a.
2? 11.17 99.95 49 23.17 24 99.79 77 48.06 2850 89.07 105 99.68 2005 8.33
11.46 99.95 50 23.78 24 99.73 78 49.33 2887 87.18 106 102.31 1761 7.00 (SI
23 11.76 2 99.95 51 24.41 24 99.76 79 50.63 311? 84.59 107 105.01 ' 2344 5.83 fl>
24 12.08 3 99.95 52 25.05 28 99.74 80 51.96 3337 82.52 103 107.79 \ji2 4.27 &
25 12.39 4 99.95 53 25.71 47 99.73 81 53.34 3973 80.30 109 110.63 1199 3.11 tn
26 12.72 3 99.94 54 26.39 40 99.69 82 54.74 4587 77.66 110 113.55 1111 2.32 a.
27 13.06 4 99.94 55 27.09 43 99.67 83 56.19 4533 74.41 111 116.55 1001 1.53 Cr
20 13.40 3 99.94 56 27.80 41 99.44 .84 57.67 5145 71.60 112 119.63 650 .91 a
2? 13.76 5 99.94 57 28.54 74 99.61 85 59.20 5353 68.18 113 122.79 468 .48 o
30 14.12 3 99.93 58 29.29 48 99.56 86 60.76 • 4937 64.62 114 126.03 253 .1.7
31 14.4? 7 99,93 59 30.06 86 99.53 87 62.36 5124 61.34 S3
32 14.33 5 99.93 60 30.86 119 99.47 86 64.01 5804 '57-V3
to
o
3
Table E.3: Size and size distribution for N-FCC catalysts of wide PSD after being flu-
idized in the reactor for 5 hours
Sample Id: 1 Date: 18.11.38 380 micron iOrifice Tube, serial * 937 Volumetric: 2007.9 microliters
True Li: Dia. Cn i D\a. Ch 128 Ci Gain Counts Press t Time tlaosed Time Preset Peak Preset Tc
:'14) 6.63 181.51 • 5.5 1.0 433 0 .0 3.7 25600 300C CD
P
tn
C
iotal 154901 ftean at 64.01 ian at

ted 65.70 Mod e at 67.43
I
:
CI'iTli Size VoU .'fee ;larger Chnl Size Volute ZLarger Chnl Size Volume ZLarger Chnl Size V.ilume ZLarger 8
5 7.36 14 100.00 . 33 15.27 ? 99.34 61 31.67 336 98.3? 89 65.70 5051 50.04 3
6 7.35 46 34 15.67 13 99.83 62 32.51 438 93.18 • 90 67.43 6400 46.','.5 tn
7.75 35 99. ?6 35 16.09 9 ??.82 63 33.37 486 97.86 91 69.21 5536 42.65 e
8 7.96 17 99.94 36 16.51 li 99.82 64 34.25 514 97.55 92 71.04 5487 39.08
9 8.17 7 99.93 27 16.95 14 99.81 65 35.15 653 97.21 93 72.91 5502 35.54 o
10 8.38 4 99.92 . 38 17.3? 15 99.80 66 36.03 1046 96.7? 94 74.84 5366 31.';tf i-S
11 8.60 3 99.92 39 17.85 14 99.79 67 37.03 1018 96.12 95 76.82 4415 28.52
12 8.03 2 99.92 40 18.32 34 99.78 63 38.01 1208 95.46 96 78.84 3001 25.67 "a
13 9.06 i3 99.92 41 18.81 23 99.76 69 39.01 1422 94.68 97 80.92 5457 23.73 CL.
14 9.30 99.92 42 19.30 18 99.74 . 70 40.04 1072 93.76 98 83.06 3827
15 9.55 3 99.91 43 19.81 35 99.73 71 41.10 1411 93.07 99 85.25 3276 20.21
17.74 IS)
16 9.80 3 44 20.34 i6 99.71 72 42.19 1609 92.16 100 87.50 2237 15.62
17 10.06
99.91
4 ' 99.91 . 45 20.87 38 99.70 73 43.30 2034 91.12 101 69.81 4032 14.18
18 10.33 1 99.91 46 21.43 55 99.67 74 44.44 2028 89.81 102 92.19- 3052 11.58 B
19 10.60 J 47 21.9? 38 99.64 75 45.62 2642 88.50 103 94.62 1414 9.61 a.
•20. 10.88 4 99.91
99.90 48 22.57 77 99.61 76 46.82 2485 86.79 104 97.12 2294 8.69 IS)
21 11.17 5 99.90 49 23.17 45 99.56 77 48.06 3521 85.19 105 99.68 1654 7.21
22 11.46 5 99.90 50 23.78 71 99.53 78 49.33 2973 82.92 106 102.31 2284 6.15
23 11.76 7 99.8? 51 24.41 97 99.49 79 50.63 3973 81.00 107 105.01 1934 4.61 tn
24 12.08 6 99.89 52 25.05 79 99.42. 80 51.96 4530 78.43 108 107.79 1045 3.36
118 ccr
25 12.39 5 99.89 53 25.71 99.37 81 53.34 4012 75.51 109 110.63 1130 2.66
26 12.72 7 99.88 54 26.3? 143 99.30 82 54.74 4338 72.92 110 113.55 0 1.75
27 13.06 7 99.88 55 27.0? 149 99.20 83 56.19 4098 70.12 111 116.55 441 1.95
28 13.40 7 99.87 56 27.80 144 99.11 84 57.67 5659 67.47 112 119.63 953 1.67
29 13.76 15 99.87 57 28.54 181 99.02 85 59.20 5311 63.82 113 122.79 1030 1.05
.3?
3?
14.12 13 99.86 53 2?. 2? 168 98.90 86 60.76 5431 60.39 114 126.03 0
JI 14.4? 8 79.85 59 30.06 272 98.79 87 62.36 5265 56.88 115 129.36 602 .39
32 14.83 10 99.85 60 30.36 344 98.61 88 64.01 5328 53.4a * to
o

Influence of Particle Size Distribution On The Performance of Fluidized Bed Reactors

Uploaded by

Copyright:

Available Formats

Influence of Particle Size Distribution On The Performance of Fluidized Bed Reactors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Influence of Particle Size Distribution On The Performance of Fluidized Bed Reactors

Uploaded by

Copyright:

Available Formats

INFLUENCE OF PARTICLE SIZE DISTRIBUTION ON T H E

P E R F O R M A N C E OF FLUIDIZED BED REACTORS

W e accept this thesis as conforming

© Guanglin Sun, 1991

The University of British Columbia

distributor on the performance of fluidized bed reactors.

The influence of particle size distribution on the hydrodynamics of fluidization was

List of Figures vii

1.1 PSD effect on the hydrodynamics of fluidized beds 2

2 Experimental Particles and Apparatus 13

2.1 Preparation of Particle Blends . 13

2.2 Experimental Apparatus 26

2.2.1 Gas supply system 26

2.2.3 Particle collection and return system 30

2.2.4 Gas Distributors 32

3 Reactor Performance at Low Gas Velocities 36

3.1 Rate Constant Determination 36

3.2 Stability of Catalyst Activity 40

3.3 Definition of Efficiency for Fluidized Bed Reactors 40

3.4 Effect of P S D on Conversion and Reactor Efficiency 41

3.5 Influence of Whether Fines or Coarse Particles are Active 49

3.6 M o d e l Predictions of P S D Effect at Low Gas Velocities 52

3.7 Influence of Overall Catalyst Activity on the P S D Effect 56

4 Reactor Performance at Higher Gas Velocities 60

4.1 Previous Work on Fluidization Performance at High Gas Velocities . . . 61

4.2 Dependence of P S D Effect on the Superficial Gas Velocity 64

4.3 Influence of Fluidization Regime on Reactor Performance 68

4.3.1 Regime influence for particles of wide size distribution 69

4.3.2 Regime influence for catalysts of narrow size distribution 73

4.4 Prediction of P S D Effect at High Gas Velocities 75

5 Effect of Particle Size Distribution on Two-Phase Behaviour 81

5.1 Particle Concentration i n Voids 82

6 Pressure Fluctuations and Regime Transitions 117

6.1 Brief review of previous work 117

7 Reactor Performance in the Distributor Region 139

7.1 Brief Review of Previous Work 139

8 Conclusions and Recommendations for Further Work 161

8.1 Overall Conclusions 161

A Minimum integral reactor length of minimize axial diffusion effects 185

C Statistical Analysis of Experimental Results 191

Cl Significance test of the influence of the PSD 191

C.2 Confidence intervals of experimental results 193

D Raw Data from Chemical Reaction Tests 195

E Typical measurement results of particle size and size distribution 204

3.1 Determination of the rate constant for 10 g samples of N-FCC catalysts

6.1 Transition velocity U for N-FCC particles of different size distributions

mm) and Az (930 to 1180 mm)

C. l Reaction conversions measured in the fluidized bed reactor for particles of

D. l Raw conversion data of ozone decomposition for particles of wide size

1.1 Coke particle size distributions. Identifying numbers on curves correspond

2.1 Dimensions of the infrasizer 15

catalyst inventory: 5.05 kg. PF and PM are corresponding results for

single-phase plug flow and perfect mixing respectively 45

catalyst inventory: 2.75 kg. PF and PM as in Fig. 3.2 46

fraction active, respectively 48

3.6 Effect of which fraction is active on conversion for bimodal catalyst; k =

4.0 to 4.2 a ; catalyst inventory: 5.05 kg

5 kg, k % 2.4 s'

5 kg, fc « 4.5 s'