Professional Documents
Culture Documents
N-Type TwoPhaseTheory
N-Type TwoPhaseTheory
Abstract-The distribution of gas flow between bubbles and the dense phase in a fluidized bed is
considered. By deriving general equations for gas and solids continuity in a freely-bubbling fluidized
bed, the possible contributions to the invisible gas flow are formulated. It is shown that the gross
assumptions inherent in “two-phase theories” of fluidization are not justified. Measurements of visible
bubble flow and bubble volume fraction can give nothing more than rough estimates for the gas flow
through the two phases.
Table 1. Summary of experimental data on the division of gas flow between phases in freely fluidized beds
Height above
Bed cross- Method of Particles Range of distributor, Range of Range of
section measure- d. K”, (U - U,“,) K_ U-G&
Reference (cm) ment Type (wL) (cm/s@ (cmlsec) (C”m,
Kl
(see I&. (1))
5 7.6 x 15.2 Gamma-rays Cracking cat 41 0.7 5-21 8,27 120,140
(Baumgarten Glass beads 74 0.7 l-21 2-18 24,56
and Pigford) Glass beads 116 1.1 5-20 20 5,11 33,40
Glass beads 119 1.6 l-23 20 1.5-7.5 9,22
Glass beads 234 5.2 4-25 20 16-4.5 7-13
6 1 x 24.2 Photography Glass beads 100 1.2 1.7-10.5 20-100 1.0-6.6
(Pyle and Glass beads 370 14.7 1.O-27.0 10-100 1.0-2.3
Harrison) Iron shot 180 14.7 3-26 20-% 1.2-2.2
7 122 x 122 Light probes Sand 190 2.5 36-17 85,206 2.1-5.8
(Whitehead
and Young)
30.8 dia Observation Sand 275 5.6 l-6 5-20 1.1-1.3
(Gelda:) from above
10 1.9x45.7 Photography Magnesite 274 8.1 5-24 15-60 1.1-3.0 4.3
(Grace and
Harrison)
30.8 dia Photos from Sand 128 1.2 2-8 10-60 1.5-4,8
(Geld;) above
1.3x68.5 Photos from Sand 128 1.5 2-8 10-80 1 .J-4.5
front
(Kelsez 1.3 x 68.5 Photography 128 1.5 0.5-18 20-80 1.1-9.4 22.47
1.3 x 68.5 Observation l-20 S-80 1.1-9.3
from above
20x30 x-ray Alumina 210 2.5 1.3-3.8 17-70 1.3-1.9
(Everet) or30x30 Photography Carbon 2% 8.0 2.4-56 12-60 1.2-1.5
QUFUtZ 135 2.8 3.3-15.7 17-70 1.2-2.1
Glass beads 323 8.0 4.8-11.2 15-45 1.2-1.5
Glass powder 268 5.5 4.1-9.6 15-60 1.5-2.1
15.2 x 30.5 Observation Sodium 1620 55 O-40 2.6-13.3 1.1-1.4 1.7-5.8
(McG:h and from above aluminate
Streattield)
2x61 Photography AlUmina 1520 58 60 5-20 1.8-2.0 3.5-5.0
(Gelda: and catalyst
Cranfield)
17 dias of 10, Sand 83 1.8 7.1 530 1.0-3.1
(Werther and 20,45,100 p&be
Molerus)
18 1.3x74 Photography Glass beads 490 23.0 4.618.4 20-80 1.0-1.4 0.3-3.2
(Motamedi and Glass beads 280 7.6 3.0-12.9 20-80 0.8-1.1 -1.2-1.2
Jameson) Glass beads 100 2.2 4.2-8.2 20-80 0.9-1.8 -1.2-7.3
Sand 270 10.2 6.1-11.2 20-80 1.2-1.6 2.2-8.1
Sand 500 20.9 8.4-16.7 20-80 1.2-1.6 1.8-5.2
19 61 dia Electrorest- Coke 220 2.1 2-13 11-102 0.7-3.1 - 8-74
(Nakamura) ivity probes
20 1x56 Photography Glass heads 213 5.3 13-23 20-130 1.0-3.4 O-12.6
(Chavarie)
and &. These values are almost always positive but of magnitude with the n-type theory. However, no
vary over a very broad range. It is now clear that unambiguous indications of the division of gas flow
when a value of n is assigned by using Eq. (l), it can be obtained from measurements of these vol-
differs from system to system (and possibly from ume changes.
observer to observer) and within a given system is In the light of these observations, a re-
dependent on height and superficial gas velocity. assessment of the two-phase theory is due. We con-
These observations indicate that the n-type theory, sider the most general case, a freely-bubbling bed
which was intended as an improvement on the orig- with variable dense phase voidage (E), variable in-
inal two-phase postulate, is itself oversimplified. terstitial gas velocity relative to particles (u,), and
A somewhat surprising feature of gas fluidized variable gas velocity within any given bubble (&).
beds is that changes in total bubble volume occur as The assumptions necessary to proceed from the
bubbles coalesce [25-283 and split [28]. These general case to the n-type theory are also pre-
changes are consistent in their direction and order sented.
On the two-phase theory of fluidization 329
SOLIDS CONTINUITY used to derive two useful identities:
It will be convenient to consider the dense phase
as two superimposed fluids, one consisting of solid
v,da = v,e da (2a)
particles and the other of interstitial gas. By taking I A-1ioi I A-PPi
ensemble averages over distances which are large
and
with respect to the dimensions of solid particles, it
is possible to treat the phases as continua [29]. Vel- 1
v,da =- v,S. da (2b)
ocities and voidage may then be treated as continu- I A-\‘,a< 1 - EnIfI A-Ziai
ous functions: this simplifies the notation, although
where
the conclusions are not dependent on this formula-
tion. The particle velocity will be denoted by v = 8. = E - c&f. (4)
0,; + v,; + v,E (where the x-axis is directed verti-
cally upwards) and the voidage will be denoted by Eqs. (2), (2a) and (2b) must be valid at any level in
E. the bed. Eq. (2a) simply shows that the time-
Consider a horizontal section XX through a fluid- average net flux of dense phase across any section
ized bed at any instant as shown in Fig. 1. The net is equivalent to the net flux of interstitial voids
moving with the particles. Eq. (2b) emphasizes the
fact that when 8. and u, are correlated (for exam-
ple, if the voidage of rising dense phase elements is
greater than that of downward moving elements)
then there can be a net transfer of gas in the dense
phase across a section due to an invisible transla-
tion of interparticulate voids. This flux is quite inde-
pendent of the interstitial gas flow relative to the
particles.
X X
UA = 2
I
Uoiai +
I A-W
vx(l - l)da
(5)
+ (u, + u,)cda
I A-&o,
Fig. 1. Schematic diagram of a horizontal section through
a bubbling fluidized bed. where UGi is the mean gas velocity across XX
within bubble i and u, is the vertical component of
the gas velocity relative to the particles. Averaging
solids flux across XX at any instant is _fA-Z,ar
t?,(l- over time and using Eq. (2), we obtain
l)da where a, is the area of the i’” bubble cut by
the section. Over a long period of time the solid par-
UA = 2 Uciai + w da + u,eda. (6)
ticles remain in the bed so that I A-Pai I A~&Ii
(iii) _fA-2oiV,Eda = Average net flux of interstitial server. The value of ki is then arbitrary, so that it
voids moving with the particles (discussed above). should not be surprising that different observers ob-
(iv) _fAeZo, u,*~E
da = Average interstitial gas flow tain different values for the apparent throughflow
relative to particles in the dense phase. This term term 3 kJJ,,,,a when estimating it by measuring Gs.
includes the gas flow due to any “parvoids” (voids If some arbitrary measure of Gs is adopted, Eqs.
too small to be considered true bubbles) which (2a), (6), (7) and (7a) may be combined to give
move relative to the particles in the dense phase.
The above formulation is adopted in order to di-
vide the gas flow in the way appropriate to an
analysis of a solid-catalyzed reaction in a fluidized
bed. The essential division is between term (ii), rep-
resenting the gas flow which crosses XX without
contacting particles, and terms (iii) and (iv), denot- which applies at any level, with es defined by
ing gas which crosses XX in intimate contact with
the particles. The distinction between bubbles in T a, = Ae.
term (ii) and parvoids included in term (iv) there-
fore depends on whether the gas in the void is Applying Eq. (6a) at two levels, 1 and 2, and sub-
effectively out of contact with the particles. Thus tracting, we obtain
the criterion distinguishing parvoids from bubbles
is somewhat arbitrary even for a given system. On
the other hand, the division between terms (iii) and
(iv) is unimportant and is adopted purely for analyt-
ical convenience. In the n-type two-phase theory,
term (ii) is further subdivided:
+[I ~,,-<a,, ” dal
da-I,,,-..>: (8)
+ U,~E da- GE da
I *,I-.821 1
where Usi is the velocity of bubble i and k,lJ,,,, is
where AGs,* is the rate of formation of bubble vol-
the mean gas velocity across ai relative to the bub-
ume between levels 1 and 2. In practice, Gg has
ble (the “throughflow velocity”). The gas flow ac-
been found to increase with heightl6, 10,201
ross XX within bubbles (term (ii)) may then be
primarily due to bubble growth on coalescence so
written
that AGslz is positive when section 2 is above sec-
~ -
tion 1. Equation (8) shows that this generation of
T UGiai = c Usiai + 2 kiUm,ai. (7)
bubble volume between two levels must accom-
pany a net change in the sum of the throughflow,
The first term is then assumed to represent the
flux of interstitial voids, and relative interstitial gas
translation of bubble voids (sometimes called the
“visible bubble flow”) written flow. It gives no indication as to which of these
terms, if any, changes most radically.
Gs = c UBiai. (7a)
MODIFIED OR n-TYPE TWO-PHASE THEORY
Unfortunately this formulation relies upon the exis- Apart from the assumption that Gg is a
tence of an unambiguous measure of UBi or G,, as uniquely-defined quantity, the modified two-phase
would exist if all the bubbles in the bed rose with theory is based on the following assumptions:
constant size, shape and velocity. In fact, bubble in-
teractions cause changes in bubble shape and (i) lacl-.Bj v, da = 0. This assumption is valid if l
velocity[30], increases of bubble volume[25-281, is constant everywhere in the dense phase (see Eq.
and “leakage” whereby gas is exchanged between 2) or if e and v, are uncorrelated (see Eq. 2a). It
bubbles which retain separate identities may be noted that this assumption is rather more
stringent than has previously been recognized. The
[7, 13,26-28,31,32]. The effect of these phen-
omena is discussed in Appendix B and it is assumption of a constant mean dense-phase void-
age, i.e. JA,,_raj8. da = 0 is not sufficient; Eq. (2b)
shown that there is no absolute measure of G, or
shows that the correct form must be
UBi.Thus the value of these quantities depends on
ihe method of measurement adopted by the ob- J-A<,-.s,vx&
da = 0.
On the two-phase theory of fluidization 331
(ii) JAcl_mU,E da = LLfA (1 - &). i.e. the average tal studies[l9] summarized in Table 1. It is clear
dense-phase interstitial gas velocity is equal to that that the fraction of the total flow in the dense phase
,at incipient fluidization. varies considerably according to the interpretation
(iii) Z kiai = /&A = (1 + n&A. This implies that adopted by different workers. The spread, from 14
ki and ai are uncorrelated. The validity of this as- to 59 per cent of the total flow, would be even grea-
sumption clearly depends upon the method of ter if recent work[2] which assumes net downflow
measuring G,, since the value of ki is determined by in the dense phase were included.
the method adopted. As the bottom row of Table 2 shows, the actual
magnitude of the invisible flow terms is unknown
With these simplifications, Eq. (6a) reduces im- and there is no particular reason, based on existing
mediately to Eq. (l), while Eq. (8) becomes experimental evidence, to prefer any of the first
four approaches. The original two phase postulate
AGslZ = Um,A(nleBj - n&J. provides one estimate for the total flows in each
phase, (even though the visible bubble flow is over-
It is usually implied that n = E - 1 is not a function estimated), but workers who use it in reactor mod-
of either height x or of superficial gas velocity. elling must recognize that the actual flows may
With this further assumption differ widely from those assumed. The n-type
theory, on the other hand, can be regarded as a
AGs I?= lJ,,,,An (C,,, - .h). conservative estimate in reactor modelling since all
the invisible flow except for U,,,,A( 1 - Es) is as-
Equation (la) does predict an increase in visible signed to the bubble phase. However, the values as-
bubble volume with increasing height above the dis- signed to n so far do no more than illustrate the
tributor for n >O, but only to the extent that efl magnitude of the deviation from the Toomey and
decreases with height. Johnstone postulate and show the need for further
work.
DISCUSSION The gas distribution problem can only be settled
The flows assumed for each phase by different by studies in which simultaneous measurements are
workers or in different versions of the two phase made of ls, some clearly stated Gs and either E
theory are listed in Table 2. The more general re- (corresponding to the definition of Ge used) or u,
sults from Eq. (6a) are also tabulated. In order to and v,. Only then will it be possible to evaluate the
illustrate the magnitude of the difference between integral terms in Eqs. (6) and (6a). Unfortunately,
the various assumptions, the values of the flows are satisfactory means of measuring the invisible flow
also listed for a particular set of values of U, U,,,,,A, components (i.e. I? and u,) have not yet been found.
&+and GB corresponding to one of the experimen- Instead workers continue to measure the only ac-
Table 2. Flows assumed in various versions of the two-phase theory. The figures in brackets (l./sec) are illustrative of
the difference between the models. These figures are based on measurements by Nakamura[ll where U,,,‘ =
2.1 cm/set, U = 14.7 cmlsec, A = 2920 cm’ and the measured values of E* and GB/A were 0.050 and 8.2 cmlsec respec-
tively. The measurements were taken 72 cm above the distributor in a bed of coke particles
interstitial flow
3. Throughflow velocity U&IA G, + Um&A A(U - U,&)- G,
of u,, (0.3) (24.0) (18.9) (4%)
4. n-Type theory U&l + n)&A A(U-Um,(l-&)) U,,,,A(l -k)
(13.4) (37.1) (5.8) (4%)
5. Eq. (6a)
G, + U, I: k,a,
- Iv. + f*,,-..,U-~ da
(23.7) (2 0) (i-z)
332 J. R. GRACEand R. CLIFT
cessible flow component (GB/A). It must be em- ured bubble flow and Toomey predic-
phasized, however, that measurements of G, IA by tion, (U - Gs /A)/u,,,,
themselves, no matter how careful, cannot clarify flow rate of gas relative to a bubble ac-
the overall distribution of gas flow between the ross a horizontal section inside the
phases since the invisible flow is substantial and bubble/ Urn@
yet its distribution remains unknown. Future work time average of k over a given horizontal
must concentrate on experimental techniques of bed cross-section
measuring the invisible flow. The present authors total number of photographic frames
recently reported one experimental method of es- analyzed for a given sequence
timating bubble throughflow velocities by means of %- 1; see also Eq. (1)
tracer particles[33]. Work in this direction is con- total time span
tinuing although the difficulties in obtaining and in- time
terpreting data are formidable. Moreover, the wide superficial gas velocity
range of values of n reported for the modified two- bubble velocity
phase theory (see Table 1) suggests that the various absolute mean gas velocity across sec-
terms may have different magnitudes in different tion ai inside bubble i
systems. This raises the unenviable prospect of superficial gas velocity at minimum fluid-
having to characterize the dependence of these ization
parameters on particle properties, superficial gas relative interstitial gas velocity
flow rate, and height above the distributor. volume of bubble i below section XX
(see Fig. 2)
CONCLUSIONS bubble volume
Gas passes through aggregatively fluidized beds particle velocity
partly in the bubble phase and partly in the dense x, y, 2 components of v (x vertical)
phase. The division of gas between the phases is of height above distributor
considerable importance in reactor modelling and
has been the subject of considerable discussion and Greek symbols
conjecture. Recent measurements of “visible bub- l dense phase voidage
ble flow rates” are subject to arbitrary interpreta- 63 volume fraction of visible bubble phase
tion due to volume and shape changes which ac- lmf bed voidage at minimum fluidization
company bubble interactions. More serious, the 8, E - Clnf
fraction of the total flow not accounted for by such
visible flow is substantial while each of the phases Subscripts
contains a component of flow (“throughflow” for i i”’ bubble cut by a given horizontal sec-
the bubble phase; interstitial flow relative to parti- tion at a given instant
cles and flux of interstitial voids for the dense j j’” photographic frame
phase) which has not been subject to experimental 1,2 levels in the bed
measurement. The original and n-type “two-phase
theories of fluidization” amount to no more than REFERENCES
untested guesses for the division of gas flow. Thus [l] DAVIDSON J. F. and HARRISON D., Fluidised
the two-phase problem of fluidization remains open Particles. Cambridee Universitv Press 1%3.
and vexatious, but the directions required for [2] KUNII D. and LEiENSPIEL-O., Fluidization En-
gineering. Wiley, New York 1%9.
future work are now clear. [3] GRACE J. R., Chem. Engng Prog. Symp. Ser. 197167
No. 116 159.
NOTATION
[4] TOOMEY R. D. and JOHNSTONE H. F., Chem.
A cross-sectional area of bed Engng Prog. 1952 48 220.
[S] BAUMGARTEN P. K. and PIGFORD R. L.,
a area on section XX
A.I.Ch.E. Jl l%O 6 115.
ai area of ilh bubble cut by horizontal sec- [6] PYLE D. L. and HARRISON D., Chem. Engng Sci.
tion 1967 22 1199.
d, mean particle diameter [7] WHITEHEAD A. B. and YOUNG A. D., Znterna-
GB average visible volumetric bubble flow tional Symposium on Fluidization, p. 294. Nether-
lands University Press 1%7.
across a given section [8] National Coal Board, Stoke Orchard, private com-
; j-, E unit vectors munication 1968.
K measure of discrepancy between meas- [9] GELDART D., Powder Technol. 1%8 1 355.
On the two-phase theory of fluidization 333
alternate measure of visible bubble flow is given by adds the volume of bubble 2 before coalescence, he must
underestimate G,. Method 1 is again unreliable, since the
rapid changes of velocity and shape during coalescence
prevent simple definition of a meaningful value for U.,.
For the situation depicted in Fig. 3(d), methods I and 2
For bubbles of constant volume, shape and velocity, it break down even more seriously. Assuming that &,a, or
is simple to show that all three methods give the same re- the volume of bubble 1 was measured in the first frame
sult. However, tine films of freely bubbling beds reveal shown, and that bubble 2 maintains a constant volume
that changes in bubble volume, shape and velocity are after the leakage is complete, we see that certain elements
common features for individual bubbles, these changes of gas will be counted twice so that the method overesti-
usually being associated with bubble interactions. To de- mates G,. Thus method I is inadequate for any of the
monstrate the difficulties inherent in the three methods, events depicted in Fig. 3, while method 2 can handle shape
consider the situations depicted in Fig. 3. changes but fails to deal with volume changes of the types
In Figs. 3(a) and (b), bubbles undergo shape changes in Figs. 3(c) and (d).
x-e-x X-x
x13xx~x (a)
cl fd
u
x---y=yx x----Q-x
xQxxax (b)
Q
(d)
0
Fig. 3. Schematic diagram of bubble shape and volume changes which complicate measurements of
visible bubble flow rates in fluidized beds. (a) Bubble shape changes, centroid remaining fixed and
volume constant. (b) Bubble shape changes, nose remaining fixed and volume constant. (c) Bubbles
coalesce and total bubble volume increases. (d) Apparent leakage of gas occurs from one bubble to
another.
without affecting the position of the centroid or nose re- For method 3, the shape changes in Figs. 3(a) and (b)
spectively. If LJ,, is based on the centroid position, then present no particular problem, while bubble volume
according to method 1, no visible bubble gas has passed changes accompanying coalescence are dealt with in a
the level of interest in situation (a); similarly, if U,, is reasonable and consistent manner. However, this method
based on the nose position, t.L is zero in situation (b) and again fails to cope satisfactorily with the situation in Fig.
so we would measure no visible bubble volume crossing 3(d); it assumes that a reduction in void volume for a
the level between these two frames. In both cases, there bubble cut by the section is due to the bubble crossing the
clearly has been a movement of visible voids across the section, whereas the leakage phenomenon can cause bub-
section, We therefore conclude that method 1 is unable to ble volume reduction due to transfer of gas from one bub-
cope with bubble shape changes when a simple estimate ble to another. On the other hand, leakage from a bubble
for IJBi is adopted. below the section to a bubble crossing the section will be
In Fig. 3(c), two bubbles coalesce to give a resultant given adequate representation by method 3.
bubble whose volume is greater than the sum of the vol- Since shape changes, bubble volume change on coales-
umes of the original bubbles. Having measured the vol- cence and splitting, and leakage phenomena can all be
ume of bubble 1 in the first frame according to method 2, occurring simultaneously in freely bubbling beds, we must
the observer is faced with the dilemma of what to do when conclude that the value of G, measured will depend on
coalescence occurs. If he discards his earlier measure- the method of measurement employed and that the value
ment and takes the volume of bubble 3, he is overes- of G, obtained is therefore arbitrary. Since leakage in the
timating the visible bubble flow by that fraction of the downwards direction appears to be less frequent than up-
growth associated with the part of the bubble which has wards leakage, method 3, although tedious, seems to give
already passed the reference section. If he goes back and the most meaningful value for Ge.