Gas and solid behavior in cracking circulating fluidized beds

249

Powder Technology 70 (1992) 249-258

Gas and solid behavior in cracking circulating fluidized beds M. P. Martin, P. Turlier, J. R. Bernard* Unitt! Mtie ELF-CNRS

35, B.P. 22, 69360 Saint Symphorien d’Ozon (France)

and G. Wild LSGC, 1 rue Grandville, 54000 Nancy, (France)

Abstract Gas and solid hydrodynamics have been studied in dilute circulating fluidized beds under conditions occurring in catalytic cracking risers. Gas radial velocity profiles and dispersions were established by a tracer technique in a cold set-up. The gas axial dispersion was determined in an industrial riser. The local concentrations of the solid phase were measured by a tomographic technique. This has allowed an assessment of the core-annulus structure of the bed and an estimate of the solid radial and axial dispersions. The axial solid concentration profiles were determined in pilot and industrial scale beds. These show an important accumulation upstream of the abrupt exit. The overall conclusion is that the gas flow can be considered to be plug flow with a radial velocity profile and a radial dispersion; the solid flow is slightly more dispersed due to the core-annulus structure and a high radial mixing.

Introduction

Circulating fluidized beds have long been employed in industrial cracking plants [l, 21. They have not been extensively studied in detail, however, until the introduction of a new generation of coal combustion plants developed using this technology. A number of authors [e.g. 3,4] showed that a dense bed with a fuzzy but definite level could exist if the solid flux is high enough or if the gas velocity is low. Rhodes and Geldart [5] showed that this level is adjusted so that the solid entrainment at the top of the bed equals the circulation rate. The dense bed cannot exist if the circulation rate is inferior to the saturation carrying capacity, which is given by an elutriation correlation such as the one obtained by Geldart [6]:

(1) where Fs and Fg are the mass fluxes of solid and gas. This ratio is called the C/O ratio (catalyst/oil) in the refining industry. Cracking catalysts are powders in Geldart’s class A with a minimum fluidization velocity of cu. 0.002 m S -’ and a terminal velocity of ca. 0.2 m s-’ in ambient air. They are circulated in the risers of the cracking plants with gaseous hydrocarbons (U, between 3 and *Author to whom correspondence should be addressed.

0032-5910/92/$5.00

20 m s-l) at mass fluxes between 200 and 1200 kg m-* s-l. The risers are vertical pipes whose diameters lie between 0.5 and 2 m, and are 20-40 m high. Liquid vacuum gas oils or residues are fed to the bottom, where they are quickly vaporized and then cracked. The riser terminates with an elbow connected to a disengager in which cracked gases and spent catalyst are separated. The elbow is mostly a blinded tee. This device avoids any erosion of the line by maintaining a permanent buffer volume of solid in the impact zone. The data given in [3, 41 on cracking catalysts and similar data obtained by the present authors show that formula (1) describes well the appearance of a dense bed, with A = 23.4 and B = 12.7. The application of this formula to conditions occurring in the riser reactor shows that the C/O should be larger than 25 to detect any dense bed at the bottom. These conditions would then vanish very quickly because of the velocity increase due to the cracking. As the C/O is in the range from 4-15, the cracking bed is obviously circulating under dilute conditions, even if some more concentrated catalysts may appear at the bottom because of the high solid flux and the medium gas velocity at the bottom, where the cracking conversion is low. Dilute solid-gas streams are usually described by pure plug flows models. This is an approximation, because a core-annulus structure of the solid flow can be detected [l, 71, as in denser beds. The purpose of this paper is to describe the properties of these dilute

0 1992 - Elsevier Sequoia. All rights reserved

250

cracking beds from a detailed assessment of information obtained from both cold set-ups and industrial plants.

Experimental The cold circulation set-up is shown in Fig. 1. The cylindrical riser is 0.19 m in diameter and 11.7 m high. The cracking catalyst is fed to the bottom of the riser via a 40” slope standpipe fitted with a butterfly valve. Air is injected at this point. At the top, the gas-solid flow is directed by a blinded tee elbow to a horizontal pipe which is connected to an upper cyclone hopper. The catalyst flows down through a pneumatic valve to the main hopper which is connected to the standpipe. There is a secondary cyclone with a return leg to the main hopper. The circulation rate is measured by closing the pneumatic valve and recording the catalyst level in the upper hopper by y-counters. The catalyst has a mean volumetric diameter of 62 pm and a particle density of 1560 kg mm3. Its minimum fluidization and bubbling velocities are respectively 0.002 8 and 0.005 7 m s-l; they remain constant during this study.

I

0.19m.

+

He sampling

return

t He sampling rt

The tomographic method using -y-counting described in [8] yields solid concentration maps which are averaged over a period of 2 h. Here it has been significantly improved [9] to give better mathematical resolution and more efficient grid sampling. The matrix associated to the natural pixels is singular, and pseudo-inversion yields the solution in the sense of least squares. For a given number of measurements, the best accuracy is obtained when the number of y-ray paths in a given direction (device translations) is &! times the number of directions (device rotations). Although this ratio could not be attained in the plant for practical reasons, it was reached when mapping the cold set-up. Most of the industrial results come from a cylindrical riser (0.94 m diameter, 26 m of vertical pipe from the feedstock injection point) with a blinded tee elbow. The catalyst is traced using radioactive 140La and the gas is traced by radioactive “Kr, as described in [lo].

Gas flow Radial properties

The well-known core-annulus structure of circulating beds is the result of gas-solid-wall interactions, and it is recognized that the low void fractions are associated with low gas velocities. It is striking to note that information on this last parameter is very sparse: most authors consider that all of the gas is flowing in the core and it is stagnant in the annulus. Only Van Breugel et al. [ll] show a measured radial velocity profile which is parabolic with no downstream flow at F,=400 kg m -2 s-l and U,=6.4 m s-l. To measure the gas velocity profile in the cold setup, a new method was used; a known steady flow of helium tracer is isokinetically injected in the centerline of the riser, 4 m above the bottom. The air-helium mixture is sampled at three elevations, 0.97, 1.57 and 4 m above this injection point, every 2 cm on the riser diameter in the plane of the cold set-up. The helium concentration is measured by gas chromatography. To determine the right helium feed rate at each elevation, it is necessary to introduce a velocity profile. Therefore, an empirical correlation is introduced; it is formally identical to the Ostwald-de Waele law [12] which describes the laminar flow of non-Newtonian liquids:

L.C He iniection

(2)

riser

Figure 2 shows that the relative gas velocity profile is flat when n =0 and triangular when n is large. Van Breugel’s experiment gives n = 2.1. Taking into account the radial symmetry, the helium mass flow rate is integrated from the experimental profiles: R

air injectioh Fig. 1. Diagram of the cold circulating

fluidized bed.

helium flow rate = 4 2m_U(r).C(r).dr s

0

(3)

251

101

3 u .c =t

Ug = 3.8 m/s Ug = 6 m/s

Ug = 5 m/s

I *I I

I 1

*

i /

I

/’ * //

/ ”

,/

,

/

.2

.4

.6

Relative

.6

1

Solid

and n is determined by identification of integrated and experimental rates. The uniqueness of the solution has been checked. Figure 3 shows the variations of n with F, and U,. When F, = 0, a turbulent flow profile is found and n increases when F, increases or U, decreases. Values superior to 5 are not significant because they influence only the central part of the riser area (Fig. 2). These results are in line with the variations of the core-annulus structure observed by Weinstein et al. [ 131. From the same experiments, the radial gas dispersion can be determined by processing the evolution of helium concentration profiles from a lower level to an upper level. This is done using a classical Fickian model:

with the boundary conditions: C(r, 2 = 0) = lower experimental upper level:

100

radius

Fig. 2. Relative gas velocity profile as a function of the index n of Ostwald-de Waele law.

helium concentration

profile;

/ :’

200 flux

(kg/s/m2)

Fig. 3. Ostwald-de Waele index as a function of the superficial gas velocity and of the solid mass flux.

r=R,

sc 6r

-0

(4)

This differential equation is numerically solved by Gear’s method included in the NAG mathematical library, and 0, is optimized to obtain the best fit with the upper concentration profile. Typical results are shown in Fig. 4: the fit is excellent at the two target levels 1.57 and 4 m in the central part or close to the wall, where the annulus is developed. Werther et al. [14, 151 studied the radial gas dispersion in a circulating fluidized bed by using the classical discontinuous core-annulus model, where these two zones are characterized by different void fractions and gas velocities. They showed that these zones must have two different dispersion coefficients so that the tracer results could be explained. The present data support Werther’s observations, since when using the model described by eqn. (4) with U(r) = Ug = cte, the helium concentration profile fits the experimental data only in the core area. It is overestimated in the annulus area, especially at the longest distance. This is simply because a radial gas convection profile must be introduced with the Ostwald-de Waele law for example. In this case, a unique D, coefficient is sufficient since Fig. 4 shows that the fit is good

252

(% vol.)

n

source profile

phenomenon could not be explained. Note that the tendency would reverse when setting U(r> = U, = cte. A radial gas dispersion coefficient of 0.002 to 0.003 m2 S -’ can be chosen for every set of conditions, as Werther et al. [14] find the same order of magnitude in the core area, like Bader [16], in the same conditions as the present ones, using a radially constant superficial velocity on the whole section of a 0.305 m diameter riser. In conclusion, a simple plug flow model with gas velocity radial profile and a single radial dispersion coefficient is able to represent the gas flow in the cold set-up. The presence of the circulating solid tends to distort the gas velocity profile and to increase slightly the gas radial dispersion coefficient.

-

1 x

._g 3 z

0.8

8

exp. profile 1.57 m. above source

I .57

-

-

A

computed profile

x

exp. profile 4 m. above source

c

computed profile 4 m. above source -

-

m. above source -

-

-

d

E z g

0.8

0

2

4

8

8

Axial properties Some experiments were done in the industrial riser by injecting pulses of radioactive krypton (50 cm3) in the catalyst flow upstream of the feedstock injection. The passage of the tracer was followed by external detectors placed along the riser. The normalized bottom outputs were processed with the classical axially dispersed plug flow model to yield calculated outputs. A flat gas velocity profile is preferred to obtain the overall reactor performance, because a more complicated model would require knowledge of this profile and of the radial dispersion. The best fit with the top experimental data was found by optimization of the actual gas velocity and the axial dispersion D, to minimize the surface difference between the calculated and the target outputs. The results are shown in Table 1. The flow in the industrial riser is more complex than that in the cold set-up. In effect, the gas velocity varies because of the expansion due to the cracking, and the feedstock jet section is not well understood. Thus the measurements are made at least 4 m above this zone,

10

Riser radius (cm) Fig. 4. Experimental and computed radial profiles of helium concentration at U,=6.8 m-’ and F,= 189 kg m-* s-‘.

everywhere, even at the highest elevation studied and close to the wall, where helium has the possibility to disperse. The IZindex is found to be 1 and 0, is 0.0023 to 0.0024 mz s-l, depending on the target level. Figure 5 shows the variations of D, with superficial gas velocity and catalyst mass flux. It remains very close to that measured in the empty pipe, with a tendency to increase when the solid is circulating. At low solid flux and U, = 6 m s-l, the fit with the proposed model is worse and D, is higher (0.005 m2 s-l). This reproducible

Ug = 3.8 m/s

q l

s

.005 -

*

3 2 .004 ‘I ti 8 ,003 -

A Ug = 5 m/s l

Ug=6m/s

q

c

m

z

---

5

10 15 Solid flux I Gas flux

20

25

Fig. 5. Variation of the radial dispersion coefficient with the operating conditions of the cold set-up.

253 TABLE

1. Results

of gas radioactive

tracing in an industrial

cracking riser Axial gas dispersion (m* s-l)

Peclet number

(m)

Actual gas velocity (m s-l)

output surface difference (%)

298 298

8-14 8-14

11.3 11.2

5.2 5.1

13.0 13.1

-

325 325

4-18 4-18

14.2 12.2

5.6 5.6

35 30

9 11

Catalyst flux (kg m -* s-l)

Height of the detectors

where the jet influence is assumed to be negligible. The reproducibility is good, as two tracings at the same conditions show. The adjustment of the computed output to the target is not perfect, since the relative difference between the computer and experimental surfaces is ca. 10%. This is probably due to the gaseous expansion between the two detectors. Another explanation may be that the radial velocity profile is not flat, as shown above, and this influences the tracer behavior. The Peclet number is defined by V&/D,. Its values are high and the equivalent stirred cell has a length on the order of 0.7-l m, since the number of equivalent cells is approximately half the Peclet number. So the overall gas flow is close to a plug flow. This conclusion is the same as that of Bader et al. [16], who state that 75% of the total gas flow occurs in the central 45% of the total area, concluding that a plug flow is reasonable. It is important to remember that the gas radial velocity profile is related to the core-annulus structure and this has certain consequences in reactor applications.

Solid flow

Radial properties The results obtained at the 4 m elevation in the cold bed will be presented first using the tomographic technique [8, 91. Figure 6 shows an example of the density map determined in the cold model with U, = 6.3 m s-l and F,=202 kg me2 s-l. Table 2 shows a comparison of the averaged solid concentrations drawn from these measurements with those obtained from the pressure drop. They are identical within experimental error. The corresponding radial cuts are shown in Fig. 7. There is a thin dense annulus of 100 to 300 kg me3, far from the minimum fluidization density (900 kg me3, and the core is 20 to 50 kg mV3. The oscillations appearing on the cuts are due to the mathematical noise of the reconstruction because of the relatively low number of scans (15 translations X 24 rotations). Table 2 also presents the ratio of the mean radial density (determined by the radial y-ray absorptions) to the mean surface density (determined by tomo-

graphy). This ratio is close to 1.4 and tends to increase with the overall density, indicating that the core-annulus structure is more pronounced at higher densities, as is seen in Fig. 7. The 24 extreme chordal measurements from the tomography are at 0.006 m from the wall tangent in the cold riser. They give average chordal densities between 130 and 180 kg me3, indicating a good axial symmetry. When using an intrusive y-ray source such as the one described in [17], the same density profile is obtained except in the segment closest to the wall (0.01 m), which indicates a concentration of 350 kg rne3. This is attributed to the presence of a solid accumulation at the end of the probe, which would lead to an overestimate of the annulus density and consequently the whole bed density. In industrial risers, the same structure exists, as shown previously by Saxton and Worley [l] and in Fig. 8 obtained by the present authors (F,= 1090 kg rn-‘s-l and lJ,-25 m s-l) with 3 rotations x 9 translations tomography. In this case, the mean surface concentration of the catalyst is 122 kg rne3; the extreme values of the extreme chordal measurements (1.9 cm to the wall tangent) are 168 and 313 kg mm3 with an average at 270 kg me3. The core average concentration is 50 kg rne3. Catalyst velocity and flux profiles were estimated [7] by combining these density measurements with those from a momentum probe. The results show high flux values at the centerline and low (but seldom negative) values toward the walls. All these experiments show clearly that the core-annulus structure exists in the dilute conditions of the catalytic cracking. A number of authors [U-20] model circulating beds by a discontinuous two-phase pattern in which the gas velocity is zero in the annulus and concentrated solids are falling down. The whole flow of gas is rising in the core and solid particles are entrained at high flux with a slip velocity equal to the particle terminal velocity. There is a radial net flux of solid exchange between the core and the annulus, with a transfer coefficient which is adjusted to fit the experimental axial pressure drop profile. These models are able to represent the local solid flux, but they do

254

Catalyst apparent density

(kglm?

70

35

ldius

0

Fig. 6. Catalyst density map in the cold model, determined

by tomography.

TABLE 2. Catalyst concentrations by pressure drop measurements and by y-ray tomography in the cold set-up, under various conditions Air superficial velocity (m s-t)

6.3

6.3

4.2

Solid mass flux (kg m-* s-l)

202

308

114

Average catalyst concentration from delta P (kg mw3)

41

6.5

115

Average catalyst concentration from tomography (kg mm3)

34

68

110

Surface to diameter mean density ratio

1.27

1.38

1.41

not give any information about the solid radial mixing which should govern the solid backmixing in the bed. In effect, a core-annulus exchange limited to the net transfer would imply that the solid would fall down through the annulus from every elevation to the bottom of the riser, thus introducing a large backmixing. This would have important consequences in reactor applications. Data on solid radial dispersion do not exist, but some sampling obtained in an industrial riser enable us to estimate its importance: Catalyst and cracked hydrocarbons have been sampled 4 m above the feedstock injection point by a technique similar to that described by Schuurmans [17]. The samples have been taken at three locations, close to the wall, at mid-radius and at the centerline. Figure 9 shows the local cracking conversion and the coke

yield for two different feedstock injection conditions. In both cases, the final coke yield is 5wt.%. The figure shows clearly that the conversion is higher toward the wall. This must be attributed to the larger catalyst concentration at this location (smaller space velocity and higher temperature). By contrast the coke yield (defined by the catalyst coke content multiplied by the overall C/O) exhibits a flat profile. As the coke is catalytically produced and remains bound to the catalyst, it may be concluded that the solid radial mixing is sufficiently efficient to provide solid constant radial properties despite the gas and solid radial gradients of concentration. This point will be confirmed in the next section. Axial properties In the cold set-up, axial profiles of catalyst concentration are plotted in Fig. 10 at U,=5.2 m s-’ and various solid fluxes. They are determined by measuring the mean density along two perpendicular diameters (one in the plane of the return hopper and one of the riser). A 137Csy-ray absorption is used. Under every condition a density increase is observed, i.e. a solid slackening several meters up-stream of the top elbow. This effect is also visible in the axial pressure experimental profile (Fig. 11) for U, = 5.2 m s-l and F, = 202 kg m-’ s-l, if it is considered that the slope is proportional to the solid concentration. With this hypothesis, the pressure profile derived from the catalyst mean diametral weight shown in Fig. 10 is also plotted. When this weight is corrected by the 1.4 coefficient found in these conditions, as shown earlier (Table 2) to account for the difference between the surface and the diameter in the core-annulus structure, the new

255

-10

-6

-6

-4

-2

0

2

4

6

6

10

Radius (cm)

Fig. 7. Diametral

solid apparent

density profile

in the cold model, determined by tomography (operating conditions in Table 2).

Catalyst apparent density k-------T

1.00 199

0

,adius 1

Fig. 8. Catalyst density map in the industrial

riser, determined

calculated pressure data become very close to the experimental ones. This confirms that under the present conditions pressure measurements always give a good indication of the void fraction. The catalyst slowing down may be explained by the presence of the blinded tee, where the solid impacts. Large clusters may be created again and they may fall down along a large distance until they are entrained again. It is not known whether the core-annulus structure is recreated. Nevertheless, Ambler et al. [21] point out that an abrupt exit increases the solid hold-up and its backmixing. They explain some of their results due to the reinforcement of the core-annulus structure with this configuration. These findings on axial properties in the cold setup have been extended to the industrial riser by using a radioactive tracer. The lanthanum contained in the

by tomography.

cracking catalyst is activated by a neutron flux, and a small sample (5 g) is injected into the feedstock injection level. Its passage is recorded by external detectors. This enables the average velocity to be determined between the detectors and then the average concentration calculated by the formula:

The results are shown in Fig. 12. The conditions are those described in Table 1 for the solid flux of 325 kg mm2 s-l. It appears that the void fraction increases in the first third of the riser, because of the solid and the gas accelerations. It starts to decrease in the second half, so that the solid average concentration is doubled in the last 8 m despite the gas continuing to accelerate due to the cracking. This is probably related to the

256

Conversion wt

J Coke yield

wt’X 40

10

30

5

. _.

_ _

o-

:.. _. _ G

I

Feedstock

injection

cund.

Feedstock

injection

cond. 2

.A’

/

o;?

20

0

-R

+R

-R

+R

0 Riser’s relative

Fig. 9. Radial profiles of cracking conversion

and coke yield in the industrial

Ug = 5.2 mls

-

I

150:

radius

riser.

* 216 kg/s/m2

-- n 186 kg/s/m2 -. - 0 156 kg/s/m2

I\

% ‘MlOO. S Ii; 50.

_ ‘..

x***-

..x

0

/ ^ _

o-.-‘A..

,..

,......

A

.’

2

0

2

4

6 (m)

6

10

solid density along the cold riser for various solid fluxes at U&=5.2 m s-l.

4

6

6

Height

Fig. 11. Variation

‘..-‘A 124 kg/s/m2 SIX 34 kg/s/m2

Height

Fig. 10. Mean diametral

...”

* *.x

of the relative pressure

(m)

along the cold riser.

10

257

30

20

Height (m) Fig. 12. Average

TABLE

3. Results

Catalyst flux (kg m -* s-Y

catalyst concentration

along the industrial

of solid catalyst radioactive Height of the detectors (m)

riser from a radioactive

tracing in an industrial Catalyst actual ‘velocity (m s-‘)

tracing.

cracking riser Axial catalyst dispersion (m’ s-‘)

Peclet number

outputs surface difference W)

298

298 325 325

8-14 8-14 4-18 4-18

8.7 9.8 10.7 11.5

design of the terminal elbow, which is of the blinded tee type. Thus, the solid flow is shown to become never established with such a classical design, neither in the 11.7 m cold set-up, nor in the 26 m industrial riser, because the presence of the elbow has repercussions on a long upstream part. The axial dispersions are computed by the same methods as for the gas. The results are summarized in Table 3. The reproducibility is still acceptable, and the adjustment of the dispersed plug flow model seems to be better than for the gas. The slip velocity is in the order of 3 m s-l. The backmixing is higher than that for the gas, though the results are more disperse. This is consistent with the core-annulus structure shown in Fig. 8. Nevertheless, the length of the equivalent stirred cell is less than 2.5 m. This confirms that the solid flow can be considered to be close to the plug flow over long distances, and that the radial mixing between the core and the annulus is high enough to consider the solid radial properties to be constant.

10.4 12.2 17.8 10.1

5.0 4.8 8 16

5 4

Ambler et al. [21] traced the solid circulating in a 0.05 m diameter X 3 m high bed with a blinded tee elbow. They confirm the higher solid holdup with this configuration and find an overall bimodal residence time distribution which is negligible in our large scale experiment. This distribution is attributed to the core-annulus structure which seems to be reinforced by the abrupt exit. These results are not inconsistent with the present ones, if the short length of their device is considered and if it is remembered that a dense bed is probably present, according to formula (1).

Conclusions The behaviors of gas and solid flows in dilute conditions such as in catalytic cracking risers are similar when comparing the cold-set up with the industrial operation. The gas flow is close to plug flow, but the small scale experiments show a parabolic radial velocity profile with a radial dispersion coefficient close to one with a zero solid flux.

258

In the median zone, the solid is more backmixed than the gas, because of the core-annulus structure. As this structure is less pronounced than in a dense circulating bed, the flow can be considered as plug flow in a first approximation. No significant refluxing solid is detected at the walls of the industrial plant, and the radial mixing between annulus and core is intense enough to consider constant radial properties. The presence of an abrupt exit (blinded tee) has an important influence, since the solid holdup is doubled in the last quarter of the riser and the solid flow becomes less established. This effect must be accounted for when designing or modeling a circulating fluidized bed reactor.

References

5 6 7 8

List of symbols

9 10

A and B C c/o C, D, Q F L n k u V

constants defined by formula (1) helium molar concentration (km01 mW3) catalyst to oil mass flux ratio solid (catalyst) apparent density (kg m-“) gas axial dispersion (m2 s-l) gas radial dispersion (m2 s-l) mass flux (kg mP2 s-l) axial length in the riser (m) index defined by formula (3) radius coordinate (m) riser radius (m) superficial velocity (m s-l) actual velocity (m s-l)

Subscripts gas gas solid (catalyst) S terminal t

11 12 13

14 15 16 17 18 19 20 21

A. L. Saxton and A. C. Worley, Oil Gas J., 68 (1970) 84. A. A. Avidan, M. Edwards, H. Gwen, 6th Int. Con$ Fluidkation, Banff, Canada, May, 1989. J. Yerushalmi, D. Turner and A. M. Squires, Ind. Eng. Chem. Process Des. Dev., 15 (1976) 47. L. Monceaux, M. Azzi, Y. Molodtsoff and J. F. Large, in P. Basu (ed.), Circulating Fluidized Bed Technology, Pergamon Press, Oxford, 1986, p. 185. M. J. Rhodes and D. Geldart, Fluidization, Engineering Foundation, New York, 1986, p. 281. D. Geldart, Gas Fluidization Technology, Wiley, New York, 1986, p. 144. M. Azzi, P. Turlier, J. F. Large and J. R. Bernard, 3rd Int. Con& Circulating Fluidized Beds, Nagoya, Japan, Oct. 1990. M. Azzi, P. Turlier, L. Garner0 and J. R. Bernard, Powder Technol., 67 (1991) 27. L. Desbat and P. Turlier, 13th IMACS World Congr. Computation and Applied Mathematics, Dublin, 1991. J. R. Bernard, H. Santos Cottin and M. Margritta, Katalistiks 5th annual FCC Symp., Vienna, May, 1984. J. W. Van Breugel, J. J. M. Stein, and R. J. de Vries, Proc. Inst. Mech. Eng., 184 (1970) 18. Cited in N. Midoux, Mecanique et Bheologie des Fluides en Genie Chimique, Lavoisier, Paris, 1980, p. 170. H. Weinstein, M. Shao, M. Schnitlein and R. A. Graff, Fluidization, Engineering Foundation, New York, 1986, p. 329. J. Werther, E. U. Hartge, M. Kruse and W. Nowak, 3rd Int. Conf. Circulating Fluidized Beds, Nagoya, Japan, Oct. 1990. J. Werther, E. U. Hartge, M. Kruse and W. Nowak, 4th World Congr. Chem. Eng., Karlsruhe, June, 1991. R. Bader, J. Findlay and T. Knowlton, Circulating Fluidized Bed Technology ZZ, Pergamon, New York, 1988, p. 123. H. Schuurmans, Znd. Eng. Process Des. Dev., 19 (1980) 267. W. Yang, 3rd Znt. Co@ Circulating Fluidized Beds, Nagoya, Japan, Oct. 1990. F. Berutti and N. Kalogerakis, Can. J. Chem. Eng., 67 (1989) 1010. M. J. Rhodes, Powder Technol., 60 (1990) 27. P. A. Ambler, B. J. Milne, F. Berutti and D. S. Scott, Chem. Eng. Sci., 45 (1990)

2179.