Solids motion at horizontal tube surfaces in a large gas-solid fluidized bed

SOLIDS MOTION AT HORIZONTAL TUBE SURFACES IN A LARGE GAS-SOLID FLUIDIZED BED J. P. K. PEELER and A. B. WHITEHEAD* Commonwealth Scientific and Industrial Research Organization, P.O. Box 312, Clayton, Victoria 3168, Australia (Received

23 March 1981; accepted 4

May

1981)

Abstract-The motion of fluidized particles at the surface of 38 mm diameter horizontal tubes, immersed in a 1.2 m square bed of silica sand (&, = 0.3 m/s), iluidized at 0.9 m/s has been observed using photographic techniques (2OOfps). It is shown that the solids motion is different for centrally located tubes and those adjacent to a side wall. Data on particle velocity and surface contact arc presented and the degree of particle contact at various zones around the tube circumference is shown to vary in a similar manner to published localized heat transfer rates. INTRODUCTION

is generally accepted that the relatively high heat transfer coefficients obtained in fluidized beds are the result of the scrubbing action or contact of solids with the heat transfer surface. Of particular interest is the transfer of heat between fluid&d beds and horizontal tubes, reviewed by Gel’perin and Einstein[l], Botterill[2] and Saxena el a1.[3]. Examining the literature, Chen[4] found little agreement among the various reported heat transfer correlations and data and, from the viewpoint of practical applicability, called for an improved understanding of the phenomenological mechanisms involved. This, he found, was especially true in the case of horizontal tubes as distinct from vertical tubes. Some investigators had previously utilised photographic techniques to study particle behaviour in various fluid bed-tube configurations as detailed in Table 1. In their early photographic study, Massimilla and Westwater[5] recorded the motion of fluidized particles at the walls of a glass column. They reported a net downward particle flow with some individual particle velocities four times greater than that of the superficial gas velocity. Glass and Harrison[6] using tine photography to study flow round a horizontal cylinder in a 2D fluid bed of thickness 1Omm at air velocities between 2 and 3 times incipient fluidization, reported the existence of three distinct flow regimes round the cylinder: viz. a thin film of air on the upstream surface, an almost stagnant defluidized region at the downstream surface and a region near both ends of the horizontal diameter of the cylinder where the upstream air flow formed into chains of bubbles whose size was small relative to the diameter of the cylinder. Using an endoscope, Noak[7] observed particle movements round a horizontal transparent tube in a 300 mm diameter bed of 600 Frn diameter glass beads. He qualitatively related these observations, viz. strongly fluidized motion at sides and a weak sliding movement on the downstream face of tube, to the variation in heat transfer coefficients measured at points round the circumference of a tube in a similar experimental layout. It

*Author to whom correspondence should be addressed.

Rowe and Everett[S] used X-ray techniques to study the influence of solid surfaces on bubbles within a three dimensional bed. Amongst the several surfaces studied were horizontal rods of 3 and 6 mm diameter on which single bubbles could be generated at gas flow rates just greater than the minimum fluidizing velocity, whilst at higher flows the bubbles almost permanently enveloped the rods, leading to diminished heat transfer. In their three dimensional bed, Roooey and Harrison[9] observed the downstream flow of sand particles associated with a 37 mm tube. They inferred that the gas cushion often observed in two dimensional equipment was not so permanent or extensive in their three dimensional bed. Likewise when testing a seven tube array they reported a decrease in the stability of the defluidized cap and reduced particle contact at the sides of the tubes. Working with a two dimensional bed, depth 19mm, Hager and Schrag[lO] photographically recorded particle flows of 560 pm glass beads downstream of cylinders of 39 and 52 mm diameter. They observed the bulk circulation pattern through the previously mentioned defluidized cap on the downstream surface and determined its size and particle residence time, defined as “total turnover time”, for a number of superficial gas velocities. They found that fresh particles entered through the top of the cap and gradually worked their way downward towards the cylinder surface, along which they scoured until they reached the horizontal points of bubble emanation and were swept back into the bulk phase. Whilst carrying out heat transfer tests on tubes, Cherrington et al. [ 1l] made visual observations in two and three dimensional units. They defined three distinct zones of solids to tube contact: an upstream zone of lighter than normal bed density but not gas shrouded, a downstream or cap zone of regular side to side sliding motion, frequently swept clean by bubbles and a third zone upstream of the cap zone where the particles retreated and advanced like “ocean surf”. This latter area coinceded with the area of highest heat transfer. An optical fibre sensing probe was used by Lyamkin et ai.[12] to detect the presence of a bubble at a number of positions around the circumference of a glass tube

78

J. P. K. PEELER and A. B. WHITEHEAD

37mm in diameter, immersed horizontally in a 3D fluidized bed of 0.2-0.8 mm glass microspheres. The light transmitted by individual fibres was converted to an electrical display and photographed at 94 frameslsec to provide information on the frequency of upstream gas film replacement, displacement of the downstream stagnant cap and bubble rise rate at the bed centre and outer wall. Although their results only reinforced much that has been stated above, the technique may have potential for rapid electronic processing of information as might be provided by computerized data logging systems. Loew et a/.[131 used a quasi-stereoscopic method for mapping particle velocities round various obstacles in a 2 D bed of large particles, 2OOC2380pm, not far from incipient conditions. In observing single obstacles they reported the three characteristic zones previously noted by Glass and Harrison above. However, when examining four tube square and three tube triangular arrays they found that a stable air pocket bridged across the lower tubes when closely spaced, but with wider spacings the individual tubes behaved similarly to single tubes. When 10 tubes were arranged in a triangular wide spaced array the bridging air pocket across the lower tubes still formed and the upper row of tubes hindered free particle flow when compared to the 3 tube array case. In this case the air bubbles, which were smaller in size than with the other arrays, sometimes divided on collision with a tube. In an effort to achieve a uniform voidage fraction round tubes in a bundle Fakhimi and Harrison1141used a capacitance technique to study the voidage fraction round tubes in a two dimensional bed. Testing both square and triangular pitches they found that a tube face to face distance of three tube diameters gave the most uniform viodage distribution. As observed in earlier two dimensional studies, they also reported a thin air “cushion” upstream of the tube and a tendency for some bubble nucleation to take place at the tube sides. Considering the foregoing investigation it is apparent that there are at least three distinct regions of interest, with respect to particle movement, around tubes in the systems studied. The behaviour of these zones varies however depending on whether the bed configuration is of the two dimensional or three dimensional type and the degree of fiuidization, i.e. gas velocity. In the 2D bed there exists a fairly persistent defluidized cap on the downstream surface of the tube, the extent and permanence of the cap depending on the fluidizing velocity; a thin gas film at the upstream face which forms into small bubbles at the tube horizontal diameter points; and a region between these two where the tube sides are contacted by particles in a fluidized state. It would appear however that quite different behaviour is connected with a three dimensional bed and, even though the maximum dimension of those beds reported was only 3OOmm,the defluidized cap and upstream gas film were found to be neither as large nor as permanent as those observed in two dimensional beds. Further, the investigations reported above are based on the operation of laboratory scale equipment and the problem of scale-up remains uncertain from the aspect of commercial scale equipment and its operating conditions.

Solids

motion at horizontal tube surfaces in a large gas-solidfluidizedbed

79

As particle movement is a function of the particle and gas properties, fluidization velocity, gas distribution and bed geometry it is necessary that experimenfs be carried out using equipment of a size that allows gas bubbles of similar size to those formed in full scale industrial systems to develop. Under these conditions bubbles typically encompass many tubes. This paper describes an investigation of particle movement at tube surfaces and relates the finding to those of other workers studying heat transfer between bed and lube. t GAS FLOW

EXPERlMENTALAPPARATUS

A fluid bed vessel of internal cross section 1.22m square, having a gas distribution plate containing 36 tuyeres described as type B by Whitehead and Dent[lS] and 92 tubes fitted through opposite side walls was used in the investigation. The tubes, having an outside diameter of 38 mm o.d., were arranged as shown in Fig. 1 in 8 rows of staggered square pattern, having 102mm horizontal and vertical centre spacings. For lhe purpose of observing and photographing particle behaviour on the tube surface, tubes were prepared containing glass window sections fitted Rush with the outside of the tube wall. Two of these tubes were inserted into the tube bundle at the positions indicated in Fig. I.

Fig. 2. Location of the circumferential zones used in the description of particle contact at the tube surface.

Adjustable end mountings for these tubes enables their rotation through 360”and lateral positioning within the bed. Using a pair of inclined mirrors, light from a mercury lamp, optic fibre light source was directed by the first mirror through the window section and an image of the area adjacent to the window received via the second mirror. In this manner flow at the tube surface could be observed or recorded using tine photography of either 24 or 200 frames per second. For the purpose of defining a particular portion of the tube surface a system of circumferential zones was adopted as shown in Fig. 2. OPERATING

38mm

102

O.D. K 122Dmm x102mm

LONG

PITCHED I

-1220mm

1

CONDITIONS

The fluid bed vessel was filled to a depth of 1.2 m with silica sand, having a particle size analysis and properties described in Table 2, and fluidized using air at a superficial gas velocity of 0.91 m/s which produces bubbles whose diameters are large relative to those of the tube. For this particular system and set of operating conditions, Sitnai et 01.[16], have determined, using paired differential pressure probes located within the tube bundle, a mean bubble diameter of 0.5 m, rising with a velocity of 1.8m/s at a frequency of 1.85Hertz. Complimentary investigations using solid tracer addition had shown that a persistent stream of descending solids was located next to the vessel wall parallel to the tube assembly whilst solid motion remote from the wall

Table 2. Properties of the bed material: silicasand

WNDOW

SECTKXI

Incipient (Superficial) fluidizing velocity

0.30 m/s

Incipient

0.39

Particle

porosity density

Bul k density Size >1190

Fig. 1. Fluid bed layout, showing dimensions and

location of the

central tube (1) and the wall tube (2).

distribution

2580

Kg/m3

1610

Kg/m3

% wt 0.3

Pm

1190

- 840

18.0

840

- 590

50.7

590

- 420

28.3

420

- 297

2.4 0.3

<297 average

particle

size

700 pm

J. P. K. PEELERand A. B. WHITEHEAD

80

was more random. Thus Tube one was located with its viewing port as shown in Fig. 1 to investigate behaviour in the main body of the bed whilst Tube two was located in the persistent solid stream at the vessel wall.

zone I does not fall away as quickly as the other zones but remains at an intermediate value. Mass transfer must be enhanced within the bubbles due to the tubes inducing the solids motion described above and also below in Stage C.

RESULTS

Centre located tube (Tube 1)

Examination of both the high speed (200fps) and low speed (24 fps) films indicated that there was a reasonably predictable behaviour pattern connected with each bubble transit. The high speed sequences were taken with a shutter adjustment such that each frame was exposed for 1.25ms and enlargement and printing of several hundred sequential frames enabled the estimation of velocity of particles moving at speeds up to 0.2 m/s from positional change on successive frames, whilst velocities of faster moving particles were estimated from streak lengths on individual frames. The degree of particle contact with the tube surface was estimated from the enlarged prints by counting the particle contact density. This information is presented in Figs. 3 and 4, where for convenience the transit of a bubble past the tube has been divided into four distinct stages: A. Dense phase movement preceding bubble arrival, B. Bubble arrival at the tube, C. Bubble envelopment of the tube, and D. Bubble departure and reappearance of the dense phase. Stage A

The initial part of the bubble cycle begins with the tube surrounded by closely packed particles, dense phase, slowly moving upwards in zones II and III at a velocity of approximately 20 mm/s. The particle contact density is similar to that of the bed in the unfluidizcd stage. Zones I and IV also experience this slow upward drifting movement although at the front and rear stagnation points the particles appear stationary except for a gentle vibrating movement. This upward drift continues, increasing gradually in velocity to round 350 mm/s when the bubble approach is indicated by a sudden increase in particle velocity. Stage B Bubble approach is first evident round the region where zones III and IV join. Here the particle velocity increases rapidly and particles streak across the tube surface. Accompanying this velocity change is a lowering of the particle contact density to a lean phase condition on zones II, 111and IV. Particles streak across these zones initially in an upwards direction but then in a random manner at velocities up to 5.6m/s. Particles in contact with the tube in zone I remain substantially intact, forming the much referred to “cap of defluidized particles”. These particles are not stationary but slide slowly downwards across zone I to either side of the tube vertical centre line until they approach zone II where they are swept off along with particles streaking across zones II and III. The particle contact density in

Stage C In comparison to the previous stage this stage is rather quiescent, the bulk of the dense phase has been removed and residual particles continue sliding downwards from zone I, moving at velocities between 250 and 500 mm/s. The particle contact density is further reduced in zones III and IV until finally zone IV has been swept free of particles. In zone II, however, the particle contact density rises as particles from zone I continue to slide down across the tube surface, They advance as a sharply defined front across zone II, having a particle contact density of intermediate value, probably the phenomenon referred to as “surf action” by Cherrington et al. This curtain of particles continues downwards across zone II until it is covered by particles at an intermediate packing density. Some particles traverse across into zone III and briefly increase the particle contact density there, before they are swept away as they reach the vertical tangent line at the zone II/III interface. The bulk of the material forming the cap on zone I is gradually depleted and towards the end of this stage the remaining particles are swept away, leaving the top of the tube bare. Stage D The quiescent period ends abruptly with the departure

of the bubble and particles streaking across zones IV and III at up to 5.0mls. The tube is quickly surrounded by the dense phase which continues its slow upward drift prior to the arrival of the next bubble. There was no evidence of a bubble “wake” moving at speeds approximating to the bubble rise velocity of approx. 1.8 m/s. Wall tube (Tube 2)

Cine-filmstaken from inside the tube in position 2 in the direction of the wall showed a predominantly downward particle flow. In general the flow occurs in a pulsing manner similar to that studied by Drinkenburg et a/.[171 and termed by them “stick-and-slip-flow”. A typical movement cycle is included in Fig. 3 beginning with the tube surface in contact with closely packed particles slowly moving downwards at velocities around 20mm/s. This continues for a short period until the dense phase comes to rest round the tube. Particles in contact with the lower half of the tube, zones III and IV quickly accelerate downwards leaving as it were a bridged-arch between the wall and the tube. This dense phase/lean phase interface slowly retreats upwards as particles break away at velocities round 100mm/s. This continues until the tube surface is left bare. After a short period of void round the tube, dense phase descends to surround the tube and moves downwards at a velocity around 20 mm/s.

Solids motion at horizontal tube surfaces in a large gas-solid Ruidized bed

-0

0.1

0.2

0.3

04

81

0.5

TIME (SEC!?.)

Fig. 3. Typical variation in surface particle velocity with time for the central and wall tubes.

‘DISCUSSION

Particle motion and it’s relationship to heat transfer In summarizing the work of a number of investigations Saxena et a[.[31 concluded that it would appear in general that at fluidizing velocities just greater than incipient, the maximum heat transfer coefficients are found in the region of the horizontal tube diameter whilst the minima occur at the upstream and downstream regions associated with the gas film and stagnant cap of solids, respectively. With jncreasing fluidizing velocity, the heat transfer coefficients round the entire circumference increase, particularly at the upstream and downstream regions, with the maximum value moving from the horizontal diameter region downstream into zone II of Fig. 2. The overall general increases were considered to be due primarily to the increased particle contact at the tube surface and the greater mobility and more frequent replacement of the stagnant cap by the increases size and number of bubbles. The extent to which the circumferential zones are contacted by the fluidized particles, shown in Fig. 4, and

ditions, that the maximum coefficient occurs above the horizontal diameter, with the minimum at the upstream stagnation point (Noak[7] and Cherrington et al.[ll]). Motion in wake region Rowe and Partridge[l9],

have shown using X-ray

techniques that small bubbles rising in silica sand with

the degree of particle contact during a bubble transit can be evalutated by integrating graphicatly under the particular curves, thus the relative particle contact for the

similar properties to that used here have a wake fraction of approx. 0.2. Estimates based on this data indicate that a sequence of approx. 15 highspeed tine frames showing random particle movement might be expected at the final stage of bubble transit past the tube. Such a sequence was not observed and the high velocity particles shown in Fig. 3 at the final bubble stage were only noted for a sequence of three frames, appearing as a narrow band, and were interpreted as particles streaking across the dense phase-lean phase interface. The present investigation is concerned with much larger bubbles, approx. OSm dia., in a system containing horizontal tubes and it is not known at present whether the absence of a substantial volume of fast moving solids immediately following the bubble void is due to the presence of the tubes or is an inherent feature of large bubbles in coarse material.

four circumferential zones, in order I to IV, is 67%, 72%, 45% and 50%. These values are in general agreement with the great bulk of data on heat transfer coefficients, especially the observations at similar operating con-

Maximum particle velocity The maximum velocity of particles, observed in stages B and D of bubble transit, is relativeIy high when com-

TIME (SEC9

Fig. 4. Variation in particle contact density at various locations on the centre tube surface during a bubble transit. CES “0,. 37, No I--F

J. P. K. PEELER and A. B. WHITEHEAD

82

pared to the measured bubble rise velocity of 1.8m/s.

However, as the terminal settling velocity of the mean sized particle is round 5 m/s and it has been previously shown that fluidizing gas passes through the bubble at velocities up to three times the incipient fluidizing velocity it is quite likely that the combination of these velocities, together with any effect contributed by the tube array, is responsible for the high degree of turbulence observed. CONCLUSIONS

For the particular system studied it is conctuded that: (1) Solid motion at the surface of a horizontal tube is dependent on tube location. Tubes located in the main body of the bed experienced a cyclic behaviour pattern as they encountered each rising bubble. This resulted in a systematic variation of the solids concentration at various circumferential locations on the tube surface. These variations corresponded to gradations in heat transfer coefficient noted by other workers. Those tubes situated in the persistent downflowing solids stream ad jacent to the vessel wall, experienced a “stick-slip” solids descent pattern with some indication of transient void formation below the tube. (2) No evidence of a wake fraction of solids moving at the bubble speed was noted. (3) The maximum speed of solids movement recorded at the tube surface was approx. 5.6 m/s. (4) The presence of tubes induced gas solid contact within the bubbles. (5) No evidence of persistent the tubes was observed.

void formation below

(6) The upper surface of the central tube was only intermittently covered by a cap of defluidized particles which was swept away during each bubble transit.

REFERENCES

[l] Gel’perinN. I. and Einstein V. G., Davidson, J. F. and Harrison D. (Eds.) Fluidizotion. Chao. . IO.Academic Press, New York 1971. ’ [2] BotterillI. S. M., Fluid-Bed Weat Transfer.AcademicPress, New York 1975. 131Saxena S. C., Grewal N. S., Gabor J. D., Zabrodsky S. S. _ and GalershteinD. M., Advnnces in Heat TmnsJer, vol. 14, p. 150.Academic Press, New York 1978. [4] Chen J. C., A&W-AK% Heat Transfer Conf., St.Louis, 1974,ASME Paper 76-HT-75. [5] Massimilla L. and Westwater J. W., A.I.Ch.E J. 1960 6(l) 135. [6] Glass D. H. and Harrison D., Chem. Engag Sci. 196419 1001. [7] Noak R., Cbemie Ing. Techn. 1970 42(6) 371. [S]Rowe P. N. and Everett D. J., Trans. Jnstif. Chem. Engrs. 197250 42. [9] Rooney N. M. and Harrison D., Keairns D. L. (Ed.) Fluidilation Technology, Vol. II, p. 3. HemispherePub. Corp. 1976. [lo] Hager W. R. and Schrag S. D., Chem. Engng Sci. 197631 657. [ll] Cherrington D. C., Golan L. P. and HammittF. G., Proc. Sth Int. Coni. on Fluidized Bed Combustion, Vol. III, P. 184. WashingtonD. C. 1977. [12] Lyamkin V. A., Gel’perin N. I., Ainshtein V. G. and Novobratskii V. L., J. Engng Phys. 197835(5)1348. [13] Loew 0.. Shmutter B. and Resnick W., Powder Tech. 1979 22 45. [14] Fakhimi S. and Harrison D., Trans. Instit. Chem. Engrs. 1980 58 125. [IS] Whitehead A. B. and Dent D. C., DrinkenburgA. A. H. (Ed.), Proc. Int. Symv. on Fluidization, p. 802. Eindhoven, 1967,Netherlands University Press. [16] Sitnai O., Dent D. C. and Whitehead S. B., submitted to Chem. Enann Sci.. Prelim. Commun. 1171Drinkenb;rg A. A. H., HuigeN. .I. J. and RietemaK., 3rd ht. Heat Transfer Conf.. Vol. IV. Pauer 147.Chicano 1966. [18] Chandran R., dhen J. ?. and Staub ‘F.W., I. Hear Trans. 1980 102(Z)152. [19] Rowe P. N. and Partridge B. A., ‘Trans. Instit. Chem. Engrs. 196543 157.