Ind. Eng. Chem. Process Des. Dev. 1985, 2 4 , 430-433
430
Heat Transfer between the Fluidized Bed and the Distributor Plate Guo-Tal Zhang' and Fan Ouyangt Department of Chemical Engineering, Oregon State University, Cowallis, Oregon 9 733 1
Heat transfer coefficients were determined for the transfer of heat from a fluidized bed to its dlstributor plate. The effects on the heat transfer coefficientof bed helght, bed temperature, particle size, and air velocity were determined. Particles used were iron, zirconia, sand, silicon and polystyrene spheres; air was used for fluidization. Rapid fluctuations of the distributor surface temperature were observed which depended on the air velocity and partlcle size. A mechanism of heat transfer between fluidized bed and the surface of the distributor plate was proposed and an empirical equation involving three dimensionless groups was correlated from 100 experimental data of five different particles.
Introduction Three types of heat transfer may occur in a fluidized bed: (a) heat transfer between gas and solid particles of the bed,(b) heat transfer between the bed and the bed wall or intern&, and (c) heat transfer between the bed and the distributor plate. While heat transfer coefficient of types (a) and (b) have been extensively reported in the literature, heat transfer between the bed and the distributor plate have only recently become a subject for investigation, primarily as a result of a new method for preparing highpurity silicon (&ha@ et al., 1982). Here cold silane enters a hot fluidized bed of silicon particles, decomposes, and deposita fine silicon dust on the particles which then grow in size
heat
SiHJcold gas) Si1 + 2H2(gas) In this operation it is imperative that the distributor plate be kept cool; otherwise decomposition will take place within the distributor plate and thereby plug it. Knowledge of the heat transfer coefficient between bed and distributor plate tell whether the plate can be kept cool enough (Levenspiel et al., 1983). A variety of particles were used in our study whose purpose it was to develop a general correlating expression between the heat transfer coefficient and the system variables. The experimental investigation has determined the effect on h of the following variables: air velocity, size of the fluidized solids, density of the fluidized solids, and bed temperature level. Mechanism of Heat Transfer between t h e Fluidized Bed and the Surface of Distributor Yoshida et al. (1969) have proposed a mechanism of heat transfer between fluidized beds and wall surfaces which includes both steady-state conduction of heat through an emulsion layer a t the wall and the unsteady-state absorption of heat by emulsion elements. However, the mechanism of heat transfer between the distributor plate and a fluidized bed has not been discussed previously in the literature. It should be pointed out that here the direction of heat transfer is opposite of the direction of air flow rather than normal to the direction of air flow, as is the case in most heat exchangers. In comparing these mechanisms we recognize that these are two contrasting situations, as shown in Figure 1. In one case there exists a film of gas at the heat transfer surface. However, in the *On leave from East China Institute of Chemical Technology, Shanghai, China. On leave from Institute of Chemical Metallurgy, Academia Sinica, Beijing, China. 0196-430518511124-0430$01.50/0
situation considered here, this film is absent, and heat is transferred to the distributor plate primarily by impingement of hot bed particles on the surface. Even though the film is absent, we may represent the heat transfer rate by a heat transfer coefficient, h, which depends on the bed geometry, gas and particle properties, and fluidization characteristics which in turn affect the collision frequency of particles with the distributor plate. This h is found as follows. Referring to Figure 2, the temperature of the distributor plate must be such that
81+ Q3
=
(1)
Q2
or rjlairCp,air(Teir,in)
+ hAdis(Tbed -
Tdis)
= rjlairCp,airTdis
(2)
therefore
- Tair,in) - Tdis)
mairCp,air(Tdis
h=
Adis(Tbed
(3)
where it is assumed that the air leaves the distributor plate at the distributor plate temperature, and the whole distributor plate is at one temperature. It must be pointed out that the temperature measurement of the air which just leaves the distributor plate is difficult. However, experimental results show that the temperature difference between distributor plate and the air just leaving it (-1 mm above the distributor plate) is never more than 3 "C. Hence the above assumption should be reasonable. Experimental Section Apparatus and Material. Figure 2 is a sketch of the fluidizing apparatus. Air at room temperature supplied by a compressor was passed upward through a bed of solids contained in an insulated 152-mm Pyrex pipe. The bed rested on a 1.5 mm thick distributor plate made of sintered stainless steel with nominal 5-pm pores, obtained from Mott Metallurgical Corp. Outlet air passed through a disengagingsection and then through a fiiter which capture any of the fines which may have elutriated from the bed. Heat was supplied to the bed by a 1000-W electric heating coil placed just above the bed surface where it would not distort the normal fluidizingbehavior of the bed, and temperatures were measured at various points in the system, as specified in Figure 2, with 0.25-mm chromelalumel thermocouples, and recorded using an Esterline Angus PD 2064 data logger. According to Levenspiel (1984), for irreguIar particles with no seemingly longer or shorter dimensions, the proper mean particle size to use for fluidization studies is dp
=
ddscreen
0 1985 American Chemical Soclety
(4)
Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 2, 1985
4-4
gas f i l m
431
Table I. Properies of Particles
cold
hot bed.
heat to wall f r o m gas and p a r t i c l e s
cold distributor plate
..
heat t o p l a t e from particles
II
0
k
4
P
I
' / / y, / / v, / x. / N, / / /
part. type sand
direction of gas f l o w
direction of gas flow
(a)
(b) Figure 1. (a) In heat transfer to bed walls or to bed internals a gas film is present at the surface. (b) In heat transfer to a distributor plate no gas film is present.
Polystyrene silicon iron zirconia
a i r to f i l t e r a
density, dcm3
min fluidiz veloc. umf, cm/s
0.0296 0.0204 0.0158 0.0133
2.61 2.61 2.61 2.61
9.2 4.6 2.9 2.4
0.80 0.80 0.80 0.80
16-40
0.0778
1.06
13.4
1.255
40-200
0.0161
2.33
3.0
0.816
50-70 70-80
0.0204 0.0158
7.8 7.8
15.6 9.9
0.46 0.46
50-70 70-80
0.0204 0.0158
5.9 5.9
12.4 7.3
0.44 0.44
part. size, mesha
mean dim, a,, cm
40-50 50-70 70-80 80-100
p..
heat cap. (at 38 "C) C,, J/(g K)
U.S.Standard.
(a) (b) (C) (d) Figure 3. Different method to measure the temperature of the surface of the distributor plate.
Figure 2. Location of temperature probes in the fluidizing apparatus and heat balance around the distributor plate.
where 4 is the sphericity, estimated to be 0.80 (Kunii and Levenspiel, 1977) for all particles used except polystyrene is the average screen diameter given by solids, and dscreen (5) Table I gives information concerning the particles which were used in this study. Procedure. The experiments reported in this paper were designed to determine the effects of the variation of air velocity, bed height, particle size, and bed temperature on the heat transfer coefficient from the bed to the distributor plate. The steady-state experimental test was used here. Air at room temperature was passed upward through a bed of particles a t a known fixed rate to obtain a fluidized bed. The amount of energy input to the heating coil above the bed was measured by a wattmeter. When the system came to thermal equilibrium, the inlet and outlet air temperatures, bed temperature, and distributor plate temperature were measured. Since temperature measurements provide basic data for determining h, it is therefore important to develop a means of measuring the temperature of the top surface of the distributor plate without undue influence of the bed particle temperature. Figure 3 shows four different methods which were looked into to measure the temperature of the top surface of the distributor plate. I t was immediately found that method (a) of Figure 3 was un-
satisfactory because the thermocouple was much too responsive to the bed particle temperature. In methods (b), (c), and (d) the thermocouples were brought through the distributor plate from below in order to eliminate the effect of the hot bed temperature, but in method (b) the temperature measurement was still affected by the hot particles above the distributor. For method (c) the temperature measuring point was not imbedded sufficiently in the distributor plate. Method (d) was the method finally used in all runs. Here the thermocouple wire was placed in a shallow groove in the top surface of the distributor plate and an epoxy cement was used to seal the gap around the thermocouple wire. It was observed that the distributor surface temperature fluctuated rapidly with the presence of air bubbles at the plate. As a consequence, a time averaging process using 8-10 ponita was necessary to more accurately establish the mean surface temperature of the plate.
Results and Discussion Table I lists the data obtained in 100 runs on 5 different types of solids. The effect of the variables on h are as follows. Bed Height. Figure 4 shows that h seems to be independent of bed height. The heat transfer coefficient rises sharply at the onset of fluidization. The initial rapid rise in h is explained by the increased solid circulation, whereas the decrease beyond the maximum is attributed to the lowered solid concentration at high gas velocity. Figure 4 is similar to Figure 10.12 of Gelperin and Einstein (1971) and leads to identical conclusions regarding the influence of bed height on the heat transfer coefficient. Such similarities suggest that the bed-distributor heat transfer is similar to the bed-wall heat transfer. Bed Temperature. To determine the effect of bed temperature on h it was necessary to make runs at different power levels or different bed temperatures and hold all other independent variables constant. This procedure
Ind. Eng. Chem. Process Des. Dev., Vol. 24, No. 2, 1985
432
0" 0
Hbed = IO cm
'-1
Particle, size (mesh) iron
0 50-70
V 70-80 A 50-70 $8 70-80 polystyrene x 16-40 18
zirconia
200
Hoed= 20cm P = 160 W d, o f S . = 1 6 1 p m X
-
= 3
u,,
0
I: f
- 1 - I0
L _ 3
J
A_--
_L_
25
2
A r Velocity
?5
73
60
50
70
Air Velocity, u. (cm/s)
uo icm/s)
Figure 4. Variation of heat transfer coefficient with air velocity for
Figure 7. Variation of heat transfer coefficient with air velocity for
different settled bed heights.
different particles and various mean sizes. C?.
i
+-
l
300r
:
0
I
: ;
=
0
0
s
i i
U
I
H o e d= 20 cm u. = 9 7 cm/s
d, o f S I = I61 pm .. . 1
21
3
50
40
1
60
I
70
1
80
I
90
I
0 2
03
1
05
/
,
I
08
*
Bed Temperature, Tbed ( O C )
Figure 5. Heat transfer coefficient is independent of bed temper-
" c
-
+ u
I00
I
I
I
I
Sand size, mesh X 40-50 V 50-70 0 70-80 A 80-100
3
I 5
, , , , I 8
IO
?0
30
40
uo
Figure 8. Display of the data and the dimensionless correlating line of eq 6.
a further increase in the flow rate of air. Because the particle size is not uniform, the h will not increase rapidly until the largest particles fluidize. As a result it was found that h jumps when the air velocity is larger than umfof the largest particles. For polystyrene particles the slope of h vs. the uocurve was shallow. This may be because its density is the smallest and its diameter is the largest of the particles tested, and because there seemed to be no clear point for onset of fluidization. Empirical Correlation. All the experimental data on five different kinds of solids were empirically correlated in terms of three dimensionless groups. The results are represented by eq 6 and are shown in Figure 8.
for 0.3
changed. These curves are similar to those of Figure 6. The heat transfer coefficient, h, increases quite drastically at onset fluidized bed; the value of h seems to reach a maximum at velocities greater than umfand then drop slightly with
2
dP pq J !
100
ature.
/
I0