HYDRODYNAMICS OF SLAT T R A Y S B. D . G I L E S ' A N D C. E. D R Y D E N
Department of Chemical Engineering, The Ohio State Uniuersity, Columbus, Ohio
Turbogrids are used extensively for gas-liquid mass transfer, but where low flow rates and/or high holdup are required for process design, sieve plate or bubble cap designs are usually specified. By using the cheap construction features of Turbogrids with narrower slot openings and conventional downcomer-weir-baffle arrangement for liquid flow control, a slat tray was designed with possible cost advantage over sieve trays and less tendency toward fouling and corrosion, particularly where wood can be used. The hydrodynamic characteristics of the slat tray, as given by pressure drop, entrainment, and leak-through, were measured experimentally. The latter two were determined with the aid of a phosphorus-32 radioactive tracer technique. The slat tray i s similar to sieve trays in hydrodynamic behavior but different from Turbogrids. Mass transfer comparisons and more hydrodynamic data on Turbogrid designs are needed.
COWENTIONAL equipment for gas-liquid
mass transfer operations consists of a vertical tower containing some type of packing or bubble cap, sieve, or Turbogrid trays. Because of the scale of many gas absorption operations, a considerable capital investment is required for the internal components of absorption towers. An example is the purification of large quantities of waste gas from atomic energy fuel reprocessing plants where a minimum scrubbing water flow is necessary. Turbogrid trays have proved advantageous for a large percentage of the gas-liquid mass transfer requirements of industry (9, 77, 78). but few characteristic operating data have been published ( 3 ) . Such a design cannot be used where low liquid rates and/or high holdup is a requirement. Thus, a combination of cheaply constructed grids or slats with closer clearances and the conventional downcomer-weir-baffle design for better liquid control would be a satisfactory compromise with possible cost advantage over sieve trays. Such tray designs have been used for replacement in a few columns where the removal of existing downcomers was troublesome (9. 77), but no design data for this type of tray have been published. Such a tray is referred to here as a slat tray, since the contacting area consists of a slat layout with small slotted clearances. These can be constructed from parallel strips or from one or more wider sections and slotted by stamping or programmed salving. Metal or Ivood is applicahle; if wood is used, due allowance must be made for swelling in selecting the final operating clearances for the long rectangular slot openings. This paper presents some characteristic performance parameters for slat trays, particularly those related to hydrodynamics, with experimental data on pressure drop, entrainment, and leak-through of a slat tray scrubber as a function of design and operating variables. Design Evaluation Program
Pressure Drop. The pressure drop across a tray is a function of a number of independent variables expressed in a general condensed form as: 0
fl(gc,
.I),Gs,
L, H , P,
U,
PL,
PL, ~ u t ,,
L ~ O ,fvs,6 s )
(1)
As summarized by a recent Russian communication (75), it is convenient to separate the total pressure drop, Ap, further into that of the tray alone, A P T , and that from the gas-liquid Present address, Hooker Chemical Corp., Niagara Falls, S . Y . 188
I&EC PROCESS DESIGN A N D DEVELOPMENT
interaction above the tray, A ~ L with , the resulting additive flow resistance equation:
lp
= APT
ApL
(2)
Dimensional analysis of Equation 1, together with a consolidation of terms held constant throughout the experiments, was used to develop a correlation equation of the form shown by Equation 2. Entrainment and Leak-Through. Entrainment is defined as the removal of drops from liquid flowing across a slat tray by gas bubbles moving through the liquid and the subsequent carry-over of this liquid to the tray above. The mechanisms in the entire sequence from bubble generation to final deposition of drops in the tray above are very complex. Sewitt. Dombrowski, and Krelman (76) described and photographed the phenomenon for sieve trays. Entrainment can be determined by injecting a tracer such as a dye ( Z ) , salt (7. 70. 76). or radioactive isotope (4, 73) in the entraining liquid and measuring the amount deposited in the liquid on the collector tray above. Leak-through or "weeping" can be measured in the same manner by placing the collector tray underneath. For these experiments. a three-tray model was built to measure both entrainment, E, and the leak-through, J , using radioactive P3?in phosphoric acid as the tracer. This method was chosen because it gave continuous measurements with recording instrumentation. Phosphorus-32 has a reasonably short half-life of 14.3 days and is a moderately strong p-emitter, thus allotving for easy internal detection without any external shielding of the equipment. Entrainment and leak-through are determined via P3* material balance equations based on the flowsheet (Figure 1). Using assumptions of equal E and J for all trays and complete liquid mixing on each tray, the resulting summary equations are : (3) (4)
W'hen E and J are small compared to L , these equations reduce to :
AIR OUT TO ATMOSPHERE
I I
I; L
l
Y
FILTER RECORDING COUNT RATE POTENTIOMETER METER
L-J b
J.
-
u
COUmNG CHAMBER
ENCLO ED PUMP
SEWER
Figure 2. Schematic diagram of slat tray column and Pa2 tracer system
Figure 1. Material balance flowsheet
Experimental Program 0AFFLE
Equipment. -4 three-tray column for contacting air and water was constructed as shown schematically in Figure 2. The sides were made of 3/8-inch Plexiglas, glued together on three sides and positioned within an angle iron framework. The front panel was gasketed and bolted for sealing, thus providing access to the three trays. Details of the tray design are shown in Figure 3. T h e slats were constructed of cypress wood measuring 4.0 X 0.37 X 0.19 inches when wet and having a clearance of 0.036 inch between slats and a free area of 5.87,. Turbogrids have three to four times this clearance, \+ith about 207, free area. T h e tray was hinged a t the \veir end to alter the tray pitch. The plastic weir Mas movable to adjust the liquid depth on the tray. A horizontal distance of l, 2 inch was always maintained between the bottom of the splash baffle and the top of the weir. The downcomer and well were used as a liquid distributor. The middle tray \\as fastened permanently in the middle of the column. T h e other two trays were movable, to provide a variable distance between trays of 0.5 to 2.0 feet. Each spacing position required I-inch NPS (nominal pipe size) water inlet and outlet holes to serve the trays. These holes were plugged by rubber stoppers when not in use. T h e trays were attached to the column walls by a cypress ledge and gasketed. -Air and water flows u w e metered by calibrated orifice plates and manometers. A calming section and 40-mesh screen were used to straighten the air stream flowing to the bottom tray. Pressure taps located 1 inch under each tray were connected to xylene fluid manometers to measure individual tray pressure drops and the total pressure drop for all three trays as a check. The P3? activity was measured \vith a Raytheon CK-1020 Geiger tube having glass walls coated with black enamel paint to reduce light sensitivity and with silicone grease to avoid adsorption of Pa?. Ordinary phosphoric acid as a 10% solution was slo\\ly added to each counting chamber to increase acidity and further reduce adsorption of phosphorus. The counting chamber in which the tube was immersed was designed for a time constant of 20 seconds or less based on inlet concentration and liquid flow rate range. The counting ratemeter and recording instrumentation had a time constant of 56 seconds. The response rate of this system was adequate for the steadystate experimental program set up for this study. Range of Variables Studied
G = column air mass velocity, 340 to 1600 lb./hr.-sq. ft. column area
G, = slot mass velocity, 5850 to 27,600 lb./hr.-sq. ft. L = liquid rate, 2500 to 5100 lb./hr.-sq. ft. column area h, = weir height, 0 to 2 inches
U
, 2
1
-
b
IN.
Figure 3.
P
0 to 2O7,, computed as cosine of angle of tray bvith horizontal interfacial tension, 29 to 72 dynes per cm. liquid viscosity, 1.00 to 3.7 cp. liquid density, 57.9 to 68.5 lb./cu. ft. actual tray spacing, 0.5 to 2.0 feet per cent open or slot area, 5.8%
= tray pitch,
= pL = pL = S = A, = u
Slat tray assembly
The physical property range of liquids was achieved by using aqueous ethanol (up to 457,) or aqueous glycerol (up to 45%). Procedure. The radiation detection equipment was turned on for a 30-minute warm-up period. In the meantime, city water was metered to all three trays and the air flow was set a t a predetermined level. After 30 minutes, a background count on all three trays was recorded and then liquid to the middle tray was changed to the P32-tagged recirculating solution. The middle tray count rate was followed until a leveling off a t a steady rate in the range of 5000 to 15,000 counts per minute was achieved. The top and bottom count rates were determined in succession with 8to 10-minute reading times often necessary to get reasonable precision (=k2OyO maximum error) when the count rate was 10W (50 to 100 c.p.m.) and only slightly above background. All pressure tap manometers were read during these counting periods. The air and liquid flow rates were then changed so that a complete set of data for a given column geometry was obtained for a series of three to six air rates for each of three to four liquid rates. -4 change in geometry, such as tray spacing, was then made and the series of determinations was repeated. VOL. 2
NO. 3
JULY 1963
189
t
i
EQUATION 11
10
G, , SLOT VELOCITY, L B / H R - S Q F T
Figure 4. trays
Pressure drop data for slat and sieve
Figure 6.
Results
The group labeled I contained variables which were not changed during the study. Group I1 contained ratios secondary in importance. The remaining dimensionless ratios were correlated for exponents by multiple regression using a digital computer. After separating terms, the powers were collected and average numerical values assigned for p@ 01* and G,-O 06. These values, along with constants p,., g,, and gL, make up the constant of the resulting equation: A f L = 0,0076j ,,0.20
,
3.0
/
$ 1
I
//
/
c
/
L0.23 p L 1 . 6 1
/
/
//
/
/
/
/
/
,2
/
,’
/’
/
/O
A ’ /
/’
.o
/’
’*
- -- BUBBLE CAP TRAY 2 0 % f r e area.
~
(
6)
:
SIEVE TRAY, t 5 ) : 8 % f r e c area, 3 / 1 6 - i n h o l e s , O - i n weir.
x
L ,
0
S L A T TRAY, EQUATION 11, 0-in weir,
axc
4000 6000 8000 l0000 LIQUID FLOW RATE , L B . / H R . - S Q F T .
Figure 5. Comparison of pressure drop characteristics for gas-liquid contactors 190
I & E C PROCESS D E S I G N A N D D E V E L O P M E N T
I
I1
Vc ’. 2.5
3.01
Entrainment data for slat trays
The pressure drop due to liquid on the tray was more complex to analyze. Based on the functional variables given in Equation 1, a dimensional analysis resulted in the following dimensionless groups :
Pressure Drop. The data on pressure drop through the slat trays without liquid cross flow were correlated by the equation APT = 8.84 X 10-8 Gs1,88 (7) which compares favorably with numerous sieve tray data recently reported by Mukhlenov and Tarat (75) and correlated by the equation APT = 8.1 X 10-8 G,’.8’ (8) The range of free area for their correlation was 7.5 to 31.5% with a constant plate thickness of 0.197 inch and t / D . values ranging from 0.625 to 2.5, compared with the slat tray free area of 5.87,: a slat thickness of 0.19 inch, and a t / D e of 2.66. T h e exponent of 1.88 also compares favorably with the 1.8 value reported by Arnold ( 7 ) for thin sieve plates with a t / D , range of 0.077 to 0.74. If the permanent head loss for the slot was simply a n abrupt contraction and expansion, the exponent on G, should have the value of 2. Some investigators have used this exponent in correlating dry sieve tray data ( 7 7 , 74).
b
SPACING, Serf - In.
EFFECTIVE TRAY
H0.48 p2l.i
(10)
This pressure drop is approximately equivalent to the total head of liquid, H , over the range of variables studied. T h e total pressure drop across a single slat tray comes from the additive resistances of the two effects as given in Equation 2. 8.84 X 10-8 G,'.** 4- 0.00765 (TO."
A$
Lo.23 ~ ~ 1 HO . .4 ~9 PZl.5 1
(11)
.411 of the experimental pressure drop data were correlated by Equation 11 with a precision of 1 4 . 6 7 , for Ap from 0 to a maximum measured value of 39.2 lb.fi'sq. ft. The similarity of total pressure drop for slat and sieve trays is shown in Figure 4 for comparable geometry and flow conditions. .4 more general comparison is made in Figure 5 for four types of contacting trays a t typical operating conditions. The slat tray was computed a t the same slot velocity as the sieve trays. T h e data for these two trays compare well up to a n L of 5000: which is close to the maximum used in this work. Equation 11 could thus be used for calculation of sieve tray pressure drop. T h e correlating equation appears to predict too low a total pressure drop for slat trays a t higher L values and should not be used for a n L above 5000. The slat tray definitely does not behave as a Turbogrid, which has three distinct regions, as pointed out by Garner (6). ,411 three types of open trays have loiver pressure drops than a bubble cap design. Entrainment. The carry-over of liquid to the top tray was observed under varying conditions of tray spacing, weir height, tray pitch, and flow rates for gases and liquids of varying physical properties. The best correlation of all the entrainment data was obtained as a function of column gas velocity, G? and effective tray spacing, Sefi,defined as the distance between the top of the aerated liquid and the tray above. The aerated liquid height was difficult to measure directly because of the unstable and turbulent motion of the fluid. A more reliable L due principally to the average height was taken from A ~ data, head of liquid on the tray and a n estimated foam density of 0.35. This value is an average of the 0.3 to 0.35 values of Gerster (8) and the 0.4 gram per cc. density used by Hunt
lo-'
I/ / %SIEVE / I
. BUBBE CAP TRAY ( 4 )
\
4 :
6 . a
--n
/
1
/
//
/
(JSLAT
( 7 7 ) . Typical data are plotted in Figure 6 and can be represented by : GZ.7
E - _- ( 2 f 1)(10-5) -
G
This correlation does not include the 24-inch tray spacing of pitch greater than zero. These latter data showed higher entrainment than expected, probably caused by the increased rolling and splashing observed as the liquid moved down the sloping tray toward the weir. Hunt ( 7 7 ) used a similar procedure for sieve tray entrainment correlation, with the resulting equation for air a t 1 atm. and a number of fluids given as:
E G
=
A
ul
9
TRAY,EQ.
(&)3.z
\+
/
LIQUID -WATER L,1485- 5170
D
WEIR, 1 in.
, L :5000
T
O e / o PITCI-1
t
1
t.
12- I N . T R A Y SPACING
L
.
~
):(
I2
.
1,
10-9
1
[L
-
x
T R A Y , E Q . 13
Q
5 U e
3.69
Liquid heights varying from 1.8 to 3.9 inches had no noticeable effect on entrainment. This is consistent with present findings. Within the precision of the data, no effect of physical properties was observed, whereas Hunt ( 7 7 ) found a significant effect of surface tension for sieve trays, as shown in Equation 13. One explanation is that surface tension plays a more important role in creating an effective perforation velocitv for gas escaping from a small circular hole than in the case of a long narrow slot. Entrainment characteristics of various trays are compared in Figure 7. Bubble cap data were derived from the recent work of Dorweiler and Burnet ( 4 ) and show consistently higher entrainment than the other trays. Sieve and slat tray correlations are nearly equivalent. Entrainment data for Turbogrids are not readily available. Those used here are approximations from the work of Simkin (79) on the relative values of Turbogrids cs. bubble caps. The odd shape of the entrainment curve shows the rather drastic function of liquid rates through the trays as the gas velocity varies. This is not characteristic of slat or sieve trays. The liquid loading point for Turbogrid operation should be above 1500 Ib. hr.-sq. ft., so the curve for L = 500 shown here is for poor gas-liquid contact with a low entrainment for all gas velocities. More
_I
w z -
(SerfP . 3
5000
2-lN.WEIR O N
TRAYS
0
i
A
TRAY SPACING 6 in. 12 in 24 in
TURBOGRID, L.500
0 w .,
i'
,I 10
G , COLUMN
,
,
,
,
,
,
, 10
lo4
G A S VELOCITY, LB/HR.-SQ.FT.
Figure 7. Comparison of entrainment characteristics for gas-liquid contactors
1003
10,m
SLOT GAS VELCCITY, G, -LB/HR.Figure 8.
S0.FT.
Leak-through data for slat trays
VOL. 2
NO. 3
JULY
1963
191
detailed entrainment data for Turbogrids are needed to make more valid comparisons. Leak-Through. Measurements of plate stability in terms of leakage or weeping made by the P32tracer technique were less precise than those for entrainment because of the lower order of magnitude of such leakage. Figure 8 shows characteristic data for the water-air system. KO quantitative leakage criteria could be found in the literature to make comparisons. With an arbitrarily chosen stability criterion of 0.1% leakage based on liquid flows, a gas slot mass velocity of 15.000 lb./ hr.-sq. ft. or a linear slot velocity of 56 feet per second is required a t a n L of 1500; this decreases to 35 feet per second for a n L of 5000. Such values are to be compared with the 25- to 40foot-per-second slot velocities reported in the literature ( 7 , 5, 72) to avoid noticeable leakage of fluid through sieve trays. From an economic optimization standpoint, high leakages may often be tolerated. However, slat tray action was inefficient below G , of 5000, since gas was not bubbling from many of the slots which were still liquid-sealed. KO attempt was made to obtain a seal point correlation.
hC
= height of crest over weir computed from Frances
h,
= weir height, feet = maximum clear liquid height on tray =
weir formula, feet
H
J = L = (MT,,) = Ab APT APL
P
= = = =
sei,
=
(TT,,,) VC
= = =
It’,
=
FVS
=
t
GREEKSYMBOLS fig, fir,
Conclusions
$JS
Slat trays behave like sieve trays in terms of pressure drop, entrainment, and leak-through. The mass-transfer and operational characteristics of both designs should also be similar. There is no hydrodynamic resemblance between slat trays and Turbogrids, even though the basic contacting surfaces are geometrically similar except for spacing between slats. The slat tray is a n ideal design for low liquid flow-high holdup requirements, combining the stable operating performance of a sieve tray with the relatively low-cost construction of the contacting area, particularly where wood can be used. Acknowledgment
The helpful suggestions of C. E. Lapple, Stanford Research Institute, in formulating this research problem are deeply appreciated. Financial assistance of the Shell Oil Co. and The Ohio State University in the form of scholarships and research funds is gratefully acknowledged. Computer time was kindly donated by the Electrochemicals Department, E. I. d u Pont de Nemours st Co., Inc., Niagara Falls, N. Y . Nomenclature
As
(BT,,,,)
D, E
fl,f2
gc
92 G
G,
192
per cent open area of slat grid bottom tray counting rate? counts per minute hydraulic diameter, feet entrainment, lb./hr.-sq. ft. column area mathematical functions gravitational constant, 32.2 lb.,-ft./lb.f-sec.2 = local acceleration of gravity, ft./sec2 = mass velocity of gas in column, lb.,/hr.-sq. column area = slot mass velocity, lb.,/hr.-sq. ft. of slot area = = = = = =
l & E C PROCESS D E S I G N A N D D E V E L O P M E N T
pQ, p L U
= = = =
viscosity of gas and liquid, respectively, lb.,/ft. sec. slat shape factor, dimensionless density of gas and liquid, respectively, lb./cu. ft. interfacial tension; dynes per cm. in Equation 13; lb., 'set.* in other equations
literature Cited
(1) Arnold, D. S., Plank, C. A., Schoenborn, E. M., Chem. Eng. Progr. 48 (12), 633 (1952). (2) Ashraf, F. A., Cubbage, T. L., Huntington, R. L., Znd. Eng. Chem. 26, 1068 (1934). (31 Brit. Chem. Eng. 2 (lo), 544 (1957). (4) Dorweiler. V.. Burnet. G.. 53rd Annual AIChE Meeting., Dee. 4, 1960. (5) Ellis, S. R. M., Moyade, H. K., Brit. Chem. Eng. 4, 342 (1959). (6) Garner, F. H., Ellis, S. R. M., Freshwater, D. C., Trans. Inst. Chem. Eng. 35 ( l ) , 61 (1957). (7) Garner, F. H., Ellis, S. R. M., Lacey, J. A.,Zbid., 32, 222 (1954). (8) Gerster, J. A., Bonnet, LV. E., Hess, I., Chem. Eng. Progr. 47, 523 (1951). ( 9 ) Giles. B. D.. Ph.D. dissertation. Ohio State Lniversitv. March’1962. ’ (10) Holbrook, G. E., Baker, E. M., Ind. Eng. Chem. 26, 1063 ’ (1934). (11) Hunt, C. D., Hanson, D. N., LVilkie, C. R., A.I.Ch.E. J . 1 (4), 441 (1955). (12) Kug’minykh, I. N., Rodionov, A. I., J . Appl. Chem. (USSR) 32, 1311 (1959). (13) Manowitz, B., Bretton, R. H., Horrigan, R. V., Chem. Eng. Progr. 51 (7), 313 (1955). (14) Mayfield, F. D., Church, W.L., Green, A. C., Lee, D. C., Rasmussen, R. W., Znd. Eng. Chem. 44, 2238 (1952). (15) Mukhlenov, I. P., Tarat, E. I., Zh. Prikl. Khim. 31, 542 (1958). (16) Newitt, D. M., Dombrowski, H., Krelman, F. H., Trans. Inst. Chem. Eng. 32, 244 (1954). (17) Petrol. Refiner 31 ( l l ) , 105 (1952). (18) Shell Development, Engineering Staff, Chem. Eng. Progr. 50 (2), 57 (1954). (19) Simkin, D. J., Strand, C. P., Olney, R. B., Zbtd., 50 ( l l ) , 565 (1954). RECEIVED for review September 19, 1962 ACCEPTED January 21, 1963 %
\
’
ft.
+
h, h,, feet leak-through, lb./hr.-sq. ft. column area liquid flow rate, lb.,/hr.-sq. ft. column area middle tray counting rate, counts per minute pressure drop across tray and liquid, lb.f/sq. ft. pressure drop across tray only, lb.f/sq. ft. pressure drop across liquid on tray, lb.f/sq. ft. tray pitch, 70, computed as cosine of angle measured upward from horizontal vertical distance between top of aerated foam and bottom of tray above, inches plate or slat thickness, feet top tray counting rate, counts per minute superficial linear velocity of gas in column, feet per second slot clearance or width of opening between slats, feet width of slat tray. feet
-
,
I
~
,
I