630
Ind. Eng. Chem. Process Des. Dev.
1980, 79, 630-635
Hydrodemetallization of Heavy Residual Oils in Laboratory Trickle-Flow Liquid Recycle Reactors R. H. van Dongen," D. Bode, H. van der Eljk, and J. van Klinken Koninklijke/Shell-Laboratorium, Amsterdam (Shell Research B. V.), 1003 AA Amsterdam, The Netherlands
Kinetics of hydrodemetallization have been studied in bench-scale fixed-bed reactors with product recycle facilities using a high metal content residual oil as the feedstock. In the liquid recycle mode of operation a continuous stirred tank reactor (CSTR) is simulated. It is shown that if the total product is recycled without adding fresh feed (batch stirred tank reactor, BSTR), the reaction kinetics of a cocurrent moving-bed reactor can be simulated. This avoids the difficulties involved in operating actual moving bed systems on a laboratory scale. The apparent reaction order for vanadium removal appears to be different in both modes of operation in the conversion regions examined. This phenomenon, which also occurs in residue desulfurization, is rationalized as being a consequence of the wide spread in reactivity of the individual metal or sulfur-bearing species.
Introduction
Nearly all crude oils contain appreciable quantities of metals-mostly vanadium and nickel-in organic structures of high molecular weight. Upon refining the petroleum these organometallic compounds end up in the distillation residues. It may be necessary to remove at least part of the metals for various reasons. For example, hydrodemetallization may serve as pretreatment for desulfurization of residues containing more than, say, 100 ppmw of metals. In this way HDS catalyst deactivation due to metals deposition is slowed down and catalyst life is prolonged. Furthermore, hydrometallization itself is a means of upgrading residual oils-the removal of metals is coupled with a marked conversion of asphaltenic structures and a substantial reduction of the viscosity of the oil. In order to be able to cope with residual oils containing many hundreds of ppmw of metals, while preserving the advantages of the trickle-flow reactor technology, a moving bed reactor would seem very useful. In such a reactor system the packed catalyst bed moves slowly down through the reactor as spent catalyst is withdrawn from the bottom and fresh catalyst is regularly sluiced in at the top (van Ginneken et al., 1975). In kinetic studies and studies on the catalyst performance proper, a moving bed may not be essential. In addition, as will be argued below, the effect of the basic process variables can adequately be studied in a fixed-bed reactor. For engineering studies a truly moving bed reactor is required. It will be obvious that a moving bed reactor places high demands on the properties of the demetallization catalyst. In addition to superior mechanical properties, a high storage capacity for metals is desirable. In process studies related to catalyst performance (activity, deactivation, useful life), and in catalyst development, extensive use is made of fixed bed trickle-flow recycle reactors to simulate the kinetics corresponding to a continuous stirred tank reactor (CSTR). See, for instance, Dautzenberg et al. (1978). In the present work the kinetics of vanadium removal from the higher metal content residual oil have been examined in both simulated CSTR and simulated movingbed reactors. The results are rationalized on the basis of differing reactivities of the individual components. E x p e r i m e n t a l Section
The present experiments were carried out in bench-scale equipment featuring separate oil and hydrogen preheating,
a 1-m 250-mL reactor fitted with a central thermowell, a high pressure gas/liquid separator, and facilities for recycling liquid product. In all cases gas once-through operation was applied. Simplified flow schemes are shown in Figures 1 and 2. In the moving-bed simulations the catalyst bed was diluted with an equal volume of 0.5-mm silicon carbide to improve catalyst/oil contact (van Klinken and van Dongen, 1980). It will be pointed out below that catalyst bed dilution is not necessary in the CSTR experiments as liquid velocities are high enough to ensure good oil/catalyst contact. The catalysts used are experimental versions of proprietry residue demetallization catalysts. In the present work catalyst beads with an average diameter of 2.5 mm have been used. Before testing, the catalysts were presulfided in situ in a stream of H2/H2S. All experiments were carried out at a standard pressure and temperature. Liquid space velocities, linear velocities, and gas flow rates differ from one case to the other as indicated. S i m u l a t e d Continuous S t i r r e d Tank Reactor In studies on reaction kinetics the recycle reactor is an established concept (Levenspiel, 1972). Basically it consists of a fixed-bed plug-flow reactor (catalyst volume being Vd with a low conversion per pass as a result of the large recycle of product (recycleratio, r, is the ratio of the recycle flow rate to that of the fresh feed). Let Cf be the concentration of the reactant, say sulfur or vanadium, in the feed which enters the system at a rate of $L kg/h. If the rate of conversion of the reactant can be described by simple power law kinetics of order n, the rate constant k for this conversion is given by the following expression for a plug-flow reactor
where the space velocity F = I $ ~ / Vand ~ , C is the concentration of the reactant in the effluent. Tkis equation reduces to the corresponding logarithmic form by taking the limit for n to 1. At infinite recycle ratio r, eq 1reduces to the kinetic expression for an ideal stirred-tank reactor
k, = F ( C f - Cp)/Cp"
(2)
For realistic recycle ratios the exponential term in the
0196-4305/80/1119-0630$01.00/00 1980 American Chemical Society
Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980
-
631
Table I. Space Velocity Excursions in One CSTR Experimenta run time, h 80
space velocity, kg/L h
330
3.5 0.25 3.5
410
1.1 1.1
fractional Va removal
apparent reaction order
0.28 0.85 0.22 0.47 0.47 0.22
3.5
1.0 1.1
1.0
a Combined liquid feed rate: 2 5 kg/L h ; standard iemperature and pressure; hydrogen feed rate: 4 0 standard L/kg combined feed; reactor loaded with 100 m L of catalyst A.
SSTMPPER
LIQUID
HIGH-PRESSURE
PRODUCT
SEPARATOR
Figure 1. Simulation of a continuous stirred tank reactor using a fixed-bed reactor with large external liquid product recycle.
Table 11. Three CSTR Experiments at Different Space Velocitya space velocity, 1.0-order rate constant, kg/L h kg/L h 1.7 0.25 1.9 1.0 1.6 3.5 ~
a
U
tI
BATCH W I S E CIRCULATION
I
Ix1
RE A CTOR
OFF-GAS
STRIPPER/ PRODUCT VESSEL
HIGH -PRESSURE
SEWATOR
Figure 2. Simulation of a moving-bed reactor using a fixed-bed reactor with batchwise circulation of the total liquid product.
brackets can be expandied into a rapidly converging Taylor series. Omitting all terms of the third degree and higher, we can give the followinlg approximation for the ratio k/k,
k *Ik, 2C,(r
+ 1)
If we do not want K and k, to differ more than 5%, so that we can safely use the simple CSTR kinetics of eq 2 for all practical purposes, the recycle ratio should be high enough to meet the following condition r + 1 >* 10n(Cf - C,)/C, (3) The experiments in the recycle reactor were carried out in such a way that the superficial liquid flow rate was constant. This was achieved by adjusting the recycle ratio r when a different space velocity was applied. With the data shown in Table I it can be verified that under these conditions, for a firsborder reaction, the above criterion is met in all cases. In is interesting to note that a deviation from ideal plug flow, as a result of axial dispersion in the catalyst bed, does not affect the above consideration. In fact, axial dispersion provides an internal contribution to the total mixing in the system and hence actuidly relaxes the criterion for sufficient recycling. The superficial liquid flow rate amounts to some 3 kg/m2 s in all of our recycle reactor experiments. This high value
~~
Conditions as in Table I, except catalyst B.
is of the order of magnitude of the liquid flow rates in commercial trickle-flow reactors so that good oil catalyst contact may be assumed to occur. Indeed it has been found that neither doubling nor halving the recycle ratio affects the degree of vanadium and sulfur removal. For the determination of the reaction order n for the removal of vanadium, the space velocity in a simulated CSTR experiment was changed a number of times in the course of the run. Table I shows the change in fractional vanadium removal (from cy1 to a,) when the space velocity is altered (from F1 to F2). At each transition the reaction order can be calculated since, rewriting eq 2 F,CYl(l - C
Y p
= F,CY,(l
- C
Y p
which is based on the fair assumption that the activity of the catalyst remains constant during the space velocity transition. It appears that first-order kinetics describe the hydrodevanadization well up to 85 wt % vanadium removal, if not higher. Further confirmation of first-order kinetic behavior was obtained in three separate experiments carried out at widely different space velocities with the same feedstock but with a slightly different catalyst. Table I1 shows that the first-order rate constants at zero catalyst age differ only slightly. First-order kinetics for vanadium removal in CSTR do not apply only to the present catalyst-oil combination; they provide an adequate description of hydrodemetallization in CSTR systems in general, including the vanadium removal inevitably accompanying residue desulfurization (Dautzenberg et al., 1978). Simulated Moving-Bed Reactor The Ideal Simulation. In the ideal cocurrent moving-bed reactor the liquid, gas, and catalyst all move in plug flow downward through the reactor, Flow rates of liquid and catalyst are denoted by 4L (kg/h) and C # J ~ (L/h), respectively. For the steady-state situation the mass balance for the reactant (vanadium) in the liquid phase over a slice of reactor (volume dV) can be written as 4 L dCL = -R dV (4) The reaction rate R may be any function of CLand Cc, the concentration of the metal reactant in the liquid and on
632
Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980
the catalyst, respectively. As all of the metals removed from the oil are deposited on the catalyst the total mass balance over the volume element dV reads 4L
dCL = -4cpc dCc
(5)
where pc is the bulk density of the fresh catalyst. If the function R(CL, Cc) were known, the performance of the reactor could be calculated using these two equations. Let us compare the above expressions with the basic relationships for a batch recycle reactor. Such a system consists of a closed loop of a fixed-bed reactor (contents VR) and a buffer vessel containing WL kg of liquid. The circulation rate is so high that the plug-flow reactor acts as a differential reactor. Then the conversion of metalin-oil to metal-on-catalyst is given by the rate expression WL dCL = -RVR dt
(6)
where t is the run time. The appropriate mass balance reads WL dCL = -V@c dCc
(7)
Comparing the mass balance over a volume element dV of the moving-bed reactor with that of a time element dt in the batch recycle reactor (BSTR), it appears that eq 5 and 7 are identical if the catalyst to oil intake ratio in the BSTR is equal to the catalyst to oil feed rate ratio in the moving-bed reactor, or @C
-
4L
VR
WL
This ratio represents the catalyst use rate U which is usually expressed in terms of kg of catalyst per kg of oil. Hence
for moving-bed and BSTR, respectively. Substitution of eq 9 in the rate expressions (4)and (6) yields Pc
dV
U
4c
-dCL = -R-
(moving-bed)
PCdCL= -R d t (BSTR)
U
(10)
Boundary conditions at VI@,= 0 for the moving bed and at t = 0 in BSTR are C, = 0 and CL = Cfeed. The boundaries being identical, the similar differential equations give similar solutions. Hence, reaction rates R and concentrations C, and CL in the moving bed a t a given value of the catalyst space time VI@, are identical with those in the corresponding BSTR at run time t, or, in terms of catalyst space velocity, l/Fc = t. It should be emphasized that catalyst space time in the moving bed is equivalent to run time in BSTR only if the same catalyst use rate is applied. From eq 9 and 10 it follows that run time t in BSTR can be translated into oil space velocity FLin the moving-bed reactor according to FL
=
PC
6
In practice the catalyst use rate U is dictated by the maximum metal load allowable on the spent catalyst, the metal content of feed, and that of the product aimed at. Once U is chosen and an oil space velocity F L is selected, eq 11 fixes the duration of the moving-bed reactor simulation run.
Table 111. Simulation of a Moving-Bed Reactor at Various Catalyst Use Ratesa experiment catalyst use rate (relative to standard) oil space velocity per pass, kg/L h number of processing steps
1 3.5
2
1.8
3 1
7
12
23
14
10
9
Standard temperature and pressure; catalyst B.
Practical Implementation. Unlike the situation in a continuous stirred tank reactor, the effect of the liquid recycle rate on the closeness of approximation of the ideal batch stirred tank reactor cannot be expressed for a general case. It should be realized, however, that if the catalyst did not deactivate as it collected more metals, the moving-bed reactor would be kinetically identical with a fixed-bed reactor. In that situation the actual rate of circulation in the BSTR is completely immaterial, provided liquid plug flow prevails in the trickle-flow reactor section. Hence, the choice or the circulation rate should be based on what is deemed acceptable in view of (anticipated) catalyst deactivation on the one hand and on practical considerations as to equipment limitations and operability on the other. For the reaction system under investigation an acceptable compromise seems to be the application of a relatively low circulation rate by batchwise reprocessing instead of continuous and rapid recycling of the oil. In the batchwise mode of operation the batch of oil is processed once through the fixed bed and the liquid product is collected. After completion of the pass the entire product is reprocessed, and so on. The experiments were set up in such a way that some 10-15 passes were sufficient to reach the target product vanadium content. Note that only a t the completion of each pass, and not at any other moment during that pass, has the correct catalyst use rate been applied so that only the run time marking the completion of a reprocessing step can be used in eq 10 and 11. A condition for reliability of results in the batchwise reprocessing mode of operation is that plug flow of the oil can be assumed to occur in the trickle-flow reactor. Mears (1971) has formulated a criterion for near plug flow in trickle-flow reactors based on the axial dispersion model. If the volume of the actual reactor is not allowed to differ more than 5% from that of the corresponding ideal plug-flow reactor with the same conversion a, the ratio of the bed height L and particle size d, should meet the following requirement L 20n 1 ->-lnd, BO l - C Y where n is the order of the reaction and Bo is the Bodenstein number expressing the axial dispersion in the bed. Taking Bo = 0.02 (Van Klinken and Van Dongen, 1980) the condition of eq 12 is only satisfied for d, < 1 mm in the present experiments. As the catalyst particle diameter is 2.5 mm, it was deemed necessary to dilute the catalyst bed with an equal volume of 0.5-mm silicon carbide particles. Good oil/catalyst contacting may be relied upon thanks to this bed dilution and the relatively high superficial liquid flow rate applied. A survey of the conditions is given in Table 111. Moving-Bed Simulation Results. Three BSTR experiments simulating performance of the moving-bed reactor were carried out, as indicated, under identical conditions except for the amount of catalyst in the reactor. Figure 3 shows that the product vanadium content falls
Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980 633 Cp/Cf
,FRACTION
'VANADIUM RETAINED IN PRODUCT
6 t
10
0.5
I
0 0
l/F,
10 20 30 CATALYST SPACE TIME (ARBITRARY UNITS)
Figure 3. Residual oil denietallization in a simulated moving-bed reactor at various catalyst use rates. C p / C f , FRACTION VANADIUM RETAINED IN PRODUCT
CATALYST USE RATE (RELATIVE TO STANDARD)
35 1.8 1
"
0.5
10 l/'FL
15
, LIQUID
2.0 SPACE TIME, hl/k9
Figure 4. Rate of vanadium removal vs. liquid space velocity a t
off more rapidly, the higher the amount of catalyst present. Run time in BSTR, or catalyst space time, 1/Fc, in the moving bed, are inconvenient time memures and therefore the data are reproduce'd in Figure 4 as a function of reciprocal oil space velocity 1 / F L or oil space time (eq 11). I t is then found that the three curves of Figure 3 merge, the differences between the three being insignificant. This implies that a t a given oil space velocity the degree of vanadium removal is the same for all three catalyst use rates. However, the actual amounts of vanadium deposited on the catalysts at a certain degree of vanadium removal differ by a factor of three-because the amounts of catalyst used differ by the same factor. Hence, the relationship between product vanadium content and oil space velocity is independent of the vanadium load on catalyst. Assuming simple power law kinetics we may write for this relationship
The apparent order n was determined using the method of least squares such that a plot of Cql-n/(n- 1)vs. l/FL fits a straight line as closely as possible (Kittrell et al.,
0
as
I
I
1.0 1.5 2.0 l/FL, LlOUlD SPACE TIME, hl/kg
Figure 5. Description of residual oil hydrodemetallization in a simulated moving-bed reactor with 1.5-order kinetics.
1966). The best fit was obtained for n = 1.5. Figure 5 shows the excellent fit obtained for 1.5-order kinetics with vanadium removal levels from 10 to 90%. These 1.5-order demetallization kinetics for plug-flow modes of operation are much more widely applicable than indicated by the specific case discussed here. Indeed, we have found that an n of about 1.5 always gives a best fit, even for cases with different residual feedstocks, other catalysts, other conditions, and even in fixed-bed liquid once-through reactors. Turning now to the results published by other workers in this field, we find that Chang and Silvestri (1974, 1976), for example, use first-order kinetics to describe hydrodemetallization over manganese modules. Oleck and Sherry (1977) report that a better description of this reaction system is obtained with second-order kinetics at moderate degrees of metal removal. In addition, the data published by Beuther and Schmid (1964) can be used to infer that second-order kinetics should be applied, and Inoguchi et al. (1971) find both first- and second-order descriptions equally food for their plug-flow reactors. As none of the reaction systems reported is completely comparable to the others, the differences in the results are hard to rationalize. In any event the present finding of 1.5-order kinetics for vanadium removal in a plug-flow reactor does not seem to conflict with any of the observations cited. Apparent Kinetics in Multicomponent Systems In the preceding sections it has been shown that vanadium removal can be described satisfactory by an apparent reaction order of 1.0 in a CSTR and 1.5 in a plug-flow reactor (PFR). In view of the rather simplistic approach to such a complex system this difference in apparent reaction order in itself is not very surprising. Indeed, a similar situation is found in residual oil desulfurization. In a CSTR a reaction order of 1.5 in unconverted sulfur describes the effect of space velocity on sulfur removal most adequately (Dautzenberg, 1978). At the same time there is plenty of evidence that residue desulfurization in a PFR obeys second-order kinetics (Beuther and Schmid; 1964; Massagutov et al., 1967). The latter relatively high reaction order has been made plausible by various investigators (van Deemter, 1964; De Bruijn, 1976). In numerical examples they illustrate how a number of parallel first-order reactions with mutually different rate constants may appear as one second-order reaction. This concept of parallel reactions is wholly compatible with the mul-
634
Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980
ticomponent nature of petroleum fractions. By reference to the case of residue hydrodesulfurization a more rigorous elaboration of this concept can be given which will explain the different reaction orders in CSTR and PFR systems. The catalytic hydrodesulfurization of individual sulfur compounds follows first-order kinetics (Frye and Mosby, 1976; Kolboe, 1969). In such a case the conversion of one component in a PFR at space time, 7, is given by 1 exp(-kr). Let f(k) dk represent the fraction of sulfurcontaining molecules which react with a rate constant between k and k dk; then the expression for the overall conversion a reads
+
a = 1-
e-kr f(k) dk
Jm
(14)
The integral represents the Laplace transform of the distribution function f(k); this function can, for that reason, be solved by inverse Laplace transformation of an experimentally observed function, 1 - a(7). An obvious form of the latter function is that for nth-order power law kinetics a = 1- [ l
+ ( n - l)knCfn-l~]l/(l-n)
(15)
0.7
0.6
0
0.5
08
I
I
a
I
I
I
1.5
1.5
- Ln (1 - a )
Figure 6. Calculated multicomponent parallel first-order reaction kinetics (circles) correlated with apparent 1.5-order kinetics for a single component (line).
It appears that an exact solution of f(k) from eq 14 and 15 is possible for apparent reaction orders n between 1and 2. The solution takes the form of a Poisson distribution
law kinetics it should be possible to describe the data with eq 2, which can be written in a linear form
where
Figure 6 shows how closely the synthesized data fit to a straight line. In the conversion range 0.2-0.8 a best fit was obtained for an apparent reaction order of 1.5. Note that this line does not pass through the origin which means that the apparent rate constant k,Cf"-l in a CSTR is a little greater than the true average rate constant kav. The above calculations illustrate that finding reaction orders of 2.0 in a PFR and 1.5 in a CSTR may be explained by a wide spread in the reactivities of basically 1.0-order hydrodesulfurization reactions. It would seem reasonable, therefore, to attribute the appearance of 1.5 order in PFR and 1.0 order in CSTR for vanadium removal to a similar spread in reactivities. Besides, the similarity of both cases would suggest that the reaction order of the individual vanadium-bearing species is in fact less than 1.0. Support for this latter suggestion has been obtained from calculations starting from a wide spread of parallel reactions obeying the Langmuir-Hinshelwood model for a first-order reaction on the catalyst surface. By allowing stronger adsorption of the reactant to the surface, the order of the reaction seems to become less than 1. The apparent reaction order in CSTR and PFR decreases accordingly, although the difference in the order remains 0.5. Only when the actual spread in the reactivities is made smaller does the difference between apparent kinetics in CSTR and PFR diminish, both approaching that of the individual components. Conclusions The liquid recycle reactor provides an adequate means of ranking residue hydroprocessing catalysts and assessing their ageing characteristics using simple continuous stirred tank reactor (CSTR) kinetics. The same liquid recycle reactor can, in principle, be used to simulate a moving trickle-bed reactor when it is operated as a batch stirred tank reactor (BSTR). In the latter mode of operation the recycle reactor can be applied most effectively in studying process variables. In a CSTR the hydrodevanadization of residual oils follows 1.0-order kinetics. In the simulated moving-bed
1 p = - n* - 1 '
The average first-order rate constant, k,, defined as k,, =
j- kf(k) dk 0
turns out to be the reduced apparent overall rate constant In the case of residue desulfurization an apparent second-order reaction has been found so that the distribution function reduces to a simple exponential form 1
Knowing the distribution function, the overall conversion in a CSTR can be found by integration over individual contributions. Similar to eq 14 we find a = l -
1
+ k7
Integration yields
where -E(-x) =
I-:
-e+ dt
Numerical values for this exponential integral can be found in the usual mathematical tables. With the help of eq 20 the conversion was calculated for a variety of values of l/k,7. If the thus synthesized data follow CSTR power
Ind. Eng. Chem. Process Des. Dev.
reactor vanadium remclval appears to obey 1.5-order PFR kinetics. We have found that this difference can be described on the basis of a spread in reactivity of the individual vanadium-bearing species. In this respect apparent reaction kinetics for the removal of heteroatoms such as vanadium and sulfur from (residual) petroleum fractions are similar. Nomenclature
C = concentration, kg/ kg d = particle diameter, m
8=space velocity, kg/(L h)
ka, = average first-order rate constant, kg/(L h) k , = nth-order rate constant (see eq 13) L = catalyst bed height,, m n = reaction order R = reaction rate, kg/('L h) r = recycle ratio, kg/kg U = catalyst use rate, kg/kg V , = reactor volume, L, WL = liquid mass in BSTR, kg a = conversion 4~ = liquid flow rate, kg/h 4~ = catalyst flow rate, L/h pc = catalyst bulk density, kg/L 7 = space time, reciprocal space velocity, L h/kg Indices C = catalyst L = liquid f = feed
635
1980, 19, 635-638
p = product
Abbreviations CSTR = continuous stirred tank reactor BSTR = batch stirred tank reactor
PFR = plug-flow reactor Literature Cited Beuther, H.; Schmid, B. K. "Proceedings, 6th World Petroleum Congress", Section 3, Paper 20, Hamburg, 1964. Chang, C. D.; Silvestri, A. J. Ind. Eng. Chem. Process Des. D e v . 1974, 13, 315. Chang, C. D.; Silvestri, A. J. Ind. Eng. Chem. Process D e s . Dev. 1978, 15, 161. Dautzenberg, F. M.; van Klinken, J.; Pronk, K. M. A,; Sie, S. T.; Wljffels, J-B. "5th International Symposium Chemical Reaction Engineering", Houston, 1978. De Bruijn, A. "Proceedings, 6th International Congress on Catalysis", Paper 834, London, 1976. Frye, C. G.; Mosby, J. F. Chem. Eng. Prog. 1967, 63, 66. Inoguchi, M., et al. Bull. Jpn. Pet. Inst. 1971, 13, 153. Kittrell, J. R.; Mezaki, R.; Watson, C. C. Ind. Eng. Chem. 1988, 58(5), 51. Kolboe, S. Can. J. Chem. 1989, 47, 352. Levenspiel, 0. "Chemical Reaction Engineering", Wiley: New York, 1972; Chapter 14. Massagutov, R. M.; Berg, G. A.; Kulinich, G. M.; Kirlllov, T. S . Proc. 7th WorM Pet. Cong. 1987, 4 , 177. Mears, D. E. Chem. Eng. Sci. 1971, 26, 1361. Oleck, S . M.; Sherry, H. S. Ind. Eng. Chem. Process Des. Dev. 1977, 16, 525. van Deemter, J. J. "Proceedings 3rd European Symposium on Chemical Reaction Engineering", p 215, Amsterdam, 1964. van Ginneken, A. J. J.; van Kessel, M. M.; Pronk, K. M. A.; Renstrom, G. Oil Gas J. Apr 28 1975, 59. van Klinken, J.; van Dongen, R. H. Chem Eng. Sci. 1980, 35, 59.
Receiued for review November 9, 1979 Accepted June 30, 1980
Gas Holdup in Gas-Liquid and Gas-Liquid-Solid Flow Reactors D. H. Y'lng,' E. N. Givens, and R. F. Welmer Air Products and Chemicals, Inc., Allentown, Pennsylvania 18 105
A recent article by Tarrer, Lee, and Guin, raised a number of questions concerning the methodology for designing a three,-phase reactor for the solvent refining of coal (SRC). Laboratory studies of two- and three-phase flow in subscale models of an SRC reactor have now resolved several of these questions. In particular, waterhitrogen and methanolhitrogen gas holdup data from our 2-in. and 5-in. diameter columns are in good agreement with the correlation of Yoshida and Akita. The presence of silica particles (-140 and 30-45 mesh) reduces gas void fraction at low superficial gas velocities. However, at higher gas flow rates (0.20 ft/s) the reduction effect of solid particles diminishes to a negligible level, suggesting that the correlation of Yoshida and Akita is equally adequate for a gas-liquid-solid system at high gas superficial velocities. Furthermore, gas distribution and liquid flow rate show no effect on gas holdup.
Introduction
Gas holdup in bubble columns has been a subject of great interest, partiwlarly in reactor column design. Numerous investigators have studied gas holdup with many different systems, covering a wide range of liquid viscosity (0.58-152.0 cP), surface tension-(22.3-?6.0 dyn/cm), and density (0.79-1.70 g/cm3). Recently, Tarrer e t al. (1978) discussed the evaluation of gas holdup in bubble columns in conjunction with a dispersion model for the solvent refined coal process. In particular, the authors examined the general. applicability of two gas holdup correlations. One was that of Hikita and Kikukawa (1974). ,, who proposed the expression tg = 0.505 V,0.47(72/~)2/3(1/p)0.05 (1)
who expressed the gas void fraction as a complex function of three dimensionless functional groups
--% (1 - E A 4
- Cl(N,o)1'8(NGa)1'12(NFr)
(2)
where
NE0 = gD2PL/g (Bond number) NG, = gD3/vL2 (Galileo number) NFr
=
vg/&3
(Froude number)
(3)
(4) (5)
~
The other correlation was from Yoshida and Akita (1973), 0196-4305/80/1119-0635$01 .OO/O
and C1 = 0.20 for nonelectrolyte solution, 0.25 for electrolyte solution. Equation 2 also predicts a holdup which is independent of column diameter, and it may be written in terms of only
0 1980 American
Chemical Society