638 Ind. Eng. Chem. Process Des. Dev. 1980, 19, 638-641

effect extended to slightly higher gas input rates (0.19 ft/s vs. 0.13 ft/s) as shown in Figure 6. This suggests that the gas turbulence required to reduce the coalescence effect is higher for larger solid particles. Increasing the average solid concentration from 15.7 wt % to 45.0 wt 5% produced no further systematic change in the gas holdup as shown by Figure 5.

Our laboratory data showed that gas holdup in bubble columns with suspended solid particles was not affected by the slurry flow rate. Three different slurry input rates, ranging from 5 to 24 cm3/s were employed in testing the slurry flow rate effect on gas holdup in the 5-in. diameter column. The results showed negligible differences between the gas holdups measured with and without slurry flow.

Both in the presence and absence of suspended solid particles, gas holdup at low gas velocities is by itself a small quantity. For instance, at a superficial gas velocity of 0.06 ft /s (corresponding roughly to the gas feed rate in the Wilsonville SRC pilot plant), the gas void fraction was reduced from 0.05 to 0.04 in the presence of solid particles as shown by Figure 5. This reduction has little effect on the liquid holdup. However, it is important to note that this change represents a 20% reduction in the gas void volume, which together with changes in bubble size could appreciably affect mass transfer rate. Consideration of the effect of solids suspension on bubble behavior may be

significant in the design of three-phase slurry reactors. Acknowledgment

The authors wish to thank J. Lopez and D. Schoene- berger for constructing and operating both columns. Nomenclature

t = superficial gas velocity, m/s in eq 1; cm/s in eq 5 and

CT = surface tension of liquid, dyn/cm p = viscosity of liquid, CP g = gravitational constant, cm/s2 D = diameter of column, cm pL = density of liquid, g/cm3 vL = kinematic viscosity of liquid, cmz/s Vp,, = bubble rise velocity, cm/s L i t e r a t u r e Ci ted Akita, K., Yoshida, F. I d . Eng. Chem. Process Des. Dev . 1973, 12, 76. Datta, R. L., Napier, D. H., Newitt, D. M. Trans. Inst. Chem. Eng. 1950, 28,

Hikita, H. Kikukawa, H. Chem. Eng. J . 1974, 8 , 191-197. Kato, Y., Nishiwaki, A., Fukuda, T., Tanaka, S. J. Chem. Eng. Jpn 1972, 5 ,

Taner, A. R., Lee, M. H., Guin, J. A. Ind. Eng. Chem. PTocessDes. Dev. 1978,

Received f o r review November 13, 1979 Accepted June 10, 1980

= gas void fraction

6

14.

112.

77, 127.

Power Requirements and Interfacial Area in Gas-Liquid Turbine Agitated Systems

Gordon A. Hughrnark

Ethyl Corporation, Baton Rouge, Louisiana 7082 1

Extensive data for flat-blade turbine impeller systems have been correlated to provide equations for power requirements and interfacial area. The gassed to ungassed power ratio is represented by

- p, = 0.10 (QV - j1I4( N2D4 )’I5

P go1 V2I3

and interfacial area by

Calculation of mass transfer coefficients from the interfacial area equation for oxygen absorption in water data indicates that the equation applies over a vessel size range of 10 to 51 000 L and for gas superficial velocities to 5.3 cmls.

Power requirements and interfacial area for the me- chanical agitation of aerated liquids have been extensively investigated. Flat-blade turbine impeller systems represent a major part of this published data. Correlations from these data for scale-up of power requirements and inter- facial area are of industrial interest. This paper presents the derivation of correlations for power requirements and for interfacial area that represent data for different liquid

0196-4305/80/1119-0638$01.00/0

physical properties and a range of impeller and vessel diameters. Data analysis is restricted to flat-blade turbines with Newtonian fluids and does not include electrolyte solution data. The analysis is also limited to systems in which the power input is provided by mechanical agitation with a minimum contribution from gas sparging. Miller (1974) proposed design methods for the combination of mechanical agitation and gas sparging.

0 1980 American Chemical Society

Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980 639

Power Requirements Michel and Miller (1962) reported aerated power data

with several liquids in 0.165 and 0.305-m diameter vessels with two different turbine impellers. A dimensional cor- relation was used to represent the data. Oyama and Endoh (1955) proposed the gassed to ungassed power ratio and an aeration number for a dimensionless correlation. Clark and Vermeulen (1963) suggested a Weber number as a correlating variable. These dimensionless groups have been used in a correlation proposed by Luong and Volesky (1979) for Newtonian. fluids

-0.38 N2D3p -’‘18

_ - PB - 0.497(&) (7) (1) P N D 3

Application of eq 1 to the extensive Michel and Miller data shows that this equation is restricted to the surface tension range of 55 to 72 dyn/cm and an impeller to vessel ratio of one-third as these conditions represent the data used to obtain the correlation. The equation is not applicable to the Michel and Miller data with carbon tetrachloride or with different impeller to vessel ratios. Power Requirement Data Analysis

The extensive data of Michel and Miller include fluids with a density range of 0.87 to 1.6 g/mL, a viscosity range of 0.8 to 28 cP, and ,a surface tension range of 25 to 72 dyn/cm. Maximum vessel diameter is 0.305 m and the range of impeller to1 vessel diameters is 0.25 to 0.46. Pharamond et al. (1975) obtained data for water with three vessel diameters (0.29,0.48, and 1 m). Impeller to vessel diameter is one-third. Luong and Volesky report data for a 0.222-m vessel with an impeller to vessel diameter of one-third. Liquids relpresent a viscosity range of 0.9 to 3.0 CP and a surface tenciion range of 55 to 72 dyn/cm. Gas rates are from 0.2 to 1.8 vol of gas per liquid volume per minute (VVM). These references provide 243 data sets with a Pg/P range of 0.31 to 0.8.

Calculated values of the power ratio with eq 1 for the 243 data sets show an average absolute deviation of 21.4% between the calculated and experimental values. Corre- lation with the model of eq 1 to obtain the least-square coefficients reduces the average absolute deviation to 16 70 between calculated and experimental values, but the coefficient for the Weber number is not statistically sig- nificant. This observation is consistent with the Michel and Miller correlation that does not include the surface tension as a significant variable. Correlations with revi- sions to the independent variables show that revision of the aeration number to include the liquid volume rather than the cube of the impeller speed improves the corre- lation coefficients. A Froude number related to the fluc- tuating velocity in the turbulent impeller stream is also effective in reducing the residual variance of the correla- tion. Schwartzberg and Treybal(1968) report fluctuating velocities for several different impeller and vessel diameters that correlate as a function of (tD)ll3. If the impeller stream diameter is defined as the impeller width, a Froude number is then W D 4 / g D i W . Thus, the model for the power ratio is of the form

Correlation of the 243 data sets with this model shows a correlation coefficient of 0.85. Coefficient values are 0 = -0.27 and y = -0.181. Correlation with 0 = - l f4 and y =

were found to only slightly increase the residual var- iance. Therefore, the resulting correlation equation is

Average absolute deviation between calculated and ex- perimental values is 8.5% using eq 3.

Bimbinet (1959) reports aerated power data for water with 10.2, 12.7, and 15.2-cm diameter impellers in 0.305 and 0.457-cm diameter vessels. Data are also reported for syrup solutions with a 10.2-cm diameter impeller in a 0.305-m diameter vessel. Maximum liquid viscosity is 90 cP. The Bimbinet data provide an additional 148 data sets. Combination of these data with the prior data result in a total of 391 data sets. Average absolute deviation between calculated and experimental values for the total data is 11.7% with eq 3. This appears to be an excellent representation of power requirements for the wide range of conditions represented by the 391 data sets. As eq 3 represents data for vessel diameters from 0.165 to 1 m, this would be expected to apply to large vessel diameters where the liquid depth to diameter ratio is typically about unity. Interfacial Area

Calderbank (1958) correlated interfacial area data for 5 and 100-L vessels with the dimensionless group (PI V)o.4a0.2/a0.6 and the square root of the gas superficial velocity. Impeller to vessel ratio was one-third. Liquid systems with a wide range of physical properties were represented. For scale-up with geometric similarity and with constant volume of gas to volume of liquid, this provides interfacial area as a function of W.2D1.3. Cal- derbank also proposed a “back-mixing” correction factor for high Reynolds numbers. Van Dierendonck et al. (1968) presented correlations for gas holdup and bubble diameter. Constant bubble diameter was reported for impeller speeds much greater than the minimum speed for dispersing gas bubbles; thus interfacial area scale-up is essentially rep- resented by the gas holdup. Scale-up with geometric sim- ilarity and with constant volume of gas to volume of liquid shows interfacial area as a function of ND0.93.

Data analyses for gas holdup and bubble diameter are presented in this paper. These can be combined to provide a correlation for interfacial area. Gas Holdup Data Analysis

Gas holdup data for water and cyclohexane with a 0.165-m vessel diameter and 0.06-m impeller diameter are reported by van Dierendonck et al. Kawecki et al. (1967) report data for water with a 0.191-m vessel and 0.0765-m impeller. Brown and Craddock (1969) also report data for water with a 0.22-m vessel and impeller diameters of 0.076, 0.10, and 0.126 m. Rushton and Bimbinet (1968) show data for a 0.304-m vessel with a 0.164-m impeller. Farritor and Hughmark (1980) report interfacial areas for p-xylene and tetradecane systems. Gas rates range from 1.2 to 36 VVM. These references provide 77 data sets with a range of gas holdup volume fractions from 0.02 to 0.17. Ex- perimental methods represent observed gas fractions and interfacial area by a chemical method.

The success in using the aeration and Froude numbers represented by eq 2 for aerated power requirements sug- gests that these dimensionless groups might also be ap- plicable to gas holdup. A Weber number based upon a gas bubble in the turbulent impeller stream is another pos- sibility. The velocity is the fluctuating velocity. This provides a model of the form

640 Ind. Eng. Chem. Process Des. Dev., Vol. 19, No. 4, 1980

Calderbank (1959) reported mass transfer data for gas- liquid systems with gas bubbles in the range of 0.2 to 0.5 cm. Sridharan and Sharma (1976) determined mass transfer coefficients by gas entrapment in an agitated system and Farritor and Hughmark report coefficients for a sparged-agitated system that indicate reasonable agreement with the Calderbank data for this bubble size range. Therefore, a bubble diameter of 0.25 cm was se- lected for use with eq 4 and this was assumed to be, con- stant for the entire data set. Correlation of the data with this model results in a correlation coefficient of 0.985 with coefficients of 0 = 0.517, y = 0.537, and a = 0.248. Standard deviations of the coefficients show that all three are statistically highly significant. Correlation with 0 =

/2, y = only slightly increased the residual variance. The resulting correlation equation is

and a = 1

Average absolute deviation between calculated and ex- perimental values for the 77 data sets is 8%. The complete Bimbinet data were obtained after this analysis. Addition of the data which were not included in the prior analysis provide a total of 208 data sets. Average absolute deviation between calculated and experimental values is 11.5% for these 208 sets which shows excellent agreement between the correlation and the data. The Bimbinet data include a 0.457-m diameter vessel which increases the vessel di- ameter range of the initial data.

Rushton and Bimbinet show a threefold increase in gas holdup with 3 CP aqueous corn syrup in comparison to water. The cyclohexane data (2.2 cP) of van Dierendonck et al. and the tetradecane solution data (2.6 cP) of Farritor and Hughmark do not show this viscosity increase so liquid viscosity is not included as a variable in eq 5.

Bubble Diameter Analysis Bubble diameter data for the air-water system are re-

ported by Lee and Meyrick (1970) and Brown and Crad- dock. Calderbank shows data for water, ethanol, and ethyl acetate. The Eotvos number provides a dimensionless representation of bubble diameter which can then be correlated with gas holdup, any of the dimensionless groups from eq 5, and the gassed to ungassed power ratio. The following equation was found to provide a good rep- resentation of the data

D2&P $112

(6) --

- 5*5 ( gDiV2J3(Pg/P)'J3 hnD3 >"' U

Average absolute deviation between calculated and ex- perimental values for the 39 data sets is 10.6%.

Equations 3 ,5 , and 6 can then be combined to provide an equation for interfacial area n = - ( : ) ' I2( &)1/3( ~ - 2 ~ 4 )0.5g2( D f l D 4 ) 0 ' 1 8 7

1.38 - gDiV2J3 a w 3

(7)

The Weber number contains D, assumed as 0.25 cm as for eq 5. It is apparent that D, as a variable from eq 6 could be used for the correlations represented by eq 5 and 7. The excellent fit of eq 5 to the data shows that this additional complexity is not justified. For scale-up with geometric

Table I. Calculated Mass Transfer Coefficients vessel no. of kL (mean), kL (std dev), size, L data sets cm/s cm/s

10 15 0.00795 0.0024 200 13 0.0131 0.0033 550 13 0.01265 0.0013

3000 12 0.020 0.006 0.00185 51000 8 0.00845

Table 11. Physical Parameters of the Agitated Systems

vessel diam-

Bimbinet (1959) 0.305, 0.457 Brown and 0.22

Calderbank (1959) 0.19, 0.51 Farritor and 0.10

reference eter, m

Craddock (1969)

Hughmark (1980) Fuchs et al. (1971) Kawecki (1967) Lee and

Meyrick (1970) Luong and

Volesky (1979) Michel and

Miller (1962) Pharamond (1 97 5)

Rushton and Bimbinet (1968)

Van Dierendonck et al. (1968)

0.21-3.33 0.191 0.304

0.222

0.165, 0.305

0.29, 0.48, 1.0

0.304

0.165

impeller to vessel diam-

eter ratio 0.333-0.50 0.34 5-0.57 3

0.333 0.50

0.4 11-0.576 0.40 0.333

0.333

0.25-0.46

0.333

0.54

0.364

similarity and with constant volume of gas to volume of liquid, interfacial area is a function of N.22Do.906.

Fuchs et al. (1971) report mass transfer data (kLa) for sparged oxygen in water with agitated vessels from 0.21 to 3.33-m diameter. Impeller to vessel diameter ratios were from 0.411 to 0.576 and gas rates were from 0.2 to 3.5 VVM. Calderbank's experimental data (1959) indicate that mass transfer coefficients (kL) are a function of liquid phase diffusivity and not of power input or gas rate for agitated vessels. Mass transfer coefficients can be calcu- lated from the kLa data with eq 7.

Table I summarizes the calculated values from the data of Fuchs et al. for the ungassed power input range of 0.2 to 5 W/L which is the range of interest for most turbine agitated systems. The data do not show a significant trend in kL with vessel size. Calderbank's data indicate kL = 0.006 cm/s for oxygen in water. The mean values of kL from Table I other than for the 3000-L vessel are ap- proximately within a factor of 2 of the Calderbank value. This indicates that eq 7 provides a reasonable approxi- mation for design over a large range of vessel sizes.

Table 11 provides the vessel diameters and impeller to vessel diameter ratios for the systems representing data for the correlations. Standard flat-blade turbines are re- ported or assumed if not reported. These have six blades with blade width of the impeller diameter. Liquid depth was generally equivalent to vessel diameter, but represent the range of 0.75 to 1.87 for the ratio of liquid depth to vessel diameter. Impeller depth was one-third to two-thirds of the liquid depth. Liquid depth and im- peller depth do not appear to be significant variables within these ranges. Conclusion

Equations are reported for gassed power input, gas holdup, and bubble diameter that represent extensive experimental data within what is probably the experi- mental error of the data for flat-blade turbines. An

Ind. Eng. Chem. Process Des. Dev. 1980, 19, 641-646 641

equation for interfacial area from combination of the gas holdup and bubble diameter equations appears to apply over a vessel size range from 10 to 15000 L and for gas superficial velocities to 5.3 cm/s. Nomenclature a = interfacial area D = impeller diameter Di = impeller blade width D, = bubble diameter g = acceleration of gravity kL = mass transfer coefficient N = impeller speed P = power input to uingassed liquid Pg = power input to gassed liquid Q = gas rate V = liquid volume Greek Letters e = energy dissipation 4 = gas holdup p = liquid density a = surface tension

Literature Cited Bimbinet, J. J., M.S. Thesis, Purdue University, Lafayette, Ind., 1959. Brown, D. E., Craddock, J., paper presented at the Symposium on Mixing,

Calderbank, P. H., Trans. Inst. Chem. Eng., 37, 443 (1958). Calderbank, P. H., Trans Inst. Chem. fng., 38, 173 (1959). Clark, M. W., Vermeulen, T., UCRL-10996, University of California, Berkeley,

Farritor, R. E., Hughmark, G. A,, Chem. Eng. Commun., 4, 143 (1980). Fuchs, R. Ryee, D. D. Y., Humphrey, A. E., Ind. Eng. Chem. Process Des.

Kaweckl, W., Reith, T., van Heuven, J. W., Beck, W. J., Chem. fng . Sci., 22,

Lee, J. C., Meyrick, D. L., Trans. Inst. Chem. fng . , 48, T37 (1970).

Mlchel, B. J., Miller, S. A,, AIChE J., 8, 262 (1962). Miller, D. N., AIChEJ., 20, 445 (1974). Oyama, Y., Endoh, K., Chem. fng. Jpn., 10, 2 (1955). Pharamond, J. C., Roustan, M., Roques, H., Chem. Eng. Sci., 30, 907 (1975). Rushton, J. H., Blmbinet, J. J., Can. J. Chem. fng . , 48, 16 (1968). Schwartzberg, H. G., Treybai, R. E., Ind . Eng. Chem. Fundam., 7, 1 (1968). Sridharan, K., Sharma, M. M., Chem. Eng. Sci., 31, 767 (1976). Van Dierendonck, L. L, Fortuin, J. M. H., Vanderbos, D., Fourth European

Received for review January 14, 1980 Accepted May 6, 1980

Institute of Chemical Engineers, Leeds, Sept 1969.

1963.

D e v . , 10, 190 (1971).

1519 (1967).

Luong, H. T., Volesky, B., AIChE J., 25, 893 (1979).

Symposium on Chemical Reaction Engineering, Brussels, 1968.

Oxidative Rapid Pyrolysis of Coal in a Dilute Phase Transport Reactor

Oto Sltnal' and David J. McCarthy Commonwealth Scientific and Industrial Research Organization, Division of Mineral Engineering, Clayton, Victoria 3 168, Australia

A process for oxidative flash pyrolysis of coals or other carbonaceous materials is described. Oxidation at low values of oxygen to coal ratio is used to raise the temperature of a preheated feedstock to pyrolysis temperature while simultaneously destroying any caking properties of the feedstock. Experimental results for oxidative pyrolysis of a high volatile bituminous coal in a laboratory pyrolyzer using both single stage and multistage additions of oxygen are given. The measured tar yields are found to be comparable with the value from the Gray-King assay of the coal.

Introduction Pyrolysis of small particulate coal under rapid heating

conditions and at a short residence time is a well known practice. Yields of' primary decomposition products, namely oil and tar, sometimes higher than the Fischer assay can be achieved. Rapid or flash heating of coal in a pyrolysis reactor is iusually accomplished by rapid mixing of coal particles with hot particulate or gaseous heat car- riers. Fluidized bed reactors with pyrolysis char as the heat carrier has been the preferred pyrolysis technique, as for example in the following processes: COED (Jones, 1973), Lurgi-Ruhrgas (Doring et al., 1975), COALCON (Martin, 1975), Clean Coke (Johnson et al., 1975), FLUPROC (Waters et al., 1978)i.

High volatile coals are the favored feedstock for rapid pyrolysis because of their potentially large yields of tar and oil, but when rapidly heated they also tend to be strongly agglomerating. In fact, some of the coals which are non- caking under slow heating conditions agglomerate when flash heated in a pyrolysis reactor. Various methods for controlling agglomeration during pyrolysis have been proposed and tried, e.g., staged heating of coal in several fluidized bed reactors (COED process), jet mixing of coal with char (Lurgi-Ruhrgas process), mechanical mixing of coal with hot large inert particles (Toscoal process) in a rotary bed reactor (Carlson et al., 1973), mechanical mixing

0 196-4305/80/ 1 1 19-064 1$0 1 .OO/O

of coal with hot char before entering a fluidized bed reactor (another version of the Lurgi-Ruhrgas process) (Marnell, 1976), pneumatic mixing of particulate coal with a large proportion of hot char in a dilute phase suspension (Garrett process) (Green, 1975), mixing of coal with hot char in fluidized beds (Phinney, 1955; Choi, 1977), partial combustion of a mixture of coal and char in a fluidized bed reactor (Parry process) (Parry et al., 1956) and for hy- dropyrolysis, the application of rocket engine techniques to achieve rapid mixing and reaction (Oberg et al., 1978).

The main emphasis in the development of these pro- cesses was toward maximizing the tar yield and simulta- neously coping with the troublesome coal agglomeration. This resulted in the development of fairly complicated processes. For example, a large recirculation of a high temperature particulate heat carrier between the pyrolysis reactor and special vessels for heating and storing the heat carrier was usually used. Most of the processes are tailored to a particular type of coal and are not suitable for the pyrolysis of other more difficult coals. In many cases it was necessary, as the last resort, to treat the coal prior to pyrolysis, for example by preoxidation (Forney et al., 1964; Curran et al., 1973).

The liquid product from pyrolysis of coal is inferior to crude oil as refinery feedstock, and expensive hydro- treatment is required to enhance the hydrogen/carbon

0 1980 American Chemical Society