Ind. Eng. Chem. Res. 1996, 35, 3597-3606
3597
On the Performance of a Cascade Crossflow Air Stripping Column Y. Akiyama, K. T. Valsaraj, D. M. Wetzel, and D. P. Harrison* Department of Chemical Engineering, Louisiana State University, Baton Rouge, Louisiana 70803
The cascade crossflow packed column is an innovative design that offers the separation advantages of countercurrent flow while avoiding flooding limitations. Liquid and gas crosssectional flow areas and path length in contact with packing may be controlled independently. These features are illustrated by studying the air stripping of methylene chloride (MeCl), 1,2dichloroethane (1,2-DCA), and methyl ethyl ketone (MEK). Stripping efficiencies in the cascade crossflow column were generally slightly smaller than in countercurrent flow at equal liquid and gas flow rates. However larger gas-to-liquid ratios were possible in crossflow, permitting larger maximum stripping efficiencies to be attained. The experimental mass transfer coefficients were smaller than predicted by the Onda correlation. Modifications to the gas-phase Onda correlation are proposed that reduce the magnitude of the average deviation between experiment and prediction for 40 tests to about 12%. Experimental values of 38 of 40 tests were within (30% of modified Onda correlation predictions. Additional applications for the cascade crossflow concept are suggested, such as vacuum distillation columns and trickle-bed reactors, which require low pressure drop and/or large gas-to-liquid ratios. Introduction Packed column gas-liquid separation processes have traditionally been operated with either countercurrent or cocurrent flow of gas and liquid. While separation factors equivalent to a number of equilibrium stages are possible in countercurrent operation, the allowable gas and liquid rates are limited by flooding. In cocurrent flow, flooding is not a problem, and gas and liquid flow rates may be as much as an order of magnitude larger. The more intense contacting achieved at such high rates may lead to higher mass transfer efficiencies (King, 1980). The separation factor, however, is limited to that corresponding to one equilibrium stage. The cascade crossflow contactor shown schematically in Figure 1 offers the separation advantages of countercurrent flow while avoiding the flow rate limitations associated with flooding. Only a portion of the column cross section is packed, with the packing held in place by two vertical screens extending the length of the column. Liquid flow is vertically downward as in the traditional countercurrent contactor. The gas plenums on either side of the packing are fitted with partial baffles which cause the gas to flow back and forth across the packing at approximately right angles to the liquid. Gas enters at the bottom and exits at the top, thereby providing overall countercurrent flow. At high gas rates, a portion of the liquid may be deflected into the gas plenum, where it collects on the baffle below. Liquid draining from the baffle encounters gas flowing in the opposite direction, which redistributes the liquid back toward the center of the packing. The alternating gas flow directions cause the system to be self-correcting and permit stable operation at conditions which would flood a countercurrent column. Additional design freedom is provided by the cascade crossflow contactor in that both the fraction of the crosssectional area filled with packing and the baffle spacing may be varied. Thus, the cross-sectional areas through which the gas and liquid flow as well as the gas and liquid path lengths through the packing may be independently controlled. In contrast, for both countercurrent and cocurrent flow, the cross-sectional flow area * To whom correspondence should be addressed.
S0888-5885(96)00178-9 CCC: $12.00
Figure 1. Schematic diagram of the cascade crossflow contactor.
for both phases is equal to the cross-sectional area of the column, and the contact path length for both phases is equal to the height of packing. Widely spaced baffles produce small gas velocity and small gas path length in contact with the packing. This results in low gas-phase pressure drop and the ability to achieve large volumetric gas flow rates with low power input. As the baffle spacing is decreased at constant gas volumetric flow rate, both the gas velocity and path length through the packing increase. In turn, the pressure drop increases but remains below that associated with countercurrent flow at the same liquid and gas loading rates. When mass transfer is controlled by the liquid-phase resistance, the variation in gas flow characteristics has little effect on mass transfer ef© 1996 American Chemical Society
3598 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 Table 1. Properties of Test Compounds at 25 °C chemical structure molecular weight density (kg/m3) vapor pressurea (mmHg) solubilitya (mg/L) Henry’s constantb a
MeCl
1,2-DCA
MEK
CH2Cl2 84.94 1.318 426.0 16700 0.1211
ClCH2CH2Cl 98.97 1.246 109.3 8700 0.0577
CH3COCH2CH3 72.11 0.803 77.5 260000 0.0053
Gossett et al. (1985). b Ashworth et al. (1988).
ficiency. In contrast, when the gas-phase resistance is significant, the mass transfer efficiency becomes a function of the gas flow characteristics. A series of earlier papers from our laboratory examined the performance of the cascade crossflow contactor when applied to the air stripping of both volatile and semivolatile compounds from water. Wood et al. (1990) studied the effect of baffle spacing on stripping efficiency and compared experimental overall mass transfer coefficients to predictions using a modified Onda (1968) correlation. The agreement was generally within the range of (20% for the more volatile compounds, such as carbon tetrachloride and chloroform, but not so good for less volatile 1,2-dichloroethane (1,2-DCA). Mertooetomo et al. (1993) studied the air stripping of 1,2-DCA and compounds of lower volatility. Wide baffle spacings were used in order to achieve the large gas flow rates required if compounds having small Henry’s constants were to be stripped with high efficiency. Pressure drop reductions of approximately 2 orders of magnitude compared to that experienced in countercurrent flow were observed. The large gas flow rates made possible by the small pressure drop were sufficient to produce high stripping efficiency for methyl ethyl ketone (MEK), a compound which normally requires the use of steam or preheated water for effective stripping. Field tests using “identical” countercurrent and cascade crossflow stripping columns were reported by Verma et al. (1994). Actual ground water contaminated with about 1 ppm 1,2-DCA was used. Laboratory results concerning reduced pressure drop, higher achievable air flow rate, and increased stripping efficiency in the cascade crossflow column were confirmed. For example, at water feed rates of 0.63 kg/s, the pressure drop-gas flow rate combination required to achieve 95% stripping efficiency resulted in approximately 35% lower power consumption in the cascade crossflow unit. In addition, at a water loading rate of 18.9 kg/m2‚s, the maximum stripping efficiency which could be achieved in the countercurrent stripper was 0.92. Because of the lower pressure drop, larger gas flows were possible in the cascade crossflow column, which resulted in a maximum stripping efficiency of 0.99. Results of a similar field comparison of countercurrent and cascade crossflow strippers for removal of 1,2-DCA have been published by Gavaskar et al. (1995). The current study has extended the previous work by examining the performance of a laboratory semibatch cascade crossflow stripper using small baffle spacings to produce a large number of crossflow passes. Stripping efficiency and gas-phase pressure drop were measured as a function of gas and liquid flow rates. Overall mass transfer coefficients were calculated from the experimental data and compared to predictions using the Onda (1968) correlation. Three organic compounds were studied: methylene chloride (MeCl), 1,2-DCA, and MEK. Selected properties of these compounds at 25 °C are summarized in Table 1.
Experimental Apparatus and Procedure The laboratory-scale semibatch cascade crossflow air stripper which has been previously described (Wood et al., 1990; Mertooetomo et al., 1993) was used. The 0.152 m (6 in.) i.d. column was packed with 0.016 m (5/8 in.) nominal polypropylene Pall rings to a packed height of 2.44 m (8 ft). The vertical screens were arranged so that the central 65% of the cross-sectional area was packed, leaving open plenums occupying approximately 17% of the cross-sectional area on either side. The spacing of the partial baffles was varied to provide 8, 24, or 48 crossflow passes through the packing. Figure 2 provides a visual image of the column geometries as well as the actual dimensions. Under ideal conditions in which the liquid flow is vertical and the gas flow horizontal, the liquid flow path was constant at the packed height of 2.44 m while the gas flow path varied from 0.8 m using 8 crossflow passes to 4.8 m using 48 passes. In selected tests the column internals were removed, allowing the entire cross-sectional area to be packed, and the unit was operated in true countercurrent fashion. At the beginning of a test, the reservoir was charged with an aqueous solution of the desired organic. The solution was continuously pumped to a distributor at the top of the column, from where it flowed downward through the packing by gravity and returned to the reservoir. Care was taken to ensure that the reservoir contents were well mixed at all times. A variable speed centrifugal blower forced air through the column, with the air entering in the bottom crossflow section and exiting from the top. Exit air containing stripped organic was discharged through the laboratory exhaust system. Column pressure drop was measured using a U-tube manometer filled with n-hexane. After the desired air and water flow rates were established, sufficient time was allowed for the column to stabilize, and liquid samples were taken simultaneously from the top and bottom of the column. Three duplicate top and bottom samples were taken at 5 min intervals at each set of operating conditions. The samples were stored in 5 mL glass vials. Care was taken to minimize exposure during transfer and to ensure that no vapor space was left in the vials. A Hewlett-Packard 5890A gas chromatograph equipped with a flame ionization detector was used for analysis. Separation was achieved using a 0.0032 m (1/8 in.) diameter by 1.92 m (5 ft) column packed with 80/120 carbopack B coated with 3% SP-1500. Helium was used as the carrier gas. Initial solute concentrations were in the range 500-700 ppm, and a test was continued until the concentration approached the lower detection limit of about 20 ppm. At this point the reservoir was recharged with additional solute and another test was initiated. Each sample was subjected to repetitive analysis until the standard deviation of the integrated peak areas was no more than 5%. Results described in the following section are based upon the arithmetic mean of the three duplicate data sets taken at constant operating conditions. Stripping efficiency, E, was calculated from the measured concentrations at the top, CLT, and bottom, CLB, of the packing using the equation
E ) 1 - CLB/CLT
(1)
The overall liquid-phase volumetric mass transfer coefficient, K1a, was calculated from the measured efficiency using (Mertooetomo et al., 1993)
Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3599
Figure 2. Cascade crossflow column geometries used in the current study.
EL L 1 ln 1( AZ) [1 - E( GH)] K a) 1
1-
L GH
(2)
Values of Henry’s constant, H, required for the calculation were taken from Table 1 except when specifically stated otherwise. Experimental Results Flow Characteristics. The liquid loading rate was varied between 8.0 and 22.4 kg/m2‚s (1.64-4.59 lb/ft2‚s) while the gas rate was varied up to the blower capacity. The maximum gas rate decreased as the liquid rate and number of crossflow passes increased, but was always significantly larger than could be obtained in countercurrent flow. This is illustrated in Figure 3 where the maximum gas loading rate, Gm)max, is plotted versus liquid loading rate, Lm. Both Gm and Lm are defined on the basis of the cross-sectional area of the column which is packed. A number of duplicate test results are shown, and the scatter is attributed to different packing characteristics which resulted when the column internals were altered and the column was repacked. At equal liquid loadings, the maximum achievable gas loadings for the crossflow configurations of 8, 24, and 48 passes were approximately 400%, 225%, and 110%
larger than the countercurrent maxima. Several of the cascade crossflow conditions were within the flooding region in countercurrent flow. On a volumetric basis, the Figure 3 results range from Gmax/L ∼ 20 for countercurrent operation at a liquid loading of 17.6 kg/ m2‚s to Gmax/L ∼ 410 for an 8-pass crossflow column at a liquid loading of 8 kg/m2‚s. The ability to achieve high gas rates is crucial if contaminants having small Henry’s constants such as MEK are to be stripped with high efficiency. The minimum G/L required for complete stripping, i.e., E ) 1, of a solute is
G ) H-1 L min
)
(3)
Thus, complete stripping of MEK at 25 °C would, according to Table 1, correspond to G/L)min ) 189, a value which cannot be reached using reasonable liquid loadings in countercurrent flow, but was easily achieved in a number of crossflow situations. Pressure drop as a function of gas loading rate and number of crossflow passes is shown in Figure 4 for a liquid loading rate of 12.8 kg/m2‚s. Countercurrent experimental data as well as the manufacturer’s correlation for countercurrent flow are included for comparison. In the midrange of the gas loading rates, the crossflow pressure drop is approximately linear with a
3600 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Figure 3. Maximum achievable gas loading rate as a function of liquid loading rate and column geometry.
Figure 4. Pressure drop as a function of gas loading rate and column geometry: Lm ) 12.8 kg/m2‚s.
slope near 2. At low gas loading the slope is less than 2 while at high gas loading the slope is greater than 2. The lower slope is attributed to the fact that gas flow is not fully turbulent at small gas loading, and the larger slope is attributed to the buildup of liquid on the baffles and the resultant bubbling of gas through the liquid pools. The most important result from Figure 4 is the very large reduction in pressure drop associated with the crossflow configuration. For example, at a gas loading rate of 1 kg/m2‚s, the pressure drop was approximately 30 times smaller using 8 crossflow passes and 3 times smaller using 48 crossflow passes than the comparable countercurrent pressure drop. This reduction in pressure drop is, of course, the reason that the larger gas rates shown in Figure 3 are possible. With 8 and 24 crossflow passes and low gas loading rates, the water appeared to fall vertically and produce complete wetting of the packing. As the gas loading increased, water began to be deflected out of the packing and accumulated on the next lower baffle. At still larger gas loading rates, increased deflection produced a larger pool of liquid on the baffle, and gas began to bubble through this pool. However, as previously noted, liquid overflowing from the baffle was deflected back toward the center of the packing by gas flowing in the opposite direction. Column operation was stable, and as will be
shown in the following section, good mass transfer characteristics were maintained. The initiation of bubbling corresponded roughly to the point at which the slope of the pressure drop curve began to increase above 2. With 48 crossflow passes, water deflection occurred even at low gas rates. At large liquid and gas loadings, a portion of the gas clearly short-circuited the ideal crossflow pattern by flowing directly around each baffle. Some gas bypassing probably occurred with the other baffle spacings, but to a much smaller extent. Foaming was observed on the baffles during MEK studies at large liquid and gas loadings, but no significant foaming occurred at any operating conditions using 1,2-DCA or MeCl. Stripping Efficiency. Stripping efficiency was studied as a function of compound volatility (Henry’s constant), baffle spacing (number of crossflow passes), and gas and liquid flow rates. A limited number of countercurrent tests were performed by removing the column internals and packing the entire cross-sectional area. While stripping efficiencies were generally slightly lower in the cascade crossflow configuration at equal gas and liquid loading rates, higher stripping efficiencies were obtained in a limited number of cases. Higher stripping efficiency coupled with lower pressure drop makes these configurations and operating conditions particularly attractive.
Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3601
Figure 5. Cascade crossflow stripping efficiency for the three test compounds as a function of the volumetric gas-to-liquid ratio: Lm ) 17.6 kg/m2‚s, N ) 8.
Figure 6. Generalized relation between stripping efficiency and (G/L)/(G/L)min: N ) 24.
Figure 5 shows the effects of compound volatility and volumetric gas-to-liquid ratio on stripping efficiency using a liquid loading rate of 17.6 kg/m2‚s and 8 crossflow passes. As expected, stripping efficiency increased with both H and G/L. Ninety-five percent stripping efficiency was achieved with MeCl at G/L ∼ 80 and with 1,2-DCA at G/L ∼ 125. Pressure drops at these conditions were ∼80 and 220 N/m2/m, respectively. The maximum G/L which could be achieved when stripping MEK was 129 because of foaming, and the stripping efficiency was 35%. In a countercurrent test using MeCl at Lm ) 17.6 kg/m2‚s and G/L ) 20, the stripping efficiency and pressure drop were 0.90 and 120 N/m2/m. Thus for MeCl, the cascade crossflow column produced an increase in stripping efficiency from 0.90 to 0.95 with a 33% reduction in pressure drop. Comparison of the stripping efficiency on the basis of the ratio (G/L)/(G/L)min, with G/L)min defined by eq 3, removes the effect of Henry’s constant and effectively collapses the data for all compounds at all liquid rates into a single curve as shown in Figure 6 using 24 crossflow passes. The knee in the curve occurs at (G/ L)/(G/L)min ∼ 2.5 and corresponds to E ∼ 0.92. Using eq 3 and the Henry’s constant data from Table 1, these values correspond to G/L ) 22, 45, and 470 for MeCl, 1,2-DCA, and MEK, respectively. These G/L values
were easily achieved for MeCl and 1,2-DCA over a range of liquid rates, but G/L ) 470 for MEK was beyond the capacity of our blower even at the smallest liquid rate studied. The effect of the number of crossflow passes at constant Lm and G/L is illustrated in Figure 7 for 1,2DCA. Data from Mertooetomo et al. (1993) for 1, 2, and 3 crossflow passes are included. The effect is greatest at small N and small G/L. For example, at G/L ) 22, E increased from 0.67 with N ) 1 to 0.84 with N ) 24, but only to 0.86 at N ) 48. This behavior is attributed to the increasing importance of the gas-phase resistance at small G/L and N. As both increase, the gas-phase resistance becomes smaller and the stripping efficiency approaches a constant value. With 48 crossflow passes, stripping efficiencies equal to or larger than and pressure drops smaller than the countercurrent values at the same Lm and G/L were found in a number of tests. The relevant data are presented in Table 2. As an example, at G/L ) 44, the stripping efficiency with 48 crossflow passes was 2 points larger, 0.93 versus 0.91, while the pressure drop was 40% smaller. G/L values as large as 160, which produced E ) 0.97, were possible with 48 crossflow passes, while the countercurrent maximum G/L was about 86 with the corresponding E ) 0.96.
3602 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Figure 7. Variation in stripping efficiency with the number of crossflow passes: 1,2-DCA, Lm ) 8.0 kg/m2‚s. Table 2. Comparison of 1,2-DCA Stripping Efficiency and Pressure Drop for the Crossflow and Countercurrent Configurations: Lm ) 8 kg/m2‚s crossflow (N ) 48)
countercurrent
G/L
E
∆P(N/m2/m)
E
∆P(N/m2/m)
22 44 86 160
0.86 0.93 0.96 0.97
11 35 136 417
0.84 0.91 0.96 naa
16 59 220 na
a
Not achievable.
Mass Transfer Coefficients The overall volumetric liquid-phase mass transfer coefficient, K1a, was calculated from experimental stripping efficiency data using eq 2. The effect of gas loading rate on K1a is shown in Figure 8 for MeCl using 8 crossflow passes at three liquid loading rates. At small Gm (e0.2 kg/m2‚s), K1a increased rapidly. In this region the dominant resistance was contributed by the gas phase. At Lm ) 8 kg/m2‚s, an abrupt transition to liquid-phase control occurred at Gm ∼ 0.5 kg/m2‚s. At the two larger liquid rates, both liquid-phase and gasphase resistances contributed to the total resistance. The effect of liquid loading rate at Gm ) 2.5 kg/m2‚s and N ) 8 is shown in Figure 9 for the three compounds. There is an almost linear dependence of K1a on Lm for both MeCl and 1,2-DCA while, for MEK, K1a is effectively independent of Lm. At this high gas rate, the liquid-phase resistance effectively controls the rate of mass transfer for both MeCl and 1,2-DCA while the gasphase resistance controls for MEK. Experimental cascade crossflow mass transfer coefficients sometimes exceeded the countercurrent values at equal gas and liquid loading rates. This situation occurred with each solute although the most significant increases were associated with MEK at G/L values approaching the maximum achievable in countercurrent flow. For example, with N ) 24, Lm ) 12.8 kg/m2‚s, and G/L ) 42, the experimental K1a of 0.00145 s-1 in crossflow was about 70% larger than the countercurrent value. Similarly, with N ) 24, Lm ) 8 kg/m2‚s, and G/L ) 84, the crossflow mass transfer coefficient was more than 50% larger. Comparison with the Onda Correlation The Onda correlation (Onda et al., 1968) for K1a was developed for countercurrent flow and has been recom-
mended as the best correlation for air stripping applications (Staudinger, 1990). However, the reliability of the Onda correlation for less volatile compounds at conditions where the gas-phase resistance is important has been questioned (Roberts et al., 1985; Gossett et al., 1985; Harrison et al., 1993). Wood et al. (1988) first modified the Onda correlation for the cascade crossflow stripper by basing the gas loading rate on the vertical flow area between opposite baffles instead of the horizontal packed cross-sectional area. This modification made the gas-phase mass transfer coefficient, kga, a function of the number of crossflow passes but had no effect on the liquid-phase coefficient, k1a. As a result, there was little effect on the overall mass transfer coefficient, K1a, for liquidphase controlled systems, but the modification was quite important for systems having large gas-phase resistance. Before comparing predicted K1a to experimental values, it is necessary to examine some characteristics of eq 2. The theoretical maximum value of E for specified flow rates and Henry’s constant is
Emax ) GH/L ) 1.0
when
GH/L < 1.0
when
GH/L g 1.0
(4)
The nature of eq 2 is such that as the stripping efficiency approaches either 0 or the eq 4 maximum, small errors in E produce large errors in K1a. This behavior is illustrated in Figure 10, where the percent error in K1a resulting from a positive 0.01 error in E is plotted versus E. The parameter values of GH/L represent the approximate range of conditions covered in the experimental tests. Thus, if the error in the experimental K1a is to be kept within reasonable limits, say (10%, it is necessary to exclude all data in which E < 0.1, and also to establish a maximum limit on E which depends on the value of GH/L. All data used in the following analysis adhere to this requirement. In addition, results of tests in which there was visible buildup of liquid on the baffles and gas bubbling through the liquid pool have been eliminated. Data from tests using 48 crossflow passes have also been excluded because of the previously described flow maldistribution. Results from 40 tests (14 involving 1,2-DCA, 10 with MEK, and 16 using MeCl) have been used in the following analysis.
Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3603
Figure 8. Effect of gas and liquid loading rates on the overall volumetric mass transfer coefficient: MeCl, N ) 8.
Figure 9. Effect of liquid loading rate on the overall volumetric mass transfer coefficient for the three test compounds: Gm ) 2.5 kg/ m2‚s, N ) 8.
Figure 10. Error in the experimental overall volumetric mass transfer coefficient caused by a positive 0.01 error in the measured stripping efficiency.
Figure 11 compares experimental values of K1a for these tests to predictions using the Onda correlation as modified by Wood et al. (1990). All but one of the experimental values are below the parity line with an
average deviation of about -20% for 1,2-DCA, -60% for MEK, and -30% for MeCl. Although flow in the cascade crossflow stripper is quite different from that in the countercurrent stripper
3604 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Figure 11. Parity plot comparing overall volumetric mass transfer coefficients from experimental data and from the Onda correlation as modified by Wood et al. (1990).
for which the Onda correlation was developed, the above results confirm the concern over the adequacy of the Onda correlation when the gas-phase resistance is important. Since the Onda correlation for the liquidphase mass transfer coefficient, k1a, is thought to be adequate, and since the liquid flow patterns in the countercurrent and crossflow columns are similar, it is reasonable, as a first approximation, to accept that k1a from the Onda correlation is correct. As a result, an experimental value of kgaH can be obtained from the following equation:
1 Hkga
) [ ) )
exp
1 K1a
-
exp
) ]
1 k1a
(5)
Onda
Of course, when the liquid-phase resistance dominates, K1a and k1a are approximately equal and the subtraction becomes meaningless. Equation 5 can be rearranged to obtain the parameter f, which is the fraction of the total mass transfer resistance contributed by the gas phase:
f)
K1a)exp Hkga)exp
)1-
K1a)exp k1a)Onda
(6)
Those tests which satisfied the previously stated requirements were re-examined to identify those in which the gas-phase resistance contributed at least half of the total resistance based upon eq 6. Fifteen tests (two involving 1,2-DCA, nine involving MEK, and four involving MeCl) satisfied all requirements. Of these, twelve involved eight crossflow passes, and the remaining three involved 24 crossflow passes. On the average, the experimental value of K1a for these fifteen tests was 54% smaller than predicted by the Onda correlation, and only one test produced an experimental K1a within (30% of prediction. Hkga)exp values for these 15 tests were calculated from eq 5, and the results were regressed to determine new constants, R and β, while maintaining the original form of the Onda gas-phase equation,
( )( )
Gm′ kg )R atDg atµg
β
µg FgDg
1/3
(atdp)-2
(7)
Regression gave values of R and β of 2.56 and 0.55, respectively, compared to Onda’s values of 5.23 and 0.7.
Using the new parameters to adjust the predicted value of Hkga produced major improvement in the MeCl and MEK results. The magnitude of the average deviation of the 16 MeCl data points was 6.5%, and all experimental points were within (20% of prediction. The 10 MEK tests, instead of clustering around the -60% parity line, produced an average deviation magnitude of 20%, and 8 of the 10 tests were within (30% of prediction. Unfortunately, the adjustment made to the Onda equation for kg overcorrected the 1,2-DCA results, changing the average difference for the 14 tests to +30%. The difference in the agreement between experimental and predicted mass transfer coefficients (using both the original and adjusted Onda parameters) for the three compounds may be the result of uncertainty in Henry’s constant. Harrison et al. (1993) compared published Henry’s constant data for these and other compounds and reported major differences. Errors in the values of H for a particular compound would propagate throughout the mass transfer coefficient analysis. In order to illustrate the importance of H, all data for 1,2-DCA was recalculated using H ) 0.070 as reported by Ehrenfeld et al. (1986). The larger value of H reduced the experimental values of K1a calculated from eq 2, increased the K1a values calculated from the Onda equation as modified by Wood et al. (1990), and increased the average deviation from about -20% to just over -30%. However, when the second modification to the Onda correlation associated with the new values for the parameters R and β in eq 7 were applied, the agreement between experiment and prediction improved markedly. The average magnitude of the difference for the 14 1,2-DCA tests was reduced to 12%, and for all 14 tests the difference was less than (30%. With this additional modification to Henry’s constant for 1,2-DCA, the agreement between experiment and prediction was quite good for all three compounds as shown by the parity plot of Figure 12. The magnitude of the average deviation for the 40 data points was 12%, and the deviation was less than (30% for 38 of the 40 cases. The two cases for which the deviation exceeded (30% were associated with the small values of K1a belonging to MEK.
Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3605
Figure 12. Parity plot comparing overall volumetric mass transfer coefficients from experimental data and from the Onda correlation with additional modifications.
Conclusions and Discussion The cascade crossflow contactor provides added flexibility in the design and operation of air strippers. With varied baffle spacing, the cross-sectional areas for liquid and gas flow may be controlled independently. Greatly decreased gas-phase pressure drop in the cascade crossflow configuration allows larger gas-to-liquid ratios with reduced power requirements. Stable operation is possible at gas-to-liquid ratios beyond the values which would cause a countercurrent column to flood. Stripping efficiencies exceeding those which can be achieved in countercurrent contactors are possible. These properties should extend the applicable range of air stripping to lower volatility compounds such as MEK which normally require steam stripping. Treatment of the larger volumes of dilute off-gas produced by stripping low-volatility compounds will be a problem if standard processes such as carbon adsorption or catalytic incineration are utilized. However, in situations where the off-gas can be discharged to the atmosphere or is used as combustion air to a boiler or incinerator, expensive off-gas treatment may be avoided. While the Onda correlation for the overall mass transfer coefficient is believed to be applicable for the air stripping of volatile compounds in countercurrent flow, the correlation is less applicable when less volatile compounds exhibiting appreciable gas-phase mass transfer resistance are to be stripped. Proposed modifications to the Onda gas-phase mass transfer coefficient equation greatly improve the level of agreement with experimental results, even for a compound such as MEK, where the gas phase contributes the majority of the total mass transfer resistance. The cascade crossflow concept may be applicable to other operations as well. One candidate is vacuum distillation, where the lower pressure drop could be important in reducing the reboiler temperature. In addition, the large variations in gas volumetric flow rate with pressure under vacuum conditions can be handled by varying the baffle spacing along the column height. This would avoid the current situation in which the diameter of the rectifying section of the column is sometimes larger than the diameter of the stripping section. Another potential application is in trickle-bed reactors where the solid (packing) serves as a catalyst as well
as promoting gas-liquid contact. These reactors are used in the array of hydrotreating operations in petroleum refining and in the liquid-phase hydrogeneration of benzene to cyclohexane. Cocurrent gas and liquid flow is typically used in order to avoid the flooding limitations of countercurrent flow. The cascade crossflow contactor would avoid flooding while maintaining the advantages of overall countercurrent flow. Nomenclature A ) column cross-sectional area filled with packing, m2 at ) dry surface area of packing per unit packed volume, m2/m3 CLB, CLT ) solute concentration in the liquid phase at the bottom and top of the packing, respectively, ppm DG ) gas phase diffusivity, m2/s dp ) nominal packing diameter, m E ) fractional stripping efficiency f ) fraction of the total mass transfer resistance contributed by the gas phase G ) volumetric gas flow rate, m3/s Gm ) gas loading rate based on the packed cross-sectional area of the column, kg/m2‚s Gmax ) maximum volumetric gas flow rate, m3/s Gm)max ) maximum gas loading rate, kg/m2‚s Gm′ ) gas loading rate based on the area between adjacent baffles available for gas flow, kg/m2‚s H ) Henry’s constant K1a ) overall liquid-phase volumetric mass transfer coefficient, s-1 kg ) gas-phase mass transfer coefficient, m/s kl ) liquid-phase mass transfer coefficient, m/s L ) volumetric liquid flow rate, m3/s Lm ) liquid loading rate based on the packed cross-sectional area of the column, kg/m2‚s N ) number of crossflow passes Z ) packed height, m Greek Symbols R ) adjusted coefficient in the Onda equation for kg β ) adjusted exponent in the Onda equation for kg µg ) gas viscosity, kg/m‚s Fg ) gas density, kg/m3
3606 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Literature Cited Ashworth, R. A.; Howe, G. B.; Mullins, M. E.; Rogers, T. N. AirWater Partitioning Coefficients of Organics in Dilute Aqueous Solution. J. Hazard. Mater. 1988, 18, 25-36. Ehrenfeld, J. R.; Ong, J. H.; Farino, W.; Spawn, P.; Jasinsky, M.; Murphy, B.; Dixon, D.; Rissman, E. Controlling Volatile Emissions at Hazardous Waste Sites; Noyes Publications: Park Ridge, NJ, 1986. Gavaskar, A. R.; Kim, B. C.; Rosansky, S. H.; Ong, S. K. Crossflow Air Stripping and Catalytic Oxidation of Chlorinated Hydrocarbons from Groundwater. Environ. Prog. 1995, 14, 3340. Gosset, J. M.; Cameron, C. E.; Eckstrom, B. P.; Goodman, G.; Lincoff, A. M. Mass Transfer Coefficients and Henry’s Constants for Packed Tower Air Stripping of Volatile Organics: Measurements and Correlations. Final Report, ESL-TR-85-18; Air Force Engineering and Service Center: Tyndall AFB, FL, 1985. Harrison, D. P.; Valsaraj, K. T.; Wetzel, D. M. Air Stripping of Organics From Groundwater. Waste Manage. 1993, 13, 417429. King, C. J. Separation Processes, 2nd ed.; McGraw-Hill: New York, 1980. Mertooetomo, E.; Valsaraj, K. T.; Wetzel, D. M.; Harrison, D. P. Cascade Crossflow Stripping of Moderately Volatile Compounds Using High Air-To-Water Ratios. Water Res. 1993, 27, 11391144.
Onda, K.; Tadeuchi, H.; Okumoto, K. Mass Transfer Coefficients Between Gas and Liquid Phases in Packed Columns. J. Chem. Eng. Jpn. 1968, 1, 56-62. Roberts, P. V.; Hopkins, G. D.; Munz, C.; Riojas, A. H. Evaluating Two-Resistance Models for Air Stripping of Volatile Organic Contaminants in a Countercurrent Packed Column. Environ. Sci. Technol. 1985, 19, 164-173. Staudinger, J.; Knocke, W. R.; Randall, C. W. Evaluating the Onda Mass Transfer Correlation for the Design of Packed-Column Stripping. J. Am. Water Works Assoc. 1990, 82, 73-79. Verma, S.; Valsaraj, K. T.; Wetzel, D. M.; Harrison, D. P. Direct Comparison of Countercurrent and Cascade Crossflow Air Stripping Under Field Conditions. Water Res. 1994, 28, 22532261. Wood, D. F.; Locicero, L.; Valsaraj, K. T.; Harrison, D. P.; Thibodeaux, L. J. Air Stripping of Volatile Hydrophobic Compounds Using Packed Crisscross Flow Cascades. Environ. Prog. 1990, 9, 24-29.
Received for review March 25, 1996 Revised manuscript received June 14, 1996 Accepted June 17, 1996X IE960178N X Abstract published in Advance ACS Abstracts, September 1, 1996.
