162 Ind. Eng. Chem. Process Des. Dev.

1982, 21, 162-170

Generalized Correlation for Mass Transfer in Packed Distillation

Columns

Jose L. Bravo and James R. Fair*

Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712

An Improved correlation of mass transfer rates for commercial-scale packed distillation columns has been de-
veloped. The correlation is based on effective area for mass transfer

a 0.498a (CaufieG)0392
Downloaded from pubs.acs.org by UNIV NACIONAL AUTONOMA MEXICO on 03/11/19. For personal use only.

where ae is effective area for transfer, a p is specific surface of the packing material, Z is height of packed bed
(ft), is surface tension (dyn/cm), and CaL and fieG are dimensionless liquid capillary number and gas Reynolds
number, respectively. Packings represented are of the randomly arranged ring and saddle types. Basic data are
from a bank of 231 evaluated runs covering eleven distillation systems. The correlation Is shown to be more reliable
and more general than previously published models, with 96% of the calculated data falling within ±20% of the
observed values. For 95% confidence in design, a 1.6 safety factor is required.

Packed columns have been used extensively for distil- from simple, naturally occurring materials, such as chunks
lation applications, especially those in which pressure drop of coke, to carefully fabricated geometric shapes such as
Ind. Eng. Chem. Proc. Des. Dev. 1982.21:162-170.

is critical. The columns have usually been filled with
randomly arranged packing elements of the ring or saddle
type, and it is on these “random packings" that most ex-
perimental investigations of mass transfer rates have been
made. While there is increasing interest in the so-called
“regular packings” (mesh, grids, etc.), their applications
tend to be limited by high cost and incomplete perform-
ance documentation. These latter packings are normally
handled on a proprietary basis through the experience of
their vendors.
In 1979 Bolles and Fair described a large data base that
they assembled from published tests of flooding, pressure
drop, and mass transfer in columns containing random
packings. They utilized the base to evaluate existing
packed column models. They did not attempt to improve
the existing models for flooding or pressure drop, but did
develop a mass transfer model that was superior to those
previously published. Their model generalized on fluid
properties and flow rates, but required two inputs: prior
validation for a specific packing type and size, and a
consistent (if not completely reliable) method for pre-
dicting the flooding condition for the system and packing
under scrutiny.
The present work represents an attempt to develop a
general design model for distillation applications, a model
that would not require previous validation for the packing
type or size and which hopefully could be extended to the
regular packings. Furthermore, the Bolles-Fair require-
ment of flooding knowledge was thought to be an encum-
brance as well as a limitation on reliability, and it was
hoped that such requirement would not be needed with
a new model. The model, if successful, could then be
applied to a variety of packing sizes and shapes while still
being general with respect to properties and flow condi-
tions; in view of the continuing development of packing
types, such a model should be well received by process
engineers and packing vendors alike.
Previous Work
A vast amount of literature dealing with mass transfer
in randomly packed columns is available. Studies have
covered a wide variety of packing types and sizes ranging
metallic or ceramic saddles, and from tiny devices for
laboratory use to large commercial packing elements with
nominal sizes of 3 in. and larger.
An extensive review of earlier data pertinent to com-
mercial operations was presented by Cornell et al.
(1960a,b). These data are mainly from studies that used
either Berl saddles or Raschig rings. Also, the data were
obtained largely from tests with unidirectional mass
transfer, gas-to-liquid or liquid-to-gas. A more recent paper
by Charpentier (1976) presents a good general overview
of additional data and of some of the approaches that
investigators have used in dealing with the prediction and
description of mass transfer in packed columns.
As noted earlier, the 1979 paper by Bolles and Fair
included an extensive data bank for a wide range of
packings, column sizes, and systems. Many of the data
in this bank are for distillation conditions and include
results of work done by Fractionation Research, Incorpo-
rated (FRI). This data bank provided all of the model
validation material for the present work. As is appropriate
for differential contacting, the Bolles and Fair model
utilizes the transfer unit approach, equivalent to the mass
transfer coefficient approach
0 =
(1)
K0GaJ>MG
where the effective interfacial area oe may differ from the
actual interfacial area.
The volumetric mass transfer coefficient KGGae in eq 1
may be divided into coefficient and area components for
fundamental analysis, and this approach has been used by
a number of workers (Shulman et al., 1955a-c; Yoshida and
Koyanagi, 1954; Onda et al., 1968,1972; Hughmark, 1980).
Most of the work was confined to Raschig ring and Berl
saddle packings.
The work by Shulman can be considered as the basis
for the present effort, as it has been for most other mass
transfer efficiency studies that describe the transfer
coefficient and the effective area.
The work by Onda and co-workers is based on the
concept of wetted packing area, as opposed to interfacial

0196-4305/82/1121-0162$01.25/0 © 1981 American Chemical Society

 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982 163

area. The work draws on the earlier efforts by Yoshida Both correlations for the individual mass transfer
and co-workers and uses the wetted area as a geometric coefficients show the same type of dependence on key
parameter in determining the Reynolds number for the variables such as diffusivity and velocity.
liquid. The equations describing kL and kG in the Shulman
The Onda and Shulman models, which form a part of model are
the basis for the present model, will be described fully in
the next section.
1.195 (ScG),-2/3
=
kGpGMRT/G (8)
Model Development
The two-film mass transfer model as well as the concept
of the transfer unit provide the basis for this work. Al-
ternate mass transfer models based on the penetration and Mp/Dl-25.1
[¥]" (ScL)10.50 (9)
-

surface renewal theories may provide a more realistic

physical description of the mass transfer mechanisms, but In eq 8 and 9 dp is the diameter of a sphere having the
at the present stage of development they have not changed same surface area as an element of packing under con-
significantly the mass transfer relationships based on the sideration.
film model. On the other hand, the model developed by Onda is
An exception to the above may be found in the value described as follows
for the exponent affecting the diffusion coefficient in the
general flux equation. The two-film theory predicts an
exponent of unity whereas the penetration theory predicts (Sc,)V«(opd'p)-2 (10)
a value of one-half. The applicable mass transfer coeffi- 5.23^
cient correlations, especially the two used in the present
work (by Shulman and by Onda), show a dependence on ( L \2/3 f
=
0.00511 I (ScL)-1/2(a pd P')0.4
-

the diffusivity intermediate between film and penetration

\OwMl/
predictions: a two-thirds exponent for the gas side and (ID
a one-half exponent for the liquid side.
Using the two-film model, and allowing for significant
resistance to mass transfer in both phases, the overall -1.45Re 0·1^^·05
transfer unit height results from individual phase transfer WeL0-2| j J J j

unit heights (12)

Hoc Hg + XHl
=
(2) where
where =
mGMLM, or the ratio of slopes of the equilibrium
=
and operating lines. Re'h
—

(13)
apPh
The integration of the general flux equation, coupled
with the definition of the number of transfer units, leads aX2
to defining expressions for the heights of individual ^ 2
(14)
transfer units SPl
Hg =
G/(kGaePMG) (3) L2
WeL (15)
Hl = L / (kLaepi) (4) OpCPL

where oe is the effective interfacial mass transfer area.

At this point it is important to note that the value for
ae in a given system is considered equal for the gas and
liquid phases, since it consists of the area through which
mass transfer occurs at the interface. It is also to be ob-
served that the value of ae is composed not only by the
wetted area over the packing but also by the area provided
by suspended and falling droplets, gas bubbles within
liquid puddles, ripples on the liquid film surface, and any
contribution from film falling on the walls of the column.
A combination of eq 2, 3, and 4 gives
ae
=
(ajío + XaJi^/HoG
and 0, a critical surface tension, = 61 dyn/cm for ceramic
packings, 75 dyn/cm for steel packings, and 33 dyn/cm
for polyethylene packings.
The data bank provided input variables to enable cal-
culation of effective areas by the Shulman and Onda
models. These areas were compared with the known
specific areas of the packings used.
The Onda correlation gave more moderate ae values than
the Shulman correlation (ae values never greatly above the
value for the specific area, ap, of the packing under con-
sideration) and also covered a wider range of packing types
and sizes as well as test systems. Therefore, it was selected
(5) as a source of kh and kG values. It was recognized that

Thus, an effective area can be calculated from available
mass transfer coefficients kG and kh, themselves based on
known values of G, L, MG, ML, P, and pL.
Clearly, the selection of a specific mass transfer coef-
ficient correlation is critical to the present work. The
correlations by Onda and co-workers (1968,1972) and by
Shulman and co-workers (1955a-c) were selected for our
analysis because they conform to the commonly accepted
functionalities
koPMa/G =
QReGuScGv
and
klC^M^/L —
future work leading to improved methods for predicting
kL and kG could lead to improvements in the present work.
Because of the nature of the hydraulics and geometry
of packed beds with gas and liquid in counterflow, a cor-
respondence between liquid holdup and effective transfer
area is to be expected; thus, the trends observed in liquid
holdup behavior at varying loading conditions should apply
for effective area. Also, an added gas kinetic energy effect
should be included for the area analysis because of its
influence on film surface rippling, liquid droplet dispersion,
(6) and the occurrence of gas bubbling in puddles. It can be
reasoned that the behavior of liquid holdup for different
flow conditions should be similar to that of the interfacial

Q'ReLrScL‘ (7) area, the latter possibly showing a greater influence from
164 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982

gas velocity and density. Accordingly, a significant gas Table I. Symbols for Figures 1-17
effect was included in the present work since it is generally
A Cyclohexane/n-Heptane 5.0 psia
practical to operate commercial packed columns at high
gas loadings and relatively close to flooding conditions.
In analyzing the various forces acting in such systems, X Cyclohexane/n-Heptane 24.0 psia
it was concluded that an improved correlation should be
able to deal with the following considerations, (a) The + n-Butane/1-Butane 165.0 psia
effective transfer area should be proportional to the liquid
rate since increased liquid velocities would provide a better
o l-Propanol/water 14.7 psia
wetting of the packing, rippling of the film surface, and
a more effective renewal of the liquid in the puddles, (b)
Systems with low surface tensions should provide increased
X Ethylbenzene/Styrene 2.0 psia
wettability as well as smaller droplets yielding in this sense
more interfacial area. On the other hand, a lower surface X Methanol/Ethanol 14.7 psia
tension would make the liquid separate more easily from
the packing at high gas rates, thus reducing the effec- Q Ethanol/water 14.7 psia
tiveness of the packing, (c) Increased kinetic energy of the
gas should tend to increase the effective area, (d) A
CD Ethylbenzene/Styrene 1.0 psia
packing with a high specific surface should yield a rela-
tively high value for the net interfacial area under similar
flow conditions, (e) A correlation for liquid maldistribution X n-Heptane/Toluene 14.7 psia
due to column height and diameter should be expected.
With these considerations in mind, and after different Y 14.7 psia
schemes were tried unsuccessfully, the following func- NH3/H20/air
tionality was proposed X 14.7 psia
02/H20/air
ß/ =
f{L,G,ZJ)c,a) (16)
* Benzene/Dichloroethane 14.7 psia
The values for aep (the effective mass transfer area for the
packing alone) were obtained by correcting the total ef-
fective area (oe) for column wall effects if i-Octane/Toluene 20.0 psia

aep
=
o8
-

«4/Dc (17) Various combinations of these dimensionless groups

The factor a allows for wall blockage by packing elements
and should not exceed unity. For the present work a value
of a = 1.0 was assumed, since for large columns the wall
correction essentially disappears.
Modification of the findings of Cornell et al. (1960a) and
of Fair and Moczek (1964) led to a corrected effective
packing area
* = ß °·4/ 0·5 (18)
where o* is an adjusted effective area suitable for analysis
or design of commercial equipment. Equation 20 proved
to be a good choice since all of the available data grouped
very nicely when the adjustment, combined with the
previously mentioned diameter correction, was applied. It
should be noted that eq 18 is dimensional, although it
could be made dimensionless by using height and surface
tension ratio to base values (e.g., Z = 10 ft and „ = surface
tension of water = 72 dyn/cm). As used in this work, Z
has units of feet and has units of dyn/cm.
In order to incorporate the liquid and gas rate effects
as well as the contributions from the viscous forces within
the liquid, the following dependence was then proposed
(19)
o*/op =
f(Cah, ReG)
In eq 19, CaL is the dimensionless capillary number
CaL = pLL/(phcgc) (20)
which provides the ratio of viscous forces to surface tension
forces. The dimensionless Reynolds number ReG is defined
as
ReG =
(21)
6G/(apMG)
which in turn provides the ratio of inertial forces to viscous
forces in the gas. Note that the length criterion is the
diameter of a sphere having the same surface to volume
ratio as the packing element under consideration.
were tried in searching for the best grouping of the data.
Most combinations showed a very significant packing
size/type dependence, and in most instances, the trends
followed by the correlating coefficients were not distin-
guishable among the different packings and were highly
random. For these cases, it was impossible to establish a
relationship between the variation in the value of the
correlating coefficients with the differences in packing size
and type.
To circumvent this problem, a function of the type
X ReG)b
o*/op =
A(Col (22)
was used, and its application showed all the data for either
distillation or absorption (each operation independently)
in what appeared to be one single group, regardless of
packing size or type. This finding led us to the proposal
of one single correlation for all of the packings studied.
The form of the equation selected meets all of the re-
quirements outlined above with the added advantage of
generality toward packing size/type. Nevertheless, this
form has some limitations, particularly in distillation using
low reflux ratios. These limitations are discussed in the
Conclusions section and some recommendations are also
made.
Data Presentation and Correlation
The functional relationship of eq 22 is shown in Figure
1 through 13 (see Table I for symbols) for the packings
listed in Table and the systems listed in Table III. The
values of a*/ap were obtained from the Bolles-Fair data
bank by using physical properties at the experimental
conditions, the geometric parameters of the experimental
column and packing, mass transfer coefficients computed
by the Onda model, and the reported equilibrium and
HETP or Hqq data. The systems analyzed, as well as their
important physical properties, are shown in Table III.
An inspection of Figures 1-13 shows clearly that the
functionality of eq 22 holds regardless of the system and
 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982 165

Figure 4. Area ratio for 1-in. ceramic Raschig rings.

1/2 '
'CERAMIC fi. fi. 1.5' 'CERAMIC R. R.

Q- -
cr :
*
c
ü

-«
" r
r,rm,| 111111-rr'r m»|-
10° 101 102 IQ3 104
.

10= 10°......l'o1 102 103 104

CAL*REG*1E 03 CALmfiEG*1E 03
Figure 2. Area ratio for 0.5-in. ceramic Raschig rings. Figure 5. Area ratio for 1.5-in. ceramic Raschig rings.

'CERAMIC R.R.
'
3

xz,x

X /
- /
:
V
/
X
/
/
/
/
z>
z ^Master correlation

_
riii
mi|-1—I II I
III]- -r I > r rrrn
-ioD 101 102 103 1q4 1 q5
CAL*REG*1E 03
Figure 6. Area ratio for 3-in. ceramic Raschig rings.
166 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982

METAL ñ.B.
’ 1
1

-
:

Master correlation /

X/
%/
X"·
X, ,X'
tt-
10° 101 io2 io3 104
CAL*ñEG*1E 03
Figure 7. Area ratio for 1-in. metal Raschig rings. Figure 10. Area ratio for 6/g-in. metal Pall rings.

1. 5
'
'METAL ñ. R. 1
'
'METAL P.R.

/
Master correlation Master correlation /
/
/

X*0
o
X/Z oj8®
X/"
x,x
ib°
M
"'io1
TTTTj-1—r
L2
lb2
i r
rrq-1—
ir>3
io:
rm|-1—r-
lb4 IO5 10u "iV 102 IQ3 ""io4 "To3
CAL*REG*1E 03 CAL*REG*1E 03
Figure 8. Area ratio for 1.5-in. metal Raschig rings. Figure 11. Area ratio for 1-in. metal Pall rings.

1.5' 'METAL P. R.
2' ' E T AL R.R.

/
/
*
c
X / */
*/
X / /
X7 X

Master correlation

/
Jq x
-
........
r"l ........I TTTT ""
10*, io2
-

io3 io4 1QJ

=
io° lb1 ib2 io3 io4 105
CAL*REG*1E 03 CAL*REG*1E 03
Figure 9. Area ratio for 2-in. metal Raschig rings. Figure 12. Area ratio for 1.5-in. metal Pall rings.
 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982 167

o
MASTER CORRELATION FOR DISTILLATION

Figure 13. Area ratio for 2-in. metal Pall rings.

Figure 14. Master correlation for distillation.
Table II. Packings Included in Data Bank and or liquid rate or both. Thus, the product CaifieG can be
Used in the Present Work
viewed as a loading parameter, but no effort here has been
sp made to generalize it in terms of experimental observations
nominal surface, of loading and/or flooding.
size, void
packing type in. material fraction ft^ft1 Figures 1-13 do not include flooding conditions, but they
do include experiments under high loadings. The increase
Berl saddles 1 ceramic 0.68 76 in the slope of the curve, at high values of CaL X ReG and
Raschig rings ‘ ceramic 0.64 112
Raschig rings 3/4 ceramic 0.72 74 a*/a- is in accordance with the well known behavior of
Raschig rings 1 ceramic 0.74 58 HETP at higher loadings.
Raschig rings l‘/2 ceramic 0.73 37 Figure 14 shows all of the distillation data in one single
Raschig rings 2 ceramic 0.74 28 plot combining all of the test systems and packings used
Raschig rings 3 ceramic 0.75 19 in the correlation. The relationship that best describes
Raschig rings 1 steel 0.86 56 these data is
Raschig rings l‘/2 steel 0.90 39
Raschig rings 2 steel 0.92 29 a*/ap =
0.498(Col X fieG)°·392 (23)
Pall rings0 =
/8 steel 0.93 104
Pall rings 1 steel 0.94 63 The fit of this equation to the data is discussed in the
Pall rings 2 steel 0.96 31 following section. Equation 23 is presented in Figures 1
°
This packing type is marketed under the names of Pall to 14 as a dashed line for comparison. No data are shown
rings, Flexirings, and Ballast rings. for distillation in 2-in. ceramic Raschig rings because the
data bank does not include distillation runs for this
experimental conditions under consideration. The data packing. After analyzing the fit for the other packings,
also line up very nicely when values for different packings especially for 1 and 1.5-in. ceramic Raschig rings it was
are plotted together. Thus, for a given value of CoL X Rec concluded that the correlation should include 2-in. ceramic
the corresponding value for o*/op appears to be the same Raschig rings. The same reasoning was applied to Berl
for different systems and/or packing type or size. saddles in the range of 1 to 2 in.
It is well known that a packed column loading/flooding The proposed correlation fits the individual data for
condition can be approached by increasing either gas rate each packing very well except for the case of 3-in. ceramic

Table III. Test Systems Considered in the Present Work

surface
pressure, gas density, liquid density, tension, rel
system psia lb/ft3 lb/ft3 dyn/cm volatility
cyclohexane/n-heptane 5.0 0.06 44 20 1.9
cyclohexané/n-heptane 24.0 0.30 42 15 1.6
n-butane/isobutane 165.0 1.76 31 6 1.3
2-propanol/water 14.7 0.11 47 17 1.6
ethylbenzene/styrene 1.0 0.016 52 27 1.4
ethylbenzene/styrene 2.0 0.03 52 26 1.3
methanol/ethanol 14.7 0.08 47 19 1.7
ethanol/water 14.7 0.09 48 22 1.4
n-heptane/toluene 14.7 0.20 41 15 1.3
benzene/dichloroethane 14.7 0.20 63 22 1.1
isoo ctane / toluene 20.0 0.28 41 14 1.2
NH3/H20/air 14.7 0.08 62 74 a
02/H20/air 14.7 0.07 62.3 73 a

a
These are absorption systems; all others are distillations.
168 Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982

MASTER PLOT FOR ABSORPTION PARITY PLOT DISTILLATION

Raschig rings with cyclohexane/ -heptane at 24 psia as

the test system. This same system also falls somewhat out
of line for 3/4-in. ceramic Raschig rings as shown in Figure PARITY PLOT DISTILLATION
3. Figure 5 indicates that at low loadings the iso-
octane/toluene system does not correlate very well. These
deviations could not be explained when analyzing the data.
Nevertheless, we do not think they seriously affect the fit
of the overall correlation due to the large number of points
(the deviated ones included) taken into consideration.
The absorption data as shown in Figure 15 also group
adequately when plotting a*/op vs. CoL X ReG. These data
show lower values of effective area when compared to
distillation data under similar conditions. This marked
difference could not be readily explained, but may be due
to the existence of vapor condensation simultaneous with
liquid bubbling under distillation conditions. These two
phenomena occurring at the same time should provide
added mass transfer surface within the bulk of each phase.
This added surface could account for part of the difference
between absorption and distillation effective areas.
On the other hand, factors influencing the mass transfer
coefficients and the use of a characteristic value for the
slope of the equilibrium line in distillation could also in-
tervene in explaining this difference in the deduced values
for ae.
Yoshida and Koyanagi (1962) and Onda et al. (1968) also
observed the difference between distillation and absorption Figure 17. Parity plot for distillation (expanded lower region).
effective mass transfer areas. Yoshida and Koyanagi offer-
a qualitative explanation of this difference through surface Reliability and Comparison with Other
tension effects since the absorption data they used had Correlations
water as the irrigating liquid, and the distillation data they In this section reference is made to observed and cal-
analyzed were for organic liquids. In the present work, culated values for the ratio a*/ap. The observed values
even though the ethanol/water system is included in the are obtained from data bank values plus the Onda corre-
correlation, the surface tensions for the liquids in the lation for mass tranfer coefficients. The calculated values
distillation data are lower than the surface tension in the are obtained by the use of the equation
absorption systems. Thus, surface tension of the irrigating (o*/Op)caicd
=
0.498(CoL X fieG)0'392 (23)
liquid could also account for the difference between the
effective interfacial areas in distillation and absorption. The statistical analysis of the correlation can be de-
Figures 16 and 17 show a parity plot for the distillation scribed as follows, (a) The total number of data points
data using the deduced values for o*/op and the calculated used in the correlation is 231, with 11 different systems
ones from eq 23 with Figure 17 representing an expanded included, (b) The average absolute deviation was found
section of Figure 16. The degree of scatter shown in these to be 22%
plots does not seem to vary with the value of o*/ap; thus av abs dev =
£abs [i(o*/ap)calcd/(a*/ap)obsd]100/n
-

the degree of loading or approach to flood does not appear

to influence the fit of eq 23. (c) The average ratio (o*/ap)calcd/(o*/op)obed was equal to

Table IV. Model Reliability Analysis and Comparison
Bolles and Fair this work
0.042 -3.42 X 10'14
OFm
S 0.323 0.286
Fs (95% confidence) 1.70 1.60
1.042. (d) For the 231 points, 96% of the calculated values
of o* were within ±20% of the observed values, 71%
within ±10%, and 40% within ±5%.
Bolles and Fair (1979) used and recommended a relia-
bility analysis procedure based on a logaritmic ratio ob-
jective function, which in the case of this work is
OF¡ = In ((o*/op)obsd/(a*/ap)calcd) (24)
The lower bound of this objective function was selected
to be the critical one since the conservative approach is
to assume that the actual design value for the effective area
will be smaller than the one obtained from the model. (For
an Hqq objective function like the one proposed by Bolles
and Fair, the upper bound will be the critical one since the
height of a transfer unit is inversely proportional to the
effective area.)
Following the procedure presented by Bolles and Fair,
the following expression for lower bound critical is obtained
(o*/op)deeign
=
e0Fme~tS(a* / ap)cM (25)
where OFm = mean of the objective function = OF¡/n,
= number of
observations, OF¡ = objective function for
the ith observation, S = standard deviation of the objective
function = [ (0- ;- OFm)2/(n l)]0·5, and t = Student’s
-
“£" for a certain degree of confidence.
The real measure of the reliability of the model is the
standard deviation or scatter of the objective function.
Thus the “safety factor”, Fs = etS, should provide the best
reliability measure for the model for a certain percent
confidence.
The analysis described above was applied to the present
model for a 95% confidence (f = 1.645) and the results as
well as a comparison with the Bolles and Fair model are
presented in Table IV.
The present model appears to be slightly more reliable
than the one by Bolles and Fair. These authors show in
their paper an improvement over the models of Onda and
Cornell et al. Based on this comparison, the present model
appears to be the most reliable of the four. It should be
noted, however, that the other three models are not spe-
cifically for distillation as the present one is, so a greater
scatter is present in those models that encompass different
mass transfer operations in packed columns.
Extension to Other Packings
The model described here is based on randomly ar-
ranged packing elements that can be characterized as
having a nominal diameter d'p. This diameter is used in
the calculation of phase transfer coefficients kg and kL.
One would expect the model to apply to packings other
than those tested so long as the packings have similar
shapes and identifiable nominal diameters. For example,
ceramic Intalox saddles were not included in the data bank
and thus did not contribute to the general development
of the model; however, they are similar to Berl saddles with
about 12% higher specific surface op. The model would
show that ceramic Intalox saddles would have a somewhat
higher mass transfer efficiency than ceramic Berl saddles,
for the same phase rates and system properties.
Extension of the model to the nonrandom packing ma-
terials would call for assumptions of equivalent diameter
that could only be validated by considerable test data. At
Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 1, 1982 169
the present time, such an extension is not recommended.
Conclusions and Recommendations
The independent treatment of the individual mass
transfer coefficients and the effective interfacial area for
mass transfer appears to provide a logical and fundamental
approach to the description of mass transfer in counterflow
gas-liquid packed column. Transfer rates within a single
phase should depend on the available concentration gra-
dient, the degree of turbulence in that phase, and on the
physical properties of the fluid. The bulk-to-interface
transfer rates (or the individual mass transfer coefficients)
should be much more heavily dependent on the system
itself than on the geometric characteristics of the packing.
On the other hand, the effective interfacial area for mass
transfer at any given point in the bed should depend only
on the relative kinetic energy of the phases, the geometry
of the packing, the wettability of the packing surface, and
the tendency toward rippling and/or droplet formation in
the liquid. The extrapolation of this “point” area to a
complete column should be some function of the overall
dimensions of the bed.
The product of the capillary number for the liquid
(which accounts for the liquid viscous and surface tension
forces) and the Reynolds number for the gas (which de-
scribes the inertial and viscous forces in the gas) seems to
provide an appropriate parameter for describing effective
areas.
Through the course of this work it was observed that
for similar flow conditions the available area for mass
transfer was substantially higher in distillation systems
than in absorption systems. The increased area in dis-
tillation could be explained by the existence of simulta-
neous and significant boiling in the liquid phase with
condensation in the vapor phase. These two effects are
thought to provide a continuous formation of interfacial
area within each phase, thus providing an increased net
effective area.
Most of the distillation data used in the correlation were
reported under total reflux conditions. Thus, a separation
of the liquid and gas velocity effects could not be studied
from the available data. The work of Clump (1953), while
not conclusive, indicated little effect of liquid/vapor ratio
on measured Hqq values. The work of Eckert and Walter
(1964) showed that for the rectifying section a decrease in
liquid/vapor ratio (operation at finite reflux) the measured
HETP decreased. An analysis of the effects of varying
liquid and vapor rates, according to the models presented
here, indicates compensating effects in the rectifying and
stripping sections of a distillation column such that the
predicted total reflux Hqq should be satisfactory for finite
reflux operation, being somewhat conservative.
When the correlation is applied to new designs, a
question of limits on packed height can arise. The data
bank included studies of mass transfer in heights up to 33
ft. However, prudent design often limits the height of an
individual section of packing to about 20 ft. Thus, the
height correction appears to be valid within the range of,
practical application.
In conclusion, an improved mass transfer model has
been developed for application to distillation in columns
containing random packings. The model is more funda-
mental than those previously published and on the basis
of testing against a large data bank is more reliable than
those currently available. More work will be needed before
the model can be extended to nondistillation systems and
to nonrandom packings.
Nomenclature
A = coefficient in eq 22
170 Ind. Eng. Chem. Process Des. Dev. 1982, 21, 170-173

a* =
corrected effective area, (ft2/ft3)(ft0·4/(dyn/cm)0·5) ScG
=
Schmidt number for the gas
ae
=effective interfacial area, ft2/ft3 Sch = Schmidt number for the liquid
oep
= effective area provided by packing alone, ft2/ft3 S = standard deviation
=
specific surface of the packing, ft2/ft3 T = absolute temperature
aw
=
wetted area, ft2/ft3 t = Student’s t
b = exponent in eq 22 u, v exponents in eq 6
—

CaL = capillary number for the liquid WeL = Weber number for the liquid
Dc = column diameter, ft Z =
height of packing in column, ft
Dg = diffusion coefficient for the gas Greek Symbols
Dl = = diffusion coefficient for the liquid
dp diameter of a sphere with the same surface area as one a = factor to correct for available column wall surface
packing element e = void fraction
dp = nominal diameter of the packing element
= X = ratio of slopes (equil. line/operating line =
mGm/Lm)
safety factor for model reliability viscosity
=
Fs µ
FrL = Froude number for the liquid density
=
p
G = gas superficial mass velocity = surface tension, dyn/cm
Gm
=
gas superficial molar mass velocity ac
=
critical surface tension, dyn/cm
g -
acceleration of gravity
Subscripts
gc
=
force/mass conversion factor G = referred to the gas
Hg = height of a gas transfer unit, ft
Hh = height of a liquid transfer unit, ft
L = referred to the liquid
GM = molar property for the gas
Hqq = height of an overall gas transfer unit, ft
HETP = height equivalent to a theoretical plate, ft Literature Cited
kG
=
mass transfer coefficient for the gas phase
Bolles, W. L; Fair, J. R. Inst. Chem. Eng. Symp. Ser. 1979, 56 3/35.
kL = mass transfer coefficient for the liquid phase
,

Charpentler, J. C. Chem. Eng. J. 1978, 11, 161.

KGg = overall gas mass transfer coefficient, based on gas Clump, C. W. Ph.D. Dissertation, Carnegie Institute of Technology, Pittsburgh,
concentration Pa., 1953.
L = liquid superficial mass velocity Cornell, D.; Knapp, W. G., Fair, J. R. Chem. Eng. Prog. 1960a, 56(8) 68.
Cornell, D.; Knapp, W. G.; Close, H. J.; Fair, J. R. Chem. Eng. Prog. 1960b,
Lm = liquid superficial molar mass velocity 56(8) 48.
m = slope of the equilibrium line Eckert, J. S.; Walter, L. F. Hydrocarbon Process. 1964, 43, 107.
= molecular weight Fair, J. R., Moczek, J. S„ Presented at las Vegas AIChE Meeting, Aug 1964.
= number of observations Hughmark, G. A. Ind. Eng. Chem. Fundam. 1980, 19, 385.
Onda, K.; Takeuchl, H.; Okumoto, Y. J. Chem. Eng. Jpn. 1968, 1, 56.
OF, = objective function for observation i Onda, K. Mem. Fac. Eng., Nagoya Unlv. 1972, 24(2) 164.
OFm = mean of the objective function Shulman, H. L; Ullrich, C. F.; Wells, N. AIChE J. 1955a, 1, 247.
P = total pressure Shulman, H. L; Ullrich, C. F.; Proulx, A. Z.; Zimmerman, J. O. AIChE J.
1955b, 1, 253.
Q,Q'= coefficients in eq 6 and 7 Shulman, H. L; Ullrich, C. F.; Wells, M; Proulx, A. Z. AIChE J. 1955c, 1,
Reg = Reynolds number for the gas (= 6G/(cy¿G) 259.
ReL = Reynolds number for the liquid (= 6L/(a,~pi) Yoshlda, F.; Koyanagl, T. Ind. Eng. Chem. 1954, 46, 1756.
ReG = modified gas Reynolds number (= ReG/6) Yoshlda, F.; Koyanagl, T. AIChE J. 1962, 6, 309.
Rel' = modified liquid Reynolds number (= fieL/6)
R = universal gas constant Received for review May 22, 1981
r, s = exponents in eq 7 Accepted September 10, 1981

Upgrading of Recycle Solvent Used in the Direct Liquefaction of

Wyodak Coal
Ronald L. Miller and Howard F, Silver*
Chemical Engineering Department, University of Wyoming, Laramie, Wyoming 82071

Robert J. Hurtublse
Chemistry Department, University of Wyoming, Laramie, Wyoming 82071

Hydrogen donor characteristics are one of the most important properties of a coal-derived recycle solvent used
In the direct liquefaction of coal. This property can be modified by direct catalytic hydrogenation to increase the
solvent hydrogen content. Results of this work using 22 solvents derived from three coals Indicate that recycle
solvent effectiveness goes through a maximum as the hydrogen content of the solvent Increases. However, the
solvent hydrogen content at which the observed maximum occurs differs for different solvents. Weight percent
hydrogen in a solvent at observed maximum liquefaction effectiveness has been correlated as a linear function
of a Watson-type characterization factor, /fw.

Introduction coal dissolver reaction conditions but also on the quality

The extent to which coal can be directly converted to of the recycle solvent used. A good recycle solvent must
gases and benzene soluble liquids depends not only on the be physicallv compatible with the coal fragments formed

0196-4305/82/1121-0170$01.25/0 © 1981 Americar· Chemical Society