Académique Documents
Professionnel Documents
Culture Documents
n - 1 = tray n - 1 King, C. J. Separation Process, 2nd ed.; McGraw-Hill, Inc; New York,
i = component i 1981; p 472
Lewis, W. K. Ind. Eng. Chem. 1936, 28. 399.
D = distillate Mohan, T. M.Tech. Thesis, I.I.T., Kanpur, India, 1981.
Literature Cited
BubbleTray Design Manual; American Institute of Chemical Engineers,
Received for review May 27, 1981
New Ycfk, 1958. Revised manuscript received October 21,1982
Colbwn, A. P. Ind. Eng. Chem. 1936, 28, 526. Accepted December 16,1982
The Soave modification of the Redlich-Kwong equation has been very effective for conelating the phase equilibrium
of systems containing nonpolar and slightly polar substances. In this work further modiflcations are introduced
which retain the simplicity and robustness of the Soave equation but extend its application to systems containing
highly polar substances such as water. Appllcations shown in the work include systems of interest in coal processing
and some common systems containing water.
Table I. Vapor Pressures of Highly Polar Substances; Comparison between Original and Modified RKS
% dev,
% dev, p = 0 p fitted
temp range,
substance T,, K P c , atm w P K max ava max ava reference
water 647.3 218.3 0.3439 0.1277 273-647 -33.6 10.2 0.9 0.3 Meyer et al. (1967)
acetone 508.1 46.38 0.307 0.0715 259-508 -9.1 2.5 0.9 0.4 Ambroseetal. (1974)
methanol 512.64 79.91 0.565 0.2359 288-513 -19.4 4.6 0.9 0.4 Ambrose and Sprake (1970)
Ambrose et al. (1975)
ethanol 513.92 60.68 0.646 0.1006 293-514 -9.4 1.5 1.7 0.7 Ambrose and Sprake (1970)
Ambrose et ai. (1975)
l-pentanol 588.15 38.58 0.585 -0.2615 348-512 17.9 5.3 1.7 0.7 Ambrose and Sprake (1970)
Ambrose et al. (1975)
l-octanol 652.50 28.23 0.598 -0.2109 386-554 17.9 4.0 4.5 2.2 Ambrose and Sprake (1970)
Ambrose et al. (1975)
Average absolute percentage deviation = Z i Ipiemfl- piCad I/(no. of data points).
prove VLE calculations of systems containing highly su- substances. He correlated the characteristic constant m
percritical substances. We show that this modification as a function of the acentric factor. The correlation used
successfully correlates the VLE of systems containing in this work was obtained by Graboski and Daubert (1978).
substances such as hydrogen, methane, and carbon dioxide It is very similar to that of Soave but is slightly better since
at temperatures much above their respective critical tem- it was regressed from a larger data set. The correlation
peratures. Incidentally, these systems are of interest in for m is given by
coal conversion processes. mi = 0.48508 + 1 . 5 5 1 9 1 ~
- ~0.15613wt (6)
A comment about the importance of fundamental con-
siderations is in order. Many authors (e.g., see Henderson, The Soave equation for a (eq 6 and only the first two terms
1979) have pointed out that the theoretical basis of the on the right side of eq 5) adequately correlates the vapor
Redlich-Kwong equation is weak. There is a valid sepa- pressures of nonpolar and slightly polar substances (Gra-
ration of the attractive and repulsive contributions, but boski and Daubert, 1978). However, it was found that the
the expression for repulsive contributions is correct only agreement with vapor pressure data of polar substances
for a one-dimensional system of hard rods. Further, the is seriously in error. By definition, the Soave equation
equation gives a universal value of the critical compres- reproduces the vapor pressures of all substances at the
sibility factor of 1/3 which is too large for all substances, critical temperature and a reduced temperature of 0.7.
leading to poor liquid molar volumes. However, perhaps Comparison with experimental data for polar substances
due to a fortunate cancellation of errors and certainly due shows that the Soave equation over-predicts the vapor
to the clever parameterization of Soave (1972), it has been pressure at reduced temperatures between 0.7 and 1.0 and
successfully used for the prediction of multicomponent under-predicts it at reduced temperature less than 0.7, or
phase equilibria in a wide variety of process design ap- vice versa. One of the simplest empirical terms which
plications. It should also be mentioned that the approach corrects this is the third term on the right side of eq 5, and
of this work will probably yield similar results with the fortunately it is found to work very well.
Peng-Robinson (1976) or Martin (Martin, 1979; Joffe, Table I shows the significant improvement in vapor
1981) equations. pressure correlation obtained by wing the polar parameter
Description of the Equation p for some common polar substances. We note that the
critical temperature and pressure used are the best ex-
Pure Fluids. The Soave equation is perimental value and the acentric factor is that which
p = - -RT
- U reproduced the experimental vapor pressure at a reduced
u - b U(U + b ) temperature of 0.7 using the Soave equation. Finally, the
polar parameter p was fitted to the experimental vapor
In order to reproduce the critical temperature and pres- pressure data.
sure, the constants a and b are given by The polar parameter p introduced in eq 5 is highly em-
pirical. It is unlikely that p can be correlated in terms of
R2Pd other quantities such as the dipole moment since ita value
ai(TCi)= = 0.42747-
Pci is probably a lumped result of many different effecta-
including the inadequacy of the Redlich-Kwong equation.
RTci (Note that ita value even changes sign as one goes to the
bi = 0.08664- (3)
Pci higher normal alcohols.) However, there is only one ad-
ditional parameter whose value could probably be obtained
Following Soave, we keep b as a constant and make a a from few experimental vapor pressure data. Further, the
function of temperature such that the vapor pressure of simple form of the Redlich-Kwong equation is retained,
the pure substance is well correlated. Thus and this is important for the computational robustness and
ai(T) aciai(T) (4) efficiency essential to process design applications.
CY for Supercritical Substances. Since the CY ex-
where ai(7') is an adimensional fador which takes the value pression was obtained by matching pure component data,
of unity at T = Tci. A good correlation for ai is given by use of the same expression a t supercritical temperatures
aP5 = 1 + mi(l - - pi(l - TR)(0.7 - TRi) (5) (T> TJ represents extrapolation into an unknown region.
Boston and Mathias (1980) have suggested the following
where T a = T/T,is the reduced temperature of compo- extrapolation at supercritical temperatures
nent i. up5 = exp[ci(l - TRd')]
The second term on the right side of eq 5 was introduced (7)
by Soave to reproduce the vapor pressures of nonpolar where
Ind. Eng. Chem. Process Des. Dev., Vol. 22, No. 3, 1983 387
3Or I I
I
eq 5 at TRi= 1; the value and first derivative have been Original RKS I---- GLL
matched and the second derivative of with respect to
Points Experimentoi D a t a
(Dymond and S m i t h ,
2000 ::
TRihas been set to zero. The Boston and Mathias ex- 19691 8
1500 -I
trapolation has been adopted in this work. >- 5
As explained in the Introduction, this work aims to IO00 f
design an equation of state to produce good agreement
with experimental phase equilibrium behavior. Given a
simple equation such as the Redlich-Kwong equation, this I 1 I I I I I
t
necessarily means a compromise, albeit small, in other 50 I00 200 500 1000 3000 5000
TEMPERATURE, K
quantities such as molar densities. Thus it is not appro-
priate to choose the supercritical a by enhancing agree- Figure 1. Virial coefficienta of nitrogen calculated by original and
modified Soave equations.
ment with PVT data. The Boston and Mathias extrapo-
lation provides better agreement with VLE data than the used. Perhaps the results of this work could be used as
Soave equation for common nonpolar binaries (Boaton and a benchmark against which the more sophisticated mixing
Mathias, 1980). Further, eq 7 produces a qualitatively rules may be evaluated.
correct high-temperature limit while the Soave equation It should also be noted that the simple mixing rules
does not. This is illustrated in Figure 1, which shows defined by eq 10-13 have been found to work very well.
second and third virial coefficients for nitrogen. The Soave For example, zero values for all binary parameters produce
a produces better agreement with experimental second good results for hydrocarbon mixtures, even those with
virial coefficient data but does go through a qualitatively large size disparities (Graboski and Daubert, 1978).
incorrect maximum a t a temperature of about lo00 K.
(We should note that this temperature is high enough to Applications
be of little practical interest and this is generally true for A wide variety of applications are shown in this work
all substances.) For the third virial coefficient it is hard which demonstrate that the simple equation of state de-
to tell whether the Soave a is better than eq 7 given the fined by eq 1-13 is flexible enough to correlate difficult
sparse available data. Again, the Soave a produces an phase behavior. No applications are shown for systems
unphysical minimum at a temperature of about 1000 K. exhibiting liquid immiscibility, but indications are that this
Extension to Mixtures. The one-fluid theory has been too can be satisfactorily correlated.
chosen to extend the pure fluid equation to mixture. Thus Experimental phase equilibrium data have been re-
the form of eq 1 is retained in the mixture case but the greased by finding optimum values of the binary parameter
parameters are obtained as simple quadratic mole fraction subset chosen through the use of the generalized least-
averages of the pure component values squares method of Britt and Luecke (1973). Values of
a = CCXiXjUi, (10) standard deviation chosen are 0.1 K for temperature and
i l 2% for the other variables. In cases where vapor compo-
sition data were not available, estimated values were used
and the standard deviation made extremely large. For
mole fraction data there is a possibility that the model
and values will move outside the range 0-1. Hence these
variables have been transformed by
solubil param,
solvent Tc, K Pc, atm W (cal/cm3)'' 2 Err. KH' Err. K s
tetralin 716.5 33.1 0.316 9.50 3.65 4.86
diphenylmethane 770.2 28.2 0.434 9.60 8.21 5.89
1-methylnaphalene 772.2 35.2 0.302 10.02 5.39 4.37
quinoline 782.0 45.0 0.333 10.90 4.31 4.62
m-cresol 705.9 45.0 0.466 10.78 3.93 2.29
m-xylene 617.0 35.0 0.331 8.82 3.80 3.76
toluene 562.1 48.3 0.212 8.92 4.85 6.42
Average absolute error in K value of hydrogen (see Table I).
Hydrogen-Containing Systems. There is ambiguity O \
62315K
Male Fraction average
Colculoted b y
modified R K S
20000 20000
4- 0000
473 15K
1
5000
3000
2000
Ln
w
IO+
Point5 Experimental Data
Calculated by modlfled R K S
I 1000
5000
3000
. 2 000
1000
5000
10-1
0 02 04 06 08
MOLE FRACTION E T H A N O L
IO
2ooo~
too0
, , , I I:f
02 04 06 08 10
Figure 4. Vapor-liquid equilibrium of the ethanol-water system MOLE FRACTION METHANOL
(data of Barr-David, 1959; and from Dechema, 1977). Figure 6. Henry constant of oxygen in the methanol-water mixed
solvent (binary parameters for oxygen-methanol: k 2 = 0.0527, kbl
5000r = -0,1732).
Poinls Experimental Data ( O e c h e m a i
I - Calculated bymodified R K S
range. Further, good description of binary phase equi-
librium does not guarantee a correspondingly improved
description of multicomponent systems. These issues will
be addressed in future work.
Acknowledgment
The author wishes to express his thanks to Dr, Joseph
F. Boston for helpful discussions. This work was part of
Advanced System for Process Engineering Project (AS-
PEN) supported by the U.S.Department of Energy under
Contract Number E(49-18)-2295.
Nomenclature
a = attraction parameter, eq 1
b = hard core volume, eq 1
c = constant, eq 7
d = constant, eq 7
k = binary parameter, eq 12 and 13
oc m = characteristic constant, eq 5
2
i p = polar parameter, eq 5
P = pressure
lo 02
04 d6 l o
R = gas constant
MOLE FRACTION METHANOL T = absolute temperature, K
Figure 5. Vapor-liquid equilibrium of the methanol-water system. u = molar volume
r = mole fraction
can correlate VLE data for the methanol-water syatem and z = transformed mole fraction, eq 14
the Henry constant for oxygen in the methanol-water Greek Letters
mixed solvent. In the latter case, the model is inadequate a = adimensional factor for a, eq 4
at low methanol concentrations but is in good agreement 6 = solubility parameter, ( c a l / ~ m ~ ) l / ~
with data for the rest of the concentration range. w = acentric factor
Conclusions Subscripts
The results shown in the work demonstrate that a simple c = critical property
Redlich-Kwong variant which is computationally robust i, j = component identifications
and economical can be successfully applied to a wide va- Literature Cited
riety of phase equilibria applications. It is expected to be Ambrose, D.; Sprake, C. H. S. J . Chem. Thefmodyn. 1970, 2, 631.
useful for process design applications. Ambrose, D.; Sprake. C. H. S.: Townsend, R. J . Chem. Themrodyn. 1974,
However, more testing must be done before the model 6 , 693.
Ambrose, D.; Sprake, C. H. S.; Townsend, R. J . Chem. lhermodyn. 1975,
can be accepted for engineering use. It is possible that the 7, 105.
simple mixing rules defined by eq 10-13 will not be able Asseiineau, L.; Bogdanlc, 0.; Videl. J. Flukl Phase qui&. 1979, 3, 273.
Barr-David, F.; Dodge. B. F. J . Chem. Eng. Data 1959, 4 , 107.
to describe highly nonideal systems (e.g., water-hydro- Boston, J. F.; Mathias. P. M. phase Equilibria in e TMrd-Generation Process
carbon immiscibility) at both ends of the composition Simulator": Resented at 2nd International Gonference on Phase Equiiibria
Ind. Eng. Chem. Process Des. Dev. 1983,22, 391-396 391
and Fluid Propertles in the Chemical Process Industries, Berlin (West), Krichevskii, I.R.; Zhavorokov, N. M.; Tsikiis, D. S. Zh. Fiz. Khim. 1937, 9 ,
17-21 Mar 1980. 317.
Brm, H. I.; Luecke, R. H. TechmmeMcs 1973, 15, 233. Martin, J. J. Ind. Eng. Chem. Fundam. 1979, 16, 81.
Chan, W. K.; Boston, J. F. "A New Alsorfthm for the Construction of P-T Mathias, P. M.; Boston, J. F.; Watansiri, S. "Effective Utilization of Equations
Emrdopes", Accepted for publicetion in Ind. Eng. Chem. Process Des. of State for Thermodynamic Properties in Process Simulation", AIChE J.
Dev. 1981. in press.
Chao, K. C.; Un. H. M.; Nagsshwar, 0. D.; Kim, H. Y.; Oiiphant. J. L.; Sebas- Meyer, C. A.; McCiintock, R. B.; Sihrestri, G. J.; Spencer, R. C., Jr. "1967
tian, H. M.; Slmnick, J. J. "Phase Equilibrium in Coal Liquefaction Steam Tables-Thermodynamic and Transport Properties of Steam C o n
Processes", Final Report, Research 367-2.Electric Power Research In- prising Tables and Charts for Steam and Water"; ASME, United Engineer-
stitute, Palo Aito, CA. 1980. ing Center, New York, Dec 1967.
Chueh, P. L.; Prausnitz, J. M. Ind. Eng. Chem. Fundam. 1967, 6 , 492. Micheison, M. L. FiuM P h a Equiib.
~ 1080, 4 , 1.
Conndiy. J. F. J. Chem. f h y s . 1962, 36,2897. Nichols, W. B.; Reamer, H. H.; Sage, B. H. AIChE J. 1057, 3 , 262.
Culberson. 0.L.; McKetta. H. J., Jr. Pet. Trans. AIM 1950, 189, 321. Peng, D. Y.; Robinson, D. B. Ind. Eng. Chem. Fundam. 1976, 15, 59.
Culberson, 0. L.; McKetta, H. J., Jr. Pet. Trans. A I M 1051, 192, 297. Reamer, H. H.; OMS, R. H.; Sage, B. H.; Lacey, W. N. Ind. Eng. Chem.
Dechema Chemistry Data Series "Vapor-Uquld Equilibrium Data Collection"; 1996, 2 8 , 1936.
Bahrens, D.; Eckermann, R., Ed.; Dechema. Deutsche Geseiischaft fur Rediich, D.; Kwong, H. N. S. Chem. Rev. 1049, 4 4 , 233.
Chemisches Apparatewesen: Frankfwt/Main, West Germany, 1977;Voi. Reid, R. C.; Prausnitz, J. M.; Sherwood, T. K. "The Properties of Gases and
1, Part 1. Liquids", 3rd ed., McGraw-HIII Book Company: New York, 1977;p 630.
Deming, E.; Shupe, L. E. fhys. Rev. 1932, 4 0 , 848. Saddington, A. W.; Krase, N. W. J. Am. Chem. Soc. 1934, 5 6 , 353.
Dymond, J. H.; Smith, E. B. "The Virlai Coefficients of Gases": Oxford-Ciar- Soave, G. Chem. Eng. Scl. 1072, 2 7 , 1197.
endon Press: Oxford, 1969. Tokunags, J. J. Chem. Eng. Data 1975, 2 0 , 41.
Gmhling, J. D.; Liu, D.; Prausnitz. J. M. Chem. Eng. Sci. 1979, 34, 951. Whiting, W. B.; Prausnitz, J. M. "Equatlons of State for Strongly Nonideai
Graboski, M. S.;Daubert, T. F. Ind. Eng. Chem. Process Des. Dev. 1976, Fluid Mixtures: Appllcation of the Local Composition Concept"; Presented
17, 443. at the AIChE Spring National Meeting, Houston, 1981.
Heidemann, R. A.; Khall. A. M. "The Calculation of Critical Points"; Presented Wiebe, R.; Gaddy, V. L. J. Am. Chem. Soc.1934, 5 6 , 76.
at the 86th National AIChE Meeting. Houston, Apr 1979. Wiebe. R.; Tremearne, T. H. J. Am. Chem. Soc. 1034, 56, 2357.
Henderson, D. A&. Chem. Ser. 1970, No. 182, 1. Williams, R. B.;Katz, D. L. Ind. Eng. Chem. 1954, 46, 2512.
Huron, M-J.; Vidal, J. HUM Phase Equiii6. 1979, 3 , 255.
Jaffe, J. Ind. Eng. Chem. Process D e s . Dev. 1981, 20, 168.
Kkk, B. S.; Zlegler, W. T. A&. Cryog. ng. 1965, 10, 160. Received for review June 29, 1981
Kiink, A. E.; Cheh, H. Y.; Amick. E. H., Jr. AIChE J. 1975, 2 1 , 1142. Accepted October 13, 1982
Procedures are devised for transforming process optimization problems Into signomlal programming problems. An
investigation of the various algorithms proposed for finding the global optimum of such a signomiai programming
problem showed the Falk algorithm to be most suitable. When the Falk procedure is coupled to an efficient
optimization program, global optima to small size problems are readily obtained. In a moderate size problem,
a close approximation of the global optimum can be obtained with reasonable computing effort.
Introduction One can be certain that the global optimum has been
The use of formal optimization or mathematical pro- obtained only if a maximization problem is a concave
gramming procedures has become quite common in process programming problem or a minimization problem is con-
engineering design. A number of large size, nonlinear, vex. We have a convex programming problem if (a) the
constrained optimization computer programs are generally function being minimized, f ( X ) is convex and (b) the
available. Such programs are capable of efficiently solving contrainta are only inequalities of the form gi(X)Ibi with
problems of the form each g i ( X ) also being convex. A concave programming
problem is the maximization of concave objective function
max ( m i d f(x) (1)
subject to g i ( X )I bi where the g i ( X )are concave. Almost
subject to all practical design problems are neither convex nor con-
g i ( m = bi (i = 1, 2, ..., 1) (24 cave and one cannot be certain that the global optimum
has been found.
gi(X){S,ZJbi +
(i = 1 + 1, 1 2, ..., m) (2b) Most designers have been willing to accept the fact that
Nonlinear problems with more than 75 variables and a they may not have located the global optimum providing
similar number of constraints can now be readily handled. the minimum or maximum they have located is consid-
A wide variety of optimal design problems have been erably better than they would have achieved without the
successfully solved by this approach. use of the formal optimization procedure. However, in
Essentially all of the available optimization programs many cases the system being designed is very costly and
are hill climbing (descending) techniques. They locate the an expediture of a considerable design effort can be jus-
constrained mRnimum (or minimum) closest to the starting tified to locate a better optimum. A common procedure
point, but they will not necessarily determine the global in this case is to repeat the optimization process for a series
maximum (or minimum) if there are multiple extrema. of different starting points in the hope that one of the
chosen starting points will be in the vicinity of the global
optimum. This technique is often successful. However,
f Case Western Reserve University, Cleveland, OH 44106. Wilde (1979) points out that in some real optimization
0196-4305/83/1122-0391$01.50/0 @ 1983 American Chemlcal Society