Академический Документы
Профессиональный Документы
Культура Документы
SCOPE
The models of multicomponent, multistage counter- phase distillation problem presented here requires the
current unit operations, such as distillation and liquid- conversion of the modeling equations into forms similar
liquid extraction, are usually restricted to two counter- to those of the two-phase distillation problem. These
current phases. A significant number of chemical engi- equations are then solved by a Newton-Raphson itera-
neering processes, however, have two partly miscible tive procedure for correction of the liquid compositions.
liquid phases and a vapor phase. This type of system The simulation model presented can be used not only
cannot be simulated adequately by assuming only two for the simulation of three-phase distillation columns,
phases. Therefore, a model of the distillation column with but also for columns with any combination of two- and
three-phase stages has been derived. three-phase stages. This is believed to be an essential
The more recent approaches to solving the modeling feature of the method, because in a distillation process
equations of two-phase multistage countercurrent opera- involving components which may form regions of im-
tions use the Newton-Raphson technique within iterative miscibility the phase type of a column is usually not
solution procedures. The calculation method for the three- known until the computation is completed.
A new model has been employed to simulate the and consists of linearizing the component mass balances
steady state distillation column. I t takes into account with the liquid compositions as the independent varia-
three phases, a vapor phase and two partly miscible liquid bles. The method can be regarded as a modified bubble
phases. Thus, the model also includes the simulation of point procedure which has a block tridiagonal form rather
a liquid-liquid phase separator within the distillation than the usual tridiagonal structure. This was found to
column. be a very efficient solution algorithm because of the highly
To describe analytically the phase separation of the composition dependent equilibrium ratios encountered in
liquid at a stage, the phase splitting parameter is intro- three-phase distillation.
duced. A three-phase stage is defined by 0 < 7 < 1, It is known that industrial distillations frequently have
a two-phase stage by 7) = 1. If 7) = 1 for all stages, three phases. These distillation columns can he simulated
the new modeling equations reduce to the well-known with the new model. The selected exampler cover the
equations of the two-phase distillation. Therefore, with simulation of stripping columns with a top phase separa-
the presented set of modeling equations, all types of tor, the very common case where low boiling components
stages can be adequately simulated. are removed by distillation from a mixture which is only
A computational scheme for the three-phase distilla- partly miscible, and a particular distillation operation,
tion problem is presented. It is based on an iterative solu- by which an additional separation effect is gained by a
tion method which uses a Newton-Raphson technique side stream phase separator.
M
C
i=1
Xf’i,n = 1, for 1 4 n 4N (5) Equations (6), (7), and (8) shows that only two of
these equations are independent; therefore, the equilib-
Phase equilibrium equations: rium ratios are interrelated by
= K‘i,n/K’‘i,n, for 14 i A M , 1 4n N
yi,n = K’i*nX‘i,n, for 1 i M, 1 nLN (6) (9)
yi,n = K”i,nx”i,n, for 1 6 i 6 M, 14 n A N (7) For the solution of Equations (1) to ( 8 ) , the follow-
xffdn = K*i,nX‘i,n, for 16 i 6 M, 16 n N (8) ing functions are assumed. The stage index n is dropped
for convenience in the following section.
The three-phase equilibrium of the vapor phase and Enthalpy function of the gas phase:
the two liquid phases ’ and ” is represented by Equa-
tions (13) to (8). The K‘i,n and K”i,,, are the vapor-liquid hG = hG(vi, vj, . . . p , T ) (10)
equilibrium ratios of component i at stage n with refer-
ence to the liquid phases ’ and ”, respectively. The
Enthalpy functions of the liquid phases:
liquid-liquid equilibrium ratio is A comparison of h’ = h‘(X‘*, X‘j, . .. T ) (11)
-;I
:G C,;
Stage n
SX If ?n = 1 for all stages, the new modeling elpations re-
F" stose n SL"
G:
SL" duce to the well-known equations for the two-phase
, , ~
distillation.
0,' Q"'
n 6-1 L" Ip Gn-1 Ln From Equations ( 17), ( 1 8 ) , and ( 19) we obtain
and
L', + S', = T,E,, for 16 n A' (20)
2
i=l
Xi,n = 1, for 1 5n L N (22)
Liquid-liquid equilibrium ratio: By using these definitions, the mass and erithalpy bd-
ances, Equations (1) and ( 2 ) , can be converted into
forms which are similar to those of the two-phase distilla-
tion.
NUMBER OF UNSPECIFIED VARIABLES AND
EQUATIONS Component i mass balance equation:
There are for the three-phase distillation the N ( 3 M Gn-lyz,n-l + L'n+lX',,n+l + LNn+lx''t,n + 1L F,~t,n
METHOD OF SOLUTION OF EQUATIONS The final solution requires the ri,n to be zero; hence, the
r"+l
i,n
are set equal to zero. The variation of the stage
Most of the model equations which describe multi-
index m of Equation (35) follows from Equation (23)
component, multistage countercurrent flow columns are because the residuals ri,n of stage n are only functions
nonlinear. Therefore, in the usual approach to solving
these e(quations, successive iteration piocedures are em-
of the variables of stages n - 1, n and n +
1. Since the
ployed. The generalized linear methods (Ortega and scaling restriction, Equation (22), has to hold, the A;k,m
Rheinboldt, 1970) treat all equations as linear. The are correlated by
well-known bubble point and sum rate methods of Friday
and Smith (1964) are examples of this class. This con-
cept was improved by Wang and Henke (1966) by
z-
M
k=l
i?rxk,m = 0, for 1L m L N (36)
developing a tridiagonal matrix method. The more re- M-1
-
cent methods of Thomich (1970), Goldstein and Stan-
field (1970), Naphtali and Sandholm (1971), Gelbe and
AXM,~ =- 2 Axk,*, for 1 m 6N (37)
k=l
Nomine (1971), and Ishii and Otto (1973) account
for the high nonlinearity of some of the equations by If Equations (35) and (37) are combined, we get
solving these with Newton-Raphson methods.
In principle, the system of equations describing the
three-phase distillation column can be solved by any
of the commonly used procedures for the two-phase
distillation problem. In the following, we will focus our for 1 L i L M - 1, l'n" (38)
attention on correction methods which use the Newton-
Raphson technique. There is an advantage to solution The partial derivatives of Equation (38) together with
procedures which use the liquid compositions (Ishii and Equation (23) become
Otto, 1973) or, equivalently, the component flow rates
(Naphlali and Sandholm, 1971) as independent variables,
obtaining the improved liquid compositions by lineariza-
AIChE: Journal (Vol. 22, No. 3) May, 1976 Page 585
TABLE2. NRTL PARAMETERS as fellows:
R&= --I (41)
Components ij Tij 7% ffij
The r is the vector of the residuals of the component mass
Butyl alcohol ( 1 1- balances; the AX is the vector of the corrections of
Water ( 2 ) 0.90047 3.51307 0.48 the actual Xk,n. The Jakobian matrix R, the matrix of
Water ( 2)-Butyl the partial derivatives, has a block tridiagonal form.
acetate ( 3 )
Butyl acetate ( 31-
Butyl alcohol ( 1 )
Butyl alcohol ( 1 )-
Water ( 2 )
Water (2)-Propyl
alcohol ( 3 )
5.04652
-0.30827
0.90047
2.7425
1.75717
1.15161
3.51307
-0.07149
0.34
0.30
0.48
0.30
Thus, Equation (41) may be written as
I
a1.1
a2,1
a1,2
%n-1
a22 %,n
a2,3
aN,N-1
%,n+1
aN,N
I:![) bN
(42)
=-
CN
Propyl alcohol (3)- The elements of the matrix in Equation (42) are quad-
Butyl alcohd ( 1 ) -0.61259 0.71640 0.30 ratic submatrices, the elements of which are
Butylacetnte (31
molelmole
where
(44)
[gE 1
-8
and
b,=
U M -1.n
D
V\ V
V V
cn =
ButylalCohd(l1 02
y2 -Oh 06 08 Water121
Fig. 3. Stage compositions of a stripping column with top phase It is obvious from a comparison of Equations (38) and
separator (Example 1). (39) how the partial derivatives in Equation (44) are
Flow rates:
F = 50 mole/h, B = 34.84 mole/h, D = 15.16 mole/h, R = 13.8 obtained. The block tridiagonal matrix system, Equation
molelh. (42), is solved and yields the &,n for 1 f i 6 M - 1,
liquid compositions 1‘nLN.
vapor compositions
6. The A&,,, for 1 6 n 6 N are computed by Equa-
I__) liquid-vapor tie line
-- - liquid-liquid tie line tion (37).
-- - binodal curve 7. The new ?i,n are computed from the following equa-
tion:
- -
xv i,n
+l - xui,n + G , n , for 1 ~i 4M , 1 ~n A N
(47)
k is a damping factor. It may be computed with a method
outlined by Thomich ( 1970).
- Steps 2 through 7 are repeated, using the new sets of
xu+1, until two independent convergence criteria for each
i,n
of the ri,n and G i , n are satisfied.
CONVERGENCE CHARACTERISTICS
- ZnA&,n= 0, for l L i A M - 1, l 4 n A N (39)
The convergence characteristics of solution procedures
In Equation (38) or (39),the partial derivatives of the which use the Newton-Raphson technique are known to
vapor compositions are treated as a function of the liquid depend on the initial estimates of the independent vari-
compositions X k and of the temperature T . Thus we have ables. This experience was also made with the presented
to write, omitting the stage index solution algorithm. In practice, however, the designer
of a distillation column first inspects the phasc equilibria,
thus getting a rough idea what the initial estimates
should therefore be. Under such condii ions the conver-
gence stability of the proposed procedure is verv satisfying.
The presented solution algorithm for three-phase dis-
tillation can be extended by increasing the number
- of
Since the temperature T is used as a dependent variable,
the system will remain near the bubble point condition. the independent variables; for example, the G,, Ln, and T,
Equation (38) may also be expressed in vector notation can be employed in addition to xz,n.
In this case we
Page 586 May, 1976 AlChE Journal (Vol. 22, No. 3)
OF
TABLE3. PRINT-OUT THE SIMULATION WITH THREE-AND
OF A DISTILLATIONCOLUMN
TWO-PHASE
STAGES(EXAMPLE 2)
N T -G Y1 Y2 Y3
L I..' L" R1 R2 33 X1' X2' x3 9 xi" X2" X3"
12 88.4 0000 0.014694 0.601846 0. 383460
12 116.00 0.030840 00619607 0.349552
11 88.9 116.00 0 0 030840 (-1 6 19607 0.349552
11 87.10 0.057763 0.647114 0.295 123
h)
h) 10 8Y.5 116.10 0 0 5 10 3 8 (I 640244 0 . 30871 8
10 870x2 0.084724 00672762 0.242514
Z 9 90.1 116022 0 07 12 78 0 6 5 Y 498 Om 269224
? 9 87.31 0.107482 0.690801 0.201717
w
Y 8 90.5 116.31 0.088372 0.673049 0.238578
8 137.43 0.130386 0.689119 0.180495
7 90.9 116.43 0.105757 0.688604 0.205639
7 137.56 0 0 145868 00702242 0 0 151890
6 91.3 116.56 00124055 0.704093 0 0 171852
6 137.66 0.161569 (10711946 0.125485
5 91.7 116.66 0.142601 00716722 0 0140677
5 137.73 128.49 9.25 00176541 0.721817 0.101642 00187443 0.705573 0.106984 0.025094 00947484 0.027422
4 92.0 116073 0 0 160279 0.727187 0.112534
4 137.19 118.88 18.91 00190069 0.729288 0.080643 00216519 0.693227 0.090254 0.023784 0.955993 0.020223
3 92.2 116.79 0.176248 0.735998 0.087754
3 137.84 111069 26.15 00201919 U.735426 0.062655 0.245862 0.682274 0.073864 0.022770 0.962448 0.014782
2 92.4 116.84 0.190233 0.743236 0.06653 1
2 137.88 106.19 31.68 0.212108 0.740407 0.047485 0.268828 0.672697 0.058475 0.021992 0.967358 0.010650
1 92.6 116.88 0.20225 7 0.749110 0 0048634
1 21000 19.11 1.89 00266935 00691971 0.041094 0.291195 00664390 0.044416 0.021393 0.971128 0.007478
il 22 23
0.130000 0.650000 00220000
NSL r s SL 5' 599 XSL 1 XSLZ XSL3 X1' x2 9 x3 ' X1" X2" X3"
12 88.4 29.00 OoO3084O 0.619607 0.349552
EXAMPLE PROBLEMS
Three example problems, which are typical to those ACKNOWLEDGMENT
encountered in the chemical industry, were selected to The authors wish to thank Dr. D. Wolf for his valuable sug-
illustrate the simulation of three-phase distillation columns. gestions.
For demonstration purposes, the number of components
has been limited to three. The calculation of the phase
equilibria for the examples is outlined in the Appendix.
The NRTL equation was used for the computation of
NOTATION
the activity coefficients. The NRTL parameters are listed
in Table 2. The vapor pressures were computed with a, b, c, d = elements of matrix or vector
the Antoine equation. B = bottom product molar flow rate
The simulation of a stripping column with a top D = top product molar flow rate
phase separator is demonstrated with the first example. F = feed stream molar flow rate
This kind of operation is used for removing small amounts f = fugacity
Page 588 May, 1976 AlChE Journal (Vol. 22, No. 3)
G = vapor stream molar flow rate Ishii, Y., and F. D. Otto, “A General Algorithm for Multistage
h = specific enthalpy Multicomponent Separation Calculations,” Can. J . Chem.
I = number of equations for a distillation column Eng., 51, 601 (1!373).
Naphtali, L., and 11.S. Sandholm, “Multicomponent Separation
K = equilibrium ratio Calculations by ]Linearization,” AIChE I., 17, 148 (1971).
L = liquid stream molar flow rate Ortega, S. M., and ’W. C. Rheinboldt, Iterative Solution of Non-
M = total number of components linear Equations in Seueral Variables, Academic Press, New
N =: total number of stages, including reboiler and York ( 1970).
condenser Prausnitz, J. M., Molecular Thermodynamics of Fluid-Phase
p =: pressure Equilibria, Prentice-Hall, Englewood Cliffs, N. J. (1969).
=: rate of heat input to a stage Renon, H., and J. M. Prausnitz, “Local Compositions in Ther-
Q modynamic Excess Functions for Liquid Mixtures,” AIChE J.,
R =: reflux stream molar flow rate 14,135 (1968).
R =: Jacobian matrix Thomich, J. F., “A New Simulation Method for Equilibrium
r =: residual of the component material balance Stage Processes,” ibid., 16,229 ( 1970).
r =: column vector of r Wang, J. C., and G.. E. Henke, “Tridiagonal Matrix for Distilla-
S =: side stream molar flow rate tion,” Hydrocarbon Process., 45, 155 (1966).
SL =: externally specified liquid side stream molar flow
rate APPENDIX A: CALCULATION METHOD FOR
T =: temperature EQUILIBRIUM RATIOS
x =I composition of liquid, mole fraction
x =: column vector of x The condition of the three-phase equilibrium for given pres-
composition of vapor, mole fraction sure ( p G = p’ = p ” ) and temperature ( T G = T’ = T”) is
y =I