.

CALCULATION OF MULTICOMPONENT DISTILLATION

OF NON-IDEAL MIXTURES BY A SHORT-CUT METHOD

EGON ECKERT and VLADIMIR HLAVACEK

Department of Chemical Engmeermg, lnstltute of Chemical Technology, Prague 6, Suchbatarova 1903,
Czechoslovakia

(Recerved 7 February 1977, accepted 21 Apnl 1977)

Abstract-The Smith-Bnnkley method has been modtied for use m computmg non-ideal m&component dlstd-
latlons The procedure proposed IS of pa.rtlcular advantage for rapld calculatmg hydrocarbon fractionators

1 INTRODUCTION
-0
Rgorous calculations of multlcomponent dlstdlatlon are
so tedious that chemical engmeers have long been look- LN+, = RD

mg for a simple short-cut method that would be capable I

of accurately and rapldly calculatmg multlcomponent -N
dlstlllatlons with muumum effort Procedures of this type
N-l
are valuable In developing the computer-aided-design
programs Despite the considerable effort expended by
process engmeers m the last three decades there 1s lack
V L
of mformatron on procedures which might be used for a I I
short-cut calculation of non-ideal multlcomponent dls-
F . M+2
tlllatlon problems -
M*l
Among the most general short-cut calculation me-
thods 1s that of Smith and Bnnkley[l] which may be M
applied to all types of equdlbnum stage dlstillatlon
processes except for cases wherem equlhbnum rela- V L
tlonshlps are non-Ideal To reduce essentmlly the I I
machme time required for a ngorous calculation of
,2
hydrocarbon fractionators using the Chao-Seader type
- 1
of equlhbrmm descrlptlons, Scheller [2] proposed a 4, --- .

simple procedure based on Cox’s idea of correlatmg the

temperature dependence of vapor pressure of a par- t
0
t~.~lar component Utlllzatlon of this approach gn!es nse
to a short-cut procedure which can also handle the Fig 1
non-ideal sltuatlons
The aim of this paper 1s to develop a short-cut method gwen by
for calculation of non-Ideal dlstlllatlon problems and to
report experience wrth apphcatlon of this procedure B,
En=exp A,+ (2)
( 0 5(TT + TF) + c, >
2 MJZI’EOD OF CALCULATION

The method proposed will be presented for a simple Bc

KS, = exp A, f (3)
dlstlllatlon column pictured in Fig 1 The approach of 05(TF+TB)+C, >
the Smith-Brinkley method IS based on an lteratlve
Improvement of the temperature TF which 1s used for Here TT and TB are the dew-pomt and bubble-point
calculatmg the average eqmhbnum dlstnbutlon co- temperatures of the outlet streams, respectively It 1s
efficlents m the strlppmg and rectdymg part of the evident that for a non-ideal mixture, where the equlh-
column If the temperature dependence of the equi- bnum coefficient depends both on temperature and
bnum coefficient K, corresponds to the Antome equation concentration, this procedure cannot be applied In the
development of the method, nomdeahty m the liquid and
vapor phases wdl be consldered, me the equdlbrmm
(1) &stibution coefficient IS qven by

then the average eqmhbrmm coefficients m the rectlfymg

@a)
and strlppmg sectlons a, and I& respectively, are
78 E ECKERT and V HLAVAEEX

For low pressure vapor-hqmd equlhbrla eqn (4a) may FT = 0 5(T, + Tzz)
often be reduced to i% = 0 S(T, + TF)
3ALGoRnmM
KT = y,$ = -y,K, (4b) A brief descnptlon of the algonthm wdl be presented
1 All preliminary calculations resultmg from spe-
A detailed analysis of the results calculated by ngorous cification of the problem will be performed, e g cal-
methods mdtcates that frequently the dependence of K* culation of vapor and hquld flow rates m both parts of the
on temperature may be reasonably well described by an column and of reflux ratio, evaluation of bottom and top
emplrlcal relation[2] pressure, speclficatlon of CT and the first guess of the
separation factor f,
2 Based on guessed composition of the top and bot-
(54 tom product the dew point TT and bubble point TB are
calculated
For low pressure eqmhbna VP/+, + 1 and since the tem- 3 Calculation of the temperature TF In the first
perature dependence of K, IS known we can approximate Iteration the feed temperature may be used or a linear
only the changes of y, m the column mterpolatlon between TT and TB with regard to the feed
tray location 1s performed In the following iterations and
BT Newton method 1s used to calculate a new guess of TF
y, =exp AT+- (5b) We denote
.>
D.-D,
Here T IS the temperature at a particular equlhbnum l=T (10)
stage Of course, both approxlmatlons inherently include
the effect of composltlon of each of the components where D, and D, are the specified and calculated top
Makmg use of eqns (5a) or (Sb) for the top and the product respectively
bottom of the column the equlhbnum coefficients KT (or For a correct guess of TF, E = 0, I e the Newton-
n) which are a by-product of calculation of dew and Raphson formula yields
bubble points may be adopted to express the unknown
coefficients AT and BT E’( TFk-’ - TKk)
TFk+’= T,* _ k-l k (11)
l --c

A: = In KS, - In $$/((TB + Cp)/( T= + CT) - 1) (6)

Ia If l k IS smaller than a pre-assigned tolerance then the
calculation IS fimshed
BT = In$$‘/(l/(T, + CZ) - l/(T, + CT)) (7) 4 Calculation of the constants At and Bt and the
average dlstnbutlon eqmhbnum coefficients Z@, and E$,
from eqns (9a) and (9b) Calculation of effective stripping
Relationships of the same type are valid for A$ and BT factors for each component m the stnppmg and rectlfy-
computed from yTt and yB, We have no additional mg section S,,, and S,,,,, respectively
mformatlon how to calculate the unknown coefficient
CT Therefore, m our calculations we have adopted the
Calhgaert-Davis relation (12)

B, The value of H, to be used depends upon the nature of

lnK,=A,+T_43 15, (8)
the feed For a feed more sumlar to liquid phase

ie C,=CT=-43 15, i=l,2, ,n

Having an estlmatjon for the constants AT and BT the (13)
averagd eqmhbrmm dlstnbutlon coefficients may be
easily calculated m each iteration For a feed more slmllar m nature to the vapor phase

BT
RX= exp AT+- cw H,=; (14)
TT+C? >

(9b) In addition, the fraction of the component f, winch will

be recovered at the bottom of the column will be cal-
culated according to expressions developed by Smith and
where Brinkley [ l]

(I - S;-“) + ZW - Snd
f.= (I- Sz-“) + R(1- S,,) + H,SEwM(l - SZ+‘)(l - Smr)/(l- Sml) (15)
 Calculation of multicomponent dlstlllatlons 79

For s,, += 1 and S,, + 1 I’Hospltal’s rule must be used Example 2

The values f, yield the recovertes of each component In this example a more difficult problem, gasohne
whtch leaves the column stablhzatlon, will be considered [3]

B-G?, = FXF& Problem statement

OX, = FX, - BX,, Feed composlt;on I component -J&i

1 methane 0 020
5 Return to pomt 2 2 ethane 0 100
After performmg these calculations reboller and con- 3 propylene 0 060
denser heat load may be evaluated 4 propane 0 125
5 lsobutane 0 035
6 n-butane 0 150
4 CALCULATION EXPERIENCE EXAMPLES 7 n -pentane 0 152
The proposed procedure has been tested for a number 8 n -hexane 0 113
of examples Some results are presented below All 9 h -heptane 0090
calculations have also been performed by USI~ a 10 n-octane 0 085
machine program solvmg rigorously all governing heat 11 n-tetradecane 0 070
and mass balances For all examples presented the value
of the constant Ct has been chosen Ct = -43 15 (I = Number of trays 11+ reboller + partial condenser
l-2, I n) Feed hquld IS fed as bollmg hqmd m stage 9,
100 moles/unit time
Pressure 2 068 lo5 Pa, the pressure drop m the coIumn
Example 1 IS not considered
A butane-pentane splitter will be considered [ l] Reflux 20
Dstlllate 31 6 molelumt time
Problem statement for the example Thermodynamic specdicatlon see Example 1
The results obtained from the rigorous and ap-
Feed bollmg hquld, feed tray location-stage 5, feed
proximate calculations are reported tn Table 2 From this
rate 100 mole/unit time
table may be mferred that the agreement between the
Feed composltlon 1 component XI? approximate and rigorous results 1s excellent
1 propane 0 0.5
2 lsobutane 0 15 Table 2 Overhead and bottom composltlons Example 2
3 n-butane 0 25
4 isopentane 0 20
5 n-pentane 0 35 1 0 0633 0 0633 00000 00
2 03164 0 3165 00000 OOCUlO
Number of trays 9 + reboller + total condenser 3 0 1834 0 1865 0 0030 00016
4 0 3737 0 3841 0 0101 0 0053
Pressure 8 274 x lOsPa t the pressure drop m the
5 0 0243 0 0216 0 0399 0 0412
column 1s not consldered) 6 0 0380 0 0276 0 2017 0 2065
Reflux 2 579 7 00009 OCOO4 0 2218 0 2220
Dtstlllate dtstlllate rate 48 9 mole/unit time 8 OOOOO OOOOO 0 1652 0 1652
Thermodynamic SpeLlfiCdtlOn The equlltbnum dlstnbu- 9 00000 00000 0 1316 0 1316
10 00 00 0 1243 0 1243
tlon coefficients are calculated from the Chao-Seader 11 00 00 0 1024 0 1024
correIatlon Corrections of flow rates using the enthalpy
balances are not performed Lf,, = 282 3 K (total condenser), TTr= 316 4 K (partial
The results obtained from the rigorous and ap- condenser), TBr = 451 3 K, TBr = 452 3 K, S, short cut
method, r rigorous method
proximate calculations are reported m Table 1
It 1s obvious that the agreement between both pro-
Example 3
cedures 1s very good
In this example a strongly nomdeal mixture ~111 be
considered [3]
Table 1 Overhead and bottom composltrons Example 1 Problem statement
Feed composltlon i component X,
1 n-hexane 03
1 0 1021 0 1021 00001 00000 2 ethanol 01
2 0 2936 0 3005 0 0125 0 0056 3 methylcyclopentane 03
3 0 4526 0 4775 0 0562 0 0317 4 benzene 03
4 0 0821 00699 0 3129 0 3249
5 0 0696 0 0501 0 6183 0 6379
Number of trays 8 + reboller + total condenser
T,,=3364K, T,,=3356K, T,.=3817K, TBr= Feed Bollmg hquld IS sent to stage 4, lOOmote/umt
383 6 K, s, short-cut method, r, rigorous method ttme
80 E ECKERTand V HLAVA~EK

Pressure 1 034 x 10’ Pa, the pressure drop m the column Table 4 Overhead and bottom cornposItions Example 3
. IS not consldered
Reflux 20
Dtsttllate 50 mole/unit time 1 0 3510 0 3504 0 4005 0 2490 0 24% 0 1994
Thermodynamic spec&atlon The equdibnum co- 2 02000 02ooo 00693 00000 00000 0 1307
efficients are calculated from eqn (4b) using the followmg 3 0 2766 0 2717 0 3418 0 3234 0 3283 0 2582
dependence P? = Pp( T) 4 0 1724 0 1779 0 1884 0 4276 0 4221 04117

T,,=3344K, T,,=3344K, T,,,=3462K, T,,=3553K,

T,,=3452K, TStd= 349 0 K, s, short-cut with nomdeahty, rd.
short-cut wlthout nomdeahty, r, rigorous
Numerical values of the constants C1, , C5 are
culatlons 1s excellent Table 4 reveals that nomdeahty of
presented m Table 3a To describe the nomdeahty in the
the solution must not be neglected m the short-cut cal-
liquid phase Wdson’s equation has been adopted
culation
5 CONCLUSIONS ANDDISCUSSION
ln~,=l-ln(z~A,,)-~~~~ (17) Frequently the number of Iterations IS less than 20 m
J kJ order to reach a 0 1% tolerence of the calculation,
where however, the storage requvements and computer time
expenditure are by a factor 100 lower m comparison with
A,, = 2 exp (-(A,, - A,,)/RT) (f@ rigorous calculatgons Evidently, the posslblhty of handl-
mg nomdeal mixtures by the Smith-Bnnkley procedure
essentially enlarges the area of application Based on our
The values of bmary interaction coefficients are m
experience with the modified Smith-Brinkley procedure
Table 3b Table 3c presents numerical values of the
it 1s possible to calculate with high preclslon the hydro-
temperature dependence of molar volumes
carbon fractionators making use of the Chao-Seader
correlation As a result, both the computer tune and
v, = al, + a2,T + as,T2 (19) computer memory are very low Hence this procedure
may be used as a sub-routme m a computer-aded-desan
Flow rates m the column have not been corrected
program However, on the other hand, we may expect
using the enthalpy balances
that for azeotroptc distillation dticultles may occur The
The results of rigorous and approximate calculations
procedure proposed cannot be used If for TT < TB both
are presented m Table 4
of the followmg relatronshlps hold K, < &, as well as
Though the system IS strongly nomdeal (n = 1 04 -
K.TF> IG, Nevertheless, also for strongly non-ideal
124, yz=33-133, y3=103-12, y,=14-13) the
solutions, the procedure suggested can be used m the
agreement between the rigorous and approximate cal-
majority of cases

Table 3a Numerical values of the constants C, NOTATION

A, B, C constants m temperature dependence of eqml-
lbrmm coefficient
B molar flow rate of bottom product
1 124 81 -715149 00 0 020 -17 0
2 135 44 -8754 09 00 0 020 -18 1 D molar flow rate of top product
3 20 697 -2731 00 -4711 00 00 f separation factor in column
4 14484 -8026 29 00 0 024 -20 3 H parameter m the Smith-Bnnkley method
K equlllbrmm dlstnbutlon coefficient
Table 3b NumerIcal values of the bmary m- L hqmd flow-rate m column
teractlon coefficients A,, - AU M number of trays below feed stage
N number of trays m column
41 1 2 3 4 number of components
i pressure m column
1 00 283 63 272 09 169 92
2 2282 99 00 222147 1297 9 PP vapor pressure of component I
3 -175 70 16153 00 97 33 %. 4c reboder and condenser heat loads
4 173 93 13147 161 44 00 R reflux ratio
sn, s#n stippmg factors m rectdymg and strlppmg part
Table 3c Numerlcal values of the coefficients
of the column, respectively
.
a,
T temperature
X molar concentration
1 al, a2, lo3 a3, V molar voIume
V vapor flow-rate m column
: 125
53 96
70 -003111
-0 14456 0 547
160 Greek symbols
3 104 27 -0 08676 0390 y activity coefficient
4 70 86 0 01491 0 159
l residuum
 Calculation of multlcomponent dlsttiafions 81

d fugaclty coefficient m vapor phase r calculated by ngorous method

A, - A,, bmary constants m W&on equation s specified value, calculated by short-cut method
A binary constant of Wilson equation T rectlfymg part of column
Y0 fugaclty coefficient m liquid phase - average value
* nomdeal behawor
Induces ’ stnppmg part of column
B bottom product
D dlstdlate product
F feed REFERENCES
c calculated value
Ill SmlthB D andBnnkleyW K,AIChE J 19406446
1 component I w Scheller W A , Personal commumcation (1975)
rd Ideal behavior of hquld and vapor phases [31 Boston J F , PhD Thesls, Tulane Umverslty 1970

CESVol 33 No I-F