You are on page 1of 2

ChemicalEngineeting Scimcr, 1975, Vol. 30, p. 1447. PergamonPress.

Printed in Great Britain

A rigorous calculation method for the minimum stages in multicomponent distillation

(Received 6 January 1975)

Dear Sirs, with various assumed values of M until the case resulting in the
Recently I had occasion to read the article “A Rigorous smallest value for N t M is found. If N and M were continuous
Calculation Method for the Minimum Stages in Multicomponent variables the required additional relation would of course be
Distillation”, by H. H. Y. Chien[l]. It appears that the algorithm d(M + N)/dM = 0. Since N and M are actually each integers it is
presented therein will not necessarily determine the minimum readily discerned that the number of such solutions required
number of stages for the reason described below. certainly does not exceed N t M, and that a solution to the
The section “Feed Stage Composition Matching”, in[l] begins problem at any reflux ratio places an upper bound on this number.
with the paragraph, Chien’s procedure in fact determines a minimumconstrained by
“As mentioned previously the total number of variables in eqn his eqn (30). This is probably illustrated by his example wherein
(1) is one more than the total number of equations in eqns (2)-(6). his procedure determines a minimum of 9.26 stages as opposed to
Therefore, only the combined value of (N+M) may be 8.44 stages obtained using Winn’s method.
determined. To determine the values of N and M separately
another equation must be defined”. IntemaGonal Business Machines Corp. D. S. BILLINGSLEY
The additional equation is then given by Chien as his eqn (30) P.O. Box 1369
and states that for instance the ratio of the mole fractions of light Houston
and heavy keys in the liquid on the feed stage, shall equal the TX 77001
corresponding ratio a liquid feed. Only by chance would any such U.S.A.
condition exist at the minimum number of stages even in a binary
mixture. Chien admits his eqn (30) is arbitrary. It is actually
incorrect for the intended purpose since the minimum number of REFERENCE
stages will be found by ignoring it and solving Chien’s eqns (2~(6) [ll Chien H. H. Y., Chem. Engng Sci. 197328 1967.

ChemicalEngineeringScience, 1975, Vol. 30, pp. 1447-1448.PergamonPress. Printed in Great Britain

Author’s rebuttal to the letter of D. S. Billingsley on the paper “A rigorous calculation method
for the minimum stages in multicomponent distillation”

(Received7 February 1975)

Dear Sirs, and


When a distillation column is operated at total reflux, the ratio of
internal liquid (and vapor) flow rates to the feed rate is infinite. T’; j=l,...,J
Therefore, it makes no difference to a total reflux column where the
feed has been introduced. Neither the separation, nor the minimum with the following equations:
stages of a column is affected by the feed location. Expressed
mathematically in the form suggested by reader Billingsley: +:+ 1-g xr’; i=l ,...,n (2)
( >
d(J)
-=o x,’ = k /‘k.‘-’
I --- k,‘xi’, i=l,...,n (3)
dM
2 k/k,‘-‘---ki’x,‘; j= 1I.. .,.l (4)
where J-M t N,the minimum number of stages.
In fact, if one rearranges eqns (2)-(6) in the following way, the
x“ = h’ (5)
number of equations equals the number of unknowns and the
variable M (or N) disappears from the variable set (1). Given x,’ = fH’. (6)
zi; i=l,...,n; X2,=1 The above equations can be solved by estimating and then
converging on the bottoms compositions instead of the feed plate
and
compositions as presented in the original paper. The advantage of
17,‘,x;’ this approach (“Lewis and Matheson” algorithm) has also been
discussed.
it is to find The arbitrary eqn (30), introduced in an attempt to provide an
estimate of the optimum feed location of a continuous column
xl; i=l,...,n based on the converged composition profiles of a total-reflux
column, is not needed in the calculation of J and therefore will not
xi’; i=l ,...,n (1)
tiect the value of J. It may be applied after the convergence of a
DIE J solution in the Lewis-Matheson approach.

1447
1448 Letters to the Editors

The difference between the result of this procedure and that of I am sorry that these points were not made clear in the paper.
Winn’s method can be attributed to the following:
Monsanto Company HENRY H. Y. CHIEN
(1) third Antoine constants for the components in the system are 800 N. Lindbergh Blvd.
not identical as assumed in Winn’s method; St. Louis
(2) fractional stage definitions used in the two approaches are MO 63166
different. U.S.A.

Chemical EngineeringScience,1975,Vol. 30. pp. 1448-1449. PergamonPress. Printed in Great Britain

D. S. Biilingsley’s reply to the author’s rebuttal of the comments on the paper “A rigorous
calculation method for the minimum stages in multicomponent distillation”

(Receioed 8 March 1975)

Dear Sirs, Y, connects the points (l/2,3/4) and 1, 1), and has the
Certainly I agree with Chien when he points out that the equation y = (1/2)x t l/2.
location of the feed makes no difference to a total reflux column.
Unfortunately his eqn (30) specifies a condition upon all the liquid Now
leaving the feed stage, not just on the feed itself. Near the end of x1N = ] .O- tN = 3/4 = xLN.
Chien’s reply he asserts “The arbitrary eqn (30). . . is not needed
in the calculation of J and therefore will not atfect the value of J”. In terms of J as needed for Chien’s revised formulation, eqns
It is agreed that eqn (30)is not needed. It is a constraint, however,
(l)‘_(6)
and can affect the value of J as the following example
demonstrates. XLN-- XL’= 314.
The notation of Ref. (1) is used. The problem will be worked
according to the reformulated eqns (l)‘-(6)’ of Chien’s reply and The subscripts 1 and L will often be supressed henceforth. Using
then reworked accordina to esns (30) (l)-(6). (17)and (18)of Ref. line Y,
(1). In the first case the-calculated minimum number of stages is
exactly 2. In the second case the result is exactly 2.3, which being x ’ = h’ = l/4
greater than 2 cannot be the minimum number of stages. For this
problem y’=x’tl/4=1/2=x2
yz=x*t1/4=3/4=x“.
z,=z,_=l/3, zz=ZH=2/3, n=2
This computation is that specified by Chien’s (revised) eqn (3)‘.It
fL.-M = l/4, X”N= l/4. may be placed in that particular form by computing

Figure 1is the McCabe-Thiele diagram. The equilibrium curve is k’=y’/x’=2

k2 = y2/x2 = 312
I.0
and then

x’ = (3/2)(2)(1/4)= 3/4. (3)

This reformulation will be omitted henceforth. The stages


computed using eqn (3)’are shown by the solid-line steps in Fig.
1.
Next Chien’s eqn (30) is applied.
x
XLF Z‘ 1
-z==-=-
XH &I 2
so that as shown in Fig. 1

XLF=; (‘2,).

0 IV4 IL 314 IO
The dotted-line steps show the stages at total reflux under the
XL
restriction of eqn (30). The necessary computations follow.

xF= yF-’ = ]/3


the 3 straight lines labeled Y,, Y, and Y,. Straight lines are used
so calculations will be simple and exact. A smooth convex curve xF-’ = (l/2)yF-’ = l/6 (line Y,)
Could be used. YF=xF+1/4=7/12 (line YJ
Y, connects the points (0,O) and (l/4, l/2), and has the YF+’ = (]/2)xF+’ +1/2=7/24+12/24
equation y = 2x = ]9/24 = x F+Z. (line YJ
Y2 connects the points
_.. (l/4,1/2) and (l/2,3/4), and has the
equatton y = x + l/4 Using eqns (17) and (18) from Ref. [1] one calculates

You might also like