You are on page 1of 11

Food Control 22 (2011) 1322e1332

Contents lists available at ScienceDirect

Food Control
journal homepage: www.elsevier.com/locate/foodcont

Modelling methanol recovery in wine distillation stills with packing columns


J. Carvallo a, M. Labbe b, J.R. Pérez-Correa b, *, C. Zaror a, J. Wisniak c
a
Department of Chemical Engineering, Universidad de Concepción, Edmundo Larenas s/n, Casilla 160-C, Concepción, Chile
b
Department of Chemical and Bioprocess Engineering, School of Engineering, Pontificia Universidad Católica de Chile, Casilla 306, Santiago 22, Chile
c
Department of Chemical Engineering, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel

a r t i c l e i n f o a b s t r a c t

Article history: Methanol is a well known toxic congener that contaminates many spirits. Chilean legislation indicates
Received 2 May 2010 that methanol content in wine distillates should not exceed 1.5 g/L absolute alcohol (a. a.). To achieve this
Received in revised form stringent limit, distillers need new tools to improve their operational policies since trial and error
31 January 2011
experiments are too slow. In this paper, a fast pseudo-stationary simulator for batch distillations in
Accepted 9 February 2011
packed bed columns was developed. Packed bed columns are preferable for batch productions of fruit
wine distillations since they have a wider range of stable hydraulic operations. The model was solved
Keywords:
using orthogonal collocations and considers a liquid vapour equilibrium model of the ternary water-
Differential algebraic system
Dynamic simulation
ethanol-methanol mixture, which accounts for most of the volatiles in wine. Simulations compared
Mass transfer favourably with our own laboratory batch distillations of water-ethanol-methanol mixtures. In addition,
Orthogonal Collocations our model simulations were able to qualitatively reproduce the results reported in the literature for
Spirits methanol recovery in experimental batch distillations. Ethanol recovery errors were within measure-
Vapour-liquid equilibrium ment errors, while methanol simulations showed a 3% bias by the end of the distillation. Hence, the
developed model is accurate and could be used to explore optimal operational policies in order to
minimize methanol content in the heart cut used in the final product.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction simulations to explore many operational strategies, predict the


impact of changes in raw materials and design new control methods.
Pisco brandy is one of the most popular alcoholic beverages in Nevertheless, minimization of toxic compounds in Pisco has not
Chile, a young distillate from Muscat wine with a distinctive fruity received major attention yet. Indeed, methanol is believed to be
character produced since colonial times. During the last decade, associated with many discomforts experienced after alcohol inges-
several studies have been carried out to define and enhance the tion, such as fatigue, thirst, headache, stomachaches, nauseas, vomits,
aromatic quality of this beverage, improve process operation and hypersensitivity to light and sound, lack of concentration and atten-
reduce costs. Like other spirits, Pisco’s characteristics depend on raw tion, tremors, excessive perspiration and hypertension (Swift &
materials, fermentation process, distillation method and ageing. All Davidson, 1998). Methanol is slowly metabolized in the human
these will determine the content of minor volatile components, so- body, producing formaldehyde and formic acid, which are extremely
called congeners, in the distillate. Congeners can be favourable toxic at high concentrations. Hence, in many countries the methanol
(terpenes, esters and some high alcohols at very low concentrations), content of alcoholic beverages is controlled. In Chile, the concentra-
off-flavours (fatty acids), or even toxic (methanol, acetaldehyde and tion of methanol should be less than 1.5 g/L of absolute alcohol (a.a.).
furfural). In Pisco distillates, oenologists normally look for high Packed columns are best suited to distil wine in batch stills since
contents of terpenes and esters (that define the typical Muscat aroma) they have a wider range of stable hydraulic operations. Therefore,
and very low concentration of fatty acids, which are normally asso- operating policies with an extended range of reflux rates are
ciated with tails aroma. Specific distillation technologies have been possible, increasing the degrees of freedom during operation.
developed to efficiently and reproducibly obtain Pisco distillates with However, packed columns came after bubble caps; hence, the
a pre-defined aromatic characteristic. Mathematical modelling has conservative Chilean spirits industry still traditionally uses this
been crucial in this development, since process engineers can use older technology to distil Pisco. This research focuses on packed
column technology with the intended use for high volume opera-
* Corresponding author. Tel.: þ56 2 3544258; fax: þ56 2 3545803. tions for fruit distilled spirits, which is very relevant for the case of
E-mail address: perez@ing.puc.cl (J.R. Pérez-Correa). a fruit producing country such as Chile.

0956-7135/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodcont.2011.02.007
J. Carvallo et al. / Food Control 22 (2011) 1322e1332 1323

Sorel (1893) developed the first general model for plate distil- packed column (1 m height, 8 cm i.d.), a partial condenser and
lation columns, assuming that the streams leaving each stage were a total condenser, all thermally insulated. The boiler is equipped
in thermodynamic equilibrium. This idealization was then with two 2.4 kW electric heaters, the column is filled with glass
extended to model packed columns using the concept of HETP (750 mL) and copper (500 mL), 5 mm Raschig rings, and the partial
(Height equivalent to theoretical plate) that includes the influence condenser consists of a copper coil (6 mm o.d.) with 140 cm2
of type, size and packing material. In a real operation, this equi- external surface mounted above the packing column. The total
librium is not achieved; hence, it is necessary to incorporate some condenser was also a copper coil, and water at 25  C approx. was
kind of efficiency term, like those of Murphree (1925), Hausen used as cooling fluid.
(1953) or Standart (1965). However, using these efficiencies in The instrumentation and control system includes six Pt-100
modelling multi-component distillation columns can yield inac- sensors to measure temperature along the rectification column, in
curate predictions. A better option in these cases is the NEQ (non- the inlet and outlet cooling water streams in the partial condenser
equilibrium) or RB (rate-based) model, which assumes that both and in the boiler. Additionally, an L-Dens Anton-Paar transmitter
phases in a packing segment are not in equilibrium. Therefore, the measured on-line the alcoholic degree of the distillate, and a sole-
state of the outlet streams is the result of simultaneous mass and noid valve (Burkert G1/8) regulates the flow rate of the cooling water
energy transfer through the interface (Krishnamurthy & Taylor, in the partial condenser. All the instrumentation of the distillation
1985). The resulting model corresponds to a system of partial equipment is linked to an OPTO-22 data acquisition and control
differential equations (space and time derivatives) coupled with system attached to a Dell Latitude Pentium 3 portable computer. A
highly non-linear implicit algebraic equations. Frequently used graphical interface allows process monitoring, changing operating
methods to solve these types of systems are finite differences, finite variables like boiler power and partial condenser cooling, and
volumes and orthogonal collocations (Villadsen & Michelsen, 1978). applying control strategies.
The method of choice for a particular application will depend on the In this study, wine was simulated as a 12% v/v hydro-alcoholic
complexity of the problem, simplicity of implementation, numer- solution with low methanol concentrations. The solution was
ical solution stability, accuracy and computational cost. prepared in the boiler by loading 34.8 kg of distilled water, 5 L of
Few published works deal with modelling wine stills. Lora, ethanol (95% v/v) and 13.3 mL of methanol (90% v/v). The distillation
Iborra, Pérez & Carbonell (1992) and Berna, Bon, Sanjuán & Mulet process begins at total reflux. Once the column temperature stabi-
(1995) developed models for wine distillation, but in continuous lizes the collection of the head cut starts. After approx. 10 min, when
columns. Osorio, Pérez-Correa, Belancic & Agosin (2004) presented the alcoholic degree of the distillate reaches 74 % v/v, the collection of
a batch distillation dynamic model that has been applied to design the heart cuts begins. At this time, the boiler heating power is
variable reflux rate policies to increase terpenes and reduce fatty reduced to 735 W and the automatic control operation is turned on
acids in Pisco (Osorio, Pérez-Correa, Biegler & Agosin, 2005). The with a pre-set alcoholic degree of the distillate. The distillation run
model considers a plate column, as used in Chilean Pisco distill- finishes after 200 min. of operation, during which distillate cuts are
eries. More recently, Batista et al. (2009), developed steady state collected every 40 min. Samples are cooled and stored until HPLC
and dynamic distillation models to analyse the production of analysis is done for ethanol and methanol content.
cachaça. For example, they used ASPEN plus software to develop Samples were analyzed in duplicate using HPLC (Merck-Hitachi
a continuous distillation model to investigate the impact of 7350) with a refractive index detector and WSD software for data
degassing on the acetaldehyde concentration and on the ethanol acquisition. The column was a Bio-Rad Aminex HPX-87H (300 
loss. These authors concluded that degassing is a good option to 7.8 mm), the oven temperature was 55  C and the injection volume
control acetaldehyde only if its concentration in the product is was 20 mL. The mobile phase was a solution of pH ¼ 2.20 prepared
slightly above the legal limit, otherwise ethanol losses are too high. with concentrated sulphuric acid (95-97 %) in Milli-Q water; the flow
In addition, degassing also affects the concentration of other rate was 0.5 mL/min. All the samples and the mobile phase were
valuable volatiles such as ethyl acetate. Alternatively, if acetalde- filtered before analysis using cellulose acetate filters (MilliporeÒ)
hyde or other undesirable volatiles concentrations are large, an with a pore size of 0.45 mm.
additional rectification column is required. In this case, a by-
product stream is obtained, although it cannot be used for spirit
2.2. Model assumptions
production. Batista et al. (2009) also simulate the production of
cachaça in a traditional alembic using special purpose software,
A NEQ model has been developed here, since it is commonly
since it is hard to develop a stable dynamic multi-component
used to effectively model packed columns.1 Considering the main
distillation model using standard simulation packages.
phenomena and the specific characteristics of small packing
Within this context, the objective of this work is to develop an
columns, the following assumptions yield a good compromise
effective, numerically stable and efficient model to simulate the
between accuracy and efficiency:
batch distillation of the ternary water-ethanol-methanol mixture, to
explore operating strategies for reducing methanol content in wine
i) Wine has been modelled as a ternary mixture of ethanol,
distillates with minimum ethanol losses. In addition, the model
water and methanol, since the objective of the study is to
considers a packed column, since it should be a better option than
establish operating strategies to reduce methanol content in
a plate column for batch distillation of spirits. Simulations obtained
the distillate; the other congeners in wine do not affect much
with the proposed model have been compared with experimental
methanol relative volatility.
distillations reported in the literature (Glatthar, Senn & Pieper, 2001)
ii) Due to its low concentration, the influence of methanol in the
and with experimental data obtained in our laboratory.
physical properties of the ternary mixture and in the vapour-
liquid equilibrium (VLE) of ethanol-water mixture was
2. Materials and methods

2.1. Process description 1


Asprion 2006; Dribika & Sandall, 1979; Higler, Krishna & Taylor, 1999; Srivastava
& Joseph, 1984; Kayihan, Sandall & Mellichamp, 1975; Kreul, Gorak & Barton, 1999;
The experimental batch distillation system used in this Krishnamurthy & Taylor, 1985; Taylor, Kooijman and Hung, 1994; Schneider, Noeres,
research consists of a stainless steel boiler (50 L), a stainless steel Kreul & Gorak 2001.
1324 J. Carvallo et al. / Food Control 22 (2011) 1322e1332

disregarded. Hence, physical properties (density, enthalpy, 2.3.3. Balances in the packed column
viscosity and surface tension) of the ternary mixture were The column is assumed to operate under pseudo-steady state
obtained from ethanol-water correlations. conditions, therefore, total mass, global ethanol, ethanol in the
iii) Column dynamics was very fast compared with boiler liquid phase, global methanol, methanol in the liquid phase and
dynamics, therefore, time derivatives in the column balances energy balances are as follows:
were neglected (Krishnamurthy & Taylor, 1985).
iv) The column operated under atmospheric pressure, hence the vL vV
¼ (9)
vapour phase was assumed ideal. vZ vZ
v) Plug flow was assumed for vapour and liquid phases, there-
fore temperature and composition radial gradients were not vxe vye vL
L$  V$  ðye  xe Þ $ ¼ 0 (10)
considered (Hitch & Rousseau, 1988; Rejl, Linek, Moucha, vZ vZ vZ
Prokopová, Valenz & Hovorka, 2006).
L vxe xe vL  
vi) Pressure drop through the column is very small, and therefore
, þ $  f$kew $aef $S$rL $ xe  xIe ¼ 0 (11)
it was neglected (Hitch & Rousseau, 1988; Rejl et al., 2006; l vZ l vZ
Srivastava & Joseph, 1984). In addition, mechanical equilib-
rium is assumed in a differential packing element; conse- vxm vym vL
quently, vapour, liquid and interface are at the same pressure. L$  V$  ðym  xm Þ$ ¼ 0 (12)
vZ vZ vZ
vii) No heat transfer resistance in the vapour and liquid phases,
therefore, bulk vapour and liquid temperatures were equal to !1
the equilibrium temperature in the interface. Rejl et al. L vxm xm vL xe 1  xe
(2006), measured temperature differences lower than 1  C $ þ $  f$ þ
l vZ l vZ ke;m $aef kw;m $aef
between vapour and liquid phases in an ethanol-water  
packing column. $S$rL $ xm  xIm ¼ 0 ð13Þ

~ L vxe
vH ~ V vye  V
vH

2.3. Mass and energy balances L$ sat $  V sat $  H ~ L $ vL ¼ 0
~ H (14)
vxe vZ vye vZ vZ
2.3.1. Balances in the boiler
The boiler is assumed as an equilibrium stage, where the liquid 2.4. Constitutive equations
hold-up is in its bubble point. Hence, the following total mass,
ethanol, methanol, and energy balances can be written: 2.4.1. Ethanol-water VLE
This was modelled using NRTL since it showed the best
dMh performance among the standard activity coefficient models:
¼ Ln  Vn (1)
dt
!2
      eaw;e ,sw;e
Ln $ xen  xeh  Vn $ Ke xeh  xeh lnðge Þ ¼ sw;e $ I
dxeh xe þ xIw ,eaw;e ,sw;e
¼ (2) !2
dt Mh  2
eae;w ,se;w
þ se;w $ $ xIw ð15Þ
      xw þ xIe ,eae;w ,se;w
I
dxm L n $ xm m V $ K
n  xh n
e m  xm
m xh ; xh h
h
¼ (3)
dt Mh !2
eae;w ,se;w
  lnðgw Þ ¼ se;w $
~
d Mh $H xw þ xIe ,eae;w ,se;w
I
h L V
~  Vn $H
¼ Ln $H ~ þQ (4) !2
dt n n h
eae;w ,sw;e  2
þ swe $ $ xIe ð16Þ
xe þ xIw ,eae;w ,sw;e
I
Symbols are defined in Table 3.

2.3.2. Balances in the partial condenser To fit the NRTL model we need VLE data. The DECHEMA database
Here, the vapour leaving the column directly contacts the contains a very large number of papers containing VLE data for the
copper cooling coil, thus the condensate falls immediately into the binary systems water-ethanol and water-methanol. The database
packing. Consequently, there is no hold-up in the partial condenser, also reduces all these papers to a set of about 20, which they define
and the resulting total mass, ethanol, methanol and energy as recommended. Some of these sets were obtained at constant
balances are: temperature and others at constant pressure. Several of them are at
atmospheric pressure. The fact that several different recommended
sets of data are at the same pressure (or temperature) is just a result
V1  L1  VD ¼ 0 (5)
which shows that there is no definite test for determining the
  thermodynamic consistency of VLE data. We can accept a set of data
V1 $ye1  L1 $xe1  VD $Ke xe1 ¼ 0 (6) using the semi-empirical tests available, but these do not allow
ranking the “acceptable” sets. They can only be grouped as
 e m
V1 $ym m
1  L1 $x1  VD $Km x1 ; x1 ¼ 0 (7) “acceptable”. Therefore, in this research we applied a more stringent
procedure, where the binary VLE data were assessed for thermo-
V L V dynamic consistency applying the L-W Wisniak and the Freden-
~  L $H
V1 $H ~  VD $H
~ Q ¼ 0 (8)
1 1 1 D 1 slund tests using PRO-VLE 2.0 software.
J. Carvallo et al. / Food Control 22 (2011) 1322e1332 1325

We analyzed the VLE data of the ethanol-water binary mixture In the equilibrium, the methanol concentrations are:
found in several publications.2 For model fitting we selected only
the data reported by Carey & Lewis (1932), since this data satisfied
 
   
both consistency tests. Moreover, this data showed the best fit for P sat T xIe
yIm ¼ Km xIe ; xIm ¼ $gm xIe $xIm (20)
all the activity coefficient models tested, being NRTL the best P
with aw;e ¼ ae;w ¼ 0:5 ,se;w ¼ 0:2622 and sw;e ¼ 1:47149.
Additionally, the ELV model includes the following equation that   yIm
*
is satisfied in the bubble point: xIm ¼ Km xIe ; yIm ¼ 
     (21)
P sat T xIe P $gm xIe
P sat P sat
xIe $ge $ e þ xIw $gw $ w ¼ 1 (17)
P P TðxIe Þ is the equilibrium temperature assuming an ethanol-water
binary system. The saturation pressure was calculated using
2.4.2. Ethanol-water-methanol VLE Antoine with parameters given in Table 1:
Using the NRTL model, the activity coefficient of methanol is
given by:   B
log P sat ½mmHg ¼ A  (22)
T½ C þ C
se;m $Ge;m $xIe þ sw;m $Gw;m $xIw þ sm;m $Gm;m $xIm
lngm ¼ . 2.4.3. Mass transfer
Ge;m ,xIe þGw;m ,xIw þGm;m ,xIm
 Some well known correlations for packed columns were used to
xIe $Gm;e xIw $sw;e $Gw;e þxIm $sm;e $Gm;e calculate the volumetric mass transfer coefficients (Billet & Schultes,
.þ I sm;e 
xe þGw;e $xIw þGm;e $xIm xIe þxIw $Gw;e þxIm $Gm;e 1993; Bravo & Fair, 1982; Meirelles & Telis, 1994; Onda, Takeuchi &
 Okumoto, 1968); however, unfortunately, they give significantly
xIw $Gw;e xIe $se;w $Ge;w þxIm $sm;w $Gm;w
.þ s m;w  different values. The Meirelles & Telis (1994) correlation was
Ge;w $xIe þxIw þGm;w $xIm Ge;w $xIe þxIw þxIm $Gm;w selected since it was derived using a ternary hydro-alcoholic
 mixture. To simplify the numerical solution of the model, this
xIm $Gm;m xIe $se;m $Ge;m þxIw $sw;m $Gw;m
.þ s m;m  correlation was rewritten as:
Ge;m $xIe þGw;m $xIw þxIm Ge;m $xIe þxIw þGw;m þxIm
(18) !0:73    
mL rL $g 1:37 PML
kij aef ¼ CM $Dij $a2p $ $ $ $L0:69
rL ,Dij sL $ap ap $mL $S
with Gij ¼ eaij sij . To fit the additional model constants, aij ; sij ,
a number of articles with binary and ternary data3 were analyzed as (23)
described above. The data reported by Dunlop (1948) for water-
For glass Raschig rings CM ¼ 0:0029.
methanol and those by Kurihara et al. (1993) for ethanol-methanol
In this expression, kij represents the corresponding binary mass
passed both consistency tests and produced the best fit. Again, the
transfer coefficient. Given the low methanol concentration and
NRTL model was the best with am;w ¼ aw;m ¼ 0:3, ae;m ¼ am;w ¼ 0:4,
considering equimolar counter diffusion, ethanol mass transfer
sw;m ¼ 0:54681, se;m ¼ 0:98777, sm;w ¼ 0:69919, sm;e ¼ 0:13039.
reduces to:
None of the ternary data analyzed satisfied both consistency tests;
therefore they were not used in the fitting.  
Considering that methanol is highly diluted in wines, then NeL ¼ f$kew $aef $S$rL $ xIe  xe (24)
xIm z0 and xIw z1  xIe . Hence, Eq. (18) can be reduced to:
This equation includes a correction factor,f, for the effective area
  to include the effect of uneven liquid flow through the packing.
  se;m $Ge;m $xIe þ sw;m $Gw;m $ 1  xIe
lngm xIe ¼   . To model methanol mass transfer, the following relationship
Ge;m $xIe þ Gw;m $ 1  xIe applicable to diluted ternary systems was used (Krishna & Taylor,
   1993):
xIe $Gm;e 1  xIe $sw;e $Gw;e
.þ   sm;e  . (19) !1
xIe þ Gw;e $ 1  xIe xIe þ ð1  xe Þ$Gw;e  
xe 1 þ xe
  
L
Nm ¼ f$ þ $S$rL $ xIm  xm (25)
kem ,aef kwm ,aef
1  xIe $Gw;e xIe $se;w $Ge;w
.þ   sm;w   
Ge;w $xIe þ 1  xIe Ge;w $xIe þ 1  xIe
2.4.4. Physical properties
Given the low methanol concentration, it was assumed that
these properties were those of a binary ethanol-water mixture.
2
Bloom, Clump & Koeckert (1961), Carey & Lewis (1932), Dalager (1969), Jones, Liquid density (Murphy & Gaines, 1974):
Schoenborn & Colburn (1943), Kojima, Ochi, & Nakazawa (1969), Kurihara,
Nakamichi & Kojima (1993), Paul (1976), Rieder & Thompson (1949) and Yang
(2002).
3
Ethanol-methanol ELV data (Amer, Paxton & Van Winkle, 1956; Delzenne, 1958;
Kurihara et al., 1993), water-methanol data (Bredig & Bayer, 1927; Dalager, 1969; Table 1
Dunlop, 1948; Green & Venner, 1955; Hughes & Maloney, 1952; Khalfaoui, Meniai Antoine constants.
& Borja., 1997; Kurihara et al., 1993; Kohoutova, Suska, Novak, & Pick 1970;
Maripuri & Ratcliff, 1972; Olevsky & Golubev, 1956; Ocon & Rebolleda, 1958; Ocon Constant Ethanol Water Methanol
& Tabeada, 1959; Othmer & Benenati, 1945; Pascal & Garnier, 1921; Ramalho, A 8.11220 8.07131 8.08097
Tiller, James, & Bunch, 1961; Uchida, Ogawa, Hirata, Shimada, & Schimokawa, B 1592.8640 1730.630 1582.271
1982; Van Zandijcke & Verhoeye, 1974; Verhoeye & Schepper, 1973), and ternary C 226.184 233.426 239.726
data (Hughes & Maloney, 1952; Delzenne, 1958; Kurihara et al., 1993; Resa,
Goenaga, Gonzalez-Olmos & Iglesias, 2006). Gmehling, Onken & Arlt, 1982.
1326 J. Carvallo et al. / Food Control 22 (2011) 1322e1332

xe $PMe þ ð1  xe Þ$PMw
rL ¼   (26)
5:1214,102 þ 6:549,103 ,xe þ 7:406,105 ,T $xe þ ð1  xe Þ$PMw =rw

Liquid viscosity (Reid, Prausnitz & Sherwood, 1977): vxi X5


¼ Aij $xj (33a)
uw ue vZ j¼1
mL ue $me $e463, T uw $mw $e84$ T
¼ þ (27)
rL re rw
vyi X5
Liquid diffusivity (Reid et al., 1977): ¼ Aij $yj (33b)
vZ j¼1
ln DL ¼ 6:35 þ 3:72$xe þ 20:23$x2e  53:03$x3e þ 29:71$x4e þ
 vLi X5
9:05$103 þ 8:83,103 $xe þ 4:81,103 $x2e þ 21:96$103 $x3e ¼ Aij $Lj (33c)
vZ j¼1
1=T
9:23$exe (28)
vVi X5
Liquid surface tension (Reid et al., 1977): ¼ Aij $Vj (33d)
vZ j¼1
s3:9
L ¼ xe $s3:9
e þ xw $s3:9
w (29)
Coefficients Aij ¼ dlj ðZi Þ=dZi and collocation points were taken
Liquid and vapour enthalpy (Neuburg & Pérez-Correa, 1994): from Finlayson (1972). Equation balances at each collocation point,
Z1.n , could be written; however, since boundary conditions are the
~ L ðxe Þ ¼ ð55:678,xe þ75:425ÞT 15208:44$xe 20602:34
H (30) same in the extreme points of consecutive space intervals, only
balances in “n-1” points are necessary. Hence, the following set of
algebraic equations must be solved at each time interval:
~ V ðye Þ ¼ 36172:03  2919:83$ye þ ð31:461  11:976$ye ÞT
H
 
þ 4:063$104 þ 0:0734$ye T 2 ð31Þ
Table 3
List of symbols.

2.5. Numerical solution by orthogonal collocations Variables

Symbol Definition Units


Here, three interior collocation points were used, namely,
A Orthogonal collocation parameter e
Z2 ¼ 0:09, Z3 ¼ 0:41 and Z4 ¼ 0:79 , with Z1 ¼ 0 (condenser) and aef Effective interfacial area m2 =m3
Z5 ¼ 1 (boiler). The interior points correspond to the roots of ap Specific parking surface area m2 =m3
a Jacobi polynomial of order 3. The solution is approximated by CM Meirelles parameter e
Lagrange interpolation polynomials,li ðZÞ, where the interpolation D Fick diffusivity m2 =s
g Gravitational constant m=s2
points, Z, correspond to the collocation points. ~
H Specific enthalpy J=mol
K/K * Equilibrium constant e
X
5 L Molar liquid flow rate mol=s
xðZÞ ¼ li ðZÞ$xi (32a) k Mass transfer coefficient m=s
l Column length m
i¼1
lðzÞ; lðzÞ Lagrange polynomial e
M Liquid hold-up mol
X
5 N Mass transfer rate mol=m2
yðZÞ ¼ li ðZÞ$yi (32b) n Element number e
i¼1 PM Molecular weight g=mol
P Pressure mm Hg
Q Heat flow W
X
5 Q1 Condenser cooling power W
LðZÞ ¼ li ðZÞ$Li (32c) Qn Boiler heating power W
i¼1 R Reflux ratio e
S Column cross-sectional area m2
T Temperature C
X
5
t Time s
VðZÞ ¼ li ðZÞ$Vi (32d) V Molar vapour flow rate mol=s
i¼1 x Mole fraction, liquid phase e
y Mole fraction, vapour phase e
Differentiating these polynomials in the interpolation points,
Z Normalized axial axis m
results:
Greek letters
Table 2 ae;w NRTL parameter e
Operating conditions for laboratory distillations. f Effective area factor e
g Activity coefficient e
Variable Value m Viscosity kg=ðm,sÞ
Initial volume of the mixture in the boiler (L) 39.8  0.3 (0.8%) r Molar density mol=m3
Initial ethanol content in the boiler (% mol/mol) 4.05  0.05 (1.2%) s Surface tension kg=s2
Initial methanol content in the boiler (% mol/mol) 0.01459  0.00012 (0.8%) s NRTL parameter e
Nominal heating power applied (W) 735 6 Volume fraction e
J. Carvallo et al. / Food Control 22 (2011) 1322e1332 1327

values, and fitted the following explicit expressions using


X
5 X
5
TABLECURVEÔ (http://www.download.com/TableCurve-2D/3000-
Aij $Lj  Aij $Vj ¼ 0 (34)
2053_4-10308067.html):
j¼1 j¼1

( 1:54$104 þ 12:28$xe;I )
  xe;I < 0:045 ye;I ¼
ye;I ¼ Ke xe;I ¼ 1 þ 17:79$x e;I (40)
xe;I  0:045 ye;I
e;I
¼ 3:3 þ 0:29$ex þ 2:45$x0:5 xe;I
e;I þ 2:94$e

X X   
5 5   X5
Li $ Aij $xej  Vi $ Aij $yej  yei  xei $ Aij $Lj ¼ 0 (35) xe;I ¼ Ke* ye;I ¼ 132:88597:623$ye;I þ1023:946$y2e;I
j¼1 j¼1 j¼1 1 (41)
782:923$y3e;I þ224:735$y4e;I
X
5 X
5  m  X
5
Li $ Aij $xm
j  Vi $ Aij $ym m
j  yi  xi $ Aij $Lj ¼ 0 (36)  
j¼1 j¼1 j¼1 T ¼ Teq xe;I ¼ 1176:02  1288:55$xe;I þ 605:73$x1:5
e;I
(42)
x e;I
  X5   X5  19:13$x0:5
e;I  1076:03$e
1 1
$Li $ Aij $xej þ $xei $ Aij $Lj
l j¼1
l j¼1 (37)
  Then, the energy balance in the boiler was rewritten:
f$ke;w $aef $S$rL $ xei  xie;I ¼ 0
2  3
~L
vH vH~ L $ xe vxe
  X5   X5 n
¼ 4 n e n 5$ n (43)
1 1 vt vxn vt
$Li $ Aij $xm
j þ $x m
i $ Aij $Lj  f,
l l
j¼1 j¼1
and replaced in the mass balance, yielding:
!1
xei 1  xei  
þ $S$rL $ xm i  xi
m;I
¼ 0 ð38Þ   ~L
ke;m $aef kw;m $aef  
Ln $ H ~  vH $ xe  xe
~L  H þ Qh
n h n
vx e h
  5   Vn ¼  ~L   (44)
~ sat xe X
vH ~ sat ye
vH 
Li $ L e i $ Aij $xej  Vi $ V e i $ ~V  H
H ~  vH $ ye xe
n h n
vx vy vx e h
j¼1
  X (39)
X
5 5
Aij $yej ~V  H
 H ~L $ Aij $Lj ¼ 0 Since the boiler is in equilibrium, the liquid enthalpy corre-
i i
j¼1 j¼1 sponds to saturation. In addition, the derivatives of ethanol
enthalpies with respect to ethanol composition were calculated in
Orthogonal collocations transform each of the ordinary differ- advance and the following expressions were fitted using
ential equation (ODE) describing the column dynamics into nþ1 TABLECURVEÔ:

sat
~ ðxe Þ 
dH L xe < 0:04 4653654$x2e þ 498970$xe  18043
¼ xe (45)
dxe xe  0:04 138175:95 þ 174:03$exe þ 99432:62$x0:5
e þ 117527:95$e

sat
implicit algebraic equations. Then, these equations plus the set of ~ ðye Þ
dH
VLE equations yields 8n þ 16 implicit equations in the whole model.
V
¼ 250220$y6e  773755$y5e þ 875541$y4e
dye
Unlike explicit equations that can be evaluated directly, solving
implicit equations demands iterative methods, considerably  436590$y3e þ 96054$y2e  11536$ye þ 2195:8 ð46Þ
increasing computing costs. Therefore, reducing implicit equations
can significantly reduce the CPU time of the numerical solution. 3. Results and discussion
Bosley & Edgar (1994) proposed two methods to reduce implicit
equations: 3.1. Model calibration

i) Transform the implicit VLE set into explicit algebraic Two distillation runs, one at zero reflux (R0) and the other at
equations. total reflux (RT), were carried out to fit the model. Table 2 shows
ii) Rewrite the dynamic energy balances (although here this is operating conditions used in both runs.
only applicable to the boiler).
3.1.1. Distillation run at zero reflux (R0)
To reduce implicit equations, first the original VLE set was Since part of the boiler heat is lost into the environment, we
numerically solved for 100 given ethanol liquid concentration define the effective heating power as the energy used in vaporization
1328 J. Carvallo et al. / Food Control 22 (2011) 1322e1332

Fig. 1. Evolution of measured and simulated boiling temperature for different effective
heating powers, Qh, at zero reflux distillation. (- - -) 600 W; ( ) 550 W;(e e)
Fig. 3. Evolution of measured and simulated distillate flow rate for different effective
500 W; (oooo) meas.
cooling rates, Ql, at total reflux distillation. (e e) 370 W; ( ) 340 W; (- - -) 300 W;
(oooo) meas.

and liquid heating. This input variable was fitted by trial and error
to minimize the difference between the experimental and simu- 3.1.2. Distillation run at total reflux (RT)
lated values of the evolution of the boiling temperature of the Here, the same heat losses estimated above were considered.
mixture (Fig. 1). It is easily seen in the figure that with an effective Then, the measured distillate flow rate was used to estimate the
power of 550 W, the simulated boiling temperature follows exactly effective cooling rate in this case. Fig. 3 shows that with an effective
the measured values, except at the beginning of the distillation run. cooling rate of 340 W, the model can better represent the evolution
Nonetheless, this initial error is very small. Consequently, heat of the measured flow rate. Since the distillate flow rate is so small, it
losses can be estimated as 185 W. is extremely sensitive to fluctuations in the cooling water flow rate
In our model, the effective cooling rate includes the column heat and temperature.
losses and the heat removed by the cooling water in the partial Unlike reflux zero, in the case of total reflux the effective area
condenser. This input variable was also fitted by trial and error to affects the alcoholic strength of the distillate, as shown in Fig. 4. For
minimize the difference between the experimental and measured f ¼ 0:83 , the model predicts an average error of 2.3% in the
alcohol content evolution in the distillate (Fig. 2). Hence, in this case measured alcohol content in the distillate.
the effective cooling rate is 80 W, representing column losses. It can
be seen at the beginning of the distillation run that the model
3.2. Model validation
responds faster than measurements, which is partly due to the
inherent delay of the L-Dens and the accumulation of distillate
The model has been compared with experimental data reported
between the total condenser and the sensor location. The other
by Glatthar et al. (2001) dealing with methanol content in pear
observed deviations are probably due to unmeasured disturbances
distillates, and with our own experimental distillations of a water-
in cooling water flow rate and temperature.
ethanol-methanol mixture.

Fig. 2. Evolution of measured and simulated alcoholic content in the distillate for Fig. 4. Evolution of measured and simulated alcoholic content in the distillate for
different effective cooling rates, Ql, at zero reflux distillation. (e e) 120 W; ( ) different effective area factors, f , at total reflux distillation. ( ) 0.83; (e e) 0.76;
80 W; (- - -) 60 W; (oooo) meas. ( ) meas.
J. Carvallo et al. / Food Control 22 (2011) 1322e1332 1329

Fig. 5. Evolution of methanol concentration in the distillate for different reflux rates, R. (a) Data extracted from Glatthar et al. (2001): ( ) 4.4; ( ) 2.9; (e e) 1.3. (b)
Simulations with our model: ( ) 4.4; ( ) 2.5; (e e) 1.6; ( ) 0.7.

3.2.1. Comparison with experimental data from literature that the simulated values agree with data reported in Glatthar et al.
From the study of Glatthar et al. (2001) is not possible to extract (2001), although the simulated reflux rate used in simulations is
detailed information about the operating conditions of the exper- a bit smaller than the experimental value.
imental column, such as heating power, cooling rate, distillation Both, simulations and experiments, show that the absolute
time, condenser hold-up, thermal insulation, etc.. Hence, operation methanol concentration in the three cuts is practically the same,
variables and parameters needed for simulations were set by trial with a little bit higher concentration in the heart cut. However,
and error to mimic the qualitative behaviour of the experimental relative methanol concentrations are 50% higher in the tail cut, due
still. Despite these approximations, the model developed in this to ethanol depletion. This behaviour has also been observed by
study was able to reproduce the observed data. Apostolopoulou, Flouros, Demertzis & Akrida-Demertzi (2005) in
Fig. 5 shows that in simulations and experiments methanol distillations of a traditional Greek spirit.
distils throughout the whole run, independently of the reflux rate;
the simulated methanol concentration varies in the same range as 3.2.2. Validation with experimental data obtained in laboratory
the measured values. Moreover, by the end of the distillation, Fig. 8 compares the measured and simulated evolution of the
simulations reproduce the maximum in methanol concentration alcoholic degree of the distillate, under inlet cooling water
observed in experimental runs, which decreases more or less temperature disturbances. Simulations were performed with a full
sharply depending on the reflux rate. Simulations predict that with dynamic model (that considered accumulation in the packing) and
a small enough reflux rate (R ¼ 0.7 in Fig. 5), this maximum in the simplified pseudo-stationary model described above. It is seen
methanol concentration disappears. On the other hand, the that both simulation curves are practically identical, except for the
maximum observed at the beginning of the distillation is not first few minutes where the simplified model starts from a much
reproduced by the model. lower alcoholic content, which then rises sharply; a similar
Moreover, Fig. 6 shows that throughout the heart cut, the rela- behaviour was observed in Fig. 5. Overall, the dynamics of the
tive concentration of methanol (on ethanol) is almost constant and column is negligible compared to the dynamics of the boiler. In
does not depend much on the reflux rate. Furthermore, at the addition, the simulated responses closely follow the measured
beginning of the tail cut (900 mL of collected distillate) the relative values.
methanol concentration increases sharply as a result of the deple- In Fig. 9, simulations and measured values were compared for
tion of ethanol. a base case using conditions defined in Table 2 and an intermediate
Fig. 7 summarizes the absolute and relative methanol concen- cooling rate. The observed alcohol content and flow rate of the
trations in the different cuts of the distillation. Here, we can also see distillate are extremely oscillatory, due to its high sensitivity to

Fig. 6. Effect of reflux rate on methanol relative concentration to ethanol. (a) Data extracted from Glatthar et al. (2001): ( ) 4.4; ( ) 2.9; (e e) 1.3. (b) Simulations with our
model: ( ) 4.4; ( ) 1.6; (e e) 0.7.
1330 J. Carvallo et al. / Food Control 22 (2011) 1322e1332

Fig. 8. Evolution of distillate ethanol content in simulated and experimental ( )


distillations under inlet cooling water disturbances. Simulations were performed with
a full dynamic model (e e) and with a simplified pseudo-stationary model ( ).

Fig. 7. Absolute and relative methanol concentration in different cuts. (a) g MetOH/Ltot;
(b) g MetOH/LEtOH. Head cut. Heart cut. Tail cut. measurement error (Fig. 9d). In turn, the simulation error of the
methanol recovery at the end of the distillation (between 160 and
200 min), where methanol concentration in the boiler is extremely
cooling water disturbances. Despite this, the model effectively low, is 3% greater than the measurement error. The experimental
reproduces the behaviour of the distillation, with average errors of data used to fit the NRTL model were far from infinite dilution.
1.8, 1.4 and 2.0%, for the alcoholic content (Fig. 9a), boiling Consequently, the VLE model may not be accurate at such low
temperature (Fig. 9b) and distillate flow rate (Fig. 9c), respectively. methanol concentrations.
Moreover, the ethanol recovery simulation error is within the

Fig. 9. Model validation; simulated ( ) vs. experimental (oooo) distillation. (a) Distillate strength. (b) Boiler temperature. (c) Distillate flow rate. (d) Ethanol ( ) and methanol
(oooo) recoveries.
J. Carvallo et al. / Food Control 22 (2011) 1322e1332 1331

3.4. Strengths and limitations of the proposed model

The model presented here is reasonably accurate and efficient;


hence, it can be integrated to dynamic optimization routines to
establish operating strategies for the heart cut that minimize
methanol content without excessive ethanol and aroma losses.
Even though there are standard methods to reduce methanol
content in spirits to achieve legal limits, they require process
modifications and would increase losses of ethanol and other
valuable aromas. Additionally, since the only model inputs are wine
concentrations and operating procedure (partial condenser cooling
rate and boiler heat input), the same model can be used to improve
the operation of the still in different years, where only the wine
concentration changes.
Still, the model will benefit from further refinement. For
instance, it cannot reproduce well the distillation process during
the head cut (where acetaldehyde and relevant esters are distilled),
Fig. 10. Sensitivity of relevant production parameters to boiler heating power (Qh) and
since during this stage the rectification column dynamics is
condenser cooling rate (Ql). Heart/Tail cut time. Ethanol recovered. Distillate important and our model assumes a pseudo steady state. A fully
strength. dynamic orthogonal collocation model is too slow to be linked to
dynamic optimization routines (which require hundreds of simu-
lations). To improve the model performance during the head cut,
3.3. Sensitivity analysis we are trying alternative integration schemes such as the methods
of lines (Schiesser & Griffiths, 2009).
The effect of 20% variations on the main input variables In addition, more congeners should be added to the model,
(heating power and cooling rate) on the heart/tail cut time, the although this should be done with care given the scarcity of
recovered ethanol and the distillate strength are shown in Fig. 10. appropriate VLE data. For example, the methanol VLE model was
The heating power and cooling rate had a higher impact on the cut fitted with published data that mainly include measurements far
time, which varied between 30 and 55 %. The distillate strength was from the infinite dilution range; therefore, prediction of methanol
not affected much (around 5% change), while the collected volume recovery is not accurate by the end of the distillation run. Conse-
varied more than 20%. quently, we are starting an experimental research to obtain reliable
Next, the effect of physical properties on the simulated distillate VLE data in the infinite dilution region for congeners relevant to
volume and strength was analyzed. Fig. 11 summarizes the most Pisco brandy.
significant effects of these changes. Other changes (not shown) had
little impact (less than 2%) on distillation performance. The most 4. Conclusions
influential properties proved to be the vapour enthalpy, molar
density and VLE constant. The vapour enthalpy and density only A simplified model for the batch distillation in packed columns
affect the distillate volume, having almost no influence on the of the water-ethanol-methanol mixture was developed. Our results
alcoholic strength. In turn, the VLE constant has a moderate influ- show that this model could be used with confidence to design new
ence in both the strength and volume of the distillate. Considering heart cut distillation strategies to reduce the methanol content of
this analysis, the model can be simplified significantly assuming fruit wine distillates and other fresh spirits. Main simplifications
constant values for viscosity, diffusivity and surface tension, since included pseudo-stationary rectification column and pre-solution
their variations do not have much impact on the simulation results. of implicit algebraic equations, yielding fast and accurate simula-
tions. The model reproduced well the observed behaviour of
methanol in experimental pear brandy distillations reported in the
literature. Moreover, model predictions compare well with our
laboratory distillation experiments. For example, simulations of
boiling temperature, ethanol content and distillate flow rate, pre-
sented average errors smaller than 2%; for ethanol this is within
measurement error. Furthermore, methanol concentration errors
were within measurement error during most parts of the distilla-
tion, covering the head and heart cuts. Only during the tail cut
methanol simulations showed a bias of 3%.

Acknowledgements

J.C. is grateful to a M.Sc. scholarship granted by the School of


Graduates and the Labour Union N 2 of the Univ. de Concepción.
M.L. appreciates a postdoc grant from the School of Engineering at
P. Univ. Católica de Chile. J.W. thanks a Visiting Professor grant from
the School of Engineering at P. Univ. Católica de Chile. J.R.P.
appreciates the support of AGAUR from the Generalitat de Cata-
lunya through grant 2007PIV-00017 and P. Univ. Católica de Chile
Fig. 11. Sensitivity of relevant production parameters to vapour enthalpy (Hv), vapour for financial support for a sabbatical stay at Department D’Engi-
density and VLE constant (Keq). Ethanol recovered. Distillate strength. nyeria Química at Universitat Rovira i Virgili. Part of this work was
1332 J. Carvallo et al. / Food Control 22 (2011) 1322e1332

financially supported by Fondecyt project #1100357-2010. All Krishna, R., & Taylor, R. (1993). Multicomponent mass transfer (1st ed.). , New York:
Wiley & Sons.
authors are grateful to the editorial assistance of L. Gingles, My
Krishnamurthy, R., & Taylor, R. (1985). A nonequilibrium stage model of multi-
Personal Editor LLC. component separation processes. Part 1: model description and method of
solution. American Institute Chemical Engineering Journal, 31(3), 449e456.
Kurihara, K., Nakamichi, M., & Kojima, K. (1993). Isobaric vapor-liquid equilibria for
methanol þ ethanol þ water and the three constituent binary systems. Journal
References of Chemical Engineering Data, 38(3), 446e449.
Lora, J., Iborra, M. I., Pérez, R., & Carbonell, I. (1992). Simulación del Proceso de
Amer, H. H., Paxton, R. R., & Van Winkle, M. (1956). Methanol-Ethanol-Acetone. Destilación para la Concentración de Aromas del Vino. Revista Española de
Industrial Engineering Chemistry, 48(1), 142e146. Ciencia y Tecnología de Alimentos, 32(6), 621e633.
Apostolopoulou, A. A., Flouros, A. I., Demertzis, P. G., & Akrida-Demertzi, K. (2005). Maripuri, V. O., & Ratcliff, G. A. (1972). Vapor-liquid equilibrium data collection.
Differences in concentration of principal volatile constituents in traditional Frankfurt: DECHEMA.
Greek distillates. Food Control, 16(2), 157e164. Meirelles, A., & Telis, V. (1994). Mass transfer in extractive distillation of ethanol/
Asprion, N. (2006). Nonequilibrium rate-based simulation of reactive systems: water by packed columns. Journal of Chemical Engineering of Japan, 27(6),
simulation model, heat transfer, and influence of film discretization. Industrial 824e827.
& Engineering Chemistry Research, 45(6), 2054e2069. Murphree, E. V. (1925). Rectifying Column Calculations. Industrial & Engineering
Batista, F. R. M., Scanavini, H. F. A., Batista, E. A. C., Ceriani, R., Meirelles, A. J. A., & Chemistry, 17(7), 747e750.
Luz, L. F. L. (2009). Chapter 3. Distillation applied to the processing of spirits and Murphy, J. A., & Gaines, G. L. (1974). Density and viscosity of aqueous hydrogen
aromas. In M. A. A. Meireles (Ed.), Extracting bioactive compounds for food sulfide solutions at pressures to 20 atm. Journal of Chemical and Engineering
products, theory and applications. CRC Press. Data, 19(4), 359e362.
Berna, A., Bon, J., Sanjuán, N., & Mulet, A. (1995). Wine aroma concentration by Neuburg, H., & Pérez-Correa, R. (1994). Dynamic and steady state modelling of
using a general purpose simulator (ProSim). Food Science and Technology a pilot binary tray distillation column. Latin American Applied Research, 24,
International, 1(2e3), 117e127. 1e15.
Billet, R., & Schultes, M. (1993). Predicting mass transfer in packed columns. Ocon, J., & Rebolleda, F. (1958). Vapor-liquid equilibrium data collection. Frankfurt:
Chemical Engineering & Technology, 16(1), 1e9. DECHEMA.
Bloom, C. H., Clump, C. W., & Koeckert, A. H. (1961). Simultaneous Measurement of Ocon, J., & Tabeada, C. (1959). Vapor-liquid equilibrium data collection. Frankfurt:
Vapor-Liquid Equilibria and Latent Heats of Vaporization. Industrial & Engi- DECHEMA.
neering Chemistry, 53(10), 829e832. Olevsky, V. M., & Golubev, I. F. (1956). Vapor-liquid equilibrium data collection.
Bosley, J., & Edgar, T. (1994). An efficient dynamic model for distillation. Journal of Frankfurt: DECHEMA.
Process Control, 4(4), 195e204. Onda, K., Takeuchi, H., & Okumoto, H. (1968). Mass transfer coefficients between gas
Bravo, J., & Fair, J. (1982). Generalized correlation for mass transfer in packed and liquid phases in packed columns. Journal of Chemical Engineering of Japan, 1
distillation columns. Industrial & Engineering Chemistry, Process Development (1), 56e62.
and Design, 21(1), 162e170. Osorio, D., Pérez-Correa, R., Belancic, A., & Agosin, E. (2004). Rigorous dynamic
Bredig, G., & Bayer, R. (1927). Vapor-liquid equilibrium data collection. Frankfurt: modeling and simulation of wine distillations. Food Control, 15(7), 515e521.
DECHEMA. Osorio, D., Pérez-Correa, R., Biegler, L., & Agosin, E. (2005). Wine distillates: Practical
Carey, S., & Lewis, W. K. (1932). Studies in distillation. Industrial & Engineering operating recipes formulation for stills. Journal of Agricultural and Food Chem-
Chemistry, 24(8), 882e883. istry, 53(16), 6226e6331.
Dalager, P. (1969). Vapor-liquid equilibria of binary systems of water with methanol Othmer, D. F., & Benenati, R. F. (1945). Vapor-liquid equilibrium data collection.
and ethanol at extreme dilution of the alcohols. Journal of Chemical Engineering Frankfurt: DECHEMA.
Data, 14(3), 298e301. Pascal, P., & Garnier, D. E. (1921). Vapor-liquid equilibrium data collection. Frankfurt:
Delzenne, A. O. (1958). Vapor-Liquid equilibrium data for ternary system methanol- DECHEMA.
ethanol-water. Industrial & Engineering Chemistry, Chemical & Engineering Data Paul, R. N. (1976). Study of liquid-vapor equilibrium in improved equilibrium still.
Series, 3(2), 224e230. Journal of Chemical Engineering Data, 21(2), 165e169.
Dribika, M. M., & Sandall, O. C. (1979). Simultaneous heat and mass transfer for Ramalho, R. S., Tiller, F. M., James, W. J., & Bunch, D. W. (1961). Vapor-liquid equi-
multicomponent distillation in a wetted wall column. Chemical Engineering librium data collection. Frankfurt: DECHEMA.
Science, 34(5), 733e739. Reid, R., Prausnitz, J., & Sherwood, T. (1977). The properties of gases and liquids, their
Dunlop, J. G. (1948). Vapor-liquid equilibrium data collection. Frankfurt: DECHEMA. estimation and correlation. New York: McGraw Hill.
Finlayson, B. (1972). The method of weighted residuals and variational principles. New Rejl, J., Linek, V., Moucha, T., Prokopová, E., Valenz, L., & Hovorka, F. (2006). Vapour
York: Academic Press Inc. and liquid side volumentric mass transfer coefficients measured in distillation
Glatthar, J., Senn, T., & Pieper, J. (2001). Investigation on reducing the methanol column: comparison with data calculated from absorption correlations.
content in distilled spirits made of Bartlett pears. Deutsche lebensmittel run- Chemical Engineering Science, 61(18), 6096e6108.
dschau, 97(6), 209e216. Resa, J. M., Goenaga, J. M., Gonzalez-Olmos, R., & Iglesias, M. (2006). Measurement
Gmehling, J., Onken, U., & Arlt, W. (1982). Vapor-liquid equilibrium data collection. and modeling of phase equilibria for ethanol þ water þ methanol at isobaric
Frankfurt: DECHEMA. condition. Journal of Chemical Engineering Data, 51(6), 2114e2120.
Green, S. I., & Venner, R. E. (1955). Vapor-liquid equilibrium data collection. Frankfurt: Rieder, R. M., & Thompson, A. R. (1949). Vapor-Liquid equilibria measured by
DECHEMA. a Gillespie Still. Industrial & Engineering Chemistry, 41(12), 2905e2908.
Hausen, H. (1953). Zur Definition des Austauschgrades von Rektifizierböden bei Schiesser, W. E., & Griffiths, G. W. (2009). A compendium of partial differential
Zwei- und Dreistoff-Gemischen. Chemie Ingenieur Technik - CIT, 25(10), equation models: Method of lines analysis with Matlab. New York: Cambridge
595e597. University Press.
Higler, A., Krishna, R., & Taylor, R. (1999). Nonequilibrium cell model for packed Schneider, R., Noeres, C., Kreul, L., & Gorak, A. (2001). Dynamic modeling and
distillation columns-The influence of maldistribution. Industrial & Engineering simulation of reactive batch distillation. Computers and Chemical Engineering, 25
Chemistry Research, 38(10), 3988e3999. (1), 169e176.
Hitch, D., & Rousseau, R. (1988). Simulation of continuous contact separation Sorel, E. (1893). La rectification de lalcohol. Paris.
processes multicomponent batch distillation. Industrial & Engineering Chemistry Srivastava, R. K., & Joseph, B. (1984). Simulation of packed bed separation
Research, 27(8), 1466e1473. processes using orthogonal collocation. Computers and Chemical Engineering, 8
Hughes, H. E., & Maloney, J. O. (1952). The application of radioactive tracers to (1), 43e50.
diffusional operations. Binary and ternary equilibrium data. Chemical Engi- Standart, G. (1965). Studies on distillation-V: Generalized definition of a theoretical
neering Progress, 48, 192e200. plate or stage of contacting equipment. Chemical Engineering Science, 20(6),
Jones, C. A., Schoenborn, E. M., & Colburn, A. P. (1943). Equilibrium Still for Miscible 611e622.
Liquids. Industrial & Engineering Chemistry, 35(6), 666e672. Swift, R., & Davidson, D. (1998). Alcohol hangover, mechanics and mediators.
Kayihan, F., Sandall, O. C., & Mellichamp, D. A. (1975). Simultaneous heat and mass Alcohol Health and Research World, 22(1), 54e60.
transfer in binary distillation-I Theory. Chemical Engineering Science, 30(11), Taylor, R., Kooijman, H. A., & Hung, J. S. (1994). A second generation nonequilibrium
1333e1339. model for computer simulation of multicomponent separation processes.
Khalfaoui, B., Meniai, A. H., & Borja, R. (1997). Thermodynamic properties of Computers and Chemical Engineering, 18(3), 205e217.
water þ normal alcohols and vapor-liquid equilibria for binary systems of Uchida, S., Ogawa, S., Hirata, M., Shimada, Y., & Schimokawa, S. (1982). Vapor-liquid
methanol or 2-propanol with water. Fluid Phase Equilibria, 127(1e2), 181e190. equilibrium data collection. Frankfurt: DECHEMA.
Kohoutova, J., Suska, J., Novak, J. P., & Pick, J. (1970). Vapor-liquid equilibrium data Van Zandijcke, F., & Verhoeye, L. (1974). Vapor-liquid equilibrium data collection.
collection. Frankfurt: DECHEMA. Frankfurt: DECHEMA.
Kojima, K., Ochi, K., & Nakazawa, Y. (1969). Relationship between liquid activity Verhoeye, L., & Schepper, H. D. (1973). Vapor-liquid equilibrium data collection.
coefficient and composition for ternary systems. International Chemical Engi- Frankfurt: DECHEMA.
neering, 9, 342e347. Villadsen, J. V., & Michelsen, M. L. (1978). Solution of differential equations by poly-
Kreul, L., Gorak, A., & Barton, P. (1999). Dynamic rate-based model for multicom- nomials approximation. New Jersey: Prentice-Hall.
ponent batch distillation. American Institute Chemical Engineering Journal, 45(9), Yang, B. (2002). Vapor-liquid equilibrium for mixtures of water, alcohols, and
1953e1962. ethers. Journal of Chemical Engineering Data, 47(5), 1324e1329.

You might also like