Professional Documents
Culture Documents
Batch Distillation
CONTENTS
4.1 Early Theoretical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.1.1 Simple Distillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.2 Operating Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
C
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
er
Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
ia
l-
Batch distillation1 is the oldest separation process and the most widely used unit
Ta
feature of batch distillation is its flexibility. This flexibility allows one to deal with
o
In the distillation process, it is assumed that the vapor formed within a short pe-
riod is in thermodynamic equilibrium with the liquid. Hence, the vapor composition
Fr
form y = f (x). The exact relationship for a particular mixture may be obtained
nc
from a thermodynamic analysis and is also dependent upon temperature and pres-
is
sure. Figure 4.1 shows an example equilibrium curve for a system consisting of CS2
and CCl4 at 1 atmosphere pressure.
Simple distillation is the simplest form of batch distillation. In this type of dis-
tillation, a still is initially filled with a feed mixture, which evaporates after heating
and leaves the still in the vapor form. This vapor which is richer in the more volatile
component is collected in the condenser at the top. This simple distillation is often
referred to as Rayleigh distillation because of Rayleigh’s pioneering theoretical work
[35]. The concept of reflux and the use of accessories such as plates and packing ma-
terials to increase the mass transfer converts this simple still into a batch distillation
column as shown in Figure 4.2a. Because this batch column essentially performs the
rectifying operation, it is often called a batch rectifier .
The basic difference between batch distillation (rectifier) and continuous distilla-
41
42 Batch Processes: Modeling and Design
1 1
0.8 0.8
0.6 0.6
y
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 1
x
C
op
FIGURE 4.1
yr
Equilibrium curve for the CS2 and CCl4 mixture at 1 atmosphere pressure
ig
ht
ed
tion is that in continuous distillation (Figure 4.2b) the feed is continuously entering
the column, while in batch distillation the feed is charged into the reboiler at the
M
beginning of the operation. This feed is heated to form vapor. This vapor is passed
at
through packing or plates where it meets the liquid coming from top (refluxed back)
er
from the condenser at the top. Mass transfer takes place in contact and one gets a
ia
distillate which is richer in the more volatile component. The reboiler in batch dis-
l-
tillation gets depleted over time. The process has an unsteady-state nature. Reflux
ratio, which is a ratio of liquid reflux to distillate is an important operating parame-
Ta
ter on which operating condition depends. If reflux ratio is kept constant throughout
yl
the operation this results in variable product composition. On the other hand, in
o
order to keep the key composition constant, reflux ratio needs to be varied. There
r&
is a third policy of operation known as the optimal reflux policy that is neither the
constant reflux policy nor the variable reflux policy. Instead, this operating policy
Fr
exploits the difference between the two operating modes. Thus, the optimal reflux
a
policy is essentially a trade-off between the two operating modes, and is based on
nc
the ability of the process to yield the most profitable operation. Thus, there are
is
distillate distillate
rectifying
reflux reflux
section
rectifying section
feed
stripping
section
reboil
C
op
bottom
product
yr
reboiler reboiler
ig
ht
ed
FIGURE 4.2
Types of distillation processes: (a) batch distillation and (b) continuous dis-
M
tillation.
at
er
ia
l-
in which each charge consists of a fresh feed stock mixed with the recycled off-
specification material from the previous charge. Figure 4.3d represents a stripping
Ta
column for separating a heavy component as the bottom product where the liquid
yl
feed is initially charged into the top. In 1994, Davidyan et al. [37] presented a batch
o
distillation column that has both stripping and rectifying sections embedded in it
r&
(Figure 4.3e). This column is called the middle vessel column. Although this column
has not been investigated completely, recent studies demonstrated that it provides
Fr
added flexibility for the batch distillation operation. Recently [38] described a new
a
column configuration called a multivessel column (Figure 4.3f) and showed that the
nc
column can obtain purer products at the end of a total reflux operation. These
is
This section presents early theoretical analysis of simple distillation, which was first
analyzed by [35].
44 Batch Processes: Modeling and Design
2 3 2 3
C
op
1 1
yr
ig
FIGURE 4.3
Various configurations of batch distillation column.
Batch Distillation 45
mole fraction of the more volatile component A in the feed. Let B be the amount
op
still, and xD the mole fraction of component A in the vapor phase. The differential
ig
giving:
at
Z B Z xB
dB dxB
= , (4.2)
er
F B xF xD − xB
ia
or: Z xB
B dxB
l-
ln = . (4.3)
F xD − xB
Ta
xF
In this simple distillation process, it is assumed that the vapor formed within a
yl
short period is in thermodynamic equilibrium with the liquid. Hence, the vapor
o
of the form xD = f (xB ). The exact relationship for a particular mixture may be
obtained from a thermodynamic analysis depending on temperature and pressure.
Fr
For a system following the ideal behavior given by Raoult’s law, the equilibrium
a
relationship between the vapor composition y (or xD ) and liquid composition x (or
nc
composition is very low for a separation process, simple distillation cannot be used
op
for separation in real practice. To obtain products with high purity, multi-stage
yr
As stated earlier, the two basic modes of batch distillation are (1)constant reflux,
at
distillate composition of the key component, respectively. The third operating mode,
ia
optimal reflux or optimal control is the trade-off between the two operating modes.
l-
It is easy to understand the two basic modes using binary distillation. This
involves use of McCabe-Thiele graphical method described below.
Ta
yl
Although simple distillation marks the first analysis of batch distillation process, the
graphical analysis presented by [39] provided the basis for analyzing batch distillation
Fr
operating modes. The difference between simple distillation and batch distillation
a
operations is the relation between the distillate composition xD and the bottom
nc
Figure 4.4. If one assumes constant molal overflow (Lj = Lj−1 = .. = L0 = L and
Vj = Vj−1 = .. = V1 = V ), then McCabe-Theile’s method can be used.
In the McCabe-Thiele method, the overall material balance (with no liquid
holdup) on the plate is considered from the condenser to the j-th plate. This leads to
the following operating line equation (Equation 4.6). Note that in this procedure, we
use the theoretical number of stages or plates. If you divide the number of theoreti-
cal plates with the plate efficiency, we get the actual plates in a plate column, and
if you multiply these theoretical stages with the Height Equivalent to a Theoretical
Plate (HETP) then you get the actual height of a packed column.
R 1
yj = xj−1 + xD (4.6)
R+1 R+1
where R = L/ dD
dt
and V = (R + 1) dD
dt
.
This operating equation represents a line through the point yj (xj−1 = xD )
Batch Distillation 47
Condenser
V1
L0
1 dD/dt
xD
j-th
plate
Vj Lj-1
C
N
op
yj xj-1
yr
ig
y xj
ht
j+1
N+1
ed
Reboiler B, xB Vj+1 Lj
M
at
er
ia
FIGURE 4.4
l-
equilibrium curve to find the x1 , for this x1 , yj + 1 can be found out from the
operating line (vertical line, Equation 4.6). So the procedure is, starting from point
Fr
(xD , xD ), Equation 4.6 and the equilibrium curve (e.g., Equation 4.4) between yj
and xj can be recursively used from the top plate 1 to the reboiler (the reboiler can
a
nc
Smoker and Rose[40] presented the first analysis of the constant reflux operation
of a binary batch distillation with no holdup. They used the Rayleigh equation in
conjunction with the McCabe-Thiele graphical method to capture the dynamics of
the batch distillation column. In their procedure, the relationship between xD and
xB is recursively determined by the McCabe-Thiele graphical method. Then, the
right hand side of the Rayleigh equation (Equation 4.3) is integrated graphically
by plotting 1/(xD − xB ) versus xB . The area under the curve between the feed
composition xF and the still composition xB now gives the value of the integral,
which is ln (B/F ). The average composition of the distillate can be obtained from
48 Batch Processes: Modeling and Design
1
1
3 A
( xD xD )
5
2 ,
0.8
4
0.6 6
9
y
8
0.4
10
C
0.2
op
R
Slope =
yr
R+1
ig
0
xB
ht
x
M
at
FIGURE 4.5
er
McCabe-Thiele method
ia
l-
Ta
xD,avg = (4.7)
F −B
o r&
Although Smoker and Rose presented the calculation method independent of time,
the time can be introduced through the vapor boilup rate V of the reboiler. The
Fr
V R+1 R+1
nc
dB = dt ⇒ T = (F − B) = D (4.8)
R+1 V V
is
This operation policy is the easiest one and is commonly used. The following
example illustrates this policy.
Example 4.2: Consider the same mixture presented in Example 4.1 but use
batch distillation column with five theoretical plates and reflux ratio equal to 5.630.
a) Find the distillate and still composition when 40 percent of the mixture is distilled
similar to Example 4.1.
b) If the initial feed is 100 moles and vapor boilup rate is 143.1 moles/hr, what is the
total time required to complete the distillation operation?
Solution: The distillation operation requires 40 percent of the mixture to be
distilled, which corresponds to 60 percent of the original feed F remaining in the
reboiler, i.e., B = 0.6F . The problem is that of determining the distillate and still
Batch Distillation 49
composition, for which the Rayleigh equation (Equation 4.3) applies. To attain the
specified requirements, the left hand side of the Rayleigh equation can be calculated
as:
F 1
ln = ln = 0.5108
B 0.6
For various values of xD the operating lines are drawn and the bottom compo-
sition is calculated using the procedure shown in Figure 4.5. From trial and error,
we discover that the initial value of xD corresponding to the still composition of
xF = 0.4 is 0.9471. The rest of the xD are chosen below this composition, and
1
various values of xB , xD − xB , xD − xB
are obtained.
Rx
Values of xB versus xD − xB are plotted in Figure 4.6 from which x B xDdx−BxB
1
F
is obtained for each value of xB . The operation is stopped when the integral is equal
to 0.5108.
C
a) The distillate composition xD at the end of the operation is 0.5681, and the still
op
2.2
ht
ed
2.1
M
2
at
1/(xD - x B )
1.9
er
ia
1.8
l-
1.7
Ta
1.6
yl
o
1.5
r&
FIGURE 4.6
nc
Z D Z B
xD dD = xB dB
0 F
Z D
xDav dD = F xF − BxB
0
F xF − BxB
xDav =
D
100 × 0.4 − 60 × 0.1
xDav = = 0.8500
40
The average distillate composition is 0.85 for the key component which is much
larger than obtained in simple distillation.
50 Batch Processes: Modeling and Design
R+1
T = × D = 1.8532 hrs
V
Variable Reflux
In 1937, Bogart [41] presented the first analysis of the variable reflux policy for a
binary system. The steps involved in the calculation procedure for the variable reflux
mode are similar to those in the case of the constant reflux mode; however, in the
variable reflux case, the reflux ratio is varied instead of the distillate composition
at each step. Moreover, the Rayleigh equation, though valid for the variable reflux
condition, takes a simplified form. Since the distillate composition remains constant
C
(remember that we are considering binary systems here) throughout the operation,
op
B xD − xF
ig
= (4.9)
F xD − xB
ht
ed
The second step is to establish the relation between R and xB using the McCabe-
Thiele graphical method. Several values of R are selected, operating lines are drawn
M
through the fixed point (xD , xD ) with the slope of R/(R + 1), and steps are drawn
at
between the operating line and the equilibrium curve to get the bottom composition
er
(xB ). This recursive scheme is repeated until the desired stopping criteria is met,
ia
and thus B and xB can be found at each value of the reflux ratio. The time required
l-
for this operation at a given product purity is calculated by plotting the quantity
(R+1)
× F(x(xD−x
−xF )
2 versus xB in the following equation and then finding the area
Ta
V D F)
under the curve. Z xF
yl
R + 1 F (xD − xF )
T = dxB (4.10)
o
xB V (xD − xB )2
r&
Example 4.3: Rework the problem in Example 4.2 for the variable reflux mode.
Fr
a
Solution: Since the distillate composition is held constant throughout the vari-
nc
able reflux mode of operation, the distillate composition xD = xDav = 0.85. For
is
the various iterates of R, we obtain the corresponding values of xB . The value of the
amount of product distilled at each xB is also calculated using the Rayleigh equation
for the variable reflux condition (Equation 4.9).
xD − xF
D = F (1 − )
xD − xB
a) The operation is stopped when the amount of product collected is greater than or
equal to 40. The still composition at D = 40 is found to be equal to 0.10.
b) The time required for this operation is calculated by plotting the quantity
R+1 F (xD − xF )
V (xD − xB )2
versus xB and then finding the area under the curve between
xB equal to 0.4 (Figure 4.7) and 0.1. The time required is found to be 1.995 hrs.
Batch Distillation 51
30
25
20
dt / dx B 15
10
0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
xB
C
op
FIGURE 4.7
yr
The optimal reflux mode is a reflux profile that optimizes the given indices of column
at
performance chosen as the objectives. The indices used in practice generally include
er
the minimum batch time, maximum distillate, or maximum profit functions. This
ia
reflux policy is essentially a trade-off between the two operating modes, and is based
l-
on the ability to yield the most profitable operation from optimal performance. The
calculation of this policy is a difficult problem and relies on optimal control theory
Ta
consider the following reflux profile for the same separation given in Examples 4.2
o
and 4.3.
r&
Remember we are using the same batch and the same column for this operation.
nc
Since in this case neither distillate composition nor reflux ratio is constant, the
is
Z D Z B
xD dD = xB dB
0 F
Z D
xDav dD = F xF − BxB
0
F xF − BxB
xDav =
D
100 × 0.4 − 60 × 0.1
xDav = = 0.8500
40
The average distillate composition in this case is found to be 0.85, which is again
the same as that of the two operating modes.
C
Similarly, the time requirement for this operation is derived from the basic ma-
op
terial balance equations as was done in the case with the variable reflux condition.
The time T required when neither distillate composition nor reflux is constant is
yr
found to be
ig
ht
Z xF
B R+1
ed
T = dxB (4.11)
xB V xD − xB
M
The value of T is found to be 1.686 hrs, smaller than for the constant and variable
at
It can be seen that this reflux profile requires the least amount of time to obtain
ia
the same amount of product with the same purity. This policy is neither variable
l-
reflux nor constant reflux but is in between the two as shown in Figure 4.8. The
Ta
policy which was specified in the above example is the optimal policy obtained by
Coward in 1967[42].
yl
o r&
50 1
Fr
45 0.95
Distillate Composition
40 0.9
a nc
35 0.85
Variable reflux
Reflux Ratio
30 Constant reflux
is
0.8
Optimal reflux
25 0.75
Variable reflux
20 0.7 Constant reflux
Optimal reflux
15 0.65
10 0.6
5 0.55
0 0.5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
FIGURE 4.8
The three operating modes
Batch Distillation 53
shortcut model[45].
op
yr
along with the reboiler and condenser dynamics. Distefano presented detailed anal-
ysis of the characteristics of differential mass and energy balances associated with
M
work forms the basis for almost all of the later work on rigorous modeling of batch
er
For an arbitrary plate j (Figure 4.4), the total mass, component, and energy
l-
balances yield the governing equations are summarized in Table 4.1. This table
lists all the equations involved in the dynamic analysis of the batch column and
Ta
the assumptions behind these equations. Note that energy balance equations are
yl
difference equations instead of differential equations, this is due to the fact that
o
severe transients. Stiff equation solver is necessary to solve these type of equations.
a
The stiffness of the system is reduced considerably when one considers zero plate
nc
holdup. This results in semirigorous model for batch distillation. This model is sim-
is
ilar to what was used earlier with McCabe-Theile method (except with additional
energy balance equations whenever necessary).
TABLE 4.1
The complete column dynamics
Assumptions
• Negligible vapor holdup,
• Adiabatic operation,
• Theoretical plates,
• Constant molar holdup,
• Finite difference approximations for the enthalpy changes.
Composition Calculations
C
(i)
dxD V1 (i) (i)
dt = HD (y1 − xD ), i = 1, 2, . . . , n
yr
Plate Dynamics
ig
(i)
dxj 1 (i) (i) (i) (i)
ht
i = 1, 2, . . . , n; j = 1, 2, . . . , N
Reboiler Dynamics
M
(i)
dxB 1 (i) (i) (i) (i)
= − xB ) − VB (yB − xB )], i = 1, 2, . . . , n
B [LN (xN
at
dt
Flowrate Calculations
er
L0 = R dD dt ; V1 = (R + 1) dt
dD
l-
On the Plates
Ta
Lj = Vj+1 + Lj−1 − Vj ; j = 1, 2, . . . , N
Vj+1 = Jj+11− Ij [Vj (Jj − Ij ) + Lj−1 (Ij − Ij−1 ) + Hj δIj ]
yl
o
j = 1, 2, . . . , N
r&
Condenser Duty
nc
QD = V1 (J1 − ID ) − HD δt ID
is
Reboiler Duty
QB = VB (JB − IB ) − LN (IN − IB ) + Bδt IB
Thermodynamics Models
Equilibrium Relations
(i) (k)
yj = f ((xj , k = 1, . . . , n), T Ej , Pj )
Enthalpy Calculations
(k)
Ij = f ((xj , j = 1, . . . , n), T Ej , Pj )
(k)
Jj = f ((yj , j = 1, . . . , n), T Ej , Pj )
Batch Distillation 55
and δt Ij ) in Table 4.1 can be assumed to be zero. These equations can be integrated
using Runge-Kutta method. This results in a zero holdup model, so this approach
can be used for simulating the semirigorous model of batch distillation.
dB V
= − , B0 = F (4.12)
dt R + 1
(k)
dxB V (k) (k) (k) (k)
= (x − xD ), xB0 = xF (4.13)
dt (R + 1)B B
The holdup effects can be neglected in a number of cases where this model ap-
proximates the column behavior accurately. This model provides a close approxima-
tion to the Rayleigh equation, and for complex systems (e.g., azeotropic systems) the
synthesis procedures can be easily derived based on the simple distillation residue
curve maps (trajectories of composition). However, note that this model involves
C
The rigorous model of batch distillation operation involves a solution of several stiff
er
differential equations and the semirigorous model involves a set of highly nonlinear
ia
equations. The computational intensity and memory requirement of the problem in-
l-
crease with an increase in the number of plates and components. The computational
complexity associated with these models does not allow us to derive global proper-
Ta
ties such as feasible regions of operation, which are critical for optimization, optimal
yl
control, and synthesis problems. Even if such information is available, the compu-
o
tational costs of optimization, optimal control, or synthesis using these models are
r&
prohibitive. One way to deal with these problems associated with these models is to
develop simplified models such as the shortcut model.
Fr
The shortcut model of batch distillation proposed by [45] is based on the assump-
a
tion that the batch distillation column can be considered equivalent to a continuous
nc
distillation column with changing feed at any instant. Since continuous distillation
is
TABLE 4.2
Time implicit model equations for the shortcut method
(i)
op
Pn
i=1 xD = 1
(k)
yr
C Estimation xD estimation
1 C1 (k)
ig
Pn αi xD (i) (k) 1
(k) xB = 1 xD =
ht
i=1 αk xB Pn
αi
C
1 xB
(i)
i=1 αk (k)
ed
x
B
Fenske Equation
M
Nmin ≈ C1
at
Underwood Equations
er
Pn αi x(i) Pn αi xD
(i)
B
i=1 αi −φ = 0 ; Rmin u +1= i=1 αi −φ
ia
Gilliland Correlation
l-
R Estimation C1 Estimation
Ta
h i
(1 + 54.4X)(X − 1)
Y = 1 − exp (11 + 117.2X)√X
yl
R − Rmin N − Nmin
X = R + 1 ; Y = N + 1
o
R Estimation
r&
R = F (M inimum H)
Fr
a nc
satisfied. The complete simulation of the shortcut method for the different operating
is
Start
Initialization
(i)
xF = x(i)
B
i = 2,3,...,n
C
op
(k)
Initialize: B and x B
yr
ig
Assume C1
ed
Hengestebeck-Geddes Eqn
M
(i)
Calculate x D i=2,3,...,n
at
New C1 (k)
l-
Calculate x
D
FUG Eqns
Ta
Calculate x B , i=2,3,...,n
o
Optimal Reflux
r&
Calculate R
Fr
Integration
Stopping
a
(k) Criteria
nc
YES
Stop
FIGURE 4.9
The flowchart for the shortcut method
58 Batch Processes: Modeling and Design
effect and highly non-ideal systems, the order of rigorous model can be reduced using
collocation approach.
desired one. However, the optimal control problems we consider here do not fall under
ig
this definition of control. Because the decision variables that will result in optimal
ht
performance are time-dependent, the control problems described here are referred to
ed
as optimal control problems. Thus, use of the control function here provides an open-
loop control. The dynamic nature of these decision variables makes these problems
M
much more difficult to solve as compared to normal optimization where the decision
at
concentration for a specified time is maximized [47, 48, 49, 50, 51]. This problem
yl
Z T Z T
dD V
max J= dt = dt, (4.14)
dt Rt + 1
Fr
Rt 0 0
a
[47] were the first to report the maximum distillate problem for binary batch
is
distillation, which was solved using Pontryagin’s maximum principle, the dynamic
programming method, and the calculus of variations. These methods are described
in Chapter 14. [48] extended this optimization model to multicomponent systems
and used the shortcut batch distillation model along with the maximum principle
for the calculation of the optimal reflux policy. [49] used the orthogonal colloca-
tion approach on finite elements and nonlinear programming (NLP) optimization
techniques over the shortcut model. Further, they extended this method to the rig-
orous batch distillation model [51] in which they considered the effect of column
holdups on optimal control policy.
• Minimum Time Problem – where the batch time needed to produce a pre-
scribed amount of distillate of a specified concentration is minimized [42, 52].
Although there are several different formulations for the minimum time problem,
[52] derived the following formulations to establish a unified theory for all the
Batch Distillation 59
[48] used a different objective function to solve the profit maximization problem
under the constant and variable reflux conditions. [49] formulated a new profit
yr
function and solved the differential algebraic optimization problem for optimal
ig
design and operation. [54] developed a detailed dynamic multifraction batch dis-
ht
tillation model and discretized the model using the orthogonal collocation method
ed
on finite elements, and finally solved the maximum profit model using a NLP op-
M
timizer. [55] considered a rigorous reactive distillation system for the maximum
conversion problem, which can also be classified as the maximum profit problem.
at
The detailed dynamic system is then reduced by using polynomial curve fitting
er
A variant of this objective function is to minimize the mean rate of energy con-
l-
sumption when the market size for the product is fixed by the current demand.
Ta
RT
QR (t)dt
o
min J = 0 , (4.17)
r&
T + ts
∗
s.t. xD,avg ≥ x ,
Fr
D ≥ D∗ ,
a
where QR is the reboiler heat duty. They used this objective function for optimal
nc
control of multivessel columns for the first time. [57] also presented the optimal
is
0.8
C
op
0.6
ethanol
yr
ig
y
0.4
ht
ed
0.2
M
at
0
0 0.2 0.4 0.6 0.8 1
er
ia
x ethanol
l-
FIGURE 4.10
Ta
the vapor-liquid equilibrium data from liquid solution models and their application
a
to distillation design. However, only during the past two decades has there been a
nc
binary distillation, the barrier is represented by the azeotropic points and in case of
a ternary system, they are represented by separatrices. Based on this information,
qualitatively similar ideal systems can be represented for these non-ideal azeotropic
systems[58, 59]. Doherty & coworkers [60, 61] in their pioneering works proposed
several new concepts in azeotropic distillation. They established the use of ternary
diagrams and residue curve maps in the design and synthesis of azeotropic continuous
distillation columns. In batch distillation, they outlined a synthesis procedure based
on the residue curve maps.
The residue curve map graphs the liquid composition paths that are solutions
to the following set of ordinary differential equations:
dxi
= x i − yi i = 1, 2, . . . , n − 1, (4.18)
dξ
where n is the number of components in the system, and the independent variable,
Batch Distillation 61
0.2
0.8
0.4
0.6
0.6
0.4
0.8
0.2
C
op
1.0
yr
0.0
H2O (100 oC) ACN (81.6 oC)
ig
FIGURE 4.11
M
warped time (ξ), is a monotonically increasing quantity related to real time. One
l-
can see that Equation 4.18 is one form of the Rayleigh equation described earlier.
Ta
The residue curve map occupies a significant place in the conceptual design stage
of column sequencing in continuous distillation, and fractions (cuts) sequencing in
yl
reaction and separation since the 1980s. For example, most of the new commercial
processes of MTBE (methyl-tertiary-butyl ether, an anti-knocking agent) are based
on continuous reactive distillation technologies.
The analysis of a reactive batch distillation model in a staged column was first
published by [63]. Using a stiff integrator for the differential and algebraic equations,
they presented a numerical solution technique for the esterification of 1-propanol
and acetic acid. They argued that in reactive distillation differential energy bal-
ances should be included in the solution. [64] developed a new solution technique
based on the orthogonal collocation method on the finite element method for the
reactive batch distillation of a packed column. The differential contactor model of
a packed column, originally designed by [65], was then reduced to the low-order
polynomials with desired accuracy. They compared the results with those from the
finite difference method and global collocation method for non-reactive packed-bed
batch distillation systems, and showed that their approach was more efficient. [66]
C
extended their previous work to the optimal campaign structure for reactive batch
op
distillation, which can offer reasonably sharp separations between successive cuts
yr
and reduce the amount of waste off-cuts. To obtain the optimal reflux policies or
ig
profiles for the maximum distillate or minimum time problem, multiperiod reflux op-
ht
timization [67] can be applied. They showed that for the same production rate, the
ed
waste generation can be significantly reduced under the optimal campaign structure.
An efficient optimization approach for reactive batch distillation using polyno-
M
mial curve fitting techniques was presented by [55]. After finding the optimal solution
at
of the maximum conversion problem, polynomial curve fitting techniques were ap-
er
plied over these solutions, resulting in a nonlinear algebraic maximum profit problem
ia
that can be efficiently solved by a standard NLP technique. Four parameters in the
l-
profit function, which are maximum conversion, optimum distillate, optimum reflux
ratio, and total reboiler heat load, were then represented by polynomials in terms
Ta
of batch time. This algebraic representation of the optimal solution can be used for
yl
A dynamic rate-based model for packed-bed batch distillation was recently pro-
r&
posed [68], in which a solid catalyst was used first in the reactive batch distillation
modeling. The pilot-scale experiments were conducted with strong anion-exchange
Fr
resins. The results were compared with the experimental data and with the results
a
from its counterpart, the equilibrium-based model. The rate-based model provides
nc
more accuracy, much higher physical significance, and more predicability of the ex-
is
perimental data even though the formulation of the rate-based model is complicated.
art in batch distillation techniques and computer simulation technology, Table 4.3
identifies the required functionality and the rationale behind it.
There are several commercial software packages for simulations, optimizations,
and/or optimal controls of batch distillation (see Table 4.3). These include Bdist-
SimOPT (Batch Process Technologies), BatchSim(Simulation Sciences), BatchFrac
(Aspen Technology, based on [69]), and MultiBatchDS (Batch Process Research
Company). Bdist-SimOPT and MultiBatchDS are derived from the academic pack-
age BATCHDIST [45]. Most of these packages except MultiBatchDS are usually
limited to conventional systems as they were developed in early or late, 80s.
C
4.7 Summary
op
This chapter presented analysis of batch distillation beginning from the first the-
yr
oretical analysis in 1902. There are three operating modes of conventional batch
ig
distillation column, namely, constant reflux and variable product composition, vari-
ht
able reflux and constant product composition of the key component, and optimal
ed
reflux policy. There is a hierarchy of models available for batch distillation. These in-
clude shortcut model for feasibility, design, optimization, and synthesis, semirigorous
M
model with negligible holdup, rigorous model including all transients with holdup
at
and reduced order model based on collocation method. There are numbers of soft-
er
ware packages available for batch distillation design, simulation, and optimization.
ia
l-
Ta
yl
o
Notations
r&
Fr
dB
bottom product flow rate or change of bottom product [mol/time]
nc
dt
C1 constant in the Hengstebeck–Geddes equation for conventional batch column
is
q′ ratio of the top vapor flow rate to the bottom vapor flow rate
QR reboiler heat duty [mass length2 time−2 ]
R reflux ratio (= L/D)
Rmin g minimum reflux ratio given by the Gilliland correlation
Rmin u minimum reflux ratio given by the Underwood equations
Rt reflux ratio as a function of time
T batch time [time]
Vj vapor stream leaving plate j [mol/time]
x liquid-phase mole fraction
xB mole fraction of liquid in the reboiler
xD mole fraction of the distillate
xD,avg average distillate mole fraction
xF liquid-phase mole fraction of the feed
y vapor-phase mole fraction
C
op
Greek Letters:
yr
α relative volatility
ig
ht
ed
M
at
er
ia
l-
Ta
yl
o r&
Fr
a nc
is
TABLE 4.3
Comparison of software packages
Features CHEMCAD BATCH BATCHSEP MultiBatchDS
C
Databank CHEMCAD ASPEN PLUS CRANIUM
op
Batch Distillation
Operations
yr
Constant Reflux Yes Yes Yes
ig
Variable Reflux No Yes Yes
Optimal Reflux No No Yes
ht
Optimal Reflux-Fixed Equation No Yes Yes
ed
Models
M
Shortcut No No Yes
Low Holdup Rigorous/Semirigorous Yes No Yes
at
Reduced Order No No Yes
er
ia
Rigorous Yes Yes Yes
Configurations
l-
Rectifier Yes Yes Yes
Ta
Semi-batch No Yes Yes
Recycle waste cut No Yes Yes
yl
o
Stripper No No Yes
r&
Middle Vessel Column No No Yes
Options
Fr
Design Feasibility No No Yes
a
Optimization No Yes Yes
nc
Reactive Distillation No Yes Yes
is
3 phase Distillation Yes Yes Yes
Uncertainty Analysis No No Yes
65
is
anc
Fr
or&
yl
Ta
l-
ia
er
at
M
ed
ht
ig
yr
op
C