Você está na página 1de 11

02638762/98/$10.00+0.

00
q Institution of Chemical Engineers
Trans IChemE, Vol 76, Part A, March 1998

OPERATION AND CONTROL OF DIVIDING WALL


DISTILLATION COLUMNS
Part 1: Degrees of Freedom and Dynamic Simulation
M. I. ABDUL MUTALIB and R. SMITH (FELLOW)
Department of Process Integration, UMIST, Manchester, UK

T
he dividing wall distillation column has been known now for some 50 years. Despite its
potential to make major savings in energy and capital costs in distillation, it has not been
widely used in practice. One of the major fears in applying the technology is uncertainty
regarding the control and operation of the arrangement. This paper investigates theoretically
the operation and control of the dividing wall column. A degrees of freedom analysis was
performed to determine the number of control loops required. Possible control con gurations
were then investigated using Relative Gain Array Analysis and dynamic simulation. The
results of these theoretical studies indicate that simple control schemes are capable of
providing stable control.
Keywords: dividing wall distillation; thermal coupling; dynamic simulation; control; pilot plant

Rudd1 2 , Cerda and Westerberg1 3 , Spadoni and Stramigioli1 4 ,


INTRODUCTION Nikolaides and Malone1 5 ). Recently, a design procedure
Distillation remains the most important method used in the which allows for optimization of the column has been
chemical industry for the separation of homogeneous proposed by Triantafyllou and Smith1 6 . By contrast, studies
mixtures, with the amount of energy used in distillation on the operational and control aspects of the Petyluk
operations being considerable. Appropriate integration of con guration have received little attention in the past.
the distillation column with the overall process can result Chavez et al.1 7 and Lin et al.1 8 reported multiple steady
in signi cant energy savings (Linnhoff et al.1 , Smith and state solutions for the Petyluk con guration. Using computer
Linnhoff 2 ) but the scope for this is often limited. Other simulation, they presented four different solutions at a
options involve the use of complex distillation arrangements speci ed re ux which have different internal liquid and
such as the side-stripper, the side-recti er or the fully vapour ows between the prefractionator and the main
thermally coupled (Petyluk) con guration. Such complex column. As they reduced the re ux ratio to a value beyond
arrangements can consume signi cantly less energy when which no feasible solution existed, a unique solution was
compared to a conventional arrangement. So far, the use found. This is the optimum combination of the internal
of complex arrangements has largely been limited to crude liquid and vapour ows which gives the minimum energy
oil distillation where the side stripper arrangement has been requirement. Wolff et al.1 9 , 2 0 performed control studies
used extensively. on the Petyluk con guration using a three point and four
The Petyluk con guration (Figure 1a) was initially point composition control. For the three point composition
introduced some 50 years ago (Brugma3 ). Theoretical control, they set up a control con guration which maintained
studies on a stand alone basis (Petyluk et al.4 , Fidskowski the composition of the three main products of the column.
and Krolikowski5 , Glinos and Malone6 and Kaibel7 ) have Using one of the possible control schemes, they were able to
shown that it is capable of achieving typically 30% of achieve satisfactory control performance, given feed ( ow-
energy savings compared with a conventional sequence. rate and composition) and set point disturbances. For the
In addition, the Petyluk arrangement can also be achieved four point composition control, they used the internal liquid
by placing a vertical wall in the middle of the column split between the prefractionator and the main column to
(Figure 1b), separating the feed from the side draw (Wright8 , control the impurity ratio in the side draw as an additional
Kaibel9 ). Thus an overall reduction in capital cost can be control loop. They discovered that a problem can occur
expected through the elimination of a column shell, reboiler within a range of the internal liquid splits whereby the
and condenser when compared with a conventional arrange- product speci cations cannot be achieved. A similar result
ment. Despite these advantages, industry has been reluctant was observed when the vapour split was used in place of
to use the Petyluk and dividing wall con gurations. This can the liquid split. Morud and Skogestaad2 1 later provided an
largely be attributed to the lack of established design explanation for this using three dimensional plots display-
procedures and the fear of control problems. ing the variation of the reboiler duty and side draw
The design of the Petyluk con guration has been studied by impurity ratio against changes in the internal liquid and
many researchers (Stupin1 0 , Fonyo et al.1 1 , Tedder and vapour splits.
308
OPERATION AND CONTROL OF DIVIDING WALL DISTILLATION COLUMNS: PART 1 309

Figure 1. Fully thermally coupled columns. (a) Petyluk column; (b) Dividing wall column.

In this paper, the complexity of the dividing wall column mixed pool with the liquid leaving having the same
will rstly be assessed by performing degrees of freedom properties as the liquid on the stage, the number of
analysis. The variables leading to the additional complexity relationships is equal to 3NC + 4. These relationships
of the column, when compared to the more established consist of a total material balance, an energy balance,
side draw column, will then be analysed to investigate the NC-1 component balances, NC distribution relation-
impact on the operation and control of the column. Dynamic ships between the vapour and liquid phase, i.e. vapour-
simulation will also be used to investigate the controll- liquid equilibrium, NC-1 concentration identities for liquid
ability of the column from a theoretical point of view. Part 2 leaving and liquid on the stage, two temperature identi-
of this paper will discuss pilot plant control studies on such ties and two pressure identities for the vapour and liquid
a column. leaving and liquid on the stage. Hence the degrees of
freedom for the unit are:
DEGREES OF FREEDOM ANALYSIS
22
No. of DOF = No. of variables - No. of
A method developed by Howard for analysing the relationships
degrees of freedom at unsteady state condition has been
used for the dividing wall column. The method is based = 5NC + 11 - 3NC - 4 = 2NC + 7
on the fact that the degrees of freedom for a system is equal The degrees of freedom for a cascade of N stages can then
to the sum of degrees of freedom of all the units in the be easily determined by treating the cascade as a system.
system minus the sum of degrees of freedom for all Since a cascade of N stages are interconnected by 2N-2
the interconnecting streams. This requires the system to be streams, the degrees of freedom are:
torn apart into smaller units with inter-stream connec-

cascade of N units= for N units - 2N - 2 streams


tions. When analysing the degrees of freedom for any of No. of DOF for Sum of DOF Sum of DOF
the units, hold up is included to account for the unsteady
state. The way the method handles hold up is by treating
it as a quantity stream with the same variables as the = N(2NC + 7)- 2(N - 1)(NC + 2)
interconnecting stream, except that a hold up quantity is
used instead of owrate.
= 3N + 2NC + 4
In a distillation column, the units consist of stages, An extra degree of freedom has to be added to the above
condenser and accumulator, reboiler and stream splitters due to the choice for number of stages in the cascade. Thus
which are connected by interconnecting streams. Each of the number of degrees of freedom for a cascade of N stages
the interconnecting streams has degrees of freedom equal are 3N + 2NC + 5.
to NC + 2 where NC is the number of components present In contrast to the conventional stage, the feed and side
in the stream. These variables consist of NC-1 concentration draw stages are slightly different. This is due to the fact that
variables, a rate or quantity variable and two other intensive an extra stream is involved. For a feed stage, the number
variables i.e. temperature and pressure. of degrees of freedom is equal to the number of degrees of
A stage in a distillation column consists of four freedom for a stage plus the number of degrees of freedom
interconnecting streams, a quantity stream for the hold-up for a stream, thus giving 3NC + 9. A side draw stage can
and a heat stream. Note that only the liquid hold-up is be represented by a stage connected to a splitter which
accounted for while the vapour hold-up is neglected. Since divides between the side draw ow (either a vapour or
there are 5 streams present and each stream has NC + 2 liquid) and the ow to the next stage. Since stages can be
variables, with a heat quantity term, Q, the number of attached to the cascade of stages prior to the side draw stage
variables in the unit is 5NC + 11. The relationships among location, only the number of degrees of freedom for the
the variables depends on the way in which the stage is splitter needs to be counted. Using the same approach as
de ned. Suppose, the stage is considered to be a single described for the conventional stage, with the exception that

Trans IChemE, Vol 76, Part A, March 1998


310 ABDUL MUTALIB and SMITH

there is no hold up, the number of degrees of freedom for a


splitter unit is found to be NC + 5.
A condenser and accumulator unit consist of two ow
streams, a quantity stream and a heat stream. Note that
the liquid hold up is considered as the quantity stream.
The number of variables in the unit comes from the
three streams and a heat quantity term are 3NC + 7, while
the number of relationships between the variables is
2NC + 2, i.e. total material balance, a heat balance, NC-
1 component balances and NC-1 concentration, a pressure
and a temperature identities for liquid leaving the unit.
Therefore, the of degrees of freedom for this unit are
NC + 5.
The reboiler unit consists of 3 ow streams, a quantity
stream and a heat stream. Again, the liquid hold up is
considered as a quantity stream. The number of variables
in the unit which comes from the four streams and a heat
quantity term are 4NC + 9. However, the number of
relationships between the variables that can be derived for
the unit is 3NC + 4 thus giving the number of degrees of
freedom to be NC + 5 (which turns out to be the same as
the condenser unit). The relationships between the
variables that can be derived for the unit consist of a total
material balance, a heat balance, NC-1 components Figure 2. Layouts of the units in a dividing wall column.
balances, NC distribution relationships and NC-1 concen-
tration, two pressure and two temperature identities for
liquid and vapour leaving the unit. additional degrees of freedom which are the liquid and
Table 1 summarizes the degrees of freedom for each of vapour splits.
the units in a typical distillation column. In the implementation of the dividing wall column, it
The dividing wall column can be represented by the is impractical to manipulate the vapour split. Hence, the
Petyluk con guration as illustrated in Figure 2. The column vapour split will be left to occur naturally. Therefore, a
is divided into 6 sections, each containing a number of degree of freedom is lost here. Unlike the vapour split,
cascaded trays. There are three feed and three side-draw the liquid split can be easily manipulated using a simple
trays located between the sections with a partial reboiler device, thus leaving the option open to the designer. The
and total condenser at the top and bottom of the column. decision whether to employ it as a manipulated variable
Table 2 gives details of the analysis of the degrees of will be assessed later.
freedom for the dividing wall column. Based on the
analysis, the number of degrees of freedom for the column
is thus 3(NS(1) + NS(2) + NS(3) + NS(4) + NS(5)+
NS(6))+ NC + 35 where NS(i) is the number of stages in COLUMN CONFIG URATION
section `i . The design of the dividing wall column can be
After taking into account of the inherent relationships treated similar to the design of the Petyluk con guration
and the product speci cations, the number of degrees of (Triantafyllou2 3 ). However, there is one major difference
freedom for the column is NC + 10. Because the feed between the two columns. This results from the xed
composition, owrate and pressure are xed, the number of position of the dividing wall which prevents the manipula-
degrees of freedom left is 9. This means that only nine tion of the internal vapour split in the column. For the
variables can be manipulated or speci ed in order to fully Petyluk con guration, the vapour split was manipulated
control the system. The 9 variables involved are shown in most of the previous studies due to the separate shell
in Table 3. When comparing this to the more estab- arrangement.
lished side draw column, the dividing wall column has 2 A steady state rigorous simulation model for
the dividing wall column was developed using the ASPEN
PLUST M package. The ternary mixture used for the
Table 1. Degrees of freedom for various units in a distillation separation consists of methanol, iso-propanol and butanol.
column. The Wilson equation was used for the prediction of vapour
Unit Degrees of freedom
liquid equilibrium.
The con guration of the dividing wall column was
1. Single phase streams NC + 2 designed using the method proposed by Triantafyllou and
2. Ideal stage 2NC + 7 Smith1 6 , using the option for minimizing the number of
3. Cascade of N ideal stages 3N + 2NC + 5 stages. The feed to the column has an equimolar composi-
4. Feed stage 3NC + 9
5. Stream splitter (no hold up) NC + 5 tion and the products speci ed to be 98.5 mol percent. The
6. Total condenser/accumulator NC + 5 arrangement on the distribution of the number of stages at
7. Partial reboiler NC + 5 different sections inside the column, together with the
operating parameter, are given in Figure 3.

Trans IChemE, Vol 76, Part A, March 1998


OPERATION AND CONTROL OF DIVIDING WALL DISTILLATION COLUMNS: PART 1 311

Table 2. Degrees of freedom analysis for dividing wall column.

Units Degrees of Freedom

Top Column Section.


Condenser/accumulator. NC + 5
Re ux splitter. NC + 5
Cascade of stages (Section 1). 3NT(1) + 2NC + 5
Vapour feed stage. 3NC + 9
Liquid draw stage. NC + 5
Dividing Wall Section.
Prefractionator side.
Cascade of stages (Section 5). 3NT(5) + 2NC + 5
Feed stage 3NC + 9
Cascade of stages (Section 6). 3NT(6) + 2NC + 5
Main Side.
Cascade of stages (Section 2). 3NT(2) + 2NC + 5
Liquid draw stage. NC + 5
Cascade of stages (Section 3). 3NT(3) + 2NC + 5
Bottom Column Section.
Liquid feed stage. 3NC + 9
Vapour draw stage. NC + 5
Cascade of stages (Section 4). 3NT(4) + 2NC + 5
Partial reboiler. NC + 5
Total DOF 3(NT(1) + NT(2) + NT (3) + NT(4) + NT (5) + NT (6))+ 27NC + 87
Restriction from 26 interconnecting streams - 26NC + 52
Total column DOF 3(NT(1) + NT(2) + NT (3) + NT(4) + NT (5) + NT (6))+ NC + 35
Restriction from inherent relationships, design speci cations and uncontrolled variables.
Pressures on all stages, reboiler, condenser and re ux splitter. NT(1) + NT (2) + + NT (6) + 9
Heat leaks on all stages and splitters. NT(1) + NT (2) + + NT (6) + 7
Holdup on all stages. NT(1) + NT (2) + + NT (6) + 3
No. of plates at each section. 6
Feed (composition, owrate and pressure). NC + 1
Total DOF restricted 3(NT(1) + NT(2) + NT (3) + NT(4) + NT (5) + NT (6) + NC + 26
Degrees of freedom for the column 9

IMPAC T OF THE LIQUID SPLIT ON THE MIDDLE according to the set ratio used. Manipulating the liquid split
PRODUCT COMPOSITION at the top of the dividing wall is a way of manipulating
the re ux ratio on each side of the wall. Triantafyllou
In the practical implementation of the dividing wall and Smith1 6 presented a procedure to optimize the re ux
column, the liquid split can easily be manipulated. One way ratio in different parts of the column in the initial design.
of achieving this is by means of simple ow controller Calculations were performed to nd the relation between
installed externally on both liquid streams returning to the liquid split and the composition of the light key
the top of each side of the dividing wall. A ratio controller (methanol) as well as the middle key (iso-propanol) in
can be used to x or to vary the two ows according to a the middle product. To achieve this, the liquid split was
speci ed ratio. However, if a xed ratio is desired, an changed at different values while keeping the vapour
internal mechanism located at the top of the dividing split constant at base case value, i.e. 1.29. Note that the
wall can serve to divide the ows to each side of the wall split is de ned as the ratio between the ow on the product

Table 3. Controlled and manipulated variables for dividing wall column.

Controlled variables Manipulated variables

1. Feed temperature Feed preheater duty


2. Column pressure Condenser cooling duty
3. Top product composition Re ux owrate
4. Middle product composition Distillate owrate
5. Bottom product composition Sidedraw owrate
6. Condenser/accumulator holdup Reboiler duty
7. Reboiler holdup Bottom product owrate
8. Light impurity in middle product Liquid split at top of dividing wall
9. Heavy impurity in middle product Vapour split at bottom of dividing wall

Trans IChemE, Vol 76, Part A, March 1998


312 ABDUL MUTALIB and SMITH

The magnitude of changes for the methanol composi-


tion seems to be very small for most of the range used for
the variation in the liquid split. Therefore, controlling the
methanol concentration using the liquid split would require
a large action in order to correct a small deviation. In
addition, varying the liquid split also affects the iso-
propanol composition by a similar magnitude to the effect
on the methanol composition. If the iso-propanol composi-
tion is controlled by another manipulated variable such as
the side draw owrate, the two control loops are bound to
interact signi cantly.

IMPACT OF THE VAPOUR SPLIT ON THE MIDDLE


Figure 3. The dividing wall column con guration and the base case PRODUCT COMPOSITION
operating parameter.
In the dividing wall column operation, manipulating
the vapour splits would be impractical. The vapour split
side to the ow on the feed side of the dividing wall. The inside the column occurs naturally according to the pressure
feed composition, owrate and temperature as well as the drop relation across the internals at each side of the
product owrates and reboiler duty were also kept constant. dividing wall, which will be discussed later. Despite this,
Figure 4 (i) shows the variation of methanol composi- calculations were made to determine the effect of chang-
tion in the middle product, while Figure 4 (ii) shows the ing the vapour split on the middle product composition.
variation of iso-propanol composition in the middle product. The idea of doing this was to investigate whether xing the
Both were subjected to variation in the liquid split. The dividing wall would lead to any major bene t being missed.
pattern of changes followed by the methanol composition Figure 5 (i) shows the variation of butanol composition
exhibits a minimum point as the liquid split is varied. The in the middle product while Figure 5 (ii) shows the variation
signi cance of this observation is that simple PID control of iso-propanol composition in the middle product. Both
cannot be applied to link these two variables to form a were subjected to variation in the vapour split. This leads
control loop. At two different locations along the curve, to the same conclusions as for the liquid split.
separated by the minimum point, the direction for the From the analysis conducted for the liquid and vapour
control action is different. Since it is not possible for a splits, it is clear that maintaining the two splits constant
normal PID controller to recognize on which side of the seems to be the preferred option. The same suggestion
curve the current operation is located, and it is not also can also be extended to the Petyluk column. However,
possible for the controller to have a variable gain sign, in doing so, the impurities and the main component
applying the controller in such a situation will lead to compositions in the middle product cannot be controlled
failure. simultaneously.

Figure 4. Variation in middle product composition subject to changes in liquid split. (a) Methanol; (b) Iso-propanol.

Figure 5. Variation in middle product composition subject to changes in vapour split. (a) Iso-propanol; (b) Butanol.

Trans IChemE, Vol 76, Part A, March 1998


OPERATION AND CONTROL OF DIVIDING WALL DISTILLATION COLUMNS: PART 1 313

Figure 6. Generalized pressure drop correlation chart.

IMPACT OF THE LIQUID SPLIT ON THE


VAPOUR SPLIT
As mentioned earlier, the vapour divides to satisfy the Figure 7. Area ratio and vapour split relationship for pressure drop
equalization between the two sides of the dividing wall. (Equal number
pressure drop equalization between the sections on each of stages at both sides).
side of the dividing wall. In the case where the number of
stages are equal at each side of the dividing wall, the vapour
split depends on the position of the dividing wall and the
liquid loading of the two sections at each side of the dividing Figure 7 shows the plot obtained, which demonstrates a
wall. The liquid loading of the two sections vary as the linear relation between the area ratio and the vapour split
liquid split is changed. required. More signi cantly, the area ratio was found to be
Before the impact of the liquid split on the vapour split almost the same as the vapour split speci ed when equal
can be assessed, some form of relation must be established number of stages were used on each side of the dividing
between the changes in the liquid loading and the resulting wall. This indicates that in order to obtain the desired
changes in the pressure drop across a packing height or vapour split, the position of the dividing wall can be set to
trays. In the present case, it was decided to demonstrate give the same area ratio as the required vapour split.
the study using packing as the internals.
Figure 6 shows the Generalised Pressure Drop Correla-
tion chart which is typically used for determining the DESIGNING FOR THE LIQUID AND THE
pressure drop across a packed height for a given liquid and VAPOUR SPLITS
vapour loading. When considering the changes in the Having given all the arguments which led to the
pressure drop with respect to the liquid loading, two regions preference for operating the dividing wall column with a
can be identi ed. In the rst region, which corresponds to constant liquid and vapour split, it is then important to
low liquid loading, the pressure drop hardly changes with know how to design for the two splits. As it is always
the liquid loading. In the second region, which corresponds desirable to operate the column as close as possible to the
to high liquid loading, the pressure drop changes more optimal state, the sensitivity of the optimum location for
signi cantly with the liquid loading. the two splits with regard to changes in the feed composition
Based on the above observation, a number of conclusions and the product speci cations needs to be assessed.
can be made. If the dividing wall column is operated within
the rst region, the liquid split changes should not affect
the vapour split. Therefore, operating within this region IMPACT OF FEED COMPO SITION ON THE
should provide some exibility in changing the liquid split. OPTIMUM LIQUID AND VAPOUR SPLITS
Conversely, if the column is operated within the second Calculations were performed to determine the variation in
region, the vapour split will vary as the liquid split changes, the liquid and vapour splits required to maintain lowest
thus offering no exibility during the operation of the energy consumption for changes in the feed composition.
column. Following this, a correlation between the pressure Three different feed compositions within the allowable
drop and the liquid and vapour loadings for a speci c range were tested and the value for the two splits that gave
type of packing (Gempak 4A, GLITSCH2 4 ), applicable minimum energy consumption were found for each case
within the rst region, was used to determine the rela- using rigorous simulation. Note that the middle product
tion between the position of the dividing wall (described impurities were not included in the product speci cations
by the area ratio between the product side to the feed side as it has already been decided not to control them using
of the column) and the vapour split speci ed. For a given the liquid and the vapour split. Table 4 presents the results
cross sectional area of the column and the number of stages obtained from the analysis. The optimum values for the
on each side of the dividing wall and using the infor- liquid and vapour splits were found to be insensitive to
mation on liquid and vapour loading at each stage obtained the variation in the feed composition.
from rigorous simulation for a speci ed liquid and
vapour split, the area ratio required to satisfy the pres-
sure drop equalization criteria can be calculated. The IMPACT OF PRODUCT SPECIFICATION ON THE
criteria can be based on either the total pressure drop OPTIMUM LIQUID AND VAPOUR SPLITS
or the average pressure drop on both sides of the dividing Following the above, calculations were performed for
wall. different product speci cations. Table 5 presents the results

Trans IChemE, Vol 76, Part A, March 1998


314 ABDUL MUTALIB and SMITH

Table 4. Optimum liquid and vapour split for changes in feed composition. Table 5. Optimum liquid and vapour split for changes in product
speci cation.
Feed composition Optimum liquid split Optimum vapour split
Product speci cation Optimum liquid split Optimum vapour split
Feed 1 3.80 1.16
(0.333, 0.334, 0.333) 98 mol percent 3.5 1.02
Feed 2 3.80 1.16 98.5 mol percent 3.8 1.16
(0.363, 0.334, 0.303) 99 mol percent 3.7 1.29
Feed 3 3.90 1.16
(0.303, 0.364, 0.333)
Feed 4 3.75 1.15
(0.303, 0.334, 0.363)
In the dividing wall column operation, the vapour split
is largely xed by the position of the dividing wall but
the liquid split has the exibility to be changed. However, as
obtained. The optimum values for the liquid and vapour discussed earlier, manipulating the liquid split can lead to
splits were found to vary as the product speci cations serious problem in the operation and control of the dividing
changed. Figure 8 shows three sets of plots where the wall column. Therefore, it is better to x the liquid split.
difference in reboiler duty between the operation of the Fixing the liquid and vapour split for a different feed
column at a speci ed liquid or vapour split and operation at composition will not be a problem as the optimum value
the optimum splits are plotted against the speci ed liquid or for the liquid and the vapour split hardly change. However,
vapour split. The plots cover the location within the it is not so straightforward when the product speci cations
optimum splits for three different product speci cations. are allowed to vary. In this case, the above nding can be
There are two plots for each product speci cation. For the used to help decision making. The obvious choice is to set
rst plot, the vapour split is kept constant at its optimum the liquid and vapour splits at the optimum values for the
value while allowing liquid split to vary. For the second highest product speci cations designed for the column. The
plot, the liquid split is kept constant at its optimum value reason for this is because of the relative importance in
while allowing the vapour split to vary. operating the column with liquid and vapour splits close
A comparison between the plots obtained for the different to the optimum values for higher product speci cations
product speci cations shows that the optimum location compared with lower product speci cations. Even if the
tends to be more sensitive to the changes in liquid and column is operated to produce lower purity products,
vapour splits as the product speci cations were increased. the penalty in terms of the energy consumption when
This was shown by the extent of the at region within operating the column with the optimum splits for the higher
the optimum location which tended to reduce slowly as product speci cations proves to be reasonably small. On the
the product speci cations were increased. Hence, it is other hand, if the column is operated with the optimum
important that the column is run close to the optimum liquid and vapour splits for lower product speci cations
liquid and vapour splits for the higher product speci- and higher purity products are to be produced, the penalty
cations otherwise a high energy penalty will have to be in the energy consumption can be extremely high and, in
paid. certain cases, the higher product speci cations might not

Figure 8. Plots of changes in reboiler duty requirement with different liquid and vapour splits.

Trans IChemE, Vol 76, Part A, March 1998


OPERATION AND CONTROL OF DIVIDING WALL DISTILLATION COLUMNS: PART 1 315

Table 6. Results from an example to illustrate the setting for the liquid and vapour splits for the dividing wall column operation.

Operating liquid Optimum liquid Energy penalty


Product speci cation and vapour splits and vapour splits (percent)

(0.98, 0.98, 0.98) 3.7 1.29 3.5 1.02 6.9


(0.99, 0.99, 0.99) 3.5 1.02 3.7 1.29 product specs. not achieved
with the operating splits

be achieved even running the column at total re ux. INTERACTION ANALYSIS


Table 6 illustrates the results from an example to explain Two control schemes that can be used for controlling
the above. the dividing wall column were considered, i.e. L-S-V and
D-S-V (where L, S, D and V refer to manipulation of the
top re ux, side draw, distillate ows and vapour ow from
COMPO SITION CONTRO L OF THE DIVIDING the reboiler respectively). These are shown in Figure 9.
WALL COLUMN Relative Gain Array Analysis (Bristol2 5 ) was used to
In order to study the control behaviour of the divid- analyse the interaction as well as determining suitable
ing wall column, a dynamic model was built using pairings between controlled and manipulated variables in
SPEEDUP T M . Similar con gurations for the column as the two control schemes.
modelled previously using the ASPEN PLUST M package Basically, the RGA is a matrix which consists of
were used. The same feed and product speci cations were elements representing the steady state gain ratio between
employed. In line with the earlier suggestion, the liquid the respective controlled and manipulated variables
and vapour split were maintained at the optimum values when all other manipulated variables are constant, divi-
obtained from the design method used (Triantafyllou and ded by the steady state gain ratio between the same
Smith1 6 ). The assumption of no heat transfer across the controlled and manipulated variables when all other
dividing wall was maintained and only decentralized control controlled variables are constant. This is represented by
was used. the equation:
The short cut dynamic model which is available in
SPEEDUP T M was used. Assumptions involved in using
k ij = (y / m )
i j mi / (yi / mj )yj
the short cut model are perfect material balance, i.e. no where k ij is the relative gain between controlled variable yi
accumulation of material in the system, and constant molal and manipulated variable mj . If k ij = 0 then yi does not
over ow. In addition, since the short cut model solves respond to mj and mj should not be used to control yi . If
stage equilibrium by lumping several stages together to k ij = 1 then yi only responds to mj and does not interact
form a section within the column, constant relative with other manipulated variables. This is the preferred case.
volatility within the section was assumed (but varied between If 0 <k ij <1.0 or 1 <k ij < then an interaction exists
sections). The maximum number of stages that can be because not only yi will respond to mj but other controlled
lumped within a section must be kept within a reasonable variables also. Finally, if k ij <0 then the interaction exists
number in order to keep the model within an acceptable between the related manipulated and controlled variables
accuracy for simulation purposes. The relative volatility is in an opposite direction and can cause instability. The
within each section was speci ed by taking the average selection of the best control con guration, as well as the
value obtained for the relevant stages from the rigorous pairings, is based on the arrangement which gives a RGA
simulation done using ASPEN PLUST M . with diagonal element values closest to unity, indicating
the least interaction.
The results from the RGA analysis are given in Table 7.
From the RGA analysis, the extent of interaction between
the separate control loops in the D-S-V con guration seems
to be lower than the traditional L-S-V con guration. In
terms of controlled and manipulated variable pairings, the
RGA seems to suggest that cross pairing should be adopted

Table 7. Results for the steady state gain array and the relative gain array.

Scheme Steady state gain array Relative gain array

2.433 - 0.274 - 2.659 11.736 - 10.057 - 0.679


L-S-V - 0.003 0.019 - 2.672 0.001 0.011 0.988
- 2.401 0.297 2.665 - 10.736 11.045 0.691
2.433 - 0.274 - 2.659 0.891 0.100 0.008
D-S-V - 0.003 0.019 - 2.672 0.001 0.007 0.992
0.014 0.014 0.001 0.107 0.892 0.001
Figure 9. Two composition control scheme for the dividing wall column.

Trans IChemE, Vol 76, Part A, March 1998


316 ABDUL MUTALIB and SMITH

Figure 10 (i). Control run for scheme L-S-V with set point changes. (ii). Control run for scheme D-S-V with set point changes.

between the middle product and the bottom product control DYNAMIC SIMULATION
loops. This means that the middle product composition
should be controlled by the vapour ow while the bottom Dynamic simulation was then performed to observe
product composition should be controlled by the side draw the behaviour of the two control schemes when subjected
ow. to set point changes and feed disturbances as shown in

Trans IChemE, Vol 76, Part A, March 1998


OPERATION AND CONTROL OF DIVIDING WALL DISTILLATION COLUMNS: PART 1 317

Figure 11 (i). Control run for scheme L-S-V with feed disturbance. (ii) Control run for scheme D-S-V with feed disturbance.

Table 8. The controllers were tuned using the open loop show the response of the controllers in the two schemes
response curve method. Figure 10 (a) and (b) show the when subjected to feed disturbances. Again stable control
response of the controllers in the two schemes when response was produced by both schemes.
subjected to set point changes. Both schemes were able to Overall, the D-S-V scheme seemed to produce better
produce a stable control response. Figure 11 (a) and (b) response compared with the traditional L-S-V scheme and

Trans IChemE, Vol 76, Part A, March 1998


318 ABDUL MUTALIB and SMITH

Table 8. Set point changes and feed disturbances used. selected for demonstrating the control using simulation. The
column was subjected to changes in product speci -
Middle product composition
Set point changes IPA: 0.9850.988 cation and feed disturbances. Simulation results indicate
that stable control was achievable by both control schemes.
Feed disturbances Combined feed composition and owrate
Composition: MeoH 0.333 to 0.363
IPA 0.334 to 0.284 REFERENCES
BuoH 0.333 to 0.353
1. Linhoff, B., Dunford, H and Smith, R., 1983, Chem Eng Sci, 38(8):
Flowrate: 1.081 to 1.190 kmol/hr
1175.
2. Smith, R. and Linhoff, B., 1988, Chem Eng Res Des, 66: 195.
3. Brugma, A. J., 1942, US Patent 2,295,256.
4. Petyluk, F. B, Platonov, V. M. and Slavinskii, D. M., 1965, Int Chem
Eng, 5(3): 561.
this is in agreement with the results derived from the RGA 5. Fidkowski, Z. and Krolikowski, L., 1987, AIChE J, 33(4): 643.
6. Glinos, K. and Malone, M. F., 1988, Chem Eng Res Des, 66 (3): 229.
analysis. More importantly, both schemes were able to 7. Kaibel, G., 1988, IChemE Symp Series No. 109: 43.
produce satisfactory control. 8. Wright, R. O., 1945, US Patent 2,471,134.
9. Kaibel, G., 1987, Chem Eng Technol, 10: 92.
10. Stupin, W. J., 1970, The separation of multicomponent mixtures
CONCLUSIONS in thermally coupled distillation systems, PhD Diss (University of
Southern California).
In this paper, studies relating to aspects of the operation 11. Fonyo, Z., Szabo, J. and Foldes, P., 1974, Acta Chim, 82: 235.
and control of the dividing wall column have been 12. Tedder, D. W. and Rudd, D. F., 1978, AIChE J, 24(2): 303.
investigated. When compared to the side draw column, 13. Cerda, J. and Westerberg, W., 1981, Ind Eng Chem Proc Des Dev,
the dividing wall column has a more complex nature, shown 20(3): 546.
14. Spadoni, G. and Stramigioli, C., 1983, 3rd Int Cong Computers and
through degrees of freedom analysis. This results from two Chemical Engineering, Paris, No 27: 43.
additional manipulated variables, the liquid and vapour 15. Nikolaides, I. P. and Malone, M. F., 1987, Ind Eng Chem Res, 26(9):
splits. However, it is impractical to manipulate the vapour 1839.
split, which is xed by the position of the dividing wall. On 16. Triantafyllou, C. and Smith, R., 1992, Trans IChemE, 70 (A2): 118.
the other hand, the liquid split can be easily varied but has 17. Chavez, R., Seader, J. D. and Wayburn, T. L., 1986, Ind Eng Chem
Fundam, 25(4): 566576.
been found to have little bene t in comparison to the 18. Lin, W. J., Seader, J. D. and Wayburn, T. L., 1987, AIChE J, 33:
complication that will be added to the operation and control 886897.
of the column. A linear PID controller cannot be used and 19. Wolff, E. A., Skogestad, S. and Havre, K., 1993, AIChE Ann Meet,
severe interactions would occur between the control loops, St. Louis, paper 195a.
20. Wolff, E. A., Skogestad, S. and Havre, K., 1994, ESCAPE 4, Dublin,
particularly for controlling the composition of the middle IChemE Symp Series No 133: 111118.
product and its impurities. In the region of high liquid and 21. Morud, J. and Skogestad, S., 1994, AIChE Ann Meet, San Francisco,
vapour loading, varying the liquid split will also affect the paper 131d.
vapour split considerably and this even affects the operation 22. Howard, G. M., 1967, Ind Eng Chem Fundam, 6(1): 86.
and control of the column further. Based on this account, it 23. Triantafyllou, C., 1991, The design optimisation and integration of
dividing wall distillation columns, PhD Thesis (submitted to UMIST
was suggested that the column should be operated and Manchester).
controlled with a constant liquid split in addition to the xed 24. GLITSCH, 1992, (Private Communication.)
vapour split. Having made the suggestion, the column can 25. Bristol, E. H., 1966, IEEE Trans Autom Control, AC-11: 133134.
then be operated and controlled in a similar manner as
the more established side draw column. Nevertheless, the
ACKNOWLEDGEMENTS
disadvantage of this proposal is that the impurities
composition in the middle product cannot be controlled. The authors would like to express their appreciation to the UK
Department of Energy, Energy Ef ciency Of ce, BP, Exxon, Glitch, ICI,
The optimum location for the liquid and vapour splits M. W. Kellogg, and Shell for nancial support of this project. The authors
was found to be insensitive to the variation in the feed would also like to express their gratitude to Dr Frigyes Lestak for his
composition but this is not so for the variation in the product contribution to the project.
speci cation. From a sensitivity study, it was found that the
lower product speci cation has a atter optimum compared
with the higher product speci cations. Therefore, it was ADDRESS
concluded that the design for the liquid and the vapour split Correspondence concerning this paper should be addressed to
should be based on the highest product speci cation when Professor R. Smith, Department of Process Integration, UMIST,
PO Box 88, Manchester M60 1QD, UK.
the column is required to produce a range of product
compositions. The manuscript was received 4 August 1997 and accepted for publication
Two composition (three point) control scheme were after revision 22 January 1998.

Trans IChemE, Vol 76, Part A, March 1998

Você também pode gostar