You are on page 1of 5

Ind. Eng. Chem. Res.

1999, 38, 451-455

451

On the Steady-State Multiplicities for an Ethylene Glycol Reactive


Distillation Column
Rosendo Monroy-Loperena and Jose Alvarez-Ramirez*,
Area de Ana lisis Aplicado and Area de Ingeniera Qumica, Universidad Auto noma
MetropolitanasIztapalapa, Apartado Postal 55-534, Me xico D.F. 09000, Me xico

An ethylene glycol column is studied in this paper, where the main objective is to describe the
bifurcation diagram with respect to the reboiler boilup ratio. The bifurcation analysis reveals
the existence of a unique equilibrium point at low reboiler heat inputs and three equilibrium
points (two stable and one unstable) at high reboiler heat inputs. Moreover, the existence of
input multiplicities at moderate and high product purity is also revealed.
1. Introduction
The importance of multiplicity of solutions in separation processes stems from the fact that one set of design
parameters can yield several operating conditions. A
consequence of this situation is that control actions must
to be taken to regulate the operation of the process.
Multiplicity of solutions in separation processes has
been a largely studied problem. While some results state
the uniqueness of solution for binary distillation,1-3
other results come to the conclusion that multiple steady
state can arise in binary distillation if the flow rates
are given on a mass basis instead of a molar basis4 or if
multicomponent mixtures are separated in processes
with an increase of the number of stages.5
Regarding reactive distillation, several works have
reported the existence of multiple steady states.6-13
Both input and output multiplicities have been described for the cases methyl tert-butyl ether and ethyl
tert-butyl ether reactive distillations.9,10 Karpilovskiy et
al.13 derived a qualitative criterion for the existence of
multiple solutions (output multiplicity) in single-product
reactive distillation, given the main features of a distillation line diagram and reaction stoichiometry. The
nature of Karpilovskiy et al.s approach (see also Pisarenko et al.6) has a lot of similarities with that used to
analyze multiplicity of steady states in chemical reactors. The singularity theory approach has also been used
to look for the origin of the multiplicities in reactive
distillation.11,12 Gehrke and Marquardt11 used a onestage reactive distillation process for the identification
of possible sources of steady-state multiplicities related
to different chemical species. Guttinger and Morari12
reported an / analysis for the prediction of multiple
steady states in multicomponent azeotropic distillations.
Although both works substantiate the importance of
multiplicities in reactive distillation processes, they
conclude that more systematic research based on rigorous models is required to identify the physical reasons
for the phenomenon leading to multiple steady states.
Despite the large amount of reported results on
multiplicity in reactive distillation, only recently have
* To whom correspondence should be addressed. E-mail:
jjar@xanum.uam.mx. Fax: +52-5-7244900. Phone: +52-5-7244649.
Area de Ana
lisis Aplicado.
Area de Ingenier
a Qumica.

their causes and implications for operation and control


been discussed.4,9,10,14 One of the first systematic approaches to the problem of controlling reactive distillation was addressed by Kumar and Daoutidis.14 The
main control objective is to control the product purity
using the reboiler heat input. As a preliminary study,
a bifurcation analysis was performed, which revealed a
region of output multiplicity and a transition from
minimum-phase behavior at moderate product purity
to a nonminimum-phase behavior at high product
purity. For the case of an ethylene glycol reactive
distillation column, a nonlinear controller was designed
that yields good performance in setpoint tracking.
Ciric and Gu15 have presented a reactive distillation
column for manufacturing ethylene glycol. Using homotopy methods, Ciric and Miao8 demonstrated the
existence of output multiplicities for a cost-optimal
column configuration (i.e., total reflux and optimal
distributed feed). The aim of this short paper is to
describe the bifurcation diagram of the ethylene glycol
distillation column with respect to the reboiler heat
input. The main task is to demonstrate the existence of
both input and output multiplicities with respect to the
reboiler heat input which in the cases of columns with
total reflux is the most commonly used manipulated
variable to regulate the product purity. In Kumar and
Daoutidiss work,14 some preliminary results on the
multiplicities of ethylene glycol reactive distillation
columns were reported. Basically, the work by Kumar
and Daoutidis focuses mainly on the design of a control
law and uses the bifurcation diagram to illustrate the
existence of unstable operating points. In this work, we
take a step ahead by showing that these columns may
present both input and output multiplicities. Moreover,
the origin of the multiplicities and the main variables
affecting the bifurcation maps are discussed. On the one
hand and in the spirit of Pisarenko et al.s work,6 the
main idea to explain the origin of the multiplicities is
that intermediate products are recycled via the separation stages, which leads to an increase of the reaction
rate and consequently to an increase in the heat
production due to exothermic reactions. On the other
hand, the main variables which may affect multiplicities
are the reboiler boilup ratio and the feed composition.
It must be pointed out that this work must be seen
as a preliminary study toward addressing the control
problem of ethylene glycol distillation columns. In fact,
Jacobsen and Skogestad4 have remarked on the impor-

10.1021/ie980424q CCC: $18.00 1999 American Chemical Society


Published on Web 01/14/1999

452 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

tance of manipulated input/regulated output bifurcation


diagrams in obtaining a priori information about the
operability of the process and possible control performance limitations. These issues will be discussed in this
paper.
2. Base Case Study
Ciric and Gu15 have presented a reactive distillation
column for manufacturing ethylene glycol (EG) and
diethylene glycol (DEG) from ethylene oxide and water.

C2H4O + H2O f C2H6O2 (EG)


C2H40 + C2H6O2 f C4H10O3 (DEG)

(1)

The following assumptions were adopted to develop


a rigorous model (see Ciric and Miao8): (1) the vapor
and liquid phases are in equilibrium on each tray; (2)
the vapor-liquid equilibrium can be considered ideal
(ideality existing between polar molecule interactions
cancel out); (3) the reactions only occur in the liquid
phase; (4) the liquid phase is homogeneous; (5) the
enthalpy of the liquid streams is negligible relative to
the heat of vaporization and the heats of reaction; (6)
the heat of vaporization and the heats of reaction are
constant; (7) the reactions only occur on the trays within
the reaction zone, and (8) the column operates at total
reflux. The steady-state model of the column is given
by material and energy balances, reaction rate equations, and summation equations. The model we use in
this work was taken from refs 8 and 15. For completeness in presentation, a brief description is given in the
Appendix.
Ciric and Gu15 used a mixed-integer nonlinear programming approach to obtain a cost-optimal column
design (see Figure 1). Because ethylene oxide is the
common reactive to both reactions (see reaction scheme
(1)), the idea is to optimize its feeding along the column.
The cost-optimal column contains 10 trays. Separation
and reaction occur on trays 5-10, and separation
without reaction occurs on trays 1-4. Reaction volumes
are distributed unevenly among trays 5-10. Water in
the amount of 26.3 kmol/h is fed onto the top tray, and
a total feed of 27.56 kmol/h of ethylene oxide is distributed among trays 6-10. The distribution of ethylene
oxide and the liquid holdup volumes among the reactive
trays was taken as in Table 2 in Ciric and Miao.8 In
addition, physical properties for chemical reactions were
taken as in Table 1 in Ciric and Miao.8
It must be remarked that the physical realization of
the optimal column with the feed distribution presented
by Ciric and Gu15 is potentially hard as the ratio of the
liquid holdup on two consecutive stages is roughly 125
(see Table 4 in Ciric and Gu15). In fact, the two-feed
column is significantly more balanced. Although this
case could be a better base case study, in the next
section we will show that the feed distribution scheme
does not have a significant effect on the qualitative
behavior of the bifurcation maps (structural stability).
Ciric and Miao8 used homotopic continuation methods
to describe multiple steady-state solutions. Using the
liquid holdup on reactive trays as a homotopy parameter, they demonstrated the existence of three steady
states for the optimal value of the boilup ratio ) 0.958.
Because all of the states of a reactive distillation column
are bounded, one can conclude that the steady state
corresponding to the middle product purity is unstable.

Figure 1. Optimal distributed feed reactive distillation column


for ethylene glycol synthesis (Ciric and Miao8).

Using eigenvalue analysis, one can also conclude that


the other two steady states (high and low product
purities) are stable. Therefore, the cost-optimal operating point corresponds to the (stable) high product purity.
In reactive distillation columns with total reflux, the
reboiler heat duty, say Qh, is taken as the manipulated
variable to regulate the product purity.14 In our case
(see the model in the Appendix), the boilup ratio (RBR)
is the portion of liquid stream leaving the reboiler tray.
Thus, is proportional to Qh and can be taken as the
manipulated variable.
To explore the effects of the distributed ethylene oxide
feed in the bifurcation map, we will also consider a
suboptimal case. Here, the distributed feed was replaced
with a single ethylene oxide feed located on tray 5. In
this suboptimal column, all of the 27.56 kmol/h ethylene
oxide is fed to tray 5, rather than distributed among
trays 5-10, and reaction volumes are setting on each
reaction stage to 0.549 m3.
Although we have called suboptimal to this feed
distribution configuration, it must be remarked that the
column structures studied are both cost-optimal but for
a different number of maximum feed points (for details,
see Ciric and Gu15).
3. Computational Aspects
The model defined in the Appendix leads to a set of
70 algebraic equations, which is highly nonlinear and
ill-conditioned (the corresponding Jacobian presents
high conditioning numbers). The bifurcation study has
been conducted using AUTO, a software package developed by Doedel.16 Among other capabilities, AUTO
generates one- and two-parameter bifurcation diagrams
for systems of nonlinear equations in the steady state.

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 453

Figure 2. Bifurcation diagram for the cost-optimal configuration.


The symbols s and denote stable and unstable steady states,
respectively.

The set of algebraic equations was solved with a


Newton-Raphson method. To reduce round-off effects
due to ill-conditioning and sparsity of the Jacobian
matrix, matrix inversion was made with the subroutine
MA28.17 Moreover, initial conditions to track steadystate branches were taken as the (unstable) middle
product purity solution corresponding to ) 0.958. This
solution was obtained via the homotopic continuation
method suggested by Ciric and Miao.8
4. Results
The bifurcation diagram between the EG product
purity (output variable) and the RBR (input variable;
recall that 0 e e 1) for the cost-optimal and the
suboptimal cases are shown in Figures 2 and 3, respectively. Note that the bifurcation diagram corresponding
to the cost-optimal case is very similar to the bifurcation
diagram corresponding to the suboptimal case. This
show that the steady-state structure displayed by the
cost-optimal configuration is structurally stable. This
means that small changes in the configuration and/or
operating conditions of the cost-optimal reactive distillation column do not drastically change the bifurcation
diagram. Moreover, structural stability of the bifurcation diagram in Figure 2 provides further evidence that
the multiple steady-state phenomenon is a result of
neither the modeling techniques nor optimized configurations.
Regarding the bifurcation diagram presented in Figures 2 and 3, the following comments are in order:
(i) Input Multiplicities. Second-order input multiplicities were obtained for EG product purities higher
than about 0.27. This implies that the same EG product
purity is obtained at two different values of the RBR.
However, the operating conditions corresponding to the
higher RBR are open-loop unstable while the other ones
corresponding to the lower RBR are stable. Hence, open-

Figure 3. Bifurcation diagram for the suboptimal configuration.

loop operation of the EG reactive distillation column is


feasible at operating conditions corresponding to low
energy consumption rates. This is not the case in the
methyl tert-butyl ether (MTBE) reactive distillation
column studied by Sneesby et al.9,10 where low and high
energy consumption operations are possible because of
a third-order input multiplicity. In such a case of
reactive distillation, the manipulated variable is the
reflux rate from a total condenser. The same MTBE
product purity is obtained at three different values of
the reflux rate. However, the steady states corresponding to low and high reflux rates are stable, while the
steady state corresponding to the middle reflux rate is
unstable.
(ii) Output Multiplicities. Output multiplicity occurs at high RBRs. At RBRs between 0.76 and 1.0, there
are three separate unique solutions to exactly the same
value of the RBR . Each solution corresponds to a
different product purity of EG. When only one steady
state is present (no output multiplicity), such a steady
state is stable. On the other hand, when three steady
states are present, the ones corresponding to low and
high product purities are stable, while the other one
corresponding to middle product purities is unstable
(saddle type). Along the branch containing unstable
steady-state solutions, open-loop operation is not possible. In fact, an arbitrarily small disturbance acting on
the process moves the operating point toward either the
high or the low product purity operating points. Along
this unstable branch, a counterintuitive phenomenon
can be observed: Increasing the product purity results
in decreased RBR. From an operating viewpoint, it is
not hard to conclude that this phenomenon is inherently
unstable.4 Moreover, because of the fact that for each
unstable steady state there exists a stable steady-state
solution with a lower reboiler boilup ratio (second-order
input multiplicity), closed-loop (controlled) operation
cannot be realized along the unstable branch. That is,
for a given setpoint of the product purity and under a

454 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

stabilizing feedback control, the manipulated variable


converges to the stable branch.
(iii) Startup of the Reactive Distillation Column.
If operating conditions are chosen such that input
multiplicity is present, the startup process can be
complex. In fact, this is because the initial regime
usually corresponds to a high concentration of reagent
and a low concentration of product and because the
column is first filled with the initial mixture. Thus, the
conditions are created for the process trajectory to be
in the region of attraction of the steady state with the
lower reaction extent.
To avoid these startup problems, the structure of the
bifurcation diagram can be exploited to obtain the
following startup guideline. Note that the bifurcation
diagram can be seen as composed of two branches (see
Figure 2). Branch A is defined for all and corresponds
to stable steady states. Branch B is defined only for high
values of the RBR ( > 0.85) and contains both stable
(low product purity) and unstable (middle product
purity) steady states. The idea is to start up at operating
conditions (sufficiently low values of the RBR ) where
only one steady state is present. Once the reactive
distillation column has reached the corresponding stable
steady state (belonging to branch A), the roboiler ratio
must be increased slowly up to its design value *. Thus,
the reactive distillation column will track branch A to
attain the steady state with high product purity corresponding to the design value *.
(iv) Cost-Optimal Operating Conditions. The
bifurcation diagram in Figure 2 corresponds to the costoptimal configuration. The optimal value of the reboiler
boilup ratio obtained by Ciric and Miao8 is ) 0.958,
whose corresponding steady states are shown in Figure
2. Under cost-optimal operating conditions, the EG
reactive distillation column presents three steady states
(output multiplicities). However, the cost-optimal steady
state corresponds to the higher product purity (xD,EG =
0.965), which is stable and the higher possible purity
that can be obtained with the lower possible reboiler
boilup ratio. Note that if increases, the energy
consumption also increases while the product purity
decreases.
(v) Mechanisms. Input and output multiplicities
arise because of the very complicated link between
separation and reaction processes or more formally
because of the considerable nonlinearity of the equations
describing these processes. Hauan et al.18 developed a
qualitative understanding of this phenomenon for the
case of MTBE production. The main idea of their
qualitative understanding is that intermediate products
are recycled via the separation stages, which leads to
an increase of the reaction rate and consequently to an
increase in the heat production due to exothermic
reactions. In this sense, multiplicity of steady states is
the way the reactive distillation column conciliates its
internal competing mechanisms, namely, heat extraction via the vapor flow rate (separation mechanism) and
heat production via the exothermic reactions (reaction
mechanisms). As the internal recycling increases, more
reagents are incorporated into the reaction stages, thus
leading to more stressed competing mechanisms. Evidence of this phenomenon is presented in Figure 4
where the internal recycling rate (backmixing) defined
by F ) L10/FT as a function of the RBR is presented.
Here FT is the total fed flow rate. The internal recycling
rate F represents the average number of times a pack

Figure 4. Recycling backmixing rate (F) as a function of the


reboiler boilup ratio (). This diagram corresponds to the costoptimal bifurcation diagram in Figure 2.

of reagents is incorporated into the separation/reaction


path. For low values of , F is on the order of 3-5,19
which are typical values for nonreactive distillation
columns. For high values of , F is on the order of 50
and larger, which evidences the fact that the reactive
distillation column behaves like a tubular reactor with
high rates of recycled reagents.20 Within a qualitative
framework, our findings agree well with those described
in Figure 1b of Karpilovskiy et al.s work.13
(vi) General Reactive Distillation Configurations. It is our belief that the input and output
multiplicity structure and competing mechanisms described in this work are typical of general reactive
distillation columns. This means that input and output
multiplicities are very likely to occur in reactive distillation columns under high internal recycling rates. We
based our conclusion on the result in this work and
previously reported ones.6,8,9,10,14,18 In our case, output
multiplicity occurs at high RBRs. In Sneesby et al.s
work9,10 for the case of MTBE reactive distillation
columns, output multiplicity occurs at high reflux rates.
Both mechanisms, high RBRs and high reflux rates,
have the effect of increasing the incorporation of reagents into the reactive stages, thus leading to a highly
stressed operation.
5. Conclusions
This short paper discussed the steady-state structure
displayed by an ethylene glycol reactive distillation
column. The cost-optimal reactive distillation configuration reported by Ciric and Miao,8 which is a total
reflux distillation column, was taken as the base case
study. The following conclusions were obtained:
1. Very similar bifurcation diagrams were obtained
for the cost-optimal and suboptimal cases. This evidences that the steady-state phenomenon is a result of
neither the modeling techniques nor optimized configurations.

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 455

2. Input multiplicities occur at moderate and high EG


product purity. The same EG product purity is obtained
at two different values of . The steady state corresponding to low energy consumption (low ) is stable,
so then open-loop operation of the EG reactive distillation column with low energy consumption rates is
possible.
3. Output multiplicities occur at high RBRs ( > 0.76).
Generically, either one or three steady states can be
present, which agrees with previously reported results
for other reactive distillation columns.9,10
Appendix
Assume that the column has N trays (counted from
bottom to top), I chemical species, and J chemical
reactions and operates at total reflux. The material
balance for chemical species i on tray k in the middle of
the column is given by

Fi,k + Vk-1xi,k-1 + Lk+1xi,k+1 - Lkxi,k - VkKi,kxi,k +


J

i,jj,k ) 0

j)1

(A.1)

where i, j, and k denote the chemical species in the


reactive system, the chemical reaction, and the tray in
the column, numbered from bottom to top, respectively.
Fi,k is the feed rate of the chemical species i to the tray
k, Ki,k is the vapor-liquid equilibrium coefficient for
component i on tray k (computed from Antoines equation), i,j is the stoichiometric coefficient of component i
in reaction j, and j,k is the extent of reaction j on tray
k. The vapor stream VN off the top tray is totally
condensed and returned to the tray:
J

Fi,N + VN-1Ki,N-1xi,N-1 - LNxi,N +

i,jj,1 ) 0

j)1

(A.2)

A portion (the reboiler boilup ratio) of the liquid


stream leaving the bottom tray is vaporized and fed
back to the tray:
J

L2xi,2 - (1 - )L1xi,1 - V1Ki,1xi,1 +

i,jj,1 ) 0

j)1

(A.3)

The extent of reaction j is calculated from the kinetic


equation

j,k ) Wkfj(xi,k,Tk)

(A.4)

where Wk is the liquid holdup volume on tray k and fj


is the kinetic rate of reaction j. The energy balance for
tray k becomes
J

Hvap(Vk-1 - Vk) -

Hjj,k ) 0

j)1

(A.5)

where Hvap is the heat of vaporization and Hj is the


heat of reaction j. Finally, the mole fractions of chemical
species in each liquid or vapor stream must sum to
unity:
I

xi,k ) 1, Ki,kxi,k ) 1

i)1
i)1

(A.6)

Equations A.1-A.6 contains N(I + J +3) + 1 variables


(N temperatures, JN extents of reactions, N liquid
streams, N vapor phase streams, IN liquid compositions,
and the RBR ). There are N(I + J + 3) equations in
the model. Therefore, the model (A.1)-(A.6) has a
degree of freedom, which we take as the RBR . Note
that 0 e e 1.
Literature Cited
(1) Doherty, M. F.; Perkins, J. D. On the dynamics of distillations process: IV. Uniqueness and stability of the steady state in
homogeneous continuous distillation. Chem. Eng. Sci. 1982, 37,
381.
(2) Lucia, A. Uniqueness of solutions to single-staged isobaric
flash process involving homogeneous mixtures. AIChE J. 1986,
11, 1761.
(3) Shridar, L. N.; Lucia A. Analysis and algorithms for
multistage separation processes: fixed temperature and pressure
profiles. Ind. Eng. Chem. Res. 1989, 28, 793.
(4) Jacobsen, E. W.; Skogestad, S. Multiple steady states and
instabilities in distillation. Implications for operation and control.
Ind. Eng. Chem. Res. 1995, 34, 4395.
(5) Gani, R.; Jorgensen, S. B. Multiplicity in numerical solution
of nonlinear models: separation processes. Comput. Chem. Eng.
1994, 18, S55.
(6) Pisarenko, Y. A., Epifanova, O. A.; Serafimov, L. A. Steady
states for a reaction-distillation column with one product stream.
Theor. Found. Chem. Eng. 1988, 4, 281.
(7) Nijhuis, S. A.; Kerkhof, F. P. J. M.; Mak, A. N. S. Multiple
steady states during reactive distillation of methyl tert-butyl ether.
Ind. Eng. Chem. Res. 1993, 32, 2767.
(8) Ciric, A. R.; Miao, P. Steady-state multiplicities in an
ethylene glycol reactive distillation column. Ind. Eng. Chem. Res.
1994, 33, 2738.
(9) Sneesby, M. G.; Tad, M. O.; Smith, T. N. Implications of
steady-state multiplicity for operation and control of esterification
columns. Inst. Chem. Eng. Symp. Ser. 1997, No. 142, 205.
(10) Sneesby, M. G.; Tad, M. O.; Datta, R.; Smith, T. N. ETBE
synthesis by reactive distillation. 2. Dynamics simulation and
control aspects. Ind. Eng. Chem. Res. 1997, 36, 1870.
(11) Gehrke, V.; Marquardt, W. A singularity theory approach
to the study of the reactive distillation. Comput. Chem. Eng. 1997,
Suppl. 21, S1006.
(12) Guttinger, T. E.; Morari, M. Predicting multiple steady
states in distillation: Singularity analysis and reactive systems.
Comput. Chem. Eng. 1997, Suppl. 21, S1001.
(13) Karpilovskiy, O. L.; Pisarenko, Y. A.; Serafinov, L. A.
Multiple solutions in single-product reactive distillation. Inst.
Chem. Eng. Symp. Ser. 1997, No. 142, 685.
(14) Kumar, A.; Daoutidis, P. Nonlinear control of a high-purity
ethylene glycol reactive distillation column. Proceedings of IFACADCHEM, Banff, Canada, July 1997; p 371.
(15) Ciric, A. R.; Gu, D. Synthesis of nonequilibrium reactive
distillation processes via mixed integer nonlinear programming.
AIChE J. 1994, 40, 1479.
(16) Doedel, E. J. AUTO: Software for Continuation and
Bifurcation Problems in Ordinary Differential Equations; Computer Science Department, Concordia University: Montreal,
Canada, 1986.
(17) Duff, I. S. MA28sA set of Fortran subroutines for sparse
unsymmetric linear equations; Harwell Report AERE-R.8730;
1980.
(18) Hauan, S.; Hertzberg, T.; Lien, K. M. Why methyl tertbutyl ether production by reactive distillation may yield multiple
solutions. Ind. Eng. Chem. Res. 1995, 34, 987.
(19) Holland, C. D. Fundamentals of Multicomponent Distillation; McGraw-Hill: New York, 1981.
(20) Matros, Y. S. Catalytic Process under Unsteady-State
Conditions; Elsevier: Amsterdam, The Netherlands, 1989.

Received for review July 2, 1998


Revised manuscript received November 6, 1998
Accepted November 9, 1998
IE980424Q

You might also like