Vous êtes sur la page 1sur 20

Computers and Chemical Engineering 25 (2001) 445 464

www.elsevier.com/locate/compchemeng

A generalized framework for computing bifurcation diagrams


using process simulation programs
Arjun Vadapalli, J.D. Seader *
Department of Chemical and Fuels Engineering, Uni6ersity of Utah, Salt Lake City, UT 84112, USA
Received 21 March 2000; received in revised form 18 December 2000; accepted 19 December 2000

Abstract
Interest in discovering multiple steady-state solutions for reaction and separation processes grew exponentially in the 1990s.
Process simulators like ASPEN PLUS, CHEMCAD, HYSYS, and PRO/II are designed to obtain at best one solution. Simulation
programs can find multiple solutions only by the expenditure of much effort. Here, a bifurcation technique using arclength
continuation is presented that can be incorporated as an add-in subroutine to a simulation program to automatically trace a
solution path, including turning points, to obtain multiple solutions with respect to a user-selected parameter. The technique is
illustrated with applications to the ASPEN PLUS process simulator. The algorithms are based on a predictor corrector
implementation where the predictors reside in add-in FORTRAN routines and the existing nonlinear equation solvers in ASPEN
PLUS equipment models serve as the correctors. Furthermore, the existing physical property packages in ASPEN PLUS are also
utilized. The method was tested successfully on an adiabatic CSTR example using the ASPEN RCSTR model, and homogeneous
azeotropic, heterogeneous azeotropic, and reactive distillation examples using the ASPEN RADFRAC model. Two of these
examples are presented here. In all four examples, the range of the selected bifurcation parameter covers a region that produces
three multiple solutions. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Multiple solutions; Bifurcation; Continuation

1. Introduction
Design of chemical processing systems that satisfy
economical and environmental requirements calls for
development of new design tools and instrumentation
for identifying the fundamental features of the processes. One of these features is the phenomenon of
multiple steady states, which is the characteristic of
numerous physiochemical processes such as chemical
reactions (catalytic and autocatalytic), mass-transfer
operations (distillation, extraction, absorption), and reactive mass-transfer processes. Multiple steady states
can take place at the levels of a single apparatus, a
group of functioning apparatuses, or even an entire
chemical process system.
Gani and Jrgensen (1994) define three types of
multiplicities. Output multiplicity, the most common
* Corresponding author. Tel.: +1-801-5816915; fax: + 1-8015859291.
E-mail address: j.seader@m.cc.utah.edu (J.D. Seader).

form, refers to the case where all input variables are


specified and two or more sets of output variables are
found. For example, in a distillation column, the molar
flow rate, feed composition, feed location, number of
stages, molar reflux rate, molar distillate rate, type of
condenser and column pressure profile are specified,
and two or more sets of product composition and
column profiles are found. Input multiplicity refers to
the case where one or more output variables are specified and multiple solutions are found for the unknown
input variables. Internal state multiplicity refers to the
case where multiple sets of internal conditions or profiles are found for the same values of the input and
output variables. All three types of multiplicities have
been found to occur in steady-state separation operations that are modeled with nonlinear algebraic and
transcendental equations. Multiplicities can also occur
in membrane and crystallization systems, in reaction
systems, and in certain cyclic separation operations,
such as adsorption, when modeled with nonlinear partial differential equations.

0098-1354/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0098-1354(01)00624-X

446

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

Usually, multiplicity occurs only for certain sets of


specifications and only over restricted ranges of certain
parameters. Often, sets of multiple solutions contain
both stable and unstable solutions. One or more of the
stable solutions is often clearly superior for practical
application and one or more of the other stable solutions is trivial or very undesirable for practical application. For design purposes, it is important to be aware
of the possibility of multiplicity and to discover all
multiple solutions within the practical domain of operating variables. Seader and Kumar (1994) discuss in
detail the reasons for the existence of multiple solutions, methods for obtaining them, and their practical
implications and significance.

2. Motivation
Chemical engineers use systems of equations to
model reactors and separation equipment. Frequently,
these equations, whether they be algebraic, transcendental and/or differential, are nonlinear in nature. Although linear algebraic equations have only one root
(or none if the coefficient matrix is singular), nonlinear
algebraic and/or transcendental equations can have
multiple roots. Except for a small number of special
cases (e.g. quadratic, cubic and quartic equations), a
direct analytical solution is not possible and one must
resort to an iterative numerical solution. The most
commonly applied numerical algorithms for systems of
nonlinear equations, as incorporated into steady-state
process simulators seek, at best, one solution. Additional steady-state solutions can sometimes be found by
making case studies, performing a sensitivity study on
one or more parameters, or varying initial guesses of
the solution. However, this procedure is not likely to
produce a smooth continuation or bifurcation curve
that includes turning points and all multiple solutions.
Simulator algorithms either require the user to provide
the initial guesses for the set of unknowns or provide a
built-in initialization procedure. In either case, an attempt is made to provide an initial guess that is within
the domain of a practical answer. This works in many
cases and fortunately, the solution obtained is often the
only practical solution. However, when multiple solutions exist in the practical domain, i.e. where the mole
fractions are between zero and one and flow rates and
absolute temperatures and pressures are positive and
real, it is desirable to have methods that find all
solutions.
In addition to iterative numerical methods, investigators have considered other mathematical tools that can
help detect multiple solutions. Some of the steady-state
process simulators (e.g. ASPEN PLUS, CHEMCAD,
HYSYS, and PRO/II) have algorithms that simultaneously adjust the recycle tear variables and the design

variables to satisfy the stationarity conditions. Unfortunately, they are of limited utility for the design of
processes that operate near or within regimes characterized by hysteresis and periodic or chaotic behavior.
They have no bifurcation algorithms that can trace
steady-state solution paths, perform a stability analysis
(evaluation of the eigenvalues of the Jacobian), and
locate the global optimum.
One method for finding multiple solutions, described
by Morgan (1987), but restricted to sets of polynomial
equations, takes the starting system of equations or a
partial reduction of it, and applies differential arclength
continuation using a special homotopy that establishes
parallel continuation paths that lead robustly to all
possible roots, real and complex. This method is widely
available in two software packages, CONSOL (Morgan, 1987) and POLSYS1H in HOMPACK90 (Watson,
Sosonkina, Melville, Morgan, & Walker, 1997). Alternatively the interval-Newton method in conjunction
with generalized bisection, as implemented in the public
domain software program INTBIS of Kearfott and
Novoa (1990) can locate multiple solutions of general
sets of nonlinear equations. This method has the advantage that the user restricts the search domain and then
INTBIS finds all of the roots within that domain. A
third method involves the use of bifurcation analysis
and singularity theory to locate higher-order singular
points in addition to limit points, such as real bifurcation points and Hopf bifurcation points, which homotopy algorithms fail to detect. AUTO ( Doedel, Wang,
& Fairgrieve, 1994) is one such package that can perform a bifurcation analysis. Doedel, Keller, & Kernevez
(1991) present an exhaustive discussion on the numerical analysis and control of bifurcation problems in
finite dimensions with several illustrative examples.
Bekiaris, Meski, Radu, and Morari (1993) used bifurcation analysis to compute the solution diagram for the
special case of homogeneous azeotropic distillation with
an infinite number of trays and infinite reflux ratio.
This method has come to be known as the infinity/infinity analysis. Results obtained from this limiting case
are then applied to the actual finite column case.
For all the above methods, investigators have written
specialized programs that include model equations,
thermodynamic routines, and in some cases, corresponding equation solvers to carry out the necessary
calculations. Also, simplifying assumptions have sometimes been made to successfully solve specific cases.
Often, these assumptions have been unsuitable for
other case studies. The aim here is to utilize existing
process models already present in simulators like ASPEN PLUS, thus saving the user the time of writing
and debugging separate and specialized codes to perform a bifurcation analysis. ASPEN PLUS was chosen
to illustrate the bifurcation methods developed here
because it is widely used and its architecture is well

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

447

suited for adding user-supplied subroutines. Before presenting the details of the bifurcation methods, it is of
interest to discuss the use of the sensitivity analysis tool
of ASPEN PLUS and its shortcomings when attempting to use it for locating multiple solutions.

an adiabatic CSTR is presented here to provide a better


understanding of the capability and limitation of the
sensitivity analysis tool.

3. Behavior of the sensitivity analysis tool of ASPEN


PLUS in the region of multiplicity

Consider the liquid-phase hydrolysis of propylene


oxide to propylene glycol in an adiabatic CSTR of fixed
volume. The influence of the inlet volumetric flow rate
of water on propylene oxide conversion and outlet
temperature is desired. This problem is solved independently with AUTO, using relevant model equations for
a CSTR and necessary physical properties, and is compared to the results obtained from a sensitivity analysis
performed with ASPEN PLUS using its built-in RCSTR model and physical property estimation models.
The reaction, which can be considered irreversible, is:

ASPEN PLUS uses sequential modular simulation,


where output streams for each operation in a process
are calculated from given, guessed, or previously computed input streams and equipment specifications for
the operation. The RCSTR and RADFRAC models of
ASPEN PLUS consist of equations that model a stirred
tank reactor and a distillation column, respectively.
Solutions of these equations depend on input stream
and equipment specifications. Bifurcation analysis of a
particular operation involves varying any one of the
input stream variables or equipment specifications (referred to as a bifurcation parameter) over a range of
values and observing its effect on output and/or state
variables. For example, in the RCSTR program of
ASPEN PLUS, material and energy balance equations,
including kinetic expressions, are solved to obtain exit
temperature and exit component molar flow rates. For
a given reactor volume, a change in the input molar
flow rate of any one of the reactants affects the values
of all computed variables.
Since neither bifurcation nor homotopy continuation
tools are present in any of the available commercial
simulators, the only way to conduct a parametric analysis of a process is to use a sensitivity analysis tool. A
sensitivity analysis of an exothermic reaction involving
the hydrolysis of propylene oxide to propylene glycol in

3.1. Adiabatic continuous stirred-tank reactor case


study

Propylene oxide (A)+ water (B)


Propylene glycol (C)

(1)

where the rate of the exothermic reaction is second-order with respect to propylene oxide, and the rate constant, k, can be expressed in Arrhenius form:
k= A e-E/RT = 9.15
1022 exp (18,700/Tr) m3/ kmol-s
(2)
where Tr, is the temperature in the reactor in K. The
problem specification is shown in Fig. 1. The feed
consists of a mixture of 1.3202 m3/h propylene oxide,
1.3202 m3/h methanol (M) and water. Methanol is
added in order to increase the bubble-point temperature
of the mixture. Varying the flow rate of water from 5.4
to 8.2 m3/h varies the total volumetric flow rate, while

Fig. 1. Simulation configuration of the adiabatic CSTR problem.

448

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

keeping the flow rates of propylene oxide and methanol


constant. The entering temperature of the feed streams
is 24C (75F) and reactor pressure is 1 atmosphere.
The volume of the reactor is 1.1356 m3 (40.1 ft3).

3.2. Comparison of ASPEN PLUS sensiti6ity analysis


with AUTO results
AUTO was used with the following material and
energy balance equations. Material balance equations
for components A, B and C assuming incompressible
flow:
(3)

QCi0 QCi RNET =0

where i is either component A, B, or C and RNET =


kC 2AVr. Q is the total volumetric flow rate in and out of
the reactor; Ci0 is the inlet molar concentration and Ci
is the exit molar concentration.
Energy balance equation for the adiabatic reactor:
3

XEB[DHR +DCP(Tr TR)] =% UiC0 Pi(Tr Ti0)

(4)

where, XEB is the fractional conversion predicted by the


energy balance equation; DHR is the heat of reaction at
reference temperature, TR; DCP is the average change in
molar heat capacity of the reacting mixture over the
temperature range of TR to Tr; Tr is the exit and reactor
contents temperature; Ui is the mole fraction of each
component in the combined feed; Ti0 is the inlet temperature of the combined feed; and C0 Pi is the average
heat capacity of each component in the inlet stream
over the temperature range of Ti0 to Tr.
To obtain an energy balance as consistent as possible
with ASPEN PLUS calculations, the following procedure was employed. The energy balance equation was
formulated with the help of the RSTOIC model of
ASPEN PLUS. RSTOIC computes the exit temperature for a specified conversion of a key component. For
a fixed set of input variables and equipment specifications, RSTOIC was solved to find exit temperatures for
different fractional conversions. Because conversion
was found to be close to a linear function of the exit
temperature, the energy balance equation was fitted to
the equation form XEB =mT + b. Energy balance lines
of this form were constructed for different total volumetric flow rates by changing the flow rates of water
alone while keeping the flow rates of propylene glycol
and methanol constant. The slope and intercept of the
energy balance plot varies as the total volumetric flow
rate changes because of a corresponding change in Ui.
Thus, slope (m) and intercept (b) are expressed as linear
functions of the total volumetric flow rate (Q). Eq. (4)
was reduced to:

(0.001379Q 0.001019)Tr + (0.7317Q 0.488)


+

RNET
=0
QCA0

(5)

Eq. (3) and Eq. (5) were solved by AUTO with initial
values of Q=8.0 m3/h, CA = 0.081 kmol/m3, CB =
34.76 kmol/m3, CC = 2.26 kmol/m3, and Tr =78C.
Since methanol does not participate in the reaction
mechanism, its mass balance equation was not included
with the rest of the mass balance equations. The four
nonlinear algebraic equations (3 mass balance and 1
energy balance) were solved to obtain the solution
vector u, which is made up of elements, CA, CB, CC and
T. The total volumetric flow rate, Q was the selected
bifurcation parameter.
The bifurcation package of AUTO generates an output file that contains parameter values (the volumetric
flow rate of water) and state variables (temperature and
concentrations), along with indicators for turning
points and algebraic bifurcation points. The sign next
to the step number determines the stability of a particular solution. Stable points are indicated with the ()
sign and unstable points by the (+ ) sign. In the
following example, turning points were found at entering water rates of 7.79 and 6.04 m3/h, with the absence
of any real bifurcation points. The bifurcation diagram
from AUTO for this problem is presented in Figs. 2
and 3 as solid lines. One observes a decrease in the
propylene oxide conversion from 96 to 65% as the
water flow rate increases from a starting point of 5.4
m3/h to the first turning point at 7.79 m3/h. Thereafter,
an unstable steady state is calculated and the conversion continues to fall at decreasing flow rates until the
second turning point is reached at 6.04 m3/h with a
conversion of 15%. Beyond the second turning point,
the flow rate again increases with a marginal drop in
conversion (15 4%) corresponding to a second stable
steady-state region. Calculations are terminated arbitrarily at a water flow rate of 8.2 m3/h. Thus, two stable
steady states and one unstable steady state exist for
total volumetric flows between the two limit points of
7.79 and 6.04 m3/h, which defines the region of multiplicity. Unique steady states exist outside the region of
multiplicity.
A sensitivity analysis run using ASPEN PLUS with
the RCSTR model was carried out for each of the two
options provided in ASPEN PLUS. In the first option,
current results are used as initial estimates for the next
step, which involved a new value of the volumetric
water flow rate. In the second option, results are reinitialized at each step, with ASPEN PLUS supplying a
new set of initial estimates to begin each step. In Fig. 2,
where no reinitialization was done at each step, the
initial water rate is 5.4 m3/h and is incremented by 0.1
m3/h for each sensitivity step. Open square markers
show the ASPEN PLUS results. Conversion decreases
gradually until the first turning point is approached at

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

449

Fig. 2. Comparison of ASPEN PLUS (with no reinitialization) and AUTO results.

Fig. 3. Comparison of ASPEN PLUS (with reinitialization at each step) and AUTO results.

a water flow rate of 7.8 m3/h. The next sensitivity step,


at a water rate of 7.9 m3/h, witnesses a precipitous
decrease in conversion to 4%. A second series of sensitivity steps starts from a water rate of 8.2 m3/h, and
decreases by increments of 0.1 m3/h. The ASPEN
PLUS results are shown as open triangle markers in
Fig. 2. Note that the conversion jumps from 4.0 to 90%
at a water flow rate of 6.6 m3/h, instead of 6.04 m3/h as
predicted by AUTO. All ASPEN PLUS results are
close to the AUTO results. Fig. 2 shows that the

ASPEN PLUS sensitivity analysis finds only one solution in the region of multiplicity and is not a reliable
tool to accurately predict turning points. In Fig. 3, the
results obtained from the second sensitivity option are
compared to the AUTO results. It can be seen that
ASPEN PLUS calculates the unstable solution sporadically at a few water flow rates. Therefore, with neither
sensitivity option is it possible to obtain a smooth
bifurcation curve with all the steady-state solution
branches in the region of multiplicity.

450

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

4. Development of arclength continuation algorithms for


tracking multiple solutions with ASPEN PLUS
The traditional way of finding all possible solutions
to a set of nonlinear equations has been to use a
homotopy or bifurcation program in which the user
supplies the model equations. In addition, property
estimation equations are necessary. The technique presented here uses the process model and the property
equations already present in the chemical process simulator, e.g. ASPEN PLUS, to achieve the goal of tracking all possible solutions in a region of multiplicity.
Solution branches are computed using an arclength
continuation technique to increment the selected bifurcation parameter rather than using sensitivity analysis,
which as shown in Figs. 2 and 3, is an unreliable tool
for tracking solutions along a particular branch. Both
natural arclength and pseudo arclength continuation
were explored. Both incorporate a predictor-corrector
method for tracking the observed variables with respect
to a bifurcation parameter. It was particularly important to provide close approximations to the output
variables so that the nonlinear equation solver of ASPEN PLUS could converge to the nearest solution. The
user supplies to ASPEN PLUS the predictor method as
an add-in, while Newtons algorithm, included in the
ASPEN PLUS simulator, is used for the corrector
method. The user-supplied predictor algorithm was
merged with ASPEN PLUS via the inline FORTRAN
option available in ASPEN PLUS.
The convergence behavior of the Newton solver in
the RCSTR model of ASPEN PLUS motivated the
selection of a natural arclength method over the
pseudo-arclength method. However, as discussed below, the latter method was used with the RADFRAC
model. The natural arclength method is used with the
RCSTR method because it solves the mass balance
equations first, with the converged results supplied as
initial estimates for solving the energy balance equation. Separate convergence parameters can be provided
for the mass balance and energy balance equations. The
NEWTON solver in RADFRAC for the classical
Naphtali Sandholm option solves the column-describing equations simultaneously with Newtons method.
Convergence is stabilized using the dogleg strategy of
Powell.
In AUTO it is assumed that a solution u0 =u(s0) is
known for some particular value of l0 =l(s0). With
AUTO the first step in bifurcation analysis consists of
determining the stationary solution branches. These are
solutions [u(s), l(s)] of the nonlinear system of
equations,
f(u, l)=0, u, f  Rn, l  R

(6)

where s denotes some parameterization. From the starting point, the branch is traced out in a stepwise manner

using the pseudo-arclength continuation technique of


Keller (1977). Assuming (uj-1, lj-1) and (u%j-1, l%j-1) have
been computed, this well-known technique determines
the next solution (uj, lj ) from the equations,
f(uj, lj )= 0,

(7a)

(uj -uj-1)Tu%j-1 + (lj -lj-1) l%j-1-Ds = 0

(7b)

where Ds is the step size along the branch and superscript T is the transpose. The advantage of using the
inflated system of Eq. (7) is its capability of computing
past limit points on a branch, because the Jacobian for
Eq. (7),

fu
u

f&0xFFFD;
, is not only nonsingular at a
&0xFFFD;

regular solution point of Eq. (6) where fu is nonsingular, but also at a simple turning point. Points where the
variable l% changes sign are located by a secant iteration scheme and are treated as potential turning points.
Once a turning point is detected, the solution branch is
continued in the new direction using pseudo-arclength
continuation. In order to implement this scheme into
ASPEN PLUS, it would be necessary to access the
derivative information fu and fl, where fu is the Jacobian matrix of all functions with respect to state variables u, and fl are the function derivatives with respect
to parameter l. Alternatively, the derivatives can be
calculated using a finite differencing scheme by directly
accessing the functions. Unfortunately, neither was possible with ASPEN PLUS and, therefore, an alternative
method was developed. The detection of a turning
point is treated the same for both algorithms as explained next in some detail. Branch switching is handled differently for the RCSTR and RADFRAC
models because the latter deals with equilibrium-stage
operations, whose temperature and composition profiles can be suitably exploited to enable the algorithm to
successfully negotiate around turning points.
The methods used for detecting turning points and
for performing branch switching are explained with the
help of Fig. 4, which is a schematic of a typical
bifurcation curve with more than one steady state,
where, u i is a component of the vector of state variables
u, and l is the bifurcation parameter. The variable u i
represents a state variable like temperature or component mole fraction, and l represents an input or equipment variable that is being varied over a range of
values. The S-shaped curve consists of three steadystate solution branches that change directions at the
two turning points, and hence between the turning
points there are three sets of state variables that satisfy
the process model. The first is between points S and A,
the second between A and D, and the third between D
and B. Solutions on branches SA and DA cease to exist
to the right of point A. Similarly, solutions on branches
AD and BD cease to exist to left of point D. For

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

parameter values less than at point E, there exists only


one unique solution, as also at points beyond B on
branch DB. The branches SA, AD, and DB can be
numbered as I, II, and III, respectively (or the reverse
notation can be used and the branches DB, AD, and
SA can be numbered as I, II, and III respectively).

4.1. Detecting a turning point in the absence of a


Jacobian matrix for both algorithms
Turning points are characterized by large changes in
state variables like temperature or component mole
fractions for a small change in the value of the bifurcation parameter. The relative change is measured as the
angle tan-1(Du i/Dl). For any two points on a solution
branch, the absolute value of the angle formed with the
X-axis will be anywhere between 0 and 90 degrees.
Unless the branch is steep with respect to the X-axis,
this angle is normally well below 90 degrees. At a limit
point, since the change in variable u i is large, the angle
is close to 90 degrees. This sudden and large change is
due to the fact that a point like B% in Fig. 4 does not
actually exist and the Newton solver converges to solution B, which lies on another solution branch, DB. The
change in variable u i defined by the difference AB is
greater than any possible change when on branch SA
for the same Dl. In the case of an adiabatic CSTR, this
change is observed for all exit concentrations and the
reactor temperature. For a distillation column, large
changes in distillate and bottoms temperature are observed. By thermodynamics, temperature and component mole fractions in the vapor and liquid are related
to each other, hence causing large changes in distillate
and bottoms product compositions.

451

The criteria for detecting turning points can thus be


summarized as:
lim tan 1(du/du)90deg

Du 0

(8)

4.2. The natural arclength method in the absence of a


Jacobian matrix
The continuation method makes use of the two most
recent converged solutions. The difference in the two
state variable vectors, uprevious and ucurrent, is divided by
the corresponding difference in the parameter values,
lprevious and lcurrent, to obtain the direction vector with
respect to the two solution sets. The following steps
summarize the method used to compute any one solution branch in the bifurcation diagram,
1. Step 1: Compute the direction vectors for all components of vector u from
du i/dl = (u icurrent-u iprevious)/ (lcurrent-lprevious)

(9)

This step is also required for detecting turning points as


discussed in the previous section.
2. Step 2: Compute the guess values u *inext for the next
solution along the branch from
u*inext = u icurrent + du i/dl (lnext-lcurrent)

(10)

where the subscripts pre6ious and current correspond to


the two most recent converged solutions, and subscript
next refers to the next converged solution along the
branch. The guess values u *inext are supplied to the

Fig. 4. Typical bifurcation curve for a chemical process.

452

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

nonlinear-equation solver and convergence is achieved


in a few steps. It is important to note that since an
adaptive step size strategy is implemented to achieve
the above steps, (lnext-lcurrent) is not equal to (lcurrentlprevious).
The RCSTR program allows the user to supply
initial estimates for flow rate and temperature variables,
which are calculated in step 2 above. These estimates
are called F-EST and T-EST respectively. In the region
of multiplicity, where more than one steady-state solution exists, the following convergence characteristics are
observed,
1. Providing estimates of component flow rates and
reactor temperature does not guarantee convergence
to a specific steady-state solution. For certain flow
rates, the steady-state solution obtained is not the
closest solution to a given set of initial estimates.
2. The temperature estimate alone is found to be sufficient to ensure convergence to a specific steady-state
solution. This behavior might be attributed to the
fact that in the RCSTR model the mass balance
equations are first solved in an inner loop, while the
energy balance equations are solved last in the outer
loop. The energy balance equations alone appear to
control the convergence of the complete system of
equations. Therefore, by only supplying different
initial temperature estimates for the reactor, it is
possible to obtain different steady-state solutions in
the region of multiplicity.
In the natural arclength algorithm developed for the
RCSTR program, a guess value for the new point C in
Fig. 4 is provided by assuming the existence of an
intermediate solution branch lying somewhere between
the two solution branches, SA and DB. The intermediate solution is found by looking in a direction opposite
to the direction traversed on branch SA. The main idea
is to provide a guess value that lies in the region of
convergence of point C, which is represented as an
elliptical region encircling point C in Fig. 4. This is
accomplished by searching at an angle less than 90
degrees in the opposite direction. This is implemented
by providing a guess value of the temperature halfway
between points A and B for a bifurcation parameter
value less than that at point A. The angle formed by the
search direction in this way makes an angle less than 90
degrees with the X-axis. The overall goal is to supply a
temperature estimate that lies in the region of convergence of the NEWTON solver of ASPEN PLUS. This
heuristic criterion was suggested after having performed
a number of tests on the algorithm. A potential problem with this approach is that the Newton solver in
ASPEN PLUS might possibly converge to a point on
either branch DB or SA. Hence, it might require more
than one attempt to successfully converge to point C.
Bifurcation diagrams for other variables like molar flow
rates or concentrations are expected to possess similar
characteristics.

Switching from an unstable branch to a stable one is


treated no differently than for the case of switching
from a stable to an unstable branch. After performing
a number of runs, it was found that providing a temperature estimate lower in value than at point D enables the solver to converge to a point on branch DB.
Similarly, providing a temperature estimate higher than
the temperature at point A causes the solver to converge to a solution on branch SA. The underlying
assumption is that any variable u i will increase or
decrease monotonically as one begins calculations from
branch SA or DB. This means that any solution on
branch SA will always be higher in value than any
solution on branch AD or DB.

4.3. The pseudo-arclength method in the absence of a


Jacobian matrix
The following method is explained with the help of a
homogeneous azeotropic distillation column studied by
Guttinger, Dorn, and Morari (1997) to separate the
ternary mixture methanol methyl butyrate toluene,
which demonstrated the existence of multiple steady
states caused by nonlinearities in vapor liquid equilibrium. Experiments on an industrial pilot-plant-size
column revealed two stable steady states for the same
feed flow rate and composition, and the same set of
operating parameters. Experimental measurements were
found to be in reasonable agreement with predictions
obtained from the solution of a model for infinite-stage
columns operating at infinite reflux ratios, referred to as
the infinity/infinity analysis. The experiments confirmed
the existence of output multiplicity, thus supporting
earlier predictions made with models by various researchers. Rigorous equilibrium-stage calculations, performed using the RADFRAC model of ASPEN PLUS
with the Wilson K-value option, confirm the existence
of the two stable steady-states.
The simulation configuration for the distillation
problem is shown in Fig. 5. The column, with 40
equilibrium stages plus a total condenser and partial
reboiler, operates at a reflux ratio of R= L/D: 2.0 to
2.5. Feed is introduced into the column at stage 22,
numbered from the condenser, at a constant flow rate
of 2.0 kg/h. The feed composition given in Fig. 5 is
similar to that used by Guttinger, Dorn, and Morari
(1997). The multiplicity range was found to cover a
distillate flow rate of 1.4 1.9 kg/h. The sensitivity
analysis tool was used initially to obtain the two stable
steady-state branches. Starting at a distillate flow rate
of 1.4 kg/h and performing calculations with the ASPEN PLUS NONIDEAL algorithm (to improve convergence), the distillate flow rate was incremented in
intervals of 0.1 kg/h until a flow rate of 1.9 kg/h was
reached. Results obtained from this run were stored
and the distillate flow rate was then decreased in incre-

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

453

Fig. 5. Simulation configuration of the homogeneous azeotropic distillation column.

ments of 0.1 kg/h from 1.9 to 1.4 kg/h. Significant


changes in distillate and bottoms compositions were
observed at distillate flow rates of 1.53 and 1.81 kg/h,
which corresponded to turning points like A and D in
Fig. 4. Solving the problem separately with AUTO,
Guttinger, Dorn, and Morari (1997) confirmed these
points as turning points and also calculated the unstable distillate and bottoms compositions. They, however, did not confirm the unstable branch with ASPEN
PLUS. Here, by using the NEWTON algorithm instead
of the NONIDEAL algorithm of RADFRAC, and by
providing initial estimates for temperature and for distillate and bottom compositions from the bifurcation
results of Guttinger, Dorn, and Morari (1997), the first
unstable solution was obtained at a distillate flow rate
of 1.70 kg/h. More points along the unstable branch
were obtained by using results of the previous run as
the initial estimate for a subsequent run.
The temperature and methanol composition profiles
of all three steady-states at a distillate flow rate of 1.58
kg/h are shown in Figs. 6 and 7 respectively, where they
are numbered I, II and III. Using the results presented
by Guttinger, Dorn, and Morari (1997), profiles I and
III are stable and profile II is unstable. In Fig. 6, the
stage temperatures along profile III of the column vary
by only 1C along the entire length of the column.
Profiles II and I vary from 65C for the distillate to
110C for the bottoms, with profile II changing
abruptly at the feed stage [22]. Other observations are:

1. All three steady-state profiles are distinct. The maximum differences in temperature and composition
for profiles III and I is found for the last few stages
at the bottom.
2. The temperature and composition values for the
first few stages from the top are nearly identical for
all three profiles.
3. The unstable profile II is characterized by temperatures that are close to the stable profile III for stages
above the feed stage. Stage temperatures for profile
III are lower than for the other two profiles for the
entire column. For stages below the feed stage [22],
stage temperatures of profile II increase suddenly
and approach those of profile I. Since the temperature transition seemed to occur close to the feed
stage, a study of the effect of feed stage location on
the unstable state profile was carried out for feedstage (f.s.) locations of 6, 11, 15, 22 and 28. Results
obtained from this study are presented in Fig. 8,
which shows a strong dependency of the unstable
profile on the feed stage location. This characteristic
behavior is exploited and used in the pseudoarclength algorithm to enable branch switching. A
similar behavior was also found for heterogeneous
azeotropic and reactive distillation examples.
The continuation method presented here also makes
use of the two most recent converged solutions. The
difference in the two state variable vectors uprevious and
ucurrent and in the corresponding parameter values is

454

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

used to obtain the pseudo-arclength with respect to the


two solution sets. The following steps compute any one
single solution branch, where ucurrent uprevious 2 is the
L2-norm of ucurrent uprevious.
3. Step 1: Compute the pseudo-arclength with vector u
and parameter l,
Ds 2 = ucurrent uprevious 2 +(ucurrent uprevious)2

(11)

4. Step 2: Compute the direction vectors with respect to


s and l,

du i/ds = (u icurrent-u iprevious)/Ds

(12)

dl/ds = (lcurrent-lprevious)/Ds

(13)

du i/dl = (u icurrent-u iprevious)/(lcurrent-lprevious)

(14)

Fig. 6. Temperature profiles of all three steady-states of the homogeneous azeotropic distillation column at a distillate flow rate of 1.58 kg/h.

Fig. 7. Methanol composition profiles of all three steady-states of the homogeneous azeotropic distillation column at a distillate flow rate of 1.58
kg/h.

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

455

Fig. 8. The unstable temperature profile for different feed stage locations of the homogeneous azeotropic distillation column.

where, du i/ds and dl/ds are required in step 3 for


calculating guess values for the next possible solution
and du i/dl is required for detecting turning points.
5. Step 3: Compute the bifurcation parameter and the
guess values ui* for the next solution along the branch,
u*inext =u icurrent +(du i/ds) Dsnext

(15)

lnext =lcurrent +(dl/ds) Dsnext

(16)

It is important to note that since an adaptive stepsize strategy is implemented to achieve this, Ds is not
equal to Dsnext, which is managed independently in the
algorithm with a maximum value supplied by the user
at the start of the run.
The RADFRAC model in ASPEN PLUS allows the
user to supply estimates of temperature and component
mole fractions for each stage as computed in step 3
above. The program variables for temperature estimates are called T-EST and those for liquid and vapor
mole fraction estimates are X-EST and Y-EST, respectively. In the region of multiplicity, where more than
one steady-state exists, it is possible to converge to a
specific steady-state solution by specifying both temperature and mole fraction estimates for some or all stages.
However, for a given set of initial estimates that are
close an actual solution:
1. The inside-out method with the default option may
converge to a solution different from the one originally sought, usually on the stable steady-state
branch.
2. The NEWTON option of the inside-out model is
more predictable and converges usually to the

nearest steady-state solution for a given set of initial


estimates. The odd convergence behavior of the
inside-out algorithm with the default option is possibly due to the fact that the model equations are
solved in two nested iteration loops. The NEWTON
option, on the other hand, solves the column equations simultaneously, using Newtons method. The
convergence of the NEWTON option is stabilized
using the dogleg strategy of Powell (1970).
All three branches can be calculated using the NEWTON option. When a limit point is exceeded, a new
solution on a different branch is obtained using the
default option of the inside-out model. The following
procedure assumes that the user will start most often on
a stable steady-state branch. This is a fairly good
assumption since the first solution point is obtained
using the default option of the inside-out algorithm,
which rarely converges to an unstable steady-state in
the region of multiplicity.
There are two possible cases of branch switching. In
the first, referring to Fig. 4, a transition is made from a
stable branch like either SA or DB to the unstable
branch AD. The characteristic feature of this case is
that the old and new solutions lie on different stable
steady-state branches and the unstable branch lies
somewhere in between these stable branches. If the
pseudo-arclength algorithm first calculates branch SA
in Fig. 4 and encounters a turning point, the default
option of the inside-out model is used to obtain point
B. Since point C lies on the unstable branch, a very
good initial estimate for point C is obtained by
combining column profiles at points A and B. Stages
above the feed stage are assigned values from the lower

456

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

steady-state profile and those below the feed stage are


assigned values from the higher steady-state profile.
This procedure was developed from the results obtained
from the homogeneous azeotropic distillation example.
With these initial estimates, the NEWTON option converges easily to point C.
When switching from an unstable branch to a stable
one, a different approach is applied. For example in
Fig. 4, if the unstable branch is being calculated, starting from point A, then the default option of the insideout model will converge to point E after passing point
D. Similarly, if the unstable branch is calculated starting from point D, then point B will be calculated after
passing point A. Since branches I and II are already
calculated, the new branch is searched for in a direction
opposite to that traversed on branch II. For the
pseudo-arclength algorithm, a shift in the column profiles is sufficient to make the NEWTON option converge
to a point on the new branch. If the default option
converges to a solution that possesses a lower steadystate profile, then the solution branch being sought is
the one with the higher steady-state profile. In such a
situation, the temperatures are shifted up, so that stages
above the feed stage have higher temperatures. A similar method is followed if the default option converges

to a higher steady-state profile. A brief discussion of the


implementation of the continuation algorithms into the
UNIX version of ASPEN PLUS, with reference to Fig.
9, is given in the appendix.

4.4. Merits of the continuation algorithm

1. The algorithms presented here make complete use of


a simulation program, thus saving the user the need
of supplying model equations and thermodynamic
routines.
2. The algorithms calculate the entire bifurcation diagram systematically, moving from one solution
branch to another. The existence of a new branch
can be confirmed by looking at the plot after calculating the first few points. If the algorithm retraces
an old solution curve, then the user can access the
stored information to restart calculations for the
new branch.
3. At any one time, the algorithms require only the
previously computed solution and the new solution.
Thus the program can be interrupted at any stage
and restarted from where it was last stopped.

Fig. 9. Implementation of the algorithms with ASPEN PLUS.

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

4. The algorithms calculate the bifurcation diagram


using an adaptive step-size algorithm. This allows
the user to start with a small step size and increase
the step size with each step until a user-supplied
maximum step size is reached.
5. Since ASPEN PLUS cannot compute values outside
the practical domain, a converged solution is guaranteed to exist in the practical domain.

4.5. Limitations of the algorithms

1. The algorithms cannot indicate immediately


whether the solution is stable or unstable. This
shortcoming can be overcome to some extent by
noting that the branches change stability at every
turning point. Therefore, a bifurcation diagram of a
particular process consisting of three branches can
have only two possible stability characteristics. The
three solution branches are either stable unstable
stable or unstable stable unstable. Most investigators have reported the former case.
2. The algorithms cannot detect exact bifurcation
points because the Jacobian generated by ASPEN
PLUS is not accessible to the user. Therefore, the
algorithms calculate only one solution path at any
one time. Homotopy continuation algorithms also
share this shortcoming.

5. Results and discussion


The algorithms discussed above were used to obtain
bifurcation diagrams for the two cases discussed above
(adiabatic CSTR and homogeneous azeotropic distillation). Vadapalli (1998) gives case studies for heterogeneous azeotropic distillation and reactive distillation
examples. The former deals with the dehydration of
ethanol using benzene as the entrainer and is performed
using the rigorous three-phase distillation algorithm
option available in the RADFRAC model of ASPEN
PLUS. The reactive distillation problem deals with the
effect of the methanol feed-stage location and the reflux
ratio on the conversion of methanol and isobutene to
methyl-tert-butyl-ether (MTBE) and its separation from
inert n-butene (1-butene). The two case studies presented here give converged results obtained from ASPEN PLUS, and deal with processes that exhibit three
steady states in the region of multiplicity. The branches
are numbered I, II and III in the order they are
calculated. The stability of the branches for some of the
cases were established by comparing results obtained
from independent calculations with the bifurcation
package AUTO.

457

5.1. Adiabatic CSTR


The natural arclength algorithm was used to calculate the bifurcation diagram for the adiabatic CSTR
problem described earlier. A maximum step length for
the bifurcation parameter (feed water flow rate) was
specified as 0.15 m3/h. A value larger than this is likely
to cause the algorithm to switch solution branches
before completing calculations on the original branch.
The minimum parameter step length was chosen based
on the degree of accuracy desired by the user to locate
a particular turning point. A value of 89.9 degrees for
the critical angle was sufficient to locate both turning
points. All exit component compositions and the exit
temperature satisfied this value of critical angle simultaneously. Calculations on a branch were started with a
value of the step length between the minimum and the
maximum, and the step length was increased by 20% at
each step until the maximum value was reached. The
kinetic rate law was available to the main program
either by linking a user-supplied FORTRAN subroutine or by directly providing these values in the input
data for the RCSTR model.
The computed bifurcation diagram of reactor temperature as a function of the volumetric flow rate of
feed water as the selected bifurcation parameter for this
problem is given in Fig. 10, where each point represents
a converged calculation by ASPEN PLUS. The algorithm is implemented in a continuous manner,
whereby the user does not interact with the execution of
the bifurcation program after supplying the required
inputs for the program. AUTO ascertained that
branches III and I are stable and II is unstable. As seen
in Fig. 10, turning points occur at 7.84 and 5.71 m3/h.
The temperature falls rapidly from 72.7 to 50.8C as the
volumetric flow rate of water increases along branch I
from the starting point of 5.87 m3/h until the first
turning point is encountered at 7.84 m3/h. Beyond that
point, the unstable steady state branch II is calculated
and the temperature continues to fall, as the flow rate
decreases, until the second turning point is reached at
5.71 m3/h. This turning point is less than the value 6.04
m3/h calculated with AUTO. Beyond the second turning point, the flow rate begins to increase with a
marginal drop in temperature (29.6 to 24C), and the
second stable steady state, branch III, is calculated.
Corresponding bifurcation diagrams of exit concentrations for propylene oxide, water, and propylene glycol
against the same bifurcation parameter are shown in
Figs. 1113 respectively. Figs. 10 and 13 look similar
because the production of propylene glycol depends on
the rate of the reaction, which in turn depends on the
temperature of the reactor. Concentration is highest for
volumetric flow rates less than 5.7 m3/h where the
temperature is higher than 70C, and is lowest when the
temperature is close to the feed temperature. Corre-

458

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

Fig. 10. Bifurcation diagram of exit reactor temperature for the adiabatic CSTR.

Fig. 11. Bifurcation diagram of the exit propylene oxide concentration for the adiabatic CSTR.

spondingly, the concentrations of both water and propylene oxide are the lowest when temperature is the
highest. All the diagrams presented in this section were
constructed from data points obtained from a single
run. In all there were 34 ASPEN PLUS simulations
carried out to solve this problem. The conversion plot
similar to Figs. 2 and 3 is shown in Fig. 14. In the
region of multiplicity, three operating points are possible. The maximum amount of propylene glycol is produced if the reactor is operated on branch I. It is best
to operate the reactor with an inlet water flow rate of

less than 5.7 m3/h because at least a 95% conversion is


guaranteed in that region. The reactor should not be
operated at water flow rates greater than 7.9 m3/h
because the conversion of propylene oxide is never
higher than 5%. In the region of multiplicity, the operating point depends on how the reactor is started up.
The authenticity of the calculated second turning point
was tested by running independent simulations of ASPEN PLUS at flow rates less than 5.71 m3/h. ASPEN
PLUS converged to the higher stable steady-state
branch for these tests. A similar procedure was carried

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

out for the turning point located at 7.84 m3/h. In this


case, flow rates greater than 7.84 m3/h were chosen. All
runs converged to the lower stable steady-state branch.
Hence, for the present problem, the algorithm successfully calculated the complete bifurcation diagram.

5.2. Homogeneous azeotropic distillation


The pseudo-arclength algorithm was used to solve
the homogeneous azeotropic distillation example discussed earlier. The choice of standard arclength, DS,

459

and the maximum arclength, DSMAX, were set to 1.0


and 2.0, respectively. This arbitrary selection of arclengths is similar to that used in AUTO. Because the
algorithm requires only the two most recently calculated solutions at any one time, it is possible to stop the
algorithm at any time during the computation of the
solution branch, make the necessary changes to the
program parameters, and then restart the calculations
from the last converged solution. Thus, the user does
not have to begin calculations from the first point
again. The 10 representative stages were chosen after

Fig. 12. Bifurcation diagram of the exit water concentration for the adiabatic CSTR.

Fig. 13. Bifurcation diagram of the exit propylene glycol concentration for the adiabatic CSTR.

460

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

Fig. 14. Bifurcation diagram of the exit propylene oxide conversion for the adiabatic CSTR.

carefully studying the temperature and composition


profiles presented in Figs. 6 and 7. Only a few important stages are selected instead of specifying initial
estimates of temperature and mole fraction for all
stages. The method is particularly useful for large
columns where specifying estimates for all stages can be
tedious to keep track of for each ASPEN PLUS run.
The solution of a distillation column model lies in
2MN + 5N +4M+ 5 space, where M is the number of
components and N the number of stages. Thus, for
large columns, many variables must be specified. By
using a number much smaller than N, the number of
required initial estimates is reduced significantly. The
scheme is effective because in most columns one stage
in a section of stages easily represents the change in the
entire section. Since the feed-stage location has a direct
influence on the shape of the unstable profile and in
turn enables branch switching, five stages above the
feed stage and five stages below the feed stage were
chosen from the 42 stages. This led to the selection of
stages 1, 5, 10, 15 and 22 above the feed stage and
stages 25, 30, 35, 40 and 42 below the feed stage. This
selection represented the column temperature and composition profiles well because the feed stage was located
almost midway between the first and last stages. The
selected bifurcation parameter in this particular problem was the distillate flow rate. A critical angle of 89.8
degrees was sufficient to locate both turning points. A
more stringent critical angle value could have been
used, but only at the cost of a smaller specified arclength and, hence, additional calculations along a particular branch.

The bifurcation diagrams of the bottoms and top


compositions compare well with those obtained from
Guttinger, Dorn, and Morari (1997). While they presented results on a mass basis, results here are on a
molar basis. Except for the bottoms composition of
methyl-butyrate, all other bifurcation diagrams calculated using the algorithm look similar to those presented by Guttinger, Dorn, and Morari (1997). For this
particular bifurcation diagram, shown in Fig. 15 in
terms of mole fractions, branches I and II intersect at a
distillate flow rate of 1.65 kg/h. Hence, the two solution
branches meet at two points. This however, is not
observed in Fig. 16, which is in terms of mass fractions
and where the two branches meet only at the turning
point of 1.81 kg/h. Multiplicities are known to exist
when some flows are specified on a mass basis (instead
of a molar basis) and are due to the nonlinear mass to
molar flow transformation, a type of multiplicity first
discussed by Jacobsen and Skogestad (1991). In this
particular problem, flow rates are expressed on a mass
basis and compositions on a molar basis. However, this
intersection is not observed for the other composition
variables. This kind of multiplicity was not reported by
Guttinger, Dorn, and Morari (1997).
Infinity/infinity and AUTO predictions made by Guttinger, Dorn, and Morari (1997) predict an overlap of
branches II and III for the distillate composition of
methanol shown in Fig. 17. Although the branches are
close to one another, our calculations show them as
distinct. This is also found to be true when plotted on
a mass fraction basis.
The bifurcation diagram for the bottoms temperature
is shown in Fig. 18. This diagram exhibits a very

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

interesting behavior along branch I as the turning point


of 1.81 kg/h is approached. When starting from a
distillate flow rate of 1.51 kg/h. The temperature rises
only slowly from 64 to 69C as the flow rate increases
to 1.78 kg/h. Thereafter, the temperature rises steeply
from 69 to 103.7C for a change of only 0.03 kg/h in
distillate flow rate. Because of this large change in
temperature for a relatively small parameter change,
the algorithm calculates smaller pseudo-arclengths and,
hence, the data points in this region are close to one

461

another. The large temperature increase can be explained by the material balance relationship between
internal vapor, reflux, and distillate flow rates. The
reflux flow rate is kept constant for the entire parameter
range and, by material balance, the vapor flow rate is
equal to the sum of the distillate and reflux flow rates.
Hence, the increased distillate demand can only be met
by an increase in internal vapor flow rate. This is
possible only by supplying more steam to the reboiler,
which in turn causes the bottoms temperature to rise.

Fig. 15. Bifurcation diagram of the methyl-butyrate bottoms composition for the homogeneous azeotropic distillation column.

Fig. 16. Bifurcation diagram of the methyl-butyrate bottoms composition for the homogeneous azeotropic distillation column (in terms of mass
fractions).

462

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

Fig. 17. Bifurcation diagram of the methanol distillate composition for the homogeneous azeotropic distillation column.

Fig. 18. Bifurcation diagram of the bottoms temperature for the homogeneous azeotropic distillation column.

The bifurcation diagram for the reboiler duty, shown in


Fig. 19, confirms this along branch I as the turning
point is approached.
A stability analysis was carried out by Guttinger,
Dorn, and Morari (1997), who found two stable steadystate branches with an unstable branch between the
two. Comparing the AUTO calculations performed by
Guttinger, Dorn, and Morari (1997) with those obtained from ASPEN PLUS, it is ascertained that
branches I and III are stable and branch II is unstable.
Column control might be difficult in the upper region

of branch I. Because this region is very close to branch


II, which is unstable, there is a likelihood that the
column might oscillate between the two steady states. If
so, it is advisable not to operate between distillate flow
rates of 1.77 and 1.81 kg/h.
As can be seen from all the bifurcation diagrams, the
rapid change observed in the bottoms temperature is
not observed in corresponding bifurcation diagrams of
distillate and bottoms compositions. If the bottoms
temperature were to be used as the sole criteria for
detecting turning points, the algorithm would be likely

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

463

to detect a turning point well before the true turning


point. It is for this reason, as explained above, that
more than one criterion is necessary to accurately detect turning points. Therefore, in addition to the bottoms temperature, distillate and bottoms mole fractions
of toluene and methyl-butyrate were used to evaluate
possible turning points. As in the case of the adiabatic
CSTR, the algorithm succeeded in calculating all three
branches predicted by AUTO and did better than the
sensitivity analysis run, where only the two stable
branches could be calculated. It is also worth noting
that branches I, II, and III do not appear in the same
order in each of the bifurcation diagrams presented in
this section. This further strengthens the need for such
an analysis of a process operating in the region of
multiplicity.

branches. The natural arclength algorithm developed


here is capable of starting at any position on the
bifurcation diagram.
Calculations for the homogeneous azeotropic distillation problem were initiated at a stable branch with the
pseudo-arclength algorithm. This is reasonable because
the user is most likely to use an inside-out algorithm,
such as RADFRAC, to find the first solution of the
problem and that algorithm almost always converges to
a stable steady-state solution. The pseudo-arclength
algorithm was successful in detecting turning points for
the homogeneous azeotropic distillation example at distillate flow rates of 1.53 and 1.81 kg/h. The turning
point at 1.81 kg/h was located with an accuracy of
0.0005 kg/h, while the turning point at 1.53 kg/h was
found with an accuracy of 0.007 kg/h.

6. Conclusions

Appendix A. Implementation of continuation algorithms


in ASPEN PLUS

Two continuation algorithms have been presented


that can be added into ASPEN PLUS to calculate
steady-state branches that exist in regions of multiplicity. For the adiabatic CSTR problem, the natural arclength algorithm is successful in detecting both turning
points and all three steady-state branches, as predicted
by the bifurcation package AUTO. The turning points
were located to a degree of accuracy of 0.02 m3/h and
can be improved upon by choosing a smaller value for
the minimum step length. Unlike the sensitivity analysis
results presented in Figs. 2 and 3, the continuation
algorithm does not ignore the presence of the intermediate unstable branch lying between the two stable

FORTRAN programs based on algorithm pseudocodes were linked with the different modules of
ASPEN PLUS via in-line FORTRAN, which can be
used to write FORTRAN statements and expressions in
design specifications, FORTRAN blocks, optimization
problems, sensitivity blocks and scaling reports. In-line
FORTRAN statements invoke most FORTRAN features, such as subroutine calls, I/O statements, array
variables, and labeled COMMONs. ASPEN PLUS
checks the FORTRAN code interactively as it is entered. Most syntax errors are detected before actually
running the simulation. ASPEN PLUS interprets most

Fig. 19. Bifurcation diagram of the reboiler duty for the homogeneous azeotropic distillation column.

464

A. Vadapalli, J.D. Seader / Computers and Chemical Engineering 25 (2001) 445464

in-line FORTRAN statements. FORTRAN that cannot be interpreted is compiled and dynamically linked
to the ASPEN PLUS module. Thus, the overhead for
in-line FORTRAN requiring compilation is small. In
addition to all the features mentioned above, in-line
FORTRAN also allows access to all stream and block
results. Program diagnostics like convergence results
that are not available from inline-FORTRAN are accessed using the Summary File Toolkit routines of
ASPEN PLUS.
To help run the continuation algorithms, three FORTRAN blocks named F-1, F-2, and F-3 were written.
Initially, the user does not use these blocks with either
the RCSTR or RADFRAC programs. As shown in
Fig. 9(a), the problem is specified in the standard
ASPEN PLUS format. After running the initial simulation, the bifurcation parameter and a set of observable
variables for the particular problem are chosen. The
appropriate interface needed to run the algorithms with
RCSTR and RADFRAC programs is created using
F-3. The interface consists of a set of files that carry the
guess values for the current step, the converged values
of the previous step and values of the program parameters like ISF, IBR, ISTEP, etc. This step is shown in
Fig. 9(b). After obtaining these files, F-3 is hidden from
the RCSTR/RADFRAC program using the Hide option for FORTRAN blocks. Next, blocks F-1 and F-2
are revealed to run as part of an overall flowsheet with
the ASPEN PLUS programs. As shown in Fig. 9(c)
block F-1 runs before the program and F-2 after. The
results of block F-2 are written into a set of output files.
These files are later copied into a set of input data files,
using a UNIX shell script. While the main job of F-1 is
to enter guess and parameter values for the next iteration, the arclength algorithms reside in block F-2. The
UNIX shell script calls ASPEN PLUS a number of
times until the entire bifurcation diagram is calculated.
Each iteration of the shell script represents one pass
through the flowsheet sequence F-1 (RCSTR or
RADFRAC) F-2. The parameter and initial estimate
values in the input file are replaced at each pass

through the flowsheet sequence. The add-in subroutines


for the UNIX version of ASPEN PLUS can be downloaded from the web site at: http://www.che.utah.edu/
 vadapall/C6.html

References
Bekiaris, N., Meski, G. A., Radu, C., & Morari, M. (1993)). Multiple
steady states in homogeneous azeotropic distillation. Ind. Engng
Chem. Res., 32, 2023.
Doedel, E. J., Keller, H. B., & Kernevez, J. P. (1991)). Numerical
analysis and control of bifurcation problems, (I) bifurcation in
finite dimensions. Intl. J. Bifurcation Chaos, 1, 493.
Doedel, E. J., Wang, X., & Fairgrieve, T. (1994). Auto94 software for
continuation and bifuration problems in ordinary differential
equations. In Applied Mathematics Report. Pasadena, California:
California Institute of Technology.
Gani, R., & Jrgensen, J. B. (1994). Multiplicity in numerical solution
of nonlinear models: separation processes. Comput. Chem. Engng,
18, S55.
Guttinger, T. E., Dorn, C., & Morari, M. (1997)). Experimental study
of multiple steady states in homogeneous azeotropic distillation.
Ind. Engng Chem. Res., 36, 794.
Jacobsen, E. W., & Skogestad, S. (1991). Multiple steady states in
ideal two-product distillation. AIChE Jl., 37, 499.
Kearfott, R. B, Novoa III, M. (1990). Algorithm 681, INTBIS, A
portable interval Newton/bisection package. ACM Trans. Math
Software, 16, 152.
Keller, H.B. (1977). Numerical solution of bifurcation and nonlinear
eigenvalue problems. Applications of Bifurcation Theory, P.
Rabinowitz, Editor, Academic Press, New York, 359.
Morgan, A. P. (1987). Sol6ing Polynomial Systems using Continuation
for Engineering and Scientific Problems. Englewood Cliffs, NJ:
Prentice Hall.
Powell, M.J. (1970). A hybrid method for solving nonlinear equations. Numerical Method for Nonlinear equations, P. Rabinowitz,
Editor, Gordon and Breach, London.
Seader, J. D., Kumar, R. (1994). Multiple solutions in process simulation. Paper presented at ASPENWORLD 94, 6 9 November,
1994, Boston, MA.
Vadapalli, A. (1998). Computing Multiple Solutions in Chemical Processes with Aspen Plus. PhD. Thesis in Chemical Engineering,
University of Utah, Salt Lake City, Utah.
Watson, L. T., Sosonkina, M., Melville, R. C., Morgan, A. P., &
Walker, H. F. (1997). Algorithm 777: HOMPACK90: a suite of
Fortran 90 codes for globally convergent homotopy algorithms.
ACM Trans. Math Software, 23, 514.

Vous aimerez peut-être aussi