Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
POLYMERIZATION REACTORS
SUBMITTED BY:
FIAZ AHMED TAHIR
(2002POLY1062)
MOHSIN ABBAS
(2002POLY1052)
SUBMITTED TO:
DR.JAVED RABBANI KHAN
DEPARTMENT OF CHEMICAL ENGINEERING
UNIVERSITY OF ENGINEERING & TECHNOLOGY
LAHORE.
Preface
Materials are more than mere components in technology; rather, the basic properties of
materials frequently define the capabilities, potential, reliability, and limitations of
technology itself. Improved materials and processes will play an ever increasing role in
efforts to improve energy efficiency, promote environmental protection, develop an
information infrastructure, and provide modern and reliable transportation and civil
infrastructure systems. Advances in materials science and engineering, therefore, enable
progress across and broad range of scientific disciplines and technological areas with
dramatic impacts on society.
Among these materials which have grown tremendously during last few decades are
synthetic polymers. Today, polymers are found in a large variety of products –e.g.,
automobiles, paints, and clothing, to name a few. Polymers have replaced metals in many
instances, and with the development of polymers alloys, applications in specialty areas
are certain grow. The new and highly specialized application of polymers, along with the
trend toward totally quality management and global competitiveness, has served to drive
up the quality expectations of the customer. These developments make it imperative to
operate the polymerization processes efficiently, which underscores the importance of
modeling and optimization of polymerization reactors.
In a polymerization reactor, raw materials are mixed at specified operating
conditions to produce polymer(s) having desired properties. The enduse properties of
interest include color, viscoelasticity , thermal properties, and mechanical properties
among others. To produce a polymer with such desired properties means that process
variables such as temperature, molecular weight, molecular weight distribution must be
tightly controlled. The manipulated variables available for controlling the variables of
interest at setpoints include the flow rates of raw materials and catalyst, temperature of
feed streams and temperature, and/or flow rated of heating/cooling mediums. Thus
mathematical modeling of polymerization reactors which relate molecular weight and
molecular weight distribution of polymers with manipulated variables is very important.
Being undergraduate students of polymer engineering we can’t model and optimize
the polymerization reactors in details because it is very difficult to model and optimize
the polymerization reactors specially on this level.
Now here is brief overview of our project.
We begin in chapter 1 with a brief overview of modeling, optimization, polymerization
techniques and polymerization reactors.
In chapter 2, we started with brief concepts of NACL, WACL, NAMW, WAMW and
MWD. We followed this with the discussion of chemistry and kinetics of various
polymerization reactions.
Chapter 3 gives the details of modeling, how to build a model, use of modeling, modeling
principles and how to model a polymerization reactor.
Chapter 4 is devoted to optimization in which we define objective function, variables,
constraints, mathematical relationships between these, definition of optimization
problems and optimization solution methodologies.
2
In Chapter 5 we shifted our attention from chemistry and kinetics of polymerization to
modeling of batch polymerization reactors. We developed models for anionic, free
radical and step growth polymerization and solved these using MATLAB programming.
Chapter 6 throw light on modeling of continuous stirred tank polymerization reactors.
The techniques used were anionic, free radical and step growth polymerization. Then we
solved these using MATLAB.
Chapter 7 is about optimization of polymerization reactors. First we have discussed what
is multi objective optimization and then we have written multi objective optimization of
polyester reactor.
We thank the Department of chemical engineering UET Lahore for their support in
this endeavor.
We pay special homage to our respective teacher Dr. Javed Rabbani Khan, who really
paid their special attention in completion of our project.
FIAZ AHMED TAHIR
MOHSIN ABBAS
3
TABLE OF CONTENTS
CHAPTERS PARTICULARS PAGE NO.
CHAPTER 1 INTRODUCTION
1.1: Modeling 6
1.2: Optimization 7
1.3: Classification of polymerization reactions 7
1.4: Polymerization reactors 8
CHAPTER 2 POLYMER REACTION ENGINEERING
2.1: Molecular weight and molecular weight distribution 9
2.2: Kinetics of anionic polymerization 10
2.3: Kinetics of cationic polymerization 13
2.4: Kinetics of free radical polymerization 14
2.5: Kinetics of step growth polymerization 16
2.6: Kinetics of copolymerization 17
2.7: Polymerization reactors 19
2.8: Reactor selection
CHAPTER 3 MODELING
3.1: What are models? 20
3.2: Reasons for developing models 20
3.3: General modeling principles 21
3.4: Classification of models 21
3.5: How to build a model 22
3.6: Modeling of Polymerization reactors 23
CHAPTER 4 OPTIMIZATION
4.1: How to define a model for optimization? 27
4.2: What makes a model hard to solve? 29
4.3: Mathematical relationships 30
4.4: Optimization solution methodologies 32
4.5: Algorithm solutions to optimization problems 32
4
CHAPTER 5 MODELING OF BATCH POLYMERIZATION REACTORS
5.1: Anionic polymerization 36
5.2: Free radical polymerization 38
5.2.1:Model for free radical batch polymerization
reactor 39
5.2.2:MATLAB solution of model 40
5.3: Step growth polymerization 45
5.3.1:Step growth polymerization in absence of catalyst 46
5.3.2:Model for step growth polymerization reactor
in absence of catalyst 47
5.3.2.1:MATLAB solution of model 48
5.3.3:Step growth polymerization in presence of catalyst 53
5.3.4:Model for step growth polymerization reactor
in presence of catalyst 54
5.3.4.1:MATLAB solution of model 55
5.3.5:Comparison of catalyzed and non catalyzed reactions
56
CHAPTER 6 MODELING OF STIRRED TANK POLYMERIZATION
REACTORS
6.1: Anionic polymerization 58
6.1.1:Model for stirred tank anionic polymerization
reactor 59
6.2: Free radical polymerization 60
6.2.1: Model for continuous stirred tank free radical
polymerization reactor 61
6.2.2: MATLAB solution of model 62
6.3: Step growth polymerization 71
6.3.1: Model for continuous stirred tank step growth
polymerization reactor 72
6.3.2:MATLAB solution of model 73
6.4:Reactor dynamics 77
CHAPTER 7 OPTIMIZATION OF POLYMERIZATION REACTORS
7.1: What is multiobjective optimization? 81
7.2: Optimization of polyester reactor 82
5
CHAPTER 1
INTRODUCTION
Synthetic polymers have grown tremendously during last few decades. Today, polymers
are found in a large variety of products ranging from common to very specialized
applications. The new and highly specialized application of polymers, along with the
trend toward totally quality management and global competitiveness, has served to drive
up the quality expectations of the customer. These developments make it imperative to
operate the polymerization processes efficiently, which underscores the importance of
modeling and optimization of polymerization reactors.
In a polymerization reactor, raw materials are mixed at specified operating
conditions to produce polymer(s) having desired properties. The enduse properties of
interest include color, viscoelasticity , thermal properties, and mechanical properties
among others. To produce a polymer with such desired properties means that process
variables such as temperature, molecular weight, molecular weight distribution must be
tightly controlled. The manipulated variables available for controlling the variables of
interest at setpoints include the flow rates of raw materials and catalyst, temperature of
feed streams and temperature, and/or flow rated of heating/cooling mediums. Thus
mathematical modeling of polymerization reactors which relate molecular weight and
molecular weight distribution of polymers with manipulated variables is very important.
1.1 Modeling
Modeling is “The representation of a physical system by a set of
mathematical relationships that allow the response of the system to various
alternative inputs to be predicted”.
Reasons for developing process models are that we can improve or understand
chemical process operations, improve the quality of produced products, increase
the productivity of existing and new processes, for operator training and process
design etc.
General modeling principles are Steady state modeling, Dynamic modeling and
Constitution relationships.
Models can be classified as. Theoretically based vs. empirical, Linear vs.
nonlinear, Steady state vs. unsteady state, Lumped parameter vs. distributed
parameter and Continuous vs. discrete variables
6
1.2 OPTIMIZATION
Reactors are often the critical stage in a polymerization process. Recently, demands on
the design and operation of chemical processes are increasingly required to comply
with the safety, cost and environmental concerns. All these necessitate accurate
modeling and optimization of reactors and processes. To optimize a reactor, first we
define an objective function, constraints, variables. Then we develop a mathematical
relationship between these. Then we solve these by different optimization solution
methodologies.
1.3 Classification of Polymerization reactions
Polymerization reactions are classified as homogeneous Polymerization and heterogeneous
Polymerization.
In homogeneous polymerization common techniques are bulk polymerization and solution
polymerization. And in heterogeneous polymerization are emulsion polymerization,
suspension polymerization, precipitation polymerization and solidphase polymerization.
We will focus our attention mostly on bulk and solution polymerization techniques.
On the basis of kinetics main classification of polymerization reactions are cationic,
anionic, free radical and step growth or condensation polymerization.
1.4 Polymerization Reactors
Polymerization reactors can be classified by the phase involved in the reaction.
Classification of Polymerization Reactors
Continuous Phase Dispersed Phase Type of
Polymerization
Polymer solution None Homogeneous bulk
or solution pzn
Polymer solution Any (e.g.
condensation
product)
Heterogeneous bulk
on solution pzn
Water or other non
solvent
Polymer or polymer
solution
Suspension,
dispersion or
7
emulsion pzn
Liquid monomer Polymer (swollen
with monomer)
Precipitation or
slurry pzn
Gaseous monomer Polymer Gas phase pzn
A polymerizer reactor will be heterogeneous whenever the polymer is insoluble in
the monomer mixture from which it was formed. If the polymer is soluble in its
own monomers, a dispersed phase polymerization requires the addition of a non
solvent (typically water) together with appropriate interfacial agents. For high
volume polymers like high volume chemicals continuous operation is generally
preferred over batch.
In a batch reactor feed is entered and product is removed in batches.
In a semi batch reactor initiator or monomer is added continuously and product is
removed in batches.
Tubular reactors are occasionally used for bulk, continuous polymerizations. A
monomer or monomer mixture is introduced at one end of the tube and if all goes
well, a high molecular weight polymer emerges at the other.
Continuous stirred tank reactors are widely used for bulk, free radical
polymerizations. The details for polymerization reactors and their kinetics are
discussed in the next chapters.
8
CHAPTER 2
POLYMER REACTION ENGINEERING
While polymerization and the reactions of polymers are in many respects similar to ordinary
chemical reactions, there are some significant differences that make the former unique in the
sense of reactor and reaction engineering. These are high viscosities and low diffusion rates
associated with concentrated polymer solutions and polymer melts. In polymerization reactors
we have to control the molecular weight and molecular weight distribution to achieve good end
use product properties. These properties include color, viscoelasticity, thermal properties and
mechanical properties. To produce a polymer with such properties means that process variable
such as temperature, molecular weight, molecular weight distribution and mooney viscosity
must be tightly controlled. In this chapter, we will briefly explain about degree of
polymerization, molecular weight and molecular weight distribution and also will explain the
kinetics of various types of polymerization reactions.
2.1: Degree of polymerization, molecular weight and molecular weight distribution
Degree of polymerization:
The no of repeat units per chain is known as degree of
polymerization and it is denoted by x, it is also known as length of polymer chain.
Molecular weight:
Molecular weight of given chain is defined as degree of
polymerization times the molecular weight of repeat unit.
We have defined the degree of polymerization for a single polymer molecule. But not all
polymer molecules within a reactor have the same degree of polymerization. Rather, a
polymer produced in a single reaction exhibits a distribution of chain lengths (degree of
polymerization). The distribution of chain lengths within a polymeric material may will be the
most important factor in determining its enduse properties. Therefore, it will be necessary to
develop a method of describing the distribution of chain lengths in a polymeric material.
As a single reaction exhibits a distribution of chain lengths, the mean of this distribution is the
number average chain length (NACL), this is the weight chain length distribution, and its
average is the weight average chain length (WACL). Respectively, there is the number
9
molecular weight distribution and the weight molecular weight distribution. Their averages are,
respectively, the number average molecular weight (NAMW) and the weight average molecular
weight (WAMW). The values of NAMW and the WAMW will not be necessarily the same.
This is because both are single number attempts to represent an entire distribution. It should be
noted that the enduse properties of a polymer are determined by the distribution of molecular
sizes which is independent of the average used to characterize it. If P
n
(the concentration of
chains containing n polymer units) is known for all values of n, the various averages may be
calculated as follows
∑nP
n
NACL= µ
n
= (2.1)
∑P
n
∑n
2
P
n
WACL= µ
w
= (2.2)
∑nP
n
∑(nw)P
n
NAMW= m
n
= (2.3)
∑P
n
∑(nw)
2
P
n
WAMW= m
w
= (2.4)
∑(nw)P
n
Molecular weight distribution :
Molecular weight distribution is defined by
polydispersity D which is given as m
w
µ
w
D =
= (2.5)
m
n
µ
n
Inspection of Eqs. (2.1) through (2.5) reveals that the polydispersity takes a value of 1 for a
monodisperse sample (one in which all of the chains are the exact same length). For any other
distribution of chain lengths, the polydispersity will be greater than 1. On the other hand, the
polydispersity varies or slightly with average chain length.
Variations in degree of polymerization (and hence in molecular weight) occur for at least three
reasons. The main mechanism by which the molecular weight distribution is broadened is
through the nature of the seriesparallel reaction mechanisms leading to chain formation.
Second mechanism is that of spatial or temporal variations in reaction conditions during
polymerization. Variations in temperature, monomer concentration, etc in any reactor, and in
residence time in a continuous reactor, affect the individual chain lengths. The final mechanism of
variation in degree of polymerization that of stochastic variations reaction rates on a molecular
level. This however has been shown to be insignificant in relation to the previous two. The import
concept, then, is that a distribution of chain lengths will result due to the nature of the reaction
10
mechanisms, even when all environmental variables (temperature monomer concentration, etc.)
are kept constant.
2.2: Kinetics of Anionic Polymerization
Addition polymerization can be carried out by a number of mechanisms. The free radical
mechanism is commercially predominant, but addition polymerization is often carried out by
anionic and cationic mechanisms. Anionic polymerization takes place via the opening of a carbon
double bond on the monomer unit. Initiation takes place with the addition of a negative ion to
the monomer, resulting in the opening of a double bond and growth at the end bearing the
negative charge. Propagation proceeds by addition of monomer units with the carbanion
remaining with the propagating chain end. Termination of a growing chain usually involves
transfer, and only results in the net loss of a growing chain if the new species is too weak to
propagate. Because termination usually involves transfer to some impurity in the system, it is
possible, with carefully purified reagents, to carry out polymerization in which termination is
lacking. The resulting species are termed living polymers and may result in extremely narrow
(essentially monodisperse) molecular weight distributions.
Anionic polymerization is employed with vinyl monomers containing electronwithdrawing
groups such as nitrile, carboxyl, phenyl, or vinyl in an aprotic nonpolar solvent. It is
characterized by high rates of polymerization and low polymerization temperatures. Strong
bases such as alkyl metal amides, alkoxides, alkyls, hydroxides, and cyanides are often used to
form the original carbanion.
(2.6)
AM
n

+M AM

n+1
Propagation (2.7)
AM
n

+B AM
n
+B

Chain transfer (2.8)
Here A

is the anion initiating the polymerization, B is the chain transfer molecule, and B

is the new anion formed by chain transfer, which may or may not be capable of initiation of a
new chain. The mechanism can be written in a form which is more concise and consistent with
the subsequent treatment for free radical polymerization as
K
AC A

+ C
+
Rate of initiation is given as(R
I
)
Initiation
k
i
R
I=
k
i
A

M (2.9)
A

+ M P
1
Initiation
11
Rate of propagation is given as(R
P
)
kp
P
n
+ M P
n+1
Propagation R
P
= k
p
P
n
M (2.10)
k
f
Rate of chain transfer
P
n
+ B M
n
+ B

Chain transfer R
T
= k
f
P
n
B (2.11)
Here P
n
is taken to mean AM

and M
n
represents AM
n
. For very fast reactions, concentration
of reactive species (in this case, ionic chains) becomes essentially constant very early in the
reaction. For this to happen here, the rates of initiation and chain transfer must reach steady
state quickly be equal. This is known as the quasisteadystate approximation (QSSA).
Based on the mechanism above and making the QSSA for P
n
, the rate of polymerization
can be written as
At steady state
Rate of initiation = Rate of termination
k
i
A

M = k
f
P
n
B
k
i
A

M
P
n
= (2.12)
k
f
B
Putting the value of P
n
in Eq.(2.10), so
_
kp k
i
A

M
2
R
P
= (2.13)
k
f
B
Rate of formation of ion A

is given as
rA

= KAC  KA

C
+
+ k
i
A

M
At steady state (rA

=0)
Rearranging KAi
A

= (2.14)
KC
+
k
i
M
Putting value of A

from Eq. 2.14 in Eq. 2.13 gives
K kp k
i
AC
M
2
R
P
= (2.15)
K k
f
B C
+
k
i
k
f
BM
By putting K k
f
= k
f
and also k
i
k
f
=0 , final eq. becomes
12
K kp k
i
AC
M
2
R
P
=
(2.16)
k
f
B C
+
QSSA is only valid for significant chain transfer to an unreactive anion, B

. In the
absence of
rapid chain transfer (or of chain transfer to a nonpropagation anion), the rate of
polymerization will continue to rise as the total number of living chains increases until
initiation is complete. Initiation is complete when all of the catalyst has been consumed;
from this point on, the number of live chains will remain constant.
The instantaneous degree of polymerization may be written as the rate of propagation divided
by the rate of chain transfer (rate of production of dead chains).
kp M
x = (2.17)
k
f
B
2.3 CATIONIC POLYMERIZATION
Cationic polymerization is another mechanism of addition polymerization. It proceeds
through chain propagation via a carbonium ion with the opening of a double bond on the
monomer unit as with anionic polymerization. The carbonium ion is formed by the reaction of
a strong Lewis acid (catalyst) with a weak Lewis base (co catalyst) followed by attack on the
doublebonded monomer unit. Termination via terminal doublebond formation and chain
transfer to monomer and polymer are dominant.
C'ationic polymerization is carried out with vinyl monomers containing electron
releasing groups such as alkoxy, phenyl, and vinyl. The system is characterized by very high
rates of polymerization. The mechanism of cationic polymerization may be written as
Initiation
(2.18)
Propagation (2.19)
Termination (2.20)
Chain transfer (2.21)
Here A is the catalyst and RH is the cocatalyst. These two species react to form the
catalystcocatalyst complex in Eq. (2.18). This complex donates a proton to the monomer,
forming a carbonium ion. Because cationic polymerization is usually carried out in a
chlorinated hydrocarbon solvent of low dielectric constant, the anion ( AR

) cannot be
separated from the carbonium ion. Rather, the two form an intimate ion pair. Propagation
takes place by the addition of monomer to the growing chain end [Eq. 2.19]. Termination
occurs with the formation of a terminal double bond and the regeneration of the catalyst
cocatalyst complex [Eq. (2.20)]. Chain transfer to monomer takes place as shown in Eq.
13
− + − +
− + − +
− +
+
− +
− + − +
− +
+ ÷→ ÷ +
+ ÷→ ÷
÷→ ÷ +
÷→ ÷ +
÷→ ← +
AR HM M M AR HM
AR H M AR HM
AR HM M AR HM
AR HM M AR H
AR H RH A
n
k
n
n
k
n
n
k
n
k
K
f
i
p
i
1
(2.21). The mechanism can be written in a form which is more concise and consistent with the
previous notation as
Initiation (2.18)
Propagation (2.19)
Termination (2.20)
Chain transfer (2.21)
Here P
n
is taken to mean HM
+
n
AR

Based on the mechanism above, the rate of polymerization may be written as the product of
the propagation rate constant, the monomer concentration, and the concentration of live
chains (P). By applying quasisteadystate approximation for Pn and following the same
procedure as in anionic polymerization on equations (2.18) to (2.21) the rate of
polymerization can be written as
K kp k
i
A(AH)
M
2
R
P
=
(2.22)
k
t
If, as is often the case in ionic polymerization, termination is negligible, the QSSA is not
applicable and the term after the second equality in Eq. (2.22) cannot be used. In this case,
the rate of polymerization will continue to rise as the total number of living chains
increases until initiation is complete. Initiation is complete when all of the catalyst has
been consumed; from this point on, the number of live chains will remain constant. The
instantaneous degree of polymerization may be written as the ratio of propagation to the sum
of the rates of termination and transfer:
M k k
M k
PM k P k
PM k
x
f t
p
f t
p
+
·
+
·
Thus, if transfer predominates, the degree of polymerization is a function only of temperature
(through k
p
/k
f
}. If termination predominates and if the activation energy for termination is
greater than the sum of the activation energies for initiation and propagation (as is often the
case), both the rate of polymerization and the degree of polymerization increase with
decreasing temperature. These conditions are the reverse of those found in free radical
polymerization and allow the attainment of high molecular weight.
14
1
1
1
P M M P
AR H M P
P M P
P M AR H
AR H RH A
n
k
n
n
k
n
n
k
n
k
K
f
i
p
i
+ ÷→ ÷ +
+ ÷→ ÷
÷→ ÷ +
÷→ ÷ +
÷→ ← +
− +
+
− +
− +
2.4 Free Radical Polymerization
Free radical polymerization is the most common of all addition polymerization mechanisms.
When free radicals are generated in the presence of unsaturated monomers, the radical adds
to the double bond and the resultant unpaired electron generates another radical. This radical
is free to react with another monomer unit, and in this way the polymer molecule grows by
adding monomer units while maintaining a free radical at the reactive end of the live
(growing) chain. Chain growth continues until the radical is terminated or transferred to
another chain. The complete mechanism can be written as follows:
I
÷→ ÷
d
k
2R Initiation (2.23)
M+R ÷ → ÷
ki
P
1
(2.24)
P
n
+M ÷ → ÷
kp
P
n+1
Propagation (2.25)
P
n
+P
m
÷ → ÷
ktc
M
n+m
Termination by combination (2.26)
P
n
+P
m
÷ → ÷
ktd
M
n
+M
m
Termination by disproportionation (2.27)
Pn+M ÷ → ÷
d
k
M
n
+P
1
Chain transfer to monomer (2.28)
P
n
+S ÷→ ÷
d
k
M
n
+S Chain transfer to solvent (2.29)
P
n
+T ÷→ ÷
d
k
M
n
+T Chain transfer to transfer agent (2.30)
P
n
+M
m
M
n
+P
m
Chain transfer to polymer chain (2.31)
P+In Q Inhibition (2.32)
This complex set of reactions may be divided into initiation, propagation, termination,
and chain transfer reactions. The rate of polymerization may be derived by applying mass
action kinetics to the elementary reactions in eqns (2.23)to(2.32).
P M R
ki
÷→ ÷ + (2.33)
1 P M P
kp
÷→ ÷ + (2.34)
2 P P P
kt
÷→ ÷ + (2.35)
By applying mass balance for R and P from equation yields
RM k fI k dt dR
i d
− ·2 /
(2.36)
(2.37)
As in the free radical polymerization, the propagation step is very short as compared to
termination and initiation so in above equation propagation term is neglected.
2
/ P k RM k dt dP
t i
− ·
15
2
/ P k kpPM RM k dt dP
t i
− − ·
At steady state, dP/dt=0 and also dR/dt=0
By applying this and solving above equations give
t
i
t
i
k
RM k
P
k
RM k
P
·
·
2
(2.38)
Finally the rate of polymerization can be written as
t
i
p p P
k
RM k
M k PM k
dt
dM
R · · − ·
(2.39)
In the case where inhibition is significant, P is given by
2 / 1
/ 1
1 2
∧
,
_
¸
¸
+
,
_
¸
¸
·
M k In k k
I fk
P
i in t
d
(2.40)
The instantaneous degree of polymerization can be defined as the rate of propagation divided by
the rate of production of dead chains (the sum of the rates of all reactions leading to dead
chains):
PT k PS k MP k P k P k
MP k
x
ft fs fm td tc
p
+ + + +
·
2 2
2 / 1
2.5 StepGrowth Polymerization
Stepgrowth polymerization involves reaction of functional groups on adjacent monomer
molecules with the evolution of water or other lowmolecularweight byproducts. The
reaction is stepwise or stepgrowth in the sense that the reaction of each functional group is
essentially independent of previous condensation reactions. There are no activated species
as in addition polymerization.
Condensation polymerizations are of two general types. AB type and AA/BB type
Experimental observation of stepgrowth polymerization yields the following general
characteristics: early disappearance of the monomer, absence of any high polymer during the
early stages of reaction, and equilibrium between polymerization and depolymerization
reactions. These observations suggest a mechanism of linear condensation in which monomer
molecules react to form dimers, the dimers react with each other to form tetramers (or with
other oligomers to form larger oligomers), and the tetramers react with other oligomers to form
longer chains. Thus, in this stepgrowth mechanism, the monomer disappears rapidly as it is
converted to a dimer. A great deal of lowmolecularweight material is formed early in the
16
reaction, and the average chain length grows slowly as the polymer chains condense to form
longer chains. Because the condensation reaction is reversible, the polymerization is always
in equilibrium with the depolymerization reaction (hydrolysis). The depolymerization can be
controlled by continuously removing the water (or other byproduct) of condensation, thus driving
the polymerization to completion. The validity of this sort of polymerization mechanism has
been verified for a large number of linear condensation polymerizations.
Rate expressions for stepgrowth polymerization can be written from massaction
kinetics once the mechanism is understood. The condensation is catalyzed by acids. Thus,
for a driven system, the rate of polymerization in the presence of an acid can be written as
A+B+H
+
P+C (2.41)
Where P is Polymer formed and C is a small molecule which is condensed out, A,B
monomers and H
+
is acid which is used to catalyze the reaction.
Rate of Polymerization can be written as:
+
· − · kABH
dt
dA
R
p
(2.42)
For a stoichiometric ratio of A and B, and assuming the acid concentration to be constant
over the reaction, the rate of polymerization may be simplified to
RP =
dt
dA
− = k' A
2
( 2 . 4 3 )
In the absence of added strong acid, an acid functional group on the monomer can catalyze
the reaction. The kinetics then becomes
RP =
dt
dA
− = k ”A
2
B ( 2. 44)
Wher e A r epr es ent s t he aci di c f unct i onal gr oup. For a s t oi chi omet r i c
r at i o of f unct i onal gr oups t hi s becomes
RP =
dt
dA
− = k ”’ A
3
( 2. 45)
The progress of the polymerization reaction can be quantified by introducing the extent of
reaction, p, defined as the fraction of A or B functional groups which have reacted at time t.
The number average chain length is given by the total number of monomer molecules initially
present divided by the total number of molecules present at time t, which can be related to
the extent of reaction as follows:
P P) ( N
N
N
N
μ
n
−
·
−
· ·
1
1
1
0
0 0
( 2. 46)
Inspection of eq. (2.46) will indicate that to obtain the necessary high number average
chain length, the extent of reaction must be well above 0.99.
2.6 COPOLYMERIZATION
One of the most important polymerization techniques is co polymerization. In this type of
polymerization two monomers with different functional groups are reacted. The
17
mechanism of co polymerization is same as that of polymerization. The steps in co
polymerization are initiation propagation and termination, as for other polymerization
techniques. The mechanism is as under
A.
I.
B.
C.
The I. Symbol is for radical. The rate of decomposition is given by (R
I
= k
I
I).The free
radicals (/.) react wi t h monomer molecules to form chain radicals. The system to be
studied consists of three types of monomers. A, B, and C; hence the initiation stage
can be symbolized. The primary chain radicals A. , B., and C. can now react wi t h
monomers and thereby create a growing chain. This growth phase is symbolized, where
A. , B., and C. now represent pol ymer chains ending with a radical attached to an A,
B. or C monomer. Since there are three monomers, there will be nine possible
reactions. The K
JK
are the
"propagation" rate constants.
A. A.
B. B. P
.
C.
C.
The fourth step in the sequence is the termination reaction where two chain radicals
react to form a "dead" polymer molecule. There are six possible reactions in t hi s last
step. Thi s sequence cont i nues t hrough these four steps unt i l all t he monomer
present is converted to polymer or t he rate of radical formation from t he i ni t i at or
decreases to zero.
18
Radical generation
. I I
i
k
÷→ ÷
Initiation
. .
.
. .
C C I
B B I
A A I
→ +
→ +
→ +
Growth
. .
. .
. .
. .
. .
. .
. .
. .
. .
C C C
B C B
A C A
C B C
B B B
A B A
C A C
B A B
A A A
CC
CB
CA
BC
BB
BA
AC
AB
AA
k
k
k
k
k
k
k
k
k
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
÷→ ÷ +
Termination
P B C
P C A
P B A
P C C
P B B
P A A
TCB
TAC
TAB
TCC
TBB
TAA
k
k
k
k
k
k
÷ ÷ → ÷ +
÷ ÷ → ÷ +
÷ ÷ → ÷ +
÷ ÷ → ÷ +
÷ ÷ → ÷ +
÷ ÷ → ÷ +
. .
. . .
. . .
. . .
. . .
. . .
Kinetics of the co polymerization is as under
Radical generation R
I
= k
I
I (2.47)
Monomer reaction rates
For monomer A
( ) . . . C k B k A k VA R
CA BA AA A
+ + ·
(2.48)
For monomer B
( ) . . . C k A k B k VB R
CB AB BB B
+ + ·
(2.49)
For monomer C
( ) . . . C k A k C k VC R
BC AC CC c
+ + ·
(2.50)
2.7 POLYMERIZATION REACTORS
There are different types of Polymerization reactors. Here three types will be considered.
1. Batch (or semi batch) reactor
2. Plug flow reactor
3. Continuous stirred tank reactor
1. Batch (or Semi batch) Reactors
The most common polymerization reactor on a numerical basis is the batch kettle. Batch
kettle may range in size from a 5gal pilot plant kettle, to a 30,000gal production kettle.
They are generally constructed of stainless steel or glass lined.
If all reactants are added at the beginning of the polymerization, the kettle is said to be
operating in the batch mode. If a reactant is added during the course of polymerization.
The kettle is said to be operating in a semi batch mode.
2. Plug flow reactors
In a plug flow reactor, each element of the reaction mixture can be viewed as an
individual batch reactor. The batch time is the residence time in tubular reactor, which is
easily calculated as the total volume of the tube divided by the volumetric flow rate.
Because no material enters or leaves the fluid element during the reaction time, all of the
kinetic relationships derived thus far for the batch reactor are directly applicable to the
plug flow reactor.Tubular reactors (approximating plug flow characteristics) are
applicable in high volume polymerizations.
3. Continuous stirred tank reactors
19
The use of continuous stirred tank polymerization (in a single CSTR or train multiple
CSTRs in series) may be warranted for high volume products. The nature of the reactor
system results in low processing costs, high throughput and in most cases a highly
uniform product. The fact that the polymerization rate is constant will contribute to
product homogeneity. Large residence time CSTR systems are not particularly flexible
and are therefore best suited to extended production runs of a small number of products.
In low residence time CSTRs (as in olefin polymerization), grade changes can be made
rapidly and low volume products can be made effectively.
CHAPTER 3
MODELING
In this chapter, we will study about use of modeling, modeling principles, how to develop
a model and at the last we will discus that how a polymerization reactor is modeled and
the techniques for solution of difference equations used for modeling purpose.
3.1 What are Models?
Models may be defined in many ways
“The process of creating a depiction of reality, such as a graph, picture,
or mathematical representation.” OR
“The representation of a physical system by a set of mathematical
relationships that allow the response of the system to various alternative
inputs to be predicted”.
The Webster dictionary defines the term model as “a small representation
of a planned or existing object ”.
From Mc Graw Hill dictionary we find the following definition, “a
mathematical or physical system obeying certain specified conditions,
whose behavior is used to understand a physical, biological or social
system to which it is analogous in some way ”.
20
Here we define a process model as “a set of equations (including
necessary input data to solve the equations) that allows us to predict the
behavior of a chemical process system”.
3.2: Reasons for Developing Models
Many reasons for developing process models:
a) Improving or understanding chemical process operations
b) Improve the quality of produced products
c) Increase the productivity of existing and new processes
d) Operator training
e) Process design
f) Safety system analysis
g) Design control system design
h) To simulate and predict real events and processes
3.3: General modeling principles
3.3.1: Steady state modeling
For steady state modeling
[Mass or energy entering in system]  [Mass or energy leaving a system] = 0
The in and out terms would then include the generation and conversion of species
by chemical reaction, respectively.
3.3.2 Dynamic modeling
In this class, we are interested in dynamic balances that have the form:
1
]
1
¸
−
1
]
1
¸
·
1
]
1
¸
system the leaving
energy or mass of Rate
system the entering
energy or mass of Rate
system within d accumulate
energy or mass of Rate
21
3.3.3: Constitution relationships: Sometimes, we need more relationships than
simple mass or energy balance, such as
Gas law e.g., ideal gas law PV=nRT
Arrhenius rate law K(T)=Aexp(E/RT)
Equilibrium relationships yi=Kixi
Heat transfer
T UA Q ∆ ·
Flow through valves
g s
P
x f C F
v
v
.
) (
∆
·
3.4: Classification of models
Models can be classified as
a) Theoretically based vs. empirical
b) Linear vs. nonlinear
c) Steady state vs. unsteady state
d) Lumped parameter vs. distributed parameter
e) Continuous vs. discrete variables
3.5: How to build a model
For convenience of presentation, model building can be divided into four phases:
1) problem definition and formulation, 2) preliminary and detailed analysis, 3) evaluation
and 4) interpretation. Keep in mind that model building is an iterative procedure.
22
Experience,
reality
Formulate model objectives,
Evaluation criteria, costs
of development
Management
Objectives
Select key variables,
Physical principles to be
applied, test plan to be use
Develop
Model
Observations,
data
Computer simulation,
Software development
Estimate
Parameters
Evaluate and
verify model
Apply model
3.6: Modeling of Polymerization Reactors
For modeling of polymerization reactors, the mass and energy balances incorporating the kinetics and
reactor type should be combined with a description of the molecular weight development to
produce a model of a polymerization reactor which can accurately describe the production rate
and character of the product. Often, however, the proper kinetic constants may be unknown. In
some cases, even the exact form of the reaction mechanism is not understood. In these cases, it
may be necessary to use parameter estimation to fit the model to experimental data. The
number of adjustable parameters should be kept low, however, or the model becomes
descriptive of particular data sets rather than mechanistic.
Any model, even if based on established kinetics, should be validated by simulation of data sets
which were not used in the estimation of parameters within the model. These data should
preferably cover operating regimes far from those of the data used for parameter estimation.
Poor agreement suggests incorrect mechanisms or a tendency of the model to simply
correlate the data from which its parameters were estimated. Validated polymerization reactor
models may be extremely valuable for design of the reactor system, optimization of the operating
parameters, simulation of potential new modes of operation, and even for the development of new
products through the modification of the product via modification of the process. In addition,
simulation studies can lead to the identification of potential stability problems such as
steadystate multiplicity and can be used to evaluate potential control schemes.
3.6.1: Mass and Energy Balances
Polymerization reactors can be modeled using the classical techniques of chemical reactor
design. Primary to this approach are mass balances over various chemical species as well
as an energy balance. In all cases, the knowledge of the chemical kinetics is used to
describe the rates of formation of various species or, in the case of the energy balance,
23
to describe the rate of heat generation via reaction. These terms are combined with
flow terms specific to the reactor in question. In the case of heterogeneous systems,
terms describing interphase mass or heat transfer may also be included. Some processes
may occur so rapidly that the species involved are assumed to be at equilibrium at all
times. By applying the principles of mass and energy balances
1
]
1
¸
+
1
]
1
¸
−
1
]
1
¸
·
1
]
1
¸
system within consumed or
generated energy or mass of rate
system the leaving
energy or mass of Rate
system the entering
energy or mass of Rate
system within d accumulate
energy or mass of Rate
We can derive the equations for mass and energy balance. A mass balance over a
monomer may be written as
·
dt
dM
Q
f
M
f
 QM  Vk
p
PM, M(0) = M
0
(3.1)
Where V is the reactor volume, Q
f
and Q are the inlet and outlet volumetric flow rates,
respectively, and P and M are taken to be concentrations of polymer and monomer
respectively.
If P is assumed constant (rapid initiation and no chain transfer), a mass balance over the
live polymer may be written as
QP P Q
dt
dP
V
f f
− · , P(0) = P
0
(3.2)
where the generation term may be negative or zero as above.
The energy balance may be written as
0
) 0 (
) ( ) (
T T
T T UA PM k H V T Q C T Q C
dt
dT
V C
j p p p f f p p
·
− − ∆ − + − · ρ ρ ρ
(3.3)
The heat of reaction (propagation), and U and A are the overall heat transfer coefficient and heat
transfer area for the cooling jacket. In view of the viscosity of polymerizing solutions and the
effect of micro mixing on molecular weight development, it may be desirable in some instances to
incorporate a more complex mixing model for the reactor.
3.6.2 Molecular Weight Distribution
In modeling of polymerization reactors it is also important to control the molecular
weight and molecular weight distribution. Therefore it is necessary to develop the
mathematical models which relate the molecular weight and molecular weight
distribution with the process variables. So in the next chapter we will explain the
molecular weight distribution.
24
Models of the form of Eqs. (3.1), (3.2), and (3.3) contains derivative. To solve these we will
have to integrate these equations. These equations can be integrated numerically (or occasionally,
analytically) from specific initial conditions to determine the transient behavior of the system.
For design purposes, it may be acceptable to consider only the steadystate solution. This may
be obtained by setting the time derivatives to zero and solving for M, P, and T. Two methods
are discussed here.
The Method of Moments
If the kth moment of the NCLD is defined as
∑
∞
·
·
1 n
n
k
k
P n λ
k=0 , 1 , 2 The NACL
may be written as the ratio of the zeroth to the first moments:
0
1
λ
λ
µ ·
n (3.5)
The variance of the NCLD is the second moment about the mean or
2
0
1
0
2
2
,
_
¸
¸
− ·
λ
λ
λ
λ
σ
n
(3.6)
Similarly, the k
th
moment of the WCLD may be written as
n
n
k
k
P n w
∑
∞
·
+
·
1
) 1 (
λ
k=1,2,… (3.7)
The WACL may be written as
w
µ
=
0
1
λ
λ
=
1
2
λ
λ
The variance of the WCLD may be written as
2
1
2
1
3
2
0
1
1
2
2
,
_
¸
¸
− ·
,
_
¸
¸
− ·
λ
λ
λ
λ
λ
λ
λ
λ
σ
w
(3.8)
The means of the NMWD and WMWD may be calculated from the NACL and WACL via
Eqs. (2.1) and (2.2). The variances of the NMWD and WMWD are functions of the variances
of the NCLD and WCLD, respectively, as follows:
2 2 2
n mn
w σ σ · (3.9)
2 2 2
w mw
w σ σ · (3.10)
Thus it may be seen that if the leading moments of the NCLD (or WCLD) are known, the mean
and the variance of any of the distributions (NCLD, WCLD, NMWD, or WMWD) characterizing
the product may be calculated directly. Once the NAMW and WAMW have been determined. It
(3.4)
25
is possible to reconstruct the entire distribution from an infinite set of moments. If the
distribution is not complex, a good approximation may be made with a finite (even small)
number of moments. In cases of simple kinetics, it may be possible to derive equations for
the development of the moments directly from the mass balances. Consider batch anionic
solution polymerization. Equation dP
n
/dt =
M k
p
−
) (
1 −
−
n n
P P
may be multiplied by n
k
and then summed for all values of n, resulting in
∑ ∑ ∑
∞
·
∞
·
−
∞
·
+ − ·
1 2
1
1
,
n n
n
k
p n
k
n
p
n k
P n M k P n M k
dt
dP
n
(3.11)
∑
∞
·
·
1
10
n
n
P P
Here
∑
∞
·
·
1
,
n
k n
k
P n λ
dt
d
dt
dP
n
k n k
n
λ
·
∑
∞
·1 (3.12)
The final term in Eq.3.11 may be evaluated for k=1,2,3 as
k=0,
0
2
1
λ
∑
∞
·
−
·
n
n
k
P N
(3.13)
k=1,
1 0
2
1
λ λ + ·
∑
∞
·
−
n
n
k
P n
(3.14)
k=2,
2 1 0
2
1
2 λ λ λ + + ·
∑
∞
·
−
n
n
k
P n
(3.15)
the equations for the first three moments becomes
, 0
0
·
dt
dλ
10 0
) 0 ( P · λ
(3.16)
,
0
1
λ
λ
M k
dt
d
p
·
10 1
) 0 ( P · λ
(3.17)
, 2
1 0
2
λ λ
λ
M k M k
dt
d
p p
+ ·
10 2
) 0 ( P · λ
(3.18)
Integration of eqs. (3.16) through (3.18) together with the monomer balance will yield an
adequate characterization of the molecular weight development under isothermal
conditions.
26
zTransforms
Another way of dealing with the differential difference equations describing the chain
length distribution is through the use of ztransforms. The use of ztransforms is common
in digital process control, but can be used to solve systems of difference equations arising
from any source. The ztransform of P
n
is defined as
), ( ) , (
0
t P z t z F
n
n
n
∑
∞
·
−
·
0 ) ( · t P
n
for n<=0 (3.18)
Note that in this case the discrete variable is chain length ( n ) and that time remains a
continuous variable. This is different than in digital control applications where the only
independent variable, time, is being discretized. However, standard tables of ztransforms can
still be used if this difference is kept in mind and the results are interpreted accordingly.
) ( ) (
n k n
P F z P F
k −
·
−
(3.19)
The moments of the NCLD can be calculated from the ztransform of P
n
(t) as
¹
)
¹
¹
'
¹
∂
∂
− ·
→ k
k
k
z k
z
t z F
t
) (ln
) , (
) 1 ( lim ) (
1
λ
(3.20)
The ztransform technique is quite similar to the generating function approach. It has an
advantage, however, in that extensive tables of ztransforms are available in the literature.
27
CHAPTER 4
OPTIMIZATION
Reactors are often the critical stage in a polymerization process. Recently, demands on
the design and operation of chemical processes are increasingly required to comply with
the safety, cost and environmental concerns. All these necessitate accurate modeling and
optimization of reactors and processes. To optimize the reactors, in this chapter we have
discussed about optimization.
How to Define a Model for Optimization?
We need to quantify the various elements of our model: decision variables, constraints,
and the objective function and their relationships.
Decision Variables
Start with the decision variables. They usually measure the amounts of resources, such as
money, to be allocated to some purpose, or the level of some activity, such as the number
of products to be manufactured, the number of pounds or gallons of a chemical to be
blended, etc.
Objective Function
Once we've defined the decision variables, the next step is to define the objective, which
is normally some function that depends on the variables. For example profit, production
rate, Conversion, yield, various costs, etc. We usually maximize or minimize objective
function such as we maximize profit, production rate, Conversion ,yield and minimize
various costs.
Constraints
You'd be finished at this point, if the model did not require any constraints. For example,
in a curvefitting application, the objective is to minimize the sum of squared differences
between each actual data value, or observation, and the corresponding predicted value.
This sum has a minimum value of zero, which occurs only when the actual and predicted
values are all identical. If you asked a solver to minimize this objective function, you
would not need any constraints.
In most models, however, constraints play a key role in determining what values can be
assumed by the decision variables, and what sort of objective value can be attained.
28
Constraints reflect realworld limits on production capacity, market demand, available
funds, and so on. To define a constraint, you first compute a value based on the decision
variables. Then you place a limit (<=, = or >=) on this computed value.
General Constraints. For example, if A1:A5 contains the percentage of funds to be
invested in each of 5 stocks, you might use B1 to calculate =SUM(A1:A5), and then
define a constraint B1 = 1 to say that the percentages allocated must sum up to 100%.
Bounds on Variables. Of course, you can also place a limit directly on a decision
variable, such as A1 <= 100. Upper and lower bounds on the variables are efficiently
handled by most optimizers and are very useful in many problems.
Policy Constraints
Some constraints are determined by policies that you or your organization may set. For
example, in an investment portfolio optimization, you might have a limit on the
maximum percentage of funds to be invested in any one stock, or one industry group.
Physical Constraints
Many constraints are determined by the physical nature of the problem. For example, if
your decision variables measure the number of products of different types that you plan
to manufacture, producing a negative number of products would make no sense. This
type of nonnegativity constraint is very common. Although it may be obvious to you,
constraints such as A1 ≥ 0 must be stated explicitly, because the solver has no other way
to know that negative values are disallowed.
As another example of a physically determined constraint, suppose you are modeling
product shipments in and out of a warehouse over time. You'll probably need a balance
constraint, which specifies that, in each time period, the beginning inventory plus the
products received minus the products shipped out equals the ending inventory  and
hence the beginning inventory for the next period.
Integer Constraints
Advanced optimization software also allows you to specify constraints that require
decision variables to assume only integer (whole number) values at the solution. If you
are scheduling a fleet of trucks, for example, a solution that called for a fraction of a truck
to travel a certain route would not be useful. Integer constraints normally can be applied
only to decision variables, not to quantities calculated from them.
A particularly useful type of integer constraint specifies that a variable must have an
integer value with a lower bound of 0, and upper bound of 1. This forces the variable to
be either 0 or 1  nothing in between  at the final solution. Hence, it can be used to
model "yes/no" decisions. For example, you might use a 01 or binary integer variable to
represent a decision on whether to lease a new machine. Your model might then
29
calculate a fixed lease cost per month, but also a lower cost per item processed with the
machine, if it is used.
Feasible Solution.
A solution (set of values for the decision variables) for which all of the constraints in the
model are satisfied is called a feasible solution. Most solution algorithms first try to find a
feasible solution, and then try to improve it by finding another feasible solution that
increases the value of the objective function (when maximizing, or decreases it when
minimizing).
Optimal Solution
An optimal solution is a feasible solution where the objective function reaches a
maximum (or minimum) value.
Globally Optimal Solution
A globally optimal solution is one where there are no other feasible solutions with better
objective function values.
Locally Optimal Solution
A locally optimal solution is one where there are no other feasible solutions "in the
vicinity" with better objective function values  you can picture this as a point at the top
of a "peak" or at the bottom of a "valley" which may be formed by the objective function
and/or the constraints. The Solver is designed to find optimal solutions  ideally the
global optimum  but this is not always possible.
Whether we can find a globally optimal solution, a locally optimal solution, or a good
solution depends on the nature of the mathematical relationship between the variables and
the objective function and constraints (and the solution algorithm used).
What Makes a Model Hard to Solve?
Solver models can be easy or hard to solve. "Hard" models may require a great deal of
CPU time and randomaccess memory (RAM) to solve  if they can be solved at all. The
good news is that, with today's very fast PCs and advanced optimization software from
Frontline Systems, a very broad range of models can be solved.
Three major factors interact to determine how difficult it will be to find an optimal
solution to a solver model:
a) The mathematical relationships between the objective and constraints, and the
decision variables
30
b) The size of the model (number of decision variables and constraints) and its
sparsity
c) The use of integer variables  memory and solution time may rise exponentially as
you add more integer variables
Mathematical Relationships
a) Linear programming problems
b) Smooth nonlinear optimization problems
c) Global optimization problems
d) Nonsmooth optimization problems
The types of mathematical relationships in a model (for example, linear or nonlinear, and
especially convex or nonconvex) determine how hard it is to solve, and the confidence
you can have that the solution is truly optimal. These relationships also have a direct
bearing on the maximum size of models that can be realistically solved.
A model that consists of mostly linear relationships but a few nonlinear relationships
generally must be solved with more "expensive" nonlinear optimization methods. The
same is true of models with mostly linear or smooth nonlinear relationships, but a few
nonsmooth relationships. Hence, a single IF or CHOOSE function that depends on the
variables can turn a simple linear model into an extremely difficult or even unsolvable
nonsmooth model.
A few advanced solvers can break down a problem into linear, smooth nonlinear and
nonsmooth parts and apply the most appropriate method to each part  but in general,
you should try to keep the mathematical relationships in a model as simple (i.e. close to
linear) as possible.
Below are some general statements about solution times on modern Windows PCs (with,
say, 2GHz CPUs and 512MB of RAM), for problems without integer variables
Linear Programming Problems  where all of the relationships are linear, and hence
convex  can be solved up to hundreds of thousands of variables and constraints, given
enough memory and time. Models with tens of thousands of variables and constraints
can be solved in minutes (sometimes in seconds) on modern PCs. You can have very
high confidence that the solutions obtained are globally optimal.
Smooth Nonlinear Optimization Problems  where all of the relationships are smooth
functions (i.e. functions whose derivatives are continuous)  can be solved up to tens of
thousands of variables and constraints, given enough memory and time. Models with
thousands of variables and constraints can often be solved in minutes on modern PCs.
If the problem is convex, you can have very high confidence that the solutions obtained
are globally optimal. If the problem is nonconvex, you can have reasonable confidence
that the solutions obtained are locally optimal, but not globally optimal.
31
Global Optimization Problems  smooth nonlinear, nonconvex optimization problems
where a globally optimal solution is sought  can often be solved up to a few hundred
variables and constraints, given enough memory and time. Depending on the solution
method, you can have reasonably high confidence that the solutions obtained are globally
optimal.
Nonsmooth Optimization Problems  where the relationships may include functions
like IF, CHOOSE, LOOKUP and the like  can be solved up to scores, and occasionally
up to hundreds of variables and constraints, given enough memory and time. You can
only have confidence that the solutions obtained are "good" (i.e. better than many
alternative solutions)  they are not guaranteed to be globally or even locally optimal.
Model Size
The size of a solver model is measured by the number of decision variables and the
number of constraints it contains.
Most optimization software algorithms have a practical upper limit on the size of models
they can handle, due to either memory requirements or numerical stability.
Sparsity
Most large solver models are sparse. This means that, while there may be thousands of
decision variables and thousands of constraints, the typical constraint depends on only a
small subset of the variables. For example, a linear programming model with 10,000
variables and 10,000 constraints could have a "coefficient matrix" with 10,000 x 10,000 =
100 million elements, but in practice, the number of nonzero elements is likely to be
closer to 2 million.
Integer Variables
Integer variables (i.e. decision variables that are constrained to have integer values at the
solution) in a model make that model far more difficult to solve. Memory and solution
time may rise exponentially as you add more integer variables. Even with highly
sophisticated algorithms and modern supercomputers, there are models of just a few
hundred integer variables that have never been solved to optimality.
This is because many combinations of specific integer values for the variables must be
tested, and each test requires the solution of a "normal" linear or nonlinear optimization
problem. The number of combinations can rise exponentially with the size of the
problem. "Branch and bound" and "branch and cut" strategies help to cut down on this
exponential growth, but even with these strategies, solutions for even moderately large
mixedinteger programming (MIP) problems can require a great deal of time.
32
Optimization Solution Methodologies
Solving and obtaining the optimum values is the last phase in design optimization.
A number of general methods for solving the programmed optimization problems relate
Various relations and constraints that describe the process to their effect on objective
function. these examine the effect of variables on the objective function using analytic,
graphical, and algorithmic techniques based on the principles of optimization and programming
methods.
Procedure with One Variable
There are many cases in which the factor being minimized (or maximized) is an analytic function of
a single variable. The procedure then becomes very simple. Consider the example, where it is
necessary to obtain the insulation thickness that gives the least total cost. The primary variable
involved is the thickness of the insulation, and relationships can be developed showing how this
variable affects all costs.
Cost data for the purchase and installation of the insulation are available, and the length of
service life can be estimated. Therefore, a relationship giving the effect of insulation thickness
on fixed charges can be developed. Similarly, a relationship showing the cost of heat lost as a
function of insulation thickness can be obtained from data on the thermal properties of steam,
properties of the insulation, and heattransfer considerati6"ns. All other costs, such as
maintenance and plant expenses, can be assumed to be independent of the insulation thickness.
Procedure with Two or More Variables
When two or more independent variables affect the objective function, the procedure for
determining the optimum conditions may become rather tedious; however, the general approach is
the same as when only one variable is involved. Consider the case in which the total cost for a
given operation is a function of the two independent variables x and y, or
C
T
= f(x,y)
By analyzing all the costs involved and reducing the resulting relationships to a simple form, the
following function might be found.
C
T
= ax +b/ xy + cy+d
Where a, b, c, and d are positive constants.
(936)
33
Graphical Procedure: The relationship among C
T
, X, and y could be shown as a curved
surface in a threedimensional plot, with a minimum value of C
T
occurring at the optimum
values of x and v. However, the use of a threedimensional plot is not practical for most
engineering determinations.
Analytical Procedure: The optimum value of x is found at the point where (dC
T
/dx)
y=yi
, is equal to
zero. Similarly, the same results would be obtained if y were used as the abscissa instead of x. If this
were done, the optimum value of y (that is yi) would be found at the point where (dC
r
/dy)
x=x
i.
, is equal
to zero. This immediately indicates an analytical procedure for determining optimum values.
Algorithm Solutions to Optimization Problems
An algorithm is simply an objective mathematical method for solving a problem and is purely
mechanical so that it can be programmed for a computer. Solution of programming problems
generally requires a series of actions that are iterated to a solution, based on a programming
method and various numerical calculation methods. The input to the algorithm can be manual,
where the relations governing the design behavior are added to the algorithm. They can also be
integrated or set to interface with rigorous computer simulations that describe the design.
Use of algorithms thus requires the selection of an appropriate programming method,
methods, or combination of methods as the basic principals for the algorithm function. It also
requires the provision of the objective functions and constraints, either as directly provided relations
or from computer simulation models. The algorithm then uses the basic programming approach to
solve the optimization problem set by the objective function and constraints.
Linear Programming Algorithm Development
To develop this form of approach for linear programming solutions, a set of linear inequalities
which form the constraints are written in the form of "equal to or less than" equations as
m n mn m m
n n
n n
b v a v a v a
b v a v a v a
b v a v a v a
≤ + + +
≤ + + +
+ + + ≤
...
........ .......... .......... .......... ..........
...
...
2 2 1 1
2 2 2 22 1 21
1 1 2 12 1 11
........ .......... .......... .......... ..........
m n mn m m
b v a v a v a ≤ + + + ...
2 2 1 1
or in general summation form
∑
·
<·
n
j
i j ij
b v a
1
i = 1,2,….,m
for
0 >·
j
v
j=1,2,……,n
where i refers to rows (or equation number) in the set of inequalities and j refers to
columns (or variable number).
34
As indicated earlier, these inequalities can be changed to equalities by adding
a set of slack variables
m n n
v v
+ +
,....,
1
(here v is used in place of S to simplify the
generalized expressions), so that
1 1 1 2 12 1 11
... b v v a v a v a
n n n
· + + + +
+
2 2 2 2 22 1 21
... b v v a v a v a
n n n
· + + + +
+
..... .......... .......... .......... .......... ..........
m m n n mn m m
b v v a v a v a · + + + +
+
...
2 2 1 1
or in general summation form
∑
·
+
· +
n
j
i i n j ij
b v v a
1
) (
i= 1,2,…,m
for
j
v
≥0 j=1,2,…..,n+m
In addition to the constraining equations, there is an objective function for the linear
program which is expressed in the form of
z=maximum(or minimum) of n n j j
v c v c v c v c .... ....
2 2 1 1
+ + + +
where the variables j
v
are subject to j
v
≥0(j=1,2,….,n+m). Note that, in this case, all
the variables above n
v
are slack variables and provide no direct contribution to the value
of the objective function.
Simplex Algorithm:
The basis for the simplex method is the generation of extremepoint solutions by starting at any
one extreme point for which a feasible solution is known and then proceeding to a neighboring
extreme point. Special rules are followed that cause the generation of each new extreme, point to
be an improvement toward the desired objective function. When the extreme point is reached
where no further improvement is possible, this will represent the desired optimum feasible
solution. Thus, the simplex algorithm is an iterative process that starts at one extremepoint
feasible solution, tests this point for optimality, and proceeds toward an improved solution. If an
optimal solution exists, this algorithm can be shown to lead ultimately and efficiently to the
optimal solution.
The stepwise procedure for the simplex algorithm is as follows (based on the
optimum being a maximum):
1. State the linear programming problem in standard equality form.
2. Establish the initial feasible solution from which further iterations can proceed. A common
method to establish this initial solution is to base it on the values of the slack variables,
where all other variables are assumed to be zero. With this assumption, the initial matrix for
the simplex algorithm can be set up with a column showing those variables that will be
35
involved in the first solution. The coefficient for these variables appearing in the matrix table
should be 1 with the rest of column being 0.
3. Test the initial feasible solution for optimality. The optimality test is accomplished by the
addition of rows to the matrix which give a value of Zj for each column, where Zj is defined
as the sum of the objective function coefficient for each solution variable.
4. Iteration toward the optimal solution is accomplished as follows: Assuming that the
optimality test indicates that the optimal program has not been found, the following
iteration procedure can be used:
a. Find the column in the matrix with the maximum value of c
j
 Zj and designate this column
as k. The incoming variable for the new test will be the variable at the head of this column.
b. For the matrix applying to the initial feasible solution, add a column showing the
ratio b
i
/a
ik.
c. . Find the minimum positive value of this ratio, and designate the variable in the
corresponding row as the outgoing variable.
c. Set up a new matrix with the incoming variable, as determined under (a), substituted for the
outgoing variable, as determined under (b). The modification of the table accomplished by
matrix operations so that the entering variable will have a 1 in the row of the departing
variable and 0s in the rest of that column. The matrix operations involve row manipulations
of multiplying rows by constants and subtracting from or adding to other rows until the
necessary 1 and 0 values are reached.
d. Apply the optimality test to the new matrix.
e. Continue the iterations until the optimality test indicates that the optimum objective
function has been attained.
5. Special cases:
a. If the initial solution obtained by use of the method given in the preceding is not feasible, a
feasible solution can be obtained by adding more artificial variables which must then be
forced out of the final solution.
b. Degeneracy may occur in the simplex method when the outgoing variable is selected. If
there are two or more minimal values of the same size, the problem is degenerate, and a poor
choice of the outgoing variable may result in cycling, although cycling almost never occurs
in real problems. This can be eliminated , by a method of multiplying each element in the
rows in question by the positive coefficients of the kth column and choosing the row for the
outgoing variable as the one first containing the smallest algebraic ratio.
6. The preceding method for obtaining a maximum as the objective function can be applied to the
case when the objective function is a minimum by recognizing that maximizing the negative of
a function is equivalent to minimizing the function.
36
CHAPTER No. 5
MODELING OF BATCH POLYMERIZATION
REACTORS
In this chapter, the focus will be shifted from chemistry and kinetics of polymerization to
the mathematical description i.e. modeling of polymerization process. In this chapter, we
will apply mass and energy balances combined with the kinetics of polymerization
reactions for batch reactor. We will focus our main attention to the molecular weight and
molecular weight distribution for different polymerization processes using batch reactors.
We will model the equations to determine the effects of operating conditions on the mean
chain length distributions and the breadth of distribution.
5.1 Anionic Polymerization
Let us consider batch anionic polymerization. Now we apply mass and energy balances to
batch reactor. As there are no inflow and outflow terms for batch polymerizer so the
equation becomes
PM Vk
dt
dM
V
P
− ·
which can be written as
(5.1)
0
0
·
·
dt
dP
dt
dP
V
( )
0
0 P P ·
(5.2)
The equation (5.1) and (5.2) can be solved to determine the time dependent behavior of
concentration of monomer and polymer.
To investigate molecular weight distribution it is necessary to develop mass balance over
the concentration of live chains of length n.
The balance for P1 is simply
3 2
2 1
P M P
P P M
P
P
k
k
÷→ ÷ +
÷→ ÷ +
(5.3)
37
PM k
dt
dM
P
− ·
( )
0
0 M M ·
1
1
MP k
dt
dP
P
− ·
similarly balance for P
2
is
) (
1 2
2
2 1
2
P P M k
dt
dP
MP k MP k
dt
dP
P
P P
− − ·
− ·
from this the balance for P
n
can be written analogous to P
2
(5.4)
we define variance
τ
as under
(5.5)
from this equations (5.1) (5.3) and (5.4) can be written as under
1
1
P
d
dP
− ·
τ
P
1
(0)=P
10
(5.6)
) (
1 −
− − ·
n n
n
P P
d
dP
τ
Pn(0)=0, n>=2 (5.7)
M(0)=M
0
(5.8)
taking the ztransform of equatins (5.6)and (5.7) one obtains
( )
( ) ( ) τ
τ
τ
, 1
,
1
z F z
d
z dF
− ·
−
( )
10
1
0 , P z z F
−
· (5.9)
equation(5.9) is separable and can be solved as
) exp( ) exp( ) , (
1
10
1
τ τ τ
− −
− · z P z z F
expanding power series in z
1
(5.10)
now by comparing with definition of z
transform the eq.(5.10) can be written as
)! 1 (
) (
) exp( ) (
1
10
−
− ·
−
n
P t P
n
n
τ
τ (5.11 )
this is poisson distribution with mean (1+
τ
) and variance
τ
.
Polydispersity can be written as
2
1
2 0
λ
λ λ
µ
µ
· · ·
n
w
n
w
m
m
D
(5.12)
the variance about mean NACL denoted by
2
n
δ can be written in term of moments as
2
0
1
0
2
2
,
_
¸
¸
− ·
λ
λ
λ
λ
δ
n
(5.13)
38
) (
1 −
− − ·
n n P
n
P P M k
dt
dP
( ) ' '
0
dt t M k
t
P
∫
· τ
P
d
dM
− ·
τ
n
n
n
z
n
P z F
−
∞
·
−
∑
−
− ·
1
1
10
)! 1 (
) (
) exp( ) , (
τ
τ τ
we know that NACL is defined as in term of moments as
0
1
λ
λ
µ ·
n
2
0
1
2
,
_
¸
¸
·
λ
λ
µ
n
(5.14)
Dividing (5.14) on (5.13) gives
2
0
1
2
2
0
1
0
2
2
,
_
¸
¸
·
,
_
¸
¸
− ·
λ
λ
µ
λ
λ
λ
λ
δ
n
n
1
2
1
0 2
2
2
− ·
λ
λ λ
µ
δ
n
n
2
1
0 2
2
2
1
λ
λ λ
µ
δ
· +
n
n
from the definition of Polydispersity given by equation (5.12) it can be written as
2
2
1
n
n
D
µ
δ
+ ·
(5.15)
from eq.(5.5)
τ µ + ·1
n
for an ionic polymerization
τ
>>1 so one can be neglected in the above eq.
τ µ ≅
n
=
2
n
δ
from eq(5.15)
2
2
1
n
n
D
µ
δ
+ ·
=1+
n
µ
1
(5.16)
Hence for high degree of polymerization D approaches unity meaning NCLD approaches
monodispersity hence ionic polymerization in absence of termination or chain transfer is
useful for creating narrow molecular weight distribution.
The above discussion was for anionic polymerization without chain transfer or
termination. How ever for anionic polymerization where chain transfer takes place.
The instantaneous degree of polymerization may be calculated as the rate of propagation
divided by the rate of chain transfer (rate of production of dead chains).
kp M
x = (5.17)
k
f
B
5.2. Free Radical Polymerization
Consider the free radical polymerization mechanism Ignoring inhibition and considering batch
solution polymerization, the proper mass balances may be written assuming constant reactor
volume and isothermal operation.:
MP k
dt
dM
p
− ·
0
) 0 ( M M ·
(5.18)
39
I k
dt
dI
d
− ·
(5.19)
The time derivatives of R and P are set to zero, and R is eliminated from the two equations
2 / 1
2
1
]
1
¸
+
·
td tc
d
k k
I fk
P (5.20)
Equations (5.18), (5.19), and (5.20) define the conversiontime behavior of the reactor. Equations
(5.18) and (5.19) can be solved.
Integrating Eq. (5.19)
∫ ∫
− ·
I
I
t
d
o
dt k
I
dI
0
t k
I
I
d
− ·
0
ln
t k
d
e I I
−
·
0
(5.21)
Now, integrating eq. (5.18)
∫ ∫
− ·
M
M
t
p
o
dt P k
M
dM
0
Pt k
M
M
p
− ·
0
ln
Pt k
p
e M M
−
·
0
(5.22)
Rate of Propagation is given by
PM k R
p p
·
2 / 1
2
,
_
¸
¸
+
·
td tc
d
p p
k k
I fk
M k R (5.23)
and
t
p
n
R
R
X ·
−
which can be written as
2 / 1
) ( 2
] [
I ktfk
M k
X
d
p
n ·
−
40
5.2.1 Model for Free radical batch polymerization reactors
a E
t
a E
d
a E
p
Time(t)
k
d
I
0
I
f
(Monomer concentration at any time (t))
M
R
p
(Rate of polymerization at any time t)
41
RT Et
t
ae k
/ −
·
RT Ed
d
ae k
/ −
·
RT Ep
p
ae k
/ −
·
t k
d
e I I
−
·
0
,
_
¸
¸
,
_
¸
¸
·
t I
k
fk
k
t
d
p
e M M
2 / 1
2 / 1
0
fI k R
d i
− ·
M I
k
fk
k R
t
d
p p
2 / 1
2 / 1
,
_
¸
¸
·
( )
2 / 1
2 fI k k
M k
d t
p
n
· µ
In model given above:
a=frequency factor
E=activation energy
5.2.2 MATLAB SOLUTION OF MODEL
Knowing the initial concentrations of radical and monomer we can solve this model.
Values of activation energies , different initiator efficiencies and frequency factor values
are available . we can solve this model using MATLAB,
EXAMPLE
Styrene is polymerized by free radical mechanism in a batch reactor. The initial
concentrations of monomer and initiator are 1 M and .001 M.
Model the reactor to determine (a) rate of initiation (b)initiator concentration
(c)rate of polymerization (d)number average degree of polymerization at any time?
At 60 degree C initiator efficiency is0.30 and all other cnstants are as under
kd=1.2*10^5 (1/sec),kp=176 (1/M s) and kt=7.2*10^7 (1/M s)
MATLAB SOLUTION
MODEL
M0=1;
I0=.001;
T=333;
f=.30;
kd=1.2*10^(5);
kp=176;
kt=7.2*10^7;
t=input('enter time in seconds at which u want ur calculation=');
I=I0*exp(kd*t); %concentration of free radicals at time t
Ri=f*kd*I; % rate of initiation
M=M0*exp(kp*(f*kd/kt).^.5*t*I.^.5); %concentration of free radicals at time t
Rp=kp*(f*kd/kt).^.5*M*I.^.5; %rate of polymerization at any time t
mun=kp*M/(2*(f*kd*kt*I).^.5); %number average degree of polymerizationat time t
Lt=I/(kd*f*I); %free radical life time
disp('concentration of free radicals= ')
disp(I)
disp('concentration of monomers=')
disp(M)
disp('rate of polymerization=')
disp(Rp)
disp('number average degree of polymerization=');
disp(mun)
disp('free radical life time=')
disp(Lt)
SOLUTION
enter time in seconds at which u want ur calculation=60
concentration of free radicals=
42
9.9928e004
concentration of monomers=
0.9999
rate of polymerization=
1.2440e006
number average degree of polymerization=
172.8977
free radical life time=
2.7778e+005
Effect of time on concentration of radical
» I0=.001;
»kd=1.2*10^(5);
» t=1:1800;
» I=I0*exp(kd*t);
» plot(t,I)
0 200 400 600 800 1000 1200 1400 1600 1800
9.7
9.8
9.9
10
x 10
4
graph b/w time and concentration of radical
concentration
of radicals
(M)
time(seconds)
43
Effect of time on rate of polymerization
M0=1;
» kp=176;
» kt=7.2*10^7;
» kd=1.2*10^(5);
» I0=.001;
» f=.30;
» t=1:1800;
» I=I0*exp(kd*t);
» M=M0.*exp(kp*(f*kd/kt).^.5.*t.*I.^.5);
» Rp=kp*(f*kd/kt).^.5*M.*I.^.5;
» plot(t,Rp)
0 200 400 600 800 1000 1200 1400 1600 1800
1.23
1.235
1.24
1.245
x 10
6
rate of polymerization graph b/w and time
rate of
polymerization
time(seconds)
(M/sec/l)
44
Effect of time on degree of polymerization
» M0=1;
» kp=176;
» kt=7.2*10^7;
» kd=1.2*10^(5);
» I0=.001;
» f=.30;
» t=1:1800;
» I=I0*exp(kd*t);
» M=M0.*exp(kp*(f*kd/kt).^.5.*t.*I.^.5);
» Rp=kp*(f*kd/kt).^.5*M.*I.^.5;
» mun=kp*M./(2*(f*kd*kt*I).^.5);
» plot(t,mun)
0 200 400 600 800 1000 1200 1400 1600 1800
172.8
173
173.2
173.4
173.6
173.8
174
174.2
174.4
graph b/w time and degree of polymerization
degree of
polymerization
time(sec)
45
5.3:Step growth polymerization
The polymerization in which polyfunctional reactants react to produce larger units in a
continuous stepwise manner.
Step growth polymerization:
1. In absence of solvent or catalyst
2. In presence of catalyst
5.3.1:step growth polymerization in absence of solvent or catalyst.
Assume the polyesterification is conducted in the absence of solvent or
catalyst and that the monomers are present in stoichiometric ratio.
Then by applying the law of conservation of mass equation for batch
reactor. As there are no inflow and outflow terms so the eqn. can be written as
2
kA
dt
dA
− · (5.24)
integrating eq. (5.24)
∫ ∫
− ·
t A
A
dt k
A
dA
o
0
2
(5.25)
kdt
A
− ·
+ −
+ −
1 2
1 2
∫ ∫
·
A
A
t
o
dt k
A
0
1
1
1 1
+ ·
· −
· −
kt A
A
A
kt A
A
A
A
A
kt
A A
o
o
o
o
o o
o
1
) (
+
·
kt A
A
t A
o
o
(5.26)
and P is given by
0
0
) (
A
t A A
p
−
·
(5.27)
and
n
µ
p p A
A
A
A
o
o o
n
−
·
−
· ·
1
1
) 1 (
µ
(5.28)
and
w
µ
p
p
w
−
+
·
1
1
µ
(5.29)
Finally we get the value of “D” by dividing the above two equations. (5.28) and (5.29).
p D
n
w
+ · · 1
µ
µ
46
5.3.2 Model for batch polycondensation reactors
(in absence of catalyst)
Activation energy of the reaction(E)
T(Temperature) Frequency factor(a)
k
Initial concentration
of monomer A(A
0
) A(t) Concentraion of
monomer at any
time(t) time (t)
A(t)
P
Monomer conversion
P
P
Weight
w
µ
n
µ
Number
Average Average
D.P D.P
Molecular Weight
Distribution
47
1
) (
0
0
+
·
kt A
A
t A
0
0
) (
A
t A A
P
−
·
P
n
−
·
1
1
µ
n
w
MWD
µ
µ
·
RT E
ae k
/ −
·
P
P
w
−
+
·
1
1
µ
5.3.2.1 MATLAB SOLUTION OF MODEL
Polyesterification is conducted in the absence of solvent or catalyst and that the
monomers are present in stoichiometric ratios.Model the reactor .given is that the
dicarboxylic acid concentration is 3 mol L^1 and polymerization rate constant is 10^2
Lmol^1 S^1
MODEL
k=10^2;
A0=3;
t=input('enter time at which u want to calculate parametersin sec=');
A=A0./(A0*k*t+1);
p=(A0A)/A0;
mun=1./(1p);
muw=(1+p)./(1p);
D=muw./mun;
disp('monomer concentration=')
disp(A)
disp('monomer conversion=')
disp(p)
disp('number average degree of polymerization is =')
disp(mun)
disp('weight average degree of polymerization is =')
disp(muw)
disp('mol weight distribution=')
disp(D)
SOLUTION
» enter time at which u want to calculate parametersin sec=1800
monomer concentration=
0.0545
monomer conversion=
0.9818
number average degree of polymerization is =
55.0000
weight average degree of polymerization is =
109.0000
mol weight distribution=
1.9818
48
Effect of time on monomer conversion
A0=3;
» t=[1:900];
» k=10^2;
» A=A0./(A0*k*t+1);
» p=(A0A)/A0;
» plot(t,p)
»
0 100 200 300 400 500 600 700 800 900
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
conversion
time(sec)
49
Effect of time on NACL and WACL
t=[1:900];
» k=10^2;
A0=3;
» A=A0./(A0*k*t+1);
» p=(A0A)/A0;
» mun=1./(1p);
» muw=(1+p)./(1p);
» plot(t,muw,'')
» hold on
» plot(t,mun)
0 100 200 300 400 500 600 700 800 900
0
10
20
30
40
50
60
muw
mun
number
avg.and
weight avg.
D.P
time(sec)
50
Effect of time on MWD
» t=[1:900];
» k=10^2;
»A0=3;
» A=A0./(A0*k*t+1);
» p=(A0A)/A0;
» mun=1./(1p);
» muw=(1+p)./(1p);
» D=muw./mun;
» plot(t,D)
»
0 100 200 300 400 500 600 700 800 900
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
MWD
time(sec)
51
Effect of stoichiometric ratio on degree of polymerization for different conversion
» P=.996;
» r=[1:.01:1.25];
» mun=(1+r)./(1+r2*P);
» plot(r,mun)
» hold on
» P=.994;
» mun=(1+r)./(1+r2*P);
» plot(r,mun,'*')
» P=.990;
» mun=(1+r)./(1+r2*P);
» plot(r,mun,'^')
» P=.980;
» mun=(1+r)./(1+r2*P);
» plot(r,mun,'')
» P=.970;
» mun=(1+r)./(1+r2*P);
» plot(r,mun,'^')
1 1.05 1.1 1.15 1.2 1.25
0
50
100
150
200
250
p=.996
p=.994
p=.990
p=.980
p=.970
effect of stoichiometric ratio and conversion on number average degree of polymerization
number average
degree of
polymerization
stoichiometric ratio
52
5.3.3 Step growth polymerization in presence of catalyst
Let the catalyst is an acid functional group. For stoichiometric ratios of acidic
functional group. By applying Eq. of mass balance
3
kA
dt
dA
− · (5.30)
Integrating the above eq.
kt
A
− ·
+ −
+ −
1 3
1 3
Applying the limits
kt
A A
2
1 1
2
0
2
· −
kt A
A
A
2
0 2
2
0
2 1· −
kt A
A
A
2
0 2
2
0
2 1+ · (5.31)
kt A
A
A
2
0
2
0 2
2 1+
·
kt A
A
t A
2
0
0
2 1
) (
+
·
(5.32)
P is given as
0
0
) (
A
t A A
P
−
·
(5.33)
n
p p A
A
µ ·
−
·
,
_
¸
¸
−
2
2
0
0
) 1 (
1
) 1 (
2
) 1 (
1
p
n
−
· µ
(5.34)
From the equations (5.30) to (5.34) we can develop a model of batch step growth polymerization
reactor in presence of catalyst.
53
5.3.4 Model for batch polycondensation reactor
(in presence of catalyst in stoichiometric ratio )
Activation energy of the reaction(E)
T(Temperature) Frequency factor(a)
k
Initial concentration
of monomer A(A
0
) A(t) Concentraion of
monomer at any
time(t) time (t)
A(t)
P
Monomer conversion
n
µ
Number
Average
D.P
54
1 0 2
0
) (
2
+
·
kt A
A
t A
0
0
) (
A
t A A
P
−
·
RT E
ae k
/ −
·
2
) 1 (
1
p
n
−
· µ
5.3.4.1 MATLAB Solution of Model
Polyesterification is conducted in the presence of solvent or catalyst and that the monomers and
acid are present in stoichiometric ratios.Model the reactor .given is that the dicarboxylic acid
concentration is 3 mol L^1 and polymerization rate constant is 10^1 Lmol^1 S^1
MODEL
k=10^1;
A0=3;
t=input('enter time at which u want to calculate parameters in sec=');
A=A0./(A0.^2*k*t+1).^.5;
p=(A0A)/A0;
mun=1./(1p).^2;
disp('monomer concentration=')
disp(A)
disp('monomer conversion=')
disp(p)
disp('number average degree of polymerization is =')
disp(mun)
OUTPUT:
enter time at which u want to calculate parameters in sec=50
monomer concentration=
0.4423
monomer conversion=
0.8526
number average degree of polymerization is =
46.0000
55
Effect of time on monomer conversion
» t=[0:300];
» k=10^1;
»A0=3;
» A=A0./(A0.^2*k*t+1).^.5;
» p=(A0A)/A0;
» p=(A0A)./A0;
» plot(t,p)
»
0 200 400 600 800 1000 1200
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
time(sec)
onversion c
Comparison of catalyzed and non catalyzed reactions
Effect of time on monomer conversion for catalyzed and non catalyzed reactions
56
» t=[0:300];
» k=10^1;
»K=10^2;
»A0=3;
» A=A0./(A0.^2*k*t+1).^.5; % for catalyzed reactin
» p=(A0A)/A0; % for catalyzed reactin
»AA=A0./(A0*K*t+1); % for uncatalyzed reactin
» P=(A0AA)./A0; % for uncatalyzed reactin
» plot(t,p)
» hold on
» plot(t,P,'+')
»
0 100 200 300 400 500 600
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
catalyzed
uncatalyzed
time(sec)
conversion
Effect of conversion on NACL for catalyzed and non catalyzed reactions
» p=[0:.04:.92];
57
» mun=1./(1p).^2;
» Mun=1./(1p);
» plot(p,mun);
» plot(p,Mun,'+');
» plot(p,mun);
» hold on
» plot(p,Mun,'+');
»
effect of conversion on NACL for catalyzed and non catalyzed reactions
» p=[0:.04:.92];
» mun=1./(1p).^2;
» Mun=1./(1p);
» plot(p,mun);
» plot(p,Mun,'+');
» plot(p,mun);
» hold on
» plot(p,Mun,'+');
»
CHAPTER 6
MODELING OF CONTINOUS STIRRED TANK
POLYMERIZATION REACTORS
58
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
50
100
150
200
250
300
catalyzed
Noncatalyzed
conversion
NACL
In this chapter we will apply the mass balance for polymerization techniques on CSTR. And we
will derive the time conversion behaviour of polymerization reactors.
6.1 Anionic Polymerization
If we assume rapid initiation and no termination or chain transfer, the monomer and total polymer
balances for anionic polymerization are given by
PM Vk QM M Q
dt
dM
p f f
− − ·
0
) 0 ( M M ·
(6.1)
QP P Q
dt
dP
V
f f
− ·
0
) 0 ( P P ·
,
∑
∞
·
·
1 n
n
P P
(6.2)
Balances on P
1
and P
n
may be written as
1 1
1
MP Vk P Q
dt
dP
V
p f f
− ·
10 1
) 0 ( P P ·
(6.3)
2
, 0 ) 0 (
) (
1
>·
·
− − − ·
−
n
P
P P M Vk QP P Q
dt
dP
V
n
n n p n nf f
n
(6.4)
If the inflow and outflow terms are set equal and the feed is assumed to contain only P1, the
following steadystate balances may be obtained from Eqs. (6.1), (6.2), (6.3), and (6.4):
f f
P P P
1
· ·
(6.5)
MP Vk M M Q
p f
· − ) (
(6.6)
1 1 1
) ( MP Vk P P Q
p f
· −
(6.7)
) (
1 −
− ·
n n p n
P P M Vk QP
(6.8)
This system of difference equations may be solved by ztransforms to give
,
) 1 (
1 f p
f
P k
M
M
θ +
·
Q
V
· θ
(6.9)
1
1
1
1
) 1 (
) 1 ( ) 1 (
−
−
− ·
1
1
]
1
¸
+ +
·
n
f
n
p
p
p
f
P
M k
M k
M k
P
Pn φ φ
θ
θ
θ
(6.10)
φ
µ
−
·
1
1
n (6.11)
φ + ·1 D
(6.12)
where as
1
1
]
1
¸
+
·
) 1 ( M k
M k
p
p
θ
θ
φ
(6.13)
from the above equations model can be developed as
6.1.1 Model for continuous stirred tank anionic polymerization reactor
59
T a E
p
V Residence
Time(θ )
Q
θ p
k
M
f
M
φ
Number
average
Weight avg.D.P D.P
Polydispersity
6.2. Free Radical Polymerization
Consider the free radical polymerization mechanism Ignoring inhibition and considering
CSTR solution polymerization, the proper mass balances may be written assuming constant
reactor volume and isothermal operation. If q is the volumetric feed rate and the
subscript/indicates feed conditions, balances over monomer and initiator yield:
60
RT Ep
P
ae k
/ −
·
Q
V
· θ
) 1 (
1 f P
f
P k
M
M
θ +
·
) 1 ( M k
M k
p
p
θ
θ
φ
+
·
φ
φ
µ
−
+
·
1
1
w
φ
µ
−
·
1
1
n
φ
µ
µ
+ · · 1
n
w
D
V MP k M M Q
dt
dM
V
p f
− − · ) (
0
) 0 ( M M ·
(6.14)
I Vk I I Q
dt
dI
V
d f
− − · ) (
(6.15)
As before, the time derivatives of R and P are set to zero, and R is eliminated from the two
equations
2 / 1
2
1
]
1
¸
+
·
td tc
d
k k
I fk
P (6.16)
Equations (6.14), (6.15), and (6.16) define the conversiontime behavior of the reactor.
At steady state, flow rate in and flow rate out are equal and there is no accumulation. There fore, the
terms dM/dt and dI/dt are zero. So Eq. (6.14) and (6.15) become
V MP k M M Q
p f
− − ) (
= 0
I Vk I I Q
d f
− − ) (
=0
by solving these
θ
p
f
k
M
M
+
·
1
(6.16)
(6.17)
6.2.1 Model for continuous stirred tank free radical polymerization reactor
T,a
61
θ
d
f
k
I
I
+
·
1
E
p
E
fs
E
fm
E
td
E
tc
E
d
f
I I
f
I
I
θ
M
f
M
S
Number
average
Weight avg.D.P D.P
Polydispersity
62
2 / 1
2
,
_
¸
¸
+
·
td tc
d
k k
I fk
P
Q
V
· θ
) 1 ( P k
M
M
P
f
θ +
·
) / 1 ) ( ( θ
α
+ + + + +
·
P k k S k M k M k
M k
td tc fs fm p
p
α
α
µ
−
+
·
1
1
w
α
µ
−
·
1
1
n
α
µ
µ
+ · · 1
n
w
D
RT E
td
td
ae k
/ −
·
RT E
tc
tc
ae k
/ −
·
RT E
fm
fm
ae k
/ −
·
RT E
fs
fs
ae k
/ −
·
RT E
P
p
ae k
/ −
·
) 1 ( f k
I
I
d
f
θ +
·
RT E
d
d
ae k
/ −
·
6.2.2: MATLAB solution of model
To determine the initiator concentration ,monomer concentration ,NACL ,WACL, MWD
and polymer concentration at different temperature inputs at constant residence time
NOTE:We have applied this model for solution polymerization of styrene with
Initiator BPO(benzoyl per oxide) solvent is benzene.Nearly all constants have been taken from
“polymer handbook” fourth edition by J.Brandrup published by John&Wiley.
Using MATLAB(mfile)
Mf=1; %moles/litre
If=.001;
f=.30;
V=500;
Q=10;
R=8.314;
S=.03;
T=input('enter the temperature in kelvin');
disp('residence time=')
theta=V/Q;
disp(theta);
disp('sec')
kp=2.16*10^7*exp(32500/(8.314*T));
ktc=2.59/2*10^3*exp(9920/(8.314*T));
ktd=0;
kfm=2167*exp(48.185*10^3/(8.314*T));
kfs=7.57*10^8*exp(9.2*10^4/(8.314*T));
kd=1.22*10^13*exp(1.206*10^5/(8.314*T));
disp('initiator concentration=')
I=If./(1+theta*kd*f);
disp(I);
disp('polymer concentration =')
P=((2*f*kd*I)./(ktc+ktd)).^.5;
disp(P)
disp('monomer concentration =')
M=Mf./(1+theta*kp*P);
disp(M)
alpha=kp*M./(kp*M+kfm*M+kfs*S+(ktc+ktd)*P+1./theta);
disp('number average degree of polymerization=')
mun=1./(1alpha);
disp(mun);
disp('weight average degree of polymerization=')
muw=(1+alpha)./(1alpha);
disp(muw)
disp('polydispersity=')
D=1+alpha;
disp(D)
63
OUTPUT:
» enter the temperature in kelvin350
residence time=
50
sec
initiator concentration=
9.9982e004
polymer concentration =
1.3084e005
monomer concentration =
0.8338
number average degree of polymerization=
1.2289e+004
weight average degree of polymerization=
2.4577e+004
TO MODEL THE EFFECTS OF TEMPERATURE
Effect of temperature on number average and weight average D.P
» T=[250:350];
» Mf=1;
» If=.001;
»f=.30;
»V=500;
» Q=10;
»R=8.314;
» S=.03;
» theta=V/Q;
» kp=2.16*10^7*exp(32500./(8.314*T));
» ktc=2.59/2*10^3*exp(9920./(8.314*T));
» ktd=0;
» kfm=2167*exp(48.185*10^3./(8.314*T));
» kfs=7.57*10^8*exp(9.2*10^4./(8.314*T));
» kd=1.22*10^13*exp(1.206*10^5./(8.314*T));
64
» I=If./(1+theta*kd*f);
.» P=((2.*f.*kd.*I)./(ktc+ktd)).^.5;
» M=Mf./(1+theta.*kp.*P);
» alpha=kp.*M./(kp.*M+kfm.*M+kfs.*S+(ktc+ktd).*P+1./theta);
» mun=1./(1alpha);
» muw=(1+alpha)./(1alpha);
» plot(T,mun)
» hold on
» plot(T,muw,'*')
»
effect of temperature on number average and weight average degree of polymerization
250 260 270 280 290 300 310 320 330 340 350
0
0.5
1
1.5
2
2.5
x 10
4
number average D.P
weight average D.P
number
average
and weight
average
D.P
temperature(kelvin)
Effect of temperature on molecular weight distribution
» T=[250:350];
» Mf=1;
» If=.001;
»f=.30;
»V=500;
» Q=10;
65
»R=8.314;
» S=.03;
» theta=V/Q;
» kp=2.16*10^7*exp(32500./(8.314*T));
» ktc=2.59/2*10^3*exp(9920./(8.314*T));
» ktd=0;
» kfm=2167*exp(48.185*10^3./(8.314*T));
» kfs=7.57*10^8*exp(9.2*10^4./(8.314*T));
» kd=1.22*10^13*exp(1.206*10^5./(8.314*T));
» I=If./(1+theta*kd*f);
.» P=((2.*f.*kd.*I)./(ktc+ktd)).^.5;
» M=Mf./(1+theta.*kp.*P);
» alpha=kp.*M./(kp.*M+kfm.*M+kfs.*S+(ktc+ktd).*P+1./theta);
» D=1+alpha;
» plot(T,D)
» plot(T,D)
»
Effect of temperature on molecular weight distribution
250 260 270 280 290 300 310 320 330 340 350
1.994
1.995
1.996
1.997
1.998
1.999
2
2.001
MWD
Temperature(kelvin)
Effect of temperature on initiator conentration
» T=[250:350];
» Mf=1;
» If=.001;
»f=.30;
»V=500;
66
» Q=10;
»R=8.314;
» S=.03;
» theta=V/Q;
» kp=2.16*10^7*exp(32500./(8.314*T));
» ktc=2.59/2*10^3*exp(9920./(8.314*T));
» ktd=0;
» kfm=2167*exp(48.185*10^3./(8.314*T));
» kfs=7.57*10^8*exp(9.2*10^4./(8.314*T));
» kd=1.22*10^13*exp(1.206*10^5./(8.314*T));
» I=If./(1+theta*kd*f);
» plot(T,I)
»
Effect of temperature on initiator concentration
250 260 270 280 290 300 310 320 330 340 350
9.998
9.9982
9.9984
9.9986
9.9988
9.999
9.9992
9.9994
9.9996
9.9998
10
x 10
4
concentration
temperature(kelvin)
To determine the initiator concentration ,monomer concentration ,NACL ,WACL ,
MWD and polymer concentration at constant temperature(30 degC) with different
residence time inputs.
Using MATLAB (mfile)
Mf=1; %moles/litre
67
If=.001;
f=.30;
R=8.314;
S=.03;
T=303;
theta=input('enter the residence time(in sec)=');
kp=2.16*10^7*exp(32500/(8.314*T));
ktc=2.59/2*10^3*exp(9920/(8.314*T));
ktd=0;
kfm=2167*exp(48.185*10^3/(8.314*T));
kfs=7.57*10^8*exp(9.2*10^4/(8.314*T));
kd=1.22*10^13*exp(1.206*10^5/(8.314*T));
disp('initiator concentration=')
I=If./(1+theta*kd*f);
disp(I);
disp('polymer concentration =')
P=((2*f*kd*I)./(ktc+ktd)).^.5;
disp(P)
disp('monomer concentration =')
M=Mf./(1+theta*kp*P);
disp(M)
alpha=kp*M./(kp*M+kfm*M+kfs*S+(ktc+ktd)*P+1./theta);
disp('number average degree of polymerization=')
mun=1./(1alpha);
disp(mun);
disp('weight average degree of polymerization=')
muw=(1+alpha)./(1alpha);
disp(muw)
disp('polydispersity=')
D=1+alpha;
disp(D)
Output:
enter the residence time(in sec)=10
initiator concentration=
1.0000e003
polymer concentration =
68
6.8490e007
monomer concentration =
0.9996
number average degree of polymerization=
539.5726
weight average degree of polymerization=
1.0781e+003
polydispersity=
1.9981
TO MODEL THE EFFECTS OF RESIDENCE TIMES
Effect of residence time on number average and weight average D.P
» theta=[0:300];
» Mf=1;
» If=.001;
»f=.30;
»R=8.314;
» S=.03;
»T=303;
» kp=2.16*10^7*exp(32500./(8.314*T));
» ktc=2.59/2*10^3*exp(9920./(8.314*T));
» ktd=0;
» kfm=2167*exp(48.185*10^3./(8.314*T));
» kfs=7.57*10^8*exp(9.2*10^4./(8.314*T));
» kd=1.22*10^13*exp(1.206*10^5./(8.314*T));
» I=If./(1+theta*kd*f);
.» P=((2.*f.*kd.*I)./(ktc+ktd)).^.5;
» M=Mf./(1+theta.*kp.*P);
» alpha=kp.*M./(kp.*M+kfm.*M+kfs.*S+(ktc+ktd).*P+1./theta);
» mun=1./(1alpha);
» muw=(1+alpha)./(1alpha);
» plot(T,mun)
» hold on
» plot(T,muw,'+')
»
69
0 50 100 150 200 250 300
0
0.5
1
1.5
2
2.5
3
3.5
x 10
4
NACL &
WACL
Residence time(sec)
Effect of residence time on MWD
» theta=[0:300];
» Mf=1;
» If=.001;
»f=.30;
»R=8.314;
» S=.03;
»T=303;
» kp=2.16*10^7*exp(32500./(8.314*T));
» ktc=2.59/2*10^3*exp(9920./(8.314*T));
» ktd=0;
» kfm=2167*exp(48.185*10^3./(8.314*T));
» kfs=7.57*10^8*exp(9.2*10^4./(8.314*T));
» kd=1.22*10^13*exp(1.206*10^5./(8.314*T));
» I=If./(1+theta*kd*f);
.» P=((2.*f.*kd.*I)./(ktc+ktd)).^.5;
» M=Mf./(1+theta.*kp.*P);
» alpha=kp.*M./(kp.*M+kfm.*M+kfs.*S+(ktc+ktd).*P+1./theta);
» D=1+alpha;
» plot(theta,D)
70
Effect of residence time on MWD
0 50 100 150 200 250 300
1.98
1.985
1.99
1.995
2
2.005
MWD
Residence time(sec)
Effect of residence time on monomer concentration
» plot(theta,M) ( in previous workspace)
0 50 100 150 200 250 300
0.988
0.99
0.992
0.994
0.996
0.998
1
1.002
Monomer
cocentration
Residence time(sec)
6.3. StepGrowth Polymerization
71
Stepgrowth polymerization kinetics can be represented by
AB + AB ⇔ ABAB + W
An analysis of the MWD during stepgrowth polymerization in a CSTR at steadystate can be begun
by making the assumption of irreversible polymerization (due, perhaps, to the continuous removal
of the condensation product, W) and writing mass balances on the growing chains:
P P k P P
p f 1 1 1
θ · −
(6.18)
nf
P
 n
P
= ½ p
k
θ
∑
−
·
−
1
1
n
r
r n r
P P
1 > n (6.19)
∑
∞
·
·
2 n
n
P P
(6.20)
Assuming the feed consists of P
1
only, solution of Eqs. (6.18) through (6.20).
by ztransform techniques indicates that the NCLD is a binomial distribution with the
following leading moments:
θ
θ
λ
p
f p
k
P k
2 / 1
1
0
] 2 1 [ 1 + + −
·
(6.21)
) 55 . 5 )( 1 (
) 54 . 5 (
1 1 2
1 1
f p f
f
P k P
P
θ λ
λ
+ ·
·
The NACL, WACL and polydispersity can then be calculated as
] ) 2 1 ( 1 [
2 / 1
1
1
f p
f p
n
P k
P k
θ
θ
µ
+ + −
·
(6.22)
f p w
P k
1
1 θ µ + ·
(6.23)
f p
f p f p
P k
P k P k
D
1
2 / 1
1 1
] ) 2 1 ( 1 )[ 1 (
θ
θ θ + + − +
·
(6.24)
The extent of reaction in a CSTR is defined as
f
p
f
f
P
p k
P
P P
p
1
2
1
1
2
1
θ
·
−
·
(6.25)
Equations (6.22)(6.24) can be written in terms of the extent of reaction as
p
n
−
·
1
1
µ
(6.26)
2 2
) 1 (
2 1
) 1 (
2
1
p
p
p
p
w
−
+
·
−
+ · µ
(6.27)
p
p
D
+
+
·
1
1
2
(6.28)
A number of important points can be made from these last equations. As with stepgrowth
polymerization in a batch reactor, the extent of reaction must be almost unity before a large NACL
can be achieved. High conversion products usually cannot be made economically in a single
CSTR because this would require the CSTR to operate at almost complete conversion, and
would require a reactor of very large volume. In addition, note that as the extent of reaction
72
approaches unity (giving a highmolecularweight product), the polydispersity increases
without bound. For both these reasons, stepgrowth polymerization is rarely carried out in a single
CSTR. A series of CSTRs or, preferably, a PFR is preferred if stepgrowth polymerization is to
be carried out continuously.
In summary, because the lifetime of a growing polymer chain is equal to its residence time
in the reactor, the effect of the residence time distribution causes extreme broadening of the
molecular weight distribution during stepgrowth polymerization in a CSTR. The constancy
of the polymerization environment, which acted to narrow the distribution in free radical
polymerization, has an insignificant effect in stepgrowth polymerization.
6.3.1: Model for continuous stirred tank step growth polymerization reactor
T a E
p
V Residence
Time(θ )
Q
θ p
k
P
1f
M
6.3.2: MATLAB SOLUTION OF MODEL
V=5000
Q=input('enter flow rate= ');
theta=V/Q
73
RT Ep
P
ae k
/ −
·
Q
V
· θ
f p
P k
w
1
1 θ µ + ·
( ) ( ) [ ]
f p
f p f p
n
w
P k
P k P k
D D
1
2 / 1
1 1
2 1 1 1
θ
θ θ
µ
µ
+ + − +
· · ·
( ) [ ]
2 / 1
1
1
2 1 1
f p
f p
n
P k
P k
θ
θ
µ
+ + −
·
kp=10;
p1f=.01;
mun=(kp*theta*p1f)./(1+(1+2*kp*theta*p1f).^.5);
muw=1+kp*theta*p1f;
D=muw./mun;
disp(mun)
disp(muw)
disp(D)
SOLUTION
V =
5000
enter flow rate= 5
theta =
1000
7.5887
101
13.3092
Effect of residence time on weight avg.degree of polymerization
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0
200
400
600
800
1000
1200
muw
time(sec)
Effect of residence time on MWD
» theta=[1:1800];
» kp=10;
74
» p1f=.01;
» mun=(kp*theta*p1f)./(1+(1+2*kp*theta*p1f).^.5);
» muw=1+kp*theta*p1f;
» D=muw./mun;
» plot(theta,D)
»
0 200 400 600 800 1000 1200 1400 1600 1800
0
2
4
6
8
10
12
14
16
18
20
MWD
residence time(sec)
effect of conversion on weight average degree of polymerization
» p=[0:.09:.90];
» muw=(1+p.^2)./(1p).^2;
75
» plot(p,muw)
» mun=1./(1p);
» D=muw./mun;
» plot(p,D)
»
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
10
20
30
40
50
60
70
80
90
100
conversion
mun
Effect of conversion on molecular weight distribution
» p=[0:.09:.90];
» muw=(1+p.^2)./(1p).^2;
» mun=1./(1p);
76
» D=muw./mun;
» plot(p,D)
»
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0
2
4
6
8
10
12
14
16
18
20
MWD
conversion
6.4: Reactor Dynamics
As we have seen, the use of a CSTR or CSTR train for polymerization reactions may be justified
in some cases by kinetic considerations. However, before implementing CSTR polymerization,
the engineer should be aware of the unique dynamics associated with exothermic and/or
77
autocatalytic reactions in a CSTR. Consider an irreversible firstorder exothermic reaction in a
CSTR. The rate of thermal energy release by reaction can be plotted versus temperature as shown
by the curve Q
g
in Fig. 5.1. At low temperatures, the reaction rate is low and the slope of Q
g
is slight. At high temperatures, the reactor is operating at a high level of conversion (low
reactant concentration), and additional increases in
Temperature,T Figure 5.1 Heat balance multiplicity during
exothermic reaction in a CSTR.
temperature result in a negligible increase in reaction rate and heat evolution. If the reactor is
jacketed, the rate of heat removal (for fixed jacket temperature) is linear with reaction
temperature. Thus, depending on operating conditions, the rate of heat removal may be represented
by the various heat removal lines marked Q
r
in Fig. 5.1. Because, at steady state, the rate of heat
generation must equal the rate of heat removal, steadystate conditions can exist only at the
intersection of the Q
g
and Q
r
curves. Depending on operating conditions (the slope and position of
the Q
r
line), there may be one or three steady states. In the case of three steady states, it may be
seen easily that the upper and lower steady states are stable because perturbations in
temperature will result in the system returning to its original position when the perturbation is
removed. The middle steady state, however, can be seen to be unstable because any perturbation will
drive the system away from the middle steady state and toward the upper or lower steady state
(depending on the direction of the perturbation). This type of heat balance multiplicity is
common in CSTR polymerization due to the highly exothermic nature of polymerization
reactions.
A similar phenomenon can be observed in isothermal free radical polymerization in a CSTR.
Figure 5.2 shows the rate of polymerization versus monomer conversion for the free radical
solution polymerization of methyl methacrylate.
78
Rate of
Polymerization
Monomer Conversion
Figure 5.2 Mass balance multiplicity during free radical solution polymerization in a CSTR.
Unlike a more common reaction in which the rate of reaction falls monotonically with
conversion, the rate of reaction rises with conversion due to the onset of the gel effect. Thus,
the system can be thought of as autocatalytic. At high conversions, the polymerization becomes
monomerstarved and the rate of polymerization falls to zero. At a fixed residence time, there
must be a specific rate of polymerization to produce a given monomer conversion. Thus, the
mass balance is represented by the dotted lines in Fig. 5.2. The slope of the mass balance line
will vary with operating conditions, but it will always pass through the origin because at zero
reaction rate the monomer conversion is zero. Inspection of Fig. 5.2 will reveal that for mass
balances (operating lines) with slopes between the two dotted lines, three steady states will exist
because an intersection of the reaction rate curve and the operating line define a steady state.
This may be better seen by referring to Fig. 5.3, where monomer conversion has been plotted
versus reactor residence time. A similar plot will result from the heat balance multiplicity.
It may be seen that, over a range of residence times, three values of monomer conversion are
possible. As before, the upper and lower steady states are usually stable, whereas the middle
steady state is not. Steadystate multiplicity can be an operational problem for a number of reasons.
If one wishes to operate at an intermediate level of monomer conversion, one may be forced to
operate in the unstable region, relying on closedloop control to stabilize the operating point.
This is tricky at best. Additionally, the steady state (upper or lower) to which the system goes
on startup will depend on how the startup is effected. A careful startup policy may be needed
to assure that the system arrives at the desired steady state. Finally, large upsets in the process
79
Residence Time = Q/t
r
Figure 5.3 Multiplicity during isothermal CSTR solution polymerization effect of solvent
volume fraction (0). Dotted line: 6 = 0.25; dashed line: 0 = 0.30; solid line: 6 = 0.35.
may move the system from the desired (upper or lower) steady state to the other (stable) steady
state. A system designed to operate at the lower steady state may not have the heat transfer
capacity to operate safely at the upper steady state. A system designed to operate at the upper
steady state will be operating way below design product yield at the lower steady state.
Additionally, the product quality (MWD, CCD, etc.) will be different for the two operating points.
The polymerization reactor designer should be aware of the potential for multiplicity and, if possible,
design the system to operate outside the region of multiplicity.
CSTR polymerization reactors can also be subject to oscillatory behavior. A nonisothermal
CSTR free radical polymerization can exhibit a damped oscillatory approach to a steadystate,
unstable (growing) oscillations upon disturbance, and stable (limit cycle) oscillations in which
the system never reaches steady state and never goes unstable, but continues to oscillate
with a fixed period and amplitude.
Damped oscillations will result in lost productivity because the product during these transients
may be offquality. Unstable oscillations will, of course, preclude continued operation Limit
cycle oscillations, while not unstable; will result in a product whose quality (MWD, CCD, etc.)
80
varies with time in a cyclic fashion. In most cases, this is undesirable. As in the case of
multiplicity, the polymerization reactor designer must be aware of the potential for oscillatory
phenomena and should attempt to specify operating conditions at which these phenomena do not
exist.
CHAPTER# 7 OPTIMIZATION OF POLYMERIZATION REACTORS
Process modeling and optimization studies in the past 50 years have contributed to the
development of reliable and efficient methods and software that include process
simulators (e.g., Hysys, Aspen Plus) and optimizers (e.g., Solver in Excel, GAMS). The
majority of these studies were for single objective optimization. However, industrial
processes are complex and have multiple objectives, therefore here we define multi
objective optimization.
What is Multi objective Optimization ?
Mostly industrial processes have multi objectives, such as to maximize valuable products,
maximize selectivity, minimize undesirable wastage, minimize raw materials and energy,
minimize environmental impact etc.; these are often contradictory objectives. Hence,
industrial processes need to be optimized for two or more objectives simultaneously. This
is known as multi objective optimization.
Figure: Pareto optimal solutions obtained from simultaneous maximization
of styrene flow rate and selectivity for an industrial styrene reactor.
81
Reactors are often the critical stage in a chemical process. Recently, demands on the
design and operation of chemical processes are increasingly required to comply with the
safety, cost and environmental concerns. All these necessitate accurate modeling and
optimization of reactors and processes. The project objectives are to develop accurate
models for selected industrial reactors, and to optimize reactors and associated operations
for multiple objectives. Key steps in our approach are: identification of important
industrial reactors and processes, simulation using rigorous models with detailed kinetics,
validation of models using industrial data and optimization using a genetic algorithm for
multiple objectives
Multiobjective optimization of a Polyester Reactor
Multiobjective optimization of a thirdstage, wipedfilm polyester reactor was carried out
using a model that describes an industrial poly (ethylene terephthalate) reactor quite well.
The two objective functions minimized are the acid and vinyl end group concentrations in
the product. These are two of the undesirable side products produced in the reactor. The
optimization problem incorporates an endpoint constraint to produce a polymer with a
desired value of the degree of polymerization. In addition, the concentration of the di
ethylene glycol end group in the product is constrained to lie within a certain range of
values. Adaptations of the nondominated sorting genetic algorithm have been developed
to obtain optimal values of the five decision variables: reactor pressure, temperature,
catalyst concentration, residence time of the polymer inside the reactor, and the speed of
the agitator. The optimal solution was a unique point. Problems of multiplicity and
premature convergence were encountered.
Process industries aim at maximizing their production capacities while simultaneously
maintaining the product quality. Usually, there exists a tradeoff between these two
requirements. This is particularly true in the manufacture of polymers where the
properties of the product are crucial and reactors have to be operated under conditions
that yield products that satisfy relatively narrow specifications. At the same time, the
operating variables must be at optimal condition to maximize the throughput. It is well
established that the average molecular weight of the polymer produced determines
several of the material’s important physical properties, for example, strength and impact
resistance. In addition, the concentrations of a few side products need to be below low
limits or between narrow limits to ensure some other properties, like color and dyeability,
to lie within specifications.
Thus, the design and operation of polymerization reactors require optimization using
multipleobjective functions and constraints, which are often conflicting.
we consider the optimization of reactors producing fibergrade polyethylene terephthalate
PET. As the petroleum/ petrochemical/ polymer sector is going through a lean period,
any form of costcutting measure, which can be basically achieved from process
optimization, would be beneficial for the polyester fiber industry. Commercially, PET is
manufactured in three stages, using continuous reactors. The first_esterification.stage is
carried out at atmospheric pressure and at 270280C. The raw materials commonly used
are a molar excess of ethylene glycol EG. and either purified terephthalic acid PTA. or
dimethyl terephthalate DMT.. Our study is based on the now popular PTA route. PTA
82
and EG are usually processed in a series of CSTRs or a plugflow reactor with a recycle
in the first stage.
A polycondensation catalyst, antimony trioxide, is injected in small concentrations
(0.030.05% by weight) into the oligomer stream leaving this reactor. The second
prepolymerization stage is carried out either in one or two agitated vessels under reduced
pressures, at about 24 kPa (1530 mm Hg) and 270280C. The degree of polymerization
DP attains a value of about 3040 in this stage. The prepolymer so produced undergoes
final polycondensation in a finishing or thirdstage, wiped film reactor in which the
pressure is maintained quite low at 0.1330.266 kPa (12 mm Hg), at temperatures of
about 285295C. Since the reaction mass is very viscous under these conditions, the
finishing reactor has a special construction to enhance mass transfer and the removal of
by product, ethylene glycol, in order to drive the reaction in the forward direction and to
yield a product having a high value of DP. The finisher is usually a jacketed cylindrical
vessel with a horizontal agitator, with large screens mounted on the latter. The reaction
mass in the thirdstage reactor is usually heated by condensing vapors in a jacket.
Table 1 presents the reaction scheme commonly used as well as the rate and equilibrium
constants some tuned, the others from the literature Ravindranath and Mashelkar, 1984,
to explain the industrial data on the wipedfilm reactor. The complete set of model
equations is summarized in Table 2. Table 3 gives the feed conditions used for
optimization of the operation of the industrial reactor. A sensitivity analysis was also
carried out on the model to develop an intuitive understanding of the behavior of the
industrial wipedfilm reactor.
This helped in developing a meaningful optimization problem involving several different
sets of objective functions and endpoint constraints associated with the requirement of
product property specifications.
The two important objective functions for this system are the minimization of the acid
and the vinyl end group concentrations, respectively in the product of the finisher. The
acid end group makes the polymer susceptible to hydrolysis during the downstream
operations and leads to breakage of the filaments during spinning, where the humidity is
very high. The vinyl end groups have been shown to be responsible for the coloration of
PET because of reactions not well understood right now, and not included in Table 1.
Hence, the minimization of these two end groups improves the quality of the polymer
product. The reduction of simultaneously increases the rate of polymerization
of the acid end group catalyzed polycondensation reaction and helps maximize the
throughput this catalytic effect is not directly incorporated in the model, however.. The
important endpoint constraint is to produce polymer with a desired value of the degree of
polymerization DP. Although the diethylene glycol DEG.end groups affect the
crystallinity and hence the melting point of the polymer unfavorably, they do improve the
dyeability of the fiber. Therefore, it is preferred to have a certain allowable range for the
concentration of DEG end groups in the fibergrade polyesters. An inequality constraint
is therefore imposed for the DEG end group concentration in the product, as per
industrial requirements. In addition, a further inequality constraint on the maximum
allowable limit for the acid end group concentration is imposed to ensure that it is not
only minimized but also lies below an upper limit.
83
84
85
Interestingly, it was found that the multiobjective optimization problem described earlier
had a unique optimum solution, and that no set of several equally good optimal points
was obtained. Also, it was found that the same values of the two objective functions were
obtained for several sets of values of the operating conditions values of the decision
variables used under optimal conditions, and so this optimization problem is associated
with what we shall refer to as ‘‘multiplicity.’’ A suitable methodology was developed to
86
study the problems associated with this multiplicity. Optimal solutions were generated for
several values of the DP of the end product that had industrial significance, and it was
found that there was considerable scatter in the results. This was due to the presence of
multiplicity. A curvefit procedure was developed to arrive at nearoptimal solutions,
which would be easier to implement in industry. We also found that the optimal solutions
obtained for the preceding problem gave the lowest residence times of the reaction mass
in the reactor, and so led simultaneously to a higher throughput from the continuous
polymerization unit. A sensitivity analysis was performed to study the effect of the
computational parameters. In addition, a few additional meaningful optimization
problems were solved, and their results compared with those of the first problem.
Formulation
The mathematical model for the present study is based on the general stepgrowth
polymerization kinetic scheme shown in Table 1. The reaction scheme consists of the
main polycondensation reaction along with several side reactions. The side reactions lead
to the formation of the extremely volatile acetaldehyde, A, the less volatile diethylene
glycol, DEG, in the ‘‘free’’ form, and water, W. The other important side products
produced are the acid and vinyl end groups, Ea and Ev, and EDEG, the DEG end groups
‘‘bound’’ form have studied in detail the effect of these side products on the polymer
quality. The various rate and equilibrium constants are also provided in Table 1. Some of
these k20, k90, K1, K5, and K8.differ from those of Laubriet et al. and have been
obtained by tuning the model for the reactor against industrial data Bhaskar et al. The
model equations massbalance equations for the liquid phase and the vaporliquid
equilibrium correlations. and values of some of the required parameters are given in
Table 2. Details of the model development and the tuning of the model
parameters for the industrial conditions are available else where Laubriet et al.; Bhaskar
et al.. In general, the statevariable equations can be written in the form
dx/dz = f(x,u) ; x(z=0)=x0
where x is the vector of state variables, given by
and u is the vector of control or decision variables. The variable, z, represents the
dimensionless axial location.
87
The initialvalue problem IVP is given by the ordinary differential equations ODEs in
Eq.1 and given in Table 2 can be integrated using appropriate values of the feed
conditions given in Table 3. An iterative solution is necessary to simulate the reactor for
any set of conditions because the liquid and vapor phases have different kinds of flow
plug and wellmixed, respectively. Values for the average mol fractions, yj, of three of
the volatile species, EG, W, and DEG, in the vapor phase need to be assumed see Table 4
for the first estimate values. The other volatile species acetaldehyde A is assumed to
vaporize as soon as it is produced, and so, no assumption for yA is required.
Then, the ODEs are integrated from the dimensionless value of z=0 (feed) end to z=1
(product end). The wellmixed vapor phase values of yj are then reevaluated and used as
the estimate values for the next iteration. It is found that the method of successive
substitutions converges for all the conditions studied. The iterations are continued till the
sum of the squares of the yj does not change by more than 1*10
12
between two
successive iterations. The model provides values of DP as well as the concentrations of
the hydroxyl end groups Eg, acid endgroups Ea, diester groups Z, vinyl end groups Ev,
DEG end groups E
DEG
, ethylene glycol EG., water W., and diethylene glycol DEG., as
functions of the axial position in the reactor after convergence is attained.
Five decision variables are used for optimization in this study. These are the
reactor pressure P., temperature isothermal, T., catalyst concentration [Sb
2
O
3
], residence
time of the polymeric reaction mass inside the reactor (θ), and the speed of the wiped
film agitator N. All of these variables can easily be changed in any industrial, wipedfilm
reactor for PET manufacture, including the one being studied, and are therefore used to
obtain the best optimal operating conditions. The multiobjective function optimization
problem described earlier and studied in this work is described mathematically
by:
88
where the subscripts out and d refer to the values at the outlet of the reactor and the
desired values of the product property, respectively. The variables, θ* and N* and
represent dimensionless values, θ/θref and N/Nref , where θref and Nref are values being
used currently in the industrial reactor being studied. These two values are confidential
and are not being provided in this article for proprietary reasons. Meaningful bounds have
been chosen on the five decision variables, u,based on industrial practice. These are given
in Eqs. 3f3j.
Several issues regarding the optimization problem formulated in this study
need to be clarified and justified at this point. One is that we have used the catalyst
concentration as a decision variable for the optimization of Stage 3 even though the
catalyst is added in the form of the glycolate at the inlet of Stage 2. This has been done
because the catalyst concentration used affects the operation of Stage 3 far more
significantly than Stage 2, which is equilibriumcontrolled and is influenced primarily
by the pressure and temperature used in that reactor. Since the catalyst is not consumed
by the reactions, we could use the same catalyst concentration at the feed to Stage 2 as
computed in our study of Stage 3 alone, and no correction is necessary. In fact, the
catalyst concentration is usually measured in industry both at the inlet of the
prepolymerizer and at the outlet of the finisher, and the difference is found to be below
about 5%. Most of this difference is believed to be the catalyst carryover vaporization
with the glycol that is removed at both stages prepolymerizer and finisher. Also, use of
catalyst concentration as a decision variable while keeping the feed to the film reactor
unchanged is justified for the same reasons product from Stage 2 depends on its pressure
and temperature only, and is independent of the catalyst concentration, since the
equilibrium constant is unchanged. The use of the residence time, θ*, as a decision
variable for the film reactor while keeping the feed composition unchanged, can be
justified using similar arguments. We assume that a change in the residence time of the
film reactor could be achieved by changing the flow rate of the reaction mass in this
reactor as well as in the previous two stages; this is the ‘‘connectivity’’ problem, rather
than by changing the holdup volume of the melt in this reactor which would lead to
changes in the value of k
1
a, something that is not well understood quantitatively. Since
the first as well as the secondstage reactors have considerable flexibility in industry, and
since both these stages are equilibrium controlled, a change in the flow rate does not
necessarily lead to any changes in the composition of the feed to the film reactor. The
choice of the five decision variables used in this study, and the use of constant feed
conditions are thus justified and give us a great deal of flexibility for optimization, which
can easily be implemented in industry. The endpoint constraint on DP Eq. 3b.is
incorporated in the form of a penalty function with a large value of the weightage factor,
w
2
(10
8
in this study) as
An additional large ‘‘penalty’’ value, Pe (10
4
in this study), is added to the objective
function, I
1
*, if either of the two inequality constraints in Eqs. 3c and 3d (the maximum
limit of the acid end group concentration at the outlet of the reactor, [E
a
]
out
, and the upper
and lower bounds on the DEG end group concentration at the outlet of the reactor,
[EDEG]out) are violated. If these constraints are satisfied, then the value of Pe in Eq. 4 is
89
taken as zero. This ensures that bad chromosomes in the genetic algorithm used for
optimization are not reproduced in the successive generations, even if several
chromosomes violating these constraints exist in the initial population that is, the
chromosomes violating Eqs. 3c and 3d die out rapidly. Also, a very large value of Pe is
added if the concentrations of three of the vaporizing species EG, W, and DEG in the
vapor phase are higher than their respective vaporliquid interfacial equilibrium
concentrations EG*, W*, and DEG* at any axial location, in order to ensure the
vaporization rather than condensation of these species at all axial locations. Minimization
of I
1
* leads to a decrease in the acid end group concentration while simultaneously
giving reference to solutions satisfying the several requirements just discussed. The use
of a penalty to ‘‘kill’’ chromosomes when important physical constraints are violated is
one of several adaptations of the basic algorithm suggested in this study.
The second objective function, I
2
, is also modified to incorporate the
penalties for the several constraints as discussed for I
1
, and is given by
Minimization of the second objective function leads to a minimum value of the vinyl end
group concentration in the product, thus reducing the product degradation due to thermal
effects and improving the color of the polymer. It can be noted that w
2
>>w
1
, w
3
. Use of
w
1
and w
3
as 10
4
leads to values of w
1
[E
a
] and w
3
[E
v
]of the order of unity. In our earlier
parametric sensitivity study of this reactor, we found that the two objective functions
used here conflicted.
90
91
92
Figures 1 and 2 show all the feasible solutions that is, those satisfying the endpoint
constraints in Eqs. 3b3d. at different values of the generation number, Ng. In
the initial generations, the feasible points are quite scattered in the [Ea]out  [Ev] out
space, but as the chromosomes evolve over the generations, the feasible points move
toward a unique point. There are over 45 almost indistinguishable points in Figure 2
(Ng=50). Therefore, the solution of our multiobjective function optimization problem
with endpoint constraints turns out to be unique, and is given by
Table 5 shows the values of the five decision variables that correspond to the optimal
point (reference run), as well as the corresponding values of several other product
characteristics.
The uniqueness of the optimal solution is checked by performing an optimization using a
single objective function. The objective functions, I
1
* and I
2
*, were taken one at a time,
and optimal solutions were generated under the same conditions as given in Eqs. 3b3j.
The (unique) optimal solution obtained was found to be identical to that obtained from
the NSGA. This means that minimizing either one of the side product concentrations,
[Ea]out or [Ev]out , simultaneously leads to a minimization of the other side product.
This is quite interesting. In fact, similar unique solutions were obtained for other choices
of DPd, as discussed later. It should be emphasized that GA (as well as NSGA) is an
extremely robust technique Goldberg, 1989 and converges to the global optimal solution
and also gives all the optimal solutions where several exist. It is possible that when a less
robust technique is used to, say, minimize I1, the corresponding value of I2 may not be at
the minimum value given by NSGA. Figure 2 (Ng=50) and Table 5 compare our optimal
solution (indicated by filled circles) with the values corresponding to the current
operation of the industrial wipedfilm reactor indicated by the open circle Bhaskar et al.,
2000. It is observed that the optimal values of both the objective functions are, indeed,
better than those obtained under the present operating conditions of the industrial reactor.
Lower values for both the acid and vinyl end group concentrations are obtained using a
lower value of the residence time (θ*=.967) and at a lower temperature. For the reactor
93
under consideration, this decrease in θ* corresponds to an increase of about 34% in the
throughput without any major design modifications. The product is also of better quality.
The lower temperature under optimal conditions would also lead to a reasonable amount
of energy savings. As discussed earlier, the increase in the throughput can easily be
achieved industrially by an increase in the rate of production of the prepolymer in Stages
1 and 2, without affecting the feed concentrations of the film reactor. In fact, an operator
in the plant has more flexibility to control the process variables in the earlier reactors than
in the finishing reactor, since the operating conditions are far less severe in those reactors,
and the product quality is not too sensitive to changes in the process variables. Optimal
solutions were then generated for several different choices of DPd ranging from 80 to 84,
values relevant to the fiber industry. These optimal solutions were generated using the
same computational parameters (for reference values, see Table 4). In all of these cases,
unique optimal solutions were obtained. The solutions are shown in Figures 35 filled
circles for different DP
d
. Some amount of scatter is observed in these plots. The scatter in
the optimal results was explored further. It has been suggested by Goldberg 1989 that
even though GAs are quite efficient in reaching the globaloptimum region, they are not
guaranteed to reach the precise location of the globaloptimum, particularly for complex
problems, and may attain premature convergence. In fact, a similar scatter was observed
earlier by Sareen and Gupta 1995.for an industrial nylon6 reactor. The scatter was found
to indicate the presence of compensating effects interaction of the several decision
variables used in this
study. We found that even in the case of the reference run Figure 2, Ng=50., use of
different values of Sr, the seed for generating the sequence of random numbers, led to the
same values of the two objective functions, but with several different sets of values of the
five decision variables. We refer to this phenomenon as the problem of multiplicity.
There are two methods of reducing this effect. One is to reduce the number of decision
variables. This is not very desirable for industrial operations, since it eliminates the
flexibility offered by the use of several decision variables where they are, indeed,
available for control. Hence, it was decided to continue with the set of five decision
variables and try some kind of ‘‘smoothing’’ of the results and obtain nearoptimal
solutions.
Fortunately, it was found that the results were not too sensitive to small
variations in the operating conditions. A linear curve fit of the plots (Figures 35) of the
five decision variables, P, T, [Sb2O3], θ*, and N*, as a function of DPd was performed.
The curvefit equations are given in Table 6. The properties, [Ea] out , [Ev]out, DP out,
[Eg]out , and [E
DEG
]out , were then predicted using these curvefit operating conditions in
the simulation code Bhaskar et al., 2000. The predicted results are also shown in Figures
35 (unfilled circles), plotted as functions of ( DP) out rather than DPd, since these two
need not be the same for the curvefitted case. Here we call
94
the solutions obtained using the curvefit procedure nearoptimal solutions. These near
optimal solutions are characterized by operating decision variables, which are less
cumbersome to implement, and also lead to better values of [Ea]out , [Ev]out , and
[E
DEG
]out . We recommend using these nearoptimal charts because of their convenience.
Smoothing the optimal solutions while keeping all the decision variables practically
available to us, is an important adaptation offered in this study. It also could be used for
other complex problems where the GA or other algorithms has premature convergence.
The optimum values of pressure and temperature predicted for the nearoptimal solutions
are about 0.133 kPa (1 mmHg) and 565 K. Figures 35 suggest that one can increase
DPout (without affecting the product quality) primarily by varying the catalyst
95
concentration and the speed of the agitator. These actions are quite commonly practiced
in industry.
We explored several other optimization problems besides that given in Eq. 3.
Since minimizing [Ea]out simultaneously minimizes [Ev]out , we tried to minimize θ*
and [Ev]out instead to see if Pareto solutions are obtained. The exact problem
studied is given by
Minimizing θ* leads to an increase in the throughput while minimizing [Ev]out (and thus
[Ea]out) improves the product quality. Once again, unique optimal solutions were
obtained for different values of DPd. In fact, the optimal solutions were exactly the same
as those obtained for the problem described by Eq. 3. It is interesting to observe that
minimizing [Ea]out and/or [Ev]out simultaneously minimizes θ* , which leads to a higher
throughput. This is not really surprising, since minimization of [Ea] throughout the
reactor (which accompanies
96
minimization of [Ea]out ) simultaneously leads to an increase in the rate of
polymerization. In fact, reaction 9 in Table 1 suggests that minimization of [Ea] and [Ev]
will lead to a suppression of this reaction, and thus to higher values of [Z], which in turn
leads to faster attainment of DPd. The simultaneous minimization of three objec
97
tive functions [Ea]out , [Ev]out , and θ* using NSGA would thus be expected to lead to a
unique optimal point, also. We next tried to fix some of the decision variables at their
current operating conditions of the industrial reactor, and solved the 2objective function
optimization problem given below:
Once again, a unique optimal solution is obtained. Interestingly, the optimal operating
temperature and the residence time were found to be the same as the values currently
used in both the industrial reactor, and in our previous study (Bhaskar et al., 2000). This
indicates that the present industrial reactor is, indeed, operating at its optimal condition
when the three decision variables, P,(Sb2O3), and N*, are constrained to lie at their
current operating values. However, improvements are possible if these constraints are
removed.
We also explored the effect on the optimal solutions of varying the feed concentrations
(while keeping DPf constant, somewhat arbitrarily, at 40.0). Since the feed to the
finishing reactor could be varied by changing the process conditions pressure,
temperature in the prepolymerizer, such a preliminary study gives us some insights about
whether it would be useful to carry out a more elaborate systemoptimization, using
models for all three reactors used in the process. Some information on this was obtained
from the work of Saint Martin and Choi 1991., who studied the effect of the feed
conditions on the DP of the product and the EG removal rate. Since DPfwas fixed at 40,
we varied [Eg]f which, in turn, alters the value of [Z]f .. The multiobjective optimiza
tion problem described by Eqs. 3a3j.is solved for two different concentrations of [Eg]f
(0.39 and 0.41 kmol/m3). The effect of other feed concentrations was not found to be too
significant and the polycondensation reaction is controlled primarily by [Eg]f and [Z]f.
The results are tabulated in Table 5, which suggests (see results for [Eg]f = 0.41
kmol/m3). that a more detailed system optimization could possibly provide higher
98
throughputs (lowerθ*), while giving values of [Ea]out and [Ev]out that are only slightly
worse than those obtained for the present case. Since NSGA involves several
computational parameters, it is very important to analyze how the final results are
affected by variations in these computational parameters. The computational parameters
involved are the number of generations, Ng, the number of chromosomes, Np,
chromosome length, I
chrom
, the seed, Sr , used for the random number generator, the two
parameters involved in controlling the niche count of each of the chromosomes (α
sh
, the
exponent controlling the sharing effect, and σ , the maximum normalized distance in u
space between any two points), crossover probability, pc, mutation probability, p
m
, and
the weightage factors, w1, w2, and w3, used in the penalty functions. The effect of Ng
was presented in Figures 1 and 2. It is clear that the maximum number of generations
(which is used as the termination criterion) should be set high enough to ensure
convergence, while not too high to avoid wastage of computational time. The population
size, Np, has to be large enough to allow diversity in the population. Too low a value of
Np leads to erroneous results. The final results were found to be unaffected if Np was
increased from 50 to 200. Therefore, Np=50 was used in this study. The length of the
chromsome, Ichrom, affects the accuracy of the solution. The length of the chromosome
used in this study was 160 bits (5 decision variables *32 bits each). Increase in Ichrom
did not lead to changes in the results. No significant change in the final values of the
objective functions was obtained if Sr was varied, but the effect of multiplicity was
observed, as described earlier. It was also observed that the final results were quite
insensitive to αsh, the exponent that controls the sharing effect (Srinivas and Deb, 1995).
The maximum nichecount distance, σ, would have played a more important role if the
solutions for this multiobjective optimization problem had been a Pareto set, since this
parameter affects the spread of solutions on the Pareto set. Since a unique optimal
solution was obtained for this problem, s has a negligible effect on the final results. The
two most important parameters that influence the course of NSGA are the crossover, pc,
and mutation, pm, probabilities. Figure 6 describes the effect of the crossover probability
in the range of 0.6 to 0.7. The rate at which the final solution is attained was different for
different values of the crossover probability, but the final results were not affected by its
choice, at least in the range studied. A crossover probability of 0.65 (reference value) was
used in this study. The mutation probability was varied from 0.0 to 0.005, and it was
found to only marginally affect the final solution. As expected, better results were
obtained if a nonzero value of the mutation probability was used. The optimum value of
[Ea]out and [Ev]out are 1.024*10
3
kmol/m3 and 7.59*10
4
99
kmolrm3, respectively, when p s0, compared to 1.0=10y3 m kmolrm3 and 7.549=10y4
kmolrm3 for values of pm of 0.004 and 0.005. A reference value of pm of 0.004 was used
in this study for the other cases. The effect of the three weightage factors, w1, w2, and
w3, was also studied. It was observed that w1and w3do not affect the final results much.
In fact, w1and w3are used only to obtain the values of the orderofunity objective
function. In contrast, w2 affects the nature of the solution tremendously. Lower values of
w2 =10
8
., do not give rise to a unique solution, since they permit and give deviations of
DPout from DPd of +0.15,0.15 w2=10
3
.. Such deviations are
100
usually not desired in the industrial scenario, at least under steadystate conditions. Figure
7 shows the variation of [Ea]out with [Ev]out for w
2
=10
3
. A considerable amount of
scatter is observed in plots of the decision variables vs.[Ea]out . Figure 7 should not be
mistaken for a Pareto set, since the requirement of DPout = DPd is not satisfied precisely
(the error is +0.15,0.15). Lower values of the penalty, Pe, are not preferred because we
do not want the ‘‘bad’’ chromosomes to get reproduced in the successive generations. If
Pe was decreased from 10
4
to 10
3
, the constraints in Eq. 3 would be violated by most of
the chromosomes in the population. The present study emphasizes the importance of
obtaining better correlations for the optomechanical properties (for example, color, melt
index, fiber strength) in terms of the concentrations of the different chemical species
present in the polymer product. A better understanding of these correlations would lead to
improved optimal designs and operation. This was emphasized by Bokis et al.1999.when
they addressed the problems involving the modeling of polymerization processes for
commercial simulators.
Conclusions
Minimization of the degradation side products, [Ea]out and [Ev]out , was carried out
using an adapted NSGA technique. P, T,[Sb2O3], θ* , and N* were used as the decision
variables. An endpoint constraint was imposed on DPout , while the concentration,
[Ea]out , was constrained to lie below a maximum value. In addition, an allowable range
was prescribed for [E
DEG
]out. A unique solution was obtained for this multiobjective
optimization problem. No Pareto set of noninferior solutions was obtained. Unique
optimal solutions were generated for a range of values of DPd that are of importance in
the polyester fiber industry. Optimal control variable charts were constructed for
producing polymers with different values of DPout that are less cumbersome to use, by
using a smoothening procedure. It was found that minimizing [Ea]out and/or [Ev]out
simultaneously leads to a minimization of the residence time,θ*. The operation of an
industrial reactor is discussed in reference to our optimal solutions.
101
102
REFRENCES
BOOKS:
1. Robert.L.Laurence,”Polymer Reaction Engineering”,
Dow chemical company.
2. Franks, “Modeling And Simulation In Chemical Engineering”,
WileyInterscience,1972
3. Morton M Denn “Process Modeling”,Longman Scientific& Technical”.
4. Octave Levenspiel “Chemical reaction engineering”,3
rd
edition,
John Wiley &Sons.
5. J.Brandrup, “Polymer Handbook” 4
th
edition, John Wiley &Sons.
6. F.Joseph Schork, “Control Of Polymerization Reactors”, Marcel Dekker,Inc.
7. Joel R. Fried, “Polymer Science and Technology”, Prentice Hall of India
1999.
8. Billmeyer, “Text book of Polymer Science ”, 3
rd
Edition,
John Wiley and Sons.
9. Max.S.Peters,”Plant Design and Economics for Chemical Engineers”,
5
th
Edition, Mc Graw Hill.
Web Sites:
1. solver.com
2. mathworks.com
3. nus.edu.sg
103
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.