8/11/2019 File 41295

1/6

Computers hem Engn

Elsevi

009 1354(95)00053-4

DYNAMIC SIMULATION OF A PVC SUSPENSI

A. DIMIAN, D. van DIEPEN, G. A. van der

Department of Chemical Engineering, University of Am

Nieuwe Achtergracht 166,1018 WV The Netherlands

ABSTRACT

The scope of this contribution is to present a comprehensive methodology

modelling of an industrial PVC suspension reactor. The paper analyses the

brings also some new insights on the polymerisation kinetics, the dynamic

action of the control system. The model is implemented in SPEEDUPTM an

optimisation of process operation and recipe development, as well as a cor

training simulator.

KEYWORDS

Dynamic simulation; Polymerisation, PVC, Batch Reactor

INTRODUCTION

Very large scale tank reactors, more than 100 m3, are currently used to pro

polymerisation. An accurate knowledge of the dynamic behaviour of the re

means to optimise the productivity, while keeping flexibility in obtaining d

dynamic model of an industrial PVC reactor has been published several ye

It considered a two-phase polymerisation model, time dependent heat trans

cooling system driven by a PID controller. Important progress has been reg

modelling the polymerisation kinetics [2,3], but the scale-up from laborator

still an open research problem. Recently Bretelle and Machietto [4] present

of PVC-S focused on operator training. Summing up, we may say that there

8/11/2019 File 41295

2/6

S428

European Symposium on Computer Aided Process Enginee

jacket works as an overtlow loop, the flowrate of the external thermal age

fulfil the reaction heat balance. An external condenser can remove the hea

monomer condensation. The reactor pressure is constant during the polym

decline after a critical conversion Xc. At this point the polymerisation ra

here the heat removal may be critical. The final conversion is slightly abo

critical conversion rc and final polymerisation zf, respectively, are charac

good operation should maximise the reaction rate while respecting produc

thermal regime.

During the polymerisation the particle morphology changes a lot, that has

behaviour of the suspension. A two phase model has been generally acce

Each drop consists of a monomer-rich phase and a polymer-rich phase, sw

effect will accelerate the reaction rate from the beginning up to a critical

complex phenomenon is essential for productivity, reactor operation and

start the simulation we evaluated the existing models [2,3]. The link betw

morphology is not yet satisfactorily modelled.

Fig. 1 Reaction system

0 60 12 la0

lime,r

Fig. 2

The overall heat balance of the reactor is:

~MiCi,~=Q,+Q,-Q,-Q,-Q,

where QR is the heat generated by reaction, QM is the heat generated by m

8/11/2019 File 41295

3/6

European Symposium on Computer Aided Process Engineerin

The inlet jacket temperature Twl is adapted continuously to heat balance r

thermal agent (steam for heating period, chilled water for cooling period). T

to be introduced explicitly in the model. In this case a cascade configuration

gives a faster reaction compared with a single controller (figure 1). The PID

reactor temperature supplies the setpoint of a slave PI, which controls the i

manipulating the flowrate F, of the external heat source.

SIMULATION RESULTS

The figure 2 describes a typical simulation run, in terms of pressure, tempe

reaction rate, for Perkadox 16l4OW 700 ppm at 57 C. The shapes of these

concentration and on temperature policy. A constant maximum reaction rat

would be desirable. This might be reahsed by multi-initiation and programm

kinetics modelling in this work has been done by using the general frame p

some modifications which will be discussed below.

The computation of polymerisation rate is based on the relation:

The two terms represent the rate of polymerisation in the monomer phase (

phase (2), respectively. The radical concentrations [R.], and [ R-l2 are com

operating conditions, the distribution of monomer between phases and the

elementary reaction steps. The fidelity of the simulation depends heavily o

modelled. According to the free volume theory, the rate constant of either e

polymer phase may be expressed by the equation:

k, =

k,, exp[-A * 11 V, -

11 V,,,,)]

where A* is a fitting parameter, VP is the polymer phase volume and V

parametric study shows that the free volume correction accounting for J?Ytff&

polymer phase gives the highest sensitivity. The most significant conclusio

The termination constant Kt2 has a strong influence on auto acceleratio

polymerisation time. A variation of 10 in the critical factor gives -20

The propagation constant Kp2 and the initiator decomposition constant

8/11/2019 File 41295

4/6

s430

European Symposium on Computer Aided Process Engineerin

This approach enables to examine directly the contribution of each elemen

necessary. The integration of approximately 250 differential equations plu

algebraic equations can be done with reasonable time in SPEEDUPm.

The evolution of the gel effect is depicted in the figure 3 by curves display

radicals in the monomer and in the polymer phase, as well as their precipita

macro radicals in the monomer phase reaches a maximum at a mid convers

zero at Xc. The number of macro radicals in the polymer phase increases c

maximum at Xc, then drops dramatically. The precipitation rate of the mac

peak at low conversion, when the polymer aggregates are still in formation

significant fall, when the particle size increases. Then the precipitation rate

peak and afterwards the curve follows the decline of radicals in the monom

Another important feature in operation is the monomer balance, depicted i

others, this shows that a significant amount of monomer in water exists fro

what is considered the critical conversion region. This may have two con

Firstly, an important contribution of condensation at the critical conversion

monomer apparently disappears as a distinct phase. Secondly, the critical

conditions may be higher than measured in laboratory.

Beat transfer

The basic characteristics are presented in figure 5. Early attempts have bee

transfer coefficient suspension/reactor wall. This is time variable because

suspension viscosity during polymerisation. Its computation is not accurate

unpredictable suspension viscosity. The figure 5 displays a realistic trend,

effect of increasing polymer phase with the volume shrinkage by polymeri

method the viscosity increases roughly by an order of magnitude, which ca

reduction in the heat transfer coefficient (-30 ). The minimum value cor

conversion where the maximum heat generation takes place. Fortunately, a

monomer available to remove heat by condensation, as the monomer balan

time the gas cap area available for removing heat by condensation increase

the reaction volume.

The reactor wall has also an important heat resistance. When all these are

transfer coefficient of the whole reactor becomes practically constant durin

Consequently, the accuracy in predicting the heat transfer coefficient of th

for design and operation. Both these aspects should better consider the dyn

than the static heat transfer contributions.

8/11/2019 File 41295

5/6

European Symposium on Computer Aided Process Engineerin

The development of the molecular weight distribution (MWD) takes place

polymer phase. Number-averaged, weight-averaged MWD, instantaneous a

the polydispersity index can be calculated from the kinetic model, using the

direct integration of differential equations for polymer species. This is poss

in SPEEDUPTM. In this case chain transfer, branching and any other type o

polymer weight distribution may be easier introduced in the kinetic scheme

essential for polymerisation rate, may become important for predicting end-

especiallyat high conversions.

The development of instantaneous Molecular Weight Distribution (MWD)

in the figure 8. Initially shorter length chains are formed in the monomer ph

phase, where the termination rate is affected by the gel effect. The initial po

slightly greater than 1.5. Then the mean molecular weight will increase, bec

contribution of the polymer phase. However, up to the critical conversion th

practically constant and close to 2, being controlled by the chain transfer to

critical conversion the situation changes significantly. The instantaneous M

to lower values, this time because of the diffusion controlled chain transfer,

monomer, but also with the polymer. The M, will decrease slower than M

the polydispersity index. However, these important changes affect only slig

MWD.

CONCLUSIONS

The paper presents a comprehensive dynamic modelling of a large scale su

combining reaction kinetics, heat transfer dynamics and control system acti

illustrate the approach. The kinetics remains the key factor. The recent mod

frame to describe both the global rate and the molecular weight distribution

complex and very sensitive regarding some parameters. Simpler models wi

be equally efficient for the overall kinetics.

The article evaluates the contribution of different factors in controlling the

heat house-keeping. Despite a sharp increase in suspension viscosity, the ov

slightly affected because of the reaction volume shrinkage. The heat remov

condensation may be more important. A cascade PID controller can keep th

within 1 . The simulation can predict quantitatively the effect of temperatu

molecular weight distribution. The presented modelling offers a rational bas

and for operation optimisation.

8/11/2019 File 41295

6/6

h

a

e

c

s

W

m

K

0

0

6

j

1

7

i

1 2 3 3

V

s

y

c

Q

Q

m

b

T

C

[

R

m

c

o

m

m

O

u

W

~

0

6

1

3

=

6

1 =

3 4

S

o

p

m

c

m

v

f

a

o

P0

N

P

0

6

1

=

1

2

i

2

3

3 3

4