A Kinetic Model for High Density Polyethylene Slurry Process

Xiaoyan Xu ,Jiahua Le & Jing Jiang

(Yangzi Petrochemical Corporation, P. R. of China)

Introduction
Yangzi Petrochemical Corporation(YPC) is a major producer of high density
polyethylene (HDPE), and operates multiple HDPE production lines in Nanjing,China.
The process is divided into eight areas: catalyst preparation, polymerization section,
separation and drying, pelletizing , packing and palletizing, low polymer treatment ,
solvent distillation and utilities and process auxiliaries. The characteristic of the process
is a series of grade products can be manufactured from high productivity catalyst .

The scope of the kinetic model for HDPE slurry process covers the polymerization
reactor, the heat removal system, the monomer-solvent separation and the recycle system
for grade 5000S. The core of the model is the polymerization reactor. This model is
fundamentals based, containing rigorous polymer science and chemical engineering
equations and capturing plant measurements, lab experimental data and the process
experience of YPC’s best process engineers. The model predictions have been tuned to
match measured plant values. Thus the fundamentals based “validated ” model gives
accurate non-linear predictions within the range of operation with which was validated
and reasonable predictions outside that range.

It has been a joint effort of YPC and AspenTech . AspenTech has modified their kinetic
scheme to accommodate the HDPE slurry process of YPC. We have been able to modify
and use POLYM ERS PLUS to build the kinetic model for meeting the need of design and
providing valuable information for production. At first, the model was built for grade
5000S on the basis of the design capacity of the plant. Then , YPC made a prediction for
grade 5000S on the basis of the existing capacity of plant and YPC extended the
model for other grades in the family.

Process Description
A simplified process flowsheet is shown in figure 1.
The raw material feed consists of a mixture of ethylene(monomer),propylene(comonomer)
and hydrogen(chain transfer agent) gases. The mixture is bubbled into the bottom of the
well mixed polymerization reactor, which is filled with hexane diluent. Fresh hexane is
fed to the reactor with the Ziegler-Natta catalyst(TiCl4) and cocatalyst. M other liquid
from the centrifuge and condensed hexane from the reactor vapor are also recycled to the
reactor. In the reactor ,ethylene/propylene mixture is polymerized in the adequately
agitated hexane phase to form polymer. As the polymer is insoluble in hexane , a slurry is
formed. The polymerization reactions are highly exothermic, and the heat of
polymerization is removed via sensible heat to heat up the feeds, a water cooled jacket ,
and vaporization of hexane. About 70% of the heat of reaction is removed by the latent
heat of vaporization of hexane.

The vapor from the reactor is partially condensed by the water cooled condenser and
flashed in the drum. The liquid from the drum is pumped back and mixed with part of
mother liquid from the centrifuge and transferred to the reactor. The vapor from the drum
pressurized by the fan is split. The first stream is mixed with the gaseous feed to the
reactor and the second stream is mixed with the vapor going to the condenser.

The product from the reactor is a slurry containing the polymer, the solvent and
remaining monomer, comonomer and catalyst.

Process Modeling

Component characterization
The first step in building a model is to define all components. Polymerization
processes involve many different types of chemical components. In addition to the
standard components in ASPEN PLUS, such as hydrogen and ethylene, there are catalyst
components, polymer components ,and wax or oligomer(low molecular weight
polymers)components.
 A list of components used the YPC HDPE model is listed in Table 1.
Table 1. List of components used in the M odel
Component ID Type Component Name
H2 CONV hydrogen
C1 CONV methane
C2 CONV ethane
C2E CONV ethylene
C3E CONV propylene
C4E CONV 1-butene
C6 CONV n-hexane
N2 CONV nitrogen
wax oligomer low molecular weight
polymer
HDPE polymer High density polyethylene
S-C2H4 segment Ethylene-R
S-C3H6 segment 1-butene-R
CAT CONV TiCl4
CCAT CONV Triethyl-aluminum

A polymer component represents a distribution of many chemical species having

different molecular weights or chain length, copolymerization compositions, etc. The
distribution of polymer chain lengths can change with time and location in the flowsheet
during the course of the polymerization. Using POLYM ERS PLUS, the polymer
molecules are characterized by the segment approach. The polymer molecules are made
of structural building blocks, or repeat units, that are referred to as segments. Further, a
set of attributes are associated with the polymer component to track and convey
information about the average molecular properties of the polymer component present in
any stream or block in the flowsheet. The segment and the attribute information is used in
calculating thermo-physical properties and reaction rates in the kinetic models. Segments
are used to characterize the molecular architecture of oligomers and polymers. Since
segments have fixed molecular structures, POLYM ERS PLUS can carry out conservation
equations to track changes in the polymer component and its attributes.

Some physical properties, such as critical temperature, critical pressure, etc. ,of the
polymer are drawn from the polymer databank POLYPCD. However, all the properties
of interest in the simulation , such as density, heat capacity, etc., are calculated during the
simulation from property parameters for the segments and using current attribute
information such as average chain length and copolymer composition for the polymer.

Also, the catalyst attributes are used by the Ziegler-Natta kinetic model to keep track of
catalyst activity in terms of the fraction of catalyst sites that are not yet active, active,
dead, and inhibited at any location in the flowsheet.

Physical Properties and vapor liquid equibrium

All unit operations in the flowsheet, except the reactor, use the CHAO-SEADER
property set to calculate pure component and mixture properties such as density, enthalpy,
fugacity coefficient and free energy. For the reactor ,the polymer NRTL model is used.
The POLYNRTL physical property option set is used to calculate polymer and solvent
activity coefficient and related physical properties such as fugacity coefficient, enthalpy,
and free energy.

The reason for choosing the CHAO-SEADER property option set for all unit operation
except the reactor is because the CHAO-SEADER model , generally recommended for
low and moderate pressure hydrocarbon systems, represents the system quite well.

The YPC process is a slurry process. The polymer, insoluble in the hexane solvent, forms
a slurry(solid-liquid mixture). The volatile components partition between the slurry and
the vapor phase. Using the RCSTR model for the reactor, one needs to model the slurry
as a homogeneous liquid phase. We chose POLYNRTL for modeling phase equilibria.

Polymerization Kinetics
The polymerization kinetics are modeled using the Ziegler-Natta kinetic scheme in
POLYM ERS PLUS. A comprehensive set of reactions that occur in catalyzed olefin
polymerization have been built into this kinetic scheme. It supports multi-site
copolymerization of any number of monomers. This scheme consists of the basic reaction
steps: chain initiation, propagation and chain transfer(spontaneous, as well as to small
molecules such as monomer and hydrogen). In addition to polymerization reactions there
are reactions affecting the distribution of catalyst sites on which the polymerization
reactions take place. These include catalyst site activation and site deactivation.
Table 2 shows the YPC reaction kinetics scheme.
Table 2 The reaction scheme of HDPE slurry process of YPC
Reaction Type Site ID Reaction Species Description
ACT-COCAT 1-n CAT CCAT Activation of catalyst
CHAIN-INI 1-n C2E Chain initiation with ethylene
PROPAGATION 1-n C2E C2E Propagation
1-n C2E C3E
1-n C3E C2E
1-n C3E C3E
CHAT-MON 1-n C2E C2E Chain transfer to C2E monomer
CHAT-H2 1-n C2E H2 Chain transfer to hydrogen
DEACT-SPON 1-n Spontaneous deactivation of catalyst

In the above table, n represents the total number of site types that adequately characterize
the observed polymer molecular weight distribution. The number of site types that are
necessary to adequately fit the polymer molecular weight or chain length distribution is
determined by deconvolution of a GPC(Gel Permeation Chromatography) curve into a
number of individual Flory distributions representing the polymer produced at each site
type.

GPC Deconvolution
The 5000s grade GPC data were used to perform the deconvolution of the composite
molecular weight distribution into molecular weight distributions for each site type on the
catalyst. The rate constants for each site type were fit to the results of the deconvolution.

Shown in figure 2 is the GPC curve for polymer from the reactor. The curve was
deconvoluted to determine the number of site types , their location and area under the
curve needed to fit the GPC data. The fit does become better as the number of site types
increases. However, given the experimental errors associated with GPC measurements,
and the need to keep the number of equations to a minimum for on-line models, it was
decided that using 6 site types provided a reasonable match to the data.

Fig. 2 GPC deconvolution results using six sites for grade 5000S, YPC plant

Table 3 gives the weight fraction of polymer produced by the different catalyst site types.
The weight fraction of polymer produced by a site type is proportional to the area under
that site’s GPC curve relative to the area under the composite polymer GPC curve. The
third column of this table gives tha average chain length of the polymer produced by the
different catalyst site types.
Table 3 results for the six site GPC deconvolution for grade 5000s
Site number Weight fraction DPN
1 0.091 494
2 0.207 174000
3 0.083 681000
4 0.181 13200
5 0.110 2630
6 0.329 52700

Simulation of the reactor

Because most of heat from polymerization has been removed by the latent heat of
solvent and we assume the mixture in the reactor is mixed well. the simulation of the
reactor has been built by the combination of RCSTR and FLASH2 by POLYM ERS
PLUS and ASPEN PLUS.
A single site kinetic model was first developed and the rate constants for the propagation
and catalyst site reactions were fitted to match the overall conversion of monomer to
polymer in the reactor. The chain transfer and cross propagation reaction rate constants
were adjusted to match the number average molecular weight and copolymer
composition. The single site model is used as the starting point for fitting multi-site
kinetics.
Par of input file shows as following:

BLOCK D201 RCSTR

PARAM VOL=90<CUM> TEMP=85 PRES=590 NPHASE=2 PHASE=1
MB-TOL=0.00001 PH-RES-TIME=2.5 PH-VOL-EST=38<CUM>
PROPERTIES POLYNRTL
BLOCK-OPTION TERM-LEVEL=7
CONVERGENCE SOLER=NEWTON
REACTIONS RXN-IDS=Z-NATTA

REACTIONS Z-NATTA ZIEGLER-NAT

DESCRIPTION “ ZIEGLAR-NATTA KINETIC SCHEME”
SPECIES CATALYST=CAT COCATALYST=CCAT MONOMER=C2E
HYDROGEN=H2 POLYMER=HDPE
ACT-COCAT 1 CAT CCAT 9.09E6 8000 1
CHAIN-INI 1 C2E 4.520E8 7000 1
PROPAGATION 1 C2E C2E 4.520E8 7000 1
CHAT-MON 1 C2E C2E 1.5E3 7200 1
CHAT-H2 1 C2E H2 7.2E6 8000 0.5
DEACT-SPON 1 1.6E-4 1000 1

Polymer property prediction

HDPE quality indices such as melt index(M I) and density have been correlated to
polymer molecular weight and copolymer compositions. A user-property subroutine was
developed to predict M I and density according to empirical correlations. The parameters
have been updated with propylene as the comonomer.

Selected results and comparison

The model was first built on the basis of the design capacity of the plant. Table 4 shows
the comparison of simulated results with the design values about the main components
in the product stream. M ass flowrates under the operation conditions match well. It
proves the reasonability of the model.
 Table 4 comparison of simulated results with the design values (unit: kg/h)
Mass H2 C2H4 C3H6 C6H14 WAX HDPE
flowrate
Calculated 660.8 14167 865.7 41744 176.2 4402.8
value s
Design 660.2 14224 879.5 42657 173.7 4387.8
values

Table 5 shows the comparison of simulated results with the measured values for some
samples . The measured values are within the required range of properties of polymer.
Also, It proves the reasonability of the model.
Table 5 comparison of simulated results with the measured values
properties M WW M WN PDI MI DENSITY
g/10min g/cm3

Calculated values 225604 10672.3 21.139 1.1 0.950

Measured values 218400 10100 22 1.0 0.954

Then, The model has been extended to the existing operating capacity and from 5000s
grade to other grades in the family to meet the need of plant further. The model fits the
other grades in the family whether under the design capacity or the existing capacity of
the plant..

Table 6 shows the productivity of grade 2200J under the existing operating capacity of
the plant. The grade belongs to the same family with that of grade 5000s.

Table 6 the productivity of grade 2200J under the existing operating capacity
Items Fresh ethylene HDPE MI DENSITY
feedrates
Unit Kg/h Kg/h g/10min g/cm3
Calculated values 5000 4892.5 5.5 0.964
Measured values 5000 4850 4.5~6.5 0.962~0.966

Sensitivity analysis
To meet the need of the production, we provided some valuable information by
sensitivity analysis. We have obtained the effect on the conversion of ethylene ,number
average molecular weight, weight average molecular weight, polydispersity index and
melt index by varying the total feedrate, the feedrate of ethylene, the concentration of
hydrogen in the feed . Fig.3 shows the effect on melt index by varying the hydrogen
massfraction in the feed. It is a part of the results from sensitivity analysis. The result
proves the melt index decreases by adding the concentration of hydrogen in the feed. And
it provided the range which we can use to reduce the cost of hydrogen when the qualities
of the polymer are required.Fig.3 Effect on M I by varying hydrogen mass fraction in
the feed.

Conclusion

A kinetic model has been successfully built for the high density polyethylene slurry
process by POLYM ERS PLUS. The results and comparisons show the reasonability of
the model. And we have been making full use of it for production , design .

Acknowledgements
I would like to thank my colleagues at Yangzi Petrochemical Company in P.R.of China
and for ASPENTECH for their support and assistance.
A Kinetic Model for High Density
Polyethylene Slurry Process

Xiaoyan Xu
Yangzi Petrochemical Corporation
P. R. of China
Feb.8, 2000
Introduction

• Yangzi Petrochemical Corporation(YPC) is a major

producer of high density polyethylene (HDPE)
• The scope of the kinetic model for HDPE slurry process
covers the polymerization reactor, the heat removal system,
the monomer-solvent separation and the recycle system for
grade 5000S.
• It has been a joint effort of YPC and AspenTech .
YPC HDPE slurry process

• Unique characteristics
- slurry reactor
- high productivity catalyst
-mild operation conditions
-75% heat of polymerization is removed by latent heat of
vaporization of solvent.
• Common characteristics
-solvent removal
-solvent recovery
What new capabilities were needed?

• Polymer phase equilibria

• polymer properties
• reactor model
• kinetic model
• model validation
• model prediction
How did we get these capabilities?

• Aspen+ provides framework for :

- SS process model
-conventional components
• POLYMERS PLUS adds the framework for:
- polymer characterization
- polymer physical properties
-user defined polymer properties
-polymerization kinetics
Process Modeling

• Component characterization
- The first step in building a model is to define all
components.
- Polymerization processes involve many different types
of chemical components.
- In addition to the standard components such as hydrogen
and ethylene, there are catalyst components, polymer
components ,and wax or oligomer(low molecular weight
polymers)components.
Table 1. List of components used in the Model
Component ID Type Component Name
H2 CONV hydrogen
C1 CONV methane
C2 CONV ethane
C2E CONV ethylene
C3E CONV propylene
C4E CONV 1-butene
C6 CONV n-hexane
N2 CONV nitrogen
wax oligomer low molecular
weight polymer
HDPE polymer High density
polyethylene
S-C2H4 segment Ethylene-R
S-C3H6 segment 1-butene-R
CAT CONV TiCl4
CCAT CONV Triethyl-
aluminum
•Polymer component

-The distribution of polymer chain lengths can change with

time and location in the flowsheet during the course of the
polymerization.
-Using POLYMERS PLUS, the polymer molecules are
characterized by the segment approach.
- A set of attributes are associated with the polymer
component and the segment are used in calculating
thermo-physical properties and reaction rates in the kinetic
models.
•polymer and catalyst component

-Some physical properties, such as critical temperature are

drawn from the polymer databank POLYPCD.
- Properties such as density, heat capacity, etc., are
calculated from property parameters for the segments and
using current attribute information such as average chain
length and copolymer composition for the polymer.
- the catalyst attributes are used by the Ziegler-Natta kinetic
model to keep track of catalyst activity.
•Physical Properties and vapor liquid equibrium

-All unit operations in the flowsheet, except the reactor, use

the CHAO-SEADER property set to calculate properties
such as density, enthalpy, fugacity coefficient.
-For the reactor ,the polymer NRTL model is used.
-The POLYNRTL physical property option set is used to
calculate polymer and solvent activity coefficient and
related physical properties such as fugacity coefficient,
enthalpy, and free energy.
•Physical Properties and vapor liquid equibrium

-The reason for choosing the CHAO-SEADER property

option set is because the CHAO-SEADER model ,
generally recommended for low and moderate pressure
hydrocarbon systems, represents the system quite well.
-The YPC process is a slurry process. The polymer,
insoluble in the hexane solvent, forms a slurry(solid-liquid
mixture). The volatile components partition between the
slurry and the vapor phase. The slurry is regarded as a
homogeneous liquid phase.
•Polymerization Kinetics

- The polymerization kinetics are modeled using the

Ziegler-Natta kinetic scheme in POLYMERS PLUS.

-This scheme consists of the basic reaction steps: chain

initiation, propagation and chain transfer(spontaneous, as
well as to small molecules such as monomer and

hydrogen) ,catalyst site activation and site deactivation.

 Table 2 The reaction scheme of HDPE slurry process
Reaction Type Site Reaction Description
ID Species
ACT-COCAT 1-n CAT CCAT Activation of catalyst
CHAIN-INI 1-n C2E Chain initiation with ethylene
PROPAGATION 1-n C2E C2E Propagation
1-n C2E C3E
1-n C3E C2E
1-n C3E C3E
CHAT-MON 1-n C2E C2E Chain transfer to C2E
monomer
CHAT-H2 1-n C2E Chain transfer to hydrogen
H2
DEACT-SPON 1-n Spontaneous deactivation of
catalyst
•GPC Deconvolution

- In the above table, n represents the total number of site

types that adequately characterize the observed polymer
molecular weight distribution.
- The number of site types that are necessary to adequately
fit the polymer molecular weight or chain length
distribution is determined by deconvolution of a GPC(Gel
Permeation Chromatography) curve into a number of
individual Flory distributions representing the polymer
produced at each site type.
Fig2. GPC deconvolution results for grade 5000S, YPC plant
Table 3 results for the six site GPC deconvolution for grade 5000s
Site number Weight fraction DPN
1 0.091 494
2 0.207 174000
3 0.083 681000
4 0.181 13200
5 0.110 2630
6 0.329 52700
•Simulation of the reactor

- The simulation of the reactor has been built by the

combination of RCSTR and FLASH2 by POLYMERS
PLUS and ASPEN PLUS.
- The single site model is used as the starting point for fitting
multi-site kinetics.
Par of input file:

BLOCK D201 RCSTR

PARAM VOL=90<CUM> TEMP=85 PRES=590 NPHASE=2 PHASE=1
MB-TOL=0.00001 PH-RES-TIME=2.5 PH-VOL-EST=38<CUM>
PROPERTIES POLYNRTL
BLOCK-OPTION TERM-LEVEL=7
CONVERGENCE SOLER=NEWTON
REACTIONS RXN-IDS=Z-NATTA

REACTIONS Z-NATTA ZIEGLER-NAT

DESCRIPTION “ZIEGLAR-NATTA KINETIC SCHEME”
SPECIES CATALYST=CAT COCATALYST=CCAT
MONOMER=C2E HYDROGEN=H2 POLYMER=HDPE
ACT-COCAT 1 CAT CCAT 9.09E6 8000 1
CHAIN-INI 1 C2E 4.520E8 7000 1
PROPAGATION 1 C2E C2E 4.520E8 7000 1
CHAT-MON 1 C2E C2E 1.5E3 7200 1
CHAT-H2 1 C2E H2 7.2E6 8000 0.5
DEACT-SPON 1 1.6E-4 1000 1
•Simulation of the reactor

- The model was first built on the basis of the design

capacity of the plant.
-The model has been extended to the existing operating
capacity.
- The model has been extended from 5000s grade to other
grades in the family to meet the need of plant further.
- sensitivity analysis
•Polymer property prediction

-HDPE quality indices such as melt index(MI) and density

have been correlated to polymer molecular weight and
copolymer compositions.
- A user-property subroutine was developed to predict MI
and density according to empirical correlations.
- The parameters have been updated with propylene as the
comonomer.
Selected results and comparison

• Table 4 shows the comparison of simulated results with the design

values about the main components in the product stream. Mass
flowrates under the operation conditions match well. It proves the
reasonability of the model.
• Table 5 shows the comparison of simulated results with the
measurement values for some samples . The measured values are
within the required range of properties of polymer. Also, It proves
the reasonability of the model.
• Table 6 shows the productivity of grade 2200J under the existing
capacity of the plant.It belongs to the same family as that of 5000S.
Table 4 comparison of simulated results with the design
values (unit: kg/h)
Mass H2 C2H4 C3H6 C6H14 WAX HDPE

flowrate

Calculated 660.8 14167 865.7 41744 176.2 4402.8

values

Design 660.2 14224 879.5 42657 173.7 4387.8

values
 Table 5 the comparison of simulated results with the

measurement values for some samples

properties MWW MWN PDI MI DENSITY

g/10min g/cm3

Calculated 225604 10672.3 21.139 1.1 0.950

values

Measured 218400 10100 22 1.0 0.954

values
 Table 6 the productivity of grade 2200J under the existing

operating capacity of the plant.

Items Fresh HDPE MI DENSITY

ethylene

feedrates

Unit Kg/h Kg/h g/10min g/cm3

Calculated 5000 4892.5 5.5 0.964

values

Measured 5000 4850 4.5~6.5 0.962~0.966

values
•Sensitivity analysis

• To meet the need of the production, we provided some

valuable information by sensitivity analysis.
• We have obtained the effect on the conversion of
ethylene, ,number average molecular weight, weight
average molecular weight, polydispersity index and melt
index by varying the total feedrate, the feedrate of ethylene,
the concentration of hydrogen in the feed .
Sensitivity analysis

• Fig.3 shows the effect on melt index by varying the

hydrogen massfraction in the feed. It is a part of the results
from sensitivity analysis.
• The result proves the melt index decreases by adding the
concentration of hydrogen in the feed.
• It provided the range which we can use to reduce the cost
of hydrogen when the qualities of the polymer are required.
Fig3. The effect on MI by varying hydrogen
massfraction
Conclusion

A kinetic model has been successfully built for the high

density polyethylene slurry process by POLYMERS PLUS.

The results and comparisons show the reasonability of the
model. And we have been making full use of it for
production and design etc.
Acknowledgements

I would like to thank my colleagues at Yangzi

Petrochemical Company in P.R.of China and for

ASPENTECH for their support and assistance.