You are on page 1of 11

Journal of Elastomers and Plastics

http://jep.sagepub.com/

Dynamic simulation of high-pressure tubular reactor for low-density


polyethylene production
Franjo Jovic, Marko Grasovec, Igor Dejanovic and Vanja Kosar
Journal of Elastomers and Plastics published online 4 June 2013
DOI: 10.1177/0095244313489904

The online version of this article can be found at:


http://jep.sagepub.com/content/early/2013/06/02/0095244313489904
A more recent version of this article was published on - Nov 16, 2014

Published by:

http://www.sagepublications.com

Additional services and information for Journal of Elastomers and Plastics can be found at:

Email Alerts: http://jep.sagepub.com/cgi/alerts

Subscriptions: http://jep.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Version of Record - Nov 16, 2014


>> OnlineFirst Version of Record - Jun 4, 2013

What is This?

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


Article
Journal of Elastomers & Plastics
1–10
Dynamic simulation of ª The Author(s) 2013
Reprints and permissions:

high-pressure tubular sagepub.co.uk/journalsPermissions.nav


DOI: 10.1177/0095244313489904

reactor for low-density jep.sagepub.com

polyethylene
production
Franjo Jović, Marko Grašovec,
Igor Dejanović and Vanja Kosar

Abstract
Low-density polyethylene, one of the most important polymer products, is commonly pro-
duced in high-pressure free radical polymerization processes. A dynamic model of the
high-pressure polymerization of ethylene initiated by oxygen in tubular reactor is intro-
duced, and a dynamic optimization problem is formulated for process start-up strategies.
The present study proposes a kinetic model based on an assumed reaction mechanism.
The model describes the rates of oxygen decomposition and propagation of free radical
ethylene polymerization. The mass and heat balance equations in an adiabatic tubular reac-
tor operated at a constant pressure of 2.4 kbars and a temperature range of 110–300 C
are presented. Simulations of polymerization process predict temperature of the reaction
mixture, response time for cooling water, and also conversion along the reactor length.
Response time was obtained using different inputs of controlled variables. Values obtained
from these simulations are compared with real data from the process unit (Polietilen,
Dioki1, Zagreb, Croatia) and a model validation is confirmed. Improvement in reactor pro-
ductivity and better understanding of few different start-up procedures is achieved.

Keywords
Low-density polyethylene, dynamic simulation, mathematical modeling, tubular polymer-
ization reactor

Department of Reaction Engineering and Catalysis, Faculty of Chemical Engineering and Technology, University
of Zagreb, Zagreb, Croatia

Corresponding author:
Vanja Kosar, Department of Reaction Engineering and Catalysis, Faculty of Chemical Engineering and Technology,
University of Zagreb, Savska c.16., Zagreb 10000, Croatia.
Email: vkosar@fkit.hr

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


2 Journal of Elastomers & Plastics

Introduction
Polyethylene (PE) is a thermoplastic polymer broadly used in consumer products (over 60
million tons are produced worldwide every year). PE is classified into several different
categories based mostly on its density and branching. One of the common PE products is
low-density PE (LDPE). LDPE is commercially manufactured in high-pressure processes
and represent one of the largest base products on the polymer market in general (tons per
year production). The technique requires a purified ethylene in the feed and oxygen or/and
peroxide as a catalyst. The high-pressure process is usually a bulk polymerization initiated
by oxygen in a very long continuous tubular reactor (length/diameter ratio is approximately
17,000). The process is usually operated at elevated pressures, for example, 2400 bars, and
temperature range from 90 to 110 C.1 As the process is carried out under rigorous operating
conditions, consequently a mathematical modeling is an attractive tool to study safely and
economically the influence of the different operational conditions on production perfor-
mance and product quality. Over the last few decades, a long list of scientists tried to describe
ethylene polymerization in high-pressure tubular reactors. Several authors have presented
different models dealing with the stationary state of this process, for example, molecular
weight distribution in LDPE polymerization,2–4; Baltsas et al.5 used the pseudokinetic rate
method to simulate high-pressure LDPE tubular reactors. However, less attention has been
paid to the dynamic behavior of this reactor, even though a dynamic model may be used to
study different start-up and shut down procedures and particularly in process optimization.
In those articles, different mathematical models and softwares have been used. For example,
Asteasuain et al.6 used the gOPT of the gPROMS simulator to optimize LDPE polymeriza-
tion processes; Bezzo et al.7 used Fluent and gPROMS tools for computational fluid
dynamics and process simulation of a batch reactor; Haefele et al.8 used dynamic process
simulator DIVA to make a detailed dynamic mathematical model of a LDPE plant; and Cer-
vantes et al.1,9 used a dynamical model for large-scale optimization of a PE plant. In this
work, a dynamic model of the high-pressure tubular process is derived and simulation stud-
ies using that model are presented. The model consisted of two main parts: kinetic and reac-
tor model. Moreover, a comparison of different start-up strategies and an investigation of
responses to disturbances from the nominal operating point are included. Finally, the results
are discussed, and a brief outlook on the results is given.

Process description
Tubular reactor for high-pressure ethylene polymerization is a typical industrial reactor,
which consists of a very long pipe (800 m) with cooling jackets (six zones; Figure 1). The
reaction mixture (ethylene monomer; initiator: oxygen; and a chain transfer agent: telo-
gen) flows through the inner pipe. The reactor model was divided into six zones (differ-
ent zone starts with a change in the level of jacket temperature).
The physical properties of each zone include density, global heat transfer coefficient
(in the first two zones, they are length dependent due to the fast reaction), and specific
heat of reaction mixture, all these properties regarded as constant in each zone. The
velocity of the reaction mixture is also considered constant in each zone (pressure drop is

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


Jović et al. 3

Figure 1. Tubular reactor for high-pressure ethylene polymerization.

neglected) and its value is obtained using the same procedure. The fluid in the jacket
enters counter flow with respect to reaction mixture (except in first zone where the flow
is cocurrent). In the first zone, reaction mixture is heated with steam, and in all the other
zones, the reaction mixture is cooled with water. The profile of jacket temperature in
each zone is calculated separately. Table 1 shows the scheme of the tubular reactor with
characteristic parameters.

Mathematical model of the process


A model of the reactor was designed10 and the following assumptions are considered:

 one-dimensional plug flow through reactor due to the small pipe radius and large
flow velocity;
 negligible pressure drop along the reactor; Dp ¼ 0;
 blow down of the reactor is neglected;
 negligible heat transfer from reactor jacket to atmosphere;
 polymer deposition on inner reactor wall is assumed and experimental data for
heat transfer coefficients are used;
 density and specific heat capacity are constant;
 free radical mechanism of polymerization is proposed;
 empirical model is used, where amount of reacted ethylene is proportional with
the square root of oxygen concentration and ethylene concentration on 3/2

Kinetic model
The proposed reaction mechanism is composed of several basic reactions, which are
considered in the kinetic model. Polymerization occurs at 165 C with molecular oxygen
(initiator) decomposition, which generates initial radical couple reaction initiation as
shown in the following equation6
kOx
O2 þ M  ! 2RMn
165 C
ð1Þ

Free radicals react with ethylene monomer molecules when reaction propagation
begins and molecules starts to grow as shown in the following equation

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


4 Journal of Elastomers & Plastics

Table 1. Design features and steady state operating conditions for the base case.

Length/diameter ratio (L/D) 17,000


Internal diameter (D) 0.047 m
Number of zones (Nz) 6
Zone axial length/diameter ratio (Lz,i/D) 1250, 1900, 4200, 4200, 4200, and 1250
Inlet temperature 1 (feed; T0,1) 90 C
Inlet temperature 2 (side; T0,2) 110 C
Inlet pressure (p) 2400 bar
Density of reacting mixture (r) 530 kg/m3
Monomer flow rate (Q) 30,000 kg/h
Oxygen flow rate (initiator) 1 0.2 kg/h1
Oxygen flow rate (initiator) 2 0.3 kg/h1
Conversion 21–25%
Specific heat of reacting mixture (cp,i) 3362 kJ/kg1
Global heat transfer coefficient (Uz,i) 4186, 2093, 420, 1340, 1130, 1130 J/m2/C/s
Cross sectional area of cooling zone (Av) 0.0061 m2

ket
2RMn þ M DH
! RMnþ1 ð2Þ

The polymerization reaction is highly exothermic; the growth of polymer chain


releases 803 kcal/kg of reaction heat. There are two proposed termination reactions of
polymerization. One is termination by combination as shown in equation (3), and the
other is termination by thermal degradation as shown in equation (4).
ktc
RMn þ RMn ! RM2n R ð3Þ

ktc
RMnþ1 ! RMn þ R ð4Þ
A chain transfer agent (telogen) is commonly fed at the reactor inlet in order to control
the molecular weight. Telogen (S) reacts with radicals producing a dead polymer chain
and an initiation radical as shown in equation (5).
ktrs
RMn þ S ! RMn H þ S  ð5Þ
Other reactions such as transfer to polymer and intermolecular scissions are also
important in this polymerization. They affect long- and short-chain branching and
weight-average molecular weights. Nevertheless, we focused only on conversion, and
so these side reactions were not included in the model.
The kinetic model derived on the basis of the above-mentioned mechanism is given
by the following equations
rOx ¼ KOx Cet COx ð6Þ

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


Jović et al. 5

1=2 3=2
rEt ¼ kEt COx CEt ð7Þ
Constants of kinetic model are given using equations (8) and (9), where p1, p2, p3, p4,
and p5 are observed constants obtained by experiment.
 p p2  p3 
kEt ¼ exp p1  ð8Þ
1000 T
 p5 
kOx ¼ exp p4  ð9Þ
T

Reactor model
The whole reactor is divided on n-segments with ‘‘j’’ as the step of integration. Mass balance for
oxygen (initiator) and ethylene for each segment ‘‘j’’ are given using equations (10) and (11).
 
qCOx Q qCOx
¼ þ rOx ð10Þ
qt A qx
 
qCEt Q qCEt
¼ þ rEt ð11Þ
qt A qx
Energy balance for reaction mixture is shown in equation (12) and water in jacket of
each zone ‘‘i’’ is shown in equation (13), which gives the temperature–time change on
which exothermic heat generated by the reaction is influenced, heat taken away with water
in jacket of each zone, and heat accumulation in bulk of mass for the each segment ‘‘j’’.
  
qT Q qT 4U T  Tv;i ðDH Þ
¼ þ  rEt ð12Þ
qt A qx Dcp  cp 

  
qTv;i Q qT 4U Tv;i  T
¼ þ ð13Þ
qt A qx Dcpv 
The energy accumulated in a 5-cm thick reactor wall is in consequence with differ-
ence of heat flow of inner and outer surface of the reactor as shown in equation (14)
qTs 4U ðT  Ts Þ 4U ðTs  T Þ
¼  ð14Þ
qt Du cps s Dv cps s
Steady state model describes the system in equilibrium (stationary) state, which is
shown in equation (15)
dC dT
¼ 0; ¼0 ð15Þ
dt dt
Equations (10) and (11) become equations (16) and (17):

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


6 Journal of Elastomers & Plastics

dCOx A
¼ rOx ð16Þ
dx Q

dCEt A
¼ rEt ð17Þ
dx Q
Temperature profiles under steady state of reaction mixture and cooling water are
given by equations (18) and (19), respectively
dT 4U ðTv  T Þ A ðDH Þ A
¼  rET ð18Þ
dx Dcp Q Cp Q

dTv 4U ðTv  T Þ A
¼ ð19Þ
dx Dcpv Q
These equations are first order differential equations, and they can be solved using
different, common computing methods, for example, Runge–Kutta IV method. Data
obtained from steady state model can be used as initial states for dynamic simulation
(Figure 2).
Temperature profiles of cooling water in all six zones and reaction mixture are solved
simultaneously (Figure 2).

Dynamic simulation
Dynamic reactor model is presented by corresponding mass as shown in equations (10)
and (11) and energy balances as shown in equations (12) and (13). These balances are
partial differential equations (PDE) and must be solved by numerical methods.
The initial conditions are shown in equation (20)
at t ¼ 0; CEt ðt ¼ 0; x ¼ jÞ ¼ CEt;0
COx ðt ¼ 0; x ¼ jÞ ¼ COx;0 ð20Þ
T ðt ¼ 0; x ¼ jÞ ¼ T0;1
Boundary conditions are defined as shown in equation (21)
at z ¼ 0; CEt ðt; x ¼ 0Þ ¼ CEt ðtÞ
COx ðt; x ¼ 0Þ ¼ COx;0
T ðt; x ¼ 0Þ ¼ T0;1 ð21Þ
at z ¼ 310 m; COx ðt; x ¼ 310Þ ¼ COx;310
T ðt; x ¼ 310Þ ¼ T0;2
Simple explicit method was chosen, and the equations were discretized along time
and axial distance coordinate using finite differences. In this method, partial derivations
are approximated with finite differences as shown in equation (20). However, this
method has truncation error of ðOðDxÞÞ.

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


Jović et al. 7

Figure 2. Profiles of temperatures and polyethylene mass flow against the reactor distance: solid
line indicates polyethylene mass flow; dash indicates cooling water temperature; and dash–dot
indicates reaction mixture temperature.
 
CEtðiþ1;jÞ  CEtði;jÞ Q CEtði;jÞ  CEtði;j1Þ
¼ þ rEt ð22Þ
Dt A Dx
or
 
Q Dt 
CEtðiþ1;jÞ ¼ CEtði;jÞ  CEtði;j1Þ þ rEt ð23Þ
A Dx
 
Q Dt
 ¼K ð24Þ
A Dx
Equation (23) gives the values of CEt, at time t þ Dt, at the grid point with spatial
coordinate
x ¼ jDx ¼ K
In equation (24), K is known as Courant number. If K ¼ 1, then equation (23) can be
written as
CEtðiþ1;jÞ ¼ CEtði;j1Þ þ rEt ð25Þ
Using this method, the PDEs were converted into algebraic equations and were solved
iteratively. Same procedure is made for the other PDEs for mass and energy balances.
Temperature profile during time and along axial distance of reactor is shown in
Figures 3 and 4 for model and measured data, respectively.
Real data are obtained from Process Unit (Polietilen, Dioki1, Zagreb, Croatia.)
Temperature sensors are placed at distinct lengths of the reactor. The reactor consists of

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


8 Journal of Elastomers & Plastics

Figure 3. Temperature profiles against time and reactor length obtained by model.

Figure 4. Start-up of the process.

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


Jović et al. 9

Figure 5. Oxygen percentage versus reaction time and reactor length.

10 m long connected reactor pipe segments. Comparison with real data (Figure 4)
showed good agreement with mathematical model data and validation of a model.
Side feed of ethylene monomer and oxygen (on 310 m of reactor) is necessary to
maintain the reactor efficiency, while reaction temperature is kept below 580 K at which
decomposition of PE occurred. Exothermal reaction can increase temperature of reaction
mixture above 580 K and can influence reactor shut down. Prediction of temperature
profiles of reactor and control of cooling water can avoid this. The main goal of this
model is to predict and avoid repeated reactor shut down. From the operational point of
view, it is important to reach maximal conversion in a short period of time. Also, during
the start-up procedure, the temperature at the reactor exit must be under 250 C to avoid
reaction movement in separator. This could induce separator decomposition and longer
period of plant shut down.
Oxygen concentration profile shows rapid decay (Figure 5). Further improvement in
avoiding this rapid decay could be by changing oxygen feed and/or include hydrogen
peroxide as an initiator. This change must be considered and also include other factors
such as reactor modification costs.

Conclusions
In this article, a dynamic model for an LDPE production plant was presented. The reactor
model is able to predict appropriately the actual dynamic state reactor behavior for a

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014


10 Journal of Elastomers & Plastics

wider range of operating conditions, especially with respect to the steady state. Con-
sequently, it improves the model predictive capabilities for key plant variables such as
conversion and temperature profiles.
Based on a dynamic model, optimal feed profiles of oxygen and ethylene are deter-
mined to minimize the transient states generated during the switching between different
steady states. Careful attention is given to start-up and less to shut down procedures. Due to
the very high pressure, safety of whole polymerization process is important during plant
start-up. Dynamic simulation can predict possible hazardous situations and avoid possible
accidents.

Funding
This research received no specific grant from any funding agency in the public, commercial, or
not-for-profit sectors.

References
1. Cervantes AM, Tonelli S, Brandolin A, Bandoni JA and Biegler LT. Large-scale dynamic
optimization for grade transitions in a low density polyethylene plant. Comput Chem Eng
2002; 26: 227–237.
2. Kim D-M and Iedema PD. Molecular weight distribution in low-density polyethylene poly-
merization; impact of scission mechanisms in the case of a tubular reactor. Chem Eng Sci
2004; 59: 2039–2052.
3. Verros GD. Calculation of molecular weight distribution in non-linear free radical copolymer-
ization. Polymer 2003; 44: 7021–7032.
4. Pladis P and Kiparissides C. A comprehensive model for the calculation of molecular
weight + long-chain branching distribution in free-radical polymerizations. Chem Eng Sci
1998; 53(18): 3315–3333.
5. Baltsas A, Papadopoulos E and Kiparissides C. Application and validation of the pseudo-kinetic
rate constant method to high pressure LDPE tubular reactors. Comput Chem Eng 1998; 22:
S95–S102.
6. Asteasuain M, Tonelli SM, Brandolin A and Bandoni JA. Dynamic simulation and optimi-
sation of tubular polymerisation reactors in gPROMS. Comput Chem Eng 2001; 25: 509–515.
7. Bezzo F, Macchietto S and Pantelides CC. A general framework for the integration of
computational fluid dynamics and process simulation. Comput Chem Eng 2000; 24: 653–658.
8. Haefele M, Kienle A, Boll M and Schmidt C-U. Modeling and analysis of a plant for the
production of low density polyethylene. Comput Chem Eng 2006; 31: 51–65.
9. Cervantes A, Tonelli S, Brandolin A, Bandoni A and Biegler L. Large-scale dynamic optimi-
zation of a low density polyethylene plant. Comput Chem Eng 2000; 24: 983–989.
10. Incropera FP, Dewitt DP, Bergman TL and Lavine AS. Fundamentals of heat and mass
transfer. New York: Jonh Wiley and Sons, 2007.

Downloaded from jep.sagepub.com at TEXAS SOUTHERN UNIVERSITY on December 7, 2014

You might also like