Você está na página 1de 10

322 Ind. Eng. Chem. Res.

2011, 50, 322331

Steady-State and Dynamic Modeling of the Basell Multireactor Olefin


Polymerization Process
Zu-Wei Zheng, De-Pan Shi, Pei-Lin Su, Zheng-Hong Luo,*, and Xiao-Jun Li
Department of Chemical and Biochemical Engineering, College of Chemistry and Chemical Engineering,
Xiamen UniVersity, Xiamen 361005, China and Polyolefin Department of Lanzhou Petrochemical Company
Limited of China National Petroleum Company, Lanzhou 730060, China

In this work, we developed the steady-state and dynamic models for the commercial polypropylene process
of Basell Spheripol technology, involving fundamental chemical engineering principles and advanced software
tools, i.e., Polymers Plus and Aspen Dynamics. The models considered the important issues of physical property
and thermodynamic model selection, catalyst characterization, and reactor model. Besides, a multisite catalyst
with traditional Ziegler-Natta polymerization kinetics was introduced to describe the broad molecular weight
distribution of the polymers produced in this polypropylene technology. Both the continuous stirred tank
reactor model and the combined plug flow reactors model were proposed to simulate the reactors. Furthermore,
we validated the models using industrial data and demonstrated application of the dynamic model to grade
change, start up, and shut down at a certain emergent accident.

1. Introduction equilibrium calculations in the process units.5 As a whole, there


are many methodologies proposed in the literature,1-27 which
1.1. Scope. The Basell Spheripol technology is one of the succeed in developing a comprehensive model for polyolefin
most widespread commercial methods to produce polypropylene. processes. These methodologies include model selection and
Commonly, its key parts are two loop reactors and one or two parameter tuning for physical and thermodynamic properties,
fluidized bed reactors (FBRs). Among them, the loop reactors catalyst characterization, and polymer properties in addition to
are used to produce homopolypropylene, ethylene-propylene the single and multiple site of traditional Ziegler-Natta po-
random copolymer, and ethylene-propylene-1-butene copoly- lymerization kinetics. However, most of the preceding works
mer, and the FBR is used to produce impact polypropylene.1,2 have focused on the polyethylene process and the gas-phase
It is well acknowledged that the loop reactor is one of the best polypropylene process. To the best of our knowledge, there is
choices for the preparation of homopolypropylene due to its no comprehensive model of the commercial bulk polypropylene
high mixing capability as well as high heat transfer/exchange process of the Spheripol technology in the literature prior to
rate.1 our work.
The objective of this work is to discuss the considerations In this work, both steady-state and dynamics models for the
and techniques needed to develop a comprehensive model for polypropylene process of Spheripol technology are developed
the polypropylene process of the Spheripol technology. Gener- using fundamental chemical engineering principles and advanced
ally, an excellent polymerization process model needs accur- software tools, i.e., Polymers Plus and Aspen Dynamics. We
ate description of physical and thermodynamic properties, phase focus not only on the accurate modeling of the reactors and
equilibrium, and polymerization kinetics, all of which were polymerization kinetics but also on the thermodynamic under-
discussed in our previous work.2 In addition, both steady-state standing and, moreover, discuss the limitation of thermodynam-
and dynamic models for the commercial bulk polypropylene ics and suggest the proper efforts needed for development of
process of HYPOL technology were developed in our previous the process model.
work.2 To date, there have also been many other models 1.2. Commercial Polypropylene Process of Spheripol
describing the olefin polymerization process.3-19 For instance, Technology. Spheripol technology was first industrialized by
Khare et al.3 presented a model for the gas-phase polypropylene Montedison in 1983. The second-generation technology was
process using a stirred-bed reactor, which was approximated developed by Montell in 1995, which is also called Basell
as a series of continuous stirred-tank reactors (CSTRs). Detailed technology at present.28,29 An application of this Spheripol
modeling methodologies, especially the reactor modeling, were technology in a chemical plant in China was studied here.
provided. Khare et al.4 also simulated slurry high-density Figure 1 shows a schematic flow chart for a typical polypro-
polyethylene processes. The nonequilibrium behavior in ethyl- pylene process of the Spheripol technology. The process includes
ene/polyethylene flash separators was detected in their work.6 bulk polymerization, degasification, gas-phase copolymerization,
Chatzidoukas et al.5 developed a dynamic model for the steaming, and drying. The prepolymerization is carried out under
commercial polypropylene process of Borstar technology, mild operating conditions to prevent particle overheating and
wherein a comprehensive kinetic model for the ethylene-1- generation of fine powder. The main polymerization then
butene copolymerization based on a two-site catalyst was proceeds in two liquid-phase tubular loop reactors. In the case
employed. The Sanchez-Lacombe equation of state (S-L-EoS) of production of impact copolymer, the polymerization reaction
was employed for the thermodynamic properties and the phase will be continued in FBR. The technology uses a titanium-based
catalyst and an aluminum-alkyl-based cocatalyst, which are
* To whom correspondence should be addressed. Tel.: +86-592-
2187190. Fax: +86-592-2187231. E-mail: luozh@xmu.edu.cn. titanium tetrachloride (TiCl4) and triethyl aluminum [Al(C2H5)3],

Xiamen University. respectively. Besides, hydrogen along with the propylene feed

China National Petroleum Company. is used as the molecular-weight control agent to produce various
10.1021/ie101699b 2011 American Chemical Society
Published on Web 12/01/2010
Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011 323

Figure 1. Simplified flow chart for a polypropylene process of the Spheripol technology.

Figure 2. Spheripol loop process (A, catalyst; B, propylene; C, hydrogen; D, product; E, coolant; 1, P200, pump; 2, P201, pump; 3, P202, pump; 4, R200,
prepolymerization reactor; 5, R201, main polymerization reactor; 6, R202, main polymerization reactor).

Table 1. Pure-Component Parameters for the PC-SAFT Equations


grades of polypropylene, and the tacticity control agents fed
along with the catalyst are commonly used to increase the no. component m (A) /kB (K) r (mol/g) refs 2, 3, 7
isotactic content of the polypropylene. 1 hydrogen 0.8285 2.973 12.53 2, 3
Figure 2 shows a simplified diagram of the Spheripol 2 ethylene 1.559 3.434 179.4 2, 3
technology using three liquid-phase tubular loop reactors for 3 ethane 1.607 3.521 191.4 2, 3, 7
production of homopolypropylene, ethylene-propylene random 4 propylene 1.960 3.536 207.2 2, 3, 7
5 propane 2.002 3.618 208.1 2, 3, 7
copolymer, and ethylene-propylene-1-butene copolymer. The 6 polypropylene 4.147 294.0 0.0228 2, 3, 8
temperature for the prepolymerization reactor is controlled at 7 catalyst 25.0 2.668 198.8 2, 3
20 and 70 C for the main polymerization reactor. The pressure 8 cocatalyst 25.0 2.668 198.8 2, 3
of those three reactors can range from 3.4 to 4.5 MPag, and it 9 electron donor 25.0 2.668 198.8
is 4.0 MPag in this discussion. The axial flow pumps are used
to provide the driving force for circulation and help to mix the On the other hand, it is wellknown that the density of slurry
slurry well in the reactors. Cooling jackets possess high heat in the tubular loop reactor not only affects the residence time
transfer rates due to their high aspect ratio. of the reagents but also is one of the most important control
1.3. Modeling Technique. As reported in refs 2-4 the model factors for protection of axial flow pump. However, we found
integrates the fundamental chemical engineering principles that using the parameters of polypropylene from the literature2,3,7,8
and advanced software tools to simulate both steady-state and would always lead to overpredicting the density in the tubular
dynamic processes. For the modeling part, there were mass and loop reactors. Therefore, we adjusted the pure-component
energy balances, physical properties, phase equilibrium, po- parameters for polypropylene and changed the mixture rule for
lymerization kinetics, and reactor modeling. We used Polymers liquid mixture molar volume (VLMX), a submodel in PC-SAFT
Plus and Aspen Dynamics developed by Aspen Tech. Co. to EOS for calculation of the density of the liquid mixture. Figures
simulate the process. 3 and 4 show comparisons of the polypropylene density and
heat capacity data obtained via the PC-SAFT EOS and that from
2. Physical and Thermodynamic Properties experiment.23 We could find that the parameters used in this
2.1. Introduction. We used the PC-SAFT equation of state work provide a better prediction for the polypropylene density
(EOS) to predict the physical and thermodynamics properties and almost the same precision for polypropylene heat capacity
for components involved in the polypropylene process. The PC- in the literature. Furthermore, the ideal mixing model (VL2IDL)
SAFT EOS,7,8 regarded as an extension of the well-known was used for VLMX for a better prediction of the liquid mixture
SAFT EOS, was developed specifically for polymer systems density in the tubular loop reactor. By using the ideal mixing
by Gross and Sadowski. For more information on the PC-SAFT model, the pure-component mole volume was calculated by PC-
EOS, readers are encouraged to refer to refs 2, 3, 6-8, and 19. SAFT EOS and the liquid mixture mole volume was calculated
In this work, most of pure-component parameters except for by VL2IDL.
polypropylene involved in the process were taken directly from 2.2. Polymer Properties. In this work, critical polymer
refs 2, 3, 7, and 8. The applied pure-component parameters for properties that are involved in the process are molecular weight
the PC-SAFT EOS are listed in Table 1. distribution and melt index. They were investigated in our
324 Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011

Figure 3. Density of polypropylene.

Figure 4. Heat capacity of polypropylene.


Figure 5. Tubular loop reactor. The locations of the monomer/prepolymer
previous work.2,21,26 Here, the same simulation method and feed streams and polymer product stream on the reactor are similar to those
on R202 described in Figure 2.
similar equations were adopted.
The above model was then compared with a CSTR model, the
3. Reactor Modeling other alternative for tubular loop reactor modeling.
For modeling the tubular loop reactor, many researchers20,21 For modeling the fluidized bed reactor R401, we considered
considered that the reactor could be treated as a continuous it as a CSTR, which was the same as the one in our previous
stirred tank reactor (CSTR) when the recycle ratio is high. In work,2 where the RCSTR module in Aspen Plus was chosen.
this work, we will discuss the effect of recycle ratio on reactor
operation and explore whether the tubular loop reactor can be 4. Polymerization Kinetics
treated as a CSTR.
As shown in Figure 5, the whole tubular loop reactor has 4.1. Introduction. The polymerization kinetics for Ziegler-
four vertical pipes with cooling jackets which are connected Natta catalysts has been studied extensively.12-15,27 Due to the
by four adiabatic elbows. Correspondingly, the combined plug multisite nature of the catalysts, it can produce polyolefin with
flow reactor model is comprised of four plug flow reactors with a broad molecular weight distribution (MWD). In this work,
cooling jackets connected by four adiabatic plug flow reactors. we developed a multisite kinetic model for the homopolypro-
Here, we chose the RPlug module in Aspen Plus to represent a pylene process and three steps were put forward in the
plug flow reactor. Figure 6 shows a schematic representation polymerization kinetics. First, a base set of the single-site
of the combined plug flow reactor model. R201-A, R201-B, reaction mechanism and kinetic parameters was developed based
R201-C, and R201-D represent the four RPlugs with cooling on the literature. Second, the kinetic parameters were finely
jackets, and R201-1, R201-2, R201-3, and R201-4 are four tuned to match the polymer production rate and average number
adiabatic RPlugs. The prepolymerization slurry is introduced of polymerization degree (DPn). Finally, the single-site kinetic
between R201-B and R201-2; the fresh propylene is fed between model was extended to multisite kinetic model using MWD
R201-A and R201-1. Pump P201 here is used for modeling the deconvolution to match the polymer polydispersity index (PDI).
axial flow pump in the tubular loop reactor, with which we can 4.2. Homopolymerization Kinetic Scheme. According to
balance the pressure in the reactor. OUT-SPLI is a stream splitter the literature,12-15 a subset of Ziegler-Natta reactions listed
module, through which we can model the reactor at different in Table 2 was used to describe homopolymeriztion reactions,
recycle ratios by adjusting the output fraction in OUT-SPLI. as follows.
Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011 325

Figure 6. Schematic flow sheet representation of the combined plug reactor model.

Table 2. Reaction Subset Used for Homopolymerization Kinetics where kini,i is the rate constant of chain initiation at the ith site
reaction no. description type and P1,i is an initiated catalyst site of type i with a single
monomer attached to it.
1 catalyst site activation
2 chain initiation
4.2.3. Chain Propagation. The polymer chains grow when
3 chain propagation the monomer continues to attach to the active catalyst sites
4 chain transfer
kp,i
5 catalyst site deactivation
Pn,i + M 98 Pn+1,i (4)
4.2.1. Catalyst Activation. In the polypropylene process of
the Spheripol technology, there is premixing of catalyst and where kp,i is the rate constant for chain propagation at the ith
cocatalyst before being put into the reactor. Therefore, catalyst site type and Pn,i and Pn+1,i are polymer chains of length n and
activation by cocatalyst was not considered in this study. Only n + 1 attached to site type i, respectively. The chain propagation
activation of catalyst by monomer and hydrogen were consid- rate constant plays a key role in the polymer molecular weight.
ered in this work and the reactions are as follows In addition, with the increase of the rate constant, the polymer
molecular weight increases linearly.
kacm,i 4.2.4. Chain Transfer. A chain transfer reaction will lead
CATi + M 98 P0,i (1) to disengaging of the growing polymer chain from the catalyst
active site. Accordingly, a dead polymer chain and an empty
active site or an initiated active site will form. Hydrogen is
kach,i
CATi + H2 98 P0,i (2) always used as a chain transfer agent in the polyolefin process,
and monomer will also cause the chain transfer reaction. On
the basis of the above description, the chain transfer reactions
where kacm,i and kach,i are the rate constants for activation of are as follows
catalyst at the ith site type by monomer and hydrogen, kth,i
respectively. Pn,i + H2 98 Dn + P0,i (5)
A max sites parameter was employed to represent the
number of catalyst sites per unit mass of catalyst within Polymer
ktm,i
Plus. The value of this parameter may range from 1.0 10-5 Pn,i + M 98 Dn + P1,i (6)
to 1.0 10-3 mol of sites per gram of catalyst. We can adjust
the production rate without affecting the molecular weight by
adjusting this parameter. where kth,i and ktm,i are the rate constants for chain transfer to
4.2.2. Chain Initiation. An active site can be initiated by hydrogen and monomer at the ith site type, respectively, and
reacting with the monomer Dn is a dead polymer chain. We adjusted these two kinetic
constants to match the polymer number-average molecular
kini,i weight.
P0,i + M 98 P1,i (3) 4.2.5. Spontaneous Catalyst Deactivation. The catalyst
active sites can undergo spontaneous deactivation to form dead
326 Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011

kds,i
Pn,i Dn + DCATi (8)

where kds,i is the rate constant of deactivation for the ith site
type and DCATi is a deactivated site for the ith site type. In
addition, production of the polymer decreases with the increase
of this deactivation rate constant.
4.3. Determination of Kinetic Parameters. In order to
accomplish the process simulation, the kinetic constants must
be obtained in advance. We suggested the determination
program/method in our previous study.2 Here, the same deter-
mination method was applied and has not been described yet
due to limited space. However, the main steps were listed.
To simplify the parameters determination for highly coupled
polymerization kinetics of the Ziegler-Natta catalyst, first, we
Figure 7. Deconvolution of a representative MWD results using five-site assume the catalyst contains a single site type which is capable
model.
of modeling the polymerization rate and Mn but not Mw or
Table 3. Deconvolution Results for a Representative Polypropylene equivalently PDI; second, we deconvolute GPC data to confirm
Sample the appropriate number of active site types and then determine
site type polymer weight fraction Mn the relative amount of polymer and the corresponding Mn
produced by each site type. Appendix A (Supporting Informa-
1 0.1677 29 550
2 0.3595 78 740 tion) describes the input variables, and Appendix B (Supporting
3 0.0934 599 580 Information) describes the initial kinetic parameters for the
4 0.3352 208 590 single-site model. Simulation targets for models with single and
5 0.0442 8810
multisite types catalyst are listed in Table 6 (Supporting
Information).
sites with no activity, and the corresponding reaction equations
There are five site types used in this work, and the
are as follows
representative sets of the results are shown in Figure 7 and Table
3. In addition, the reaction rate constants for multisite
kds,i
P0,i DCATi (7) Ziegler-Natta catalyst employed in this study are listed in Table
4.

Table 4. Reaction Rate Constants for Multisite Ziegler-Natta Catalyst


reaction site no. comp l comp 2 pre-exp activation energy (kcal/mol) order reference temp. (C)
act-H2 1 catalyst H2 0.002677 6.4 1 69
act-H2 2 catalyst H2 0.002677 6.4 1 69
act-H2 3 catalyst H2 0.002677 6.4 1 69
act-H2 4 catalyst H2 0.002677 6.4 1 69
act-H2 5 catalyst H2 0.0018 6.4 1 69
act-mon 1 catalyst C3H6 1.00 10-6 7.64 1 69
act-mon 2 catalyst C3H6 1.00 10-6 7.64 1 69
act-mon 3 catalyst C3H6 1.00 10-6 7.64 1 69
act-mon 4 catalyst C3H6 1.00 10-6 7.64 1 69
act-mon 5 catalyst C3H6 1.00 10-7 7.64 1 69
chain-ini 1 C3H6 125.775 5.52 1 69
chain-ini 2 C3H6 269.625 5.52 1 69
chain-ini 3 C3H6 70.05 5.52 1 69
chain-ini 4 C3H6 251.4 5.52 1 69
chain-ini 5 C3H6 33.15 5.52 1 69
propagation 1 C3H6 C3H6 125.775 5.52 1 69
propagation 2 C3H6 C3H6 269.625 5.52 1 69
propagation 3 C3H6 C3H6 70.05 5.52 1 69
propagation 4 C3H6 C3H6 251.4 5.52 1 69
propagation 5 C3H6 C3H6 33.15 5.52 1 69
chat-mon 1 C3H6 C3H6 0.075 65.42 1 69
chat-mon 2 C3H6 C3H6 0.061 65.42 1 69
chat-mon 3 C3H6 C3H6 0.0021 65.42 1 69
chat-mon 4 C3H6 C3H6 0.022 65.42 1 69
chat-mon 5 C3H6 C3H6 0.067 65.42 1 69
chat-H2 1 C3H6 H2 7.5 10.7 0.5 69
chat-H2 2 C3H6 H2 6.1 10.7 0.5 69
chat-H2 3 C3H6 H2 0.21 10.7 0.5 69
chat-H2 4 C3H6 H2 2.2 10.7 0.5 69
chat-H2 5 C3H6 H2 6.7 10.7 0.5 69
deact-spon 1 2.30 10-5 1 1 69
deact-spon 2 2.30 10-5 1 1 69
deact-spon 3 2.30 10-5 1 1 69
deact-spon 4 2.30 10-5 1 1 69
deact-spon 5 2.30 10-5 1 1 69
Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011 327

Figure 8. Simple control schematic for the bulk polymerization unit.

5. Dynamic Modeling
5.1. Introduction. The dynamic model is developed by
exporting the steady-state model from Polymer Plus into Aspen
Dynamics with complementary information, for example, reactor
and some other vessels dimensions and geometries. Meanwhile,
the default PID control schemes are configured. Some additional
control schemes are needed for reliable modeling of the dynamic
process, including setup, grade transition, and the emergent shut
down. Furthermore, the CSTR model which was used in the
steady-state model had been discussed earlier. Therefore, the
CSTR used in the dynamic model is not discussed here.
5.2. Main Control Scheme. Figure 8 shows an illustrative
control schematic for the bulk polymerization reactors. It
included the following controllers. (1) Level and temperature
controllers were used for the three tubular loop reactors. The
lever controllers kept the reactors level by adjusting the output Figure 9. Temperature distribution along the tubular loop reactor.
of the slurry. The temperature controllers adjusted the heat
capacity of the reactors to keep the reactor at constant temper- decreased along the pipe obviously at this section due to the
ature. (2) As for R201 and R202, density control was very significant temperature difference between reactant and cooling
important for these two reactors and propylene feed rate was water. As the temperature of the cooling water became higher,
adjusted instead of the reactor density to control the polymer the above temperature difference became smaller, which is
mass fraction of the output stream of the reactors. (3) Total shown in Figure 10. In addition, the decrease of the reactant
polymer production rate was controlled by adjusting the catalyst temperature along the pipe became less evident, even increased
feed flow rate. (4) Three FSplit modules (Sp0, Sp1, Sp2) after at R201-A. Therefore, it is noted that that a good control of the
three tubular loop reactors were used to split the slurry from temperature and flow rate of cooling water is very important
the three reactors to the high-pressure blow-down vessel (D601) for reactor temperature control. From Figure 9 one can also find
in the case of emergent shut down. that there was an obvious temperature decrease at the following
two positions: the first one was between R201-A and R201-1
because of the feed of fresh propylene, and the second one was
6. Simulation Results
between R201-B and R201-2 because of the feed of prepoly-
6.1. Reactor Models. In this work, both the CSTR model merization slurry from R200. Furthermore, from Figure 9 one
and the combined PFR model for reliable modeling of the knows that the temperature scale on the y axis showed that the
tubular loop reactors were first developed. Results from the data varied only from about 69.950 to 70.175 C over a distance
combined PFR model were discussed and compared with those of 230 m. From an engineering point of view, this result showed
from the CSTR model. effectively a totally isothermal reactor for all practical purposes.
Figure 9 shows the temperature distribution along the loop The fact that the temperature profile is totally isothermal is
reactor. From Figure 9 one can find that the temperature helpful to produce a resin with excellent inherent properties,
increased obviously at four adiabatic elbows of R201-1, R201- i.e., Melt Index, MWD, etc.
2, R201-3, and R201-4 due to high reaction heat. As for the Figures 11-13 described the effect of recycle ratio on the
other four vertical pipes with cooling jackets, R201-D was reactor operation. According to Figure 11, one can obtain that
positioned as the inlet of cooling water, the temperature the temperature distribution was uneven at a low recycle ratio;
328 Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011

Figure 10. Coolant temperature distribution along the tubular jacket.

Figure 12. Effect of the recycle ratio on the polymer molecular weight
Figure 11. Effect of the recycle ratio on the loop reactor temperature distribution.
distribution.
differences were caused not only by the temperature contrast
however, when the recycle ratio reached 20, the temperature but also by the hydrogen concentration difference between the
distribution would become uniform. In practice, at low slurry two models. Figure 13 shows the effect of the recycle ratio on
recycle flow rate, the fresh propylene feed affected the reactor the hydrogen concentration distribution; the same tendency with
temperature greatly and a sharp decrease of temperature can be Figure 12 can be found. With the increase of the recycle ratio,
found from 71 to 66 C. Meanwhile, at the adiabatic elbows the hydrogen concentration obtained from the combined PFR
the temperature increased obviously due to the long residence models approached that obtained from the CSTR model.
time. On the contrary, the temperature decreased dramatically Figure 14 presents the effect of the recycle ratio on the
at the vertical pipes with cooling jackets because of the smaller polymer production rate and DPn. From Figure 14 one knows
reaction heat along with the lower recycle flow rate. that with the increase of the recycle ratio the polymer producing
Figure 12 shows the effect of the recycle ratio on the polymer rate increased but the DPn decreased. In practice, at a high
molecular weight distribution. When the recycle ratio was 5, recycle ratio the residence time distribution of polymers is close
the MWD was obviously different from that obtained from the to that in CSTR, which leads to the decrease of the breakage
CSTR model. When the recycle ratio was up to 20, the MWD possibility of catalyst.2 Accordingly, catalysis activity can be
calculated by the combined PFR model was always consistent released efficiently and the polymer producing rate increases.
with the increase of the recycle ratio but there were are still In addition, the change of the DPn was the same with the MWD
some differences from the result of the CSTR model. The shown in Figure 12.
Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011 329

Figure 15. Polymer grade-transition strategy preformed by a step change


Figure 13. Effect of the recycle ratio on the hydrogen molar fraction in hydrogen flow rate.
distribution.

Figure 14. Effect of the recycle ratio on the polymer production rate and
DPn. Figure 16. Heat duty of the two reactors in the start-up process.

Table 5. Comparison of the Modeling Results with the Data from


Plant
in Figure 15. From Figure 15 one observes that the polymer
Mn in each reactor stepped to a smaller stage. Due to a lower
R201 R202
hydrogen concentration in R201, a larger polymer Mn can be
properties predicted plant predicted plant obtained, which is in line with the reality of the plant
density (kg/m3) 545.5 546.7 583.7 584.8 observation. Besides, as the hydrogen concentrations in the two
Mn 66 557 65 300 48 604 49 000 reactors were adjusted simultaneously and separately, it took
Mw 325 207 317 700 253 413 259 400 almost the same time for the two reactors to achieve steady
PDI 4.89 4.86 5.21 5.29
state. The model can be used to optimize the grade transition
process to minimize the transition time and the amount of off-
As a whole, by using the combined PFR model, one can
specification polymer produced by adjusting the control strategy.
obtain some detailed information on the reactors and it is useful
for design of the reactor. However, as for process modeling, As for the start-up process, a step feed of catalyst was
the CSTR model can be used to get appropriate results such as considered. It is very important to control the reactor temperature
the polymer production rate and MWD. Therefore, the following in the start-up process. Figure 16 shows the demanding heat
process modeling results were based on the CSTR model for duties of R201 and R202. It was a positive value at the initial
saving the simulation time. period of the start-up process, as the reaction heat was not
6.2. Model Validation. In our investigation, the plant data sufficient due to a low catalyst concentration; the hot water was
have been used to fit the kinetic parameters. The polymer required to sustain the reactors temperature at 70 C. As the
properties obtained from polymer samples were also incorpo- process proceeded, the catalyst concentration increased and the
rated to validate our model. Table 5 shows the comparison of reaction rate became larger; meanwhile, cold water was then
the modeling results with the plant data in terms of slurry needed, and the value of heat duty changed to negative. Besides,
density, MW, PDI, and MI in the reactors, demonstrating that the heat duties were almost the same for the two reactors because
the simulated results agree well with the data obtained from of the same polymer production rate as shown in Figure 17.
the plant. Here, we also point out that the polymer producing rates shown
6.3. Model Application. Several typical dynamic processes, in Figure 17 were the total polymer flow rate from the two
i.e., grade transition, start up, and shut down at certain emergent reactors. The polymer from R202 was produced by both R201
accidents, were simulated via the dynamic model. and R202, and we can see it was almost twice as much as the
For the grade transition process, we considered a step increase polymer producing rate of R201. Therefore, the polymer
of the hydrogen feed rate, and the simulation results are shown production rate was almost the same in the two reactors.
330 Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011

kinetics models based on single and multisite catalyst were


established by fitting plant data and GPC deconvolution results.
With conceptual control schemes added to the well-established
steady-state modules, the enhanced model system can be used
to simulate the dynamic behavior.
Though the modeling results were in good agreement with
the plant data, more data of different scenarios or product grade
were still important for further validation and improvement of
the flexibility of our models. The model was used to simulate
dynamics behavior such as grade transition, start up, and
emergent shut down. Those applications could potentially benefit
operators as well as process engineers.

Acknowledgment
The authors gratefully acknowledge China National Petroleum
Figure 17. Polymer production of the two reactors in the start-up process. Corporation, China Lanzhou Petrochemical Company, China
Lanzhou Petrochemical Company Research Institute (particu-
larly Zhao, X. T., the Senior President; Zhu, B. C., the Senior
Associate President), and Dalian Nationalities University (par-
ticularly Dr. Li B.H.) for supporting this work. We also thank
the anonymous referees for comments on this manuscript. The
simulation work was implemented by advanced software tools
(Polymers Plus and Aspen Dynamics) provided by China
Lanzhou Petrochemical Company Research Institute and Dalian
Nationalities University.

Supporting Information Available: Appendix A (input


variables and simulation targets for models for catalysts with
single and multiple site types); Appendix B (initial Kinetic
Parameters for the Single-site Model). This material is available
free of charge via the Internet at http://pubs.acs.org.

Figure 18. Flare gas flow rate in the shut-down process. Nomenclature
In addition, the shut-down process at a certain emergent CATi ) inactive catalyst site of type i
accident was also simulated via our dynamic model in order to Cp ) heat capacity, kJ/kg
find out the flare line load when the plant was at certain Dn ) inactive polymer chain containing n monomer segments
emergent accident. The disable reactions option in the reactor DPn ) number average of the polymer degree of polymerization
module was selected to stop the reaction by modeling the DPw ) weight average of the polymer degree of polymerization
injection of reaction terminating agent carbonic oxide, and the DCATi ) deactivated catalyst site of type i
reactant in the tubular loop reactors was discharged to the high- kach,i ) rate constants for the activation of catalyst at the ith site
pressure discharge vessel D601 in 20 min. According to the type by hydrogen, s-1
industrial data collected from a chemical plant in China, injection kacm,i ) rate constants for the activation of catalyst at the ith site
of carbonic oxide is in order to stop polymerization in the reactor type by monomer, s-1
for a little poison lead to greatly decrease the catalyst activity. kds,i ) rate constant for spontaneous deactivation of catalyst site
Therefore, the above section/option was reasonable for the shut- type i, s-1
down simulation. Figure 18 shows the flare flow rate during kini,i ) rate constant for active site initiation at site type i, L/mol s
the shut-down process in the emergent accident. From Figure kp,i ) rate constant for initiated active site propagation at site type
18 one can find that when the emergent shut-down began at i, L/mol s
1 h, the flare flow rate grew rapidly and the maximum flow kth,i ) rate constant for chain transfer to hydrogen for site type i,
rate can reach 90 T/h. After all of the slurry in the tubular loop L/mol s
reactors was discharged to D601, the flare flow rate dropped ktm,i ) rate constant for chain transfer to monomer for site type i,
rapidly. The flare flow rate grew again at about 1.5 h. That is L/mol s
because of the heating and steaming of D601. The nitrogen gas M ) monomer component
was used to dry the polypropylene at about 4 h, and the flare MI ) melt index, g/10 min
gas flow rate grew again. MWD ) molecular-weight distribution
Mn ) number-average molecular weight, g/mol
Mw ) weight-average molecular weight, g/mol
7. Conclusions
n ) number of monomer segments in the polymer (degree of
In this work, we described the strategy and methodology for polymerization)
modeling the polypropylene process of the Spheripol technology. PDI ) polymer polydispersity index
The improved PC-SAFT EOS was used to predict the thermo- P0,i ) activated catalyst site of type i
physical properties and phase equilibrium of the system, and P1,i ) initiated catalyst site of type i
both CSTR and the combined PFR model were developed for Pn,i ) live polymer chain containing n segments attached to catalyst
modeling the tubular loop reactor. In addition, polymerization site type i
Ind. Eng. Chem. Res., Vol. 50, No. 1, 2011 331
r ) size parameter for polymer species in the PC-SAFT EOS; ratio principles and illustrative examples, polypropylene. Chem. Eng. Sci. 1996,
of the characteristic chain length to the number-average molecular 51, 48594868.
(15) Kissin, Y. V. Multicenter nature of titanium-based Ziegler-Natta
weight, mol/g catalysts: comparison of ethylene and propylene polymerization reactions.
T ) temperature, C J Polym. Sci. Polym Chem. 2003, 41, 17451754.
VL2IDL ) the ideal mixing model for liquid mixture molar volume (16) Arriola, D. J. Modeling of Addition Polymerization System. Ph.D.
VLMX ) liquid mixture molar volume, m3/kg Dissertation, University of Wisconsin, Madison, WI, 1989.
/k ) the segment energy parameter in the PC-SAFT EOS, K (17) Feng, L. F.; Li, F. Y.; Gu, X. P.; Wang, J. J.; Tang, Z. W.; Liu, B.
Steady-state modeling of commercial liquid phase bulk polypropylene
) the segment diameter in the PC-SAFT EOS, process with polymers plus. Petrochem. Technol. 2005, 34, 237241 (In
Chinese).
(18) Soares, J. B. P.; Hamielec, A. E. Deconvolution of chain-length
Literature Cited distribution of linear polymers made by multiple-site-type catalyst. Polymer
1995, 36, 22572264.
(1) Hong, D. Y. Polypropylene-Principle, Process and Technology;
(19) Bokis, C. P.; Ramanathan, S.; Franjione, J.; Buchelli, A.; Call,
China Petrochemical Press: Beijing, 2002 (in Chinese).
M. L.; Brown, A. L. Physical properties, reactor modeling, and polymer-
(2) Luo, Z. H.; Su, P. L.; Shi, D. P.; Zheng, Z. W. Steady-state and
ization kinetics in the low-density polyethylene tubular reactor process. Ind.
dynamic modeling of commercial bulk polypropylene process of Hypol
Eng. Chem. Res. 2002, 41, 10171030.
technology. Chem. Eng. J. 2009, 149, 370382.
(20) Zacca, J. J.; Ray, W. H. Modelling of the liquid phase polymeri-
(3) Khare, N. P.; Luca, B.; Seavey, K. C.; Liu, Y. A.; Sirohi, A.;
zation of olefins in loop reactors. Chem. Eng. Sci. 1993, 48, 37433765.
Ramanathan, S.; Lingard, S.; Song, Y. H.; Chen, C. C. Steady-state and
(21) Luo, Z. H.; Zheng, Y.; Cao, Z. K.; Wen, S. H. Mathematical
dynamic modeling of gas-phase polypropylene processes using stirred-bed
modeling of the molecular weight distribution of polypropylene produced
reactors. Ind. Eng. Chem. Res. 2004, 43, 884900.
in a loop reactor. Poly. Eng. Sci. 2007, 47, 16431649.
(4) Khare, N. P.; Seavey, K. C.; Liu, Y. A.; Ramanathan, S.; Lingard,
(22) Eneida, A. D. E.; Rubens, M. F.; Pramo, A. M.; Jose, C. P.
S.; Chen, C. C. Steady-state and dynamic modeling of commercial slurry
Modeling and simulation of liquid phase propylene polymerizations in
high-density polyethylene (HDPE) processes. Ind. Eng. Chem. Res. 2002,
industrial loop reactors. Macromol. Symp. 2008, 271, 814.
41, 56015618.
(23) Sato, Y.; Yamasaki, Y.; Takishima, S.; Masuoka, H. Precise
(5) Chatzidoukas, C.; Perkins, J. D.; Pistikopoulos, E. N.; Kiparissides,
measurement of the PVT of polypropylene and polycarbonate up to 330C
C. Dynamic simulation of the borstar@ multistage olefin polymerization
and 200 MPa. J. Appl. Polym. Sci. 1997, 66, 141150.
process. Comput.-Aided Chem. Eng. 2003, 14, 593598.
(24) Buchelli, A.; Call, M. L.; Brown, A. L.; Bokis, C. P.; Ramanathan,
(6) Buchelli, A.; Call, M. L.; Brown, A. L.; Bokis, C. P.; Ramanathan,
S.; Franjione, J. A. Comparison of thermodynamic equilibrium versus
S.; Franjione, J. Nonequilibrium behavior in ethylene/polyethylene flash
nonequilibrium behavior in ethylene/polyethylene flash separators. Presented
separators. Ind. Eng. Chem. Res. 2004, 43, 17681778.
at the Aspen Technology users group meeting, Houston, TX, Feb 4-7,
(7) Gross, J.; Sadowski, G. Perturbed-chain SAFT: An equation of state
2001.
based on a perturbation theory for chain molecules. Ind. Eng. Chem. Res.
(25) Flory, P. J. Molecular size distribution in linear condensation
2001, 40, 12441342.
polymers. J. Am. Chem. Soc. 1936, 58, 18771884.
(8) Gross, J.; Sadowski, G. Modeling polymer systems using the
(26) Zheng, Y. Modeling for propylene polymerization in liquid phase
perturbed-chain statistical associating fluid theory equation of state. Ind.
loop reactor. Masters Thesis, Xiamen University, China, 2006.
Eng. Chem. Res. 2002, 41, 10841093.
(27) Buchelli, A.; Topliss, R. J.; Prewitt, L.; Palas, R. F.; McCullum,
(9) Orbey, H.; Bokis, C. P.; Chen, C. C. Polymer-solvent vapor-liquid
A. Plant simulator improves PP technology transfer. Hydrocarbon Process.
equilibrium: Equation of state versus activity coefficient models. Ind. Eng.
1999, 78, 8596.
Chem. Res. 1998, 37, 15671578.
(28) Urdampilleta, I.; Gonzalez, A.; Iruin, J. J.; de la Cal, J. C.; Asua,
(10) Lu, C.; Zhang, M. H.; Jiang, S. J.; Song, D. Z. The application of
J. M. Origins of product heterogeneity in the Spheripol high impact
Aspen Plus softerware on the large scare polypropylene process. Qi Lu
polypropylene process. Ind. Eng. Chem. Res. 2006, 45, 41784187.
Pertochem. Technology. 2006, 34, 404409 (In Chinese).
(29) Stolarski, L. Modern technologies of polyolefins production in
(11) Feng, L. F.; Li, F. Y.; Gu, X. P.; Tang, Z. W.; Liu, B.
Basell Orlen Polyolefins Company. Part I. Spheripol process. Polimery
Thermodynamic property simulation of propylene-hydrogen-polypropylene
2005, 50, 894898.
system using PC-SAFT equation of state. Petrochem. Technol. 2005, 34,
(30) Wang, J.; Luo, Z. H.; Zhu, Y. Kinetics of propylene polymerization
152156 (In Chinese).
catalyzed with CS-1 catalyst. Chem. Eng. (China) 2009, 37, 4247.
(12) Choi, K. Y.; Ray, W. H. The dynamic behavior of continuous
(31) Su, P. L. Flowsheet Simulation for Bulk Propylene Polymerization
stirred-bed reactors for the solid catalyzed gas phase polymerization of
Based on Polymers Plus. Masters Thesis, Xiamen University, China, 2009.
propylene. Chem. Eng. Sci. 1988, 43, 25872599.
(13) Caracotsios, M. Theoretical modeling of Amocos gas-phase ReceiVed for reView August 10, 2010
horizontal stirred bed reactor for the manufacturing of polypropylene resins. ReVised manuscript receiVed October 11, 2010
Chem. Eng. Sci. 1992, 47, 25912596. Accepted November 12, 2010
(14) Zacca, J. J.; Debling, J. A.; Ray, W. H. Reactor residence time
distribution effects on the multistage polymerization of olefins-I. Basic IE101699B

Você também pode gostar