Reactor Modeling and Recipe Optimization ofPolyether Polyol Processes
Spring 2013 EWO Meeting
Yisu Nie1 Lorenz T. Biegler1 Carlos M. Villa2 John M. Wassick2
1Center for Advanced Process Decision-makingDepartment of Chemical Engineering
Carnegie Mellon University
2The Dow Chemical Company
March 13, 2013
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 1 / 12
IntroductionProject timeline
Development of systematic optimization methods of batch processes
Nov. 2009 - Dec. 2011I Proof-of-concept: integration of scheduling and dynamic optimizationI Y. Nie, L.T. Biegler, and J.M. Wassick. Integrated Scheduling and Dynamic Optimization of Batch Processes Using
State Equipment Networks. AIChE Journal, 2012.
Jan. 2012 - Dec. 2012I Application on polyether polyol processes at DowI Homopolymerization: polypropylene glycol
1 First-principle reactor model: population balances, kinetics, VLE, etc.2 Hypothetical optimization case study: over 40% batch time reduction
I Y. Nie, L.T. Biegler, C.M. Villa, and J.M. Wassick. Reactor Modeling and Recipe Optimization of Polyether Polyol
Processes: Polypropylene Glycol. submitted for publication, 2013.
I Copolymerization: block copolymer polyol1 Reactor modeling with the method of moments2 Hypothetical optimization case study: over 40% batch time reduction
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 2 / 12
IntroductionPolyether polyol process description
Key ingredientsI Epoxides (ethylene oxide (EO), propylene
oxide (PO))
OO OOI Molecules containing active hydrogen atoms
(alcohols, amines)
OH N
H
HI A basic catalyst (KOH)
Typical procedures usedI Starters are first mixed with catalyst in the liquid phaseI Alkylene oxides in the liquid phase are fed in controlled ratesI The reactor temperature is controlled by the heat exchangerI Allowed maximum reactor pressure guarded by the vent system control valve
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 3 / 12
IntroductionPolyether polyol process description
Key ingredientsI Epoxides (ethylene oxide (EO), propylene
oxide (PO))
OO OOI Molecules containing active hydrogen atoms
(alcohols, amines)
OH N
H
HI A basic catalyst (KOH)
N2
Monomer
Reactor
Typical procedures usedI Starters are first mixed with catalyst in the liquid phaseI Alkylene oxides in the liquid phase are fed in controlled ratesI The reactor temperature is controlled by the heat exchangerI Allowed maximum reactor pressure guarded by the vent system control valve
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 3 / 12
Process Dynamic ModelingModeling reaction kinetics
Many polyether polyol products are copolymersI Example: EO capped diols
H2C
H2C
O OOH
nP nE
O OOH
nP nE
PO block EO block
EO/PO block copolymer
Challenge: the effect of chain ends on chain growth
Propagation scheme Rate constant
. . .EO∗ + EO→ . . .EO∗ kEEp
. . .EO∗ + PO→ . . .PO∗ kEPp
. . .PO∗ + EO→ . . .EO∗ kPEp
. . .PO∗ + PO→ . . .PO∗ kPPp
Terminal unit is designated by ∗
kEEp > kEP
p > kPEp > kPP
p
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 4 / 12
Process Dynamic ModelingModeling EO/PO copolymerization processes
Reaction scheme: EO/PO copolymerInitiation:
G0,0 + EOkEi−→ GE
0,1
Q0,0 + EOkEi−→ QE
0,1
G0,0 + POkPi−→ GE
1,0
Q0,0 + POkPi−→ QE
1,0
Propagation:
GPnP,nE
+ POkPPp−→ GP
nP+1,nE(nP + nE > 1)
QPnP,nE
+ POkPPp−→ QP
nP+1,nE(nP + nE > 1)
GPnP,nE
+ EOkPEp−→ GE
nP,nE+1 (nP + nE > 1)
QPnP,nE
+ EOkPEp−→ QE
nP,nE+1 (nP + nE > 1)
GEnP,nE
+ POkEPp−→ GP
nP+1,nE(nP + nE > 1)
QEnP,nE
+ POkEPp−→ QP
nP+1,nE(nP + nE > 1)
GEnP,nE
+ EOkEEp−→ GE
nP,nE+1 (nP + nE > 1)
QEnP,nE
+ EOkEEp−→ QE
nP,nE+1 (nP + nE > 1)
Transfer:
GPnP,nE
+ POkt−→ DP
nP,nE+Q0,0 (nP,nE > 0)
GEnP,nE
+ POkt−→ DE
nP,nE+Q0,0 (nP,nE > 0)
QPnP,nE
+ POkt−→ RP
nP,nE+Q0,0 (nP,nE > 0)
QEnP,nE
+ POkt−→ RE
nP,nE+Q0,0 (nP,nE > 0)
Exchange:
GPnP,nE
+DPmP,mE
ke−→ DPnP,nE
+GPmP,mE
(nP,nE,mP,mE > 0)
GPnP,nE
+DEmP,mE
ke−⇀↽−ke
DPnP,nE
+GEmP,mE
(nP,nE,mP,mE > 0)
GEnP,nE
+DEmP,mE
ke−→ DEnP,nE
+GEmP,mE
(nP,nE,mP,mE > 0)
QPnP,nE
+RPmP,mE
ke−→ RPnP,nE
+QPmP,mE
(nP,nE,mP,mE > 0)
QPnP,nE
+REmP,mE
ke−⇀↽−ke
RPnP,nE
+QEmP,mE
(nP,nE,mP,mE > 0)
QEnP,nE
+REmP,mE
ke−→ REnP,nE
+QEmP,mE
(nP,nE,mP,mE > 0)
GPnP,nE
+RPmP,mE
ke−⇀↽−ke
DPnP,nE
+QPmP,mE
(nP,nE,mP,mE > 0)
GPnP,nE
+REmP,mE
ke−⇀↽−ke
DPnP,nE
+QEmP,mE
(nP,nE,mP,mE > 0)
GEnP,nE
+RPmP,mE
ke−⇀↽−ke
DEnP,nE
+QPmP,mE
(nP,nE,mP,mE > 0)
GEnP,nE
+REmP,mE
ke−⇀↽−ke
DEnP,nE
+QEmP,mE
(nP,nE,mP,mE > 0)
1
MaterialStarter Low MW diolCatalyst KOHMonomer EO and PO
Notation
Gn growing product chains (PnO−K+)
Dn dormant product chains (PnOH)
Qn growing unsat chains (UnO−K+)
Rn dormant unsat chains (UnOH)
Pn CH3(PO)nUn CH2 = CHCH2(PO)n
Indexn,m repeating units (PO and EO)
SuperscriptE,P terminal units
Polyol chains in the reactorliving dormant
normal GP,GE DP,DE
unsaturated QP,QE RP,RE
total products X = GP + GE + DP + DE
total byproducts Y = QP + QE + RP + RE
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 5 / 12
Process Dynamic ModelingPopulation balances vs. method of moments
Population balance model (PBM)I Differential equations
d(V [XPnP,nE
])
dt=V (k
PPp [G
PnP−1,nE
] + kEPp [G
EnP−1,nE
])[PO]− V (kPPp [PO] + k
PEp [EO])[G
PnP,nE
]
d(V [XEnP,nE
])
dt=V (k
PEp [G
PnP,nE−1] + k
EEp [G
EnP,nE−1])[EO]− V (k
EPp [PO] + k
EEp [EO])[G
EnP,nE
]
d(V [YPnP,nE
])
dt=V (k
PPp [Q
PnP−1,nE
] + kEPp [Q
EnP−1,nE
])[PO]− V (kPPp [PO] + k
PEp [EO])[Q
PnP,nE
]
d(V [YEnP,nE
])
dt=V (k
PEp [Q
PnP,nE−1] + k
EEp [Q
EnP,nE−1])[EO]− V (k
EPp [PO] + k
EEp [EO])[Q
EnP,nE
]
I Algebraic equations
XSnP,nE
nc = GSnP,nE
(ni + nu), S ∈ {P,E}
YSnP,nE
nc = QSnP,nE
(ni + nu), S ∈ {P,E}
nc total moles of catalystni total moles of initiatornu total moles of unsaturated chains
Pros and cons
+ Revealing full MWD spectrum− Computationally demanding
2000 MW diol > 1400 differential equations + 1400 algebraic equations
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 6 / 12
Process Dynamic ModelingPopulation balances vs. method of moments
Definition of moments: kth order moment of polymeric species
λk =
∞∑nP+nE>1
(nP + nE)kXnP,nE , ζk =
∞∑nP+nE>1
(nP + nE)kGnP,nE , k = 0, 1, . . .
µk =
∞∑nP+nE>1
(nP + nE)kYnP,nE , νk =
∞∑nP+nE>1
(nP + nE)kQnP,nE , k = 0, 1, . . .
Moment model (MM)I Differential equations
dλPkdt
= V −1(kPi G0PO+ kPPp
k−1∑i=0
(ki
)ζPi PO+ kEP
p
k∑i=0
(ki
)ζEi PO− kPE
p ζPk EO)
dλEkdt
= V −1(kEi G0EO+ kEEp
k−1∑i=0
(ki
)ζEi PO+ kPE
p
k∑i=0
(ki
)ζPi EO− kEP
p ζEk PO)
dµPkdt
= V −1(kPi Q0PO+ kPPp
k−1∑i=0
(ki
)νPi PO+ kEP
p
k∑i=0
(ki
)νEi PO− kPE
p νPk EO)
dµEkdt
= V −1(kEi Q0EO+ kEEp
k−1∑i=0
(ki
)νEi PO+ kPE
p
k∑i=0
(ki
)νPi EO− kEP
p νEk PO)
I Algebraic equations
λSknc = ζSk (ni + nu), S = {P,E}
µSknc = νSk (ni + nu), S = {P,E}
Pros and cons
+ Model size does not increase with polymer MW
0th − 2nd moments = 12 differential equations + 12 algebraic equations
− No detailed distribution
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 7 / 12
Copolymerization Case StudyProduction of EO capped diol
Process specifications
I Charge conditionInitiator: PPGCatalyst: KOHMonomer: PO and EO
I Process constraints*F Product molecular weight > 2000 g/molF Product unsaturation value 6 0.02 mmol/g polyolF Heat exchanger load 6 UA(T − Tw) kWF Adiabatic end temperature 6 (Tb + 90)◦CF Capping ratio > 70% (% of EO ended chains in total chains)
* The set of constraints has been chosen to illustrate the use of dynamic optimization technology and does not include all
possible constraints or necessarily reflect the true capability of the plant.
Model calibration
0.0
0.2
0.4
0.6
0.8
1.0
0 0.2 0.4 0.6 0.8 1
Sca
led
reac
tor
pres
sure
Normalized time
Pressure (MM)Pressure (PBM)Plant data
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 8 / 12
Copolymerization Case StudyProduction of EO capped diol
Process specifications
I Charge conditionInitiator: PPGCatalyst: KOHMonomer: PO and EO
I Process constraints*F Product molecular weight > 2000 g/molF Product unsaturation value 6 0.02 mmol/g polyolF Heat exchanger load 6 UA(T − Tw) kWF Adiabatic end temperature 6 (Tb + 90)◦CF Capping ratio > 70% (% of EO ended chains in total chains)
* The set of constraints has been chosen to illustrate the use of dynamic optimization technology and does not include all
possible constraints or necessarily reflect the true capability of the plant.
Optimization resultsI Optimization model statistics and solution
Opt. soln MW (g/mol) Unsat (mmol/g) Capping(%) # of var. # of con. CPU (s)
0.580 2000 0.02 70 11, 248 11, 121 22
F Optimized with the moment modelF Batch time reduced by 42.0% (base case batch time normalized to 1)F Product quality constraints are satisfied at the end of the operationF More details...
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 8 / 12
Copolymerization Case Study ResultsOptimal control profiles of the process
Reactor temperature (top) and monomer feed rate (bottom) profiles
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
0 0.2 0.4 0.6 0.8 1
Tem
pera
ture
[T−T
b]
Normalized time
End of optimized recipe
OptimizedBase case
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0 0.2 0.4 0.6 0.8 1
Sca
lued
feed
rat
e
Normalized time
Optimized POOptimized EOBase case POBase case EO
Important remarksI Piecewise linear control profiles with continuity on finite element boundariesI Rising temperature for PO digestion and EO feedingI Flat feed rates for both monomers: no need for ramping
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 9 / 12
Copolymerization Case Study ResultsProcess constraint profiles
Adiabatic temp. (top) and hxn duty (bottom)
40
50
60
70
80
90
100
0 0.1 0.2 0.3 0.4 0.5 0.6
Adi
abat
ic e
nd te
mp.
[T−T
b]
Normalized time
Adiabatic end temp.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0 0.1 0.2 0.3 0.4 0.5 0.6
Sca
led
heat
rem
oval
Normalized time
Heat removal dutyCooling capacity
Important remarksI Constraint on the adiabatic temp. is inactiveI Heat exchanger capacity is the major constraining factor
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 10 / 12
Copolymerization Case Study ResultsOptimal product molecular weight distributions (MWD)
MWD of the product (left) and unsat (right) polymers
0 5
10 15
20 25
30 0 2
4
6
8
10
12
14
0
50
100
150
200
250
Product population
[mol]
Number of PO
Number of EO
Product population
[mol]
0
50
100
150
200
250
0 5
10 15
20 25
30 0 2
4
6
8
10
12
14
0
1
2
3
4
5
6
7
Unsat population
[mol]
Number of PO
Number of EO
Unsat population
[mol]
0
1
2
3
4
5
6
7
Important remarksI Similar optimal solution obtained from the PBM
Opt. soln MW (g/mol) Unsat (mmol/g) Capping(%) # of var. # of con. CPU (s)
0.588 2000 0.02 70 191, 408 185, 147 5221
I Optimization over PBM is computationally expensive
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 11 / 12
ConclusionsProject timeline
Development of systematic optimization methods of batch processes
Jan. 2012 - Dec. 2012I Application on polyether polyol processes at DowI Homopolymerization: polypropylene glycol
1 First-principle reactor model: population balances, kinetics, VLE, etc.2 Hypothetical optimization case study: over 40% batch time reduction
I Copolymerization: block copolyol1 Reactor modeling with the method of moments2 Hypothetical optimization case study: over 40% batch time reduction
Jan. 2013 - May. 2014I Simultaneous scheduling and dynamic optimization
F Multiple reactors and possible incorporation of finishing trainsF Real-time constraints on shared resources
I Methodology generalization and further extensions
Thank you
I am glad to take your questions
Yisu Nie (Carnegie Mellon University) Spring 2013 EWO Meeting March 13, 2013 12 / 12