incremental model identification of reaction systems cpac...incremental model identification of...
TRANSCRIPT
§Incremental Model Identification
of Reaction Systems
Julien Billeter
Ecole Polytechnique Fédérale de LausanneLaboratoire d’Automatique
Switzerland
CPAC / ATOCHEMIS Workshop 25 – 27 March 2013, Rome – Italy
Outline
o Motivation
o Concept of vessel extent
o Homogeneous reaction systems
o Incremental model identification
o Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
2
Model reductionSeparate fast / slow dynamics
o Discard redundant (invariant) states• What is the minimal number of states (variants) ?• Batch reactors: S → R extents• Open reactors: S → R+p+1 vessel extents• Open G-L reactors: S → R+pm+p+1 vessel extents
o Separate fast/slow dynamics• Rates are fast/slow, not individual concentrations→ work with extents, not concentrations…
3Bhatt et al, Comp. & Chem. Eng. (2011), submittedM. Amrhein, PhD dissertation n°1861 (1998), EPFL, Switzerland
Outline
Motivation
o Concept of vessel extent
o Homogeneous reaction systems
o Incremental model identification
o Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
4
DefinitionsExtent vs vessel extent of reaction
o Extent of the i-th reaction :number of moles produced by the i-th reaction
5
o Vessel extent of the i-th reaction : number of moles produced by the i-th reaction still in vessel
( ) ( ) ( ),
1, , , 0 0
s ir i s v i r it n r tνξ ξ= = =
Differential extent of reaction
( ),r i tξ
( ) ( ) ( )( ) ( ) ( ), ,, , 0 0out
r i v i ru t
r im t ixx t r t xt= − =
Differential vessel extent of reaction
( ),r ix t
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
DefinitionsExtent vs vessel extent of mass transfer
o Extent of the j-th mass transfer :mass transferred by the j-th mass transfer
6
o Vessel extent of the j-th mass transfer : mass transferred by the j-th mass transfer still in vessel
( ) ( ) ( ), , 0 0m j j m jt tξ ζ ξ= =Differential extent of mass transfer
( ),m j tξ
( ) ( ) ( )( ) ( ) ( ), ,, 0 0outu t
m j j m jjm t mx t t t xxζ −= =
Differential vessel extent of mass transfer
( ),m jx t
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
General concept of vessel extent
A vessel extent indicates the amount of material (number of moles, mass, volume), associated with one phenomenon, that is still in the vessel…
7
( ) ( ) ( )( ) ( ) ( ), ,, , 0 0outu t
f im tf i f i f ix tx t t xξ −= =
Differential vessel extent (i-th phenomenon f )
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Outline
Motivation
Concept of vessel extent
o Homogeneous reaction systems
o Incremental model identification
o Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
8
Homogeneous reaction systemsMole balance equations
9
( ) ( ) ( ) ( )( ) ( ) ( )T
0, 0outiv i mn
u tn tt t t t= + − =n n nr uN W n
( ) ( ) ( ) ( )( ) ( ) ( )
( )T , , m t tS w t V tt t V t tρ= = = nm 1 M n c
(S × R) (S × p)(R × 1) (p × 1)(S × 1)
Homogeneous reaction system consisting of S species,R independent reactions, p independent inlets and one outlet
Mole balances for S species
Mass, volume and concentrations
( ), in in tW u
( ) ( ), outt u tn
N ( )v tr0n
(S × 1)
Amrhein et al, AIChE J. 56 (2010) 2873
Homogeneous reaction systemsDecomposition in variants and invariants
o Condition:
o Vessel extents (variants) and q = S – R – p –1 redundant information (invariants)
10
( )T0rank 1in R p = + + N W n
( ) ( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( ) ( )( ) ( )( ) ( )
( ) ( ) ( ) ( )( ) ( )
1 1
T
T
T T T
T
, 0
, 0
1 , 0
0
out
out
out
R R p
p R p
R p
q p
o
R
ut
q
u tr v in in r rm t
u tin v in in in inm t
u tic v in in ic icm t
u tiv v in in ivm t
R
p
t t t t
t t t t
x t t t x t x
t t t t
×
×
× ×
× ×
= + − =
= + − =
= + − + =
= + −
I 0
0 I
0 0
0 0
x R N r R W u x x
x F N r F W u x x
q N r q W u
x Q N r Q W u
0
0
x
( ), 0 qiv =x 0
( )( )( )( )
( )( )T 0 r
in
ic
iv
tt tx tt
= −
x Rx F n nq
Qx
L
Amrhein et al, AIChE J. 56 (2010) 2873
Homogeneous reaction systemsDecomposition in variants and invariants
o Condition:
o Vessel extents (variants) and q = S – R – p –1 redundant information (invariants)
o Reconstruction:
11
( )( )( )( )
( )( )T 0 r
in
ic
iv
tt tx tt
= −
x Rx F n nq
Qx
L
( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( )( ) ( ) ( )
( ) ( )( )
( )( ) ( ) ( )
( )
, 0
,
0
0
, 0
out
ou
out
t
out
u tr rm t
u tin inm t
r
in
iu t
v R
in p
u tm t
q
ic icmc t
iv
t t t
t t t
t x t x
t
x
= − =
= − =
= − =
=
−
x x
x x
x
x u
0x
r 0
0
( )( ) ( ) ( ) ( )1
0T
0+ +inr in ict t txt−
− = x xN nn Wn L
( )T0rank 1in R p = + + N W n
Amrhein et al, AIChE J. 56 (2010) 2873
Homogeneous reaction systemsOrthogonal spaces in 4-way decomposition
12
TQ Q
inW F
TN R
space of inlet extentsspace of reaction extents
space of invariants
q
p
R
T0n q
space of initial conditions
1
S-dim space of the number of moles
Amrhein et al, AIChE J. 56 (2010) 2873
Homogeneous reaction systemsTransformation to RV form
o When ,
the numbers of moles are rearranged in Reaction Variant (RV) form:
o The R vessel extents of reaction are then computed as:
13
( ) ( ) ( )( )T0rank 1 and knownin oi tn uR p t u t < + + N W n u
( ) ( ) ( ) ( ) ( )( )TRV0 1i in in cr t xt t tt = = − − +N nx xn Wn
Amrhein et al, AIChE J. 56 (2010) 2873
( ) ( ) ( )T RVr t t
+= nx N
Outline
Motivation
Concept of vessel extent
Homogeneous reaction systems
o Incremental model identification
o Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
14
Kinetic investigation From measurements to rate expressions
15
① Simultaneous approach② Incremental approach (rate-based)③ Incremental approach (extent-based)
Bhatt et al, Chem. Eng. Sci. 8 (2012) 24
Incremental model identificationExtent-based method
o The kinetic problem is decomposed into sub-problems of lower complexity that are solved individually.
o The model identification proceeds in two steps:
o Transformat ion to extents (v+ iv)Computation of the contribution of each dynamic effect (reaction, inlets and outlets) as extents
o Model ident i f i ca t ion (Parameter estimation)Individual model identification of each effect from its corresponding extent with the integral methodof parameter estimation.
16Bhatt et al, Ind. & Eng. Chem. Res. 50 (2011) 12960
Extent-based model identificationModel identification and parameter estimation
A dynamic model is postulated for each extent of interestand a regression problem is solved individually using theintegral method of parameter estimation.
Example: fitting of R extents of reaction
17Bhatt et al, Ind. & Eng. Chem. Res. 50 (2011) 12960
Extent-based model identificationModel identification and parameter estimation
A dynamic model is postulated for each extent of interestand a regression problem is solved individually using theintegral method of parameter estimation.
Example: fitting of R extents of reaction
18
s.t. ( ) ( ) ( )( ) ( ), , , , ,
ˆ ˆ, , ,outu tr i r i v i r i r im tx t r t x t= −θ θ
, , ,L Ur i r i r i≤ ≤θ θ θ
( ),ˆ 0 0r ix =
1, , i R= ( ) ( ),
2
, , ,ˆmin ,r i
r i r i r ix t x t−θ
θ
Bhatt et al, Ind. & Eng. Chem. Res. 50 (2011) 12960
Homogeneous reaction systemsEthanolysis of phthalyl chloride in a CSTR
Ethanolysis of phthalyl chloride (A) comprising seven species(S = 7), three reactions (R = 3), two inlets (p = 2) and 1 outlet
19
( ), ,, in A in Aw u t
( ) ( ), outt u tn
N0n
( ), ,, in B in Bw u t
− − = − − − −
=
T,
,
1 1 1 1 0 0 0 0 1 1 1 1 0 0
0 1 0 1 0 1 1
0 0 0 0 0 00 0 0 0 0 0in A
inin B
ww
N
W
+ → ++ → ++ +
A B C DC B E DD B F G
( )v tr
Extents of reaction ?N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Homogeneous reaction systemsEthanolysis of phthalyl chloride in a CSTR
20
T
=
RFqQ
L
0, , inN W n
Number of moles
R = 3 Extents of reaction
p = 2 Extents of inlet
1 Extents of outlet
Each extent of reaction can then be modeledindividually, that is, independently fromall the other phenomena / extents…
Outline
Motivation
Concept of vessel extent
Homogeneous reaction systems
Incremental model identification
o Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
21
Fluid-Fluid reaction systemsMole balance equations
22
( ) ( ) ( ) ( ) ( )( ) ( ) ( ),
, ,T
, , 0,, 0out b
bv b b in bb mu t
b b bm tb in b bt t t t t= ± + − =N Wn nWr ζ nu n(Sb × Rb) (Sb × pm)(Rb × 1) (pm × 1)(Sb × 1)
Fluid-Fluid reaction system consisting of:
o Phase L: Sℓ species with Rℓ reactions, pm mass transfers, pℓ inlets and 1 outlet
o Phase G: Sg species with Rg reactions, pm mass transfers, pg inlets and 1 outlet
Mass transfer described by various models…
Mole balances on phase B
(Sb × 1)(Sb × pb)(pb × 1)
gp inlets in G Outlet from G
out,guin, in,, g gW u
lp inlets in L
in, in,, W u
Outlet from L
out,u
Phase G
Phase L
, , g g gm Vn
, , m Vn
, g± ζ ζMass transfer
m,
m,g
WW
Rb reactions, pm mass transfers, pb inlets, 1 outlet { } { }, , ,B L b∈ ∈ G g
Bhatt et al, Ind. & Eng. Chem. Res. 49 (2010) 7704
Fluid-Fluid reaction systemsDecomposition in variants and invariants
o Condition:
o Vessel extents (variants) and qb = Sb – Rb – pm – pb –1 invariants
( )T, 0, ,rank 1 b in b bm b bmb R pp = + + + N W nW
( ) ( ) ( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( ) ( ) ( )( ) ( ) ( ),
,
T, , ,
T, , , , , ,
, , ,
,
,
, 0
, 0
b
R R p R pb bout
m b b
m
p R p p pm b
out
b
b
b
m m
b
b
u tm
u tr b b b v b b m b
b b b v b b m b b b in b in b m b in bm
b b in b in b r b r bm
i
R
t
t
pt t t
t t t
t t
t t× ×
× ×
= ±
= ± + −
+ − =
=I 0 0
0 I 0
x R N r R W 0ζ
x M N r M W ζ M W u
R W u x
0
x
x
x
x
( ) ( ) ( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( ) ( ) ( )( ) ( )( ) ( )
1 1
,
1
T, , , , , ,
T T T T, , , , , , ,
, 0
1 , 00
b
p R pb p p bb m
R p pb m
out
b
ou
b
t b
b
u tn b b b v b b m b b b in in b in b in bm t
u tic b b b v b b m b b b in b in b ic b ic
p
bm t
t t t t t
x t t t t x t x× ×
× × ×
= ± + − =
= ± + − + =0 I0
0 0 0
F N r F W ζ F W u x x
q N r q W ζ q W u
0
( )( )( )( )( )
( )( )0,T
bm b
r b
b b bin
bic
biv
t
ttt
x tt
= −
x Mx R
F n nxqQx
L { },b∈ g
Fluid-Fluid reaction systemsDecomposition in variants and invariants
o Condition:
o Vessel extents (variants) and qb = Sb – Rb – pm – pb –1 invariants
( )T, 0, ,rank 1 b in b bm b bmb R pp = + + + N W nW
( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( )( ) ( ) ( )
( ) ( ) ( )( ) ( ) ( )
( ) ( )( )
( )( ) ( ) ( )
,
,
,
,
, , , ,
, , , ,
, , ,
, , ,, 0
, 0
0 0
, 0
,
out b
b
out
b
out b out b
out b
b
mb
b
b
b
u tr b v b r b r bm t
u tin b in b in b i
u tm b b
n bm tu t u t
ic b ic
m b in bm t
R
b ic bm t m t
p
p
t t t
t t t
x t x t x
t t t=
= − =
=
= −
−
=
= − − =
x ζ x
x r x x
x u x x
0
0x
0
( )( )( )( )( )
( )( )0, b
r b
in b b
m
b
bic
iv b
b
t
t tx t
t
t = −
x Mx R
x F n nqQx
L { },b∈ g
( )( ) ( ) ( ) ( ) ( )1
T,, , ,, , , 00 ,+b
m b m bb b r b in bb in ic bb bt t t t x t−
− = +±n W xx xN W nnL
( ),with biv b qt = 0x
Fluid-Fluid reaction systemsOrthogonal spaces in 5-way decomposition
25
Tb bQ Q
, bin bW F
TbbN R
space of inlet extentsspace of reaction extents
space of invariants
bq
bp
bR
T0, bbn q
space of initial conditions
1
Sb-dim space of the number of moles in phase B
, bm bW Mmp
space of mass-transfer extents
Bhatt et al, Ind. & Eng. Chem. Res. 49 (2010) 7704
Fluid-Fluid reaction systemsTransformation to RMV form
o When
, the numbers of moles are rearranged in
Reaction Mass-transfer Variant (RMV) form:
o The Rb vessel extents of reaction and pm extents of mass transfer
are then computed as:
26
( )T, 0, ,rank 1 b in b bm b bmb R pp < + + + N W nW
( ) ( ) ( ) ( )( )
( ) ( ) ( )( )
T T, ,
,
,, ,
,
, ,0
R
,
MV
1
r b
in b ic
m b m b
in b b
r b m b
b
bm b
b
tt
t tt
txt t
= ± = ±
= − − +
xx x
xN W N W
W nx
n
n
Bhatt et al, Ind. & Eng. Chem. Res. 49 (2010) 7704
( )( , andin b tu
( ) ), knownout bu t
( )( ) ( )T,
,
RMV, r b
mm b b
b
tt
t+
= ±
N Wx
nx
Fluid-Fluid reaction systemsChlorination of butanoic acid in a CSTR
Chlorination of butanoic acid comprises Sℓ = 5 (BA, MBA, DBA, Cl2, HCl) and Sg = 3 (Cl2, HCl, air) species, Rℓ = 2 reactions,pℓ = 1 and pg = 1 inlets and 2 outlets
27
in,HClW ( )out , ,u tn
Phase G
Phase L
2Clζ
Cl2, HCl, Air
BA, MBA, DBACl2, HCl
HClζin ,HClu
2in,ClW2in,Clu ( )out, ,g gu tn Gas-phase
measurements
Liquid-phasemeasurements
Extents of reaction ?
+ ⎯⎯→ ++ ⎯⎯→ +
2
2
R1
2 R2
cat
cat
BA Cl MBA HCl
BA Cl DBA HCl
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Fluid-Fluid reaction systemsChlorination of butanoic acid in a CSTR
28
T
=
RMfqQ
L
, , 0,, , , m inN W W nNumber of moles
in liquid phase
R = 2 Extents of reaction
pm = 2 Extents of m.t.
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Fluid-Fluid reaction systemsChlorination of butanoic acid in a CSTR
29
o Identification of the rate expression for the main reaction R1
o Rate expression candidates
o Identified rate expression
( )
( )
( )
( )
2
2
2
2
11 ,BA ,Cl
21 ,Cl
31 ,B
1
A ,Cl ,MBA4
1 ,BA ,Cl ,
1
1
1 MBA
r c cr cr c c cr c c c
kkkk
====
( )2
41 ,BA ,Cl ,MBA1. 3543r c c c=
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Fluid-Fluid reaction systemsChlorination of butanoic acid in a CSTR
o Identified rate expressions
o Results of curve fitting (2% noise level)
30
( )( )
2 2 22
2
2
2
1
2
Cl
HC
1 ,BA ,Cl ,MBA
2 1 ,Cl
*Cl ,Cl Cl ,Cl
*HCl ,HCl ,HCl H ll C
s w
s w
r c c cr r c
A V M c c
A
k
k V c
kk
M c
ζζ
=== −
= −
Parameter Simulated value Estimated value 95% Confidence interval
k1 1.3577 1.3543 [1.3207 – 1.3879]
k2 0.1 0.105 [0.0884 – 0.1216]
kCl20.666·10-4 0.594·10-4 [0.514·10-4 – 0.674·10-4]
kHCl 0.845·10-4 0.813·10-4 [0.763·10-4 – 0.863·10-4]
N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerland
Outline
Motivation
Concept of vessel extent
Homogeneous reaction systems
Incremental model identification
Fluid-Fluid reaction systems
o Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
31
Recent extensionsRank augmentation of conc. data by calorimetry
o Homogeneous reaction systems
o Fluid-Fluid reaction systems
32
( ) ( )( ) ( ) ( )RV
VT
T
R T
: :a aaug aug
rr r
rQ tt
t tt
= = = − Δ
n NNH
nx x
( ) ( )( )
( )( )
( )( )
T,T T
, ,
TRMV, , , , ,,
, ,
RMV : : augrm
a m a r ra
m mma
rug Q
tttt t
tt
= = = −Δ − Δ
N W x xnx x
n NH H
( ) ( ) ( )T RVau ugg ar tt
+= nx N
( )( ) ( ) ( )RMT
,, V
,aug a
rg
mu
tt
t+
=
nxx
N
Srinivasan et al, Chem. Eng. J. 207-208 (2012) 785
Recent extensionsConstruction of calibration models in spectroscopy
33Billeter et al, Anal. Chim. Acta 767 (2013) 21
Recent extensionsPrediction of concentrations from spectral data
34Billeter et al, Anal. Chim. Acta 767 (2013) 21
Outline
Motivation
Concept of vessel extent
Homogeneous reaction systems
Incremental model identification
Fluid-Fluid reaction systems
Extensions to calorimetry and spectroscopy
o Perspectives: distributed reaction systems
35
PerspectivesDistributed reaction systems
36
( )*(BC, ODE) g tc
( ) (BC, ODE)tn
( ) ( ) ( ) ( ) ( )( ) ( )( ) ( )
( )
2
2T
*
0PDE: , , , , 0, (IC,0 (BC)
, (
)
BC)
d ddt dx
gt
V t
t x t x t x x
t
xt t
δ
= + ==
= n
c N r D c c cc c
c
G LFilm F(1-dim)
( ),t xc
0 δ
Bonvin et al, TFMST, Lyon, July 13–16, 2013 (to be published)
How to decouple reaction and diffusion phenomena?
x
PerspectivesDistributed reaction systems
37
( )0(BC) x t=c ( ) (BC)x L t=c
( ) ( ) ( ) ( ) ( ) ( )( ) ( )( ) ( )
2
2 0
0
TPDE: , , , , , 0, (IC)
,0 (BC), (BC)
d d dx tdt dx d
x
x L
xt x t x v t x t x
t tt c t
x x
δ
=
=
=
= − +
==
=c N r c D c c
c cc
c
0 L
Bonvin et al, TFMST, Lyon, July 13–16, 2013 (to be published)
( ),t xc
How to decouple reaction, convection and diffusion phenomena?
1-dim
x
Laboratoire d’Automatique
38
Prof. Bonvin, Prof. Longchamp, Prof. Jones, Dr. MER Karimi6 Postdocs + 20 PhD students + technical / administrative staff
Thank you for your attention
39
Model reductiono M. Amrhein, PhD dissertation n°1861 (1998), EPFL, Switzerlando Bhatt et al, Comp. & Chem. Eng. (2011), submitted
Transformation to variants/invariants (extents)o N. Bhatt, PhD dissertation n°5028 (2011), EPFL, Switzerlando Amrhein et al, AIChE J. 56 (2010) 2873o Bonvin et al, TFMST, Lyon, July 13–16, 2013 (to be published)
Incremental model identificationo Bhatt et al, Ind. & Eng. Chem. Res. 49 (2010) 7704o Bhatt et al, Ind. & Eng. Chem. Res. 50 (2011) 12960o Bhatt et al, Chem. Eng. Sci. 8 (2012) 24
Rank augmentation by calorimetric datao Srinivasan et al, Chem. Eng. J. 207-208 (2012) 785
Incremental kinetic modeling of spectroscopic datao Billeter et al, Anal. Chim. Acta 767 (2013) 21