Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
1
Diesel Aftertreatment System Analyses with DOC and SCR Models
Santhoji KatareChemical Engineering Department
Ford Research & Advanced EngineeringDearborn, MI
10th CLEERS workshop, May 1-3, 2007Dearborn, MI
2
Acknowledgements
Diesel aftertreatment modeling
Jeong Kim
Paul Laing
Joseph Patterson
Diesel catalysis
Giovanni Cavataio
Douglas Dobson
Christine Lambert
Cliff Montreuil
Scott WilliamsAdvanced Diesel program
Michael Goebelbecker
Kevin Murphy
Peter Nelson
3
urea
DOC
NOx sensor
SCR
DOCDOC
CDPF
Spray Target
to engine intake
exhaustflow
Recent progress & key challenges
C. Lambert et al., DOE funded project, 2001 – 2005
• Typical AT system for lean NOx control: DOC + SCR + CDPF• Several challenges but good progress• Models facilitated progress• Current modeling efforts to support vehicle programs
Low exhaust THigh NOx
NH3 slip NOx remake
Non-uniform ureadistribution
AgingHC & S poisoning
CDPF: Catalyzed Diesel Particulate Filter
DOC : Oxidize HC & CO; Generate NO2SCR : Reduce NOx (using NH3 + NO2)CDPF: Control PM
4
Outline
Diesel Oxidation Catalyst (DOC) model• Model calibration• Prediction of post-DOC NO2 on a vehicle
Selective Catalytic Reduction (SCR) model• Model calibration• Effect of stored NH3 on SCR performance• Prediction of tailpipe NOx and NH3 slip in a vehicle
DOC+SCR system applications• Importance of post-DOC NO2 on SCR performance• Optimization of urea injection strategy
5
hA[Tgas-Twall]
kA([ ]gas-[ ]wall)
Conduction along substrate
Surface reaction rate, R(from empirical map)
hA(Twall-Tamb)
ΟΟΟΟ2222, , , , HC, CO, PM,NO, NO2 m, Tgas
Heat Transferto Atmosphere
SIMTWC – Heat transfer equations + catalyst conversion data
Computationally non-intensive semi-empirical modeling approach
Laing et al., SAE 1999-01-3476
6
DOC model – Hybrid approach to leverage data
“Information hooks”
Conversion map(Data)
HC storage/release(PDEs)
Rate inhibition(Correlations)
Thermal balances(PDEs)
DOC ModelPredict
vehicle data
X
Tcat
Monolith channel
Optimizer(Agent-based)
Katare & Laing, SAE 2006-01-0689
7
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
150 200 250 300 350 400 450 500 550 600
Inlet Temperature [C]
Con
vers
ion
[-]
Typical pulsator map
SV: 35k 1/h Degreened DOC
COHC
NO NO2
Key input to the DOC model
8 Goralski et al., SAE-2000-01-0654
2 nonlinear equations & 5 parameters
)exp( RTEkk ddod −=
oa Nk θ,,
θθθdsa kCk
dtd −−= )1(
( ) ( ) 01 =+−−−= NkNCkCCSkdt
dCdsasfc
s θθ
At the bulk-solid interface
In the zeolite phaseAdsorption
Desorption
HC trap model
9
HC (-)
Tf, in (C)
ka 1.06 x 104 m3/mol/skdo 3.7 x 102 1/sEd 9.26 kcal/molN 1.4 mol/kgθθθθo 0.32 --
DesorptionLight-off
Model can predict HC adsorption, desorption and light-off
θθ θθ HC
[-]
Tf, in [C]
L/20
L/2
L
L: Brick length
θθ θθ HC
[-]
Tf, in [C]
L/20
L/2
L
L: Brick length
Total amount of the trap material (N) could be optimized
Parameter estimationusing in-house
“stochastic optimizer” *
HC
con
vers
ion
HC
sur
face
cov
erag
e
Line: Model; Dots: Data
Mid brick
End brick
* Katare & West, Complexity, 11, 4, 2006
10
Tail pipe
Model
Engine out
Engine outDataModel
Euro test time (s)
CO
HC
DataModel
Euro test time (s)
Predicting engine dynamometer CO and HC emissions
Model predictions reasonable
2.2 L Puma enginePost DOC
Post DOC
11
NO oxidation chemistry is complex
DOC oxidizes NO to NO2 – critical for SCR operation
0.00
0.20
0.40
0.60
0.80
1.00
100 150 200 250 300 350 400 450 500 550
Catalyst T [C]
NO
2/N
Ox
[-]
Eqmno H2O
4.5% H2O
Inhibition0.2% CO + 0.2% HC
aged with 300 ppm S only
aged with 300 ppm S + 15 mg P/gal
P
sootPos
t DO
C N
O2/
NO
x
Feed: 400ppm NO + 8% O2
12
0.0
0.2
0.4
0.6
0.8
1.0
0 100 200 300 400 500 600
Predicting post DOC NO2 on a Ford developmental vehicle
Engine exhaust flow and T
Map
F250 FTP
Model can predict DOC-out NO2:
Useful information for the downstream SCR
NO NO2
COHC
X
FG NOx
NO2 model
NO2 vehicle data
Inlet T [C]
T DOC in
FTP time [s]
DO
C in
let T
[C]
Nor
mal
ized
NO
xan
d N
O2
0.0
0.2
0.4
0.6
0.8
1.0
0 200 400 600 800 1000 1200 1400 1600 1800 20000
100
200
300
400
120k miles aged DOC
13
Outline
Diesel Oxidation Catalyst (DOC) model• Model calibration• Prediction of post-DOC NO2 on a vehicle
Selective Catalytic Reduction (SCR) model• Model calibration• Effect of stored NH3 on SCR performance• Prediction of tailpipe NOx and NH3 slip in a vehicle
DOC+SCR system applications• Importance of post-DOC NO2 on SCR performance• Optimization of urea injection strategy
14
Introduction to Selective Catalytic Reduction (SCR)
• Base metals such as Cu and Fe on honeycombs• Selectively reduce NOx to N2• Aqueous urea sprayed onto the catalyst• Urea thermal decomposition + hydrolysis to NH3 on the catalyst
When feed NO = NO2 SCR reaction is faster
OHNONONH 2223 6444 +→++ Standard SCR
OHNNONONH 2223 64224 +→++ Fast SCR
NOx
NH3
urea N2O
Undesired products
Desired product
15
SCR NOx performance increases with NO2 in the feed
NO2 information from DOC model is key for simulating SCR performance
DF242 - 64hr @ 670C - SV = 30k/hr - Nom. 350ppm NOx/350ppm NH3
0
10
20
30
40
50
60
70
80
90
100
100 150 200 250 300 350 400 450 500 550 600 650
Catalyst Temperature (ºC)
Gro
ss N
Ox
conv
ersi
on (%
)
100% NO
80% NO - 20% NO2
50% NO - 50% NO2
SV: 30k 1/h120k miles aged SCR
Catalyst saturated with NH3before measuring NOx conversion
16
Temperature programmed desorption of NH3
ka 0.61 m3 / mol skdo 2 x 105 1/sEdo 20 kcal/molαααα 0.45 -ΩΩΩΩ 0.23 mol NH3 / kg wc
ModelData
Initial T: 150°C120k miles aged SCR
NH3 on NH3 off
Temkin isotherm captures NH3 ads/des data reasonablyModel calibrated at 150°C and verified at other T and CNH3
17
Calibrated model over predicted vehicle NOx conversion
DF242 - 64hr @ 670C - SV = 30k/hr - Nom. 350ppm NOx/350ppm NH3
0
10
20
30
40
50
60
70
80
90
100
100 150 200 250 300 350 400 450 500 550 600 650
Catalyst Temperature (ºC)
Gro
ss N
Ox
conv
ersi
on (%
)
100% NO
80% NO - 20% NO2
50% NO - 50% NO2
Thermal balances(PDEs)
SCR ModelOver predicted
vehicle NOx conversionMonolith channel
Optimizer(Agent-based)
Model showed the right trends but had to assume(1) low urea to NH3 conversion or (2)very high HC deactivation to explain vehicle data
Model
Data
T
18
0
10
20
30
40
50
60
70
80
90
100
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Cumulative NH3 exposed, g/L
% N
Ox
Con
vers
ion
103 C
151 C
198 C244 C
341 C
NOx performance increases with increasing NH3 storage
Currently a standard protocol for screening catalysts
Previous data collectedat this saturation limit
19
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0 200 400 600 800 1000 1200 1400 1600 1800 2000
120k FTP time [s]
TP N
Ox
[g/m
i]
0
50
100
150
200
250
300
350
400
T S
CR
in [C
]
!" Model TP NOx
Vehicle data TP NOx
120k miles aged SCR
Model predicts vehicle Tail Pipe (TP) NOx
20
0
10
20
30
40
50
60
70
80
90
100
0 200 400 600 800 1000 1200 1400 1600 1800 2000
120k FTP time [s]
NH
3 sl
ip [p
pm]
Model Vehicle data
120k miles aged SCR
NH3 slip prediction
Model – experiment closure: Model Experiment Data More accurate model
21
Outline
Diesel Oxidation Catalyst (DOC) model• Model calibration• Prediction of post-DOC NO2 on a vehicle
Selective Catalytic Reduction (SCR) model• Model calibration• Effect of stored NH3 on SCR performance• Prediction of tailpipe NOx and NH3 slip in a vehicle
DOC+SCR system applications• Importance of post-DOC NO2 on SCR performance• Optimization of urea injection strategy
22
System: DOC + SCR Models
Predict XNOx & NH3 slip
HC storage/release(PDEs)
Rate inhibition(Correlations)
Thermal balances(PDEs)
DOC Model
Conversion map(Data)
X
Tcat
Monolith channelOptimizer
(Agent-based)
Conversion map(Data)
Thermal balances(PDEs)
SCR Model
Monolith channel
X
Tcat
DOC Model
SCR Model
T & NO2/NOx
23
Optimum DOC for an efficient SCR ?
80%
82%
84%
86%
88%
90%
92%
94%
96%
0.0 0.1 0.2 0.3 0.4 0.5 0.6
NO2/NOx (First 200s)
% S
CR
NO
x C
onve
rsio
n
Very little NO2 produced by current DOC formulations
Ideal DOCNO = NO2
100% ESV
200% ESV
120k miles aged DOC+SCR
DOEvehicle
DOC
50% ESV
Post DOC NO2/NOx (First 200s of an FTP)
% S
CR
NO
xco
nver
sion
Increasing DOC volume increases NO2But SCR performance also requires- DOC with low thermal inertia- Rapid warm up
(tradeoff - HC v.s. DOC NO2 v.s. SCR XNOx)
Model provided insight into system design issuesKatare et al., Ind. Eng. Chem. Res., 46, 2007, 2445
24
Urea injection strategy as a function of vehicle speed over an FTP test
NO
xco
nver
sion
[%]
Normalized NH3 slip [-]
Post DPF
Post SCR
Region 1Very low
urea injection
Region 2Optimum!
Region 3Limited by SV & T
SCR DPF
Assuming 100% NH3 to NOxconversion
over DPF
Each point corresponds to a simulatedurea injection strategy that max [XNOx/NH3 slip]
Possible to use models to come up with effective urea injection strategies
25
Summary
Math models enable effective analysis and afford insight into diesel AT systems
Models leverage laboratory data to predict vehicle experiments
Optimization & statistical tools enhance model development process
DOC modelPredicted post DOC NO2 on a vehicleExtrapolated fresh catalyst information to predict aged catalyst performance
SCR modelMotivated the need for analyzing SCR performance = f (NH3 storage)Predicted vehicle NOx conversion and NH3 slip
DOC+SCR system analysisHighlighted importance of NO2/NOx from DOC for SCR performanceOptimization of the urea injection strategy (trade-off between NOx conversion & NH3 slip)
26
Models
Exp
erim
ents
Adapted from “Drawing Hands”, M C Escher, 1948
Experiments and models drive each other leading to insight
27
Model calibration with NH3 storage information required
NH3 storage decreases with increasing temperature
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 100 200 300 400 500 600
Catalyst Temperature [C]
NH
3 st
ored
[g N
H3
/ L S
CR
]
Increasing T reduces NH3 storage ability
More accurate picture of the catalystCatalyst as a storage cup
28
Stochastic algorithms for parameter estimation & optimization
• Do not get trapped in local optima – more robust and efficient than a local optimizer• Examples: Genetic algorithms, Particle swarm optimization, Differential evolution
Fitness Calculation, Parent Selection
Fitn
ess
OperatorsNew
Population
“ Survival of the fittest ”
Evolution
Initial Population (random)
Katare & West, Complexity, 11, 4, 2006
In-house built optimizer – can be customized
AT models are nonlinear and multi-dimensionalLocal optimizers may not be efficient
29
1
50
15
180
Optimizer Iterations
Katare, SAE 2006-01-0690
DB
CBAk
kk
→
→→3
21
30
Practical utility of the model: Case study
Aged DOC inlet T for exotherm generation given fresh DOC data ?
Burn soot off DPF by down stream injection of fuel (HC)
Tin Tout
Fuel injector
Fresh DOC
Engine
HC slip
Ford vehicle
Performance: (Tout – Tin) / HC slip v.s. (Tin, SV)
31
0.0
0.2
0.4
0.6
0.8
1.0
0 100 200 300 400 500
Catalyst T [C]
HC
co
nve
rsio
n [
-]
0
50
100
150
200
250
300
0 50 100 150 200 250 300DSI Data exotherm [C]
Mo
del
exo
ther
m [
C]
Calibrating model to fresh DOC data using pulsator map
Model calibrated with “fresh” DOC dataPredict “aged” DOC performance
120k milesaged DOC
Fresh DOC(Tuned)
Fresh DOC(Pulsator)
Fresh DOC vehicle data at different Tin
32
0
50
100
150
200
250
300
350
400
450
100000 150000 200000 250000 300000 350000
Space velocity [1/h]
Extrapolating model to predict performance of aged DOC
Aged catalyst can not generate exotherm
when Tin < 300 C
254 C
329 C
305 C
417 C
324 C 367 C331 C
310 C
Tin
Model established limits to exotherm generation strategy
(Tou
t –Ti
n) /
HC
slip Model – Fresh DOC
Data – Fresh DOC
Model – Aged DOC