diesel aftertreatment system analyses with doc and scr models

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




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


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




22

System: DOC + SCR Models

Predict XNOx & NH3 slip

HC storage/release(PDEs)

Rate inhibition(Correlations)


DOC Model


X

Tcat

Monolith channelOptimizer

(Agent-based)



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

diesel aftertreatment system analyses with doc and scr models

Documents