109/11/2019 MODEGAT 2019

GPF modeling supported bysize-resolved filtration efficiency measurements

Michail Mitsouridis, Grigorios Koltsakis, Zissis SamarasAristotle University ThessalonikiJeremy Gidney, John GoodwinJohnson MattheyChristian MartinAVL List GmbH

• Target: Particle free gasoline engines with cost-effective GPFsHow can we use GPF modeling to support system optimization?

• Background: GPF modeling shares many common features and challenges as the long established DPF modeling.However, there are some important differences:– Lower PM/PN emissions -> operation in depth filtration

– Lack of O2 and NO2 for thermal regeneration

– Higher exhaust gas temperatures -> thermal shock risks

– How to combine TWC functions?

– ...

2

Background & motivation

09/11/2019 MODEGAT 2019

3

Spherical collector filtration model

09/11/2019 MODEGAT 2019

dgrain =3

2

1−εpore

εporedpore

Agglomerate structure = f(fpart, Rppart, ρppart, Rmc)

• fpart: Particle fractal dimension [-] → agglomerate geometry standardization

• Rppart: Primary particle radius [m]

• ρppart: Primary particle density [kg/m3]

• Rmc: Aggregate cluster mobility radius [m] → the diameter of a spherical particle with the same mobility (ease to have constant motion) as the agglomerate in question

porous wall unit cell

deposit layer

washcoat

grainEwall = 1 − e−ηg

3

2

1−εpore,0

εpore,0

Δw

dgrain,0 , Δw: node thickness

ηg = ηg,0 + 1 − ηg,0 erf3nload

2Wcap , Wcap=

εpore,0 − εpore

εpore,0

ηg,0 = ηD + ηR

ηD = nclean,D ∙ 3.5εpore,0

Ku

1

3PenPe , Pe =

uw∙dgrain,0

Dpart

ηR = nclean,R ∙ 1.5εpore,0

Ku

R2

1+R m, m =3−2∙εpore,0

3∙εpore,0, R =

2Rc

dgrain,0

dpore

dgrain

4

Filtration model calibration

09/11/2019 MODEGAT 2019

The size-resolved filtration efficiency is well captured by the model for particles > 30 nm.

Uncoated GPF Coated GPF: 50 g/l

nclean,D = 0.35

nclean,R = 5

Particulate Filter Test System• 250 kg/h, 240°C

• GPF cell structure: 12mils, 300cpsi

• Calibration of:

– nclean,D: Diffusion mechanism correction factor

– nclean,R: Interception mechanism correction factor

nclean,D = 0.25

nclean,R = 4.5

5

Model application & validation with thinner wall filters

09/11/2019 MODEGAT 2019

Uncoated GPFs Coated GPFs: 50 g/l

nclean,D = 0.35

nclean,R = 5nclean,D = 0.25

nclean,R = 4.5

The filtration model is predictive for filters of different wall thickness.

This coating does not have a major impact on filtration efficiency (FE).

Model parameters: based on the calibration with the 12/300 GPF

6

Effect of soot loading

09/11/2019 MODEGAT 2019

nload = 4.5 nload = 8

The model captures reasonably well the filtration efficiency increase due to progressive soot accumulation.

12mils, 300cpsi

Uncoated GPF Coated GPF: 50 g/l

7

Total PN filtration efficiency vs soot loading

09/11/2019 MODEGAT 2019

Increasingwall thickness

Increasingwall thickness

The filtration model predicts the rate of FE increase as function of collected soot amount.

Uncoated GPFs Coated GPFs: 50 g/l

8

Filtration efficiency prediction in real exhaust

09/11/2019 MODEGAT 2019

CC: 50 g/l UF: 100 g/l

Engine sweep tests0-700kg/h

T=300-800°C

2.0 l TGDIengine test bed UF: 100 g/l

CC: 50 g/l

The filtration model performs well in the whole engine map.

Engine sweep tests with two exhaust line configurations.

Initial soot loading: 0 g/l

9

Towards predictive filtration modeling

The available database was used to derive correlation functions for the filtration model

tuning parameters

09/11/2019 MODEGAT 2019

10

Pressure drop modelingSoot-free filters

09/11/2019 MODEGAT 2019

2.2 l Diesel engine(Daimler OM646)

UncoatedEffect of wall thickness

CoatedEffect of washcoat loading on wall permeability

dP

dw=

ሻμ ∙ v(w

kw

kw = kw,0 ⋅ 1 + C4p0pμ

T

Mg

kw,0: Clean wall permeabilityC4: Slip correction factor

Same kw,0 Increasing kw,0

11

Pressure drop modelingSoot loaded filters

09/11/2019 MODEGAT 2019

The effect of soot in the wall on deltaP is strongly affected by the microstructure after coating.

The empirical model may have a compromised accuracy in the transition region.

dP

dw=

ሻμ ∙ v(w

kw, kw =

1

1kw,0

+ C1 ⋅ ρP + C2 ⋅ ρP2⋅ 1 + C4

p0pμ

T

Mg

dP

dw=

ሻμ ∙ v(w

kp, kp=

ρp,0 ∗ k0

ρp1 + C4

p0pμ

T

Mg

kw: Soot loaded wall permeabilityρp,l: Reference wall soot loading

ρp,0: Soot reference density

k0: Soot permeability

12

Towards predictive pressure drop modeling

Correlations of wall permeability for clean and loaded wall state were derived

for a range of 0 to 500 g/l wall.

09/11/2019 MODEGAT 2019

Reference wall soot loading: 2.5 g/lwall

Is it possible to avoid filter permeability recalibration for each new filter configuration, i.e. various washcoat loadings (grams per liter of filter), cell- and wall micro-structures?

13

Model application in RDE cycle conditions

(100g/l w/c loading)

UF: 100 g/l

CC: 50 g/l

(50g/l w/c loading)

1.8l TGDI enginechassis dynamometer

RDE protocol

• During the RDE test, soot oxidation may take place during fuel cut-off events and high temperatures.

• The UF: 100 g/l has a lower regeneration potential compared to the CC: 50 g/l filter.

13MODEGAT 201909/11/2019

14MODEGAT 201909/11/2019

Combined model validationLoss of filtration efficiency due to wall cleaning

CC: 50 g/l washcoat

PN slip: 14%

0.18

RDE high temperature cycle with preloaded GPF (=0.4 g/l)

PN slip (normalized over total emitted engine out PN) appears after soot

loading goes below ~0.2 g/l

Due to high temperature and frequent fuel cutoffs, soot is

oxidized (simulation result only)

The GPF combined model (filtration, transport, reaction phenomena) manages to predict the observed PN slip

15MODEGAT 201909/11/2019

withfuel cut-off events

withoutfuel cut-off events

To validate that the PN slip is associated with wall cleaning, the test was repeated de-activating fuel cut-off events.

Deactivation of fuel-cut off mode drastically reduces the PN slip.The model predictions are in line with the measured observations.

Verification of modelDeactivation of fuel cutoff

CC: 50 g/l washcoat

PN slip: 2%

PN slip: 14%

16MODEGAT 201909/11/2019

Model application exampleUnderfloor vs close-coupled positioning

CC: 50 g/l washcoat UF: 100 g/l washcoat

0.18

Same RDE cycle run with UF and higher w/c loading filter. Initial soot loadings: CC filter: 0.4 g/l, UF filter: 0.1 g/l

PN slip: 14%

UF filter exposed to lower temperature -> less soot is oxidizedThe remaining soot in the UF filter is sufficient to maintain very low PN slip

17

Pressure drop prediction accuracy in RDE cycle

Good prediction accuracy under complex operating conditions

09/11/2019 MODEGAT 2019

CC: 50 g/l washcoat UF: 100 g/l washcoat

Data zoom in high flow rate operating mode

18

Ash transport testing & modeling

09/11/2019 MODEGAT 2019

Ash & soot loading250°C187 kg/h

Layer ashdetachment

Redeposition aslayer ash

Deposition asplug ash

SootAsh

Testing Modeling in axisuite(previously parameterized)

Oven regeneration• 650°C in air• 187kg/h• 3 hours

19

Ash loaded model validation

09/11/2019 MODEGAT 2019

UF: 50 g/l

Engine sweep tests with filters loaded at various ash levels

The model may be used for lifecycle analysis for various filter designs and driving profiles

20

Ash model validation in RDE transient cycles

09/11/2019 MODEGAT 2019

23g/l ash

42g/l ash

0g/l ash

UF: 50 g/l

• In view of the increasing gasoline after-treatment requirements and complexity, a practical/balanced modeling approach was developed.

• Based on a small database of calibrated models, correlation functions were developed to support permeability and filtration modeling for various washcoat loadings, cell- and wall micro-structures.

• Due to the low engine-out emissions, it is important to maintain sufficient soot amount in the filter to reduce the risk of PN slip. Regenerations during fuel cutoff are of key importance.

• Inclusion of ash transport modeling allows a life cycle analysis and assessment of GPF designs and application-specific model-based optimization.

• Current-future work: – Improve filtration efficiency predictions for very small particles

– Advanced filtration models accounting for pore size distribution

– Model application in non-uniformly coated substrates: zoning & layering

21

Summary & conclusions

09/11/2019 MODEGAT 2019

The results were obtained within the project UPGRADE. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 724036.

22

Acknowledgements

09/11/2019 MODEGAT 2019

We gratefully acknowledge:- the financial support of EU - the great work of our colleagues at LAT/AUTh, AVL and JM involved in the measurements. - The valuable support of NGK Europe GmbH providing many of the test samples used in

this work

Thank you very much!

09/11/2019 23SIA Powertrain, Rouen 2014

MODEGAT 2019

Michail Mitsouridis

mmitsour@auth.gr