Effect of Fuel Composition on PM Emissions

LEV III WorkshopLEV III Workshop May 18th, 2010

J. Jetter, Honda R&D Americas, Inc.

LEV III Workshop May 18th, 2010 2 Overview

• Particulate Matter (PM) is a significant concern.

• Recent efforts to reduce PM focus on engines• Recent efforts to reduce PM focus on engines, not fuel. – Current LEV III proposal: 10mg 6mg 3mg/miCurrent LEV III proposal: 10mg 6mg 3mg/mi

• Honda theorized that better fuel specifications co ld red ce PMcould reduce PM. – This was confirmed by research and testing.

• Honda’s PM Index can lead to fuel specifications which reduce previously unaccounted variations in PM measurement

LEV III Workshop May 18th, 2010 3 Background

F l f l ti h f d ff t t il i i i f• Fuel formulation can have a profound effect upon tailpipe emissions of particulate matter (PM).

• This effect is not directly addressed in CaRFG3 market fuel specifications or in the proposed CaRFG3 certification fuel specifications, although some regulated properties (e.g. T90) do affect PM emissions.

• The fuel effect on PM is also not addressed in the current CaRFG3The fuel effect on PM is also not addressed in the current CaRFG3 predictive model.

Proposed CaRFG3 Certification Fuel Specifications

Pollutant Predictions Units

Oxides of Nitrogen (NOx) gm/mile

Exhaust Hydrocarbons (HC) gm/mile

Pollutants Included in Predictive Model

Evaporative Hydrocarbons (HC) Percent Change (Candidate Fuel Relative to Reference Fuel)

Exhaust Potency-Weighted Toxics (PWT) mg/mile

Evaporative Benzene mg/mile

Exhaust CO (Adjustment Factor for O )

gm/mile Oxygen)

Source: CARB’s “Preliminary Discussion Paper,” February 2010 Source: “California Procedures for Evaluating Alternative Specifications for Phase 3 Reformulated Gasoline using the California Predictive Model,” August 2008

LEV III Workshop May 18th, 2010 4 “PM Index” Concept

• Over the last year, Honda has been working on a model that can predict trends y , g pin tailpipe PM emission levels based on fuel properties. We call this model the “PM Index.”

• Overview of the development work:Overview of the development work:

• Pure hydrocarbons were added to an Indolene base fuel at nominally 10% by weight; i.e., each test fuel consisted of the base fuel spiked with one hydrocarbon.

PM emissions were measured over the NEDC cycle for each fuel PM number (in place• PM emissions were measured over the NEDC cycle for each fuel. PM number (in place of PM mass) was used as the metric to improve accuracy at lower PM emission levels.

• PM emissions were determined to primarily be a function of the DBE* and vapor f th f l tpressure of the fuel components:

C10 DBE 6

C10 DBE 7C10 DBE 5

C10 DBE 2

1.20E+12

1.60E+12

2.00E+12

m 9

0EM

) [#]

Alkene & AlkaneAromaNaphthalenes

∑ ⎟⎟⎠

⎞⎜⎜⎝

⎛×

+=

n

ii Wt

KPVDBE

IndexPM)443(

1

• This research, and subsequent test results, will be presented at SAE in October C12 DBE 0

C9 DBE 0

C9 DBE 5

C8 DBE 4

C9 DBE 4

0 00E+00

4.00E+11

8.00E+11P

N (I

ncre

ase

from

∑=

⎟⎠

⎜⎝i iKPV1 )443(.

will be presented at SAE in October.(2 papers)

C9 DBE 00.00E+00300 350 400 450 500 550

Chemical boiling point [K]

*DBE (Double Bond Equivalent) is essentially an indication of the number of double bonds or rings present in the molecule, and can be easily calculated from the molecular formula.

Example of test fuels, in this case arranged by BP, DBE, and PM number increase from the base fuel.

LEV III Workshop May 18th, 2010 5 Confirmatory Tests Performed by Honda

• EUDC cycle • Test fuels and market fuels

• 120 km/h steady-state • Test fuels and market fuels

RDX NEDC

2E+12

2.5E+12RDX 120km/h steady state

1.4E+13

1.6E+13

Test fuels and market fuels• Turbo-charged PFI vehicle

• Test fuels and market fuels• Turbo-charged PFI vehicle

5E+11

1E+12

1.5E+12

2E+12

PN [#]

4E+12

6E+12

8E+12

1E+13

1.2E+13

PN [#]

0

5E+11

0.00000 0.50000 1.00000 1.50000 2.00000PM INDEX

0

2E+12

0.00000 0.50000 1.00000 1.50000 2.00000PM INDEX

25.00

]

• Engine bench, 3000rpm @ full load • Market fuels • Normally-aspirated PFI engine

• PM Index correlated well with PM

R2 0 977410.00

15.00

20.00

conc

entra

tion

[mg/

m3

number and mass.

• The correlation held regardless of engine technology and cycle.

R = 0.9774

0.00

5.00

0 1 2 3 4 5

PM INDEX

Car

bon

g gy y

LEV III Workshop May 18th, 2010 6 Testing Performed at SwRI – Background

• Three test fuels were chosen, based on the range of PM Indices calculated for US k t f l Th F i b k d L V f l t i d f thUS market fuels. The Fairbanks and Las Vegas fuels were retrieved from the market in drum quantities.

EPA Certification Las Vegas Fairbanks

90% 100%

y Indolene

test fuel

test fuel (no PM Index spec)*

(top 5% of US PM Indices)

(bottom 5% of US PM Indices)

60% 70% 80%

Pro

babi

lit test fuel (sample, not

spec)

• The PM Index was calculated for >400 US

30% 40% 50%

umul

ativ

e

The PM Index was calculated for 400 US fuels, based on DHA analyses performed by Southwest Research Institute for Honda.

PM Index of CA market fuels:

0% 10% 20%

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Cu

Los Angeles San Francisco

PM Index of CA market fuels:

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5PM Index More tailpipe particulates (predicted)More tailpipe particulates (predicted)

* Since there is no PM Index specification, certification fuel from different suppliers, or a different batch from the same supplier, will have different characteristics and a different PM Index.

LEV III Workshop May 18th, 2010 7

• PM Index of the test fuels compared to the distribution of PM Indices worldwide.

Testing Performed at SwRI – Background

• Note that the US has the widest range of PM Indices, and also some of the highest values.

EPA Certification* Las Vegas Fairbanks

80%90%

100%

bilit

y

USA

50%60% 70% 80%

ive

Pro

bab

World wideIndia Japan EU

USA

30% 40% 50%

Cum

ulat

USA China

Russia Brazil

0% 10% 20%

5.00.0 1.0 2.0 3.0 4.0PM Index

* Since there is no PM Index specification, certification fuel from different suppliers, or a different batch from the same supplier, will have different characteristics and a different PM Index.

LEV III Workshop May 18th, 2010 8

• Factor of 4 difference in FTP PM mass emissions between Fairbanks

SwRI Testing – PM Mass Results

Factor of 4 difference in FTP PM mass emissions between Fairbanks fuel and Las Vegas fuel.

• Variations in the PM Index of the cert fuel can make the difference between a pass and a fail.between a pass and a fail.

910 LEV II (Current)

67 8 9

mg/

mi) Fairbanks EPA Cert Las Vegas

LEV III (2014) d

34 5 6

Mas

s (m LEV III (2014) proposed

LEV III (2017)

01 2 3

PM LEV III (2017) proposed

• Test vehicle: 2009 2.0L GDI, LEV2 ULEV (not a Honda product)

0 FTP

LEV III Workshop May 18th, 2010 9

• FTP PM number emissions from Las Vegas fuel >3x higher than

SwRI Testing – PM Number Results

FTP PM number emissions from Las Vegas fuel >3x higher than those from Fairbanks fuel.

• Variations in the PM Index of the cert fuel can make the difference between a pass and a fail.between a pass and a fail.

910

12)

Fairbanks EPA Cert Las Vegas

67 8 9

rt./m

i x 1

01

LEV III (2014) d

34 5 6

mbe

r (p

a LEV III (2014) proposed

LEV III (2017)

01 2 3

PM N

um LEV III (2017) proposed

0 FTP

• Test vehicle: 2009 2.0L GDI, LEV2 ULEV (not a Honda product)

LEV III Workshop May 18th, 2010 10 SwRI Testing – Correlation with PM Index

PM Mass

3.0

m)

PM Mass Measurement vs. PM Index

Total PM from filter

R² = 0.9988 2.0

2.5

ons

(mg/

km

Total PM from filter(40 CFR Part 1065)

Solid PM only (from AVL MSS)

R² = 0.9998 1.0

1.5

M E

mis

sio ( )

0.0

0.5

0 1 2 3

FTP

PM

Fairbanks fuel

Las Vegas fuel

EPA Cert fuel

PM Index

( )• Excellent correlation between PM Index and actual (measured) PM mass emissions.

LEV III Workshop May 18th, 2010 11 SwRI Testing – Correlation with PM Index

PM Number

12

12)

PM Number Measurement vs. PM Index

US06

R² = 0.99848

10

rt./k

m x

101

R 0.9984

4

6

ssio

ns (p

ar FTP

• Results shown are for solid PM > 23nm. R² = 0.9795

0

2

PM E

mis (EU PMP-type measurement.)

Fairbanks fuel

Las Vegas fuel

EPA Cert fuel0

0 1 2 3 PN Index

• Good correlation also achieved between PM Index and actual (measured) PM number emissions.

LEV III Workshop May 18th, 2010 12 Summary & Recommendations

• Fuel formulation can have a significant effect upon tailpipe emissions of particulate matter (PM).

• The “PM Index” provides a method by which tailpipe PM emission trends can be predicted, based on fuel composition.

Recommendations:

• Control the Certification Fuel: • The CaRFG3 certification fuel specification should include the

PM Index (or similar).

• Control the Market Fuel:• Control the Market Fuel:• Market fuel specifications (including the CaRFG3 predictive

model) should also include the PM Index. Improvements to market fuel specs will affect real-world PM emissionsmarket fuel specs will affect real-world PM emissions.

Backup Slides

14

X9V (GDI PM)AR Report

March 9th, 2010 PM Mass – Cumulative Emissions FTP

Las Vegas fuel

Fairbanks fuel

EPA Cert fuel

15

X9V (GDI PM)AR Report

March 9th, 2010 Solid PM Number – Cumulative Emissions FTP

LEV III proposal (2014) Las Vegas fuel

EPA Cert fuel

Fairbanks fuelFairbanks fuel

LEV III proposal (2017)

• Control of PM emissions throughout Phase 1 of the FTP (not just the cold-start period) is critical.

LEV III Workshop May 18th, 2010 16 Other Test Fuel Properties

FairbanksDistillation Curve Carbon Number DistributionEPA CertLas Vegas

• DHA analysis indicated th t th t j it fthat the vast majority of HCs >C9 were aromatic.

• The T90 point does not predict the fact that the PM emissions

C ti l i d f PM i i Th L V f l

were relatively close for the Fairbanks and EPA Cert fuels.

Chemical Composition • Conventional wisdom for PM emissions: The Las Vegas fuel

would be relatively low (10% ethanol content and moderate aromatic level) and the Fairbanks fuel relatively high (high aromatics). Actual results were the complete opposite.

• It is difficult to judge the PM-producing potential based only on these traditional properties; assumptions are conflictingproperties; assumptions are conflicting.

• Honda’s “PM Index” is a better predictor.