EPAct/V2/E-89: Assessing the Effect
of Five Gasoline Properties on Exhaust
Emissions from Light-Duty Vehicles
Certified to Tier 2 Standards
Final Report on Program Design and
Data Collection
&EPA
United States
Environmental Protection
Agency
-------
EPAct/V2/E-89: Assessing the Effect
of Five Gasoline Properties on Exhaust
Emissions from Light-Duty Vehicles
Certified to Tier 2 Standards
Final Report on Program Design and
Data Collection
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
and
National Renewable Energy Laboratory
U.S. Department of Energy
and
Coordinating Research Council
NOTICE
This technical report does not necessarily represent final EPA decisions or
positions. It is intended to present technical analysis of issues using data
that are currently available. The purpose in the release of such reports is to
facilitate the exchange of technical information and to inform the public of
technical developments.
United States
Environmental Protection
Agency
EPA-420-R-13-004
April 2013
-------
CONTENTS
T
TT
TTT
TV
V
VT
VTT
Executive Summary
Introduction
Test Fuel Specification and Blending
Test Vehicle Fleet Sizing and Selection
Test Procedures, Data Collection, Issues Encountered
Data Reporting
Summary and Closing Remarks
5
6
8
40
49
73
76
APPENDICES
A. Re-Design of Fuel Matrices for EPAct Program
B. SwRI EPAct/V2/E-89 Fuel Blending Report
C. Fuel Round Robin Procedure for Sample Handling
D. Identification of Extreme Values for Round Robin Laboratory Tests
E. Spreadsheets of Detailed Round Robin Fuel Inspection Data
F. Summary of Oil Sampling Dates and Mileages
G. Fuel Sampling Procedure for Carryover Experiments
H. Fuel Blending Experiment to Characterize Carryover Effects
I. Additional Fuel Carryover Experiments
J. Refueling Location Experiments
K. Detailed Measurement and Analysis Methods
L. LOD/LOQ Method
M. Details of Ford Explorer Oil Level Incident
N. EPAct NMOG Calculation Protocol
O. Text of SwRI Phase 3 Testing Report
P. Derivation of FID Response Factors for Oxygenated Species
-------
Abbreviations Used in this Report
API American Petroleum Institute
CARB California Air Resources Board (also ARE)
CFR Code of federal regulations
CH4 Methane
CO Carbon monoxide
CC>2 Carbon dioxide
CRC Coordinating Research Council
EPA US Environmental Protection Agency
DOE US Department of Energy
DNPH Dinitrophenylhydrazine
DVPE Dry vapor pressure equivalent (equivalent to RVP)
EO Gasoline containing no ethanol
E10 Gasoline blend containing 10 vol% ethanol
FBP Final boiling point
FID Flame ionization detector
FTP Federal test procedure
HPLC High performance liquid chromatograph
MON Motor octane number
MSAT Mobile Source Air Toxics
NMHC Non-methane hydrocarbons
NMOG Non-methane organic gases
NOX Nitrogen oxides
NREL National Renewable Energy Laboratory
OBD Onboard diagnostics
(R+M)/2 Average of RON and MON
RON Research octane number
RVP Reid vapor pressure
SwRI Southwest Research Institute
T50 Temperature at which 50 vol% of a fuel has been evaporated (°F)
T90 Temperature at which 90 vol% of a fuel has been evaporated (°F)
THC Total hydrocarbons
-------
I. EXECUTIVE SUMMARY
This report describes program design and data collection in the EPAct/V2/E-89 light duty
gasoline vehicle fuel effects study, which examined the exhaust emission impacts of changes in
five fuel properties (ethanol, T50, T90, aromatics, and RVP (specified as DVPE)) over a range
covering current market fuels and potential mid-level ethanol blends. Testing was performed by
Southwest Research Institute, and program sponsors were the U.S. Environmental Protection
Agency, U.S. Department of Energy via the National Renewable Energy Laboratory, the
Coordinating Research Council, and the Lubrizol Corporation.
The fuel matrix consisted of 27 fuels arranged according to a partial factorial design
optimized for selected interactions of interest, plus an E85 fuel tested on a subset of vehicles.
The test vehicle fleet consisted of 15 new light duty cars and trucks of 2008 model year selected
from among high sales makes and models to provide a representative sample of the fleet of
properly-operating vehicles meeting the U.S. Federal Tier 2 emission standards.l Given the
relatively low level of emissions from these vehicles, a number of design and procedural steps
were undertaken to minimize the impacts of measurement variability and other artifacts on data
quality.
Data collected included typical regulated pollutants by bag and second-by-second plus
speciated emissions for a subset of tests and bags. Data for 16 relevant onboard diagnostic
parameters were also captured for each test on a second-by-second basis. Complete data were
generated for 926 tests, with 30 additional tests containing valid measurements for regulated
r\
emissions. Sponsors agreed that analysis of the resulting dataset would be performed and
published independently.
1 An additional (16th) vehicle performed testing on the E85 fuel.
2 Program data is available via the EPA OTAQ website at http://www.epa.gov/otaq/fuelsmodel.htm
-------
II. INTRODUCTION
Prior fuel effects models, such as the EPA Predictive Model and Complex Model,3 were
based on data taken on 1990s-technology vehicles meeting the Tier 0 and Tier 1 emission
standards, levels an order of magnitude higher than current (Tier 2-compliant) vehicles. With the
fleet turning over to much lower-emitting vehicles, the Agency and stakeholders were interested
in generating a coherent body of updated fuel effects data on which policy could be based going
forward. Recognizing this issue, Congress, in Section 1506 of the Energy Policy Act of 2005,
directed EPA to produce an updated fuel effects model representing the gasoline vehicle fleet at
the time of the study. Carrying out other statutory requirements, such as an anti-backsliding
assessment of the effects of increased renewable fuels use on air quality as outlined in Section
209 of the Energy Independence and Security Act of 2007, were also dependent upon updated
fuel effects data.
In September 2007, Southwest Research Institute (SwRI) began conducting work in San
Antonio, Texas, on a series of tasks and assignments for EPA related to the statutory
requirement. By January 2009, SwRI had completed Phases 1 and 2 the EPAct program. These
pilot phases, described in a separate document, involved testing of 19 light duty cars and trucks
(subsequently referred to as the "EPAct fleet") on three fuels, at two temperatures.4
In March 2009, SwRI began work on Phase 3 (the main program), which was jointly
supported by EPA, the U.S. Department of Energy (DOE) through the National Renewable
Energy Laboratory (NREL), and the Coordinating Research Council (CRC). In addition,
Lubrizol Corporation provided material and analytical support. This report covers work
conducted for Phase 3, also known as EPAct/V2/E-89, which involved the testing of fifteen of
the high-sales Tier 2 compliant vehicles from the EPAct Phase 1 fleet using twenty-seven test
fuels covering typical ranges of five fuel parameters. Phase 3 also included testing of the four
3 See "Technical Support Document: Analysis of California's Request for Waiver of the Reformulated Gasoline
Oxygen Content Requirement for California Covered Areas", document number EPA420-R-01-016. See also 40
CFR 80.45.
4 EPAct program Phases 1-2 are described in a memorandum to the Tier 3 docket EPA-HQ-OAR-2011-0135.
-------
flexible-fuel vehicles (FFVs) from the EPAct fleet on an E85 fuel (three of these vehicles were
among the 15-vehicle test fleet). Phase 3 data collection was completed in June 2010.
-------
III. TEST FUEL SPECIFICATION AND BLENDING
A. Design of Fuel Matrix
1. Selection of Fuel Parameters for Investigation
Gasoline is a mixture of hundreds of hydrocarbon species, with boiling points spanning
the range of 100-400°F, and molecular size ranging from four to ten carbons. Depending on the
source of the parent crude oil and configuration of the refinery producing the gasoline, the
proportions of particular chemical species can vary widely. This is acceptable so long as a
number of key bulk properties of the mixture remain within a range that ensures acceptable
operation of vehicles and other equipment designed to use it. Over the past few decades, EPA
and other agencies have also found it useful to regulate some of these properties to control
harmful emissions.
While an exhaustive study of effects of gasoline properties on emissions could include
dozens of parameters, producing results with an acceptable level of statistical power would
require such a large and complex design that it would be impractically lengthy and expensive.
To narrow the scope a survey was made of available data for recent technology vehicles, with
special attention paid to the expected magnitude of emission changes caused by fuel properties,
as well as how important characterization of an effect might be in a regulatory context.
Databases used for estimating the magnitude of emission impacts included those associated with
the EPA Complex and Predictive models, the California Air Resources Board Predictive Model,
the CRC E-67 and E-74b test programs, as well as the 2005-6 MSAT test program conducted by
EPA and several automakers. Table III-l lists the parameters initially considered as candidates
for study and summarizes selection criteria related to each.
-------
Table III-l. Fuel parameters considered for inclusion in the study.
Fuel Parameter
Aromatics
Ethanol
RVP
T50
T90
Sulfur
Olefins
Octane number
Drivability index
Total oxygen
Polyaromatics
Data Availability for
Latest Technology
Vehicles
Little
Some
Some
Little
Little
Some
Little
Little
Little
Little
Little
Expected Magnitude
of Emission Impact
High
Uncertain
Uncertain
Uncertain
Uncertain
High
Uncertain
Low
Low
Uncertain
Uncertain
Fuel Policy Relevance
High
High
High
Low
Low
High
Uncertain
Low
Low
Low
Uncertain
Initial assessment of costs and program duration suggested it would not be feasible to
include more than five fuel parameters and their interactions. Based on a combination of the
criteria shown, the five parameters listed at the top of this table were chosen as the study targets.
Two other parameters, sulfur and olefin content, were also of interest; examination of sulfur
would require a different program design due to involvement of the exhaust catalyst, and olefins
were deferred for study later.5'6
2. Selection of Fuel Parameter Ranges and Levels
Fuel parameters may have linear or nonlinear impacts on emissions. To capture a
nonlinear impact, three or more treatment levels of a given parameter must be included in the
study design. Statistical models of data from prior studies suggested that T50, T90, and ethanol
content may have nonlinear impacts on emissions. Based on this information, EPA originally
planned to test three levels of ethanol spanning the range of EO-E10 (the limit for legal market
fuels at that time). During the design process, DOE (via NREL) offered additional funding to
add El 5 and E20 fuels to broaden the database to include fuel blends that might be expected
5 A subsequent study on the effects of olefins using the same 15-vehicle test fleet as this study was published as
Maryam Hajbabaei, Georgios Karavalakis, J. Wayne Miller, Mark Villela, Karen Huaying Xu, Thomas D. Durbin,
Impact of olefin content on criteria and toxic emissions from modern gasoline vehicles, Fuel, Volume 107, May
2013, Pages 671-679, ISSN 0016-2361, 10.1016/j.fuel.2012.12.031.
6 A subsequent study on the effects of sulfur content using the same makes and models as this study was published
by EPA as document EPA-420-D-13-003, "The Effects of Gasoline Sulfur Level on Emissions from Tier 2 Vehicles
in the In-Use Fleet".
-------
appear in the market in the future. After this adjustment, four levels of ethanol were selected at
0%, 10%, 15%, and 20% by volume. Five levels of T50 were chosen to allow detailed
characterization of any interaction with ethanol. Finally, to examine potential nonlinear impacts
of T90, two fuels were added with an intermediate level of this parameter.. 7 The remaining two
parameters, aromatic content and RVP (reported in this program as DVPE), had two levels.
The parameter ranges to be covered for T50, T90, aromatic content, and RVP were
selected to represent the range of in-use fuels based on a review of the Alliance of Automobile
Manufacturers' 2006 North American Fuel Survey. As the emissions tests were to be performed
at the nominal temperature of 75°F, summer survey data was used. Since the effect of fuel
changes on emissions was expected to be small in comparison to other sources of test-to-test
variability, the span of fuel parameter ranges was maximized in order to increase the likelihood
of discerning statistically significant results. Test fuel parameter ranges were originally drafted
to span roughly the 5th to 95th percentiles of survey results for in U.S. gasoline, though some test
fuel parameters were adjusted after the actual blending process began (discussed further in
section III.B). An intermediate level of T50 in EO fuels was selected to coincide with the high
level of T50 in E10 fuels. Similarly, an intermediate level of T50 in E10 fuels was selected to
coincide with the low level of T50 in EO fuels.
For El 5 and E20 fuels, the T90, aromatic content, and DVPE ranges selected for EO and
E10 fuels were applied. A single level of T50 was selected for E20 blends based on the
information obtained from CRC report No. 648 (2006 CRC Hot-Fuel-Handling Program) as well
as petroleum industry sources which indicated that it was largely independent of the hydrocarbon
fraction of the fuel and would not deviate more than several degrees from 160°F due to the
o
presence of a large fraction of ethanol. At the time this fuel matrix was being designed, no
information was available on distillation properties of E15 fuels. Two levels of T50 were
7 The intermediate level of T90 occurs along one edge of the fuel domain in Phase 3. Statistical analysis of any
nonlinear T90 effect was intended to rely on a fuel used in Phase 1 of the program as an additional source of data for
the intermediate T90 level.
8 As ethanol blend level moves beyond 10 vol%, T50 becomes increasingly correlated (inversely) with ethanol
content. At E15, the two can be manipulated independently with some effort within a relatively limited range. By
E20, the behavior of the center of the distillation curve (where T50 lies) is dominated by ethanol's boiling point, and
thus T50 cannot be moved outside a narrow range around 165°F. Thus, T50 and ethanol should only be understood
to be independently blended parameters at E10 and below.
10
-------
selected for the E15 fuels, the low level equal to the lowest T50 assumed for E10 fuels and the
high level being a linear interpolation between the highest T50 of E10 fuels and the sole T50
level of E20 fuels. Table III-2 shows the levels and range of parameters chosen during the initial
design of the Phase 3 fuel set. Some of these parameters were adjusted during subsequent
blending steps (discussed further in section III.B).
Table III-2. Summary of initial Phase 3 test fuel set design.
Fuel Parameter Number of Levels Values to Be Tested
Ethanol (vol%) 4 0, 10, 15,20
T50 (°F) 5 150, 160 (E20 only), 190, 220, 240
T90 (°F) 3 300,325,340
Aromatics (vol%) 2 15,40
RVP (psi) 2 7, 10
3. Statistical Design of the Fuel Matrix
Studies involving multiple levels of multiple variables are ideally conducted in such a
way that each parameter is varied independently through all its levels while the others are held
constant, thus generating data from all possible combinations of treatment levels (referred to as a
factorial design). Given the levels of treatments outlined above, the Phase 3 fuel set would
require some 240 fuels to be blended and tested to examine each point of a factorial matrix. This
clearly being impractical, use of a partial factorial design was utilized. This arrangement uses a
subset of all the possible fuel blends to allow characterization with statistical confidence of the
main effects (i.e., the fuel parameters themselves) plus a pre-selected subset of interactions
between the main effects.9 Interactions of interest were chosen based on models from prior
studies as well as engineering judgment, and are as follows: ethanol interactions with each of the
other fuel parameters, plus T50-squared and ethanol-squared. With the five main effects plus
these six interactions, the design was to be optimized for 11 potential model terms.
9 What is sacrificed in a partial factorial design is the ability to properly characterize all interactions between the
study parameters. Interactive effects beyond those for which the design is optimized are confounded with others to
varying degrees, and thus conclusions about them can't generally be made with statistical confidence.
11
-------
Selection of the best subset of fuels for a partial factorial design is a computationally
intensive process, for which specialized software has been developed. This work was done by
SwRI statisticians Robert Mason and Janet Buckingham. A matrix of 27 fuels as shown in Table
III-3 was eventually constructed to maximize statistical performance over the desired range of
parameters. The statisticians' report describing the matrix development process is available as
Appendix A.10
10 The statisticians' report covers matrix design through several stages of evolution of program objectives and
funding levels. The fuel matrix shown in Table 9 of the report was eventually used for Phase 3, with the exception
offuels 17-19, which were tested in earlier phases of the program. ThePhaseS matrix of 27 fuels was intended to
stand on its own for modeling purposes, with optional inclusion offuels 17-19 in the database as "centered" fuels.
The development offuels 17-19 is described in the SwRI EPAct/V2/E-89 Fuel Blending Final Report attached as
Appendix B.
12
-------
Table III-3. Phase 3 fuel matrix resulting from partial factorial design.
Test Fuel
Number3
1
2
O
4
5
6
7
8
9
10
11
12
13
14
15
16
20
21
22
23
24
25
26
27
28
30
31
T50,
°F
150
240
220
220
240
190
190
220
190
220
190
150
220
190
190
220
160
160
160
160
160
160
150
190
190
150
160
T90,
°F
300
340
300
340
300
340
300
300
340
340
300
340
340
340
300
300
300
300
300
340
340
340
340
340
300
325
325
Ethanol,
vol.%
10
0
10
10
0
10
0
0
0
10
10
10
0
0
0
10
20
20
20
20
20
20
15
15
15
10
20
DVPE,
psi
10
10
7
10
7
7
7
10
10
7
10
10
7
7
10
7
7
7
10
7
10
10
10
7
7
10
7
Aromatics,
vol.%
15
15
15
15
40
15
15
15
40
40
40
40
40
15
40
40
15
40
15
15
15
40
40
15
40
40
40
aFuels 17-19 were tested in an earlier phase of the program. Fuel 29 is an E85 fuel
not included in the statistical matrix design.
4. Test Fuel Specification
In addition to the five fuel properties that were the focus of this program, the test fuel
specifications included a number of parameters which were known or suspected to affect
emissions. In an attempt to produce as controlled an experiment as possible, the levels of those
parameters were specified within certain bounds across all test fuels. Examples of these
properties include olefm and sulfur content and octane number. An olefin specification of 7.0 ±
1.5 vol% was used, based on the U.S. average computed from the Alliance of Automobile
13
-------
Manufacturers' 2006 Summer North American Fuel Survey. Because of its important impact on
benzene emission performance, the fuel benzene content specification was set at 0.62 ±0.15
vol%, the level of the refinery average gasoline benzene standard effective on January 1, 2011.
The sulfur content specification of 25 ± 5 mg/kg (ppmw) was selected to ensure that the level
remained within a reasonably narrow range capped by the current refinery/importer annual
average standard of 30 mg/kg. The minimum (R+M)/2 octane specification of 87 was based on
minimum requirements of test vehicles selected for this program. Blending tolerances for all
properties are shown in Table III-4. Note that there was no upper limit on octane number, so its
value tends to be correlated with ethanol content since ethanol has a high octane number and its
level varied widely across the blends.
Special attention was paid to the distribution of aromatic molecule sizes because of
potential impacts on particulate matter emissions. An attempt was made to maintain consistent
proportions of aromatic-containing molecules with seven or more carbon atoms. The ratios of
2:2:2:1 were chosen for C7:Cg:C9:Cio aromatics based on detailed hydrocarbon analysis data for
commercial gasolines that was available to EPA at the time of fuel blending. As a practical
matter of meeting the distillation targets, the proportions had to be adjusted to include more C?
and Cg aromatics for fuels with a combination of low T90 and high aromatics.
The fuel specification also included T10, FBP, oxidation stability, copper strip corrosion
and solvent-washed gum content requirements taken from the ASTM D4814 Standard
Specification for Automotive Spark-Ignition Engine Fuel. Furthermore, a limit on total content
of oxygenates other than ethanol was adopted to safeguard the test fuels against such
contamination. Finally, a number of uncontrolled fuel properties used in emissions test
calculations were appended to the fuel specification table to make sure they were measured prior
to the launch of the emissions test program and in the EPAct/V2/E-89 Fuels Round Robin to be
conducted. They included carbon, hydrogen and oxygen content, density, and heat of
combustion.
A single E85 fuel (fuel number 29) was also incorporated in this program for testing in
flexible fuel vehicles. This fuel was not blended along with other fuels used in Phase 3 of the
14
-------
EPAct/V2/E-89 Program, but was obtained from CRC. Its properties are provided in Section
HID.
B. Test Fuel Development and Blending
1. Development and Verification of Test Fuel Formulations
All fuels tested in Phase 3 of the EPAct/V2/E-89 Program were formulated by EPA in
conjunction with Haltermann Solutions, Houston, Texas (with the exception of fuel 29, the E85
blend). To facilitate the formulation process, Haltermann made a detailed set of property data
for their gasoline blending components available to EPA for use in designing fuels for this
program. (These data were deemed confidential business information and are not provided in
this report.) The majority of these components were blendstock streams taken from various
points in refinery operations (e.g., reformate, alkylate, isomerate, light naphtha, etc.), ensuring
that the resulting test fuels contained the typical range of components found in actual market
fuels.
In the development of each test fuel formulation, EPA used a computational blending
model to define and adjust the blend recipe from the available components while Haltermann
prepared and characterized lab-scale hand blends. Distillation by ASTM D86 method, aromatics
and olefms by D1319, DVPE by D5191 and ethanol content by D5599 were always measured at
this stage. Sulfur by D5453 was measured if its predicted level was > 27 mg/kg, benzene by
D3603 when its predicted content was >0.70 vol% and octane numbers by D2699 and D2700 if
predicted knock index (RON+MON)/2 was <88.0. The use of distillation stills equipped with
OptiDist or equivalent technology was required for all E10, E15, and E20 fuels.n
The distillation parameters, DVPE, as well as the aromatic and olefinic content of hand
blends were measured by Haltermann and verified by a third-party laboratory of Haltermann's
11 Accurately measuring T50 becomes increasingly difficult as more ethanol is blended into the fuel. Older still
models were found to generate results with unacceptable levels of variability. OptiDist is a trade name denoting
enhanced automated control of distillation parameters that produces better results than traditional methods.
15
-------
choice (typically Core Laboratories, Dixie Services, BSi-Inspectorate, or Saybolt). Single
measurements were allowed for benzene, sulfur, and ethanol content, as well as MON and RON,
though confirmation could be performed at one of the third-party laboratories at Haltermann's
discretion.
As the development of fuel formulations progressed, it became clear that some property
values in the initial fuel matrix design (Table III-3) could not be met within acceptable
tolerances. They included:
• T50 target of 150±4°F for the El 5 fuel 26
• T50 target of 160±4°F for all E20 fuels
• DVPE target of 6.65±0.15 psi for the vast majority of "low" DVPE fuels
• RON target of 93±2 for many fuels
• MON target of 85±2 for many fuels
• (RON + MON)/2 target of 87±2 for many fuels
The initial blending experiments also revealed that the upper T50 limit for El5 fuels was
as high as 220°F, considerably higher than the 190°F target assumed for fuels 27 and 28 in the
absence of relevant information in the technical literature.
Finally, a decision was made to replace the aromatic content level of 40 vol% with a
target of 35 vol% to facilitate blending of "high" aromatic content fuels. It was also thought that
the lower target would be more representative of future market fuels.
As a consequence of the observations and decisions mentioned above, the following
modifications were incorporated in the fuel specification:
• T50 target of 160±4°F for the E15 fuel 26
• T50 target of 220±4°F for the El5 fuels 27 and 28
• T50 target of 165±4°F for all E20 fuels
• DVPE range of 7±0.25 psi for the "low" DVPE fuels
16
-------
• Aromatic content target of 35±1.5 vol% for all "high" aromatic content fuels
• Separate RON and MON targets were eliminated, replaced by (RON + MON)/2
target of >87 for all fuels
These specification changes are reflected in Table III-4, and the resulting domain of
ethanol content vs. T50 of the test fuel matrix is shown in Figure III-l.
17
-------
Table III-4. Final test fuel specification for blending.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol Content
Total Content of Oxygenates Other
Than Ethanol
T10
T50
T90
FBP
DVPE
Aromatics
Olefins
Benzene
S
(RON + MON)/2
C
H
O
Water Content
Net Heat of Combustion
Oxidation Stability
Copper Strip Corrosion, 3h at 122°F
Solvent-Washed Gum Content
UNIT
-
"API
wl.%
vol. %
OF
°F
°F
°F
psi
wl.%
vol. %
vol. %
mg/kg
-
mass %
mass %
mass %
mg/kg
MJ/kg
minute
-
mg/lOOml
METHOD
D4052
D4052
D5599
D5599
D86
D86
D86
D86
D5191
D1319
D1319
D3606
D5453
Calc.
D5291 mod*
D5291 mod*
D5599
E1064
D4809
D525
D130
D381
BLENDING
TOLERANCE
NA
NA
E0:<0.1;
E10:±0.5;
E15:±0.5;
E20:±0.5;
ESS: ±2
-
-
±4
±5
-
±0.25
±1.5
± 1.5
±0.15
±5
-
-
-
-
-
-
-
-
-
TESTFUELS
1
Report
Report
10
<0.1
<158
150
300
<437
10
15
7
0.62
25
>87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
-------
Table III-4 Cont. Final test fuel s
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol Content
Total Content of Oxygenates Other
Than Ethanol
T10
T50
T90
FBP
DVPE
Aromatics
Olefins
Benzene
S
(RON + MON)/2
C
H
O
Water Content
Net Heat of Combu stion
Oxidation Stability
Copper Strip Corrosion, 3h at 122°F
Solvent-Washed Gum Content
UNIT
-
"API
wl.%
vol. %
op
°F
°F
op
psi
W>1.%
vol. %
vol. %
mg/kg
-
mass %
mass %
mass %
mg/kg
MJ/kg
minute
-
mg/lOOml
pecification for blending.
METHOD
D4052
D4052
D5599
D5599
D86
D86
D86
D86
D5191
D1319
D1319
D3606
D5453
Calc.
D5291 mod*
D5291 mod*
D5599
E1064
D4809
D525
D130
D381
BLENDING
TOLERANCE
NA
NA
E0:<0.1;
E10:±0.5;
E15:±0.5;
E20:±0.5;
ESS: ±2
-
-
±4
±5
-
±0.25
±1.5
± 1.5
±0.15
±5
-
-
-
-
-
-
-
-
-
TEST FUELS
15
Report
Report
0
<0.1
<158
190
300
<437
10
35
7
0.62
25
>87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
87.0
Report
Report
Report
Report
Report
>240
-------
20
15 -
10
0
150 160 170 180 190 200 210 220 230 240 250
T50,°F
Figure III-l. Nominal ethanol content vs. T50 domain of EO-E20 fuels (Final version)
Overall, 21 different blending components were used in this program, between 9 and 16
per fuel. The process of developing formulations for each test fuel required as many as seven
lab-scale iterations in some cases with particularly challenging combinations of fuel parameters.
Once the T10, T50, T90, FBP, DVPE, knock index as well as the aromatic, olefmic,
benzene, sulfur and ethanol content requirements of the specification shown in Table III-4 were
met, the remaining parameters listed in that specification were measured. If they also met the
requirements listed in Table III-4, the final specification for the bulk blend of the fuel was issued
by EPA and NREL using the template shown in Table III-4. In each such bulk blend
specification, the approved hand blend results were used as targets for the distillation parameters,
DVPE as well as the aromatic and olefmic content. The specification shown in Table III-4
20
-------
included T30 and T70 to make sure that the distillation curves of the hand blends were closely
reproduced in the bulk blends. This specification and the corresponding fuel formulation (blend
recipe) were subsequently provided to Haltermann for use in the preparation of the bulk blends.
The use of hand blending results as targets for the bulk blends was done to accommodate
setting of realistic targets and tolerances for the fuel blending subcontractor to follow. As shown
in Section HID, this approach still kept the parameters in question within tight blending
tolerance limits.
21
-------
Table III-5. Bulk blend specification template.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol Content
Total Content of Oxygenates
Other Than Ethanol
Distillation
T10
T30
T50
T70
T90
FBP
DVPE
Aromatics
Olefins
Benzene
S
(RON+MON)/2
C
H
O
Water Content
Net Heat of Combustion
Oxidation Stability
Copper Strip Corrosion, 3h at
122°F
Sol vent- Washed Gum Content
UNIT
-
°API
vol. %
vol. %
°F
°F
°F
°F
°F
°F
psi
vol. %
vol. %
vol. %
mg/kg
-
mass %
mass %
mass %
mg/kg
MJ/kg
minute
-
mg/100
ml
METHOD
D4052
D4052
D5599
D5599
D86
D86
D86
D86
D86
D86
D5191
D1319
D1319
D3606
D5453
Calc.
D5291 mod*
D5291 mod*
D5599
El 064
D4809
D525
D130
D381
BLENDING
TOLERANCE
-
-
E0:<0.1
E10:±0.5
E15:±0.5
E20: ±0.5
-
±5
±5
±4
±5
±5
-
±0.15
±1.5
±1.5
±0.15
±5
-
-
-
-
-
-
-
-
-
SPECIFICATION
Report
Report
Per Table III-4
<0.1
Value based on
hand blend data
Value based on
hand blend data
Value based on
hand blend data
Value based on
hand blend data
Value based on
hand blend data
<437
Value based on
hand blend data
Value based on
hand blend data
Value based on
hand blend data
0.62
25
>87.0
Report
Report
Report
Report
Report
>240
-------
2. Bulk Blending and Analysis of Test Fuels
Haltermann prepared the bulk blends and adjusted their properties until they met the bulk
blend specifications. Distillation parameters, DVPE, MON, RON as well as the aromatic,
olefmic, benzene and S content were measured during bulk blending by Haltermann and verified
by another laboratory, usually Core Laboratories, Dixie Services, BSi-Inspectorate, or Saybolt.
Single measurements were required for the remaining parameters listed in Table III-5 and could
also be performed at one of the mentioned laboratories at Haltermann's discretion.
Once the bulk blend specification was met, a sample of the fuel was shipped to SwRI for
confirmatory testing. If SwRI analytical results fell inside the test method reproducibility limits
for all parameters listed in Table III-5, EPA and NREL approved the bulk blend for shipment to
SwRI. If they fell outside the method reproducibility limits, further tests and/or blend
adjustments followed until the requirements of the specification were met.
C. Shipping, Storage and Handling of Test Fuels
For contingency purposes the procured quantities of test fuels (500 gallons each)
exceeded the anticipated needs to complete the program by 20%. The fuels were shipped by
Haltermann to SwRI in epoxy-lined, 55-gallon drums and stored on site in a temperature-
controlled facility (Figure III-2). The storage temperature for unopened drums was 70°F ± 5°F.
Any drums that were to be opened (for vehicle fueling or sampling) were cooled down to a
temperature of <50°F at a dedicated cold storage facility located behind the emissions laboratory
next to the vehicle refueling bay. The temperature of both fuel storage facilities was
continuously recorded, and was verified at least once a day.
Upon arrival at SwRI, all fuels received independent identifiers which included the
EPAct/V2/E-89 fuel number, a SwRI fuel code, and a project-specific supplementary three-letter
code (Table III-6). All fuel drums and corresponding work requests included all three
designators in an effort to assure the correct fuel was being used at any point in the test program.
23
-------
Additionally, each individual drum received a sequential number. These unique alphanumeric
designations assigned to individual drums were recorded and verified by two individuals each
time a test vehicle was fueled.
Each time a full drum was opened, the properties of its contents were verified using a
portable PetroSpec gasoline analyzer. Based on these results, no mislabeling of fuel drums
occurred during this program.
With the exception of drum content verification using the PetroSpec analyzer, the 200
gallons of fuel 29 (E85) provided by the CRC were handled in the same manner as the EO-E20
fuels.
24
-------
Table III-6. Test fuel identifiers used by SwRI.
EPAct/V2/E-89
FUEL NUMBER
1
2
3
4
5
6
1
8
9
10
11
12
13
14
15
16
20
21
22
23
24
25
26
27
28
29
30
31
SwRI FUEL CODE
EM-6995-F
EM-6953-F
EM-7053-F
EM-6996-F
EM-7061-F
EM-7092-F
EM-6954-F
EM-6936-F
EM-6955-F
EM-7093-F
EM-7055-F
EM-6997-F
EM-6965-F
EM-6956-F
EM-6957-F
EM-7056-F
EM-7057-F
EM-7058-F
EM-7001-F
EM-7059-F
EM-6998-F
EM-7073-F
EM-7094-F
EM-7095-F
EM-7096-F
EM-6975-F
EM-7060-F
EM-7074-F
SwRI FUEL NAME
SAT
ELP
FLG
HOU
MCI
IND
JMJ
BWI
KAW
LNK
MIA
MLS
CLF
BNA
OAK
OSH
PHX
RNO
SLC
SFO
TEX
TUL
YAK
BOS
NBA
E80
BUF
GPZ
25
-------
Figure III-2. Constant-temperature storage of unopened fuel drums
D. Fuels Round Robin
The EPAct/V2/E-89 Fuels Round Robin was launched in October 2009 with the support
of the Coordinating Research Council (CRC). Its objective was to supplement the fuel
inspection data previously generated for the drum blends by Haltermann, Core Laboratories,
Dixie Services, BSi-Inspectorate, Saybolt, and SwRI with additional results, especially for the
parameters most critical to this program, i.e. T50, T90, DVPE, and ethanol and aromatic content.
The complete lists of parameters which were measured in the 27 EO through E20 fuels and the
E85 fuel 29 are provided in Tables III-6 and III-7, respectively.
Round robin participants included BP, Chevron, ConocoPhillips, EPA, ExxonMobil,
Marathon, PAC (distillation equipment manufacturer) and Shell. The total number of
laboratories which measured the properties of fuels used in this program thus increased to 14.
Tables III-6 and III-7 show the total number of test results which were eventually generated for
26
-------
each fuel using the specific tests. It is worth noting that at least six T50, T90, DVPE, ethanol
content and aromatic content results were available for each EO-E20 fuel.
Table III-7. Parameters measured in EO-E20 fuels in EPAct/V2/E-89 fuels round robin.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol
Total Content of Oxygenates Other Than Ethanol
Distillation
DVPE (EPA equation)
Aromatics
Olefins
Benzene
S
RON
MON
C
H
O
Net Heat of Combustion
Water
Lead
Copper Strip Corrosion
Solvent Washed Gum Content
Oxidation Stability
TEST METHOD
D4052
D4052
D5599
D5599
D86
D5191
D1319
D1319
D3606
D5453
D2699
D2700
D5291 mod.*
D5291 mod.*
D5599
D4809
E-1064
D3237
D130
D381
D525
TOTAL NUMBER
OF TEST RESULTS
PER FUEL
4
4
6
2-4
6-12
9-10
7
7
4-5
6
4
4
4
4
3-6
3
1-2
0-1
1
1
1
* Method adapted by individual laboratories to testing of gasolines
27
-------
Table III-8. Parameters measured in E85 fuel during EPAct/V2/E-89 fuels round robin.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol
Methanol
Distillation
DVPE (EPA equation)
Benzene
S
c
H
0
Water
Net Heat of Combustion
Solvent Washed Gum
Unwashed Gum
Acidity (as acetic acid)
pHe
Inorganic Chloride
Copper
TEST METHOD
D4052
D4052
D5501mod.*
D5501mod.*
D86
D5191
D5580
D5453
D5291mod.**
D5291mod.**
D5501mod.*
E203
D4809
D381
D381
D1613
D6423
D7319
D1688 modified as outlined in D4806
TOTAL NUMBER
OF TEST RESULTS
PER FUEL
8
8
5
7
11
10
1
3
4
4
5
6
1
2
2
1
2
1
1
* Method adapted by individual laboratories to testing of E85 fuels
** Method adapted by individual laboratories to testing of gasolines
The instructions issued to the participants at the outset of the round robin included special
requirements regarding test methods D86, D5501 and D5291. More specifically, they were
requested to:
• Use distillation stills equipped with enhanced distillation rate control technology to
perform D86 distillations on ethanol containing fuels. It turned out that the round robin
participants who met this requirement all used PAC stills featuring OptiDist technolog
They were also instructed to perform charge volume scans for improved measurement
precision. Seven round robin participants complied with this request.
>gy-
28
-------
• Generate C and H content data using the D5291 test method modified for use with
gasoline. The standard version of this method is "not recommended for analysis of
volatile materials such as gasoline". However, some contract and petroleum industry
laboratories are known to have adapted this method to testing of volatile petroleum
products and that support on this issue was available from test equipment manufacturers.
Overall, four participants used such a modified D5291 method in this program.
• Generate ethanol content data for the E85 fuel 29 by means of the D5501 test method
modified for use with E85 fuels. The standard version of this method is applicable to
denatured fuel ethanol containing 93-97 mass % of ethanol. In response to this request,
two of the participating laboratories adapted the ASTM D5501 method to ethanol levels
typical of E85 fuels. To this end, they utilized draft #4 of the ASTM 5501-09 Proposed
Revision of Standard Test Method for Determination of Ethanol Content of Mid and
High Level Ethanol Fuels by Gas Chromatography (valid over the 20-100 mass % range)
which at the time was being balloted by ASTM Subcommittee D02.04. It was also
determined that the three laboratories which had previously generated D5501 data on the
E85 fuel 29 for the CRC also used their own, in-house versions of the D5501 method
adapted to testing of E85 fuels.
To ensure consistency with the best industry practice, a Procedure for Sampling and
Handling of Gasoline Samples was developed with oil company assistance for use in the round
robin. It was provided to SwRI and is included in this report as Appendix C.
Finally, to enable blind testing, a new set of designations was assigned to the test fuels.
The only fuel which was recognizable to the participants prior to its analysis was the E85 fuel 29
which was subjected to a different battery of tests than the other 27 fuels.
The fuel samples reached round robin participants by mid-October 2009. By mid-March
2010, all promised data were received by the EPA. Shortly thereafter, a blinded set of round
robin results was made available to all participants for review, so that each could determine if
their data were correctly entered into the EPAct/V2/E-89 database. At the same time, the EPA
29
-------
and NREL performed a preliminary review of the whole data set, identified results which were
obviously in error and requested retesting by the respective laboratories.
Following the approval of round robin database entries by all participants and the
incorporation of any results of retesting along with fuel property data generated earlier in the
program, the EPA statistician identified outliers in the data using the methodology described in
ASTM El 78-08 Standard Practice for Dealing with Outlying Observations. The details of this
methodology as applied to the Phase 3 EPAct/V2/E-89 fuel data set have been summarized by
the EPA statistician and are included in this report as Appendix D.
The final, blinded EPAct/V2/E-89 fuel property data set is provided in Appendix E. Red
cell fill color in Appendix E tables indicates outliers. Orange fill color denotes data which were
not used in the calculation of parameter averages. The latter occurred when a T50 and/or a T90
result was an outlier or when sample contamination was suspected.
The final set of Phase 3 EPAct/V2/E-89 test fuel properties to be used in emissions
modeling is provided in Tables III-8 and III-9. Their values were computed using the round
robin test data provided in Appendix E following the elimination of outliers. Table III-9 lists the
properties of EO-E20 fuels, while Table III-10 those of the E85 fuel 29.
30
-------
Table III-9. Final properties of EO-E20 Phase 3 EPAct/V2/E-89 fuels based on round robin results.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol
Total Content of Oxygenates Other
Than Ethanol
Distillation IBP
5% evap
1 0% evap
20% evap
30% evap
40% evap
50% evap
60% evap
70% evap
80% evap
90% evap
95% evap
FBP
DVPE (EPA equation)
Aromatics
Olefms
Saturates
Benzene
S
RON
MON
(RON+MON)/2
c
H
O
Net Heat of Combustion
Water
Lead
Copper Strip Corrosion
Solvent Washed Gum Content
Oxidation Stability
UNIT
S/cm
"API
vol. %
vol. %
op
op
op
op
op
op
°F
op
op
op
°p
°F
°F
psi
vol. %
vol. %
vol. %
vol. %
mg/kg
mass %
mass %
mass %
MJ/kg
mass %
g/i
mg/lOOml
min.
TEST METHOD
D4052
D4052
D5599
D5599
E10 E15 and E20 fuels)
D5191
D1319
D1319
100-D1319Aromatics-
D1319Olefms - D5599Ethanol
D3606
D5453
D2699
D2700
D5291 mod.*
D5291 mod.*
D5599
D4809
E-1064
D3237
D130
D381
D525
FUEL
1
0.7211
64.6
10.03
<0.10
92.9
112.5
117.3
123.9
131.2
139.9
148.9
172.3
224.1
254.6
300.2
334.5
368.0
10.07
15.4
7.6
67.0
0.62
30
94.8
86.3
90.6
81.70
14.02
3.9
41.950
0.071
1A
<0.5
>240
2
0.7220
64.3
<0.10
<0.10
83.5
105.4
121.7
154.4
190.6
218.5
236.7
252.7
271.7
305.9
340.1
353.0
375.3
10.20
14.1
6.8
79.1
0.51
23
96.0
88.6
92.3
85.12
14.43
<0.1
43.960
0.010
<0.001
1A
<0.5
>240
3
0.7350
60.8
10.36
<0.10
106.4
136.0
141.7
148.9
155.0
175.1
217.5
230.2
243.6
257.1
295.9
334.4
368.9
6.93
15.0
7.6
67.0
0.61
22
98.0
87.6
92.8
81.61
14.17
3.9
41.536
0.059
1A
<0.5
>240
4
0.7346
60.9
9.94
<0.10
89.9
115.9
126.3
140.9
151.7
161.2
221.9
245.9
270.0
303.5
337.5
352.0
369.8
10.01
15.5
6.8
67.8
0.54
21
97.1
87.6
92.4
82.21
14.12
3.7
41.952
0.077
1A
1.5
>240
5
0.7573
55.2
<0.10
<0.10
94.1
128.6
145.4
172.6
199.4
222.1
237.0
247.2
258.5
273.1
300.0
323.5
357.8
6.95
34.7
6.9
58.4
0.51
24
96.7
86.3
91.5
86.58
12.92
<0.1
42.948
0.014
<0.003
1A
<0.5
>240
6
0.7342
61.1
10.56
<0.10
106.7
130.4
135.9
142.6
148.3
153.4
188.5
228.2
267.7
310.1
340.4
352.7
369.2
7.24
15.0
8.8
65.6
0.68
23
96.3
86.6
91.5
81.52
14.21
4.0
41.785
0.073
1A
<0.5
>240
7
0.7208
64.6
<0.10
<0.10
100.1
127.6
137.0
149.0
161.7
176.6
193.1
210.2
228.6
251.5
298.4
329.3
361.8
7.15
17.0
7.5
75.5
0.55
23
91.2
84.2
87.7
85.16
14.25
<0.1
43.735
0.019
<0.001
1A
<0.5
>240
8
0.7191
65.1
<0.10
<0.10
83.7
108.1
123.4
151.6
185.1
204.4
221.1
233.5
246.4
264.0
303.1
330.5
360.9
10.20
15.7
6.4
78.0
0.50
23
95.5
87.8
91.7
85.12
14.32
<0.1
44.037
0.020
0.001
1A
<0.5
>240
9
0.7454
58.2
<0.10
<0.10
85.3
105.1
115.1
130.3
147.2
167.7
192.8
224.7
260.3
292.2
341.8
363.5
384.7
10.30
35.8
6.2
58.0
0.54
23
94.5
84.8
89.7
87.03
12.82
<0.1
43.209
0.009
<0.001
1A
<0.5
>240
10
0.7644
53.4
9.82
<0.10
104.7
130.0
136.3
144.3
151.0
161.6
217.1
261.5
290.4
317.5
340.2
354.3
372.4
7.11
34.0
6.1
50.1
0.52
25
98.5
87.2
92.9
83.47
12.83
3.6
41.210
0.067
<0.003
1A
<0.5
>240
11
0.7596
54.6
10.30
<0.10
92.0
115.4
124.4
137.6
148.1
156.5
189.3
231.1
251.4
270.0
298.6
325.0
360.8
9.93
35.0
6.9
47.8
0.54
24
97.8
85.6
91.7
83.68
12.61
3.7
41.175
0.066
1A
0.5
>240
12
0.7517
56.5
9.83
<0.10
91.3
110.7
116.9
125.0
133.8
142.8
152.2
198.5
275.1
307.9
339.8
357.7
375.9
10.13
34.8
6.9
48.5
0.57
19
100.4
88.0
94.2
83.32
12.68
3.6
41.373
0.066
<0.003
1A
<0.5
>240
13
0.7540
56.0
<0.10
<0.10
96.6
127.0
139.8
158.7
178.2
199.9
222.5
245.2
269.8
303.5
337.9
354.4
377.5
6.92
34.1
6.3
59.6
0.51
23
95.8
85.8
90.8
86.76
13.15
<0.1
43.171
0.014
<0.001
1A
1.5
>240
14
0.7223
64.2
<0.10
<0.10
100.4
126.5
135.5
147.3
160.0
175.1
192.8
212.0
237.3
280.1
338.5
354.5
377.5
7.14
16.9
8.5
74.6
0.52
24
91.5
84.6
88.1
85.28
14.29
<0.1
43.519
0.015
<0.001
1A
<0.5
>240
* Method adapted by individual laboratories to testing of gasolines
31
-------
Table III-9 Cont. Final properties of EO-E20 Phase 3 EPAct/V2/E-89 fuels based on round robin results.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol
Total Content of Oxygenates Other
Than Ethanol
Distillation IBP
5% evap
10% evap
20% evap
30% evap
40% evap
50% evap
60% evap
70% evap
80% evap
90% evap
95% evap
FBP
DVPE (EPA equation)
Aroma tics
Olefms
Saturates
Benzene
S
RON
MON
(RON+MON)/2
C
H
0
Net Heat of Combustion
Water
Lead
Copper Strip Corrosion
Solvent Washed Gum Content
Oxidation Stability
UNIT
3
g/cm
"API
vol. %
vol. %
°F
°F
°F
OF
°F
°F
°F
°F
OF
op
°F
op
op
psi
vol. %
vol. %
vol. %
vol. %
mg/kg
-
-
-
mass %
mass %
mass %
MJ/kg
mass %
g/1
-
mg/lOOml
min.
TEST METHOD
D4052
D4052
D5599
D5599
E10 E15 and E20 fuels)
D5191
D1319
D1319
100 - D1319Aromatics -
D131901efins - D5599Ethanol
D3606
D5453
D2699
D2700
-
D5291 mod.*
D5291 mod.*
D5599
D4809
E-1064
D3237
D130
D381
D525
FUEL
15
0.7428
58.8
<0.10
0.10
84.7
105.5
115.6
130.5
146.6
166.3
189.7
216.2
243.0
265.9
299.4
329.3
363.7
10.23
35.3
7.2
57.4
0.54
24
95.0
84.9
90.0
86.88
12.79
0.1
43.108
0.012
<0.001
1A
0.5
>240
16
0.7636
53.6
10.76
0.10
104.5
133.0
139.2
147.8
155.1
172.1
218.8
237.5
251.9
268.6
300.6
330.8
365.6
7.12
35.6
6.8
46.9
0.62
23
101.0
88.3
94.7
83.40
12.66
3.9
41.013
0.066
-
1A
1
>240
20
0.7425
58.9
20.31
0.10
107.9
137.3
142.6
149.7
155.3
159.6
162.7
179.9
234.8
253.1
298.7
336.6
371.9
6.70
15.2
7.4
57.1
0.61
22
101.9
89.3
95.6
78.06
14.01
7.6
40.057
0.138
<0.003
1A
<0.5
>240
21
0.7754
50.8
20.14
0.10
106.3
134.7
141.3
150.3
157.1
162.6
167.6
217.3
255.2
275.3
305.0
331.3
360.5
7.06
35.5
7.1
37.3
0.61
22
101.4
87.5
94.5
79.90
12.43
7.1
39.285
0.128
0.009
1A
0.5
>240
22
0.7371
60.3
20.51
0.10
89.8
118.8
129.6
144.3
153.7
159.5
163.2
167.2
233.9
253.6
297.3
334.5
369.9
10.21
15.0
6.9
57.6
0.59
21
101.8
89.3
95.6
78.24
13.85
7.7
40.031
0.113
0.004
1A
<0.5
>240
23
0.7476
57.6
20.32
0.10
109.0
133.3
138.9
146.2
152.3
157.8
162.5
171.6
270.9
311.4
338.2
350.0
364.6
6.84
15.9
7.5
56.4
0.63
21
97.4
86.8
92.1
78.34
13.86
7.5
39.915
0.112
O.003
1A
0.5
>240
24
0.7422
58.9
20.51
0.10
89.7
115.9
126.9
142.8
153.2
160.4
165.1
172.9
266.1
305.5
338.1
350.3
368.2
10.12
15.3
7.3
56.9
0.62
21
100.8
88.6
94.7
78.47
13.86
7.6
40.114
0.108
0.005
1A
0.5
>240
25
0.7702
52.0
20.03
0.10
89.0
113.7
125.5
142.1
153.3
160.9
166.9
191.3
281.6
310.3
337.9
352.7
371.8
10.16
35.2
6.6
38.1
0.65
26
102.2
88.3
95.3
80.62
12.38
7.2
38.855
0.117
0.001
1A
0.5
>240
26
0.7593
54.6
15.24
0.10
88.7
109.6
117.1
127.8
138.6
149.8
160.3
174.7
277.0
306.5
338.7
356.7
377.3
10.21
35.6
6.5
42.7
0.62
23
101.7
88.5
95.1
81.48
12.45
5.6
40.384
0.088
O.003
1A
0.5
>240
27
0.7434
58.6
14.91
0.10
104.8
135.3
142.3
152.0
158.0
163.8
221.5
265.1
274.9
311.3
340.3
351.9
372.2
6.97
14.9
7.4
62.9
0.56
26
100.8
89.2
95.0
80.27
14.01
5.5
41.062
0.090
O.003
1A
0.5
>240
28
0.7699
52.1
14.98
0.10
103.9
136.3
144.2
154.0
160.2
165.8
216.6
240.2
251.6
268.4
298.8
327.3
363.2
6.87
34.5
7.0
43.5
0.59
24
102.7
89.4
96.1
81.78
12.62
5.4
40.383
0.091
O.003
1A
0.5
>240
30
0.7508
56.8
9.81
0.10
90.9
110.3
116.7
125.4
133.9
143.1
152.9
197.2
267.3
294.6
323.8
341.8
366.1
10.23
35.5
6.5
48.2
0.58
23
100.5
88.1
94.3
83.17
13.00
3.6
41.304
0.086
-
1A
0.5
>240
31
0.7742
51.1
20.11
0.10
105.8
132.5
139.1
147.7
155.1
161.3
167.3
214.0
271.6
297.0
325.2
342.1
365.6
6.98
35.5
6.8
37.6
0.60
25
101.7
88.2
95.0
79.90
12.49
7.2
39.391
0.143
O.003
1A
0.5
>240
'* Method adapted by individual laboratories to testing of gasolines
32
-------
Table 111-10. Final properties of the E85 Phase 3 EPAct/V2/E-89 fuel.
PROPERTY
Density, 60°F
API Gravity, 60°F
Uncorrected Ethanol
Uncorrected Methanol
Ethanol
Methanol
Estimated Hydrocarbon
Content
Distillation IBP
5% evap
10% evap
20% evap
30% evap
40% evap
50% evap
60% evap
70% evap
80% evap
90% evap
95% evap
FBP
OWE (EPA equation)
Benzene
S
c
H
0
Water
Net Heat of Combustion
Solvent Washed Gum
Unwashed Gum
Acidity (as acetic acid)
pHe
Inorganic Chloride
Copper
UNIT
g/cm
°API
mass %
mass %
vol. %
vol. %
vol. %
OF
OF
op
op
op
op
op
op
op
op
op
op
op
psi
vol. %
mg/kg
mass %
mass %
mass %
mass %
vol. %
MJ/kg
mg/lOOml
mg/lOOml
mass %
-
mg/kg
mg/1
TEST METHOD
D4052
D4052
D5501 mod.*
D5501 mod.*
D5501 mod.*
D5501 mod.*
100 - D5501Ethanol - E203Water
D86
D5191
D5580
D5453
D5291 mod.**
D5291 mod.**
D5501 mod.*
E203
E203
D4809
D381
D381
D1613
D6423
D7319
D1688 modified as outlined in D4806
Average
0.7797
49.8
79.59
0.01
77.15
<0.01
22.14
99.0
132.9
154.3
167.6
170.3
171.2
171.8
172.1
172.5
172.9
173.9
176.2
265.8
8.92
0.12
16
57.74
12.80
27.19
0.93
0.72
30.058
1.9
1.8
0.0021
8.08
Not detected
0.02
* Method adapted by individual laboratories to testing of E85 fuels
** Method adapted by individual laboratories to testing of gasolines
33
-------
K»
o
>
r\
+J
c
-------
400
Fuel Designations:
257
9 13 14
15
0 10 20 30 40 50 60 70 80 90 100
% Evaporated
Figure III-4. Distillation curves of EO Phase 3 EPAct/V2/E-89 fuels.
35
-------
Fuel Designations:
i
11
PH
o
CL>
I
H
3
12
4
16
6
30
10
400
350
300
250
200
150
100
50
0 10 20 30 40 50 60 70 80 90 100
% Evaporated
Figure III-5. Distillation curves of E10 Phase 3 EPAct/V2/E-89 fuels.
36
-------
Fuel Designations:
PH
o
-
CL>
dn
g
-------
11
10
123456789 101112131415162021222324252627283031
Fuel Designation
Figure III-7. DVPE of Phase 3 EPAct/V2/E-89 fuels.
38
-------
40
30
3
o
u
o
o
i-l
20
10
123456789 101112131415162021222324252627283031
Fuel Designations
Figure III-8. Aromatic content of Phase 3 EPAct/V2/E-89 fuels.
39
-------
IV. TEST VEHICLE FLEET SIZING AND SELECTION
A. Fleet Selection
Emission behavior of vehicles originally meeting Tier 0 and Tier 1 emission standards
(generally, those sold in the 1990s) is well-characterized in databases supporting the Complex
Model for gasoline production compliance and the Agency's California oxygenate waiver
19
decision. Together, these databases contain thousands of measurements taken on hundreds of
vehicles and fuel blends. At the time this program was being designed, only a limited amount of
fuel effect data existed for vehicles meeting the Tier 2 standards, data that would be required by
Congress as highlighted in Section II. Thus, the focus of this program was on filling that data
gap.
1. Statistical Determination of Number of Vehicles
Proper design of an experimental study should ensure that the effect being investigated
has a good chance of being detected if it exists. This likelihood of detection is referred to as the
statistical power of a study. Estimation of power requires specification of a size of effect
deemed to be meaningful, as well as information about the variability or noise that will occur in
the measurements. These inputs may be informed by previous experiments or, if no such data is
available, more arbitrarily chosen based on some assumptions about the behavior of the test
subjects.
A power analysis was performed during the initial design of this program to estimate the
number of vehicles required to detect a fuel effect of various sizes for both hydrocarbon and
NOx emissions. The statistical methods were based on those presented by Snedecor and
Cochran, and are consistent with work done for the Auto/Oil Air Quality Improvement Research
Program (AQIRP) in the early 1990s.13'14 Table IV-1 shows estimates of vehicle-by-fuel
12 See footnote 1.
13 Snedecor G.W. & Cochran W.G., "Statistical Methods", Iowa State Univ Press, 8th Ed.,1989.
14 See Painter, J.L. & Rutherford, J.A., "Statistical Design and Analysis Methods for the Auto/Oil Air Quality
Improvement Research Program", SAE paper number 920319.
40
-------
variability and repeat measurement error derived from data collected in recent emission
programs.
Table IV-1. Data used in vehicle selection power analysis.
Vehicle-by-fuel variability (% as COV)
Test-to-test repeatability (% as COV)
Study
CRC E-67
2005-6 MS AT
Value used
CRC E-67
2006 CARB
2005-6 MS AT
CRC-E74b
Value used
NOx
2
14
14
20
22
15
22
22
NMHC
18
13
18
19
17
17
20
20
The prior test programs shown here were conducted in a variety of configurations and
were included to give an idea of the range of values expected. CRC E-67 and E-74b represent
testing on different vehicles and fuels performed at the same laboratory. The 2005-6 MSAT
program tested different vehicles and fuels at different laboratories. The 2006 CARB data
represent testing of the same vehicle on the same fuel at different laboratories. Examination of
both NOx and hydrocarbon emissions allowed the design to accommodate the more restrictive of
the two.
These two measures of variability were combined to compute the vehicle-by-fuel
standard error was computed as follows, where n is the number of replicates assumed.
= variability2 +
repeatability2
n
41
-------
The initial vehicle selection done for Phase 1 of the program assumed a power of 0.90 and a
significance level of 0.05 for the statistical tests. Thus, the following equation was used to
estimate the number of samples (test vehicles) required, denoted here as r. 15
r = 1 + 10.5 2
r
^vehicle by fuel
[relative difference
Table IV-2 shows an example of the results run for different numbers of replicates and relative
differences detectable between two treatments (e.g., 0.25 = 25% fuel effect).
Table IV-2. Results of power analysis for number of vehicles to test.
Replicates
1
2
3
cp
'-'^vehicle by fuel
0.270
0.230
0.215
Relative difference
0.05
0.10
0.15
0.20
0.25
0.40
0.05
0.10
0.15
0.20
0.25
0.40
0.05
0.10
0.15
0.20
0.25
0.40
Vehicles at 0.90 power
614
155
70
40
26
11
446
113
51
29
19
8
389
98
45
26
17
8
A relative fuel effect of 25% was selected during the design of AQIRP as a reasonable
size of effect to investigate. Using this level of effect, a design of 19 vehicles with two replicates
for each fuel-vehicle combination was originally chosen to meet a 90% study power. This
number seemed large, considering that anti-backsliding policy discussions at the time of this
study design were focusing on NOx emission changes on the order of 5-10% that occurred when
' For more details on this calculation, see Snedecor & Cochran, 1989, p. 102-105.
42
-------
ethanol was added to gasoline in previous studies on older vehicles. However, there was an
expectation that through careful design and execution, this program could achieve better
measurement variability and repeatability than suggested by previous programs, which would
allow adequate statistical power with smaller sample size (something that could not be known
until after data collection was underway). Additionally, the program budget would not have
accommodated the hundreds of test vehicles suggested by the calculation to detect relative
differences below 10%.
2. Criteria Based on Sales, Engine Size, Technology, etc.
The Phase 1 fleet of 19 test vehicles was chosen with the intent of being representative of
latest-technology light duty vehicles being sold at the time the program was being launched. In
terms of regulatory standards, the test fleet was to conform on average to Tier 2 Bin 5 exhaust
levels and employ a variety of emission control technologies, to be achieved by including a range
of vehicle sizes and manufacturers.
Engine family sales data obtained from EPA certification and Wards databases was
analyzed to generate a list of high-sales vehicles as candidates for inclusion.16 Grouping sales
data by engine family allowed additional transparency and flexibility in choosing test vehicles
that represent a wider group than one specific make and model. The resulting test fleet, shown in
Table IV-3, was used in Phases 1-2 of the program and the engine families represented were
expected to cover more than half of new vehicle sales for MY 2008. No criteria were used by
the sponsors to select the individual test vehicles for lease, and thus this selection was effectively
random.
16 Engine family is a term used in manufacturing and certification to describe a combination of a base engine and
after-treatment system that may be used in several vehicle makes and models offered by a manufacturer.
43
-------
Table IV-3. Test vehicle fleet used in Phases 1 and 2 (19 vehicles).
Model
Year
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
Make
GM
GM
GM
GM
Toyota
Toyota
Toyota
Toyota
Ford
Ford
Ford
Ford
Chrysler
Chrysler
Chrysler
Honda
Honda
Honda
Nissan
Brand
Chevrolet
Chevrolet
Saturn
Chevrolet
Toyota
Toyota
Toyota
Toyota
Ford
Ford
Ford
Ford
Dodge
Dodge
Jeep
Honda
Honda
Honda
Nissan
Model
Cobalt
Impala FFV
Outlook
Silverado FFV
Corolla
Camry
Sienna
Tundra
Focus
Taurus
Explorer
F150 FFV
Caliber
Caravan FFV
Liberty
Civic
Accord
Odyssey
Altima
Program
ID
CCOB
CIMP
SOUT
CSIL
TCOR
TCAM
TSIE
TTUN
FFOC
FTAU
FEXP
F150
DCAL
DCAR
JLffi
HCIV
HACC
HODY
NALT
Engine
Size
2.2L 14
3.5LV6
3.6LV6
5.3L V8
1.8LI4
2.4L 14
3.5LV6
4.0L V6
2.0L 14
3.5LV6
4.0L V6
5.4L V8
2.4L 14
3.3LV6
3.7LV6
1.8LI4
2.4L 14
3.5LV6
2.5L 14
Engine Family
8GMXV02.4025
8GMXV03.9052
8GMXT03.6151
8GMXT05.3373
8TYXV01.8BEA
8TYXV02.4BEA
8TYXT03.5BEM
8TYXT04.0AES
8FMXV02.0VD4
8FMXV03.5VEP
8FMXT04.03DB
8FMXT05.44HF
8CRXB02.4MEO
8CRXT03.3NEP
8CRXT03.7NEO
8HNXV01.8LKR
8HNXV02.4TKR
8HNXT03.54KR
8NSXV02.5G5A
Tier 2
Cert Bin
5
5
5
5
5
5
5
5
4
5
4
8
5
8
5
5
5
5
5
3. Reduction of Fleet for Phase 3
Due to budget constraints that arose after beginning Phase 3, the fleet was reduced to 15
vehicles from the 19 used in Phases 1-2. (One additional vehicle, the Dodge Caravan, was used
only for E85 testing to provide a total of four FFVs tested in the program.) A study power
analysis was repeated using data from 15 vehicles in Phase 1, and results suggested a power in
the range of 0.7-0.8 for detecting a 25% relative difference at a significance level of 0.05.
(Relaxing the significance level to 0.10 will increase the study power. Statistical analysis of
Phase 1 data found significant fuel effects smaller than 25%, so this shouldn't be understood as a
lower limit of detectable effects, but rather as a screen for the largest effect that is unlikely to be
missed.)
Reduction of the fleet meant choosing four vehicles to eliminate. Primary considerations
in this process included retaining high-sales engine families, a balance of vehicle and engine
sizes, and maintaining representation of all manufacturers originally included in order to
represent a range of technologies and emission control strategies. There was also consideration
44
-------
of the fact that changes in the test fleet could shift the average program results. As a screen for
this issue, all nineteen vehicles were ranked according to their NOx and NMHC sensitivity to
fuel ethanol level based on the Phase 1 data, with the intent being to avoid removing several
vehicles with similar emissions behavior (though the two pollutants could provide conflicting
direction). Table IV-4 shows the resulting set of vehicles used to generate the Phase 3 data.
Note that the Dodge Caravan was used only for E85 testing, so that 15 vehicles were used to
produce the matrix of 27 fuels.
Table IV-4. Reduced vehicle fleet used for Phase 3.
MAKE
GM
GM
GM
GM
Toyota
Toyota
Toyota
Ford
Ford
Ford
Chrysler
Chrysler
Chrysler
Honda
Honda
Nissan
MODEL
YEAR
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
BRAND
Chevrolet
Chevrolet
Saturn
Chevrolet
Toyota
Toyota
Toyota
Ford
Ford
Ford
Dodge
Dodge
Jeep
Honda
Honda
Nissan
MODEL
Cobalt
Impala FFV
Outlook
Silverado FFV
Corolla
Camry
Sienna
Focus
Explorer
F-150FFV
Caliber
Caravan FF v
Liberty
Civic
Odyssey
Altima
VEHICLE
NAME
CCOB
CIMP
SOUT
CSIL
TCOR
TCAM
TSIE
FFOC
FEXP
F150
DCAL
DCAR
JLIB
HCIV
HODY
NALT
ENGINE
2.4L 14
3.5L V6
3.6L V6
5.3L V8
1.8L 14
2.4L 14
3.5L V6
2.0L 14
4.0L V6
5.4L V8
2.4L 14
3.3L V6
3.7L V6
1.8L 14
3.5L V6
2.5L 14
ENGINE
FAMILY
8GMXV02.4025
8GMXV03.9052
8GMXT03.6151
8GMXT05.3373
8TYXV01.8BEA
8TYXV02.4BEA
8TYXT03.5BEM
8FMXV02.0VD4
8FMXT04.03DB
8FMXT05.44HF
8CRXB02.4MEO
8CRXT03.3NEP
8CRXT03.7NEO
8HNXV01.8LKR
8HNXT03.54KR
8NSXV02.5G5A
EPAT2
BIN
5
5
5
5
5
5
5
4
4
8
5
8
5
5
5
5
CA
CERT
NA
L2
L2
NA
U2
U2
U2
U2
NA
NA
NA
NA
NA
U2
U2
L2
PHASES
STARTING
ODOMETER
4,841
5,048a
5,212"
5,347b
5,019"
4,974b
4,997
5,150a'b
6,799C
5,523a
4,959
5,282
4,785
4,765
4,850
5,211b
3 — These vehicles were added to the Phase 3 test matrix at a later date. Prior to their inclusion in the matrix, they received on-road
miles every other week.
b - These vehicles were included in an FTP interim test program (EPA WA 1-09) con ducted between Phases 1 and2.
c - During Phase 1, the initial 4,000 miles of vehicle break-in was conducted with the wrong crankcase lubricant viscosity grade. An
additional 2,000-mile break-in was conducted with the correct lubricant viscosity grade.
— Dodge Caravan FFVwas only tested with E85
4. Vehicle Procurement and Delivery
45
-------
Vehicles for this program were obtained brand new and delivered to the SwRI facility in
San Antonio, Texas, where testing took place. All vehicles were leased by SwRI for two years at
the initiation of Phase 1 of the V2/EPAct/E-89 program. Due to changes and additions to the
overall program, the term of the two-year leases expired prior to the completion of all Phase 3
testing. The Coordinating Research Council then purchased the test vehicles and made them
available to the test program for the remainder of its duration.
B. Initial Preparation and Storage
1. Mileage Accumulation and Conditioning of Engine Oil
Prior to commencement of emission testing for Phase 1 of the program, each vehicle was
brought up to 4,000 odometer miles, the intent being to avoid variability in emissions behavior
occurring during engine break-in. This was accomplished by operating the vehicles on mileage
accumulation dynamometers over the Standard Road Cycle using a non-oxygenated,
commercially-acquired, 87 octane gasoline (see Table IV-5).
Engine crankcase oil meeting GF-4 specifications was provided for the program by the
Lubrizol Corporation. Engine oil was drained and replaced with the appropriate manufacturer-
recommended viscosity grade at the start of mileage accumulation, and again after 2,000 miles.
The 2,000-mile fill remained in the test vehicles throughout Phases 1, 2, and 3 of the program.
Each vehicle's oil sump was overfilled by 12 ounces at the 2,000-mile oil change to
accommodate oil samples to be taken during the course of the program (see Section V.C.2 for
more details on oil sampling). The vehicle odometer readings at the start of Phase 3 are included
in Table IV-4.
46
-------
Table IV-5. Properties of mileage accumulation fuel.
PROPERTY
Density, 60°F
API Gravity, 60°F
Ethanol Content
IBP
T10
T50
T90
FBP
DVPE
Aromatics
Olefins
Benzene
S
(R + M)/2
Net Heat of Combustion
UNIT
. 3
g/cm
°API
vol. %
°F
°F
°F
°F
°F
psi
vol. %
vol. %
vol. %
mg/kg
-
Btu/lb
METHOD
D4052
D4052
D5599
D86
D86
D86
D86
D86
D5191
D1319
D1319
D3606
D5453
Calc.
D4809
Valero
RUL
0.7329
61.5
<0.1
82
109
194
342
416
11.1
26.2
7.7
0.95
15.9
87.5
18,734
a. Effect of Oil Age on Emissions
A study completed by EPA and Lubrizol shortly before the start of this program
examined the effect of oil age on emissions.17 The study involved accumulating 2000 miles on
two low-mileage, Tier-2-compliant vehicles using a non-ethanol fuel, with emission
measurements taken immediately after oil change and at 500, 1000, and 2000 miles. The only
significant effect of oil age observed was for PM, which dropped by nearly half after 500 miles
in one of the vehicles. Both vehicles showed a general trend of PM reduction out to 2000 miles,
though differences between each test interval were not large or statistically significant.
The study also looked at the effect of mileage accumulation using ethanol-blended fuel
on emissions and oil quality. After the initial 2000-mile accumulation period on EO, the fuel
17 Christiansen, M., Bardasz, E., and Nahumck, B. Impact of Lubricating Oil Condition on Exhaust P articulate
Matter Emissions from Light Duty Vehicles. Society of Automotive Engineers International Journal of Fuels and
Lubricants, December 2010 issue 3: pp. 476-488.
47
-------
supply was changed to E10, and another 3000 miles were accumulated. No significant changes
were seen in emissions between the beginning and end of the period, and an analysis of oil
quality (including metals, acid and base numbers) done by Lubrizol didn't show oil degradation
beyond the normal range for the accumulated mileage.
Though this study was relatively small, it was intended to serve as a screening,
confirming that 2000 miles of oil age should be sufficient to stabilize emissions, and that use of
ethanol fuels didn't appear to cause unacceptable levels of oil degradation.
2. Vehicle Storage Conditions During the Program
Vehicles being actively tested were stored indoors in a temperature-controlled soak area
adjacent to the test cell. All vehicles had batteries trickle-charged for at least 12 hours prior to
testing. Due to the nature of the randomized test matrix, as well as the incremental addition of
test vehicles to the program, certain vehicles were not involved in active testing for several
weeks at a time. Those not scheduled to test for approximately two weeks or more were
generally stored outdoors in a parking area near the test facility. In an attempt to minimize
vehicle maintenance issues due to extended inactivity, those vehicles were operated by an
experienced driver once every two weeks over an on-road course around the perimeter of the
SwRI campus. Prior to each drive, each vehicle received a brief visual inspection to ensure
proper tire inflation and fluid levels. One "lap" was completed, which was approximately 8
miles in length and about 20 minutes in duration. Speed limits ranged from 35 to 45 mph, and
the drive included six traffic signals and two stop signs. This task was conducted using a non-
ethanol fuel.
48
-------
V. TEST PROCEDURES, DATA COLLECTION, ISSUES ENCOUNTERED
There are many things that can affect vehicle emissions performance, with the fuel effect
potentially being a relatively small factor. Thus, an important consideration in fuel effects
studies is the need to minimize measurement artifacts, which include variability, error, or bias in
the results that are unintended in the design of the study. Any such artifacts become confounded
with the fuel effect itself, which can lead to loss of statistical power to resolve an effect or,
perhaps worse, detection of an effect that doesn't exist. As new vehicles have lower and lower
emission levels, more and more sources of measurement artifacts have the potential to become
influential in the results of the study. A great deal of effort was taken in the design and
execution of this program to minimize artifacts related to procedures, including following a
particular sequence of steps in vehicle handling before, during, and between emission tests, as
well as use of the same driver and test cell for all emission tests. Such procedures are described
in more detail in this section. (Another potential source of error in fuel effects models is fuel
property measurement error; the fuels round robin process described in Section III was done to
minimize this.)
A. Test Cycle, Conditions, and Schedule
All vehicle/fuel combinations were tested using the California Unified Cycle, also known
as the LA92. This cycle was chosen because it contains higher speeds and acceleration rates
representative of a wider range of typical driving behavior than what is covered in the FTP cycle.
For this program, the LA92 was conducted as a three-phase, cold-start test in a manner similar to
18
the FTP. Test procedure quality controls included criteria for maximum deviation in speed and
time from the test cycle driving trace to minimize effect of driver behavior on test-to-test
variability.
Testing was conducted during the day shift while vehicle preparation, fuel changes, and
conditioning were conducted during a second shift. All vehicle soaks and tests were conducted
18 In order to supplement data being collected during Phase 4 of this test program, the four FFVs were also tested
over the FTP cycle when operating on E85.
49
-------
at a nominal temperature of 72°F. The representative bulk oil temperature of a vehicle's sump
was stabilized to 72°F ± 3°F prior to conducting any emission test to minimize any effect of
temperature on variability.
SwRI made a number of modifications and optimizations to the test cell air handling
system during the initial phases of the program in an effort to maintain vehicle intake air
humidity at 75 ± 5 grains H^O/lb dry air while tests were being conducted, consistent with
certification procedures in 40 CFR Part 86, to minimize the impacts on NOx emissions.19
During Phase 3 and later portions of the program, this target was met 95% of the time, with some
exceptions when outdoor weather conditions were rapidly changing. SwRI provided a humidity
quality check metric within each individual test file and flagged entries in the test log where
humidity was outside the desired range for more than five percent of the time. EPA and NREL
provided guidance to SwRI regarding whether or not any individual test should be repeated.
Under Phase 1 of the program, SwRI determined and verified PM sample flow rates that
provided proportionality. Those same flow rates were used for Phase 3. The CVS blower was
turned on approximately 20 minutes before each emission test in an effort to ensure tunnel
stability.
1. Test Matrix
The test matrix was designed to test each vehicle/fuel combination in a randomized order
to minimize any effects of biases or artifacts that may not have been addressed through other
provisions in design or procedures (e.g., possible effects of season, weather, changes in test fuel
properties over time, vehicle or instrument drift, etc.). During the first nine weeks of testing,
EPA specified partially randomized vehicle/fuel assignments that alternated between EO and
higher-ethanol blends in an effort to determine the amount of conditioning necessary to allow a
vehicle's fuel control system to adapt to a new ethanol concentration. Development of the
vehicle conditioning procedure is discussed in further detail in Section V.B. Once this issue was
19 The calculation of NOx emission rates includes a humidity correction factor, but providing consistent test
conditions are preferable to relying on this factor alone.
50
-------
resolved, vehicle/fuel assignments were made using a spreadsheet tool that tracked which
combinations had been tested and chose new assignments randomly from the remaining options.
Testing started with five EO fuels as shown in Table V-l. Additional fuels were added to
the test matrix based on both the requirements of the vehicle conditioning assessment and on fuel
availability (fuel blending and delivery took several months). All fuels became available to the
randomization algorithm by the 12 week of testing. Due to initial funding limitations, only ten
vehicles were included in the original Phase 3 test plan. Two additional vehicles were added to
the matrix in the 25* week of testing, and three additional vehicles were added in the 37* week
of testing. These additional vehicles were added to the test randomization scheme with
randomized fuel/vehicle pairings from the outset.
Table V-l. Schedule of addition of fuels and vehicles to the test matrix.
Phase 3
Week
Weekl
Week 2
WeekS
Week 4
Week5
Week 6
Week?
WeekS
Week 9
Week 10
Week 12
Week 25
Week 37
Week 55
Week 60
Fuels
Added
2, 7, 8, 9 and 15
None
None
1, 12, 13
None
22,24
None
3,4,5, 11, 14, 16,20,21,23,30
None
None
6, 10,25,26,27,28,31
None
None
29 (E85)
Vehicles3
Added
CCOB, TCAM, FEXP, DCAL,
HODY
CSIL, TSIE, DLIB, HCIV, NALT
None
None
None
None
None
None
None
None
None
FFOC, SOUT
CIMP, F150, TCOR
DCAR
End of Phase 3 testing
Vehicle/Fuel
Assignments
EPA
Randomized for
rest of program,
except for E85
Last fuel tested
a - Vehicle designations are explained in Section IV.
Due to imperfect test-to-test repeatability and the possibility of individual tests to
significantly influence results, each vehicle/fuel combination was tested at least twice, back-to-
back, with the second replicate usually performed on the following day. When replicates were
split over weekends, an additional prep cycle was conducted as to maintain a 12-36 hr soak
period. After two tests were completed and the acquired data passed all quality control
verifications, the need for a third test was determined according to the variability criteria shown
51
-------
in Table V-2. If the ratio of any of the specified pollutants (THC, NOx, or CO2) on a pair of tests
for a given vehicle/fuel combination exceeded the levels shown in Table V-2, a third test was
conducted and a note was made in the test log.
Table V-2. Repeatability criteria for third replicate.
Dilute Gaseous
Emission
CO2
NOX
THC
Criterion For Requiring a Third
Test (Composite Cycle
Emissions)
Ratio of higher / lower > 1 .03
Ratio of higher / lower > 2.7
Ratio of higher / lower > 2.0
These repeatability criteria were generated based on variance levels found in Phase 1
data, with a target of performing a third replicate for approximately 5% of fuel/vehicle pairs in
Phase 3. This figure was chosen with the goal of covering a small number of measurements that
might be statistical outliers.
Since emissions performance on current technology vehicles is dominated by what occurs
shortly after start-up, engine cranking times between replicates were screened for inconsistency.
If a test differed in cranking time from a previous replicate by more than one second, its
procedure log and emissions data were reviewed by EPA and NREL to determine if an additional
replicate should be performed.
In the end, additional replicates were performed for approximately 3% of vehicle/fuel
combinations due to both repeatability and cranking criteria. (Some additional replicates beyond
these were done as necessary to cover void tests or other procedural issues.) In the end, 926 fully
valid tests were performed with all measurements as specified in the scope of work. Thirty
additional tests had valid data for regulated pollutants but were missing other subsets of data
(e.g., speciation) due to measurement or data quality issues; inclusion of these (denoted as
"makeup tests" and containing an "x" in the test number) brings the total number included in the
database to 956.
52
-------
B. Fuel Change and Vehicle Conditioning Procedures
1. Learned Fuel Trim
The vehicles tested in this program all employed learned fuel adjustments (or trim) to
continuously adjust the amount of fuel delivered for proper combustion. Most vehicle
manufacturers began using such controls during the 1990s, and today nearly all new vehicles use
microprocessor algorithms of varying sophistication to optimize vehicle performance and meet
emission standards.
When the combustion process requires a change in air/fuel ratio, such as occurs when
ethanol blend level changes, the engine controller must adjust the fuel trim to re-optimize engine
and emission performance. This "re-learning" process requires operation of the vehicle for a
certain period of time in several speed and load modes. Since this test program used multiple
fuels with widely varying properties, sufficient prep procedures were needed to ensure emission
behavior was stabilized after a fuel change.
During all testing and prep procedures, vehicles were connected to an on-board
diagnostics (OBD) scanner to capture data for parameters of interest, including those related to
fuel trim. For more discussion of OBD data collection, refer to section V.D.3.
A vehicle fuel change and conditioning procedure was initially developed during Phases
1 and 2. After those emissions and OBD data were analyzed, it was evident that the conditioning
sequence was not sufficient for all vehicles to fully adapt after a change from E20 to EO fuel.
Therefore, the beginning of Phase 3 included a study (done concurrently with emission testing)
to reassess and optimize the vehicle conditioning procedures. Long-term fuel trim (LTFT) and
short-term fuel trim (STFT) OBD parameters were monitored during successive two-phase (bag
1-2) LA92 test cycles and were analyzed for stabilization. Based on the results of this study, the
CCOB, NALT, HCIV, HODY, and TCAM were all conditioned with five successive two-phase
53
-------
LA92s starting in the third week of testing. For all the remaining vehicles, three LA92s was
sufficient. Prior to the third week, all vehicles were conditioned with three LA92 preps.
The final vehicle fuel change and conditioning procedure is given in Table V-3. OBD
data, including LTFT and STFT, were collected during all conditioning runs and made available
to program sponsors as part of the program metadata.
2. Sulfur Purge
It is well known that fuel (and lubricant) sulfur content affects the performance of the
exhaust aftertreatment catalyst. Vehicles emitting at lower and lower levels of regulated
pollutants tend to rely on increasingly active catalysts, and thus there was concern that even a
modest amount of sulfur contamination over time could produce a drift in emissions performance
of the test vehicles that would be confounded with the fuel properties being studied. Thus, a set
of high-speed and load cycles were performed after each fuel change in an effort to reset the
vehicle catalysts back to some relatively "clean" baseline. The specific procedure used was
taken from Appendix C of the CRC E-60 program report and consists of a series of alternating
90
high and low speed cruises with hard accelerations in between.
20 Durbin T.D., et al. (2003). The Effect of Fuel Sulfur on NH3 and Other Emissions from 2000-2001 Model Year
Vehicles, Appendix C. Report number E-60. Coordinating Research Council, Alpharetta, GA. Available at
www.crcao.org.
54
-------
Table V-3. Fuel change, conditioning, and test execution sequence.
STEP
DESCRIPTION
1
Drain vehicle fuel completely via fuel rail whenever possible. When switching to E85 only, drive vehicle
to fully warm up engine.
Turn vehicle ignition to RUN position for 30 seconds (60 seconds when switching to E85) to allow
controls to allow fuel level reading to stabilize. Confirm the return of fuel gauge reading to zero.
Turn ignition off. Fill fuel tank to 40% with next test fuel in sequence. Fill-up fuel temperature must be
less than SOT.
Start vehicle and execute catalyst sulfur removal procedure described in Appendix C of CRC E-60
Program report. Apply side fan cooling to the fuel tank to alleviate the heating effect of the exhaust
system. Engine oil temperature in the sump will be measured and recorded during the sulfur removal
cycle.
Perform four vehicle coast downs from 70 to 30 mph, with the last two measured. If the individual run
fails to meet the repeatability criteria established in Phases 1 and 2 of the program, the vehicle will be
checked for any obvious and gross source of change in the vehicle's mechanical friction.
Drain fuel and refill to 40% with test fuel. Fill-up fuel must be less than 50°F.
Drain fuel again and refill to 40% with test fuel. Fill-up fuel must be less than 50°F.
Soak vehicle for at least 12 hours to allow fuel temperature to stabilize to the test temperature.
Move vehicle to test area without starting engine. Start vehicle and perform three 2-phase (bags 1 and 2)
LA92 cycles. During these prep cycles, apply side fan cooling to the fuel tank to alleviate the heating
effect of the exhaust system. Following the first two prep cycles, allow vehicle to idle in park for two
minutes, then shut-down the engine for 2-5 minutes. Following the last prep cycle, allow the vehicle to
idle for two minutes, then shut down the engine in preparation for the soak.
10
Move vehicle to soak area without starting the engine.
11
Park vehicle in soak area at proper temperature (75 °F) for 12-36 hours. During the soak period, maintain
the nominal charge of the vehicle's battery using an appropriate charging device.
Move vehicle to test area without starting engine.
12
13
Perform LA92 cycle emissions test.
14
Move vehicle to soak area without starting the engine.
15
Park vehicle in soak area of proper temperature for 12-36 hours. During the soak period, maintain the
nominal charge of the vehicle's battery using an appropriate charging device.
Move vehicle to test area without starting the engine.
16
17
Perform LA92 emissions test.
18
Determine whether third replicate is necessary, based on data variability criteria (see Table V-2).
19
If a third replicate is required, repeat steps 14, 15, 16 and 17.
20
If third replicate is not required, return to step 1 and proceed with next vehicle in test sequence.
a - Vehicle coastdown repeatability criteria referred to in Step 5 were provided by EPA as follows:
• maximum difference of 0.5 seconds between back-to-back coastdown runs from 70 to 30 mph
• maximum ±7 percent difference in average 70 to 30 mph coastdown time from the running average for a
given vehicle
b - Some vehicles received only two fuel drains and fills, i.e. Step 7 was skipped. See section V.F.4 for details.
c - Conduct five 2-phase LA92 test cycles for the following vehicles: CCOB, NALT, HCIV, HODY, and TCAM.
3. Fuel Carryover in Vehicle Fuel Tanks
55
-------
On May 27, 2009, (the 11th week of testing) SwRI noticed low levels of ethanol in
exhaust speciation samples collected from the Nissan Altima on Fuel 13, which contained no
ethanol. Samples were pulled from the drum used to fuel the vehicle, and from the vehicle's fuel
tank, and analyzed with a PetroSpec analyzer. The drum sample showed no ethanol while the
vehicle's fuel tank showed 1.5 vol% ethanol, suggesting ethanol-containing fuel had carried over
from a previous test in the Altima. The fuel tank sample was sent to EPA for analysis by ASTM
D5599 method and was found to contain 1.44 vol% of ethanol. This was equivalent to a fuel
carryover rate of 7.2% following two drains and 40% fills, meaning that approximately 3 gallons
of the previous fuel (an E20) remained in the Altima's tank after it has been drained via the fuel
rail. From this point forward, except for mid- and end-point drift check tests, the Altima
received three fuel flushes during the fuel change sequence.
Investigating further, SwRI examined the speciation results for other vehicles and found
measurable levels of ethanol when testing with an EO fuel that was immediately preceded by an
E20 fuel. To better understand this situation, fuel samples were collected during tests where fuel
changed from an E15/E20 to an EO fuel. D5599 analyses of the samples by SwRI and EPA
suggested that the following percentages of the previous fuel were retained in the tanks of the
test vehicles following fuel changes consisting of two drains and 40% fills:
• Honda Odyssey: 8.8 vol%
• Toyota Sienna: 5.0vol%
• Honda Civic: 4.2vol%
• Nissan Altima: 6.1 vol%
• Toyota Camry: 5.3 vol%
• All remaining vehicles: 2.1 to 3.2 vol%
Based on these results, EPA and NREL directed SwRI to prepare several 95%/5% and
5%/95% blends of the test fuels with the most extreme combinations of distillation properties
and ethanol content to determine the effect of 5% fuel carryover on T50, T90 and RVP. The fuel
sampling procedure used during these experiments is given in Appendix G, while the test matrix
56
-------
and findings are given in Appendix H. The results of these experiments generally suggested that
significant (nonlinear) deviations of T50 and T90 would not be expected as a result of carryover,
but that shifts in RVP as high as 0.5 psi could be expected for EO fuels experiencing ethanol
contamination.
Since the initial and subsequent carryover samples from the Altima were relatively
inconsistent (3.7% vs. 7.2% carryover), SwRI performed additional refueling experiments with
the Altima, Odyssey, Camry, Civic, and Sienna to determine the variability of the measurements.
These experiments showed that a third fuel flush was effective in reducing fuel carryover to less
than one percent. The procedure and results are given in Appendix I. Based on these results,
starting on August 1, 2010, SwRI incorporated a third fuel drain and fill into the vehicle change
procedure for the Altima, Odyssey, Sienna, Civic, and Camry to address larger carryover
observed in their fuel systems.
Fuel carryover was characterized for the Focus, Outlook, Impala, F-150, and Corolla
before they were added to the test matrix. Based on results, these five vehicles received triple
drains and fills during fuel changes.
During investigation of fuel carryover, questions also arose about the impact of refueling
location. Two sites were used during the drain and fill procedures, one with a noticeable slope
toward the front of the vehicle, and the other toward one side. (Sloped pavement is typical in
refueling areas, related to spill and storm water control.) Starting in August 2009, each vehicle
was assigned a particular refueling location to attempt to limit any variability in carryover this
may have caused. Prior to this, refueling may have occurred in either location. To assess the
magnitude of effect of refueling location, additional experiments were conducted with the
Silverado, Camry, Sienna, Caliber, Civic, Odyssey, and Altima. Results did not show any clear
evidence of such effect. Additional details are given in Appendix J.
C. Crankcase Oil Level Monitoring
57
-------
Crankcase oil levels were checked monthly to ensure vehicles were operating as expected
and that sufficient volume was available for collection of samples for analysis. No abnormalities
in oil levels were noticed, with the exception of the Ford Explorer as described in section F.I.
Oil samples of four ounces in volume were collected from each vehicle at five points
during Phase 3: at the start and end of the 2,000-mile lubricant break-in (i.e., around odometer
readings of 2,000 and 4,000 miles), and following emissions testing of the 3rd, 15th, and 27th
fuels in the Phase 3 test sequence. The oil samples were shipped in batches to Lubrizol for
analysis. To accommodate the oil samples taken over the course of the program, each vehicle's
sump was overfilled by 12 ounces during the oil change at the mid-point of the 4,000-mile
vehicle break-in. A summary timeline of oil changes and samples by approximate mileage is
shown in Figure V-l, and details of dates and mileages at collection from each vehicle are given
in Appendix F.
9
D
V
a
VEHICLE MILEAG
2000
4000
5000
6500
8000
Break-in, oil aging
Ph.l Ph. 2
nr
Phase 3
Figure V-l. Summary of oil changes and sampling by odometer mileage.
D. Emissions Measurements and Other Data Collection
Both bag-level and continuous emission measurements were collected during vehicle
testing. On-board diagnostics (OBD) parameters relevant to engine operation and emissions
were also recorded during testing.
58
-------
All emission tests were conducted using a Horiba 48-inch single-roll electric chassis
dynamometer, which simulates the loads of acceleration, deceleration, and drag experienced by a
vehicle when driving a test cycle. The computerized dynamometer control system uses a pre-
defined mathematical formula with coefficients specific to each vehicle to simulate its behavior
on the road. Vehicle manufacturers generate target road load coefficients for a vehicle (called
target coefficients), which can then be used to compute set coefficients appropriate create an
accurate simulation in other laboratories. The set coefficients for this program were derived
91
from targets reported by manufacturers in EPA's on-line certification database. These values
were approved by EPA prior to the initiation of testing, and are shown in Table V-4.
Differences in electrical and mechanical characteristics and calibration of dynamometer
equipment between test sites could be a source of variability in engine load that a vehicle
experiences during a test, and thus the emissions it produces. Similarly, the vehicle driver has
considerable influence on behavior of the engine during transient portions of the test cycle, also
affecting emissions. Therefore, in an effort to minimize variability within the dataset, a single
test site and a single driver were used for all data collected in the program, something that is
unusual for a program of this size. Different drivers were used for sulfur purges and vehicle
conditioning. Tension in the tie-downs used to secure the vehicle against the dynamometer roll
was also maintained within a specified range for each test to minimize variability in tire drag and
rolling friction.
21 More details on certification data are available at http://www.epa.gov/otaq/verify/index.htm
59
-------
Table V-4. Vehicle chassis dynamometer settings.
MODEL
YEAR
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008
MAKE
GM
GM
GM
GM
Toyota
Toyota
Toyota
Ford
Ford
Ford
Chrysler
Chrysler
Chrysler
Honda
Honda
Nissan
BRAND
Chevrolet
Chevrolet
Saturn
Chevrolet
Toyota
Toyota
Toyota
Ford
Ford
Ford
Dodqe
Dodqe
Jeep
Honda
Honda
Nissan
MODEL
Cobalt
Impala FFV
Outlook
C1 500 Silverado FFV
Corolla
Camry
Sienna
Focus
Explorer
F150 FFV
Caliber
Caravan FFV
Liberty
Civic
Odyssey
Altima
NAME
CCOB
CIMP
SOUT
CSIL
TCOR
TCAM
TSIE
FFOC
FEXP
F150
DCAL
DCAR
JLIB
HCIV
HODY
NALT
ETW,
Ibs
3,125
3,875
5,000
5,500
2,875
3,625
4,500
3,000
4,750
5,250
3,500
4,750
4,250
3,000
4,750
3,500
TARGET COEFFICIENTS
A,
Ibs
21.51
19.87
38.61
28.80
22.10
29.16
38.41
27.66
32.35
27.26
52.75
35.94
29.53
23.18
28.70
47.47
B,
Ibs/mph
0.5409
0.4397
0.3921
0.8005
0.1500
0.1659
0.0249
0.2892
0.6076
0.9495
-0.3153
0.6505
0.4040
0.1904
0.6915
-0.4531
C
Ibs/mph2
0.01521
0.01752
0.02818
0.03219
0.01886
0.01844
0.02946
0.01697
0.02716
0.02932
0.02826
0.02155
0.02955
0.01699
0.02167
0.02414
SET COEFFICIENTS
A,
Ibs
4.22
8.320
19.860
18.130
8.080
10.110
16.270
15.240
14.350
4.300
15.990
18.470
9.410
8.120
11.170
19.710
B,
Ibs/mph
0.20100
0.11210
0.07430
0.31630
-0.02580
-0.15630
-0.12110
0.07660
0.43360
0.83540
-0.20400
0.30710
0.13330
0.05150
0.24850
-0.30660
C
Ibs/mph2
0.017055
0.018601
0.030294
0.035662
0.020902
0.019592
0.029718
0.018743
0.028153
0.029383
0.025692
0.023981
0.031781
0.017724
0.024710
0.021358
ROAD
LOAD HP
@ 50 mph
11.5
11.4
17.2
19.9
10.2
11.1
15.1
11.3
17.4
19.7
14.4
16.3
16.5
10.0
15.7
11.4
An additional step taken to ensure data quality was the performance of multiple coast-
down checks for each vehicle during each fuel change prep procedure (described in Table V-3,
step 5). These checks involve bringing the vehicle up to 70 mph on the dynamometer and
recording the time it takes to coast down to 30 mph while in neutral. This provides a screen for
friction or miscalibration issues that can affect program results. If a difference was found of
more than 0.5 s between two back-to-back checks, or more than 7 s from a running average, the
results were flagged for follow-up investigation. These values were chosen based on criteria
used for quality control in certification testing. While there were occasional deviations noted, no
significant, reproducible issues were found with coastdown times.
1.
Emission Measurements
The LA92 test cycle used in this program was broken into three portions, or bags, as
shown in Figure V-2. The bag-level emissions measured and reported in all bags and replicates
were THC, NMHC (by FID), NOX, NO2, CO, CO2, and paniculate mass. In addition to the
dilute, bagged exhaust samples, continuous raw tailpipe exhaust mass emissions rates were
measured on a second-by-second basis for all replicates for THC, CFLj, CO, NOx, CO2 and O2.
A subset of tests also had measurement of speciated alcohols, carbonyls, and hydrocarbons, as
well as NMOG and NMHC values calculated to adjust for partial response of the test cell FID to
oxygenated species. All tests and replicates also have reported composite values computed using
60
-------
FTP weighting factors (43% bag 1+2 plus 57% bag 2+3), primarily as a matter of convention as
99
no parallel factors exist for weighting portions of the LA92.
500
1000 1500
Time(s)
2000
Figure V-2. Speed vs. time schedule of California Unified Cycle (LA92).
a.
Methods
Gaseous emissions were determined in a manner consistent with EPA protocols for light-
duty emission testing as given in the CFR, Title 40, Part 86. A constant volume sampler was
used to collect proportional dilute exhaust in Kynar bags for analysis of carbon monoxide (CO),
carbon dioxide (€62), total hydrocarbons (THC), methane (CH4), and oxides of nitrogen (NOx).
Dilution air flow entering the sampling system was measured with a smooth approach orifice,
and a critical flow venturi measured bulkstream dilute exhaust flow. Measured dilution air flow
EPA performs inventory modeling using modeling techniques that break emission rates down by individual
operating modes. Therefore the application of fixed weighting factors to emission bags to create a single composite
result representing average driving is primarily done for historical reasons.
61
-------
was subtracted from the bulkstream flow to calculate raw exhaust flow to determine continuous
raw mass emission rates.23
During Phases 1 and 2 of the test program, continuous raw NFb measurements were
made. Results showed NH3 spikes of several hundred ppm during testing of many of the
vehicles. These concentrations were sufficient to cause poisoning of the NO2-to-NO converter in
the continuous raw NOx analyzer. In an attempt to minimize this problem, prior to the start of
Phase 3 SwRI installed two NHa adsorbers in series upstream of the continuous raw emission
measurement sample train. These adsorbers were changed daily. Additionally, the NO2-to-NO
converter was purged with 5,000-ppm (nominal) NOx for five minutes following every test in an
effort to reverse any NFb poisoning of the converter that may have occurred during testing. The
NOx analyzer was then purged for another three minutes with zero nitrogen prior to initiating the
normal pre-test zero-span sequence.
For the measurement of PM (particulate matter) mass emissions, a proportional sample of
dilute exhaust was drawn through Whatman Teflon membrane filters after secondary dilution
with HEPA-filtered air. A single filter was used for each bag, with both clean and dirty filters
being weighed three times after a period of stabilization in a temperature and humidity-
controlled clean room. Replicate weights were required to be within ±2 jig or the result was
void. The PM sampling method was compliant to CFR, Title 40, Part 1065. Several parameters,
such as filter face velocity and temperature and dilution ratio, were monitored and reported for
each test.
Sampling and analysis of alcohols was done in a manner similar to CARB method 1001,
"Determination of Alcohols in Automotive Source Samples by Gas Chromatography". Dilute
exhaust was bubbled through glass impingers containing deionized water. Immediately
following the test, samples were sealed and stored below 40°F until analysis. Most samples were
analyzed on the day they were collected, but no later than within six calendar days.
23 An electronic flow measurement device (referred to as an EFM) was used to measure raw exhaust flow during
Phase 1 of the program, and results were found to be sufficiently similar to the SAO subtraction method. Given this
fact, as well as cost and maintenance issues related to the EFM, it was not used in Phase 3.
62
-------
Sampling and analysis of carbonyl compounds was conducted in a manner similar to
CARB method 1004, "Determination of Aldehyde and Ketone compounds in Automotive Source
Samples by High Performance Liquid Chromatography". Dilute exhaust was drawn through
cartridges containing DNPH, a compound that selectively binds to carbonyl groups to form
stable complexes with known properties. The DNPH media were extracted using acetonitrile
within 15 minutes of collection, and the extracts sealed and stored immediately at a temperature
below 40°F. Most of these extracts were analyzed on the day they were collected, but no later
than within three calendar days. An effort was made to detect the presence of a tautomer of
acrolein, acrolein-x, which can be a measurement artifact. No acrolein-x was found in any
exhaust sample. Additionally, storage of alcohol and carbonyl samples was segregated to
prevent any cross-contamination of samples.
During carbonyl speciation work in Phases 1-2 of the program, SwRI noticed that
background levels of some species (e.g., acetone and acetaldehyde) were relatively high and/or
shifted in time in ways that were suspicious. SwRI began tracking daily media blanks and found
that new, unused cartridges contained background levels of some species that were similar to the
levels found in dilute exhaust. Thus, a protocol was developed for determining limits of
quantitation (LOQs) of results based on media blank levels. These LOQs were determined for
each compound on each test day using recent blank measurements. A detailed description of this
process is given in Appendix L.
Sampling and analysis of C2-Cu hydrocarbons was conducted in a manner similar to
CARB method 1002/1003, "Procedure for the Determination of C2-Cu Hydrocarbons in
Automotive Exhaust Samples by Gas Chromatography". During the analysis of C2 - C4
hydrocarbons, special consideration was given to 1,3-butadiene. Because of the instability of
1,3-butadiene, the analysis of C2 - C^ hydrocarbon samples collected during Bag 1 of a test cycle
was initiated within one hour of collection. The speciation of Cs - Cu hydrocarbon samples
collected in Bag 1 of the test cycle was completed within 4 hours of collection.
The following daily sequence was used for the analysis of VOC samples:
63
-------
• VOC samples collected during Bag 1 of the test cycle were analyzed first, in the
sequence of vehicle tests.
• If a vehicle requiring VOC sampling during all three bags of the test cycle was tested,
the Bag 1 was analyzed first, followed immediately by the Bag 3 sample and finally
by the Bag 2 sample.
• Background samples were analyzed last, in the sequence of vehicle tests.
A summary of measurement methods is shown in Table V-5. Additional details on
measurement equipment and related QA procedures are given in Appendix K.
Table V-5. Exhaust emission measurement methods.
Constituent
Total hydrocarbon
Methane
Carbon monoxide
Carbon dioxide
Oxides of nitrogen
Nitric oxide
Oxygen
Paniculate matter
Non-methane hydrocarbons
Non-methane organic gases
Nitrogen dioxide
Ci - Ci2 HC speciation
Alcohols
Carbonyls
Analysis Method
Heated flame ionization detector (bag, modal)
Gas chromatography (bag, modal)
Non-dispersive infrared analysis (bag, modal)
Non-dispersive infrared analysis (bag, modal)
Chemiluminescence analysis (bag, modal)
Chemiluminescence analysis (bag only)
Magnetopneumatic detector (modal only)
Part 1065 gravimetric measurement (bag only)
Calculated from THC and CH4 (bag, modal)
Calculated as specified in section VLB (bag only)
Calculated from difference of NOX and NO (bag only)
Gas chromatography (bag only)
Gas chromatography (bag only)
Liquid chromatography (bag only)
64
-------
b. Schedule of Speciated Emission Measurements
Speciated volatile organic compounds (VOCs) measured in this program included Ci -
Ci2 hydrocarbons, light alcohols, and carbonyls (aldehydes and ketones). Due to the additional
time and expense required, the scope of speciation covered a subset of vehicles and fuels.
Alcohols and carbonyls were speciated during bag 1 for all tests (vehicles, fuels, replicates). Ci -
Ci2 hydrocarbons were speciated for the first replicate on all vehicles for a subset of twelve fuels
(3, 4, 6, 7, 10, 13, 14, 21, 23, 27, 28, and 31) chosen to provide, as nearly as possible, useful
comparisons between different levels of ethanol, aromatics, T50, and T90. In addition, all types
of speciation were carried out for bags 2-3 for the subset of 12 fuels on a subset of five vehicles
(Civic, Corolla, F150, Impala, Silverado) intended to represent the range of sizes and
technologies present in the larger test fleet. This information is summarized in Tables V-6 and
V-7. All types of speciation were also carried out for all bags of all E85 tests (fuel 29).
Table V-6. Speciation schedule by fuel, vehicle, and bag (excluding E85 tests).
Vehicle
CIMP, CSIL,
F150, HCIV,
TCOR
All others
Speciation Type
Alcohols,
Carbonyls
Hydrocarbons
Alcohols,
Carbonyls
Hydrocarbons
Replicate 1
Bagl
All fuels
Subset shown in
Table V-7
All fuels
Subset shown in
Table V-7
Bags 2-3
Subset shown in
Table V-7
Subset shown in
Table V-7
-
-
Replicate 2+
Bagl
All fuels
-
All fuels
-
Bags 2-3
-
-
-
-
65
-------
Table V-7. Summary of fuel properties for speciation subset (excluding E85 tests).
Fuel
3
4
6
7
10
13
14
21
23
27
28
31
Ethanol
vol%
10.4
9.9
10.6
<0.10
9.8
0.10
0.10
20.1
20.3
14.9
15.0
20.1
T50
°F
218
222
189
193
217
223
193
168
163
222
217
167
T90
°F
296
338
340
298
340
338
339
305
338
340
299
325
DVPE
psi
6.9
10.0
7.2
7.2
7.1
6.9
7.1
7.1
6.8
7.0
6.9
7.0
Aromatics
vol%
15.0
15.5
15.0
17.0
34.0
34.1
16.9
35.5
15.9
14.9
34.5
35.5
2.
OBD Data Collection
Onboard diagnostic (OBD) system data were acquired at 1 Hz from each vehicle during
all emissions tests using a DBK70 data acquisition system. On rare occasions, OBD data
collection failed during a test; this was not grounds for repeating a test. This dataset includes the
following parameters:
• Engine RPM
• Vehicle speed
• Engine load
• Short term fuel trim-bank 1
• Long term fuel trim-bank 1
• MIL status
• Absolute throttle position
• Engine coolant temperature
• Short term fuel trim-bank 2
• Long term fuel trim-bank 2
• Fuel/air commanded equivalence ratio
66
-------
• Alcohol fuel percent (if available)
• Manifold absolute pressure
• Spark advance
• Engine control module voltage
• Air flow rate from mass air flow sensor
3. Data Quality Control
a. SwRI Procedures
A number of automated data quality checks were used by SwRI during collection and
processing of data. The individual test files contain a QA tab showing a number of quantitative
checks including: zero/span results for emission analyzers; PM filter weight gains, face
velocities, and temperatures; bag-modal emission mass agreement; status of OBD channels on
the vehicle CAN bus; test cell humidity. The test log file shows the status of additional checks
such as repeatability ratio for requirement of a third replicate, long crank time flag, and pass/fail
for chemistry and PM analyses.
b. EPA Procedures
Typically within about one week of completion of a test, SwRI would place the data file
on a secure FTP site for review by the program sponsors. On approximately a weekly basis,
EPA staff loaded the new data files into a database and ran various queries to look for other
potential data quality anomalies. Any questions or issues were then raised during the weekly
conference call with SwRI and other involved parties.
E. Drift Check Tests
Measurements may be affected by any of a variety of sources of drift, generally
understood to be a systematic or progressive shift in results due to vehicle wear, instrumentation
calibration changes, fuel weathering over time, etc. A rigorous scheme for detecting drift in a
67
-------
test program such as this one would involve a large amount of redundant testing, for instance
adding replicates for each vehicle/fuel combination at several points throughout the program.
Needless to say, this would have increased cost and effort significantly, and did not seem
justified given the quality control procedures already in place to minimize variability, as well as
the fact that vehicle/fuel combinations were being tested in a randomized sequence.
Nonetheless, as a screen for any gross drift that might affect the program results, each vehicle
was re-run on a fuel from early in its sequence again at the mid-point and end.
Due to concerns with vehicles properly adapting to different ethanol contents in the test
fuels, the following drift check procedure was conducted. To ensure that the test vehicles were
similarly adapted during start-, mid-, and end-point testing, the test matrix was manipulated so
the immediate history prior to mid- and end-point vehicle testing was substantially similar to that
at the beginning of Phase 3. Specifically, all vehicles were operated on two successive EO fuels
for the first three weeks of testing, which had been immediately preceded by operation on an E20
fuel at the end of Phase 2 of the program. The second EO fuel for each vehicle was designated as
the drift check fuel. With the assistance of EPA, SwRI scheduled mid- and end-point testing to
be immediately preceded by an E20 or El 5 fuel and then an EO fuel with properties substantially
similar to the first Phase 3 fuel on which each vehicle was tested.
Due to the scheduling of end-point drift checks, five of the original ten vehicles
completed testing during Week 34, while the remainder of the original ten vehicles completed
testing during Week 37. (The Ford Explorer did not complete end-point testing until much later
in the program due to a MIL issue described in Section V.F.2.)
Various statistical analyses were conducted, including graphical assessment of emission
trends, modeling measurements as a function of odometer reading, and modeling measurements
as a function of the time of test (beginning, middle, end).
No consistent drift in time-of-test measurements was found. For example, in 120
pairwise comparisons of beginning, middle and end measurements (10 measurements x 4 phases
x 3 pairwise tests) using Tukey multiple-comparison-adjusted p-values, only one comparison
68
-------
was found to be statistically significant: bag 1 CH4 means between beginning and end test runs.
This sole significant result could easily have occurred by chance. Therefore, drift in emission
behavior did not appear to occur in the test program.
F. Issues encountered with Specific Vehicles
1. Ford Explorer Oil Issues
An incorrect oil viscosity was used in the Ford Explorer during break-in. Ford specifies
5W-30 grade for the 4.0L V-6 engine and 5W-20 for the 4.6L V-8. The test vehicle was
equipped with the 4.0L V-6, and was incorrectly filled with the 5W-20 oil at both the start of
mileage accumulation and at the 2,000-mile oil change. The vehicle had accumulated 4,000
miles when this error was discovered. After discussing this situation with all involved parties,
the vehicle received a single flush with 5W-30 oil (2 drains and 2 fills with oil filter changes) and
an additional 2,000 miles were accumulated on the Ford Explorer to break-in the correct oil.
There also appeared to be an oil level issue with the Explorer. When the oil sample was
collected following testing of the 15th fuel, the technician noticed that the oil level was below
the minimum oil level on the dip stick. Following extensive discussions with all sponsors, an
additional 20 ounces of fresh crankcase lubricant were added to the Ford Explorer before
resuming testing. Details of this incident are given in Appendix M.
As a result of this situation, starting in the 22nd week of Phase 3, the oil level on all
vehicles was checked monthly. These checks were taken on a level floor inside the emissions
lab following a minimum 12-hour soak at room temperature (72°F ± 2°F). Initial results showed
the Toyota Camry oil level was between 1/4 and 1/8 of the distance from the fill level to the full
level on the dipstick. The oil level of this vehicle was monitored weekly, but did not change
during the rest of the program. No other vehicles had oil level issues.
2. Fuel System Evaporative Leak Check Failures (Explorer)
69
-------
During the 27 week of testing, the Ford Explorer illuminated a malfunction indicator
light (MIL - a.k.a. "check engine light") for diagnostic trouble code (DTC) P0455-Evaporative
Emission System Leak Detected (Gross Leak/No flow). This started a series of troubleshooting
events that are summarized in Table V-7. On-road testing seemed to indicate that code was not
due to the fuel change procedure or operation of the vehicle on the chassis dynamometer.
Following extensive discussions that included input from Ford technical staff, the team decided
that the MIL would not have an adverse affect on emissions testing, and the vehicle was placed
back into the test matrix.
70
-------
Table V-7. Troubleshooting of the Ford Explorer evaporative system MIL
DATE
9/16/2009
9/17/2009
9/23/2009
9/26/2009
9/27/2009
10/5/2009
10/12/2009
10/13/2009
10/16/2009
10/26/2009
10/28/2009
11/20/2009
11/21/2009
11/22/2009
12/1/2009
12/8/2009
ACTION
Fuel change to E20 Fuel 31; key off
MIL light during vehicle conditioning, E20 Fuel 31; PO455-Evaporative Emission System Leak
Detected (Gross Leak/ No Flow)
Vehicle sent to dealer; performed a smoke test; Canister vent solenoid replaced
Fuel change to E20 Fuel 21; key off
MIL light during vehicle conditioning, E20 Fuel 21; PO455-Evaporative Emission System Leak
Detected (Gross Leak/ No Flow)
Vehicle sent to dealer; performed a smoke test; capless fuel filler door was cleaned as it had dirt and
grime
Fuel change to E20 Fuel 21; key off
MIL light during vehicle conditioning, E10 Fuel 12; PO455-Evaporative Emission System Leak
Detected (Gross Leak/ No Flow)
SwRI performed an IDS test and a smoke test and a leak test by pressurizing the evap system and
found no leaks.
FEXP was taken to the test track where we ran 9 WOT up to 70 mph. The MIL did not light. The next
day we ran through the three LA 92 (2 -bag) prep sequence on the dyno. The MIL did light
approximately 500 seconds (~24 miles) into the third LA 92. This means the fuel change procedure is
probably not the cause for the MIL.
Vehicle was driven on road approximately 50 miles. Pending code P0422, but no MIL light
Vehicle sent to the dealer; The EVAP system was smoke tested and the capless fuel assembly was
replaced. I was told that the technician drove the vehicle for more than 10 miles to confirm that the
code did not reappear.
Fuel change to E15 Fuel 28; key off
MIL light during vehicle conditioning, E15 Fuel 28; PO455-Evaporative Emission System Leak
Detected (Gross Leak/ No Flow)
Dealership performed IDS Diagnosis, PO455 code. EVAP test found capless retainer broken.
Replaced retainer and retested ok.
MIL light during vehicle conditioning, E15 Fuel 28; PO455-Evaporative Emission System Leak
Detected (Gross Leak/ No Flow)
Following Phase 3 testing, SwRI performed additional evaporative system leak checks
with an IDS scan tool per instructions given by Ford technical staff, and the data files were
forwarded to Ford for review. Subsequent coordination among SwRI, Ford technical staff, and
the local Ford dealership allowed us to determine that the fuel tank pressure sensor had an
internal fault causing the signal to become erratic during vehicle operation, after which the
vehicle was repaired.
71
-------
3. Transmission Module Malfunction (Outlook)
A MIL illuminated during the second test of Fuel 16 with the Saturn Outlook. The codes
found were Tran Control Sys Malfunction and U0073 - Control Module Comm. Bus Off. The
MIL and same codes had occurred previously while operating the vehicle on the mileage
accumulation dynamometer during the initial vehicle break-in. At that time, the vehicle was
taken to the dealership where the codes were cleared, and the MIL did not recur until testing Fuel
16.
Review of the emissions results from the two tests on Fuel 16 did not show a significant
difference. With EPA's and NREL's approval, the codes were cleared and the vehicle was
placed back into the test program. Four days later, the MIL came on again during testing, with
the same DTCs. This time the driver noticed that the vehicle's engine was revving higher than
usual at cruising speeds and shifting hard during the first two bags. Bag 3 did not have the same
issues. With EPA's and NREL's approval, the vehicle was taken to the dealer for diagnosis, but
they were not able to find any problems. There was concern that the DBK system used to collect
OBD data may have somehow been interfering with proper vehicle operation. The next set of
tests was conducted without OBD data acquisition, and the MIL did not illuminate. All
subsequent tests were conducted without OBD data acquisition, and the MIL did not illuminate.
72
-------
VI. DATA REPORTING
A. Products and Formats
This program generated a large amount of primary data covering fuels, emissions, and
test conditions, available as follows:
• Individual test files (956 Microsoft Excel files, 3.6 MB each)
• Phase 3 testing log file (1 Microsoft Excel file, 845 KB)
• Phase 3 E85 testing log file (1 Microsoft Excel file, 37 KB)
• Test vehicle fleet data file (1 Microsoft Excel file, 46 KB)
• Fuel property data file with round robin results (1 Microsoft Excel file, 58 KB)
• Fuel speciation results (28 PDF files, 270 KB each)
Summary database files are also available covering gaseous, PM and speciation results.
1. Overview of Individual Test File Contents
The individual test files were produced in Microsoft Excel 2003 format, and use a
standardized layout of six tabs with consistent row/column locations of particular data fields
across all files. A summary of the contents of the six spreadsheet tabs is as follows:
Output-BagSummary
• Pretest data: date, time, test number, vehicle and fuel identification
• Test conditions: phase times, distances, test cell temp, humidity, crank times
• Emission results: concentration and g/mi results by bag and weighted for all pollutants
Output-QA
• Data and process quality checks, including: bag-to-modal comparison, PM collection
parameters, analyzer zero/span check results, OBD CAN data channel status by
parameter, test cell humidity time in-spec, average power and work by bag
73
-------
Output- UnregSummary
• Summary of min/max/average by phase of modal data for various species and pollutants
• Note: This tab was assembled during Phase 1 when additional modal speciation was
being done; for Phase 3 only "TP Raw" results are valid here (others may contain zeros
or other values that are not necessarily accurate).
Output-ModalData
• Second-by-second measurements of OBD and CVS parameters, tailpipe raw
concentrations, computed mass emission rates and cumulative totals
Chemistry Summary
• Mass and concentration data for alcohol, carbonyl, and gaseous hydrocarbon data, as
available for a given test, after filtration by LOQs
• Limits of quantification (LOQs) applied to speciation data
• Concentration results for the subset of compounds being used to calculate NMOG results,
after filtration according to LOQs
DataProcessingNotes
• File review status
• FID response factors used in NMOG calculations
• Miscellaneous notes on procedure and calculations
2. Treatment of Zeros and Missing Values
Speciation results in the Chemistry Summary sheet that were reported as zero after falling
below the LOQ are denoted as "-99". In other sheets, non-detects are simply reported as a
numerical zero value. Missing data omitted for some reason other than a non-detect, such as due
a voided sample, are indicated with the period character (".").
B. Calculation of NMOG and NMHC
74
-------
NMOG (non-methane organic gases) is defined as the sum of all organic emission
species minus methane. It can be compared to NMHC (typically defined as FID THC minus
methane), which contains an incomplete accounting of oxygenated emission species like
formaldehyde, acetaldehyde, and ethanol due to the way the hydrocarbon analyzer operates.24
NMOG results can be generated either by fully speciating the exhaust and summing the masses
of all compounds, or by starting with a FID-based NMHC result and then correcting it using
speciation results for alcohols and carbonyls. A non-oxygenated NMHC total can be generated
in the same way. Where alcohol and carbonyl speciation results were present (bag 1 in nearly all
tests, and bags 2-3 in a subset per Table V-6), SwRI generated NMOG and non-oxygenated
NMHC results based on corrected FID measurements as described in detail in Appendix N. For
tests where there was no speciation, the calculation was performed assuming oxygenated species
masses were zero, effectively setting the NMOG and NMHC values equal to the traditional
NMHC definition of FID THC minus methane.
24 See Appendix P for details on FID response to oxygenated compounds.
75
-------
VII. SUMMARY AND CLOSING REMARKS
This program collected regulated and unregulated emission data, by bag and second-by-
second, for 15 new, properly-operating light duty vehicles of 2008 model year using a
statistically-designed partial factorial matrix of 27 gasoline test fuels covering typical market
ranges of ethanol, T50, T90, aromatics, and RVP. Complete data were generated for 926 tests,
with 30 additional tests containing valid measurements for most emissions.
During the execution of this program a number of questions about best practices for
emission studies were raised, and experiments were carried out to answer some of them.
Observations include:
• New vehicles contain sophisticated engine control systems that can have significant
effects on emission behavior. Specific fuel change and prep procedures must be followed
to isolate to the greatest extent possible the impacts of fuel changes from transient
behaviors resulting from learning processes occurring in engine controls.
• Vehicle fuel tanks vary widely in how much fuel remains after being drained by the fuel
pump, which can result in fuel carryover and increased variability of actual in-tank fuel
properties during testing. Careful fuel change procedures can accommodate this.
• As vehicle emission levels get lower and lower, more care must be taken to reduce
sources of measurement error and variability. Speciation of carbonyls was one area
found to be especially subject to this issue since DNPH cartridges contain background
levels for some species of interest similar in magnitude to dilute emission levels. In
addition, dilute gaseous emission measurements such as NOx may have sample
concentrations similar in magnitude to background for some portions of the test (e.g.,
bags 2-3).
76
------- |