Final Determination of Hot Running Emissions From FTP Bag Emissions


           United States         Air and Radiation        EPA420-R-01-059
           Environmental Protection                  November 2001
           Agency                        M6.STE.002
&EPA    Final Determination of
           Hot Running Emissions from
           FTP Bag Emissions
                                   y£u Printed on Recycled
                                   Paper

-------
                                                          EPA420-R-01-059
                                                            November 2001
Final Determination of Hot  Running Emissions from
                      FTP Bag Emissions

                           M6.STE.002
                               Ed Glover
                            David Brzezinski
                               Phil Enns
                     Assessment and Standards Division
                    Office of Transportation and Air Quality
                    U.S. Environmental Protection Agency
                                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 which
      may form the basis for a final EPA decision, position, or regulatory action.

-------
Introduction

This document describes our efforts to develop a simple model for estimating hot running
505 (HR505) emissions from FTP data. The HR505 is an extra exhaust emissions "bag"
performed immediately following the third "bag" of the standard FTP. This new "bag" is a
duplicate in terms of speed/time to the first and third "bags". The only difference between the
"bags" is the FIR505 does not contain an engine start.

The correlation between the FIRS05 and the FTP is based on special testing done by EPA.
In this program, vehicles were tested on both the HR505 and the FTP with the FTP first and the
HR505 following immediately afterward. Because the testing process was sequential, ambient
test conditions, fuel properties, and vehicle operator variables were controlled to minimize their
effects. These data allow the development of a linear correlation of the form:

HR505 = f(FTP Bagl, FTP Bag2, FTP Bag3)

This correlation form was chosen because of its simplicity and the very high level of
correlation which is achieved. Other variables such as model year and fuel injection type, and
differences between the various "bags" were tried; also, other fits such as a non-linear fit were
tried, but were not used. None produced appreciably better correlation. The correlation between
the HR505 and the FTP is important because relatively few data points are available on the
HR505; however, many FTP data points exist, and can thus be used to calculate simulated
HR505 results.

The HR505 was developed to allow the separation of the emission effects of vehicle start
with the effects of hot running operation. This split will allow the separate characterization of
start and running emissions for correction factors such as fuel effects and ambient temperature.
It also allows a more precise weighting of these two aspects of exhaust emissions for particular
situations such as parking lots and freeways. MOBILE6 will allocate vehicle exhaust emissions
to either those associated with engine start (start emissions) or those associated with travel
(running emissions).

More information regarding start emission and running emissions and the role of the
HR505 can be found in the accompanying EPA document entitled "Determination of Start and
Running Emissions Deterioration" - M6.EXH.001. This document describes in more detail the
methodology and equations used to calculate start and running emissions using the HR505
results.
Sample Selection and Data

The sample for this analysis came from EPA emission factor testing performed at the
Automotive Testing Laboratories, Inc., in Ohio, and from testing performed at the EPA Lab in
Ann Arbor, Michigan. The Ohio lab performed 50 of the 77 vehicle tests, and the Ann Arbor

-------
lab performed the remaining 27 vehicle tests. All of the Ohio vehicles were recruited at
Inspection and Maintenance (I/M) lanes run by the State of Ohio, and were tested in an as-
received condition (without repairs). Many of these vehicles were I/M failures, and produce
excessive emissions (not a random sample). However, except for one vehicle which had an
intermittent problem, some care was taken to exclude vehicles with obvious problems (i.e.,
start/stall problems) that would bias the results. The Ann Arbor vehicles were recruited from
extensive mail solicitations of the general public, and were also tested in an as-received
condition. The sample contained a total of 77, 1983 through 1996 model year vehicles. It
comprised both cars and trucks, and was weighted predominately toward late model year vehicles
and newer technology. Since the use of the HR505 regression equation in the MOBILE6 model
was primarily on 1981-1993 model year vehicles (and post 1993 model years for CO emissions
only), the use of the 1994, 1995 and 1996 model year vehicles in this analysis assumes that the
correlation between the Hot505 and the individual bag emission results from the newer vehicles
is similar to the correlation for the 1981-1993 model years. Table 1 (a separate Excel
spreadsheet titled 'F505.xls') shows the emissions and model year data on all 77 vehicles.

All of the vehicles were tested using the FTP procedure, including an extra test segment
(bag) which did not include an engine start. The first, third and extra bag samples from this
testing all used the identical driving cycle, sometimes referred to as a "505", since it lasts 505
seconds. The "extra" bag, which uses a 505 but does not include an engine start is the HR505.
Appendix A at the end of this document contains additional details regarding the test procedure
and vehicle recruitment.

The test program data are shown in Table 1 for all of the 77 vehicles. It shows the FTP
emissions (by bag) and the results of the HR505 measurement for total hydrocarbons (HC),
carbon monoxide (CO), oxides of nitrogen (NOx), and non-methane hydrocarbons (nmHC).
The non-methane hydrocarbon emissions were calculated from the total hydrocarbon emissions
by subtracting a methane measurement which was made during all of the tests. All emissions
in the tables are reported in grams per mile.

Prior to curve fitting, examination of the data indicated that vehicle #16, a 1989 Buick
LeSabre, was an extreme outlier in terms of HR505 CO emissions. This vehicle's running 505
CO emissions were measured at 53.8 grams per mile (g/mi); however, Bag 1 (4.71 g/mi)andBag
2 (3.33 g/mi) CO emissions were much lower. This was peculiar since both Bag 1 and Bag 3 are
expected to be larger (or only slightly smaller due to testing variation) than the running 505
results. This is because both of those bags contain an engine start in addition to running
emissions. Examination of the vehicle showed a problem with the block learn multiplier test,
indicating that there was probably an intermittent failure of the closed-loop fuel control system
on this vehicle. Because of the intermittent nature of the failure and the very large discrepancy
between the hot running 505 and the other bags, Vehicle #16 was removed from the model fitting
for all three pollutants. The intermittent nature of the failure, and the fact that it is not "start"
specific is a problem because in theory it should have the same probability of occurring in either
Bag3 or the HR505 or both cycles. Thus, probabililistically a large negative is as likely as a large
positive effect. If either of these were to be added to the a small sample size, bias would occur.

-------
As a result, the vehicle was removed from the analysis. Table 2 shows the emissions statistics
for the sample with and without vehicle #16.
Vehicle #219 was also a very high CO emitter with higher emissions for the HR505 than
for Bag3. It was retained in the analysis because no intermittent performance problem could be
identified. Also, the absolute difference between the HR505 and Bag3 is large for Vehicle #219,
but the relative difference is not as greater as the relative difference for Vehicle #16.
Analyses

Several models to predict HR505 emissions versus FTP emissions were fitted using least
squares regression analysis. The regressions included simple linear regressions as well as non-
linear and logarithm transformed regressions. They utilized several dependent variables such
as the individual FTP bag results and the model year. In choosing a final model, several
formulations were considered. Beginning with Bags 1, 2 and 3 for all three pollutants and the
vehicle model year parameter as independent variables, standard variable selection methods were
applied in order to reduce the number of predictors. Not surprisingly, the best models include
Bag 2 and Bag 3 of the pollutant being predicted. Models using these two variables account for
a high percentage of the variation in the dependent variable. While the Bag 1 logs of emissions
are not statistically significant, it was decided to include this variable in the final models in order
to more fully utilize the available information. The model year variable was found to be adding
little to the predictive power of the model and be non-significant; thus, it was dropped from the
model. For all three pollutants, the final model is the transformed value of the linear fit of the
logs of Bags 1, 2, and 3:

HR505 = Exp[ (A * LN(Bag 1)) + (B * LN(Bag 2)) + (C*LN( Bag 3)) + D + LogTrans Factor]

where A, B, C, D, and the LogTrans Factor are unknown constants. Table 3a shows the
coefficients for the above formulation for each pollutant along with the R-square and T
significance statistics.

LogTrans Factor is a logarithm transformation constant. Numerically, it is the mean
squared error of the regression divided by 2. It is added to the predicted value of HR505 to
account for the bias which occurs when the data distribution is changed from log to linear. This
bias occurs because the 'logged' distribution is approximately normal with the mean equal to the
median. However, the linear distribution of emission data is positively skewed with an unequal
mean and median. The LogTrans Factor is an approximation technique to overcome the bias.
It is referenced in Kendall & Stuart, "The Advanced Theory of Statistics", 1967. The individual
values of this log transformation constant are shown in Table 3a.

A similar linear regression model of the HR505 versus the three FTP bags in linear space
(non-log transformed) were also performed. The results are shown in Table 3b. Although these

-------
regressions produced significant T statistics and generally higher r-squared values than the log
transformed models, they were not selected based on the diagnostics from the regression residual
P-P plots. These standardized P-P plots are shown in Charts 1 and 2 for HC. Similar plots were
obtained for CO and Nox. These plots suggest a non-normal distribution of the regression
residuals when working in linear space (the residuals do not follow a 45-degree line). When
transformed into log space the distribution becomes more normal (better approximate a 45-
degree line). Thus, the fundamental assumptions of linear regression are more closely met by
transforming the data into natural log space.
Conclusion

The regression coefficients presented in Table 3a are EPA's best estimate of the
correlation between FTP Bagl, Bag2 and Bag3 emission results versus the Hot Running 505
emission results. It is proposed that this correlation will be used to generate Hot Running 505
emission factors and start emission factors from standard FTP test data.
Response to Stakeholder and Peer Review Comments

Significant stakeholder comments were not received for this document. However, three
separate paid and independent peer reviewers were used, and provided the following comments.
Their comments were either addressed directly in the document or are discussed below.

1. One reviewer thought that the sample of 77 vehicles was insufficient, and that it was too
heavily weighted towards late model year vehicles and high emitters.

Given the high cost of FTP type emission testing, a small vehicle sample size is usually
the norm. Compared with other vehicle testing programs, a sample of 77 vehicles is
fairly large. The skewness towards late model year vehicles could not be avoided in the
testing program due to a desire to test newer model year vehicles (1994-1996) for other
purposes. The inclusion of newer model years also proved beneficial since these
regression equations were extended for use in post 1994 model year CO emission factor
development. The inclusion of a higher percentage of high emitters in the sample than
in the overall fleet may also add some uncertainty to the analysis. However, limited
analysis using a dummy high emitter classification variable suggested that emitter
classification was not statistically significant.

2. One reviewer questioned whether other models for predicting Hot Running 505 results
or determining cold start emission effects were considered for use. One other method
which was considered, consisted of examining the second by second emission results
from the FTP cycles, and determining how long in the FTP cycle (in terms of either cycle

-------
       time or cycle percentage) it takes to warm up a vehicle and light off the catalyst. This
       approach may have allowed for the estimation of cold start or hot start in terms of a
       percentage of Bagl or Bag3, respectively.  Unfortunately, this approach could not be
       pursued due to a complete lack of second by second data on the FTP cycle.

       Another proposed approach was to use only the third Bag of the FTP  cycle in the
       correlation and drop Bags 1 and 2. This approach was analyzed and found to produce
       statistically inferior results to the logarithm based regression of all three FTP bags.

3.      A reviewer also suggested that Table 1 should be updated to include the Hot Running
       505 emission factor produced from this study. Table 1 was not updated; however, this
       comment has been incorporated in this final draft as an attached Excel Spreadsheet titled
       'F505.xls'.

-------
Table 2
Sample Descriptive Statistics

FTPHC
FTP CO
FTP NOx
Bag 1 HC
Bag 2 HC
Bag 3 HC
Running 505 HC
Bag 1 CO
Bag 2 CO
Bag 3 CO
Running 505 CO
Bag 1 NOx
Bag 2 NOx
Bag 3 NOx
Running 505 NOx
(Bag 1 HC - Running 505 HC)
(Bag 1 CO - Running 505 CO)
(Bag 1 NOx - Running 505 NOx;
Sample without vehicle #16
Mean
1.35
19.66
1.16
1.83
1.29
1.11
0.91
23.57
20.02
16.02
15.88
1.56
0.92
1.32
1.19
0.92
7.70
0.37
Std
Dev
2.19
39.90
1.17
2.03
2.51
0.22
1.80
34.68
47.22
32.55
37.24
1.23
1.09
1.38
1.33
1.02
20.01
0.64
Min
0.07
0.58
0.08
0.29
0.01
0.01
0.01
2.54
0.00
0.04
0.04
0.22
0.01
0.04
0.01
-3.17
-93.98
-3.62
Max
11.55
203.43
5.58
10.78
14.37
10.77
11.04
150.03
253.71
162.32
224.70
5.76
5.63
5.79
5.47
5.99
120.22
1.88
Full Sample (77 cases)
Mean
1.34
19.43
1.15
1.82
1.27
1.10
0.92
23.33
19.76
15.86
16.37
1.55
0.91
1.30
1.17
0.90
6.96
0.37
Std
Dev
2.18
39.68
1.16
2.02
2.50
0.21
1.79
34.52
46.96
32.37
37.24
1.22
1.09
1.37
1.33
1.03
20.90
0.63
Min
0.07
0.58
0.08
0.29
0.01
0.01
0.01
2.54
0.00
0.04
0.04
0.22
0.01
0.04
0.01
-3.17
-93.98
-3.62
Max
11.55
203.43
5.58
10.78
14.37
10.77
11.04
150.03
253.71
162.32
224.7C
5.76
5.63
5.79
5.47
5.99
120.22
1.88

-------
                    Table 3a
   Final Model Regression Coefficients (log-log)
Dependent Variable  LN(Running 505 HC)
LN(Bag 1 HC)
LN(Bag 2 HC)
LN(Bag 3 HC)
(Constant)
Log Trans Factor

R Square    0.9531
Coefficient
0.2236 (A)
0.5010 (B)
0.3333 (C)
-0.5065 (D)
  0.0733
T Test Sig
  0.0658
  0.0000
  0.0110
  0.0000
Dependent Variable

LN(Bag 1 CO)
LN(Bag 2 CO)
LN(Bag 3 CO)
(Constant)
Log Trans Factor
R Square 0.9410
LN(Running 505
Coefficient
0.0005071 (A)
0.4304 (B)
0.5375 (C)
-0.0674 (D)
0.099

CO)
T Test Sig
0.9958
0.0000
0.0000
0.7250


Dependent Variable  LN(Running 505 NOx)
LN(Bag 1 NOx)
LN(Bag 2 NOx)
LN(Bag 3 NOx)
(Constant)
Log Trans Factor

R Square     0.9220
Coefficient
0.0209 (A)
0.4655 (B)
0.5328 (C)
0.0416 (D)
  0.0747
T Test Sig
 0.8685
 0.0001
 0.0001
 0.6267

-------
                 Table 3a Con't
   Final Model Regression Coefficients (log-log)
Dependent Variable  LN(Running 505 NMHC)
LN(Bag 1 nmHC)
LN(Bag 2 nmHC)
LN(Bag 3 nmHC)
(Constant)
Log Trans Factor

R Square     0.9487
Coefficient
0.4162 (A)
0.5379 (B)
0.2232 (C)
-0.6634 (D)
  0.1986
T Test Sig
  0.0144
  0.0000
  0.0371
  0.0000
                     10

-------
                    Table 3b
 Alternative Model Regression Coefficients (linear)
Dependent Variable  (Running 505 HC)
(Bag 1 HC)
(Bag 2 HC)
(BagSHC)
(Constant)
R Square     0.9644
Coefficient
-0.1472 (A)
0.4487 (B)
0.4918 (C)
0.0609 (D)
T Test Sig
  0.0039
  0.0000
  0.0000
  0.3112
Dependent Variable

(Bag 1 CO)
(Bag 2 CO)
(Bag 3 CO)
(Constant)
R Square 0.9806
(Running 505 CO)
Coefficient
-0.3452 (A)
0.3480 (B)
0.9700 (C)
1.5050(D)


T Test Sig
0.0000
0.0000
0.0000
0.0685

Dependent Variable  (Running 505 NOx)
(Bag 1 NOx)
(Bag 2 NOx)
(Bag 3 NOx)
(Constant)
R Square     0.9785
Coefficient
-0.0989 (A)
0.1770 (B)
0.9027 (C)
-0.0123 (D)
T Test Sig
  0.0424
  0.0168
  0.0001
  0.7667
                       11

-------
Chart 1 - Log HR505 HC
P-P Plot of Standard Residuals
1.00
.Q 75'
2
CL
| .50
O
T3
-------
Appendix A

Amendment 1 : Work Assignment 1-03
Contract 68-C5-0006

Statement of Work

Inventory Cycle Data Collection
I. BACKGROUND

EPA's "MOBILE" computer model is used by regions, states, and
municipalities in estimating in-use emissions from mobile sources. This model was
derived from data obtained from previous testing programs around the country and
most recently from data obtained at the EPA's National Vehicle and Fuel Emissions
Laboratory in Ann Arbor and from operating laboratories and I/M lanes in
Hammond IN and Phoenix AZ. EPA has the responsibility of updating its model to
provide the latest information on regional driving patterns and modeling strategies
for current driving behaviors.

This work assignment will gather emissions data from light-duty vehicles
(LDV) being run on various inventory cycles (ICs) to provide additional information
for the MOBILE database. Each 1C models an atypical (e.g., non-standard road
conditions, traffic congestion, non-FTP speeds) LDV trip. Changes in a vehicle' s
expected emissions when it is operated over one of these ICs are used to calculate
area-specific emissions for the LDV fleet within the MOBILE model. Exhaust
emission measurements will also be conducted.

II. OBJECTIVE

Several ICs as detailed in Appendices X, Y, and Z shall be run on vehicles
recruited at a centralized I/M facility. This will allow EPA to add more fleet
characteristics emission data to its MOBILE model. A secondary purpose shall be
to gather data on cold start emissions using a ST01 start cycle. All vehicles shall
receive a FTP exhaust emissions test, as well.

III. RECRUITMENT

The contractor shall recruit a total of 50 vehicles that have completed an I/M
test lane: 1) 35 light-duty vehicles and 5 light-duty trucks from model year 1988
and newer; 2) 5 light-vehicles from pre-1988 model year; and 3) 5 light-duty trucks
from 1988 to present light-duty cars. The vehicles will be a naturally occurring mix
of carbureted and fuel injected systems. Every attempt will be made to locate at

13
-------
- 14 -
least half of each sample failing the I/M240 test with either high NOx (oxides of
nitrogen) or high combined HC-CO (hydrocarbon-carbon monoxide) emissions, but
not both. The vehicles shall be recruited as shown in the table below:

Model Year Pass Fail NOx Fail HC-CO

1988-Newer 50% (12) 12.5% (4) 37.5% (9)

IV. LANE TESTING

The I/M240 test will be run on each vehicle. The Contractor shall use the
results from the state contractor's test. These tests will form the basis for vehicle
recruitment. These tests will be performed over the entire 239 seconds of the I/M240
(no fast pass or fast fail allowed) and the composite HC, CO, and NOX results in
grams per mile shall be recorded and reported. The lane procedures are shown in
Appendix XI.

V. LABORATORY TESTING

The Contractor shall perform the ST01 start cycle (the first 258 seconds of
EPA's SC03 cycle), the "area-wide" inventory cycle (similar to CARB's "Unified "
cycle), CARB's LA92, the New York City Cycle, and 11 other inventory cycles (see
detail in section "VI TEST SEQUENCE" of this work assignment). The ST01 cycle
shall be run as a cold start test and all cycle data shall be collected modally second-
by-second on a twenty-inch (20") roll dynamometer. The Contractor also shall
perform a cold-start FTP (exhaust) test on each LDV with an additional fourth bag
505 on a Clayton dynamometer and the data collected non-modally. A flowchart
showing the sequence of events is included as Attachment 1.

VI. TEST SEQUENCE

The test sequence shall include:

1) Cold ST01 start cycle (see Appendix Q of the Statement of Work)

2) A hot start LA-4 to measure and qualify bag vs. modal (second by second).

3) All of the following cycles for each test vehicle, run in random order for each
LDV:

14
-------
- 15 -
1. LOS A-C Freeway Trace (8.60 mins) ;
2. LOS D Freeway Trace (6.80 mins);
3. LOS E Freeway Trace (7.77 mins);
4. LOS F Freeway Trace (7.45 mins);
5. LOS G Freeway Trace (6.52 mins);
6. Ramp (4.43 mins)
7. LOS AB Arterial Trace (12.28 mins);
8. LOS CD Arterial Trace (10.48 mins);
9. LOS EF Arterial Freeway Trace (8.40 mins);
10. Local Roadways (8.75 mins);
11. Areawide Non-Freeway
12. LA92
13. NYCC
14. High-Speed

3) Cold-start FTP (exhaust portion) (see Appendix F, FTP SEQUENCE)
VII. REPORTING REQUIREMENTS

A. Weekly Reports

All of the raw and processed data will be reported according to the basic
contract and the attached formats. Submittal of these data will be on a weekly basis
and may be made using electronic transfer either by modem or over the Internet. A
spreadsheet for each task will be submitted that includes sufficient information to
identify the vehicle being tested and the results of each individual test performed. A
narrative description which notes any unusual problems encountered or identifies
any maintenance performed shall be included as part of the weekly report.

A narrative summary of the week's activity will be included in the normal
weekly report for each active work assignment under this contract. This will include
the number of vehicles tested to date along with any significant observances for that
week. A table showing the overall status of the work assignment will also be
included and updated each week. This narrative may also be submitted
electronically over the Internet.

Recruitment statistics shall also be included in this report. These statistics
will include a count of each and every vehicle owner approached. The data shall be
15
-------
- 16 -
broken down month by month (when sufficiently far into work assignment) into
those vehicles that were ineligible to participate, those who agreed to be tested but
were not, and those who refused to participate in the program. These three groups
are to be further broken down into specific reasons for the vehicle not participating.
The contractor shall attempt to achieve as close to 100% participation as possible.

B. Monthly Reports

Monthly reporting will be as required by the contract and will include a
summary of all work performed under the above subject tasks as well as results of
all calibrations on all equipment used.

C. Final Report

The final report shall be a narrative describing the testing in detail and
including any changes made during the performance of the work assignment.
Furthermore, the final report shall contain a summary of any problems encountered
and their resolution. It shall also list all tests and test results on all canisters in the
program.

Recruitment statistics shall also be included in this report. See Weekly
Reports for specifics on the reporting of recruitment statistics.

Within 30 calendar days after completion of the last test sequence performed
for this work assignment, the contractor shall submit for technical and editorial
review by the Project Officer a draft final report in both written and electronic
formats. The written draft shall be typed, double-spaced, and shall include all
illustrations, tables, drawings, charts, data sheets, and any other pertinent material
required in the approved final report. The Project Officer will notify the contractor
of approval or rejection of the draft report within 30 calendar days and shall
provide comments citing any changes, corrections, or additions required for
approval. Within 30 calendar days after receipt of the comments, the contractor
shall submit to the Project Officer a final report in both electronic and written
formats. The written report shall include the single spaced original manuscript and
five copies of the approved final report.
16
-------
- 17 -
Appendix XI
Test Procedures
IM Lane Procedures
An I/M240 test will be run on each vehicle at a centralized, i.e., state-
mandated, testing facility. The I/M240 testing facility must be within 100 miles of
the Contractor' s vehicle testing facility. The Contractor shall recruit vehicles for
this WA on the basis of the results of the state contractor's I/M240 test. In each
case, the composite HC, CO, and NOx results in grams per mile shall be recorded
and reported with any purge and/or pressure data. A potential test vehicle must be
on-site at the Contractor' s testing facility within twenty-four (24) hours or by close-
of-business the day following its recruitment from a centralized I/M240 facility.

TEST FUEL

During this work assignment, all vehicles shall be tested with the same lot of
indolene-type fuel which complies with Code of Federal Regulations (CFR) §86.113-
91, having a preferred RVP of 9.0 psi (not to exceed 9.05 psi and not to be less than
8.70 psi). The Contractor shall measure and record the RVP of the fuel dispensed
at each vehicle' s fueling prior to the ST01 cycle run (see Appendix 1). The
contractor must provide EPA with a complete analysis of each lot of the test fuel.
The contractor must obtain approval of the Project Officer before using any test
fuel.

INITIAL TEST CONDITIONS

Each vehicle will be pre-conditioned as per CFR §86.132-96 (a)(l); a LA-4
pre-conditioning drive shall be performed.

The data shall be recorded continuously and reported in second-by second
increments in comma separated form (C.V.) on a completed vehicle basis for modal
testing. For a FTP test, data shall be reported in as described in CFR §86.135-94.
The procedures used to calculate the HC emissions shall comply with §86.144-78.
17
-------
- 18 -
BETWEEN-CYCLE TRANSITIONS

The Contractor shall use a random number generator to randomize the test
sequence order of the fourteen cycles (1 through 14), for each of the 50 test vehicles.
The acceleration rate found at the end of each cycle will be extended for 10 seconds
past the end of the sample period. The acceleration rate found at the beginning of
the next cycle will be extended for 10 seconds prior to the start of that cycle. A forty
second transition period will be used to connect the extended speeds, for a total of 60
seconds between cycles. A "worst case" transition of 0 mph to 80 mph in 40 seconds
would result in an acceleration/deceleration rate of 2.0 mph/sec. There shall be no
emission measurements done during these transitions, but they will be documented
with speed versus time data.

Each test vehicle shall have a unique driving schedule for whole test
program based on the above random test sequence of test cycles. The cycles will be
combined into groups of two or three. If the cumulative time for the first group two
cycles is less than thirty minutes, the next cycle test sequence shall be added to that
group.

Bag samples will be collected at the same time the dilute modal samples are
collected and measured. The bag samples will be analyzed following the completion
of the group's two or three driving cycles.

The test vehicle shall be preconditioned prior to each group of cycles with an
un sampled hot transient phase (hot 505) of the FTP if less than one hour has
transpired since the last vehicle operation. An un sampled "LA-4" shall be
performed if that period exceeds one hour and less than four hours.

All subsequent vehicles will follow the same procedure until all fifty LDVs
have been tested on the test sequence.
18
-------
- 19 -
APPENDIX F

FTP SEQUENCE

Upon completion of set of the ICs, the vehicle is soaked as long as necessary
or overnight to achieve the specified FTP test start temperature. The vehicle will
then undergo a cold start FTP (exhaust portion) as shown in CFR §86.135-94.
Immediately following the hot transient phase (hot bag 3) of FTP, the contractor
will perform a repeat hot 505 without a key off and restart. The contractor shall use
a special driving cycle consisting of two consecutive 505 cycles form the FTP.

The CVS system used during the FTP test shall maintain the tail pipe
exhaust pressure to within ± 1 inch H2O of the pressure experienced by the tail pipe
with no attachments during the FTP cycle. Care shall be taken to verify the device
used to measure the pressure in the line is one which does not itself alter the
pressure significantly. The system shall be tested using both a large displacement
(more than 4L) and a small displacement (less than 1.7L) engine. This will verify
that the system functions properly under different extremes of exhaust volume.
Results of this test shall be reported to and discussed with the Project Officer prior
to initiation of testing.

Modal versus Bag Data Analysis and Quality Control

Each 1C and hot LA-4 shall include both bag and modal test results. The
contractor shall compare the difference between all Bag and Modal emissions. They
shall report the comparisons to the Project Officer to be reviewed for each cycle.
The bag vs. modal comparisons for the hot LA-4 test shall be within ± 5%.
19
-------